Midas NFX Analysis Manual

Analysis Manual
Modeling, Integrated Design & Analysis Software
1
MIDAS Family Programs consist of a group of software packages for structural analysis and
simulations for the purpose of design developed by MIDAS IT Co., Ltd.
MIDAS Family Programs and all associated documentation are copyrighted and protected by
the software protection law.
For any enquiry concerning the programs or related materials, please contact:
MIDAS Information Technology Co., Ltd.

MIDAS IT Tower, 633 Sampyeong-dong, Bundang-gu
Seongnam, Gyeonggi-do, 463-400, Korea
Phone: +82-31-789-2006 Fax: +82-31-789-2110
[email protected] www.MidasUser.com
M odeling, Integrated Design & Analysis Software
Trademarks and Registered Trademarks referred to in this User’s Guide are as follows:
AutoCAD is a registered trademark of Autodesk, Inc.

Excel is a trademark of Microsoft Corporation.
IBM is a registered trademark of International Business Machines Corporation.
Internet Explorer is a trademark of Microsoft Corporation.
MIDAS is a registered trademark of MIDAS Information Technology Co., Ltd.
Sentinel is a trademark of Rainbow Technologies, Inc.
Windows is a trademark of Microsoft Corporation.
PROGRAM VERIFICATIONS AND PRECAUTIONS BEFORE GETTING STARTED
MIDAS Family Programs produce accurate analysis results based on up-to-date theories and
numerical and graphical simulation techniques. The programs have been verified by thousands
of examples and comparative analyses with other software during the development.
Since the initial development in 1989, MIDAS Family Programs have been accurately and
effectively applied to an uncountable number of real life projects.
Due to the complex nature of structural analysis and design programs, which are based on
extensive theories and design knowledge, the sponsors, developers and participating
verification agencies do not assume any rights or responsibilities concerning benefits or losses
that may result from using MIDAS Family Programs. The users must understand the bases of
the programs and the User’s Guide before using the programs. The users must also
independently verify the results produced by the programs.
DISCLAIMER
The developers and sponsors assume no responsibilities for the accuracy or validity of any
results obtained from MIDAS Family Programs (midasNFX, midasFX+, midasCivil, midasGen,
midasGTS, midasFEA, SoilWorks and any other associated programs).
The developers and sponsors shall not be liable for loss of profit, loss of business, or any other
losses, which may be caused directly or indirectly by using the MIDAS Family Programs due to
any defect or deficiency therein. The user shall comply with the End User License Agreement
supplied with the programs.
Analysis Manual
midas NFX
2
2
Analysis Manual
1. Introduction 7
1.1 Unit System / 9
1.2 File System / 10
1.3 Notation / 12
2. Node/DOF/Coordinate System 15
2.1 Nodal Points and Degree of Freedom / 15
2.2 Coordinate System / 17
2.3 Techniques of Finite Rotation / 20
3. Elements 23
3.1 Formalization of Finite Elements / 25
3.2 Shape Function / 27
3.3 Supplementation of Locking Phenomenon / 35
3.4 Rod Elements / 46
3.5 Bar Elements / 49
3.6 Pipe Elements / 56
3.7 Membrane Elements / 61
3.8 Shell Elements / 68
3.9 Surface Elements / 78
3.10 Plane strain Elements / 82
3.11 Axisymmetric solid Elements / 85
3.12 Solid Elements / 90
3.13 Layered Shell Elements / 97
3.14 Layered Solid Elements / 102
3.15 Other Elements / 108
3.16 Geometric Rigidity / 118
3.17 Thermal Conduction Elements / 122
3.18 Stress Error / 126
3
4. Material 129
4.1 Properties of Elastic Materials / 130
4.2 Properties of Plastic Materials / 135
4.3 Properties of Hyper-elastic Materials / 142
4.4 Properties of Conduction Materials / 149
4.5 Composite Material Lamination Theory / 150
4.6 Composite Material Failure Criteria / 153
5. Algorithm 157
5.1 Simulataneous Equations Solving Algorithm / 157
5.2 Eigenvalue Sampling Algorithm / 161
5.3 Effective Mass and Mode Reiteration Algorithm / 164
5.3.1 Effective Mass
5.3.2 Mode Superposition
5.4 Dynamic Response Algorithm / 169
5.4.1 Time Integration Algorithm
5.4.2 Frequency Response
5.4.3 Random respose
5.4.4 Response Spectrum
5.5 Non-linear Finite Elements Solving Algorithm / 188
5.6 Strain/Stress Calculation Algorithm Considering Large
Deformation / 194
5.7 Extensional Time Integration Algorithm / 199
5.8 Contact Conditions / 206
5.9 Fatigue Interpretation / 214
5.9.1 Stress-life Method
5.9.2 Strain-life Method
5.10 Stress Linear Interpolation / 224
5.11 Topology Optimizations / 228
5.11.1 Objective Functions and Sensitivity
5.11.2 Material Interpolation Scheme
5.11.3 Optimality Criteria
4
6. Load/Constraint 235
6.1 Loads / 235
6.2 Constraint Conditions / 240
6.3 Thermal Load/Boundary Conditions / 245
6.4 Pecularity Error / 250
6.5 Non-linearity of Load / 252
5
6
Introduction
1. Introduction
This manual introduces technical contents and theories behind the analysis using
midas NFX. The contents of this manual include detailed descriptions of all the
midas NFX’s functions, but the range of functions available for use in reality may
differ among various versions and modules. The manual is intended to present
the technical aspects required for the operation of and the theories implemented in
the development of midas NFX in order to enable the user to properly use and
understand midas NFX. However, it excludes very complex contents or any basic
materials that are discussed in detail in general engineering publications.
midas NFX is a general purpose finite element analysis program that has been
developed on the basis of C++, and it provides the following analysis fuctions:
Table 1.1 Analysis Functions of midas NFX
Linear static structural analysis

Modal analysis
Linear buckling analysis
Nonlinear steady-state heat transfer Analysis
Nonlinear transient heat transfer analysis
Linear transient analysis (direct/modal)
Frequency response analysis (direct/modal)
Response spectrum analysis
Nonlinear static structural analysis
Nonlinear dynamic analysis (explicit time integration)
Steady-state flow analysis (CFD)
Transient flow analysis (CFD)
Also, two or more analysis functions shown in Table 1.1 can be combined to handle
7
Analysis Manual
such analyses as prestressed analysis or thermal stress analysis.
The procedures used to solve problems using midas NFX and other general
purpose finite element analysis programs are as follows:
► Define Analysis Model (nodes, elements, mesh, loads, boundary conditions)

► Define Analysis Type (analysis cases)
► Execute Analysis (solve)
► Analyze Results (post-mode)
This manual centers around the algorithms, tehnical contents and theories that are
necessary to understand the above categories.
8
Introduction
1.1 Unit System
In order to define an analysis model, relavant information such as the sizes of

structures and the properties of materials are required. Such information
regarding the physical quantities are generally defined in a specific unit system.
midas NFX allows unit changes in force/length/energy/time at any time in the
process of defining the analysis model. Once the analysis model is defined, the
units for the force/ length/energy/time must be unified to be consistent with either
the English or SI unit system.
Table 1.1.1 Units used in English/SI Systems
Physical Quantity English SI

Location, Length, inch meter
Displacement
Modulus of Elasticity lbf/inch2 Newton/meter2
Moment inch-lbf Newton-meter
Force lbf Newton
Mass lbf-sec2/inch kilogram
Time second second
Stress lbf/inch2 Newton/ meter2
Inches and feet are widely used for the units for length in the English unit system.
The English unit system differs from the SI unit system in that it often uses weight
more than mass. In this case, the mass-weight relationship ( mass  weight / g )
based on gravity acceleration must be applied. midas NFX automatically applies
this relationship to analysis rendering appropriate result values that include a time
unit (example: Hz).
9
Analysis Manual
1.2 File System
midas NFX creates and saves a number of files in the process of a finite element
analysis during which various temporary files are created and deleted. The list of
files used in midas NFX and their descriptions are as follows:
Table 1.2.1 Main Files of midas NFX
File Name Format Description

ModelName.nfxa Binary Model Data File (analyst)
ModelName.nfxd Binary Model Data File (designer)
ModelName_AnalysisName.mec ASCII Interpreter Input File
ModelName_AnalysisName.log ASCII Analysis Execution Record File
ModelName_AnalysisName.out ASCII Analysis Results Data File
Analysis Results Data File
ModelName_AnalysisName.nfxp Binary
(for Post-processing)
The midas NFX’s result files consist of two types, the .out file in the ASCII format
and the .nfxp files in the binary format. The results demanded by the user are
basically included in the .nfxp file, which is used to analyze the results via the post-
processor. The .out file may include similar results as the .nfxp files, and the
inclusion is controlled by the user’s option.
The list of temporary files, which are generated during the analysis in midas NFX,
and their descriptions are as follows:
Table 1.2.2 Files generated during Analysis
File Name Time of Generation/Description
InputName.DASM#.bin Generated in all analyses, data related to finite

elements
InputName.FACT#.bin# Generated when the multi-frontal solver is selected,
matrix data
InputName.EIGS#.bin# Generated when eigenvalues are analyzed,
Lanczos’ iterative calculation data
10
Introduction
The temporary files that are generated during the analysis are saved in the scratch
folder whose basic values are identical to the location of the model file.
11
Analysis Manual
1.3 Notation
This manual, uses both the matrix and component notations. Because the matrix
notation is very useful in expressing the second order tensor, it will be used
whenever possible.
The values required to express theories include scalar, vector, second order tensor
or matrix and fourth order tensor. The matrix notation is expressed as follows:
Table 1.3.1 Matrix Notation
Scalar u
Vector u , 
2
nd
order tensor, Matrix A , *
4th order tensor C
When bold fonts are used, vectors and tensors are notated the same, the
distinction between which is based on their contents. The matrix notation is very
effective in grasping the physical meaning or the relationship among the physical
quantities. But when operations are too complicated to be expressed using the
matrix notation or operations among components are required, it is more helpful to
use the component notation. The component notation is based on coordinate
systems, each of which may be defined by base vectors. The base vectors are
written as ei , (i  1,2,3) in a three-dimensional space, and the base vectors may
not be orthogonal to one another. Using the base vectors, a specific vector, u ,
can be expressed as follows:
u  u1e1  u 2e2  u3e3 (1.3.1)

ei : Base vector
i
u : Component
In the component notation, the subscript i represents a covariant base vector or a
12
Introduction
i
component, and the superscript represents a contravariant base vector or a
component. As they are generally identical in an orthonormal coordinate system,
they are not differentiated.
When the component notation is used, it is more convenient to use the summation
convention for a repeated index as shown below.
u  u i ei (1.3.2)
Similarly, the second order and fourth order tensors can be expressed using the
component notation as follows:
A  Aijeie j , C  C ijkl eie jek el (1.3.3)
In this manual, the summation convention is assumed to have been applied unless
noted otherwise.
13
Analysis Manual
14
Node/DOF/Coordinate system
2. Node/DOF/Coordinate System
2.1 Nodal Points and Degrees of Freedom
Nodes and elements determine the size and shape of a finite element model and
are considered the starting point of all analyses. The model defined by nodes and
elements is equivalent to the physical phenomenon expressed in terms of
numerical equations of the matrix form. The unknowns in the matrix equation are
physical quantities such as displacement, rotation and temperature, which are
referred to as the degrees of freedom.
For example, 3 displacement and 3 rotational degrees of freedom are assigned to

a node in a structural analysis problem. These 6 degrees of freedom are shown
below.
2
u2
u1 1
u3
3
Figure 2.1.1 Displacement and rotational degrees of freedom in a

rectangular coordinate system
In general, each degree of freedom is expressed using the following symbols:

► DOF 1  T1  u1
► DOF 2  T2  u2
► DOF 3  T3  u3
► DOF 4  R1  1
15
Analysis Manual
► DOF 5  R2  2
► DOF 6  R3  3
Each node retains a coordinate system, which represents the kinematic directions,
and such system is referred to as the nodal displacement coordinate system. The
aforementioned degrees of freedom follow the directions of the nodal coordinate
system assigned to each node. All the nodes are basically defined to express the
kinematic directions in reference to the global coordinate system. The
temperature degree of freedom is not relevant to the nodal displacement coordinate
system since it has no direction.
16
2.2 Coordinate System
midas NFX uses the following types of coordinate systems:

► Rectangular coordinate system
► Cylindrical coordinate system
Cylindrical
Rectangular
coordinate system
coordinate system
z z
Origin Origin
y r
x

Figure 2.2.1 Rectangular and cylindrical coordinate systems
When the kinematic directions of a node are defined in the cylindrical coordinate
system, the degrees of freedom are as follows:
► DOF 1 = displacement in r  direction
► DOF 2 = displacement in   direction
► DOF 3 = displacement in z  direction
► DOF 4 = rotation about r  direction
► DOF 5 = rotation about   direction
► DOF 6 = rotation about z  direction
In order to use the finite element method to properly model and analyze a given
problem, various coordinate systems are required. Examples are one that is used
to define the aforementioned nodal displacement directions and one that is used to
17
Analysis Manual
specify the directions of an anisotropic material. A specific coordinate system is

also used to extract the result values. midas NFX uses the following coordinate
systems to model and analyze structures.
Table 2.2.1 Types of coordinate systems used in midas NFX
Types of Coordinate Systems Description
A single standard, rectangular coordinate

GCS: Global Coordinate System
system, which expresses the entire model
NDCS: Nodal Displacement Rectangular/cylindrical coordinate systems,

Coordinate System which depict the nodal kinematic directions
A rectangular coordinate system, which is

ECS: Element Coordinate System determined by the locations of nodes that
constitute an element
Rectangular/cylindrical coordinate systems,

MCS: Material Coordinate System which define the directions of a material
applied to elements
ERCS: Element Result Coordinate Rectangular/cylindrical coordinate systems for

System producing element results
EFCS: Element Formulation A coordinate system used for finite element

Coordinate System formulation, which is either GCS or ECS
The element formulation coordinate system is used in the solver and is irrelevant to
using midas NFX, but helps understand the contents of this manual. In case of
heat transfer analysis, the nodal degree of freedom is temperature. As such,
selection of a nodal displacement coordinate system has no effect on the analysis.
18
However, other coordinate systems such as the material coordinate system or the
element result coordinate system may affect the analysis and the results. The
element coordinate system, material coordinate system and element result
coordinate system are described in detail in Chapter 3.
w
v
y z
x y
u
y x
x
z
y
GCS x
Figure 2.2.2 Various coordinate systems used in midas NFX
19
Analysis Manual
2.3 Representation of Finite Rotation
Geometric nonlinear analysis, which includes finite rotation, requires a variable

representing the magnitude of rotation. For this reason, midas NFX uses the
rotation vector as the degrees of freedom of the corresponding node. In other
words, the values of the nodal degrees of freedom 4 to 6 generated as a result of
the geometric nonlinear analysis correspond to each component of the rotation
vector. The rotation vector, θ , consists of the magnitude, θ , and the direction,
e  θ / θ . Figure 2.3.1 shows the rotation vector, θ , which is rotaed about the e
axis by the angle, θ (radian).
θ Axis of rotation
θ
z e
θ
y
x
Figure 2.3.1 Direction and magnitude of a rotation vector
In finite rotation, it must be cautioned that a compound rotation, which consists of

sequential application of a number of rotation vectors, is not made up of the sum of
all the vectors. When the rotations of θ and θ are sequentially applied, the
final rotational value, θ* , retains the property of θ*  θ  θ . Also, the

commutative law does not hold true in this case. Figure 2.3.2 shows a different
rotational value when θ is applied prior to θ . A number of methods are used
to calculate the compound rotation, one of which is using the rotation matrix. midas
NFX calculates the rotation matrix using the quaternion product. The quaternion,
q , has the following relationship with the rotation vector, θ .
20
q  (q0 , q)  (cos( θ / 2), sin( θ / 2)e) (2.3.1)
The product of two quaternions can be calculated as follows:
q*  q q  (q0q0  q  q, q0q  q0q  q  q) (2.3.2)
q* : Quaternion corresponding to θ*
q : Quaternion corresponding to θ
θ (90) θ (90)
Initial configuration Final configuration

θ (90) θ (90)
Figure 2.3.2 Examples of compound rotations, which invalidate the

commutative law
21
Analysis Manual
22
Elements
3. Elements
For structures that are analyzed by the finite element method, it is essential to
understand the types and the characteristics of all available elements. The types
of elements provided in midas NFX can be categorized on the basis of their shapes
and characteristics as follows:
• Scalar Element
The scalar element retains one node, the motion of which is defined relative to the
ground point and retains either deformation or kinetic energy. It is possible to
define the scalar element by two nodes, but the shape information such as the
distance between the nodes is not used.
• 1-Dimensional Shape
It retains two nodes in a straight line shape and uses the shape information such as
the relative distance between the nodes.
It retains 3, 4, 6 or 8 nodes in a triangular or rectangular shape, which may be of
curvature in space.
It retains 4, 5, 6, 8, 10, 13, 15 or 20 nodes in a shape of a tetrahedron, pentahedron
or hexahedron. A pentahedron element can form a wedge or a pyramid shape.
• Special Element
It is a special purpose element, and midas NFX provides lumped mass.
• Rigid Body/Interpolation Element

It is an element that can define the rigid body motion or relative motion by
interpolation between nodes. It retains similar characteristics as the multi-point
23
Analysis Manual
constraint.
• Joint Element
It is used to simulate various relative motions between two points. A ground point
– node or node – node constitute the two nodes.
Table 3.1 Classification based on the characteristics of elements
Element Type Characteristics

Spring (1/2 nodes)
Scalar Element Mass (1/2 nodes)
Damper (1/2 nodes)
Rod (2 nodes)
Bar (2 nodes)
1-Dimensional Shape
Pipe (2 nodes)
Bush (1/2 nodes)
Membrane (3/4/6/8 nodes)
Plane Stress Shell (3/4/6/8 nodes)
2-Dimensional Shape Surface (3/4/6/8 nodes)
Plane strain (3/4/6/8 nodes)
3-Dimensional Stress
Axisymmetric solid (3/4/6/8 nodes)
3-Dimensional Shape Solid (4/5/6/8/10/13/15/20 nodes)
Special Element Concentrated mass
Rigid Body / Rigid element (rigid body, rigid bar),
Interpolation Element Interpolation element
Joint Element Classified by joint types (1/2 nodes)
The elements listed in Table 3.1 are categorized only on the basis of structural
analysis perspective. The physical behaviors reflected in each element in heat
transfer analysis are explained in 3.17.
24
Elements
3.1 Finite Element Formulation
The variation theory, which includes all the equations based on the theory of linear
elasticity, is known as the Hu-Washizu’s 1 , 2
variation principle expressed as
follows:
 Gext   (T  u)σ   εT (Dε  σ)   σT (u  ε)d  (3.1.1)


 Gext : Virtual work due to an external force

u : Displacement
σ : Stress
ε : Strain
D : Stress-strain relationship matrix
 : Strain-displacement relationship operator
The above equation is in the most general form, which includes the equilibrium
equation, constitutive equation and compatibility condition. Assuming that the
relationship between strain, ε , and stress , σ , is always satisfied by the
constitutive equation, the equation becomes the Hellinger-Reissner’s3,
4
principle
as follows:
 Gext   (T  u)σ   σT (u  D1σ)d  (3.1.2)


In addition, assuming that the relationship between ε and u is satisfied by the
1
Hu, H.C., “On some variational principles in the theory of elasticity and the theory of plasticity,” Scintia Sinica, Vol. 4, 1955
2
Washizu, K., On the Variational Principles of Elasticity, Aeroelastic and Structural Research Laboratory, MIT, Technical Report,
1955
3
Hellinger, E., “Der allgemeine Ansatz der Mechanik der Kontinua,” Encyclopadie der Mathernafischen Wissenschaften, Vol. 4,
1914
4
Reissner, E., “On a variational theorem in elasticity,” Journal of Mathematical Physics, Vol. 29, 1950
25
Analysis Manual
compatibility condition, the equation becomes the general principle of virtual work.
 Gext   (T  u)Dud  (3.1.3)


Let us consider the integral field limited to one element in order to apply the finite
element method to the minimum potential energy principle. If the displacement, u ,
is interpolated by a shape function within a single element, it can be expressed as,
uh  Nde (3.1.4)
N : Shape function
e
d : Degrees of freedom of element
Using the strain-displacement relationship, εh  uh  Bde , the principle of virtual

work can be expressed as follows:
 Gext   dT F   dT   BT DBd  d   dT Kd (3.1.5)

 e 
This equation is suitable for the analysis of elastic structures exhibiting infinitesimal
displacements, and K is independent of the displacement, d , in linear analysis.
The stiffness of an individual element is expressed as K e , shown as,
K e   BT DBd  (3.1.6)
e
The finite element formulation process used in heat transfer analysis is explained in
3.17.
26
Elements
3.2 Shape Function
Defining an element starts from the hypothesis of the displacement field based on
the shape function. The displacement field’s hypothesis is also required for the
element to which numerous element performance-enhancing methods such as a
mixed formulation are applied. Unless noted otherwise, the displacement field
described in this section identically applies to the element required for heat transfer
analysis. The index described in this section, which is used to easily express the
shape function, does not follow the summation convention.
• 1-Dimensional shape
► 2-Node shape function
1  i
Ni  , 1    1
2
► 2-Node Hermite shape function

N1  1  3 2  2 3 , N2  l  2l 2  l 3 , N3  3 2  2 3 , N4  l 2  l 3 , 0    1
l : Element length
► 3-Node triangle
N1  1     , N2   , N3  
► 6-Node triangle
N1  1     1  2  2  , N2    2  1 , N3    2  1
N4  4 1      , N5  4 , N6  4 1     
27
Analysis Manual
(0, 1)
3
5
6
1
2 
(0, 0) 4 (1, 0)
Figure 3.2.1 Node locations of a triangular element

and natural coordinate system
► 4-Node rectangle
1
Ni  1  i 1  
i 
4
► 8-Node rectangle
1
Ni  1  i 1  
i  (i  i  1)
, i  1,2,3,4
4
Ni 
1
2
 
1   2 1  
i 
, i  5,7
Ni 
1
2
 
1   2 1  i  , i  6,8
28
Elements
(1, 1)
4 3
7
8  6
1 5 2
(-1, -1)
Figure 3.2.2 Node locations of a rectangular element

► 4-Node tetrahedron
N1  1       , N2   , N3   , N3  
► 10-Node tetrahedron
1  1 1
N1  2(1       )(       ) , N 2  2     , N3  2 (  ) ,
2  2  2
1
N 4  2 (  ) N5  4 (1       ) , N6  4 , N7  4 (1       ) ,
2 ,
N8  4 (1       ) , N9  4 , N10  4
29
Analysis Manual
4 (0, 0, 1)
10
8
9
(0, 0, 0) 7 3

1 (0, 1, 0)
5
6
2
(1, 0, 0)

Figure 3.2.3 Node locations of a tetrahedron element and natural coordinate system
► 6-Node pentahedron
1
Ni  (1     )(1   i ) , i  1,4
2
1
Ni   (1   i ) , i  2,5
2
1
Ni   (1   i ) , i  3,6
2
1
Ni  1      (1   i )( i  2  2 ) , i  1,4
2
1
Ni   (1   i )( i  2  2) , i  2,5
2
30
Elements
1
Ni   (1   i )( i  2  2) , i  3,6
2
Ni  2 (1     )(1   i ) , i  7,13
Ni  2 (1   i ) , i  8,14
Ni  2 (1     )(1   i ) , i  9,15
N10  (1     )(1   2 ) , N11   (1   2 ) , N12   (1   2 )


15
4 (0, 0, 1)
6
14
13
(0, 0, 0) 12 
5 10
1 9 3
11 (0, 1, -1)
7
8

2
(1, 0, -1)
Figure 3.2.4 Node locations of a pentahedron (wedge) element

A 5-node or a 13-node pentahedron retains the pyramid shape and widely uses the
degenerated shape function based on combining nodes. But this shape function
5
is known to have problems in numerical integration, and as such midas NFX uses
the following forms:
5
Bedrosian, G., “Shape functions and integration formulas for three-dimensional finite element analysis,” International Journal f
or Numerical Methods in Engineering, Vol. 35, 1992
31
Analysis Manual
1 
Ni  {(1  i )(1  
i )    ii } , i  1,2,3,4
4 1
N5  
1 
Ni  (i  
i  1){(1  i )(1  
i )     } , i  1,2,3,4
1 
i i
4
(1     )(1     )(1  
i  )
Ni  , i  6,8
2(1   )
(1     )(1     )(1  i   )

Ni  , i  7,9
2(1   )
 (1  2i   )(1  2

i  )
Ni  , i  10,11,12,13
(1   )
N5   (2  1)
5
(0, 0, 1)
12
13
10 11
4 3
8
 
(1, 1, 0)
9 (0, 0, 0)

7
1 (-1, -1, 0) 6 2
Figure 3.2.5 Node locations of a pentahedron (pyramid) element

32
Elements
► 8-Node hexahedron
1
Ni  (1  i )(1  
i )(1   i ) , i  1,2,3,...,8
8
► 20-Node hexahedron
1
Ni  (1  i )(1  
i )(1   i )(i  
i   i  2) , i  1,2,3,...,8
8
1
Ni  (1   2 )(1  
i )(1   i ) , i  9,11,17,19
4
1
Ni  (1   2 )(1  i )(1   i ) , i  10,12,18,20
4
1
Ni  (1   2 )(1  i )(1  
i ) , i  13,14,15,16
4
8 19 7
20 
17 6 18
5
 15
16

13 14
12 4 11 3
10
1 9 2
Figure 3.2.6 Node locations of a hexahedron element and natural

coordinate system
In order to use the above shape functions and apply them to the formulation
process discussed in 3.1, numerical integration is required. The numerical
integration is necessary to calculate the stiffness matrix, mass matrix, load vectors,
internal forces, etc. midas NFX uses two numerical integration methods, Gauss
integration method and Lobatto integration method.
33
Analysis Manual
Table 3.2.1 Types of numerical integration methods and applicable elements
Numerical
Integration Matrix Type Applicable Elements
Method
All elements using numerical

Structural Element
Stiffness integration
Matrix Heat Transfer All elements using numerical

Gaussian
Element integration
Quadrature
Consistent Mass All elements
Mass
All elements using diagonal
Matrix Lumped Mass
scaling6
Structural Element -
Stiffness
Heat Transfer
Matrix -
Element
Consistent Mass -
Lobatto
3-node triangle
Quadrature
Mass 4-node rectangle
Matrix Lumped Mass 4-node tetrahedron
6-node pentahedron
8-node hexahedron
6
Hinton, E., Tock, T. and Zienkiewicz, O.C., “A Note on mass lumping and related processes in the finite element method,” Ea
rchquake Engineering and Structural Dynamics, Vol. 4, 1976
34
Elements
3.3 Compensation for locking phenomenon
Elements formulated by only the assumed displacement method are generally

known to produce very low accuracy in results. The reason for that is the locking
phenomenon, and it is very important to increase the accuracy in solutions for the
applicability of a finite element program. midas NFX uses the following methods
to enhance the precision of each element. Each method is not used
independently, but two or more methods may be used in combination depending on
the element in question.
• Mixed Hybrid Formulation

The mixed hybrid formulation, combined with the variation theory or displacement,
can be categorized in many different ways depending on the assumed component.
midas NFX uses the assumed stress method and the mixed u-p method.
The variation equation whose unknowns are displacement and stress based on the
Hellinger-Reissner principle is as follows:
 Gext   (T  u)σ   σT (u  D1σ)d  (3.3.1)


If the above equation is substituted with the displacement, uh  Nde and the stress,
σ  Pβe according to the shape function of an element, the right hand side of the
equation becomes as follows:
 deT QT βe   βeT (Qde  Pβe ) (3.3.2)
where, Q and P are expressed as,
Q   PT Bd e (3.3.3)
e
P   PT D1Pd e (3.3.4)
e
35
Analysis Manual
Since β e is assumed to be discontinuous between elements, it may be eliminated

in the element as follows:
βe  P1Qde (3.3.5)
When this equation is substituted into Equation (3.3.2), the element stiffness
becomes as follows:
K e  QT P1Q (3.3.6)
The most important aspect that influences the performance of an element is the
proper selection of the function, P , which assumes the stress7.
 xx  1 0 0  0 
 
σ   yy   Pβ  TPβ  T 0 1 0 0   β
ˆ (3.3.7)
  0 0 1 0 0 
 xy 
T , in this case, is the coordinate transformation matrix of the contravariant stress

components expressed as,
 j112 j212 2 j11 j21    

  
σ  Tσˆ   j122 j222 2 j12 j22     (3.3.8)
 j11 j12
 j21 j22 j11 j22  j12 j21    
Each component of the transformation matrix is calculated from Jacobian, and the
value at the center of the element is primarily used.
7
Pian, T.H.H. and Sumihara, K., “Rational approach for assumed stress finite elements,” International Journal for Numerical Met
hods in Engineering, Vol. 20, 1984
36
Elements
 x y 
     j11 j12 
J 
y   j21 j22 
(3.3.9)
 x
   

The mixed u-p method assumes only the hydrostatic stress or pressure, p rather
8
than assuming all the components of stress, σ . As such, the method has been
used to resolve the locking phenomenon in incompressible materials. As shown
below, the stress tensor is decomposed into deviatoric stress and pressure and
uses the Hu-Washizu variation principle.
σ  σ dev  pI
(3.3.10)
p  Ktr (ε)
σ dev : Deviatoric stress

K : Bulk modulus of elasticity
tr ( ) : Trace
• ANS: Assumed Natural Strain

Since ANS does not deviate too much from the classical displacement assumption
method, the assumed natural strain has been widely used due to the ease of
9, 10, 11
application. Especially many applications can be found for shell elements .
In theory, the assumed natural strain is founded on the Hu-Washizu principle, but
when it is applied to the finite element method, it can be considered as a type of the
B-bar method.
8
Zienkiewicz, O.C., Rojek, J., Taylor, R.L. and Pastor, M., “Triangles and tetrahedral in explicit dynamics codes for solids,” Comp
uter Methods in Applied Mechanics and Engineering, Vol. 43, 1998
9
MacNeal, R.H., “Derivation of element stiffness matrices by assumed strain distribution,” Nuclear Engineering and Design, Vol.
70, 1982
10
Hughes, T.J.R. and Tezduyar, T.E., “Finite elements based upon Mindlin plate theory with particular reference to the four-nod
e bilinear isoparametric element,” Journal of Applied Mechanics, Vol. 48, 1981
11
Bathe, K.J. and Dvorkin, E.N., “A formulation of general shell elements-The use of mixed interolation of tensorial component
s,” International Journal for Numerical Methods in Engineering, Vol. 22, 1986
37
Analysis Manual
 Local transverse direction
D
4 3
x
x Integration
point x
A
C  x
1 2
B
Figure 3.3.1 Assumption of shear strains in transverse directions
For example, let us apply the assumed natural strains to the shear strains in the
transverse directions of a 4-node shell element as shown above. Among the
natural coordinate system components,   Z is known to be accurate at B and D
while   Z is accurate at A and C. Using these values, the strains at the
integration points are interpolated as follows:
1 1
z  (1   ) Bz  (1   ) Dz (3.3.11)
2 2
1 1
 z  (1   ) Az  (1   ) Cz (3.3.12)
2 2
The strains in a natural coordinate system may be transformed into a space

coordinate system using the following transformation equation:
 xz    z 
γ     T T   (3.3.13)
 yz    z 
T above is the coordinate transformation matrix of the following covariant

components:
38
Elements
j j21 
T   11
j22 
(3.3.14)
 j12
Each component of the transformation matrix is calculated from Jacobian. The

assumed natural strain method only changes the strains using Equations (3.3.11) –
(3.3.13), so it is equivalent to modifying the B matrix by the classical
displacement assumption method as follows:
ε  u  Bde (3.3.15)
• EAS: Enhanced Assumed Strain

The enhanced assumed strain is very similar to the method using the incompatible
mode12 and produces identical results. But the difference is that it is theoretically
founded on the Hu-Washizu principle and starts from assuming strain, not
displacement13. The following variation equation of Hu-Washizu assumes the 3
components (displacement, strain and stress) as unknowns:
 Gext   (T  u)σ   εT (Dε  σ)   σT (u  ε)d  (3.3.16)


Strain, ε , is assumed to be the sum of the compatible term calculated from

displacement and the incompatible term (enhanced assumed strain).
ε  u  ε (3.3.17)
If the above equation is substituted into Equation (3.3.16), it becomes,
 Gext   (T  u)D(u  ε )   ε T (Du  Dε  σ)   σT εd  (3.3.18)


12
Taylor, R.L., Beresford, P.J. and Wilson, E.L., “A non-conforming element for stress analysis,” International Journal for Numeric
al Methods in Engineering, Vol. 10, 1976
13
Simo, J.C. and Rifai, M.S., “A class of mixed assumed strain methods and the method of incompatible modes,” International
Journal for Numerical Methods in Engineering, Vol. 29, 1990
39
Analysis Manual
If the stress distribution and the incompatible term of the strain are assumed to be
orthogobnal, the equation then consists of only the displacement and the enhanced
strain.
 Gext   T  uD(u  ε )   ε T (Du  Dε )d  (3.3.19)


If the displacement by the shape function of an element and the enhanced strain
are assumed as uh  Nde and ε  Gαe respectively, substituting them will lead the
above equation to,
 deT K edd de   deT K ed αe   αeT Ke d de   αeT K

e
αe (3.3.20)
K edd represents the classical element stiffness by a displacement assumption, and

K ed and K 
e
are as follows:
K ed   BT DGd e (3.3.21)

e
e
K   GT DGd e (3.3.22)
e
With the assumption that α e is discontinuous between elements, and since no

work is done by external forces for  α e , it can be eliminated with an element as
follows:
e 1 e
αe  K K d d e (3.3.23)
Substituting this equation into Equation (3.3.20), the element stiffness becomes as
follows:
40
Elements
e 1 e
K e  K edd  K ed K K d (3.3.24)
The most crucial aspect in determining the performance of an element is to select

an appropriate function, G , which assumes the enhanced strain.
• Linked Interpolation
The linked interpolation is usually used in shell elements and has two cases for
14
performance enhancement : one in the transverse direction and the other in the in-
plane direction. In particular, the linked interpolation for the in-plane direction
performance enhancement is well known as the membrane element, which
considers the Allman’s drilling rotation.15 midas NFX uses the linked interpolation
of a shell element, which retains 6 degrees of rotational freedom as proposed by
Sze.16

4
θ1
u5
5 
2
θ2
Figure 3.3.2 Nodal rotations and higher-order shape function
14
Zienkiewicz, O.C., Xu, Z., Zeng, L.F., Samuelsson, A. and Wiberg, N., “Linked interpolation for Reissner-Mindlin plate elements
: Part I – a simple quadrilateral,” International Journal for Numerical Methods in Engineering, Vol. 36, 1993
15
Allman, J., “A compatible triangular element including vertex rotations for plane elasticity analysis,” Computers and Structure
s, Vol. 19, 1984
16
Sze, K.Y., Sim, Y.S. and Soh, A.K., “A hybrid stress quadrilateral shell element with full rotational D.O.F.S,” International Journa
l for Numerical Methods in Engineering, Vol. 40, 1997
41
Analysis Manual
For example, let us apply the linked interpolation to the displacement field of a 4-
node shell element as shown above. If the displacement of the element’s neutral
plane is separated into the part of the general 4-node shape function and the part of
the nodal rotations, it is expressed as follows:
 u   uI NI   y(j  z j  )z i z (j i   yi

yj )
 I 1, 2 , 3 , 4
i
    1 
u   v    vI NI   HJ z(ji x j )x i x( ji z j) , z i (3.3.25)
 w   w N 8  x(   ) y (   ) J  5 , 6 , 7 , 8
   I I  ji yj yi ji xj xi
NI : Shape function
HJ : Higher-order (8 node) shape function
uI , vI , wI : Nodal displacement
 j : Nodal rotation
x ji , y ji , z ji : Distance between nodes
( x ji  x j  xi , y ji  y j  yi , z ji  z j  zi )
The node numbers, i, j , represent the nodes adjacent to the mid node, J , of a
side, and have the relationships, i  mod( J  1,4)  1 and j  mod( J ,4)  1 .
Ultimately, the linked interpolation assumes the displacement of the side’s mid node
using the nodal rotations of a lower-order element and applies this value to the
higher-order shape function. Since the deformed shape of each side retains the
secondary characteristic, the linked interpolation eliminates the locking
phenomenon and also satisfies the compatibility conditions with the adjacent
elements.
• Reduced Integration Method

The strains at the integration points to which a lower order of integration is applied
are known to be more accurate than values at other points. 17, 18 Also, the
17
Barlow, J., “A stiffness matrix for a curved membrane shell,” Conf. Recent Advances in Stress Analysis, Royal Aeron. Soc., 19
68
18
Barlow, J., “Optimal stress locations in finite element models,” International Journal for Numerical Methods in Engineering. V
ol. 10, 1976
42
Elements
locking phenomenon in an element occurs generally due to unnecessary orders of

the strains, so the higher-order deformation shapes can be eliminated through
numerical integration. However, reduced integration sometimes deteriorates the
numerical characteristics of the stifffness matrix, which may induce the spurious
zero energy mode (hourglass mode).
•Stabilization Technique for Reduced Integration

Generally, the strains of a 3-dimensional lower-order element can be approximated
as,
ε  u  (B0  B1 ( ,, ))de (3.3.26)
The method of eliminating the element’s locking phenomenon by applying lower

order integration is identical to using only B 0 , which in turn is the same as
evaluating the strains at the center of the element. When the strains at the center
of the element are used, there is a problem of not being able to satisfy the patch
test in which a uniform deformation is applied. In order to resolve this, B 0 is
generally substituted with B 0 .19 B 0 , which corresponds to an average strain,
satisfies the following equation:
1
Ve e
B0  Bd e (3.3.27)
If only B 0 corresponding to the average strain is used, the deformation energy for
the aforementioned spurious zero energy mode is not considered. Since this
phenomenon is very severe in a lower-order element, a stabilization technique
becomes necessary. Various methods exist in the stabilization technique called
“hourglass control”, and midas NFX uses the physical stabilization technique
19
Flanagan, D.P. and Belytschko, T., “A uniform strain hexahedron and quadrilateral with orthogonal hourglass control,” Interna
tional Journal for Numerical Methods in Engineering, Vol. 17, 1981
43
Analysis Manual
20
proposed by Puso. B1 is expressed in the natural coordinate system in order to
calculate the stabilized strains.
B1   B  B   B  B   B   B (3.3.28)
If all of the strains in the above equation are used, the effect of the reduced
integration disappears. As such, it is common to exclude some shear strain terms.
When the average strain and the stabilization technique are applied, the identical
effect of selective reduced integration can be gained. In addition, the calculation
speed significantly improves since the numerical integration process is replaced
with the equation below.
Ve
e
[]d e 
8 
[]d d d (3.3.29)
• Non-comforming Element
A non-conforming element uses a method of decomposing strains in order to satisfy
the compatibility conditions between elements into an integral form. The
aforementioned EAS method can be also considered as a type of non-conforming
elements. Generally, the compatibility conditions between elements in an integral
form can be expressed as,
e
ui*, j d    ui n j dS
e
(3.3.30)
u* : Assumed displacement inside an element
u : Assumed displacement on the outer face of an element
nj : Vector normal to the outer face of an element (direction cosine)
j : j -direction differential
20
Puso, M.A., “A highly efficient enhanced assumed strain physically stabilized hexahedral element,” International Journal for N
umerical Methods in Engineering, Vol. 49, 2000
44
Elements
The displacement assumed inside an element consists of the part using a general
shape function and other additional parts.
u*  Nde  Pλ (3.3.31)
The displacements assumed on the outer surfaces of an element also consist of

the part using a general shape function and other additional parts. The difference
is that the assumed displacements are expressed in the form of interpolation of the
nodal displacements.
u  Nde   Mde (3.3.32)

M : Added shape function (  N )
In the above equation,  is an arbitrary coefficient, a most appropriate value of

which is based on the element’s convergency or accuracy of the solution.
Substituting Equations (3.3.31) and (3.3.32) into (3.3.30), λ can be calculated
using d e . Using the calculated λ , the strain of an element is expressed as,
ε  Bde  (P)λ (3.3.33)
A non-conforming element is equivalent to modifying the B matrix in the

displacement assumption method as,
ε  u  Bde (3.3.34)
45
Analysis Manual
3.4 Rod Element
A rod element is a 1-dimensional element defined by two nodes. It is generally

used to model members such as space trusses or a diagonal braces whose lengths
are relatively long compared to their cross-sectional areas.
M x , x
N xx ,  xx
ECS  x
N xx ,  xx
M x , x 1
Figure 3.4.1 Coordinate system of a rod element and stresses/strains
• Coordinate System
In the element coordinate system (ECS) of a rod element, the direction of the x-axis
points from Node 1 to 2. Its finite element is formulated in reference to ECS.
• Degrees of Freedom
A rod element retains the displacement and rotational degrees of freedom in and
about the x-axis direction in ECS respectively.
ui  ui  , θi   xi  (3.4.1)
• Stresses and Strains

A rod element expresses axial deformation and torsion defined in ECS as shown in
Figure 3.4.1.
N  N xx  , ε   xx  (Axial force and strain) (3.4.2)
46
Elements
T  M x  φ  x 
, (Torsional moment and strain) (3.4.3)
• Loads
The types of loads applied to a rod element are as follows:
Table 3.4.1 Types of loads applied to a rod element
Load Types Descriptions
Gravitational force Applied to the density of materials and nonstructural masses
Rotational inertia force Applied to the density of materials and nonstructural masses
• Element Results
When a rod element is used, the element result items are as shown below, and the
reference coordinate system is always in ECS.
Table 3.4.2 Result items of a rod element
Result Items Descriptions
Axial stress Location: at the center of element,  xx
Location: at the center of element

Stress
Torsional stress Calculated from the torsional stress
coefficient, c (   Tc / J )
Axial strain Location: at the center of element,  xx

Strain
Torsional strain Location: at the center of element
Axial force Location: at the center of element, N xx

Force/Moment
Torque Location: the center of element, M x
Strain energy Location: at the center of element
Misc. Total percent energy Location: the center of element
Energy density Location: at the center of element
47
Analysis Manual
• Nonlinear Analysis
A rod element can consider geometrical nonlinearity and may be of an elasto-
plastic material. The result items of nonlinear analysis are as follows:
Table 3.4.3 Result items of a rod element in nonlinear analysis

Equivalent stress Caclulated on the basis of plastic models,
Stress  eq

Plastic status
Elasticity/Plasticity, 0 / 1

Equivalent strain
Calculated on the basis of plastic models,  eq
Strain
Effective plastic strain
ep
48
Elements
3.5 Bar Element
A bar element is a 1-dimensional line element defined by two nodes. It is mainly

used when a long member relative to its cross-sectional dimensions is subjected to
bending deformation. If the ratio of the length to the width or the height of the
cross-section is less than about 5, the effects of shear deformation become greater.
In such a case, it is recommended that shell or a solid elements be used instead of
a bar element.
In the element coordinate system (ECS) of a bar element, the direction of the x-axis
points from Node 1 to 2. The directions of y-axis and z-axis in ECS are
determined using a reference node or a reference vector. Figure 3.5.1 shows the
method of determining the x-y plane using a reference node. In such a case, the
reference node must not be on the x-axis in ECS. Figure 3.5.2 shows the method
of determining the x-y plane using a reference vector, which also cannot be defined
parallel to the x-axis in ECS. Its finite element is formulated in reference to ECS.
ECS  z
ECS  y
ECS  x
Reference node 2
1
z
y
GCS x
Figure 3.5.1 ECS of a bar element defined by a reference node
49
Analysis Manual
ECS  z
ECS  y
ECS  x
1
z
y
GCS x
Figure 3.5.2 ECS of a bar element defined by a reference vector
A bar element retains the displacement and rotational degrees of freedom in and
about all the axis directions in ECS respectively.
ui  ui  
wi  , θi   xi  yi  zi
T T
vi (3.5.1)

A bar element can consider axial and shear deformations, bending and torsion
defined in ECS as shown in Figure 3.5.3. If the Euler theory is used in which no
shear deformations are considered, 0 is specified for the shear area factors.

M   
M   y  , κ   y  (Bending moment and curvature) (3.5.3)
M z   z 
T  M x  , φ  x  (Torsional moment and strain) (3.5.4)

Q   
Q   y  ,    xy  (Shear force and strain) (3.5.5)
Qz   zx 
50
Elements
M z , z
ECS  z
Qz ,  zx
ECS  y ECS  x
M x , x
N xx ,  xx
2
Qz ,  zx
Qy ,  xy
My, y
M z , z
N xx ,  xx 1
Qy ,  xy
M x , x
My, y
Figure 3.5.3 Coordinate system of a bar element and stresses/strains
• Loads
The types of loads applied to a bar element are as follows:
Table 3.5.1 Types of loads applied to a bar element
Distributed load acting on a specific range between two nodes

Bar element load of an element or concentrated load acting on a specific location
between two nodes of an element
Cross-sectional average temperature that causes deforomation

Bar element
in the axial direction
temperature load
Cross-sectional temperature gradient that causes bending
Bar element pre-load Load that causes initial axial force (bolt load)
As shown in Figure 3.5.4, bar elements can be loaded in the form of a concentrated
load or a distributed load in either element coordinate system ECS or global
51
Analysis Manual
coordinate system (GCS). When the load direction is specified in GCS, the
magnitude can be scaled using the angle between the x-axes of GCS and ECS
(Figure 3.5.5).
p2 m2
p1 m1
1 2
P M
1 2
Figure 3.5.4 Examples of bar element loads
z  L  L
GCS x
Figure 3.5.5 Scaling bar element loads based on load directions
• Element Results
When a bar element is used, the element result items are as shown below, and the
reference coordinate system is always in ECS.
52
Elements
Table 3.5.2 Result items of a bar element
Axial stress Location: A-B ends  xx

Location: A-B ends
Torsional stress
Calculated from the torsional stress coefficient c (   Tc / J )
Location: A-B ends
Point stress
Stresses due to bending at user-defined locations (C, D, E, F)  xx
Stress
Location: A-B ends
Max/Min stress Max/Min of the sum of axial stress and point stress among the
locations C – F
Bar Max/Min
Max/Min of the Max/Min stresses at A-B ends
Axial strain Location: A-B ends  xx
Torsional strain Location: A-B ends

Location: A-B ends
Point strain
Strains due to bending at user-defined locations (C, D, E, F)  xx
Strain Location: A-B ends
Max/Min strain Max/Min of the sum of axial strain and point strain among the
locations C – F
Bar Max/Min
Max/Min of the Max/Min strains at A-B ends
Axial force Location: A-B ends N xx
Force/ Bending moment Location: A-B ends My , Mz
Moment Torque Location: A-B ends Mx
Shear force Location: A-B ends Qy , Qz
Total percent
Misc. Location: at the center of element
energy
53
Analysis Manual
Mz
ECS  z
ECS  y ECS  x Qz
Mx
N xx
B
Qz
My Qy
Mz
N xx A
ECS  y
Qy
Mx
My F C
Recovery point ECS  z

(I-section)
E D
Figure 3.5.6 Result output locations and directions of a bar element
• End Release
End release condition is used when mutual constraint is absent for a motion in a
specific direction such as a pin connection. Because end release is defined in
ECS, the relationship between the coordinate systems must be examined and used
if a connection release needs to be applied in GCS. Also, releasing nodes
generates additional degrees of freedom without constraints at the corresponding
nodes, which may have an effect on the stability of the entire structure.
Rotation DOF
released Sliding joint
Pin joint
Translation DOF
released
Figure 3.5.7 Examples of end releases
54
Elements
• Offset
An offset can be used when the neutral axis of a bar element is separated from a
node or when the neutral axes of the connecting elements are not coincident. An
offset is applied in nodal coordinate system (NCS) specified at a node of a bar
element. If the offset is specified in the axial direction of the element, the element
length is considered to have been changed.
Figure 3.5.8 Examples of offsets
A bar element can only consider geometric nonlinearity and cannot use nonlinear or
inelastic materials. Therefore, there are no additional result items associated with
nonlinear analysis.
55
Analysis Manual
3.6 Pipe Element
A pipe element is a 1-dimensional line element defined by two nodes. Although it

retains similar properties as a bar element, it can also consider internal pressure
within a pipe. The cross-section of a pipe element is assumed to be in the shape
of a cylinder, as shown in Figure 3.6.1.
Wall thickness ( twall )
Outer diameter (d out )
Figure 3.6.1 Definition of cross-section of a pipe element
The ECS of a pipe element is defined in the same manner as that of a bar element,
and the finite element is formulated in reference to ECS.
A pipe element retains displacement and rotational degrees of freedom in all the
axis directions in ECS.
ui  ui 
wi  , θi   xi  yi  zi 
T T
vi (3.6.1)

Identical to a bar element, a pipe element can consider axial and shear
deformations, bending and torsion defined in ECS.
56
Elements
M   
M   y  , κ   y  (Bending moment and curvature) (3.6.3)
M z   z 
T  M x  , φ  x  (Torsional moment and strain) (3.6.4)
Q   
Q   y  ,    xy  (Shear force and strain) (3.6.5)
 z
Q  zx 
• Loads
The types of loads applied to a pipe element are as follows:
Table 3.6.1 Types of loads applied to a pipe element
Distributed load acting on a specific range between two nodes

Bar element load of an element or concentrated load acting on a specific
location between two nodes of an element
Cross-sectional average temperature that causes

Bar element
deforomation in the axial direction
temperature load
Cross-sectional temperature gradient that causes bending
Pipe element internal Load that causes deformation of a pipe in the circumferential
pressure or longitudinal direction
57
Analysis Manual
The effects of pipe element’s internal pressure on the deformation of a structure

vary with whether or not the ends of the element are closed. The equation below
shows the stress in the element’s circumferential direction generated by internal
pressure.
 hoop 

0
rm P sin  d
(3.6.6)
2twall
P : Internal pressure of a pipe element

rm : Average radius of a pipe (dout  twall ) / 2
Stress in the circumferential direction creates an initial stress in the longitudinal

direction based on the Poisson’s ratio, and its magnitude is as follows:
 L   hoop (3.6.7)
If the ends of a pipe element are closed, the internal pressure load acts additionally
in the longitudinal direction.
FL   rm2 P (3.6.8)
• Element Results
When a pipe element is used, the element result items are as shown below, and
the reference coordinate system is always in ECS.
Table 3.6.2 Result items of a pipe element
Location: A-B ends

Longitudinal stress The sum of bending stress (maximum bending direction) and
Stress longitudinal stress (  xx )
Torsional stress
Shear stress due to torsion (  x )
58
Elements
Location: A-B ends

Hoop stress
Stress due to internal pressure (   )
Location: A-B ends
Max principal stress
Calculated using 3 stress components ( P1 )
Location: A-B ends
Max shear stress
Calculated using 3 stress components (  max )
Location: A-B ends
Von-mises stress
Calculated using 3 stress components (  v )
Location: A-B ends
Longitudinal strain The sum of bending strain and longitudinal strain (  xx )
Strain due to temperature gradient is not reflected
Torsional strain
Shear strain due to torsion (  x )
Location: A-B ends
Hoop strain
Strain Strain due to internal pressure (   )
Location: A-B ends
Max principal strain
Calculated using 3 strain components ( E1 )
Location: A-B ends
Max shear strain
Calculated using 3 strain components (  max )
Location: A-B ends
Von-mises strain
Calculated using 3 strain components (  v )
Location: A-B ends

Axial force
N xx
Location: A-B ends

Bending moment
Force/Mo My , Mz
ment Location: A-B ends
Torque
Mx
Location: A-B ends
Shear force
Qy , Qz
Misc. Total percent energy Location: at the center of element
59
Analysis Manual
A pipe element can only consider geometric nonlinearity and cannot use nonlinear
or inelastic materials. Therefore, there are no additional result items associated
with nonlinear analysis.
60
Elements
3.7 Membrane Element
A membrane element is a triangular or a quadrilateral element consisted of 3, 4, 6

or 8 nodes located on a plane. Membrane elements are mainly used to model a
thin plate of a uniform thickness, and the elements retain a 2-dimensional stress
condition.
In the element coordinate system (ECS) of a triangular membrane element, the z-
axis is defined in the direction perpendicular to the element plane, and the x-axis is
defined in the direction from Node 1 to Node 2. The z-axis of a quadrilateral
membrane element is also defined in the direction perpendicular to the element
plane. But the x-axis is determined in the direction that bisects the angle formed
by the diagonal from Node 1 to 3 and the diagonal from Node 4 to 2. Its finite
element is formulated in reference to ECS.
ECS  z
ECS  z
4
8
3 7
3
6 1
5
4 6 ECS  y
1 ECS  y
5
ECS  x ECS  x
2 2
Figure 3.7.1 Coordinate system of membrane elements
In order to assign an orthotropic material to a membrane element, the material’s

principal axes must be oriented in predefined directions for which material
coordinate system (MCS) is used. midas NFX provides largely two ways to define
the material directions for a membrane element. The first method uses a rotational
angle from the line segment connecting Nodes 1 and 2 as shown in Figure 3.7.2.
61
Analysis Manual
4 3
MCS  y MCS  x

1 2
Figure 3.7.2 Defining material axes of a membrane element using an angle
The second method uses a specific coordinate system in which the x-axis is
projected onto the plane of the element, which is then assumed to become the
principal axis of the material. This method of projecting the x-axis is identically
applied to defining the element result coordinate system (ERCS) to check the
element results.
y
User-defined
material
coordinate x
z
Projection
4
3
MCS  x
Figure 3.7.3 Defining material axes of a membrane element

using a material coordinate system
62
Elements
A membrane element retains the displacement degrees of freedom in the x-axis
and y-axis directions in ECS.
ui  ui vi 
T
(3.7.1)
When using the option to consider the rotation about the direction perpendicular to
an element based on a linked interpolation, an additional degree of freedom is
retained as follows:
θi   zi  (3.7.2)

Since the basic assumption for a membrane element is at the state of 2-
dimensional stresses, the strains and the resultant forces in the in-plane direction
defined in the ECS can be considered as shown in Figure 3.7.4.
 N xx   xx 
   
N   N yy  , ε   yy  (In-plane resultant forces and strains) (3.7.3)
N   
 xy   xy 
N yy ,  yy
N xy ,  xy
ECS  y
N xx ,  xx N xy ,  xy
N xy ,  xy ECS  x N xx ,  xx
N xy ,  xy
N yy ,  yy
Figure 3.7.4 Stresses/strains of a membrane element
63
Analysis Manual
• Loads
The types of loads applied to a membrane element are as follows:
Table 3.7.1 Types of loads applied to a membrane element

Pressure load Distributed load acting on the plane or the side of the element
Element temperature Element temperature that causes a deformation in the in-plane
load direction
A membrane element does not retain stiffness in the transverse direction, but the
transverse properties of loads and masses acting on the element are considered.
• Element Results
The membrane elements in midas NFX generate identical result outputs as the
shell elements. While the shell elements provide element results at two locations
(top and bottom) in the thickness direction of the elements, the membrane elements
provide identical results at the top and the bottom because stresses and strains in
the thickness direction are always the same.
• Element Thickness
The thickness of a membrane element can be defined in midas NFX as shown in
Figure 3.7.5. It is cautioned that only the thicknesses at the vertices can be
defined for higher order elements (6 or 8 nodes).
64
Elements
t4
4
t3
t1 3
t3 1
3
t1
1
t2
t2
2 2
Figure 3.7.5 Defining thicknesses of membrane elements
• Selection of Element Techniques

midas NFX uses a number of membrane element types depending on the
techniques used for enhancing the element performance. The following table
summarizes the types used in midas NFX and the corresponding finite element
techniques and integration methods. The shaded items are the defaults.
Table 3.7.2 Performance enhancement techniques used for membrane elements
Stiffness Lumped
No. of Degrees of Element Matrix Mass
Shape Type
Nodes Freedom Technique Numerical Calculation
Integration Method
Displacemen
1 point Lobatto
t assumption
Full Linked
Rotation 3 points Lobatto
integration interpolation
Triangle 3 about out-of-
plane Linked
direction interpolation,
Hybrid 3 points Lobatto
considered Hybrid
method
65
Analysis Manual
Full Displacemen
2X2 points Lobatto
integration t assumption
Reduced
Reduced
integration
integration 1X1 point Lobatto
(Stabilization
(stabilized)
technique)
Hybrid
Rectangle 4 Hybrid 2X2 points Lobatto
method
Full Linked
Rotation 2X2 points Lobatto
integration interpolation
about out-of-
plane Linked
direction interpolation,
Hybrid 2X2 points Lobatto
considered Hybrid
method
Scaling of
Displacemen
Triangle 6 3 points diagonal
t assumption
terms
Scaling of
Full Displacemen
3X3 points diagonal
integration t assumption
terms
Scaling of
Reduced Reduced
Rectangle 8 2X2 points diagonal
integration integration
terms
Scaling of
Hybrid
Hybrid 3X3 points diagonal
method
terms
The characteristics of the element techniques are noted below with some
cautionary notes.
► 3-node element: The accuracy of solution decreases as a result of significant

reduction in the flexibility of the element unless a linked interpolation, which
includes rotation about the out-of-plane direction, is used.
66
Elements
► 4-node element: The accuracy is generally high except for the isoparametric
element that uses only the displacement assumption method.
► 6-node element: The performance of the element may be noticeably reduced if a
node, which exists on a side of the element, is not located at the center of the side.
► 8-node element: All techniques generally show accurate results. An element
that uses the reduced integration shows similar performance as an element that
uses the hybrid method while retaining outstanding efficiency in calculation, but a
spurious zero energy mode may appear.
A membrane element can consider geometric nonlinearity and use elasto-plastic
materials. The result items from nonlinear analysis are identical to those of a shell
element.
67
Analysis Manual
3.8 Shell Element
A shell element is a triangular or a quadrilateral element consisted of 3, 4, 6 or 8

nodes located on a curved surface. Shell elements are mainly used when a thin
structure such as a pressure vessel is subjected to bending deformation. It can
consider 2-dimensional stress condition, bending and shear deformations.
Since a shell element is often located on a curved surface, the nodes may not exist
on the same plane, the notion of which must be reflected in the definition of ECS.
In the ECS of a triangular shell element, the direction from Node 1 to Node 2
defines the x-axis. The z-axis is defined in the normal direction of the x-axis and
the vector oriented from Node 1 to Node 3. In the case of a quadrilateral element,
the x-axis is determined in the direction that bisects the angle formed by the
diagonal from Node 1 to 3 and the diagonal from Node 4 to 2. The normal
direction of these two vectors defines the z-axis. Its finite element is formulated in
reference to ECS.
ECS  z
ECS  z
4
8 7
3
6 3
1
5
4 6 ECS  y
1 ECS  y 5
ECS  x ECS  x
2 2
Figure 3.8.1 Coordinate system of shell elements
In order to assign an orthotropic material to a shell element, the material’s principal

axes must be oriented in predefined directions. The MCS of a shell element is
defined in the same manner as that of a membrane element, using a
68
Elements
rotational angle from the line segment connecting Nodes 1 and 2 or a specific
coordinate system.
• Curved Surface Modeling

The type of shell elements used in midas NFX assumes that an inherent
21
perpendicular vector called a director exists at each node. The deformation is
expressed by the movement of this vector. Because the direction of the rotational
degree of freedom is defined relative to the director at a node, no moment in the
direction of the director exists as an internal force. This vector may be
perpendicular to the element’s surface but may not be true when a curved surface
is modeled as shell elements.
Surface normal t
Element normal n1 1 2 Element normal n 2
Shell 1 Shell 2
Figure 3.8.2 Angle between shell elements modeling a curved surface
For example, the perpendicular vector of a curved surface can be calculated as

shown below when a small kink angle exists between adjacent elements.
t
n i
(3.8.1)
n i
t : Vector perpendicular to curved surface
ni : Vector perpendicular to the element’s surface
21
Simo, J.C. and Fox, D.D., “On a stress resultant geometrically exact shell model. Part I : Formulation and optimal parametrization,”
Computer Methods in Applied Mechanics and Engineering, Vol. 72, 1989
69
Analysis Manual
If the angle,  , formed by t and n i exceeds the allowable limit, it is deemed not
as a part of the curved surface but as an actual bent structure, and the vector on
the curved surface will not be defined. If the vector perpendicular to the curved
surface is not defined at a node, the vectors perpendicular to the element surfaces
will be used as the directors.
Creating perpendicular vectors to a curved surface to accurately reflect the
geometrical shape contributes to the accuracy of results to a great extent. But it
must be cautioned when a half or ¼ of a cylindrical shape is modeled using the
symmetric condition. At the side to which the symmetry condition is assigned,
Shell 2 as shown in Figure 3.8.2 does not exist. This results in inability to obtain
geometrically accurate vectors perpendicular to the curved surface. In such a
case, it is better not to even create the vectors perpendicular to the curved surface.
A shell element retains displacement degrees of freedom in all x-, y- and z-axis
directions in ECS.
ui  ui wi 
T
vi (3.8.2)
θi   xi  yi  (3.8.3)
As described in the preceeding section, a director is either a vector perpendicular to

a curved surface or an element surface. If 6 DOF (including drilling DOF) option is
used, which represents 6 degrees of freedom for each node, the rotational degrees
of freedom are defined in all the directions.
θi   xi  yi  zi 
T
(3.8.4)

As shown in Figure 3.8.3, a shell element can consider a 2-dimensional stresses
and bending and shear deformations defined in ECS. The shell element in midas
NFX always considers shear deformations.
70
Elements
 N xx   xx 
   
N   
 xy   xy 
 M xx   xx 
   
M   M yy  , κ   yy  (Bending moments and curvatures) (3.8.6)
M   
 xy   xy 
Qzx   zx 
Q  , γ    (Shear forces and shear strains) (3.8.7)
 yz 
Q  yz 
M xy ,  xy
N yy ,  yy
N xy ,  xy M yy ,  yy
Qzx ,  zx Qyz ,  yz
M xx , xx
ECS  y Qzx ,  zx
M xy ,  xy N xx ,  xx
N xy ,  xy
N xy ,  xy
ECS  x N xx ,  xx M xy ,  xy
M xx , xx Qyz ,  yz
M yy ,  yy N xy ,  xy
N yy ,  yy
M xy ,  xy
Figure 3.8.3 Stresses/strains of a shell element
• Loads
The types of loads applied to a shell element are as follows:
71
Analysis Manual
Table 3.8.1 Types of loads applied to a shell element
Pressure load Distributed loads acting on the plane or the sides of the element
Element temperature that causes deformation in the in-plane

Element temperature
direction
load
Temperature gradient that causes bending deformation
• Element Results
The shell elements in midas NFX generate element results at two locations (top
and bottom) in the thickness direction of the elements. The result items for shell
elements are shown below, and the user may specify the reference coordinate
system. The types of coordinate systems that the user can select are ECS, MCS
and a specific coordinate system.
Table 3.8.2 Result items of a shell element
Location: Top/bottom, vertices/element center

In-plane stress
 xx ,  yy ,  xy
Normal stress
 zz
Stress Principal stress
P1 , P2 , Principal stress directions

Von-Mises stress
v
Max shear stress
 max
72
Elements

Fiber distance
Location of stress calculation in the thickness direction
Location: Vertices/element center

Maximum values
Maximum value between top/bottom ( P1 , P2 ,  v ,  max )

Calculated using  v or P1 , P2 and limit stress
Safety factor
Calculated only for isotropic materials
Minimum value between top/bottom

In-plane strain
 xx ,  yy ,  xy
Normal strain
 zz
Principal strain
E1 , E2 , Principal strain directions
Strain
Von-Mises strain
v
Max shear strain
 max
Maximum values
Maximum value between top/bottom ( E1 , E2 ,  v ,  max )
In-plane force
N xx , N yy , N xy
Force/ Location: Vertices/element center
Bending moment
Moment M xx , M yy , M xy
Shear force
Qzx , Qzy
Strain energy Location: Element center
Misc. Total percent energy Location: Element center
Energy density Location: Element center
73
Analysis Manual
M xy
N yy
N xy M yy
Qzx Qyz
M xx
ERCS  y Qzx
M xy ,  xy N xx
N xy
N xy
ERCS  x N xx M xy
M xx Qyz ,  yz Top
M yy N xy
N yy
Bottom
M xy
Figure 3.8.4 Result output directions of a shell element
• Element Thickness and material
Thicknesses at the vertices can be defined for a shell element as defined for a
membrane element. The material and effective thickness for bending and shear
deformations can be also specified. For example, if t is the thickness defined for
the in-plane behavior, the following values can be specified:
► 12I / t 3 : Ratio of bending stiffness calculated using the actual bending stiffness,
I and the thickness, t .
► ts / t : Ratio of the actual shear deformation thickness t s to t
The above effective thickness is used only to calculate the stiffness and internal
forces and is not applied in calcaulting the mass matrix. For any other cases of
reflecting the mass effects such as gravitational forces and rotational inertia forces,
the membrane material (in-plane thickness) is used.
• Offset
An offset can be used when the neutral plane (mid-surface) of a shell element is
separated from a node or when the neutral planes between the connecting
elements are not coincident. The offset of a shell element can have a fixed value
74
Elements
in the direction of the director within the element.
•Selection of Element Techniques
midas NFX uses a number of shell element types depending on the techniques
used for enhancing the element performance. The variety of different types stems
from the use of different techniques for deformations in different directions. For
example, techniques used for the in-plane and transverse directions are different.
The following table summarizes the types used in midas NFX and the
corresponding finite element techniques and integration methods. The shaded
items are the defaults.
Table 3.8.3 Performance enhancement techniques used for shell elements

Nodal Element Technique Stiffness Lumped
No. of Degrees (In-plane / Matrix Mass
Shape Type
Nodes of Transverse Numerical Calculation
Freedom Directions) Integration Method
Full Displacement
1 point Lobatto
integration assumption/ANS
5 DOF
Hybrid method/ANS+
hybrid method
Triangle 3
Full
Linked interpolation 3 points Lobatto
integration
6 DOF
Linked interpolation+
hybrid method
Full Displacement
2X2 points Lobatto
integration assumption/ANS
Reduced
Reduced
integration/ANS
5 DOF integration 1x1 point Lobatto
Rectangle 4 (Stabilization
(stabilized)
technique)
Hybrid method/ANS+
hybrid method
6 DOF Full Linked interpolation 2X2 points Lobatto
75
Analysis Manual
integration
Linked
Hybrid interpolation+hybrid 3X3 points Lobatto
method
Scaling of
Displacement
Triangle 6 5 DOF 3 points diagonal
assumption/ANS
terms
Scaling of
Full Displacement
3X3 points diagonal
integration assumption /ANS
terms
Scaling of
Reduced
Rectangle 8 5 DOF Reduced integration 2X2 points diagonal
integration
terms
Scaling of
Hybrid method
Hybrid 3X3 points diagonal
/ANS+hybrid method
terms
cautionary notes.
► 3-node element: 5DOF element is not suitable for expressing the behavior in the
in-plane direction.
► 4-node element: 6DOF element that uses only the linked interpolation is
sensitive to mesh shapes. 6DOF element that uses a hybrid method most
accurately expresses the torsional behavior.
► 6-node element: This element results in large displacement in the transverse
direction compared to other elements. The performance of the element may be
noticeably reduced if the node, which exists on a side of the element, is not located
at the center of the side.
► 8-node element: All techniques generally show accurate results. An element
that uses the reduced integration shows similar performance as an element that
uses the hybrid method while retaining outstanding efficiency in calculation, but a
spurious zero energy mode may appear.
76
Elements
A shell element can consider geometric nonlinearity and use elasto-plastic

materials. When an elasto-plastic material is applied, different material behaviors
will appear depending on the location in the thickness direction. In this case, the
resultant forces in the in-plane direction, bending moment or the stiffness must be
numerically calculated for which midas NFX performs the Simpson integration.
Especially for the application of an elasto-plastic constitutive equation, which
calculates stresses from strains, the assumption of a planar stress condition without
shear deformations is based. The result items from nonlinear analysis are shown
below.
Table 3.8.4 Result items of a shell element in nonlinear analysis
Location: Top/bottom, integration points

Equivalent stress
Calculated on the basis of plastic models,  eq
Stress
Location: Top/bottom, integration point
Plastic status
Elasticity/plasticity, 0 / 1
Location: Top/bottom, integration point

Equivalent strain
Strain
Location: Top/bottom, integration points
Effective plastic strain
ep
77
Analysis Manual
3.9 Surface Element
A surface element is a triangular or a quadrilateral element consisted of 3, 4, 6 or 8

nodes located on a plane. Since it retains no stiffness or mass but only provides
element results such as stresses and strains, the behavior of the entire analysis
model is not influenced. Surface elements are very useful to check the surface
stresses of a thin plate as shown in Figure 3.9.1. Since a surface element retains
a 2-dimensional stress condition, it plays a similar role as a very thin membrane
element.
Surface element
Tetrahedral
mesh
Figure 3.9.1 Tetrahedral mesh and surface elements
Defining the ECS of a surface element is completely identical to that of a
membrane element. When using an orthotropic material, the methods used to
define MCS and ERCS to check the element results are also identical to those of a
membrane element.
A surface element retains the displacement degrees of freedom in the x-axis and y-
axis directions in ECS.
ui  ui vi 
T
(3.9.1)
78
Elements

Since the basic assumption for a surface element is at the state of 2-dimensional
stresses, the strains and stresses in the in-plane direction defined in the ECS can
be considered as shown in Figure 3.9.2.
 xx   xx 
   
σ   yy  , ε   yy  (In-plane stresses and strains) (3.9.2)
   
 xy   xy 
 yy ,  yy
 xy ,  xy
ECS  y
 xx ,  xx  xy ,  xy
 xy ,  xy ECS  x  xx ,  xx
 xy ,  xy
 yy ,  yy
Figure 3.9.2 Stresses/strains of a surface element
• Loads
Since surface elements must not affect the entire structure model, no loads can be
applied and no mass effect exists. However a thermal load due to change in
temperatures affect the total model in calculation, and the thermal expansion is
reflected to correctly calculate the stresses and strains.
• Element Results
The surface elements in midas NFX generate in-plane stresses and strains among
others, but since the thickness is not defined, the resultant forcees do not exist.
The result items for surface elements are shown below, and the user may specify
79
Analysis Manual
the reference coordinate system. The types of coordinate systems that the user
can select are ECS, MCS and a specific coordinate system.
Table 3.9.1 Result items of a surface element

In-plane stress
 xx ,  yy ,  xy
Location: vertices/element center
Principal stress
Von-Mises stress
Stress v
Max shear stress
 max
Safety factor Calculated using  v or P1 , P2 and limit stress

In-plane strain
 xx ,  yy ,  xy
Principal strain
Strain
Von-Mises strain
v
Max shear strain
 max
A surface element can consider geometric nonlinearity and use elasto-plastic
materials. The result items from nonlinear analysis are shown below.
80
Elements
Table 3.9.2 Result items of a surface element in nonlinear analysis
Result Items Description
Location: Integration points

Equivalent stress Calculated on the basis of plastic

Stress models, eq
Plastic status
Equivalent strain Calculated on the basis of plastic
Strain

models, eq
Effective plastic Location: Integration points
ep
strain
81
Analysis Manual
3.10 Plane Strain Element
A plane strain element is a triangular or a quadrilateral element consisted of 3, 4, 6

or 8 nodes located on a plane. Plain strain elements are mainly used to analyze
long structures maintaining a uniform cross section such as a dam or a tunnel.
Because of the presence of stresses in the thickness direction, it is not strictly at
the state of 2-dimensional stresses. Plane strain elements can be modeled in 3-
dimensional space, but are generally placed on a specific plane (x-y, x-z, y-z) in
GCS based on the element properties.
The ECS of a plain strain element is defined in the same manner as that of a
membrane element, and the finite element is formulated in reference to ECS. In
order to use an orthotropic material for a plane strain element, the principal axis of
the material must be oriented in an appropriate direction. The method of defining
the direction of the material (MCS) is identical to that of a membrane element.
The material properties ( E33 ,  23 ,  31 ) in the transverse direction must be entered to
obtain valid stress and strain results.
A plane strain element retains displacement degrees of freedom in x- and y-axis
directions in ECS.
ui  ui vi 
T
(3.10.1)
When using an option to consider the rotation about the direction perpendicular to
an element based on a linked interpolation, an additional degree of freedom is
retained as follows:
θi   zi  (3.10.2)
82
Elements

Although a plane strain element expresses a 2-dimensional stress condition and
transversal perpendicular stresses, its finite element formulation only uses the 2-
dimensional stresses and thus reflects the strains and resultant forcees in the in-
plane direction defined in ECS.
 N xx   xx 
   
N   
 xy   xy 
• Loads
The types of loads applied to a plane strain element are as follows:
Table 3.10.1 Types of loads applied to a plane strain element
Applied to the density of materials and nonstructural

Gravitational force
masses
Rotational inertia Applied to the density of materials and nonstructural
force masses
Distributed load acting on the plane or the sides of the
Pressure load
element
Element temperature Element temperature that causes deformation in the in-
load plane direction
When a temperature load is applied to a plane strain element, the thermal

expansion in the thickness direction is restrained due to planar deformation
conditions, which in turn cause a further increase in planar deformations by the
Poisson’s effect.
• Element Results
The plane strain elements in midas NFX generate identical result outputs as the
83
Analysis Manual
shell elements. While the shell elements provide element results at two locations
(top and bottom) in the thickness direction of the elements, the plane strain
elements provide identical results at the top and the bottom because stresses and
strains in the thickness direction are always the same.
• Element Thickness
A plane strain element must be of a uniform thickness in order to sastisfy the planar
deformation conditions. The thickness of a plain strain element can be directly
defined in midas NFX, or else a unit (1) thickness is assumed.
•Selection of Element Techniques

The types of the plane strain elements that can be used in midas NFX and their
characteristics are identical to those of the membrane elements.
A plane strain element can consider geometric nonlinearity and use elasto-plactic
materials. The result items from nonlinear analysis are identical to those of a shell
element.
84
Elements
3.11 Axisymmetric Solid Element
Axisymmetric solid elements are used to analyze structures in which parameters

such as shape, material and load condition satisfy rotational symmetry conditions
with respect to a specific axis. Axisymmetric solid elements cannot be used in
combination with other types of elements and must be modeled on the x-z plane in
reference to GCS. The types of axisymmetric solid elements are triangular and
quadrilateral shapes consisted of 3, 4, 6 or 8 nodes.
The ECS of an axisymmetric solid element is defined in the same manner as that of
a membrane element, and the finite element is formulated in reference to GCS.
In order to assign an orthotropic material to an axisymmetric solid element, the

material’s principal axes must be oriented in predefined directions in which case
MCS is used. midas NFX provides largely two ways to define the material
directions for an axisymmetric solid element. The first method uses a rotational
angle from the x-axis in GCS as shown in Figure 3.11.1.
4 3
GCS  z
MCS  y MCS  x
2
 1
1 GCS  x
Figure 3.11.1 Defining material axes of an axisymmetric solid element

using an angle
The second method, similar to a membrane element, uses a specific coordinate

system in which the x-axis is projected onto the plane of the element, which is then
assumed to become the principal axis of the material. This method of projecting
85
Analysis Manual
the x-axis is identically applied to defining the element result coordinate system
(ERCS) to check the element results.
An axisymmetric solid element retains the displacement degrees of freedom in the
x-axis (radial direction) and z-axis directions in GCS.
ui  ui wi 
T
(3.11.1)

An axisymmetric solid element considers strains and stresses defined in GCS and
its components are as follows:
 xx   xx 
   
   
σ     , ε     (stresses and strains in in-plane/circumferential directions) (3.11.2)

  zz   zz 
  zx   zx 
 zz ,  zz
4  zx ,  zx
 xx ,  xx
GCS  z
 xx ,  xx
 zx ,  zx
 zz ,  zz
  ,  2
1
GCS  x
Figure 3.11.2 Stresses/strains of an axisymmetric solid element
86
Elements
• Loads
The types of loads applied to an axisymmetric solid element are as follows:
Table 3.11.1 Types of loads applied to an axisymmetric solid element

Gravitational force Applied to the density of materials
Rotational inertia force Applied to the density of materials
Distributed load acting on the side of the element (considering

Pressure load the axis-symmetry, the load is applied to the outer surface of a
structure)
• Element Results
Axisymmetric solid elements in midas NFX generate result values with respect to a
reference coordinate system specified by the user. The types of coordinate
systems that the user can select are ECS, MCS and a specific coordinate system.
When defining the directions of the element result components, the in-plane
directions of the element are x, y , and the circumferential direction is  .
Table 3.11.2 Result items of an axisymmetric solid element

Stress component
 xx ,  yy ,   ,  xy
Location: Vertices /element center
Principal stress
P1 , P2
Stress Von-Mises stress
v
Max shear stress
 max
Safety factor Calculated using  v or P1 , P2 ,   and limit stress
87
Analysis Manual

Strain component
 xx ,  yy ,  ,  xy
Principal strain
Strain E1 , E2
Von-Mises strain
v
Max shear strain
 max
midas NFX uses a number of axisymmetric solid element types depending on the
Table 3.11.3 Performance enhancement techniques used for axisymmetric

solid elements
Stiffness
Lumped mass
No. of Element Matrix
Shape Type Calculation
Nodes Technique Numerical
Method
Integration
Displacement
Triangle 3 3 points Lobatto
assumption
Displacement
Full integration 2X2 points Lobatto
assumption
Rectangle 4
Hybrid
method
Displacement Scaling of
Triangle 6 3 points
assumption diagonal terms
Rectangle 8 3X3 points
88
Elements
An axisymmetric solid element can consider geometric nonlinearity and use elasto-
plastic materials. The result items from nonlinear analysis are shown below.
Table 3.11.4 Result items of an axisymmetric solid element in nonlinear analysis

Equivalent stress
Calculated on the basis of plastic models,  eq
Stress
Plastic status

Equivalent strain
Strain
strain ep
89
Analysis Manual
3.12 Solid Element
Solid elements are primarily used to model structures of significant volumes such
as an automobile engine or a thick wall. The types of solid elements in midas NFX
are tetrahedron, pentahedron and hexahedron shapes consisted of 4, 5, 6, 8, 10,
13, 15 or 20 nodes. Pentahedron elements can be of wedge and pyramid shapes.
The ECS of a tetrahedron element is identical to applying the ECS definition rules
of a membrane element to the triangular shape formed by Nodes 1, 2 and 3. The
ECS of a pentahedron wedge element is identical to applying the ECS definition
rules of a membrane element to the triangular shape formed by the midpoints
between Nodes 1 and 4, 2 and 5, and 3 and 6. The ECS of a pentahedron
pyramid element is identical to applying the ECS definition rules of a membrane
element to the rectangular shape formed by Nodes 1, 2, 3 and 4. For a
hexahedron element, the vectors located close to the ECS are defined as follows:
► r: Vector pointing from the center point among Nodes 1, 5, 8 and 4 to the
center point among Nodes 2, 6, 7 and 3
► s: Vector pointing from the center point among Nodes 1, 2, 6 and 5 to the
► t: Vector pointing from the center point among Nodes 1, 2, 3 and 4 to the
The orthogonal coordinate system placed closest to the above three vectors
becomes the ECS of the hexahedron element.
90
Elements
4
15
6
4
10 13 14
9 12
8 10 5
3
7 ECS  z 9 11 3 ECS  z
6 ECS  y
5 ECS  y
1 1
8
7
2
ECS  x 2 ECS  x
5 8 19 7
20
ECS  z
3 12 6
ECS  z
5 18 ECS  y
10 16 15
4 8 17
11
9
3 ECS  y 13
14
4 11
3 ECS  x
7 12
1 6
10
2 ECS  x 1 9 2
Figure 3.12.1 Coordinate system of solid elements
A solid element retains displacement degrees of freedom in all x-, y- and z-axis
directions in GCS.
ui  ui wi 
T
vi (3.12.1)

A solid element considers stresses and strains defined in GCS with the following
components:
 xx   xx 
   
 yy   yy 
 zz   
σ    , ε   zz  (Stresses and strains) (3.12.2)
  xy   xy 
  yz   yz 
   

 zx   zx 
91
Analysis Manual
 zz ,  zz
 yz ,  yz
 zx ,  zx
 zx ,  zx
 xy ,  xy
 xy ,  xy  xx ,  xx
 yy ,  yy
GCS  z  yz ,  yz
GCS  y
GCS  x
Figure 3.12.2 Stresses/strains of a solid element
• Loads
The types of loads applied to a solid element are as follows:
Table 3.12.1 Types of loads applied to a solid element
Gravitational force Applied to the density of materials
Rotational inertia force Applied to the density of materials
Pressure load Distributed load acting on the face of the element
• Element Results
Solid elements in midas NFX generate result values with respect to a reference
coordinate system specified by the user. The types of coordinate systems that the
user can select are ECS, MCS and a specific coordinate system.
92
Elements
Table 3.12.2 Result items of a solid element


Stress component
 xx ,  yy ,  zz ,  xy ,  yz ,  zx
Principal stress
P1 , P2 , P3 , Principal stress directions
Von-Mises stress
v
Stress Max shear stress
 max
Octahedral stress
o
Mean pressure
p0
Safety factor Calculated using  v or P1 , P3 and limit stress
Strain component
 xx ,  yy ,  zz ,  xy ,  yz ,  zx
Principal strain
E1 , E2 , E3 , Principal strain directions
Von-Mises strain
Strain v
Max shear strain
 max
Octahedral strain
o
Mean compression
c0
Misc. Total percent energy Location: Element center
Energy density Location: Element center
93
Analysis Manual

midas NFX uses a number of solid element types depending on the techniques
used for enhancing the element performance. The following table summarizes the
types used in midas NFX and the corresponding finite element techniques and
integration methods. The shaded items are the defaults.
Table 3.12.3 Performance enhancement techniques used for solid elements
Stiffness
Lumped mass
No. of Element Matrix
Shape Type Calculation
Nodes Technique Numerical
Method
Integration
Displacement
Full integration 1 point Lobatto
assumption
Tetrahedron 4
EAS, u-p hybrid
Enhanced 4 points Lobatto
method
Displacement
assumption
Reduced
Wedge Reduced
6 integration
(pentahedron) integration 1X1 point Lobatto
(Stabilization
(stabilized)
technique)
Hybrid Hybrid method 3X2 points Lobatto
Full integration 4X2 points
Pyramid Reduced Reduced Scaling of
5 1X1 point
(pentahedron) integration integration diagonal terms
Scaling of
Hybrid Hybrid method 4X2 points
diagonal terms
Displacement
Full integration 2X2X2 points Lobatto
assumption
Hexahedron 8
Reduced Reduced
1X1X1 point Lobatto
integration integration
94
Elements
(stabilized) (Stabilization
technique)
Hybrid Hybrid method 2X2X2 points Lobatto
Full integration 4 points
Tetrahedron 10
Incompatible Scaling of
Enhanced 4 points
element diagonal terms
Full integration 3X3 points
Wedge Reduced Reduced Scaling of
15 3X2 points
(pentahedron) integration integration diagonal terms
Scaling of
Hybrid Hybrid method 3x3 points
diagonal terms
Pyramid Displacement Scaling of
13 9X3 points
(pentahedron) assumption diagonal terms
Full integration 3X3X3 points
Reduced Reduced Scaling of
Hexahedron 20 2X2X2 points
integration integration diagonal terms
Scaling of
Hybrid Hybrid method 3X3X3 points
diagonal terms
cautionary notes:
► 4-node element: Displacement results are similar irrespective of which technique
is used. The element that uses the EAS and u-p hybrid method shows more
accurate stress results.
► 6-node element: The element that uses a hybrid method shows superior
performance for thin structures.
► 8-node element: The element that uses a hybrid method or reduced integration
shows superior performance for structures subjected to bending.
► 10-node element: All techniques generally show similar results, but incompatible
95
Analysis Manual
elements show relatively flexible results for thin structures.

► 20-node element: All techniques generally show accurate results. For thin
structures, elements applied with a hybrid method show excellent performance.
A solid element can consider geometric nonlinearity and use elasto-plastic and
hyperelastic materials. The result items (using elasto-plastic materials) from
nonlinear analysis are as follows:
Table 3.12.4 Result items of a solid element in nonlinear analysis (elasto-

plastic materials)

Equivalent stress
Calculated on the basis of plasticity models,  eq
Stress
Plastic status

Equivalent strain
Strain
strain ep
96
Elements
3.13 Layered Shell Element
Layered shell elements are used to effectively analyze thin structures of laminated
layers comprising different materials or different principal axes in the thickness
direction such as composite materials and sandwiches. The basic coordinate
system, curved surface model and the degrees of freedom are identical to those of
the general shell elements. A layered shell element is a triangular or a
quadrilateral element consisted of 3, 4, 6 or 8 nodes. That is, a layered shell
element, which is based on the primary shear deformation theory, has the identical
finite element formulation as the shell element. But its constitutive equation uses
the lamination theory, which will be introduced in Chapter 4.3.
• Shear Stiffness Calculation & Shear Stress Restoration in the Transverse

Direction
Unlike an isotropic material, the stiffness calculation in the transverse direction,

which is based on the shear correction factor, cannot be generalized for a
laminated composite material. Therefore, it is desirable to calculate the shear
stiffness in the transverse direction using a number of assumed deformation
shapes and stress equilibrium equations.22 The use of the series of the above
procedures also renders an advantage of the possibility of restoring shear stress in
the transverse direction.
The in-plane stresses based on the primary shear deformation theory and the
transverse shear stresses based on the 3-dimensional stress equilibrium equations
are expressed as follows:
 x 
 
 y   C (ε o  zκ )
(k )
(3.13.1)
 
 xy 
22
Rolfes, R. and Rohwer, K., “Improved transverse shear stresses in composite finite elements based on first order shear defor
mation theory,” Internatinal Journal for Numerical Methods in Engineering, Vol. 40, 1997
97
Analysis Manual
 xz 
    z  1 0 0  ( k ) 0 0 1  ( k ) 
     0    C (εo, x  zκ , x )  0 1 0 C (εo, y  zκ , y )  d (3.13.2)

 yz   0 0 1    
If the effects of the in-plane forces are ignored in the constitutive equation of a
layered shell ( N  0 ), the strain and curvature at the neutral plane are expressed
as follows:
ε o   A 1Bκ
(3.13.3)
κ  D*1M
where, D*  D  BT A1B . Using Equation (3.13.3), the transverse shear stresses

can be expressed again as follows:
 xz 
  1 0 0  0 0 1 
   F ( z )M , x   0 1 0  F ( z ) M , y (3.13.4)

 yz   0 0 1   
F( z )  
 0
 z
C( k ) d A1B  
 z
 0 
 C( k ) d D*1 (3.13.5)
When cylindrical bending behaviors about the x- and y-axes are assumed, the in-
plane differential values and the transverse shear forces can be expressed in a
simple relationship.
Qxz   M x , x 
     (3.13.6)
 yz 
Q  M y , y 
Using the above, the transverse shear stresses are expressed as,
 xz   F11 F32  Qxz 

   
F22  Qyz 
(3.13.7)
 yz   F31
That is, the transverse shear stresses are determined by the transverse shear
forces and the shape function, F( z ) , which is based on the properties, thicknesses
98
Elements
and the lamination angles of the materials constituting the laminated plate. Using
this, the transverse shear stiffness, G , which is calculated by integration based on
the lamination theory, can be expressed as follows:
1
  F F32 T F F32  
H     11  G 1  11 
F22  
(3.13.8)
  F31 F22   F31 
• Element Results
The layered shell elements in midas NFX generate results at the center or top and
bottom of each ply in the thickness direction. Also, the results include the
maximum and minimum values of the entire laminated plate. Stress and strain
results are produced in the principal axis directions of the material in each ply.
Table 3.13.1 Result items of a layered shell element
Location: Center/top-bottom of each ply, maximum/minimum

In-plane stress of laminated plate, vertices/element center
 11 ,  22 ,  12
Transverse shear
of laminated plate, vertices/element center
stress
1z , 2 z
Stress Principal stress of laminated plate, vertices/element center
Von-Mises stress of laminated plate, vertices/element center
v
Max shear stress of laminated plate, vertices/element center
 max
99
Analysis Manual

In-plane strain of laminated plate, vertices/element center
11 ,  22 ,  12
Principal strain of laminated plate, vertices/element center
Strain
Von-Mises strain of laminated plate, vertices/element center
v
Max shear strain of laminated plate, vertices/element center
 max
In-plane force
N xx , N yy , N xy
Force/
Momen Bending moment
M xx , M yy , M xy
t
Shear force
Qxz , Qyz
Failure
Misc. index/strength
Tsai-Wu, Tsai-Hill, Hoffman, max-strain, max-stress or
ratio
LaRC02
A layered shell element can consider geometric nonlinearity and use elasto-plastic
materials. When using elasto-plastic materials, the number of Simpson integration
points can be defined at each ply. The result items from nonlinear analysis are as
follows:
100
Elements
Table 3.13.2 Result items of a layered shell element in nonlinear analysis

Equivalent stress integration points of laminated plate
 eq
Calculated on the basis of plastic models
Stress
Plastic status integration points of laminated plate
Location: Location: Center/top-bottom of each ply,
Equivalent strain maximum/minimum integration points of laminated plate

Calculated on the basis of plastic plasticity model eq
Strain
Effective plastic
integration points of laminated plate
strain ep
101
Analysis Manual
3.14 Layered Solid Element
Layered solid elements are used to model composite materials and are suitable for
thick structures for which the use of layered shell elements is inappropriate. midas
NFX assumes solid elements that retain composite material properties as layered
solid elements. The types of available layered solid elements are pentahedron
wedge and hexahedron shapes. A layered solid element can consist of 6, 8, 15 or
20 nodes. The definition of a coordinate system, selection of the degrees of
freedom and the types of loads are identical to those of a solid element.
The ECS of a layered solid element is defined identically to that of a solid element.
The laminating order of materials is assumed to be in the direction from the bottom
to top in reference to the z-axis of ECS. Since layered solid elements are usually
applied to modeling laminated plates of composite materials, orthotropic materials
in MCS are frequently used. As shown in Figure 3.14.1, the x-axis is projected on
the neutral plane of an element, and the resulting direction is assumed to be the
principal axis of the material.
y
User-defined
material
coordinate x
z
Projection
8
7
4
MCS  y 3
5
MCS  x
6
1 Mid-plane
Figure 3.14.1 Defining material axes of a layered solid element

using a coordinate system
102
Elements
• Integration Method in the Thickness Direction
Since layered solid elements are assumed to be laminated with multiple layers of
different materials in the thickness direction, the amount of numerical calculation
increases if the general Gauss integration is used. In order to resolve such
problem, midas NFX expands the matrix B and Jacobian, J , in reference to the
natural coordinate system,  and performs analytic integration.
The decomposition of the displacement-strain relationship in reference to  is as
follows:
ε  Bu  (B0   B1   2B2 )u (3.14.1)
If Jacobian, J , is decomposed in the same manner, the stiffness matrix can be

calculated as follows:
 Tk
   B    
T
K 0   B1   2B2 C( k ) B0   B1   2B 2 J 0   J1   2 J 2 d d d (3.14.2)
Bk
 k
 Bk ,  Tk : Coordinates at top/bottom of each ply 
C (k ) : Stress-strain relationship matrix of each ply
• Modifying the Equation of Stress-Strain Relationship
The use of 6- or 8-node layered solid elements formulated by the general

isoparametric method cannot result in accurate solutions. In addition, a locking
phenomenon takes place, which leads to the application of various performance
enhancement techniques described in Section 3.3.
Since layered solid elements are often used to model thin laminated plates, the
stress is hardly generated in the thickness direction. Yet the layered solid
elements retain the characteristics of solid elements, which is the reason for
generating vertical stresses due to any applied loads in the thickness direction. In
order to consider these phenomena simulataneously, the 6- and 8-node layered
23
solid elements in midas NFX use a modified form of the stress-strain relationship
23
Sze, K.Y., Lo, S.H. and Yao, L.Q., “Hybrid-stress solid elements for shell structures based upon a modified variational functio
nal,” International Journal for Numerical Methods in Engineering, Vol. 53, 2002
103
Analysis Manual
equation to maintain a uniform stress in the thickness direction.
• Element Results
The layered solid elements in midas NFX generate results at the center or top and
bottom of each ply in the thickness direction. Also, the results include the
maximum and minimum values of the entire laminated plate. Stress and strain
results are produced in the principal axis directions of the material in each ply.
Table 3.14.1 Result items of a layered solid element

Stress component of laminated plate, vertices/element center
 11  22  33  12  23  31
, , , , ,
Principal stress of laminated plate, vertices/element center
P1 P2 P3
, , , Principal stress directions
Von-Mises stress of laminated plate, vertices/element center
v
Stress
Max shear stress of laminated plate, vertices/element center
 max

Octahedral stress of laminated plate, vertices/element center
o

Mean pressure of laminated plate, vertices/element center
p0
104
Elements

Strain component of laminated plate, vertices/element center
11  22  33  12  23  31
, , , , ,
Principal strain of laminated plate, vertices/element center
E1 E2 E3
, , , Principal strain directions
Von-Mises strain of laminated plate, vertices/element center
v
Strain
Max shear strain of laminated plate, vertices/element center
 max

Octahedral strain of laminated plate, vertices/element center
o

Mean
compression c0

Failure
Misc. index/strength
Tsai-Wu, Tsai-Hill, Hoffman, max-strain, max-stress or
ratio
LaRC02
midas NFX uses a number of layered solid element types depending on the
105
Analysis Manual
Table 3.14.2 Performance enhancement techniques used for layered solid elements
Stiffness Lumped
No. of Matrix Mass
Shape Type Element Technique
Nodes Numerical Calculation
Integration Methods
Displacement
assumption/ANS
Full integration 3 points Lobatto
Modification of stress-
Wedge 6 strain relationship
Hybrid method/ANS
Hybrid Modification of stress- 3 points Lobatto
strain relationship
Displacement
assumption/ANS
Modification of stress-
Hexahed
8 strain relationship
ron
Hybrid method/ANS
Hybrid Modification of stress- 2X2 points Lobatto
strain relationship
Scaling of
Displacement
Wedge 15 3 points diagonal
assumption
terms
Scaling of
Displacement
Full integration 3X3 points diagonal
assumption
Hexahed terms
20
ron Scaling of
Reduced
Reduced integration 2X2 points diagonal
integration
terms
cautionary notes:
► 8-node element: Since the natural strain assumption is chosen, all the
106
Elements
techniques generally show accurate results.

► 20-node element: For thin structures, the element that uses the reduced
integration shows excellent performance, but the spurious zero energy mode may
occur.
A layered solid element can consider geometric nonlinearity and use elasto-plastic
materials. When using elasto-plastic materials, the number of Simpson integration
points can be defined at each ply. The result items from nonlinear analysis are as
follows:
Table 3.14.3 Result items of a layered solid element in nonlinear analysis

Equivalent stress integration points of laminated plate
Calculated on the basis of plastic models  eq
Stress Location: Center/top-bottom of each ply, maximum/minimum

Plastic status

Equivalent strain
Strain  eq

Effective plastic
strain
ep
107
Analysis Manual
3.15 Other Elements
In addition to the general structural elements explained in the preceeding sections,

midas NFX also provides other element types including scalar element, bush
element, special element and rigid body/interpolation element.
• Scalar Element
A scalar element can express deformation energy or kinetic energy generated

when a node moves relative to a contact ground point, or it can consider a relative
movement between two nodes. Scalar elements are largely classified into spring,
mass and damper elements, all of which are identically expressed except for the
differences in retaining deformation energy and kinetic energy or expressing
resistance to velocity.
A 1-node spring element is used to consider an adjacent structure located at the

model’s boundaries with elastic boundary conditions or to prevent singularity errors
as a temporary measure. A 2-node spring element acts to elastically connect two
nodes. Spring elements are used to represent the relative elastic movements of
adjoining planes of independently constructed models. The stiffness of a spring
element is specified with respect to NDCS of the corresponding node.
SRY
SY
SX
SRX
Nodal Point
SZ
z
y
Z SRZ
Y
x
X
Figure 3.15.1 Example of a spring element in a nodal displacement

coordinate system
108
Elements
A mass element can be used to model a part of a structure, which retains no

stiffness effects. The mass of a mass element is also specified with respect to
NDCS of the corresponding node. A damper element is used in dynamic
structural analysis. The damping value is also specified in NDCS.
A spring element produces element results as shown in Table 3.15.1. However,

because displacement and rotational degrees of freedom can be selectively defined,
it is not possible to consistently convert units, which must be considered when the
results of stress and force/moment are reviewed.
Table 3.15.1 Result items of a spring element
Location: Element center

Stress Stress component Calculated from stress coefficient, s
(   sN or   sM )
Force/moment Location: Element center

Force/Moment
component N or M

Misc.
Total percent energy Location: Element center
• Bush Element
A bush element retains both spring and damping functions and is generally
consisted of two nodes. It can be categorized as a 1-dimensional element since it
uses the location information of the nodes. A bush element retains the same
properties as a scalar element if it is defined at a single node. Especially in case
of a 2-node bush element, ECS changes with the movements of both ends, thereby
enabling to consider geometric nonlinearity.
A bush element differs from a scalar element due to the diversity in shape
definitions. Figure 3.15.2 shows a general shape of a bush element that can be
defined by various input values.
109
Analysis Manual
ECS  x
Spring/Damper
ECS  y ECS  z Location
2
S3
z
y
1 S1 S2
GCS x
Figure 3.15.2 Shape definition of a bush element
Six components of the stiffness and damping of a bush element can be defined in
each direction of ECS. The element results are similar to those of a spring
element and include strain results.
I
Felastic  f (u I )
(3.15.1)
I
Fdamping  f (u I )
I I
Felastic , Fdamping : Internal force and damping force applied to the Ith DOF
u , u
I I
: Relative displacement and relative velocity of the Ith DOF
Felastic Fdamping
Multi-linear
Coulomb-damping
s N
u u
Figure 3.15.3 Example of defining nonlinear stiffness and damping of a

bush element
The above figure shows examples of a stiffness model exhibiting multi-linear

characteristics and a Coulomb damping model. The nonlinear stiffness model
110
Elements
elastically behaves, and the internal force function and damping force function must
always pass through the origin. The results for bush elements are similar to those
of spring elements and include strain results.
• Special Element
A type of special elements available in midas NFX is lumped mass. A lumped

mass is a 1-node element, but it is different from the 1-node mass element
previously explained in that offsets from the node and the mass rotational inertia
moment tensor, I ij , can be defined as follows:
I xx    ( y 2  z 2 )dV , I yy    ( z 2  x 2 )dV , I zz    ( x 2  y 2 )dV

(3.15.2)
I xy    xydV , I xz    xzdV , I yz    yzdV
 : Density
x, y, z : Distance from the center of gravity
• Rigid Body/Interpolation Element
The rigid body element and interpolation element are the types of elements that
mutually restrain the relative movements among the related nodes. The
restraining node is called the independent node, and its degrees of freedom are
called the independent degrees of freedom. The restrained nodes are referred to
as the dependent nodes, and their degrees of freedom are referred to as the
dependent degrees of freedom. A rigid body element functions to restrain
geometric relative movements of a number of nodes by a single node. A number
of dependent nodes are thus connected to one single independent node. The
equation of the inter-relationship between the independent node and the dependent
nodes is written as,
u D  u I  r  θ I  u I  (x)  θ I
(3.15.3)
θD  θI
u D , θD : Displacement and rotation of a dependent node
111
Analysis Manual
u I , θI : Displacement and rotation of an independent node
x : The vector oriented from a dependent node to an
independent node ( xI  xD )
Among the six degrees of freedom of a dependent node, specific degrees of

freedom can be selected to be restrained by the independent node, thereby
creating selective rigid body elements in specific directions. The following is an
example of a restrained rigid body motion on the x-y plane.
u D  u I   zI Δy , v D  v I   zI Δx ,  zD   zI (3.15.4)
4 3 4 3
1
2
y 2
1
x x
Independent node
Figure 3.15.4 Example of a rigid body motion on a plane
An interpolation element is an element by which a single node behaves relatively to

the movements of a number of other nodes. A number of independent nodes are
thus connected to one single dependent node. An interpolation element is used to
distribute force or mass among a number of nodes and renders a weaker
restraining force due to a fewer number of restrained nodes compared to a rigid
body element. For example, the process of distributing forces will be reviewed in
order to understand the displacement relationship between a dependent node and
a number of independent nodes on the x-y plane.
112
Elements
w3
w1 FD
r1 FD
r3
C.G. Reference
r2
point
w2 MD  e  F D
e MD
Figure 3.15.5 Relationship between the center of gravity of independent nodes

and forces acting on a dependent node
Figure 3.15.4 shows a dependent node located at a distance, e , away from the
center of gravity of three independent nodes with the weight factors, wi . The
D D
force, F and moment, M , applied to the dependent node will act on the center
of gravity of the independent nodes as the force, FD and the moment,
M  e F .
D D
The force and moment acting on the center of gravity of the
dependent node can be distributed among the independent nodes as follows:
wi w (M D  e  F D )  ri
Fi  F D  i (3.15.5)
 wi w r
2
i i
The above equation of force relationship can be converted into the following
displacement and rotation relationships:
wi wi ey (ryi  rxi ) wi wi ex (rxi  ryi )

u D  ui (  ), v D  vi (  ) (3.15.6)
 i  wi ri 2
w  i  wi ri 2
w
vi rxi  ui ryi
D  wi (3.15.7)
w
2
i ri
Ultimately, the average motion of the independent nodes determines the movement
of the dependent node. Because of such characteristic, a fewer number of
restrained degrees of freedom is generated compared to those of a rigid body
element.
113
Analysis Manual
• Joint Element
A joint element is used to constrain relative motions between two points based on
various joint types that have been pre-defined. The two points whose relative
motions are constrained can consist of a ground point - node or node – node.
Generally, the two points connected by a joint element often are a specific node
included in each different rigid body element. In such a case, the joint element
can be useful in simulating the relative motions between the two rigid body
elements.
In case of linear analysis, the joint element can be considered to be an expanded

form of the rigid body element expressed in Equation (3.15.3). In a rigid body
element, the constraint equations for constrained nodal DOFs are in a form, which
can be immediately eliminated because whether or not a DOF is constrained is
defined relative to the nodal coordinate system of a slave node. Unlike a rigid
body element, the constraint equations of a joint element are defined relative to the
ECS of the element. This necessitates converting the constraint equations in a
form that can be readily eliminated. Therefore, the master DOFs and slave DOFs
in a joint element are not determined in the process of defining the element.
Rather, they are appropriately selected inside by considering the overall constraint
conditions in the process of transforming the constraint equations given in terms of
ECS into a form that can be eliminated. The overall constraint equation for all
c
DOFs, u included in all joint elements is expressed as follows:
Cuc  0 (3.15.8)
Excluding the DOFs to which single-node constraints have been applied and
selecting as many slave DOFs as the number of constraint equations (the number
of rows of matrix C ), Equation (3.15.8) can be expressed in terms of the master
DOFs, u I , and slave DOFs, u D , separated as follows:
u D 
 
CD CI    0 (3.15.9)
I
u 
 
If the rank of the matrix C D is identical to the number of constraint equations (full
114
Elements
rank), the slave DOFs can be expressed in a form that can be eliminated by the
Gaussian elimination method as follows:
u D  CD 1CI u I (3.15.10)
If the rank of the matrix C D is smaller than the number of constraint equations, the
system is at an over-constrained state. In this case, the congruence between the
over-constraints and the original constraints can be distinguished based on whether
or not the remaining parts from the elimination process are all 0. By integrating
the selection of slave DOFs and elimination process, much more complex
constraint relationships of joint elements in linear analysis can be quite simply
defined compared to rigid body elements. Also, the process can accurately
identify whether or not the DOFs are over-constrained or these over-constraints are
congruent with original constraints so that appropriate actions can be taken.
The relative motions (DOFs) permitted on the basis of different joint types are
defined in reference to ECS of the elements, and especially for a joint element
consisted of node - node in geometric nonlinear analysis, ECS changes following
the rotation of the element. Like a bush element, spring stiffness and damping can
be assigned to the DOFs allowing the relative motions in a joint element. For
constrained DOFs, element forces of joint elements produce the constraint forces in
reference to ECS, and for DOFs permitted for relative motions, element forces in
reference to ECS are generated like bush elements.
Table 3.15.2 Result items of a joint element

Calculated from stress coefficient s
Stress Stress component
(   sN or   sM )
Calculated from strain coefficient e
Strain Strain component
(   e u or   e θ )
Force/moment
Force/Moment N or M
component
115
Analysis Manual
midas NFX provides the following types of joint elements: join, spherical, cylindrical,
slot, revolute, planar, translational, universal and general.
Table 3.15.3 summarizes the allowable relative DOFs for each joint type, and
Figure 3.15.6 illustrates the schematic diagrams of a few typical joint types
individually focused on their relative motions. A universal type joint element
retains ECS of a reference point and ECS2 of a slave point. And the x-axis of
ECS and the z-axis of ECS2 satisfy the orthogonal condition during deformation.
Table 3.15.3 Joint element types and allowable DOFs
Joint Type Allowable Relative DOFs
Join -
Spherical x , y , z
Cylindrical w , z
Slot u , x , y , z
Revolute z
Planar u , v , z
Translational u
Universal x , z
General User defined
116
Elements
ECS-z
Z
Spherical Cylindrical
Cylindrical
Spherical
ECS-zZ
ECS-x
X
ZECS-z
Y
ECS-y
Revolute
Revolute Planar
Planar
ECS2-zZ
XECS-x
ECS-x
X
Translational
Translational Universal
Universal
Figure 3.15.6 Joint types and motions
117
Analysis Manual
3.16 Geometric Stiffness
Geometric stiffness, also known as stress stiffness, is a type of stiffness resulting

from a change in internal forces due to a change in the geometric shape of a
structure having internal forces. Geometric stiffness is used in linear buckling
analysis and geometric nonlinear analysis. The following element types in midas
NFX consider geometric stiffness:
Table 3.16.1 Types of elements that consider geometric stiffness
Degree of Freedom
Element Types Internal Force Components
Components
Rod Axial force N xx v, w
Bar Axial force N xx v , w ,  y , z
Pipe Axial force N xx v , w ,  y , z
Membrane, Plane
In-plane resultant forces N xx , N yy , N xy u, v, w
strain
In-plane resultant forces N xx , N yy , N xy
Shell Bending moments M xx , M yy , M xy u , v , w , x ,  y

Shear forces Qzx , Qyz
In-plane stresses  xx ,  yy ,  xy
Axisymmetric solid u, w
Circumferential stress  
Stress components
Solid u, v, w
 xx ,  yy ,  zz ,  xy ,  yz ,  zx
In cases of spring, bush and rigid elements and contact conditions, geometric
stiffness is considered only in geometric nonlinear analysis.
• Geoetric Stiffness Calculation for Structural Elements
midas NFX calculates geometric stiffness based on the updated Lagrangian

formulation, which assumes the Jaumann stress rate as the objective stress rate.
118
Elements
For example, the internal force of a solid element is calculated from the stress and
virtual deformation as follows:
 ui fi    ij Dij dV (3.16.1)
1  ui  u j
 Dij : Virtual deformation (  )
2 x j xi
Since the slope of the tangent of an internal force is considered the stiffness, taking
another variation, the terms to be integrated then become,
d ij Dij   ij d Dij (3.16.2)
The variation of the integral domain is ignored in the above equation. A solid
element is fixed regardless of structural deformation because EFCS is GCS.
Therefore, d Dij  0 , and the stress increment due to the objective stress rate in
the first term is as follows:
d ij  dwik kj   ij dwjk  Cijkl dDkl (3.16.3)
1  ui  u j
 wij : Incremental spin (  )
2 x j xi
If Equations (3.16.2) and (3.16.3) are substituted into Equation (3.16.1), the
following tangent stiffness can be attained:
 ui Kij du j    DijCijkl dDkl   ij ( Lki dLkj  2 Dik dDkj )dV (3.16.4)
 Lij : Incremental displacement gradient  Dij   w ij
The first term of the integrand value is known as the material stiffness, and the
second term is the geometric stiffness. As the equation shows, the geometric
119
Analysis Manual
stiffness is proportional to the magnitude of stress. Therefore, linear buckling

analysis becomes possible through the use of such property.
• Geometric Stiffness Calculation for Rigid Body Element
Geometric stiffness for a rigid body element is created by the forces acting on the
dependent nodes. If the force and moment applied to a dependent node are f s
and m s respectively, the virtual work due to the force and moment then becomes,
 W  f s   u s  ms   θs (3.16.5)
 u s ,  θs : Displacement and rotation of a dependent node
Taking another variation on the virtual work, a basic equation to calculate the
stiffness can be attained.
dW  f s  d us  ms  d θs  f s  d us (3.16.6)
In order to transform the displacement of a dependent node into the displacement

and rotation of the independent node, the following equation is used:
 us   um   θm  (xs  xm ) (3.16.7)
If the above equation is varied again and substituted into Equation (3.16.6), the
stiffness can be calculated.
dW  f s  d us  f s  ( θm  (dθm  (xs  xm ))) (3.16.8)
The geometric stiffness of a rigid body element is composed of the force applied to
a dependent node and the relative distance and is related to the rotational degree
of freedom of the independent node.
When some of the degrees of freedom of the dependent nodes are released from
restraint, the directions of the released degrees of freedom also rotate following the
120
Elements
movement of the rigid body element. Figure 3.16.1 shows that the node
displacement coordinate system (NDCS) of a dependent node moves with the
rotation of the independent node, and the released restraints also follow the
changing coordinate system. Or, when selective degrees of freedom are
restrained using a rigid body element, the directions of restraints are not steadily
maintained, but are continuously changed.
Rotated NDCS z'

z
x'
NDCS y
x y'
Translation
Rotation
Figure 3.16.1 Directions of degrees of freedom of dependent nodes

changing with the rotation of the independent node
121
Analysis Manual
3.17 Heat Conduction Element
The heat transfer analysis in midas NFX is based on the equilibrium equations for
transient state and steady state. This section explains the spatial discritization
founded on the finite element method and the time integration method. A heat
conduction element basically retains linearity, but when temperature dependency is
assigned, nonlinearity becomes predominant. Also, the shape function and the
numerical integration are not too different from those of structural elements.
Because only one degree of freedom (temperature) is associated with each node,
the analysis can be performed at a relatively lower cost compared to structural
analysis.
• Finite Element Formulation
The transient energy equilibrium equation, which considers an increase/decrease in

internal thermal energy caused by heat flux transferred to boundaries, internal heat
generation and specific heat, is as follows:
 qdS   rd    cTd  (3.17.1)
q : Heat flux
r : Quantity of heat generated per unit volume
c : Specific heat
 : Mass density
The relationship between heat flux and temperature founded on the Fourier’s Law
is expressed as follows:
T
fi  kij (T )  kij (T ) g j (3.17.2)
x j
kij (T ) : Conductivity
122
Elements
gj : Temperature gradient
If the Fourier’s Law is substituted into the energy equilibrium equation and a
variation is taken, the following equation is attained:
 T T
 cT Td    xi kij (T )
x j
d    qext TdS   r Td 
q 
(3.17.3)
qext : Heat flux flowing in from outside
If the temperature is interpolated as a shape function of the form of T  Ni (x)Ti ,

the above equation becomes the following nonlinear simultaneous equation
consisted of nodal temperature, which includes a temperature differential term with
respect to time.
C(Ti )T  K(Ti )T  R(qext , r ) (3.17.4)
In order to calculate the time history of temperature distribution based on the above
equation, the backward difference method is applied. This method being a type of
implicit time integration can perform analysis of long time intervals and does not
generate any oscillation of solutions. The equation using the backward difference
method is as follows:
 C(Ti (t  t ))  C(Ti (t  t ))
  K (Ti (t  t ))  T(t  t )  T(t )  R(qext , r )  0 (3.17.5)
 t  t
The Newton-Raphson method is used to calculate the time history of temperature

distribution by repeatedly applying the above nonlinear equation to calculate the
temperature at the next time step.
123
Analysis Manual
The conductivity and capacitance matrices are calculated respectively as follows:
► 1-dimensional element (cross-sectional area: A )
Ni N j
Kije   k 
AdL , Cij   cNi N j AdL
e
x x
► 2-dimensional element (thickness: t )
Ni N j
Kije   kkl tdA , Cije    cNi N jtdA , k , l  1,2
xk xl
► 3-dimensional element
Ni N j
Kije   kkl dV , Cije    cNi N j dV , k , l  1,2,3
xk xl
• Relationship with Structural Elements
Unlike a structural analysis model, thermal loads and thermal boundary conditions
must be defined in a heat transfer analysis model. Except for the thermal loads
and boundary conditions, its modeling does not differ from that of a structural
analysis. Heat transfer analysis thus can be performed using the same structural
elements such as bar, shell and solid elements. The following table summarizes
the relationship between the structural elements and the heat conduction elements.
Table 3.17.1 Relationship between heat conduction elements

and structural elements
Heat Conduction Elements Structural Elements
1-dimensional Element Rod, Bar, Pipe
Membrane, Shell, Plane strain, Axisymmetric solid

2-dimensional Element
Layered shell
3-dimensional Element Solid, Layered solid
In addition to the elements in the above table, the total behavior of the temperature
degrees of freedom can be simulated with the rigid body or rigid bar elements.
124
Elements
• Analysis Results of Heat Conduction Elements
The analysis results of heat conduction elements in midas NFX are produced in a
user-defined reference coordinate system. The use of coordinate systems and the
methods of using the elements are identical to those of structural elements.
Table 3.17.2 Analysis result items of heat conduction elements

Flux component
fx , f y , fz

Flux resultant
Element f  f x2  f y2  f z2
thermal results Location: Element center

Thermal gradient component
gx , g y , gz

Thermal gradient resultant
g  g x2  g y2  g z2
125
Analysis Manual
3.18 Stress Error
In the finite element analysis of a structure, the size and distribution of the elements
have the greatest influence on the results. Generally, the appropriateness of finite
element mesh can be assessed by stress errors, but it is impossible to calculate
accurate stress errors. The stress distribution calculated in the finite element
method, however, retains a typical shape as shown in Figure 3.18.1 through which
the stress errors are calculated.
2
 1
Weighted average
wi i
Element 1
Exact stress  ex x
Figure 3.18.1 Typical shape of stresses calculated

using finite element method
Stress at a node is assumed to be the average value of the stresses calculated at

the adjoining elements.
 ex  wi i (3.18.1)
i : Stress of adjoining elements

wi : Weight ( 1 / N )
N : Number of adjoining elements
Generally, the stress calculated using the above equation is known to be more
accurate than any of the stresses calculated at the adjoining elements, and it is
assumed to be the correct solution. The stress error is then calculated using RMS
126
Elements
(Root Mean Square) of the differences between this value and the stress at each
adjoining element.
e i 
1  ( i   ex )2
(3.18.2)
 ex N
 ave is the RMS value of the stresses obtained from the entire domain of a model
relative to volume.
vi i2
 ave  (3.18.3)
 vi
vi : Element volume
The stress components used for calculating the stress error is the Von-Mises stress,
which is calculated for solid, membrane, plane strain and shell elements.
The stress error can be used to determine the elements that need be refined in
adaptive finite element analysis.
127
Analysis Manual
128
Materials
4. Materials
The materials available in midas NFX can be largely classified into structural
materials and conduction materials, each of which retains either constant material
properties or properties that change with temperature. Temperature dependent
properties can be defined for all material properties.
129
Analysis Manual
4.1 Properties of Elastic Materials
The types of linear elastic materials available in midas NFX are isotropic,
orthotropic and anisotropic materials. In addition, rigid materials are also
supported for the convenience of calculations. Orthotropic materials are defined
and used in 2-dimensional and 3-dimensional stress conditions. The following
table summarizes the relationships between the elastic materials and the elements
that can be used in midas NFX.
Table 4.1.1 Materials that can be used for each element
Element Types
Axisymmetric
Plane strain
Membrane
Surface
Material Types
Shell
Solid
Solid
Pipe
Rod
Bar
Isotropic ∨ ∨ ∨ ∨ ∨ ∨ ∨ ∨ ∨
2D orthotropic ∨ ∨ ∨ ∨
3D orthotropic ∨ ∨
3D anisotropic ∨
In order to obtain appropriate results using 2-dimensional orthotropic materials for

plane strain elements or axisymmetric solid elements, proper material properties
related to the perpendicular direction to the elemnt faces must be defined.
• Rigid Material
A rigid material retains the characteristic of a rigid body in which absolutely no
elastic deformation takes place. It is not calculated as a real material, but rather
NFX converts and processes all the elements having rigid materials into rigid
elements. All the elements retaining the identical rigid material individually behave
as a single rigid element whether or not the elements are continuously connected
or separated from one another. Also, when different rigid materials share a node,
130
Materials
all the elements retaining these rigid materials are processed as a single rigid
element, which behaves following the movement of an independent master node.
A rigid material also retains a modulus of elasticity, Poisson’s ratio and density
among which only the modulus of elasticity and Poisson’s ratio are used to
calculate the contact stiffness in case of contact, without causing any deformation
of the material. Rigid materials can be assigned to any element types, but cannot
be used for laminated composite elements.
• Isotropic Material
An isotropic material retains identical properties in any specific direction and can be
used for all the elements in midas NFX. Using the modulus of elasticity, E ,
Poisson’s ratio,  , and thermal expansion coefficient,  , the stress-strain
relationship of a 3-dimensional isotropic material is expressed as follows:
 E (1   )  E (1   )  E (1   ) 
 (1   )(1  2 ) 0 0 0 
(1   )(1  2 ) (1   )(1  2 )
 
 E (1   )  E (1   ) 
 xx  
0 0 0 
(1   )(1  2 ) (1   )(1  2 )   T 
     xx
 yy   E (1   )   yy  T 
0 0 0 
 zz   (1   )(1  2 )   zz  T 
     xy  (4.1.1)
 xy   symmetric
E
0 0  
 yz   2(1   )    yz 
      
  zx   E
0  zx 
 2(1   ) 
 E 
 
 2(1   ) 
The stress-strain relationship equation of an isotropic material in 2-dimensional

stress conditions is as follows:
131
Analysis Manual
 E E 
 0 
 xx  1   1  2
2
  xx  T 
   E E  
 yy    0   yy  T  (4.1.2)
   1   1  2
2
   xy 
  
xy
E  
0 0
 2(1   ) 

The stress-strain relationship with respect to the transverse shear is as follows:
 zx 
  zx 
 G 0   
     (4.1.3)
 zy 
  zy 
  0 G  
An isotropic material can be expressed by two material constants selected among

E , G and  . Given the two values, the remaining constant can be then
determined. For example, the shear modulus, G , retains the following
relationship:
E
G (4.1.4)
2(1   )
The range of Poisson’s ratio that is physically reasonable for an isotropic material is
as follows:
1.0   0.5 (4.1.5)
• Orthotropic Material
An orthotropic material forms a symmetry of material properties about three planes
that are perpendicular to one another. midas NFX provides orthotropic materials
for both 2-dimensional and 3-dimensional shape elements. The properties of an
orthotropic material are defined in the material’s principal axes. The stress-strain
relationship equation in 3-dimensional stress conditions is expressed as,
132
Materials
1   23 32  21   31 23  31   21 32 

 EE 0 0 0 
E2 E3 E2 E3
 11     11  11T 
2 3
       32   12 31

0   22   22 T 
1 13 31
 22  
0 0
E1E3 E1E3  
 33       T 
 
      12 21  33 33
0   
1
  12   0 0  12 
1 E2 
 23   
E (4.1.6)
 23 
   symmetric G12 0 0  
  31   G23 0 
   31 
 
 G31 
1   12 21   23 32   31 13  2 21 32 13

E1E2 E3
The stress-strain relationship equation of an orthotropic material in 2-dimensional

stress conditions is as follows:
 E1  21E1 
1    0 
1   12 21
 11   12 21   11  11T 
    12 E2 E2  
 22    0   22   22 T  (4.1.7)
   1   12 21 1   12 21   12 
 12    
0 0 G12 
 
 
The stress-strain relationship with respect to the transverse shear is as follows:
 31  G31 0   31 
    (4.1.8)
 23   0 G23   23 
Generally, an orthotropic material must satisfy the following properties:
E2 E E E E E
 212   122  1  322  3  232  2  132  1  312  3 (4.1.9)
E1 , E2 , E2 , E3 , E3 , E1
1 12 21  23 32  3113  2 21 3213  0 (4.1.10)
133
Analysis Manual
• Anisotropic Material
midas NFX permits the use of anisotropic materials only for the 3-dimensional
shape elements. The stress-strain relationship in 3-dimensional stress conditions
is as follows:
 11  G11 G12 G13 G14 G15 G16   11  1T 

   G22 G23 G24 G25 G26   22   2 T 
 22  
 33   G33 G34 G35 G36   33   3T 
    (4.1.11)
  12   symmetric G44 G45 G46   12   4 T 
 23   G55 G56   23   5 T 
    
  31   G66   31   6 T 
134
Materials
4.2 Properties of Plastic Materials
midas NFX provides plastic material models as the representative models of

inelastic materials. A plastic material model generally uses the properties in the
elastic range to express the material behavior in the plastic range, which is referred
to as an elastoplastic material. The elastoplastic theory used in midas NFX does
not reflect the strain rate (rate independent), and the behavior in the plastic range is
not related to the material temperature. Also, with the assumption that the elastic
strain is small, the following additive decomposition forms the basis of this model.
dε  dεel  dε pl (4.2.1)
el
dε : Elastic strain increment
dε pl : Plastic strain increment
Here, the strain increment can be defined as follows:
1 du du T
dε  [ ( ) ]  dD (4.2.2)
2 x x
A plastic model that does not reflect the strain rate and temperature generally
shows an elastic behavior within the inequality shown below.
f (σ, q)  0 (4.2.3)
q represents the hardening parameter, and its number varies with the type of
plastic material models. Once the stress of an elastoplastic material reaches
beyond the elastic range, plastic strain is generated, and the rate of increase in the
material’s stress only reflects the elastic strain.
dσ  Cel : (dε  dε pl ) (4.2.4)
Cel : Elastic tangent modulus
The magnitude and direction of the plastic strain generated at this point are defined
135
Analysis Manual
as,
g
dε pl  d  (4.2.5)
σ
g is known as the plastic flow potential, and  is a scalar value, which increases
with the plastic deformation. In some of the elastoplastic materials, the direction of
the plastic flow is perpendicular to the yield surface, f , and this is referred to as
an associative plastic flow.
f g
 (4.2.6)
σ σ
Together with the plastic deformation, the hardening parameters also change.
dq  d h(σ, q) (4.2.7)
The function, h , defines the hardening law and retains a relationship with the yield
surface based on the consistency condition, which must be satisfied when a plastic
flow takes place.
f f f f
df  : dσ   dq  : dσ  d   h  0 (4.2.8)
σ q σ q
g
Substituting Equation (4.2.5) into Equation (4.2.4) results in dσ  Cel : (dε  d  ),
σ
which is then substituted into Equation (4.2.8) to calculate d  as follows:
f
: Cel : dε
d  σ (4.2.9)
f f g
 h  : Cel :
q σ σ
136
Materials
g
By substituting the above equation into dσ  Cel : (dε  d  ) , the elastoplastic
σ
tangent modulus can be calculated as follows:
g f
)  ( : Cel )
(Cel :
C C 
pl elσ  σ (4.2.10)
f f g
 h  : Cel :
q σ σ
The stresses of an elastoplastic material can be calculated by integrating Equations

(4.2.4), (4.2.5), (4.2.7) and (4.2.8). Such an integration method is referred to as a
return mapping, which is a process of calculating σ n1 from the stress, σ n using
the strain increment, ε . Among a number of theories for return mapping, midas
NFX uses the implicit backward Euler method. For von Mises plastic materials,
midas NFX specifically uses the raidaul return mapping.1
Equations (4.2.4), (4.2.5), (4.2.7) and (4.2.8) are expressed as incremental

variables as follows:
σ n 1  Cel : (ε n 1  ε npl1 )
g
ε npl1  ε npl  
σ n 1 ( 4.2.11)
q n 1  q n  h n 1
f n 1  f (σ n 1 , q n 1 )  0
The above equations are a series of nonlinear simultaneous equations, which are
merely used to calculate the scalar increment,  , representing the quantity of a
plastic flow and can be calculated using the Newton-Raphson Method. The intial
 value is assumed to be 0. This is the same as calculating the stress, σ trial
n1 ,
based on the assumption of elasticity when the quantity of the plastic flow is
unknown. The plastic flow is generated when the stress, σ trial
n1 , violates Equation
1
Simo, J.C. and Taylor, R.L., “Consistent tangent operators for rate independent elastoplasticity,” Computer Methods in Applied
Mechanics and Engineering, Vol. 48, 1985
137
Analysis Manual
(4.2.3). And the stresss reaches the yield surface from σ trial
n1 through iterative
calculations of  as shown in Figure 4.2.1.
g
Plastic flow σ
Yield surface σ trial

n1
σn
σ Return mapping
σ n1
Figure 4.2.1 Direction of plastic flow and return mapping
The following table summarizes the relationships between the elastoplastic

materials and elements that can be used in midas NFX.
Table 4.2.1 Plastic materials available to use for each element
Element Types
Axisymmetric
Plane strain
Membrane
Yield
Surface
Shell
Solid
Solid
Pipe
Rod
Bar
Conditions
von Mises √ √ √ √ √ √ √
von Mises √ √ √ √ √ √ √
(with Hardening)
• von Mises Yield Condition
In the von Mises yield condition, yielding is assumed to have taken place when the
second invariant, J 2 , of the deviatoric stress reaches a specific value. This
condition is used most widely for expressing the plastic behaviors of metalic
138
Materials
materials. A perfect plastic yield condition without reflecting hardening can be

expressed by the following equation:
3
f (σ)  3J 2   y  σ dev : σ dev   y  0 (4.2.12)
2
σ dev : Deviatoric stress
y : Yield stress
Since only the deviatoric stresses are used to express the yield condition, it is
suitable to express ductile materials in which yielding occurs regardless of the
hydrostatic pressure. The von Mises yield surface in a 3-dimensional stress
condition exhibits a cylinder with a radius of 2 / 3 y parallel to the hydrostatic
pressure axis.
von Mises yield surface
1 1   2   3
2
y
3
3
1   2   3  0
π-plane
2
Figure 4.2.2 von Mises yield surface in a principal stress coordinate system
139
Analysis Manual
The plastic flow in the von Mises yield condition is assumed to be the correlated
flow, and the change in the plastic strain is as follows:
f 3
dε pl  d   d σ dev (4.2.13)
σ 2σ dev : σ dev
► Hardening Parameter
midas NFX supports isotropic hardening, kinematic hardening and combined

hardening models for the von Mises yield condition. The isotropic hardening
model assumes that the initial yield surface uniformly expands so as to maintain the
center axis of the initial yield surface.
3
f (σ, q)  3J 2   y (q)  σ dev : σ dev   y (q)  0 (4.2.14)
2
2
Isotropic hardening
Initial yield surface
· 1
Figure 4.2.3 Change in yield surface of an isotropic hardening model
140
Materials
Since the hardening parameter of an isotropic hardening model consists of the

 
effective plastic strain as q  e p   , the yield stress due to hardening is also
given as the function,  y (e p ) of the effective plastic strain and uses the input
hardening function, hy (e p ) as it is.
A combined hardening model assumes that the expansion and movement of the
yield surface simulataneously occur due to the plastic deformation taking place.
The yield surface of a combined hardening model can be defined using the yield
stress and back stress.
3
f (σ, q)  Σdev : Σdev   y (q)  0 (4.2.15)
2
Σ dev : σ dev  α
α : Back stress
2
Combined hardening
Initial yield surface
· · 1
Kinematic hardening
Figure 4.2.4 Change in yield surface of a combined hardening model
The hardening parameter of a combined hardening model becomes the effective

plastic strain and back stress.
141
Analysis Manual
e p 
q  (4.2.16)
 α 
.
midas NFX calculates the yield stress from the specified hardening function using
the combination variable, c , as follows:
 y  chy (0)  (1  c )hy (e p ) (4.2.17)
The combination variable, c  0 , corresponds to isotropic hardening, and c  1

corresponds to kinematic hardening. The plastic strain of the combined hardening
and the rate of change in back stress, which follows the Ziegler’s hardening law,
can be expressed as follows:
3
dε pl  d  Σdev (4.2.18)
2Σdev : Σdev
dhy
dα  c dε pl (4.2.19)
de p
• Hardening Curve
A hardening curve representing the plastic material properties is generally obtained

from experiments for which uniaxial tension/compression tests or pure shear tests
are widely used. The hardening curve in midas NFX is composed of entering a
true stress-plastic strain curve. The process of conversion from test results is as
follows:
► True Stress-True Strain Calculation
In case a load-displacement curve is known, the true strain and true stress can be
 L0  d   L  Pe
  log    log   ,   (4.2.20)
 L0   L0  A0
142
Materials
L0 , L : Lengths before & after deformation
A0 : Cross-sectional area before deformation
If the engineering stress and strain are known, the true stress and strain can be
  log 1   E  ,    E e (4.2.21)
 E , E : Engineering strain & stress
► True Stress-Plastic Strain Calculation
Since plastic strain takes place at the moment when a material starts to yield, the
plastic strain can be calculated as follows:

e p     el    (4.2.22)
E
E : Modulus of elasticity
143
Analysis Manual
4.3 Properties of Hyper-elastic Materials
Hyper-elasticity represents a property in which elastic deformation is maintained

through large deformation in the order of hundreds of %. Rubber typically retains
such a characteristic. The elastic behavior of a hyper-elastic material is nonlinear,
and the relationship between the deformation and stress is given in the form of
energy potential. midas NFX allows the use of hyper-elastic materials only for the
solid elements, no other element types. Since a hyper-elastic material retains a
very high bulk modulus, it has strong incompressibility, which leads to the
importance of selecting elements. In particular, the use of isoparametric elements
is not recommended for analysis using lower-order elements.
midas NFX provides polynomial, Ogden and Blatz-ko models. Functions to easily
obtain material constants from the stress-strain relationship are provided, which is
obtained from uniaxial tension/compression tests, biaxial tension tests, planar strain
tests and simple shear tests. Generally, the Ogden model is known to simulate
rubber materials better than the polynomial model, but it is important to find and
use a model that fits the best with test data.
A hyper-elastic or Green elastic material defines the stress-strain relationship from

the energy potential.
  C w  E 
S2  (4.3.1)
C E
S : 2nd PK stress
C : 2E  I (right Cauchy Green tensor)
E : Green strain
Since a hyper-elastic material uses the energy potential, it shows an identical

nd
stress condition irrespective of the path of deformation. The 2 elastic tensor
nd
corresponding to the stiffness between the Green strains and the 2 PK stresses,
144
Materials
can be calculated as follows:
 2  C  2w  E
CSE  4  (4.3.2)
CC EE
• Polynomial Model
A polynomial model defines the energy potential as follows:
Na Nd
U  A  J  3  J
i  j 1
ij 1
i
2  3 
j
D J
i 1
i 3  1
2i
(4.3.3)
Aij , Di : Material constants
In the above equation, J 3 is the 3rd invariant of the deformation gradient, and
st nd
J1, J 2 are the 1 and 2 invariants of the Cauchy Green tensor with the bulk strain,
J 3 , removed.
midas NFX provides the material degree, Na , Nd up to the 5th. If Na  1 , it is
expressed as the Mooney-Rivlin model.
Nd
U  A10  J1  3  A01  J 2  3  D  J
i 1
i 3  1
2i
(4.3.4)
If A01  0 , it is expressed as the neo-Hooknean model.
Nd
U  A10  J1  3  D J
i 1
i 3  1
2i
(4.3.5)
Regardless of this, the initial shear stiffness, 0 , and the bulk modulus, K 0 , are
defined as,
0  2  A10  A01  , K0  2D1 (4.3.6)
145
Analysis Manual
• Ogden Model
The energy potential of the Ogden model is defined from the principal stretches as
follows:
Na Nd
    D (J
i
U (1)i  (2 )i  (3 )i  3  i 3  1) 2i (4.3.7)
i
i 1 i 1
i , i , Di : Material constants
In the above equation, i is a value resulted from removing the volume change
from the principal stretches, which correspond to the eigenvalues of the Cauchy
Green tensor.
CNi  i2Ni (no summation) (4.3.8)
midas NFX provides the material degree, Na , Nd up to the 6th. The Ogden model
can also express the Mooney-Rivlin model or neo-Hookean model. The initial
shear stiffness, 0 , and the bulk modulus, K 0 , are defined as follows:
Na
  ,
1
0  i i K0  2D1 (4.3.9)
2 i 1
• Blatz-ko Model
The Blatz-ko model is mainly used for foam materials, and the property of
incompressibility is not strong. The material constant retains only the initial shear
stiffness,  , and the energy potential is defined as follows:
  I2 
U   2 I3  5  (4.3.10)
2  I3 
st nd
I1, I 2 , I3 are 1 , 2 and 3rd invariants of the Cauchy Green tensor. The equation
is in the form of potential without having separated the part pertaining to the
146
Materials
volumetric change.
• Material Constant Calculation using Stress-Strain Data

midas NFX enables the user to easily obtain the material constants of a hyper-
elastic material through the functions converting test data. The least square
method is used to obtain the material constants using the calculation of
approximation. Linear or nonlinear least square method is used depending on the
material model.
midas NFX calculates the material constants using the inputs from five different test
methods, which are uniaxial tension tests, equibiaxial tension tests, simple shear
tests, pure shear tests and volumetric compression tests. In the case of the
polynomial model or the Ogden model, the material is assumed to be
incompressible during the data processing except for the cases of volumetric
compression tests, and it must be cautioned that all the data inputs are engineering
stresses and engineering strains. Table 4.3.1 summarizes the strains and
stresses that must be entered for each test method.
Table 4.3.1 Test data for material constant calculation
Strains
Test Methods Strains Stresses
(Blatz-ko)
  / 0
Uniaxial tension   / 0 T  F / A0
  t / t0
  / 0
Equibiaxial tension   / 0 T  F / A0
  t / t0
 
Simple shear   T  F / A0
  t / t0
  / 0
Pure shear   / 0 T  F / A0
  t / t0
Volumetric
  (V / V0 )1/3   (V / V0 )1/3 T P
compression
147
Analysis Manual
► Definition of Approximate Error

In order to apply the least square method, the errors in stresses between the test
data and the values calculated from the material model must be quantified. midas
NFX uses both relative and absolute errors. The following shows a relative error.
The stresses, Ti ( A) , calculated from the material model are divided by the test
values, Titest .
2
n  Ti  A 
E  1 

i 1 

Titest 
(4.3.11)
If a relative error is used, the error in the proximity of the stress being 0 is heavily
reflected, which results in good agreement with the test data in the part with small
deformation. The following shows an absolute error, which uses the differences
between the calculated stresses, Ti ( A) and the test values, Titest .
 T  A  Titest 
2
E i (4.3.12)
i 1
If an absolute error is used, the errors from all domains are evenly considered. So
the material constants that are in good agreement with the test data can be
obtained.
► Singular Value Exclusion Method

While using the least square method, a large margin of errors may occur in
calculating the material constants due to incomplete test data. But such a risk can
be resolved through singular value exclusion in the linear least square method. If
the singular value exclusion is incorrectly used, the material constants obtained
may not coincide with the distribution of important test data. As such
approximation results must be always checked and controlled.
148
Materials
4.4 Properties of Conduction Materials
Materials used for the conduction elements are classified into isotropic and
anisotropic materials. Conduction materials are coupled with structural materials
for material definitions. The table below summarizes the relationship between
structural materials and conduction materials.
Table 4.4.1 Relationship between conduction and structural materials
Conduction Materials Structural Materials
Isotropic material Isotropic material
2-Dimensional orthotropic material

Anisotropic material 3-Dimensional orthotropic material
3-Dimensional anisotropic material
The relationship between the heat flux and temperature gradient for an isotropic
conduction material is as follows:
 fx  k 0 0   g x 
    
 f y    0 k 0  g y  (4.4.1)
   0 0 k   g 
 fz   z
The relationship between the heat flux and temperature gradient for an anisotropic
conduction material is as follows:
 f1   k11 k12 k13   g1 

    
 2
f    k22 k23   g 2  (4.4.2)
f  k33   g3 
 3  SYM
149
Analysis Manual
4.5 Composite Material Lamination Theory
The theory of composite material lamination entails the process of calculating the
average material properties of a number of laminated layers (plies or laminae) of
anisotropic materials. midas NFX uses the lamination theory for the layered shell
elements to calculate the internal forces, stiffness, etc. The plies built up in a
laminated composite are all different materials having different material principal
axes.
material  axis 2
material  axis1
material  axis1
material  axis 2
Figure 4.5.1 Principal axes of a laminated composite
At each ply, the material properties are determined by the parameters such as the
direction of fibers and the ratio of fibers to matrix. The material is by and large
orthotropic in which the principal axis (first direction) and the direction of fibers
generally coincide. The basic assumptions of the lamination theory are as follows:
► Plies constituting a laminated plate are completely adhered to one another.
► An adhesive bond retains no thickness. Displacements between the plies are
continuous, thus no shear deformations in the adhesive parts take place.
► Strain in the thickness direction exhibits a linear distribution.
150
Materials
The strain in the thickness direction and in-plane stresses at each ply are
expressed as,
ε( z )  εo  zκ
(4.5.1)
σ (i ) ( z )  C(i )  εo  zκ 
ε0 , κ : In-plane strain and bending
C(i ) : Elastic stiffness of the i th ply
The in-plane resultant force and bending moment are calculated through integration
in the thickness direction.
N 
 xx 
  h /2 h /2
N   N yy  
 
 h /2
σ (i ) ( z )dz 

 h /2
C(i )  ε o  zκ  dz

 N xy 

(4.5.2)
M 


xx 
 h /2 h /2
M   M yy  
 
  h /2
zσ (i ) ( z )dz 
  h /2
zC(i )  ε o  zκ  dz
 M xy 
 
Expressing the above in a matrix form,
 h /2 h /2 
 N  
 

 h /2
C(i ) dz
  h /2
zC(i ) dz 

εo   A B  εo 
    (4.5.3)
M   h /2 h /2   κ   B D  κ 
 
(i )
 zC dz z 2 C(i ) dz 
  h /2  h /2 
The matrices A , B and D represent the in-plane stiffness, bending stiffness

and in-plane/bending coupling stiffness for each laminated plate and serve as the
fundamental information to evaluate the stiffness of layered shell elements. The
transverse shear stiffness can be calculated with the assumption that the
transverse shear strain is constant in the thickness direction, founded on the first-
order shear deformation theory.
151
Analysis Manual
 h /2 
 
 x 
Q 
Q   
 h /2
 xz dz 
  zx 

  G  

(4.5.4)

 y 
h /2

 yz 

  h /2
 xy dz 


If the temperature factor due to the operational environment is considered, the

constitutive equation includes an average thermal expansion coefficient as follows:
 h /2 h /2 
 N   A B  εo  
 

 h /2
C(i )α (i ) dz
 h /2
zC(i )α (i ) dz 
T
  0 
  (4.5.5)
M   B D   κ   h /2 h /2   T1 
 
(i ) (i )
 zC α dz z 2C(i )α (i ) dz 
  h /2  h /2 
T0 : Average temperature change
T1 : Temperature gradient in the thickness direction
152
Materials
4.6 Composite Material Failure Criteria
The key to the strength analysis of a composite material laminated plate lies with
the failure criteria for the laminated plate. The safety of the laminated plate is
determined for the given stress or strain conditions on the basis of the failure
criteria. A number of failure criteria currently exist, but no known failure equation
exists, which can predict accurate failure in various laminated plates, specifically
various composition materials and lamination angles. Therefore, when choosing
failure criteria, various aspects such as material characteristics and existence of
test values (strength data and parameters) required by the failure criteria, must be
considered.
The types of failure criteria provided in midas NFX include the maximum stress
failure criteria and maximum strain failure criteria, both of which retain no
correlation between stress and strain components in failure calculations. Also
provided are the Tsai-Hill, Hoffman and Tsai-Wu failure criteria, which contain the
coupled terms, and the NASA LaRC02 failure criteria. Among these, the LaRC02
failure criteria have the advantage of providing the information of failure modes.
• Non-interactive Failure Criteria: Maximum Stress & Maximum Strain
Table 4.6.1 Non-interactive Failure Criteria
Direction of Transverse
Failure Criteria Shear Direction
Principal Axis Direction
Max stress X    x  X Y    y  Y S   s  S
Max strain  Ex   x  Ex  E y   y  E y  Es   s  Es
• Quadratic Failure Criteria Expanding Isotropic von-Mises Failure Criteria to

Anisotropic Materials: Tsai-Hill, Hoffman & Tsai-Wu
 x2  y2  s2 1 1  1 1 
FI   2Fxy x y       x     y (4.6.1)
XX  YY  S 2
 X X   Y Y 
153
Analysis Manual
Table 4.6.2 Quadratic failure criteria
Failure Criteria Uniaxial Strengths Fxy Fxy* (all materials)
1 Y
Tsai-Hill X  X , Y  Y   0.014    0.008
2X 2 2X

1 1 YY 
Hoffman X  X , Y  Y  0.041    0.022
2X X  2 XX 
*
Fxy
Tsai-Wu X  X , Y  Y  1  Fxy
*
1
XX YY 
• LaRC02 Failure Criteria

The LaRC02 failure criteria, developed at the NASA Langley Research Center
founded on the Hashin failure criteria, are based on assumed failure planes. The
failure theory provides the failure information for not only each ply of a fiber
composite material but also for the fibers and matrices constituting the composite
material. The status of failure of constituents of a composite material is based on
the following equations:
Table 4.6.3 Equations of LaRC02 failure criteria

Matrix tension  22  0 Matrix compression,  22  0
11  Y  11  Y 
Matrix
2 2
    2 2 2 2
cracking FI M   22    12L    eff mT    eff mL    eff T    eff L 
 Y  S  FI M  T    FI M  T   
 S   SL   S   SL 
       
Fiber tension 11  0 Fiber compression, 11  0
 22m  0  22m  0
Fiber failure 11
FI F  12m   L 22m  m   m 
2 2
1T FI F  FI F   22    12L 
SL  Y   S 
   
154
Materials
In the above table, the effective stresses and transverse shear strengths are
expressed as follows:
 T   cos  (sin    T cos  )

 eff 22
 (4.6.2)
 eff  cos  ( 12   L 22 cos  )
L

 cos  0 
S T  Y  cos  0  sin  0   (4.6.3)
 tan 2 0 
 0 represents the angle of a failure plane under a uniaxial compression load in the
direction of the matrix for which 53° is generally used.  is determined to
maximize the failure index and  , 
T L
are as follows:
 T  1 / tan 20 ,  T  S cos 20

L
(4.6.4)
Y  cos2 0
Also, the stresses of a failure plane are as follows:
11m  cos2 11  sin 2  22  2sin  cos12

 22m  sin 2 11  cos2  22  2sin  cos12 (4.6.5)
12   sin  cos11  sin  cos 22  (cos   sin  )12
m 2 2
  L  L  
 1  1  4  S   L  S  
   X   
  (G12  X ) c c  X  
  12 ,   tan 1   (4.6.6)
G12  11   22   SL  
 2  L  
  X  
   
Based on the composite material failure criteria, midas NFX produces the FE failure
index, failure index, k , or strength ratio, R , to evaluate whether or not a
composite material has failed. The FE failure index is directly extracted from the
failure equation. Accordingly, there may be a case in which no physical meaning
155
Analysis Manual
can be derived depending on the given failure criteria. If the state of stress at
failure or the stress satisfying the failure equation is σ , then the failure index, k ,
1
is determined to satisfy σ  σ FE , and similarly the strength ratio, R , is
k
determined to satisfy σ  Rσ FE . In other words, the failure index, k , and the
strength ratio, R , are helpful in determining how much of the given stress condition
must be distributed for a failure to occur.
156
Algorithm
5. Algorithm
5.1 Systems of Equation Solvers
A system of equation solver is a method of finding u , the solution to a linear matrix

equation as below.
Ku = p (5.1.1)
A system of equation solver is used not only for linear static structural analysis, but
also for all other types of analyses such as eigenvalue, buckling, dynamic and
nonlinear analyses. Generally, two types of solvers exist: a direct solver, which is
based on the Gauss elimination or LU decomposition; and an iterative solver in
which errors are minimized and solutions are converged through iterative
calculations. A direct solver is widely used in structural analysis because finding
solutions is stable without being affected by the numerical characteristics of the
matrix. But it rapidly tends to demand a significant memory space and a large
amount of calculations for a large problem in which case an iterative solver
requiring relatively less memory is more desirable. For structural analysis,
however, it must be cautioned that an iterative solver may not provide the desired
solutions due to the numerical characteristics of the matrix, or the amount of
iterative calculations may become significant in arriving at converged solutions.
midas NFX provides a function that automatically selects between the direct solver
and iterative solver depending on the scale of the problem required to be analyzed.
A direct solver obtains the solutions to a system of equations in two steps. The
first step involves matrix decomposition, and the second step pertains to the
forward-backward substitution (FBS) process. The LU decomposition method,
which is applied to a general asymmetrical matrix, can be applied to decompose a
matrix in the following form in the case of symmetrical stiffness matrix, K ,
generated in the process of finite element analysis.
157
Analysis Manual
LLT u  p or LDLT u  p (5.1.2)
L : Lower triangular matrix
D : Diagonal matrix
Generally, a matrix decomposition method that includes D is required when the

stiffness matrix is not positive definite. midas NFX uses a matrix decomposition
method in the form of LLT (Cholesky decomposition) for linear static structural
analysis and uses a matrix decomposition method in the form of LDLT for
eigenvalue analysis or nonlinear analysis since positive definiteness cannot be
guaranteed.
The important aspect of using a direct solver is to properly use the sparsity of a
matrix. Generally, the stiffness matrix, K , generated in the process of finite
element analysis is a sparse matrix containing numerous 0 terms, and the amount
of calculations and required memory space significantly change depending on the
method of utilizing the sparsity. midas NFX supports a dense solver for solving a
general dense matrix, which does not utilize the sparsity of the matrix. In addition,
midas NFX supports a multi-frontal solver as the basic direct solver, which can
remarkably reduce the amount of calculations and memory space while
appropriately utilizing the sparsity of the matrix.
The multi-frontal solver orders the degrees of freedom to minimize the amount of
calculations and memory space utilizing the sparsity of the matrix. The matrix is
then divided into a number of frontal matrices based on the ordering data, which
are subsequently decomposed. Figure 5.1.1 schematically depicts an effective
order of calculations in a rectangular element mesh based on the ordering of the
degrees of freedom. The algorithm of recursive bisection is used to implement the
ordering of the degrees of freedom. The forward substitution takes place in the
same order as the matrix decomposition while the backward substitution is
calculated in the reverse order.
158
Algorithm
1 2 1 4 1 2 1
3 3
1 2 1 1 2 1
Figure 5.1.1 Example of the order of matrix decomposition

of a multi-frontal solver
The multi-frontal solver implemented in midas NFX does not assemble and save
the stiffness matrix for the entire domain. The multi-frontal solver in midas NFX
thus requires less memory space compared to other general multi-frontal solvers.
It also supports the out-of-core analysis functionality to additionally utilize the hard
disk for calculations, which automatically becomes activated when the memory
becomes insufficient to resolve a large size problem.
An iterative solver uses a method, which reduces errors in approximate solutions

obtained through an iterative process leading to convergence. It is crucial to
rapidly reduce the convergence errors in a small number of iterations. Generally,
the number of iterative calculations depends on the preconditioning technique.
midas NFX uses the SA (smoothed aggregation) AMG (algebraic multi-grid) method,
which is known as a stable preconditioning technique, irrespective of element
1
shapes. The AMG method uses multi-grids so the number of iterations is not
greatly influenced by the number of degrees of freedom. It also exhibits stable
convergence when using elements consisted of displacement and rotational
1
Vanek, P., Mandel, J. and Bresina, M.,”Algebraic Multigrid by Smoothed Aggregation for Second and Fourth order Elliptic Problems,”
Computing, Vol. 56, 1996
159
Analysis Manual
degrees of freedom at each node such as shell elements. The iterative solver
using the AMG method automatically composes multi-grids, which are created by
the sets of neighboring nodes and the degrees of freedom representing each set of
nodes. Figure 5.1.2 shows an example of nodal sets composing a multi-grid.
Figure 5.1.2 Example of nodal sets composing a multi-grid
As explained earlier, the performance of a direct solver or an iterative solver

changes with the scale of the problem to be analyzed for which midas NFX
provides a function making an automatic selection between the two. When the
function of automatic selection of an equation solver is opted, an appropriate solver
is used on the basis of the problem size. The direct solver using the dense matrix
is selected for a small scale problem. The multi-frontal solver is selected for a
mid-scale problem. And the AMG iterative solver is selected for a large scale
problem. The criteria for automatic selection are determined by considering the
following aspects:
► If the empirical condition is known: determined on the basis of the nodes or the
number of elements specified by the user
► If the empirical condition is unknown: determined by the program based on the
number of degrees of freedom of the model and the system memory size
160
Algorithm
5.2 Eigenvalue Extraction
Eigenvalue extraction is the core algorithm for normal mode analysis and linear
buckling analysis. The form of eigenvalue extraction problems generated from
those two types of analyses is written as,
Ki  i Bi  0 (no summation) (5.2.1)
K : Stiffness matrix
B : In normal mode analysis – mass matrix ( M )

: In buckling analysis – geometric stiffness matrix ( K G )
In midas NFX, the eigenvalue extraction methods are linked to the systems of
equation solvers. The multi-frontal solver, which is the default system of equation
solver, uses the Lanczos iterative method. If the dense matrix solver is selected,
the direct method using the dense matrix is also used for eigenvalue extraction.
Each method retains the following characteristics:
• Lanczos iteration
► It is suitable for large scale problems.
► The Sturm sequence check option is recommended for possible missing
eigenvalues.
• Direct method using dense matrix

► It is suitable for small scale test models. Its performance can be drastically
3
reduced for the number of degrees of freedom at the level of 10 .
► Eigenvalues are not omitted.
The Lanczos iteration is a method, which obtains approximate eigenvalues using a

tridiagonal matrix incidental to the process of creating Krylov subspace
2
span(V1, V2 ,..., Vk ) . A block tridiagonal matrix is sometimes used to effectively
calculate eigenvalues.3 The Lanczos iteration is suitable for large scale problems,
2
Hughes, T.J.R., The Finite Element Method, Prentice-Hall International, Inc., New Jersey, 1987
3
Cullum, J. and Donath, W., “A Block Lanczos algorithms for computing the q algebraically largest eigenvalues and a corresponding
161
Analysis Manual
and its calculation speed is very fast because the size of the tridiagonal matrix is
maintained similarly to the level of the number of eigenvalues to be obtained.
Nevertheless, the use of the option to check missing eigenvalues is recommended
for finding possible omission.
The direct method using a dense matrix undergoes the process of stiffness matrix
decomposition, tridiagonal matrix generation and eigenvalue calculation. Since
the tridiagonal matrix generation and eigenvalue calculation are carried out for the
entire matrix, any omission of eigenvalues does not take place, but it is still
unsuitable for solving large scale problems.
• Range of eigenvalue calculation

Generally in a normal mode analysis, the number and the range of eigenvalues are
determined by considering modal participation factors or modal effective mass or
on the basis of the frequency domain of interest. Since the minimum critical load
is the important result item in buckling analysis, it is customary to calculate only a
few smallest critical loads. Once the number and the range of eigenvalues to be
calculated are determined, they can be specified using the data entry shown below.
Table 5.2.1 Specification method for the number and range of eigenvalues
Variable Definition Number of

Range of Eigenvalues
( v1 , v2 , N Input or No Input) Eigenvalues
v1 , v2 , N v1  v  v2 Max. N
v1 , No Input, N v1  v Max. N
No Input, v2 , N v  v2 Max. N
No Input, No Input, N   v   Max. N

v1 , v2 , No Input v1  v  v2 All eigenvalues
v1 , No Input, No Input v1  v All eigenvalues
No Input, v2 , No Input v  v2 All eigenvalues
No Input, No Input, No Input   v   All eigenvalues
eigenspace of large real symmetric matrices,” Proc. 1974 IEEE Conference on Decision and Control, IEEE Computer Society, 1974
162
Algorithm
In the above data entry, v1 and v2 represent frequencies (Hz) in normal mode
analysis and critical load factors in buckling analysis.
• Results of eigenvalues calculation

The eigenvectors, which are the results of an eigenvalue problem, satisfy Equation
(5.2.1) even if the magnitudes change.
a(Ki  i Bi )  Ki  i Bi  0

(5.2.2)
i  ai
As such a method of consistently expressing the magnitudes of the calculated

eigenvectors becomes necessary. midas NFX uses a process of eigenvector
normalization satisfying the following equations depending on the analysis types:
► Normal mode analysis: iT Mi  1
► Buckling analysis: max(i )  1 (excluding rotational degrees of freedom)
Even though the direct method with a dense matrix is used, the eigenvalue
calculation algorithm only approaches approximate solutions whose accuracy
cannot be guaranteed. Therefore, midas NFX adopts the following values as the
results of eigenvalue calculation in order to check the accuracy of the calculated
eigenvalues and eigenvectors.
Table 5.2.2 Calculation results other than eigenvalues and eigenvectors
Result Items Calculation Methods
Generalized mass bi  iT Bi
Generalized stiffness ki  iT Ki
iT1Ki iT1Bi
Orthogonality loss  i  max( , )
ki bi
Ki  i Bi
Error measure ei 
Ki
163
Analysis Manual
5.3 Effective Mass and Mode Superposition
5.3.1 Effective Mass
After calculating the natural frequencies, natural periods and mode shapes through
normal mode analysis, the results can be used to calculate important data such as
modal effective masses and modal participation factors. The directional
participation factor of the i th mode is written as i , which is calculated as
follows:
1 T
i  i MT ,   1,2,3,4,5,6 (no summation)
mi (5.3.1)
mi  iT Mi (generalized mass)
 : Direction of degree of freedom (1~3: displacement, 4~6: rotation)
T is a matrix that shows the magnitude of a directional rigid body behavior and is
defined to retain the following characteristics with respect to each node:
1 0 0 0 z  z0 y0  y   e1 
0
 1 0 z0  z 0 x  x0  e2 
0 0 1 y  y0 x0  x 0  e3 
    , e   (5.3.2)
0 0 0 1 0 0  e4 
0 0 0 0 1 0  e5 
  
0 0 0 0 0 1  e6 
x0 , y0 , z0 , represent the rotational center, and midas NFX enables the user to
specify the rotational center at a specific node or at the mass center of the entire
model. The modal effective mass is also defined for each direction, which can be
simply calculated using the modal participation factor.
  (i ) mi
mieff 2
(5.3.3)
If the effective masses of all the modes are added, the sum becomes equal to the
164
Algorithm
mass of the entire model excluding the nodes assigned with constraint conditions.
5.3.2 Mode Superposition
Mode superposition can be used in midas NFX for dynamic response analysis.
Mode superposition does not directly solve a linear dynamic equilibrium equation
as shown below. But instead the mode equilibrium equation with the problem size
reduced by eigen mode shapes obtained through eigenvalue analysis is solved.
Mu(t )  Cu(t )  Ku(t )  f (t ) (5.3.4)
Using the eigen mode shapes, Φ , the displacement in a spatial coordinate system,
u  t  , is expressed as a combination of mode displacements, ξ  t  , as follows:
u  t   Φξ  t  , Φ  1 2 ... N  (5.3.5)
Using this, the dynamic equilibrium equation (5.3.4) can be shown in reference to
the mode coordinate system.
[ΦT MΦ]ξ(t )  [ΦT CΦ]ξ(t )  [ΦT KΦ]ξ(t )  ΦT f (t ) (5.3.6)
Generally when mode superposition is used, only some of the major lower-order
modes are used with the exclusion of higher-order modes in order to compose the
eigen mode shapes, Φ . Equation (5.3.6) as such is an approximation to
Equation (5.3.4), which suggests that a sufficient number of eigen modes must be
included in the calculation to represent the true physical displacements, otherwise,
the accuracy in the calculation is compromised.
In case the mode damping matrix, ΦT CΦ , is 0, the mode equilibrium equation

(5.3.6) is independently expressed in reference to each mode as follows:
mii (t )  kii (t )  pi (t ) (5.3.7)
165
Analysis Manual
th
mi : i mode mass
th
ki : i mode stiffness
th
pi : i mode load
i : i th mode displacement
Using mode superposition, the equilibrium equation can be thus reduced to retain
the number of unknowns equal to the number of calculated eigen modes.
Especially the analysis can be very effective if the mode equilibrium equation
becomes completely separated between modes.
• Processing the damping term

In the mode equilibrium equation (5.3.6), which has been reduced using the eigen
modes, if the mode damping matrix ΦT CΦ is diagonalized, and coupling is
removed, all the modes can be separately expressed for each mode in the same
manner as Equation (5.3.7).
mii (t )  bii (t )  kii (t )  pi (t ) (5.3.8)

bi : i th mode damping
It is also expressed as follows:
1
i (t )  2 iii (t )   2i (t )  pi (t ) (5.3.9)
i
mi
 i  bi  2mii  : Modal damping ratio
i2  ki mi : Modal frequency
In midas NFX, the damping value for each mode can be entered differently based
on frequencies, in which case the modal damping value is added to the modal
damping matrix, ΦT CΦ , which is consisted of mass-proportional damping,
stiffness-proportional damping, structural damping and other general damping
values. Accordingly, if the modes can be separated out of the mode equilibrium
166
Algorithm
equation, then the modal damping matrix, ΦT CΦ , is a diagonal matrix. This

means that each proportional damping coefficient and structural damping are
constant in all the elements in the case where no damping elements (bush or
damper) exist. Otherwise, the calculation must consider the coupling between the
equilibrium equations of all the modes resulting from un-diagonalized modal
damping terms.
• Enforced motion
When an enforced motion is given in mode superposition, this cannot be directly
applied to the mode equilibrium equation. midas NFX applies such an enforced
motion through the process shown below. First, the equilibrium equation (5.3.4) is
divided into the degrees of freedom associated with the given enforced motion and
those unrelated to the given enforced motion.
 M11 M12   u1   C11 C12   u1   K11 K12   u1   f1 

            (5.3.10)
M 21 M 22  u2  C21 C22  u2  K 21 K 22  u2  f2 
u1 : Displacements of the unconstrained degrees of freedom
u2 : Displacements of the degrees of freedom constrained by enforced
motions
f1 : Loads acting on the unconstrained degrees of freedom
f2 : Restraint forces acting on the degrees of freedom constrained by
enforced motions
From the above equation, the displacements, u1 , which are unconstrained by

enforced motions, are separated into quasi-static displacements, u1qs , and dynamic
relative displacements, y ,
u1  u1qs  y
1
(5.3.11)
u1qs  K11 K12u 2
Rearranging the above for the dynamic relative displacements, y , then the
following can be expressed:
167
Analysis Manual
1
M11y  C11y  K11y  f1  M11K11 K12  M12 u2  (5.3.12)
In this equation, the damping-related term on the right hand side is ignored.
Expressing the equation in terms of the mode relative displacements, x, from
y  t   Φ11x  t  using mode superposition,
[Φ11
T
M11 Φ11 ]x  [Φ11
T
C 11Φ11 ]x  [Φ11
T
K11 Φ11 ]x  Φ11
T
f1   M11 K111K12  M12  u2 
 
1
u1  u1qs  y  K11 K12u2  Φ11x (5.3.13)
When K11 retains singularity due to the presence of a rigid-body mode in the
structure, the singularity can be eliminated by properly shifting the stiffness matrix,
K11 , using the mass matrix, M11 .
• Residual Vector
As mentioned previously, errors are generated by excluding higher-order modes
from the eigen mode shapes, Φ , when mode superposition is used. In order to
reduce such errors, midas NFX uses residual vectors, R, which are perpendicular
to the existing eigen modes for the mass matrix, M, and the stiffness matrix, K, as
follows:
R  K 1(I  MΦΦT )F (5.3.14)
F above generally consists of load vectors and also includes damping forces if
damping elements exist. midas NFX uses the method proposed by Dickens et al.
4
in which augmented mode shapes orthogonal to one another are first obtained
from the residual vectors, R, which are then added to the existing eigen mode
shapes, Φ , to apply mode superposition.
4
J.M. Dickens, J.M. Nakagawa, and M.J. Wittbrodt, “A Critique of Mode Acceleration and Modal Truncation Augmentation Methods for
Modal Response Analysis” Computers & Structures, Vol 62, No. 6, 1997, pp. 985-998
168
Algorithm
5.4 Dynamic Response Solution
5.4.1 Time Integration
midas NFX can use direct time integration and mode superposition in order to
obtain the transient response of a linear motion equation as given in Equation
(5.3.4). An implicit method is used for direct time integration for linear problems.
• Implicit direct time integration

midas NFX uses the  method (HHT-  ) proposed by Hilber, Hughes and Taylor
HHT-  is a generalized form of the Newmark
5
for implicit direct time integration.
6
method and retains a controllable numerical damping effect. It can thus control
nd
high-frequency noises and retains the 2 order accuracy for the time steps
identically to the Newmark method. The HHT-  method uses the dynamic
equilibrium equation in a modified form as follows:
Man 1  (1   H ) Cvn 1  f int, n 1  f ext , n 1    H Cvn  f int, n  f ext , n   0 (5.4.1)
a n1 and v n1 above represent the acceleration and velocity vectors at the time
step, n  1 respectively, and  H [1 3,0] is a coefficient, which determines the
numerical damping effect. When initial stresses and the effects of non-
mechanical strains such as due to thermal expansions of materials are considered,
the internal forces from linear analysis are expressed as the following equation,
which includes the product of the stiffness matrix and the degrees of freedom.
f int, n 1  Kun 1  f nonmech,n 1  f int,0 (5.4.2)
If the time difference equation from the Newmark method is introduced, the
5
H.M Hilber, T.J.R. Hughes, and R.L. Taylor, “Improved Numerical Dissipation for Time Integration Algorithms in Structural Dynamics,”
Earthquake Engineering and Structural Dynamics, Vol 5, No. 3, 1977, pp. 283-292
6
N.M. Newmark, “A Method of Computation for Structural Dynamics,” ASCE Journal of the Engineering Mechanics Division, Vol. 5, No.
EM3, 1959, pp. 67-94
169
Analysis Manual
velocities, displacements and accelerations at the steps, n, n  1 are expressed as

the following relationships:
v n 1  v n  t  a n 1  (1   )a n 
1 (5.4.3)
u n 1  u n  tv n  t 2  2 a n 1  (1  2 )a n 
2
Recomposing the equilibrium equation (5.4.1) using Equations (5.4.2) and (5.4.3),
the following form of the system of equations having the displacements as
unknowns at the time step n  1 is derived.
K eff u n1  f eff

1 (1   H )
K eff  M C  (1   H )K ,
t 2 t
f eff  f int,0  (1   H ) f ext ,n1  f nonmech ,n1    H f ext ,n  f nonmech ,n   (5.4.4)
 1 1 n  1  
M un  v   1 a n  
 t t  2
2
 
 (1   H ) n  (1   H )  n    
C u   1 v  t (1   H )   1 a n    H Ku n
  t     2   
The right hand side, f eff , of Equation (5.4.4) is determined by external forces and
by displacements, velocities and accelerations, which have already been calculated
at the time step n . Once the right hand side is determined, the displacement
vectors, u n1 , at n  1 can be calculated using the solutions to systems of
equations explained in the preceding section. Then velocities and accelerations at
n  1 are obtained by substituting the calculated displacements into the Newmark
difference equation (5.4.3). The transient response of a structure is established
through time integration in a series of the iterative process of the above.
As for the effective stiffness ( K eff ) on the left hand side of Equation (5.4.4), the
once decomposed matrix can be reused if the time step is constantly maintained.
The reuse of the decomposed matrix renders effective analysis by repeating only
170
Algorithm
the forward and backward substitution processes.

The HHT-  time integration retains unconditional stability if   (1  2 H ) / 2 and
  (1  2 H ) / 2 and becomes specialized by the Newmark method using the
average acceleration if  H  0 . midas NFX uses  H  0.05 as the default
value.
• Automatic time step control in implicit direct time integration

Linear transient response analysis by direct time integration generally uses fixed
time steps. This is to avoid increases in calculation costs resulting from the re-
decomposition of the stiffness matrix associated with changes in time steps and
noises generated from sudden changes in time steps. For these reasons, it is
more appropriate to decide fixed time steps prior to performing an analysis. midas
NFX provides a convenient method, which enables the user to specify the number
of time steps within a period based on the least-order flexible vibration mode (or the
longest flexible period). Also, in order to effectively control noises created by
sudden changes in loads, it includes an automatic time step control method, which
is based on half-step residuals.
• Damping Effect
The types of damping that midas NFX considers are mass-proportional damping,
stiffness-proportional damping and structural damping. As mentioned in Section
5.3.2, modal damping also exists, which is applicable only to mode superposition.
The damping effect in linear dynamic response analysis is applied to the damping
matrix, C , in the following form:
 1
C   ej Mej   ej K ej  K  e  ej K ej  B (5.4.5)
d d
 ej : Element mass-proportional damping coefficient of the j element

th
 ej : Element stiffness-proportional damping coefficient of the j element

th
M ej : Element mass matrix of the j element

th
K ej : Element stiffness matrix of the j element

th
 , d : Structural damping coefficient and structural damping dominant
171
Analysis Manual
frequency for the entire domain

 ej : Element structural damping of the j element
th
de : Element structural damping dominant frequency
B : Damping matrix due to damping elements (damper, bush, etc.)
• Application of mode superposition

In order to use time integration using mode superposition, the mode equilibrium
equation (5.3.6) can be re-written as follows by setting the mass to 1.
pi
i (t )  C ij j (t )  i2i (t )  pi (t )  pi (t  t )  t
t (5.4.6)
C ij  [C]ij  [Φ CΦ]ij
T
Time integration that uses mode superposition is classified into two types
depending on the state of coupling of the modal damping matrix, C ij .
① Uncoupled damping system

Once the modal damping matrix, C ij , is diagonalized and the coupling is
eliminated, the response is independently analyzed for each mode. The
displacements and velocities at each time step are determined by those at the
preceding time step as shown in the following equation. The modal integration
i th
coefficients, C ij and b , for the i mode can be found by first obtaining the
particular solution and the homogeneous solution and applying the initial condition
(displacements and velocities at the preceding time step).
in 1   a11i a12i  in   b11i b12i   pin 

 n 1    i i  n
 i i   n 1 
(5.4.7)
i   a21 a22  i  b21 b22   pi 
② Coupled damping system

If the coupling of the modal damping matrix is not eliminated, the coupling between
modes must be considered, and independent analysis for each mode cannot be
172
Algorithm
performed as previously explained. In such a case, midas NFX separates the

modal damping matrix into diagonal components ( Cdiag ) and non-diagonal
components ( Coff ) and treats the damping forces caused by non-diagonal
components as external loads for analysis.
C  Cdiag  Coff (5.4.8)
By doing so, the mode displacements become independent and the mode
velocities become coupled from which the following system of equations is
composed. If the time steps are fixed, the equations can then be solved without
re-decomposing the matrix in the same manner as the direct time integration.
I B12 Coff  ξ n1   A11

T
A12  ξ n   B11 B12  p n  Coff ξ n 
T
     n   
0 I  B CToff  ξ n1   A 21 A 22  ξ  B 21 B 22   p n1 
 22 
A  diag (a
i
) (5.4.9)
B  diag (b )
i
• Initial condition for mode superposition

Given the initial displacements and velocities, the initial displacements,  i0 , and the
initial velocities,  i0 , in modal coordinate systems are defined as below. If all the
modes are used, equations are established, and if selective modes are used,
approximate relationship equations are established.
1 T
0  i M u 0
i
mi
(5.4.10)
1 T
0  i M v 0
i
mi
i th
: i Eigen mode shape
u0 : Initial displacements
v0 : Initial velocities
173
Analysis Manual
5.4.2 Frequency Response
Frequency response analysis is a method of calculating the response of a structure

subjected to a load vibrating at a constant frequency. All the loads in frequency
response analysis are defined in a frequency domain and expressed as functions of
the excitation frequency. That is, if the angular excitation frequency is  , the
loading in frequency response analysis can be shown using a complex harmonic
function.
f  t   f   eit (5.4.11)
The response to the loading also can be expressed in the same form.
u  t   u   eit (5.4.12)
Using this, the equation of motion is expressed in the following form:
 2M  iC  K  u    f   (5.4.13)

 
The loads and displacements here are all expressed in complex numbers.
midas NFX can produce complex number data in two forms: magnitude/phase
angle or real component/imaginary component. If the value of a complex number
is expressed in magnitude/phase angle, the magnitude shows the maximum value
within the vibration period of the corresponding load or displacement, and the
phase angle shows the location (angle) where the maximum value takes place
within the vibration period. On the other hand, If the value of the complex number
is expressed in real/imaginary components, the real component becomes the
magnitude of the corresponding load or displacement at the starting point of the
vibration period, and the imaginary component becomes the load or displacement
at the ¼ period (  / 2 ) and shows the change in value following the vibration period.
The relationships between the magnitude/phase angle and real
174
Algorithm
component/imaginary component are shown in the following equations:
u  ur2  ui2 : Magnitude
  tan 1 (ui / ur ) : Phase angle

ur  u cos : Real component
ui  u sin  : Imaginary component
• Direct frequency response analysis

If a direct method is used for frequency response analysis, the frequency response,
u   , can be obtained by solving Equation (5.4.13) as a system of equations.
When no damping is present, Equation (5.4.13) becomes a system of equations of
real numbers. But with the presence of damping, the equation becomes a system
of equations of complex numbers. Accurate solutions can be obtained if a direct
method is used. But for each excitation frequency, the system of equations must
be recomposed and solved, resulting in calculations becoming very inefficient for
a relatively large problem or a problem with a large number of excitation
frequencies.
• Modal frequency response analysis

In order to apply mode superposition to frequency response analysis, Equations
(5.4.11) and (5.4.12) are substituted into the modal equilibrium equation (5.3.6),
which will now become,
 2ΦT MΦ  iΦT CΦ  ΦT KΦ ξ()  ΦT f ()  p() (5.4.14)
If the modal damping matrix, ΦT CΦ , is 0 or the coupling is eliminated by

diagonalization, Equation (5.4.14) is completely separated by modes like Equation
(5.3.8), thereby enabling to simply obtain the i th modal displacements, i ( ) .
175
Analysis Manual
pi ( )
i ( )  (5.4.15)
 2mi  ibi  ki
If the coupling of the modal damping matrix is thus eliminated by mode

superposition as described, eigenvalue analysis is performed only once, and
calculations for each excitation frequency can progress very efficiently. Its
accuracy may be somewhat compromised in comparison to that of a direct method,
but the calculations become very efficient when the problem is rather large and a
large number of frequencies is involved. If in contrary the coupling of the modal
damping matrix remains, the system of equations, which has been reduced to the
number of eigen modes, must be calculated for every excitation frequency, resulting
in the significant reduction in efficiency.
5.4.3 Random Response
An object subjected to a random load that fluctuates irregularly from a loading

source shows a very complex response with time. The magnitude of the random
response is explained through statistical characteristic values such as average,
standard deviation and probability. Examples of random response analysis
include ground motions due to earthquakes and swaying of tall buildings due to
winds. Such motions are mainly described using the power spectral density.
Figure 5.4.1 Random load
176
Algorithm
The random response analysis in midas NFX pertains to the post-processing of the
results of frequency response analysis. First, the transfer function, which
represents the ratio of the output to the input, is calculated as a frequency response
for a unit load condition, which is then multiplied by the power spectral density of
the loading source in order to obtain the power spectral density of the response. If
a number of load conditions are involved, their frequency responses are
simultaneously analyzed in the common frequency domain. Each load used in
frequency response analysis represents a distinguishable random loading source,
which can be applied to a number of nodes or elements. In addition to the power
spectral density of the response, random response analysis results include RMS
(root mean square).
• Statistical characteristic value of random response

The cross-correlation function, Rij ( ) , of physical quantities, ui and u j , at two
nodes, i and j , is defined as Equation (5.4.16). Particularly, the value at two
identical nodes is referred to as an autocorrelation function, R j ( ) . Here, R j (0) ,
being the time average value of u 2j , is used for damage analysis of structures.
1 T
T 0
Rij ( )  lim ui (t )u j (t   )d
T 
(5.4.16)
1 T
R j ( )  lim  u j (t )u j (t   )dt
T  T 0
The cross spectral density, S jk ( ) , of physical quantities, u j and uk , at two

nodes, j and k , and the power spectral density, S j ( ) , of the physical quantity,
u j , are defined as Equation (5.4.17).
2 T  it T
T 0
S jk ( )  lim e u j (t )dt  eit uk (t )dt
T  0
2
(5.4.17)
2 T
S j ( )  lim  e it u j (t )dt
T  T 0
Using the Fourier integral, the autocorrelation function and the power spectral
density can show a Fourier transform pair.
177
Analysis Manual
1 
R j ( ) 
2 
0
S j ( )cos(t )d (5.4.18)
• Calculation of random response

The transfer function, H ja ( ) , is a structural characteristic, which determines the
response, u j ( ) , at the node, j , when the loading source, Qa ( ) , is applied to
the node, a .
u j ()  H ja ()Qa () (no summation) (5.4.19)

u j ( ) : Fourier transform of the physical quantity, u j (t )
Qa ( ) : Fourier transform of the loading source, Qa (t )
The power spectral density of a response, S j ( ) , retains a relationship shown in

Equation (5.4.20) with the power spectral density of the loading source, Sa ( ) . If
a number of loading sources are statistically independent without having any
correlations with one another, the power spectral density of the total response
becomes the simple sum of the power spectral densities for the responses of all the
independent loading sources.
2
S j ( )  H ja ( ) Sa () (5.4.20)
If a number of loading sources statistically retain correlations, the strength is

expressed by the cross spectral densities, and the power spectral density of the
response can be obtained as shown in Equation (5.4.21).
S j ( )   H ja () H jb () Sab () (5.4.21)

a b
H jb ( ) : Complex conjugate of the transfer function, H jb ( )
The loading source in random response analysis does not necessarily have to be a
178
Algorithm
force at a single node. Even if a number of applied forces are completely coupled,
all of them can be combined and treated as one loading source. Also, its
response can become a specific physical quantity such as an internal force, stress,
displacement, velocity or acceleration.
• RMS
The RMS of a random response complying with a normal distribution with the
average of 0 shows a standard deviation of 1  . The magnitude of the response
during 68.3% of the response time can be assessed to be within RMS. If the
magnitude of the response is increased by two or three times, the RMS shows a
standard deviation of 2  or 3  , and the magnitude of the response during
95.91% and 99.73% of the response time respectively can be assessed to be
within RMS. The magnitude of RMS is calculated as shown in Equation (5.4.22)
using (5.4.23).
12
 1  
RMS=R j (0)1 2  
 2

0
S j ( )d 

(5.4.22)
12
   2 S ( )d 

N0   0 
j  (5.4.23)
 
 0 j
S ( )d 

Sxx
von Mises
Stress
0.0
Time
Figure 5.4.2 von Mises stress of dynamic analysis
179
Analysis Manual
• von Mises Stress

Unlike other responses, the probability distribution of von Mises stresses of a
structure subjected to a dynamic load as shown in Figure 5.4.2 is neither a normal
distribution nor has an average value of 0. The power spectral density of the von
Mises stress thus cannot be calculated by using the power spectral density of the
7
stress component. midas NFX uses the method proposed by Segalman et al. in
order to calculate the power spectral density of von Mises stresses in random
response analysis. This method requires phase information and can be used only
in complex number stresses.
S vM (i )  trace([ A]  [ S (i )])

S ( f m )  s sT (5.4.24)
s   H sx
T
H sy H sz H sxy H sxz H syz 
sT : Complex number stress component
s : Complex conjugate stress component
The matrix, A , in 3-dimensional stress conditions is as follows:
 1 0.5 0.5 0 0 0
 
 0.5 1 0.5 0 0 0
 0.5 0.5 1 0 0 0
[ A]    (5.4.25)
 0 0 0 3 0 0
 0 0 0 0 3 0
 
 0 0 0 0 0 3 
5.4.3 Response Spectrum
Response spectrum analysis is one of the methods of assessing the response of a

base motion (identical shaking of the nodes constrained by boundary conditions).
7
Segalman D. J., C. W. G. Fulcher, G. M. Reese, and R. V. Field, Jr., “An Efficient Method for Calculating RMS von Mises Stress in a
Random Vibration Environment,” Sandia Report, SAND98-0260, 1998
180
Algorithm
Response spectrum analysis especially is a most widely accepted analysis method

in assessing the structural response of a structure for seismic design. Since this
method assumes linearity in the system of response and assesses only the
maximum response, analysis using the time integration explained in Sections
(5.4.1) or (5.7) is appropriate for problems in which nonlinearity is dominant or the
results reflecting simultaneity at specific time steps are essential.
The maximum response is derived from a modal combination for which the modal
participation of the responses of individual modes corresponding to a pre-defined
spectral function is reflected. Because the simultaneity of the maximum
responses of individual modes is absent, and the maximum response is calculated
by a combination of the individual maximum responses, the results of response
spectrum analysis can be viewed as an approximation to those from time
integration. Therefore, if a spectral function is defined for a specific excitation
acceleration or a specific seismic wave, the response spectrum analysis results in
an approximate maximum value of linear transient response analysis for the
corresponding acceleration input. However, it is more common to obtain analysis
results for seismic design by using a design response spectrum, which has been
established from the statistics of historical seismic waves taken place in a specific
region or country.
• Spectral responses of individual modes

The dynamic equilibrium equation for response spectrum analysis is the same as
Equation (5.3.6), and the maximum response for each mode using the spectrum
data can be written as,
imax  max i (t )  i S D i ,  i 

imax  max i (t )   i SV i ,  i  (5.4.26)
 i
max
 max i (t )   i S A i ,  i 
SD i ,  i  : Displacement spectrum data
SV i ,  i  : Velocity spectrum data
181
Analysis Manual
S A i ,  i  : Acceleration spectrum data
i th
: Modal participation factor of the i mode in each direction
Substituting Equation (5.4.25) into Equation (5.3.5), the participation factor for the
maximum values of displacement, velocity and acceleration can be expressed in
terms of the spectrum data.
uimax  i i S D i ,  i   i i S A i ,  i  / i2

vimax  i i SV i ,  i   i i S A i ,  i  / i (5.4.27)
a max
i  i i S A i ,  i 
A point of spectrum data is defined as the absolute maximum response value for
each mode in reference to the natural period (or a natural frequency) and also
includes the effects of a modal damping ratio. A response spectrum based on a
specific acceleration history is expressed in a very complex-shaped graph due to
the diversity in the magnitudes of maximum responses by periods. In the case of
a design response spectrum, however, it is generally expressed as a combination
of simple straight lines on a logarithmic scale as shown in Figure 5.4.3.
500
Pseudo Acceleration, S A
300
200
  0.02
100
  0.05
70   0.10
50
30
102 101 100 101
Period(sec)
Figure 5.4.3 Example of an acceleration response spectrum
182
Algorithm
• Modal combination method

If the maximum physical quantity for each mode (maximum value of each
component such as displacement, stress, member force or reaction force) is called
Rimax , and the maximum value of a physical quantity is assumed to be the sum of
the maximum values of individual modes, the maximum modal values can be
simply added. However, the maximum values of individual modes do not
necessarily occur at the same time step. Simple linear superposition of the
maximum values of individual modes to obtain the maximum physical quantity will
result in a very conservative estimate.
N
Rmax   Rimax (5.4.28)
i 1
Therefore, it is necessary to introduce a modal combination method that can

approximately assess the maximum value. A number of modal combination
methods have been proposed to the extent of reflecting modal interference or
damping effects, but no method will provide suitable approximations for all cases.
For that reason, it is necessary to understand the characteristics of the proposed
modal combination methods.
① Summation of absolute values (ABS)
N
Rmax   Rimax (5.4.29)
i 1
This method assumes that the responses of all the modes retain an identical phase,
and that all the absolute maximum values of individual modes occur at the same
time. As a result, the most conservative results are provided.
183
Analysis Manual
② Square root of the summation of the squares (SRSS)
 R 
N
max 2
Rmax  i (5.4.30)
i 1
This method provides proper results when the modes are sufficiently separated
from one another.
③ Naval research laboratory method (NRL)
 R 
N
max 2
Rmax  Rmmax  i (5.4.31)
i 1, i  m
This method takes a form in which only one mode ( m ) is separated from the
SRSS method to retain an absolute maximum value. Like the SRSS method, the
NRL method provides proper results when the modes are sufficiently separated
from one another.
The preceding methods are effective only when modes are not adjacent to and
sufficiently separated from one another. In order to alleviate the limitation, the
regulatory guide 1.92 (1976) published by the Nuclear Regulatory Commission
(NRC) proposes methods that can properly assess the maximum values even if a
number of modes are adjacent to one another.
④ Ten percent method (TENP)
N  i 1 
Rmax   R i
2
 2 Ri R j  (5.4.32)
i 1  j 1 
This method includes the effects of all the modes whose frequencies are within
10% in the SRSS method. If the frequencies of two modes, i, j ( j  i) , satisfy the
following, then the two modes are considered to be adjacent to each other within
10% of the frequencies.
184
Algorithm
i   j
 0.1 (5.4.33)
i
⑤ Complete quadratic combination method (CQC)
N i
Rmax   R  R
i 1 j 1
i ij j (5.4.34)
where, ij is a cross-correlation coefficient between modes, defined as,
8  i j ( i  rij m )rij 3 2
ij  (5.4.35)
(1  rij 2 )2  4 i j rij (1  rij 2 )  4( i2   j2 )rij 2
rij : Frequency ratio
If i  j in Equation (5.4.34), ij  1 irrespective of the damping ratio, and if the

damping ratio is 0, ij  1 for all the modes and the results will become identical
to those of SRSS. If the damping ratios of the two modes are the same, it can be
simplified to Equation (5.4.26).
8 2 (1  rij )rij 3 2
ij  ( i   j   ) (5.4.36)
(1  rij 2 )2  4 2 rij (1  rij ) 2
• Signs of modal combination results

Since the modal combination methods generate the modal results in absolute
values, all the results of response spectrum analysis always retain positive (+)
values. But appropriate signs are necessary for the results pertaining to
directionality such as reaction forces and deformed shapes. The most general
method to restore the signs of combined results is to follow the signs of a major
mode. A major mode is defined as the mode whose direction is closest to the
defined direction (load direction) of the spectrum among the modes with largest
mass participation factors.
185
Analysis Manual
• Correction of spectrum data

As shown in Equation (5.4.25), spectrum data is in the form of a function related to
natural frequencies and modal damping ratios. Because of the problem that the
user has no knowledge of the frequencies prior to performing analysis, spectrum
data is defined in a tabular format with uniform intervals. Accordingly,
interpolations are used when reading spectrum values corresponding to the
frequencies or periods of a structure. A linear interpolation on a logarithmic scale
best expresses the spectrum responses with changes in natural periods, which is
the reason for its wide acceptance. Even for the spectrum data constructed for
different damping ratios, linear interpolations on a logarithmic scale are carried out
for the structure’s modal damping ratio as is done for varying natural frequencies.
If the spectrum data is constructed for only one damping ratio, no data exists for
interpolations in which case a special correction method for a single damping ratio
is required. Japanese Road Bridge Specifications (2002) proposes a correction
factor for a variation in damping ratio as shown in Equation (5.4.37).
1.5
CD      0.5 (5.4.37)
40  1
2.0
Correction factor, CD
1.5 1.5
CD     0.5
40  1
A  0.05, 1.0 
1.0
0.5
0.0
0.0 0.1 0.2 0.3 0.4 0.5
Damping ratio, 
Figure 5.4.4 Correction factors for varying damping ratios
186
Algorithm
When the damping ratio is 0.05, CD  1 (point A ). Equation (5.4.37) implicates

the correction factor when the damping ratio of the spectrum data is 0.05.
Therefore, if the damping ratio of spectrum data is not 0.05, the ratio of the
correction factor corresponding to each damping is applied as the final damping
correction factor as shown in Equation (5.4.38).
CD   i 
Rimax  Rimax (5.4.38)
CD  spectrum 
187
Analysis Manual
5.5 Nonlinear Finite Element Method
A nonlinear finite element method is a process in which an accumulated

incremental solution of iterative calculation is led to converge to the correct solution
and progresses as shown in Figure 5.5.1.
t t
fext
gi
fint,i
t
fext ui  ui 1
ui 1
t t t
u u u
Figure 5.5.1 Accumulated incremental solution and the converging process
of nonlinear finite element method
t t
In the diagram, t fext and fext represent external forces at times t and t  t
respectively, and the solutions at those time steps and the incremental solution can
be expressed in the following relationship:
t t
u t u  u (5.5.1)
u : Incremental solution generated between the time increment, t
When iterative calculations for nonlinear analysis take place in the time increment
zone, t , the accumulated incremental solution is as follows:
n
u   ui or ui 1  ui   ui 1 (5.5.2)
i 1
188
Algorithm
ui : Accumulated incremental solution generated until

th
the i iterative calculation
 ui 1 : Accumulated incremental solution generated in the
th
i 1 iterative calculation
 ui 1 is calculated through a linear system of equations using the tangential

stiffness matrix, K i 1 , as follows:
 ui 1  K i 11 gi (5.5.3)
gi : Residual force or unbalance force
The residual force, g i , is expressed as the difference between the external force,
t t
fext , and the internal force, fint ,i , as follows:
gi  t t fext  fint ,i (5.5.4)
The iterative calculation process, Equation (5.5.2) to (5.5.4), is repeated until the
user-specified convergence criteria are satisfied. The convergence criteria are
evaluated by a change in quantity such as member force, displacement or energy.
• Line search
midas NFX provides a line search function in order to enhance the performance of
the basic iterative method explained above. The basic concept of line search is to
improve the accuracy in the process of adding the incremental solution,  ui 1 ,
calculated from Equation (5.5.3), to the accumulated incremental solution by
introducing a scalar value,  . In this case, the accumulated incremental solution
is calculated by the following method:
ui 1  ui   ui 1 (5.5.5)
189
Analysis Manual
If ui 1 calculated by the above equation is assumed to satisfy the equilibrium

condition, a line search problem becomes the problem of finding  with its
differential being 0, using the principle of stationary total potential energy at the
state of equilibrium.
s( )   ui 1T g( )  0 (5.5.6)

Potential energy
Acceptable range

tan (s( ))
1
Exact solution
 uTi1g( )  0
Figure 5.5.2 Conceptual diagram of line search algorithm
If the differential value of energy, s( ) , is assumed to become linear in relation to

 , then  that satisfies Equation (5.5.6) is calculated as follows:
s(  0)
2  (5.5.7)
s(  1)  s(  0)
The slopes when  is 0 and 1 respectively can be expressed as follows:
s(  0)   ui 1T gi
(5.5.8)
s(  1)   ui 1T gi 1
Because the assumptions for the line search algorithm in reality are not exactly
190
Algorithm
satisfied, s( ) calculated from Equation (5.5.7) is not generally 0. midas NFX
repeatedly applies the procedures explained above until the value of
s( j ) / s(  0) is calculated to be less than a specific value defined by the user.
• Initial stiffness, Newton Raphson and Modified Newton Raphson methods

The iterative methods in nonlinear analysis can be classified into initial stiffness,
Newton Raphson and modified Newton Raphson methods depending on the time
of calculation. The initial stiffness method continuously maintains the tangential
stiffness calculated at the start time of analysis. The Newton Raphson method re-
calculates the tangential stiffness at every iterative calculation while the modified
Newton Raphson method calculates the tangential stiffness when changes in
external forces take place. The calculation and matrix decomposition of a
tangential stiffness matrix requires a great amount of calculation time. For that
reason, the initial stiffness and modified Newton Raphson methods can provide a
calculation speed faster than the Newton Raphson method as long as no
convergence problems exist. midas NFX does not explicitly distinguishes between
the initial stiffness and Newton Raphson methods, but the time of tangential
stiffness re-calculation is defined, which leads to gaining the effects of all the
iterative methods.
• Automatic re-calculation of stiffness matrix

It is important to select an appropriate time for a tangential stiffness calculation
depending on the characteristics of a particular problem to be analyzed including
the degree of nonlinearity and the evenness of the converged solution. midas
NFX provides an automatic tangential stiffness update as a means of solving
nonlinear finite element problems, which re-calculates the stiffness matrix at an
appropriate time by comprehensively evaluating the characteristics of convergence
or divergence. Generally, stiffness matrix re-calculations are performed if the
following conditions are met:
► If the expected number of iterative calculations is greater than the maximum
number of iterations specified by the user
► If the solution is deemed to have diverged
191
Analysis Manual
• Convergence criteria
The criteria for determining the convergence of a solution in the iterative methods
are based on the force norm, displacement norm and energy norm.
giT gi
Force norm ratio  (5.5.9)
fint,iT fint,i
 ui T  ui
Displacement norm ratio  (5.5.10)
uiT ui
 ui T g i
Energy magnitude ratio  (5.5.11)
ui T fint,i
midas NFX determines whether or not the solution has converged by comparing the
limit provided by the user with one or more of the norms.
• Divergence criteria and Load level bisection

The divergence of a solution is used as an important reference in the automatic re-
calculation of the stiffness matrix and is determined on the basis of the divergence
rate, Ei .
 uiT g i
Ei  (5.5.12)
 ui T gi 1
If the absolute value of the divergence rate is greater than 1 ( Ei  1 ), the solution
to nonlinear analysis can be assessed to have likely diverged. In such a case,
necessary algorithmic measures are taken such as stiffness matrix re-calculation or
load bisection.
Load bisection is used if it is determined that the increment of the current load step
is too large to obtain a converged solution, the examples of which include when the
solution diverges or the number of iterations required for convergence exceeds the
maximum number specified by the user. The inappropriate magnitude of the load
increment can be flexibly countered through bisecting the existing load increment
and re-starting the iterative calculations. midas NFX automatically bisects a load if
192
Algorithm
necessary until the maximum bisection level specified by the user is reached.
• Automatic control of time increment

In order to increase the efficiency of nonlinear analysis, midas NFX includes a
function, which automatically controls the magnitude of time increment based on
the convergence of nonlinear analysis. The magnitudes of time increment and the
maximum increment are determined by the user’s input. When the function of
automatic control of time increment is used in nonlinear analysis, the number of
iterations required to reach convergence in the preceding increment step is based.
The time increment at the next step will be thus increased or decreased.
t i 1  ns t i (1  ns  ns ,max ) (5.5.13)
The increment modification factor ( ns ) above is limited to a natural number, which

is designed to obtain a nonlinear solution from the start time intended by the user or
the magnitude of the load. The increment modification factor ranges from 1, which
represents the scale of the initial increment as the minimum to the value provided
by the user ( ns ,max ) as the maximum.
193
Analysis Manual
5.6 Stress/Strain Calculations considering Large Deformation
Geometric linear analysis considers no change in shape before and after

deformation in defining stresses and strains. Sstrains in geometric linear analysis
is generally defined as follows:
1 u u
ε [  ( )T ] (5.6.1)
2 X X
u : Displacement
X : Coordinates before or after deformation
Strains can be defined in various ways in geometric nonlinear analysis considering

large deformation. And there exists a stress that corresponds to each strain for
defining virtual work.
• Defining strains
The types of strains used in midas NFX are Green strain or Green-Lagrange strain
and the rate of deformation or strain rate. The Green strain tensor, E , is defined
as,
ds 2  dS 2  2dX  E  dX (5.6.2)
X : Coordinates of a structure’s specific location prior to

deformation
The Green strain can be viewed as a value representing the difference between the
square of infinitesimal length, dS , before a deformation and the square of
infinitesimal length, ds , after the deformation. The Green strain tensor can be
defined by a deformation gradient as follows:
1
E  ( FT  F  I ) (5.6.3)
2
194
Algorithm
Since the Green strain does not occur in a rigid body motion, it is an appropriate
measure of deformations. In cases where energy potential exists as in
hyperelastic materials, midas NFX uses the Green strains for the purpose of
convenience in calculations, but the resulting values provided to the user are
principal stretches and the strains calculated by using their directions.
ε  ln(1 )n1n1  ln(2 )n2n2  ln(3 )n3n3 (5.6.4)
i : Principal stretch
ni : Vector in the principal stretch direction (after
deformation)
The rate of deformation used for all other materials other than hyperelastic
materials is defined by the velocity gradient as follows:
1 v
D  (L  LT )  sym[ ] (5.6.5)
2 x
v : Velocity vector
x : Coordinates of a structure’s specific location after
deformation
That is, it corresponds to the symmetrical part of the velocity gradient tensor. The
rate of deformation can be viewed as a value representing the rate of change in the
square of an infinitesimal length.
ds 2
 2dx  D  dx (5.6.6)
t
The rate of deformation also does not occur when only a rigid body motion exists
and retains the following relationship with the Green strain.
D  FT  E  F1 (5.6.7)
195
Analysis Manual
Since the rate of deformation represents a rate of change with respect to time, it is
common to time-integrate and use it as strain. When analysis is performed for
materials other than hyperelastic materials considering the geometric nonlinearity,
midas NFX integrates the rate of deformation over time yielding the strain.
• Defining stress
When a deformation geometrically becomes large, stress can be also defined in
various ways. The types of stresses used in midas NFX are Cauchy stress and
2nd PK (Piola-Kirchhoff) stress.
n  σ  d   df  td  (5.6.8)
The Cauchy stress of a shape prior to a deformation can be transformed into the
2nd PK stress as follows:
S  JF1  σ  FT (5.6.9)

n
n0
df
df
F 1df
d
d0
Reference configuration Current configuration
Figure 5.6.1 Shapes before/after deformation and the directions

of forces for defining stresses
nd
The 2 PK stress retains no clear physical meaning, but because it is useful to
196
Algorithm
describe the equation of motion paired with the Green strain, it is mainly used to
define the behaviors of rubber materials that retain energy potential. Especially for
nd
hyperelastic materials, the 2 PK stress can be calculated from the Green strain,
which renders no need for separate stress integration. Even for such hyperelastic
materials, however, the stress values provided to the user in Equation (5.6.9) are
the Cauchy stresses calculated through an inverse transformation. For all other
materials, the stress and strain integration method explained below is used.
• Integrating stress rate and strain rate

While elasto-plastic and viscoelastic materials are not provided with energy
potential, a constitutive equation expressed in a relationship of the strain rate and
objectives stress rate is used. The Jaumann stress rate used in midas NFX is
defined as,
σ J  σ  w  σ  σ  wT (5.6.10)
The strain rate and objective stress rate are related by the constitutive equations of
the materials.
σJ  C : D (5.6.10)
In order to calculate the stress at the step, n  1 , using the stress and strain
increment calculated at the step, n , the central difference method is reflected in
Equation (5.6.10) and the rotation of the structure is considered.
σn 1  R  σn  RT  C : ε (5.6.11)
By calculating the rotational increment, R , as below, incrementally objective

8
conditions can be satisfied.
8
Hughes, T.J.R. and Winget, J., “Finite rotation effects in numerical integration of rate constitutive equations arising in large deformation
analysis,” International Journal for Numerical Methods in Engineering, Vol. 15, 1980
197
Analysis Manual
1 1
R  (I  W)1 (I  W) (5.6.12)
2 2
In particular, it must be cautioned that the strain increment and increment spin
calculations are performed in the n  1/ 2 shape.
1 u u T 1 u u T

ε  ( [ ] ), w  ( [ ] ) (5.6.13)
2 xn 1/ 2 xn 1/ 2 2 xn 1/ 2 xn 1/ 2
The integration of strain rates is calculated identically to Equation (5.6.11) using the
rotational increment of the structure, R .
198
Algorithm
5.7 Explicit Time Integration
Nonlinear transient response analysis in midas NFX is founded on explicit time

integration. In particular, midas NFX uses the central difference method, which is
most widely known among the explicit time integration methods, and lumped mass.
• Central difference
In order to apply the central difference in reference to a time axis, time steps are
divided into n  1/ 2 , n, n  1 steps.
t n 1/ 2  t n 1  t n , t n 1/ 2 
2
t  t  , t n  t n 1/ 2  t n 1/ 2
1 n 1 n
(5.7.1)
The displacement at the step, n  1 , can be calculated from the velocity at the step,
n  1/ 2 .
un 1/ 2  v n 1/ 2 
t
1
n 1/ 2 u n 1
 un  , un 1  u n  t n 1/ 2 v n 1/ 2 (5.7.2)
The velocity at the step, n  1 / 2 , can be calculated from the acceleration at the
step, n , as follows:
un  an 
1
t n
 vn 1/ 2  vn 1/ 2  , vn 1/ 2  vn 1/ 2  t nan (5.7.3)
The acceleration, a n , can be calculated through spatial discretization as follows:
Man  f n  f ext (un , t n )  f int (un , t n ), an  M1 f ext (un , t n )  f int (un , t n )  (5.7.4)
Consequently, the explicit time integration using the central difference signifies
iterative calculations from Equations (5.7.2) to (5.7.4). While using a mass matrix
199
Analysis Manual
in the form of a diagonal matrix, the cost of inverse matrix calculations in Equation
(5.7.4) becomes negligible. The above procedures are repeated at time
increments, which allow effective simulations of general dynamic behaviors of a
structure such as transient and quasistatic characteristics.
• Critical time step

The algorithm of explicit time integration is relatively simple and robust, but retains
the characteristic of conditional stability. In other words, when the magnitude of
the time step exceeds the critical value, the solution obtained by explicit time
integration diverges. As such, the critical time step or the minimum stable time
step is calculated on the basis of the smallest value among the stable time steps of
the individual elements included in the analysis model.
2
t  tcrit , tcrit   min te  (5.7.5)
max e
max above is the maximum frequency of the linearized total system, and te
represents the minimum stable time step of each element.  retains a default
value of 0.85, which is a scale factor introduced to improve the stability of explicit
time integration.
2  L 
te   min  e  (5.7.7)
max
e
 cd 
Le above represents the characteristic length of an element, which can be

generally seen as the shortest straight length passing through the inside of the
element. Dilation wave speed, cd , is calculated by material properties, and an
example for 3-dimensional elements is as follows:
M K  4 / 3
cd   (5.7.8)
 
M : Dilation modulus of elasticity
200
Algorithm
K, : Bulk modulus, shear stiffness
Because the critical time step is often very small compared to physical time, the
calculation time for internal forces of each element must be very short in order to
effectively execute explicit time integration. Therefore, those elements retaining
higher-order shape functions are almost never used. Even for the elements of
lower-order shape functions, the use of element tehchniques requiring least amount
of calculations is desired. Table 5.7.1 summarizes the element techniques and
the number of nodes, which can be used in explicit time integration in midas NFX.
Table 5.7.1 Element types that can be used in explicit dynamic analysis
Element Types Element Techniques & Number of Nodes
Rod 2 nodes
Bar 2 nodes
Pipe 2 nodes
3 nodes, Displacement assumption/ANS

Membrane, Shell 4 nodes, Reduced integration (Stabilization
technique)
3 nodes
Surface 4 nodes, Reduced integration
6 nodes
4 nodes
5 nodes, Reduced integration
6 nodes, Reduced integration (Stabilization
Solid technique)
8 nodes, Reduced integration (Stabilization
technique)
10 nodes, Incompatible element
201
Analysis Manual
• Artificial bulk viscosity

In an explicit dynamic analysis, especially for analyzing a structure with high-speed
motion, the solution of the central difference must be stabilized using the artificial
bulk viscosity. Bulk viscosity effects can be reflected by changing the internal
forces of elements.
Pressures caused by bulk viscosity are applied differently depending on whether
the volumetric strain rate is a positive number or a negative number, the rate of
which takes place in a compressed form. The pressures are also divided into two
parts, which are linearly proportional to the volumetric strain rate and proportional to
the square of the volumetric strain rate. For example, the pressures for 3-
dimensional elements are calculated as follows:
  L (C L   C1 cd  kk ) if  kk  0
2
p   e 0 e kk (5.7.9)
  Le (C1cd  kk ) if  kk  0
The pressures calculated in the above equation are reflected in the stresses of the
elements thus changing the internal forces. The pressures caused by the viscosity
effects however are not the stresses generated from the strains and constitutive
equations of the elements. So the pressures are reflected only in internal force
calculations. C0 , C1 are the coefficients required to calculate the artificial bulk
viscosity whose default values are 1.5 and 0.06 respectively. For those elements
subjected to bending such as bar and shell elements, artificial bulk viscosity for
bending deformation is additionally reflected in Equation (5.7.9).
• Damping effects
The types of damping considered in midas NFX include mass-proportional damping,
stiffness-proportional damping and structural damping. If damping effects are
considered, the equilibrium equation of explicit dynamic analysis (5.7.4) is
n 1/ 2
Man  f ext (un , t n )  f int (un , t n )  fdamp (5.7.10)
202
Algorithm

  eMe v en 1/ 2   ( e 
de 
n 1/ 2
fdamp ) C : ε d (5.7.11)
e
e e
e : Mass-proportional damping coefficient

e : Stiffness-proportional damping coefficient
,  e
d : Structural damping coefficient, Dominant frequency
C : Modulus of elasticity of material
Mass-proportional damping, stiffness proportional damping and structural damping

are calculated at each element level similarly to artificial bulk viscosity.
Damping in the central difference tends to make the critical time step small.
Generally, the damping effect on changing the time step is reflected through the
critical damping ratio.
2
te  (  2 1  ) (5.7.12)
max
e
The critical damping ratio,  , is calculated by reflecting mass-proportional,
stiffness-proportional and structural damping.
 e (e   / de )
  (5.7.13)
2 2
A change in the critical time step is caused not only by damping, but also by the
artificial bulk viscosity explained earlier. Since damping and bulk viscosity
generally make the time step small, it is not recommended to use an unnecessarily
large value.
• Mass scaling
Mass scaling is a method introduced to reduce the calculation cost in explicit
dynamic analysis. As previously explained, the critical time step is determined by
203
Analysis Manual
the element size. As such, if the sizes of some elements are very small or
become small due to deformations, the stable time step for the global analysis is
determined to be very small by the elements, which results in significant increases
in analysis time and costs. The mass scaling method artificially increases the
masses of these elements, leading to reductions in dilation wave speeds and as a
result an increase in the stable time step.
It must be cautioned however that changes in mass and inertia force of the entire
system due to an excessive use of mass scaling are reflected to the extent that will
change the dynamic characteristics. midas NFX provides the following methods
for applying mass scaling:
► A method of consistently scaling the mass of a global model

► A method of setting the time steps to a specific value
Also, considering the increase in calculation cost due to mass scaling, midas NFX
includes a function to control the frequency of mass scaling.
• Addition of joint conditions

A joint element is an element that constrains relative motions between two nodes of
elements in predefined conditions. Nonlinear transient response analysis using
explicit time integration uses a penalty method to constrain two nodes to predefined
relative movements.
If the joint constraint equation is defined as C , the constraint forces on nodes i

and j are broadly expressed as follows:
C C
fi  kC , f j  kC (5.7.14)
xi xj
k : Penalty coefficient
C : Joint constraint equation
204
Algorithm
The penalty coefficient above, k , must be an appropriate value in order to secure

the stability of explicit time integration and at the same time to accurately simulate
the specified relative behaviors. midas NFX automatically calculates the penalty
coefficient by using the critical time step value at the current step and the
eigenvalues of the linearized equation of joint motion.
205
Analysis Manual
5.8 Contact Conditions
Contact analysis is based on the assumption of a non-penetration condition under

which two objects in space can touch, but cannot pass through each other. From
the physical point of view, contact analysis corresponds to nonlinear behaviors or
conditions. The types of contacts include general contact, which analyzes a
collision between objects considering the friction generated by the collision; rough
contact without considering sliding as shown in Figure 5.8.1; welded contact in
which analysis is performed for two objects bonded at the initial stage of the
analysis as shown in Figure 5.8.1; and sliding contact, which considers only sliding
in the tangential direction. Among these types, the welded contact and sliding
contact can be viewed as linear conditions, which are assigned depending on the
locations of two objects at the beginning of analysis.
Rough contact
General contact
Figure 5.8.1 Concepts of general contact and rough contact
206
Algorithm
Welded contact
Figure 5.8.2 Concept of welded contact
• Relationships between contact conditions and analysis types

midas NFX can use welded and sliding contact conditions between initially
adjoining objects for linear structural analysis and heat transfer analysis. And
midas NFX can also use interpolation elements, which are similar to the contact
conditions. Interpolation elements connect slave nodes to the nodes of a master
segment and restrain the relative motion between them.
In nonlinear analysis (static or dynamic), general contact and rough contact can be
used in addition to welded contact, sliding contact and interpolation elements.
General contact and rough contact are considered nonlinear conditions, and their
behaviors are different depending on whether or not geometric nonlinearity is
considered in terms of analysis techniques. When geometric nonlinearity is
considered, the possibilities of contact with all master segments are considered on
the assumption that large displacements can take place. On the contrary, when
geometric nonlinearity is not considered, only the contacts between the master
segments and slave nodes adjoined within the distances initially specified by the
user are considered for calculations.
The types of contact are node to surface contact, surface to surface contact and
single surface contact among which the single surface contact can be only applied
207
Analysis Manual
to general contact and rough contact. Node to surface contact has the advantage
of using shorter calculation time, but the accuracy of the solution relatively
decreases because a node of the master object tends to penetrate into a slave
object. On the contrary, surface to surface contact demands a greater amount of
calculation time, but it can simulate the structure’s behaviors more accurately than
other types because it satisfies the non-penetration condition accurately relative to
other types of contact.
• Searching contact surfaces

Contact search uses the slave node/master segment algorithm. The algorithm
determines the status of contact depending on the proximity of a slave node to the
master segment or the degree of penetration of a slave node into the master
segment. Generally, the object containing the slave nodes and the object of
master segment can be switched. However, from the numerical point of view,
more accurate analysis results can be obtained by assigning the master segment to
an object with relatively large stiffness or coarse mesh.
The process of global search takes place in reality in order to determine whether or
not slave nodes are in contact with the master segment. The global search
represents the process of determining potential slave nodes of an object or contact
surface, which may collide in space. Contact search is performed only on the set
of slave nodes and the master segment, which have been previously identified by
the global search.
In order to assess whether or not a node comes in contact with a surface, the slave
node must be orthogonally projected onto the master segment as shown in Figure
5.8.3. If r is defined to represent a vector to Point A projected from the origin,
and x s is defined to represent a vector from the origin to the slave node, the
equations below can be obtained.
208
Algorithm
r c ,c 
  x s  r c ,c   0

(5.8.1)
r c ,c 
  x s  r c ,c   0

c ,c  above is the location of the contact point (A) expressed in a natural
coordinate system of the master segment. c ,c  satisfying the above equations
can be numerically calculated using the Newton Raphson method. The coordinate
increment,  c , c  required to apply the Newton Raphson method is as follows:
 r   r 
     
   r r       
 r        r

      r c ,c   x s  (5.8.2)
   
    

When the initial condition of c ,c    0,0 is used, the above equation readily
converges if the location of the contact point (A) or the slave node (B) is not too far
from the master segment. Now, the penetration of the slave node into the contact
segment is examined. If the slave node has penetrated, a force (contact force)
proportional to the depth of penetration is added to the slave node and the contact
segment.

nA
Figure 5.8.3 Perpendicular relationship between a slave node and master

segment
209
Analysis Manual
• Calculating contact forces

The displacement relationship between a slave node and the master segment
determined to be in contact is constrained by a penalty method. midas NFX
defines the gap and contact force between the nodes as Equations (5.8.3) and
(5.8.4).
g N   xB  x A   n A (5.8.3)
f C  kN g N if g N  0 (5.8.4)
kn : Penalty coefficient
x A , xB : Location vectors of point (A) and slave node (B) on the master
segment
nA : Perpendicular vector of point (A) on the master segment
The penalty coefficient, k N , used in the above equation retains the effect of
assigning elastic stiffness between the master segment and slave nodes. The
non-penetration condition is roughly satisfied depending on the magnitude of the
value. In the condition of welded or sliding contact, a contact force is assigned so
that the effect of preventing initially adjoined contact segment and slave nodes from
separation is reflected even when g N is a positive number. midas NFX uses
Equation (5.8.5) to automatically calculate the penalty coefficient.
f s Ki Ai2
solid elements : ki 
Vi
plane / shell elements : ki  f s M i Ai1/2
(5.8.5)
fs : Proportional coefficient
Ki : Bulk modulus
Mi : Expansion modulus of elasticity
Ai : Area
Vi : Volume
210
Algorithm
The proportional coefficient, f s , in the above equation is determined differently

based on the types of contact or analysis. The user can modify the proportional
coefficient in order to effectively satisfy the non-penetration condition or contact
condition. In a welded condition, a force that resists sliding in the horizontal
direction is assigned. to a contact segment whose magnitude is as follows:
 u B  u A   t xA 
 
gT    (5.8.6)
   y 

B A A
u u t
f T  kT gT (5.8.7)
kT : Penalty coefficient
A A
t ,t
x y : Horizontal vector at point (A) on the master segment
In the welded condition, the penalty coefficient, kT , is used same as k N . The

contact force or the resistance force against sliding in linear elasticity as expressed
in Equations (5.8.4) and (5.8.7) are not suitable for general contact and rough
contact conditions used in nonlinear analysis. Since the stiffness of general
contact and rough contact drastically changes from 0 to k N depending on the sign
of the gap, the phenomenon of many oscillations can occur until a converged
solution is obtained. In order to compensate for such phenomenon, midas NFX
uses modified contact forces as follows:
fC 0 if g N  d1
f0 g g
fC  [( N  1)(exp( N  1)  1)] if  d 2  g N  d1 (5.8.8)
(exp(1)  1) d1 d1
f0 d d
fC  [( 2  1)(exp( 2  1)  1)]  k N ( g N  d 2 ) if g N  d 2
(exp(1)  1) d1 d1
211
Analysis Manual
fC
fC 0 f C  exponential f C  linear
d1 d2 gN
Figure 5.8.4 Relationship between the gap and modified contact forces
The force resisting against sliding in the horizontal direction occurs only when a
force in the perpendicular direction is acting, which presents more complex
nonlinearity. Accordingly, Equation (5.8.7) is modified and applied as follows:
kT f C
fT  gT (5.8.9)
kN d2
By making the force in the horizontal direction proportional to the contact force in
the perpendicular direction as expressed in the above equation, discontinuity of
forces can be minimized, which can be generated at the moment when contact
suddenly disappears. In case of rough contact, Equation (5.8.9) used as is,
whereas friction can be additionally considered in general contact. The failure
function is defined as follows in order to reflect sliding caused by friction.
f  fT   f C  0 (5.8.10)
 : Coefficient of friction
212
Algorithm
The behavior satisfying the above equation becomes an elastic motion. If a

motion in the horizontal direction becomes excessive and no longer satisfies the
above equation, sliding occurs. Comparing Equations (5.8.9) and (5.8.10), the
relative displacement in the horizontal direction within a specific distance, d 2 , is
elastic motion, and the relative displacement beyond the limit is assumed to be in
sliding motion. Since general contact considering friction generates asymmetrical
stiffness matrix, the numerical calculation efficiency is significantly reduced. Also,
a large coefficient of friction (over 0.3 – 0.4) or a high modulus of elasticity in the
horizontal direction, kT , may cause convergence problems.
• Breaking-weld
The breaking-weld function in midas NFX is the same as welded contact, which
constrains the relative motion between a slave node and the master segment until
the contact force on the slave node satisfies the failure function (5.8.11). However,
if the contact force does not satisfy Equation (5.8.11) during the analysis, the
relative constraint between the slave node and the master segment is released
becoming identical to general contact.
 max  CF n ,0    CF s 
2 2
    s  1 (5.8.11)
 Ffn   Ff 
 
F fn : Perpendicular failure force
s
F f : Horizontal failure force
CF n : Contact force in the perpendicular direction
CF s Contact force in the horizontal direction

:
The contact force in the perpendicular direction in Equation (5.8.11) only considers
tension, not compression. If the failure force is specified for only one direction,
vertical or horizontal, the contact force also is considered in that direction only.
That is, only the contact force in the perpendicular direction is considered if only
F fn is provided, while only the contact force in the horizontal direction is
s
considered if only F f is provided.
213
Analysis Manual
5.9 Fatigue Analysis
Fatigue failure is a phenomenon in which a member fails when loading smaller than
the member’s yield strength is repeatedly applied. Fatigue analysis can be
performed on the basis of stress (stress-life method) and strain (strain-life method).
The stress-life method is mainly used for fatigue analysis under the condition of a
low stress level relative to the yield stress. The strain-life method on the other
hand is used in fatigue analysis under the condition of a high stress level at which
plastic deformations take place. Generally, the life obtained through the strain-life
method is relatively more accurate than that obtained through the stress-life
method. Especially when the stress level is relatively high as in a stress
concentration zone, fatigue analysis using the strain-life method is appropriate.
This manual summarizes the stress-life method, which uses an S-N curve and the
strain-life method, which uses an E-N curve.
5.9.1 Stress-life Method
The stress-life method predicts the extent of fatigue under a given loading history
using the relationship between the number of loading cycles (N) and stress
amplitude (S) at the time when failure occurs from a cyclically applied constant load.
The S-N curve shows the relationship between the stress amplitude (S) occurring
from a cyclic loading of constant amplitude applied to the structure and the number
of cycles to failure (N) when the stress of the corresponding amplitude is repeated.
For a fatigue analysis using the stress-life method, a linear elastic analysis on the
structure is performed first, and then equivalent stresses such as von Mises
stresses or principal stresses are obtained. These stresses are then applied to an
S-N curve to predict the number of loading cycles required to reach the fatigue
failure.
Each material retains different S-N curves based on various conditions. But since
fatigue tests cannot be performed for all the predictable cases, correction factors
are applied to the S-N curve, which has been obtained from a standard uniaxial
214
Algorithm
fatigue test. Also, the structures in general are subjected to cyclic loadings of
variable amplitudes. Individual stress amplitudes are extracted from the cyclic
stresses of variable amplitudes using the rainflow-counting technique and are
applied to the S-N curve.
The advantages and disadvantages of the stress-life method are as follows:

► Fatigue analysis can be performed by a relatively simple algorithm.
► Calculations are simple, and the analysis time is short.
► Since elastic deformations are dealt with, it is not valid for the cases of large
plastic deformations.
• Cyclic loading
When a stress of constant amplitude is regularly applied as shown in Figure 5.9.1,
the stress amplitude,  a , and the average stress,  m , can be calculated as
follows:
 max   min
a  (5.9.1)
2
 max   min
m  (5.9.2)
2
Figure 5.9.1 Cyclic loading amplitude and average stress
When a stress of constant amplitude with the average stress being 0 is regularly
repeated, the fatigue life is calculated using an S-N curve in the stress-life method.
215
Analysis Manual
An S-N curve typically retains the shape as shown in Figure 5.9.2, connecting two
points, P and Q by a straight line. Point P is obtained by the number of cycles
( N 0 ) causing fatigue failure when the stress amplitude in the magnitude of 90% of
the maximum allowable stress ( Su ) is applied. Point Q is calculated using the
fatigue limit stress amplitude ( S e ), which induces no fatigue failure, or the stress
6
amplitude, which induces failure upon applying cyclic loading repeated 10 times.
Generally for steel, N0  1000 and Se  0.5Su can be assumed.
1.0
P
0.9
0.8
0.7
Se
Su
0.6
Q
0.5
0.4
102 103 104 105 106 107 108
Life to failure, N (cycles)
Figure 5.9.2 Example of classical S-N curve
By defining b for the slope of the straight line on the S-N curve shown in Figure
5.9.2, the number of cycles to fatigue failure, N , pertaining to a cyclically applied
specific stress amplitude, S , can be calculated as Equation (5.9.4).
b
 log S  log Se  (5.9.3)
log Ne  log N
1
 S b
N  Ne   (5.9.4)
 Se 
• Miner’s cumulative damage theory

When a number of stress amplitudes exist, the degree of damage of a material is
calculated by accumulating individual damages from each of the stress amplitudes.
216
Algorithm
Using the number of cycles corresponding to a specific stress amplitude, ni , and

fatigue life, N i , the cumulative damage is calculated as,
ni
Damage   (5.9.5)
i Ni
The fatigue life of a structure is the reciprocal of the damage.
1
Life  (5.9.6)
Damage
• Average stress effects

If average stresses,  m , are different, the fatigue lives are also different as shown
in Figure 5.9.1, even if the stress amplitudes,  a , are identical. In order to
account for the effects of average stresses, Goodman and Gerber each proposed
the following equations. 9
► Goodman (England, 1899)
a m
 1 (5.9.7)
Se Su
► Gerber (Germany, 1874)

2
a  
  m  1 (5.9.8)
Se  Su 
It must be cautioned here that Se is a modified value for considering the average
stress. Using the modified Se , the S-N curve is reconstructed to reflect the
effects of the average stress.
9
Bannantine, J., Comer, J. and J. Handrock, Fundamentals of Metal Fatigue Analysis, Prentice-Hall, 1990
217
Analysis Manual
5.9.2 Strain-life Method
The number of cycles causing failure due to an increase in stress resulting in

plastic deformations is drastically reduced in comparison with a case in which no
plastic deformations take place. In order to consider such a phenomenon, the
degree of fatigue failure is examined from the strain-life relationship, which is
referred to as the strain-life method. In order to calculate plastic strains at the
state of cyclic loading, the history of strains obtained through a nonlinear analysis
can be directly used. Alternatively, the Neuber’s Law can be applied to the stress
results of a linear elastic analysis to obtain the plastic strains at the state of cyclic
loading. 10 Linear elastic analysis in conjunction with the calculation of plastic
strains by the Neuber’s Law was widely used in the past due to relatively less
demand for calculations. For more accurate strain calculations however, the
nonlinear history analysis method has been the preference lately. Similar to the
stress-life method, the strain-life method is also suitable for cases with plastic
deformations taking place in certain stress concentration zones rather than the
cases of entire structures developing plastic strains.
• Linear elastic analysis and strain calculation using the Neuber’s Law
The magnitude of a cyclic strain under the action of cyclic constant loading is
obtained by summing the elastic deformation and the plastic deformation as
expressed in Equation (5.9.9). The magnitude of the elastic deformation (  e ) is
calculated from the change in stress (  ) and the modulus of elasticity ( E ). The
magnitude of the plastic deformation (  ) is calculated using the change in
p
stress, cyclic deformation strength coefficient (   ) and the coefficient of cyclic

strain hardening ( n ). Since the change in stress used here is also a stress, which
reflected plastic deformation as shown in Figure 5.9.3, it must be calculated by
using the Neuber’s Law expressed in Equation (5.9.10). The stress obtained from
the results of linear elastic analysis corresponds to K f S in Equation (5.9.10).
10
Neuber, H. (1961) "Theory of Stress Concentration for Shear-Strained Prismatic Bodies with Arbitrary Nonlinear Stress-Strain Laws,"
Journal of Applied Mechanics, Trans. ASME, Vol. E28, p. 544-550.
218
Algorithm
The change in strain in Equation (5.9.9) is obtained by an iterative method because

the magnitude of change in stress includes the magnitude of change in strain.
The magnitude of change in stress due to cyclic loading of variable amplitudes is
calculated using the rainflow-counting technique in the same manner as the stress-
life method.
1
  e  p     n
    (5.9.9)
2 2 2 2 E  2  
1 1
K f  S e     
2 2
(5.9.10)
Kf : Stress concentration factor

S : Magnitude of change in nominal stress (magnitude of change in
stress at a location where stress concentration effects do not exist)
The Neuber’s Law differently defines the relationship between the elastic stress
and strain as shown in Equations (5.9.11) and (5.9.12) depending on the types of
equivalent stress to be applied.
► von Mises equivalent stress
 K f S   
2
S S
e   , (5.9.11)
3G  3E  3G
 
 
2 1  v 
► Maximum shear stress or maximum principal stress
K S 
2
S
e   
f
, (5.9.12)
E E
219
Analysis Manual
 e  p
 B
A

2
O 
  
C 2
2 2
Figure 5.9.3 Classical stress-strain curve when plastic deformation occurs
• Strain calculation using nonlinear history analysis

In case a strain history used in the strain-life method is directly calculated through
nonlinear analysis, the cyclic deformation strength coefficient (   ) and the
coefficient of cyclic strain hardening ( n ) are not used. Instead the nonlinear
material properties are specified for nonlinear analysis, and the nonlinear analysis
results are used. The strain is the sum of elastic strain and plastic strain. Each
strain can be calculated using the principal strains ( 1 ,  2 ,  3 ) as shown in Equations
(5.9.12) – (5.9.14), depending on the type of equivalent strains. Equivalent stress
or nonlinear analysis results are used to calculate the effects of the average stress.
In order to consider the changes in stress and strain simultaneously, the multi-
components rainflow-counting technique, which uses a number of components
simultaneously, is applied.
220
Algorithm
► von Mises equivalent strain

1
 e 1   e  e  2   e  e  2   e  e  2  2
  1  2    2  3    3  1  
2 2 1     2 2   2 2   2 2  

1
(5.9.12)
2  1p  2p    2p  3p    3p 1p  
2 2 2 2
 p
           
2 3  2 2   2 2   2 2  
 
► Maximum shear strain
 e  1e   2e  2e   3e  3e  1e 
 maximum  or or 
2  2 2 2 
 
(5.9.13)
 p  1p   2p  2p   3p  3p  1p 
 maximum  or or 
2  2 2 2 
 
► Maximum principal strain
 e 1e

2 2
(5.9.14)
 p 1p

2 2
• E-N curve
Figure 5.9.4 represents an E-N curve, which shows the relationship between the
strain and the number of cycles causing fatigue failure as expressed in Equation
(5.9.15). By applying a number of individual strain amplitudes to the E-N curve,
which have been extracted through rainflow-counting, each number of cycles and
the corresponding individual damage level are obtained. As was done in the S-N
technique, the Miner’s cumulative damage theory is used to linearly accumulate
individual damage levels and predict the final damage and life.
  f
'
  2 N    'f  2 N 
b c
(5.9.15)
2 E
221
Analysis Manual
 f
100
Total strain
  f
c
 2 N    f  2 N 


b c
101
2
2 E
Strain amplitude
102
b
103
Elastic strain
 f
104 E
Plastic strain
105
100 101 102 103 104 105 106 107
Reversals to failure, 2N
Figure 5.9.4 Relationship between strain (  ) and the number of cycles to

fatigue failure ( N )
The advantages and disadvantages of the strain-life method are as follows:

► Using the results of linear elastic analysis, the method can be applied to fatigue
analysis for a case in which partial plastic deformations occur.
► It is possible to calculate using the results of nonlinear history analysis.
► Calculations are complex and the analysis time is long.
• Average stress effects

If the average stress,  m , is different as shown in Figure 5.9.1, the fatigue life is
also different even if the magnitude of change in strain in a structure is identical.
In order to consider the effects of average stress in the strain-life method, the
following proposed equations are used:
► SWT (Smith-Watson-Topper, 1970)
  f 
' 2
 2 N    'f  'f  2 N 
bc
 max 
2b
2 E (5.9.16)

 max  m
2
222
Algorithm
► Morrow (1968)
  f   m 
'
  2 N    'f  2 N 
b c
(5.9.17)
2 E
223
Analysis Manual
5.10 Stress Linearization
Stress linearization signifies a calculation process that expresses complex stress

conditions in terms of equivalent membrane stresses and bending stresses. It is
mainly used for the design of pressure vessels and is based on the ASME B&PV
code.
Stress linearization methods differ depending on the stress conditions of finite

element models and can be largely classified into the 3-dimensional stress state
and axisymmetric stress state. Solid elements and plane strain elements are
assumed to be at the 3-dimensional stress state, and axisymmetric solid elements
are assumed to be at the axisymmetric stress state. The section to which
linearization is applied can be specified anywhere by the user, but a section is
generally defined in the thickness direction in a finite element model.
When Points A and B are defined in a 3-dimensional model as shown in Figure

5.10.1, a section and a coordinate system for linearization can be determined.
The direction pointing from Point A to B is defined to be the x-axis. The y-axis is
assumed to exist on a plane formed by the x-axis and GCS y-axis. The
membrane stress is the average value of the stresses acting in the linearization
section.
1 t/2
t t / 2
σm  σdx (5.10.1)
The bending stress in the form of equivalence to the 1st moment of the stresses
acting on the section can be calculated as follows:
6 t/2
t 2 t / 2
σbB  σbA  xσdx (5.10.2)
224
Algorithm
A x
B
Figure 5.10.1 Defining the section and coordinate system for stress linearization (3-
dimensional model)
• Axisymmetric stress state

In an axisymmetric model as shown in Figure 5.10.2, individual components of the
stresses are calculated by different methods. The coordinate system for stress
linearization in an axisymmetric model is determined solely by Points A and B.
Linearization in an axisymmetric model basically uses the same principle as for a 3-
dimensional model. However, because the neutral axis has moved to the outer
side from the center of x  0 , slightly different methods are used to calculate
individual stress components.
The meridional stress corresponds to  y , and the following equations are used to
calculate the membrane stress and bending stress:
1 t/2
RC t t / 2
 ym   y Rdx (5.10.3)
x  x f t/2
 yb,  t/2  ( x  x f ) y Rdx ,   A, B (5.10.4)

t / 2
(x  xf ) 2
Rdx
t / 2
xf : Distance between the center of the section and the

neutral surface
The hoop stress corresponds to  z , and the following equations are used to
225
Analysis Manual
1 t/2 x
t t / 2
 zm   z (1  )dx (5.10.5)

x  x f t/2 x
 zb,  t/2 x 
t / 2
( x  x f ) z (1 

)dx ,   A, B (5.10.6)
 ( x  x f ) (1 
2
)dx
t / 2 
xf : Distance between the center of the section and the

neutral surface
The radius of curvature,  , is a value defined not by the shape of the model, but
by the user. If the radius is not specified, the program assumes that no curvature
exists. Also, it must be cautioned that x f is different from the value used to
calculate the meridional stress.
GCS  z
y B
x
xf

A
RA
RC
Neutral surface
RB
GCS  x
Figure 5.10.2 Defining the section and coordinate system for stress linearization
(axisymmetric model)
The thickness stress corresponds to  x , and the following equations are used to
226
Algorithm
1 t/2
t t / 2
 xm   x dx (5.10.7)
 xb,   x, A   xm ,   A, B (5.10.8)
The shear stress is assumed to retain a parabolic shape, and 0 is assumed at the
two endpoints, A and B. Therefore, the bending stress component does not exist,
and only the membrane stress component is calculated by the following equation:
1 t/2 R
 xym 
t  t / 2
 xy dx
RC
(5.10.9)
227
Analysis Manual
5.11 Topology Optimization
Toplogy optimization is a layout optimization problem, which determines the

distribution of materials most suitable to a given objective. It is primarily used to
produce a fundamental basis for the designer’s engineering judgment at the
conceptual design stage or generate ideas for new alternatives. In addition,
toplogy optimization can lead to more realistic design in reflection of various
constraints associated with the problem at hand.
In order to express the distribution of materials in toplogy optimization, density

variables of the finite elements created for analysis are used. The element density
of 1 represents a part that requires the element, while 0 represents a part that does
not require the element. Unlike general design optimization, the only design
variable is the element’s density, which determines whether or not the material is
used. As such, the user does not specify separate design variables but composes
an optimization problem using only the combinations of objective functions and
constraints. This section explains the types of toplogy optimization problem
composition and optimization techniques provided by midas NFX.
5.11.1 Objective Functions and Sensitivity
The problem composition method of toplogy optimization is a method of finding

design variables that maximize or minimize the objective functions under the
specified constraints like general optimization methods. In case an objective
function is minimized, the toplogy optimization problem composition can be
expressed as Equation (5.11.1).
Minimize: c  x 
(5.11.1)
subject to : h i  x   0 (i  1,..., M )
228
Algorithm
c : Objective function
hi : Constraints
x : x   x1, , xe , , xN  0  xe  1.0,  e  1, , N  ( xe :요소밀도)
x   x1, , xe , , xN  0  xe  1.0,  e  1, , N  ( xe : element density)
M : Number of constraints
The types of toplogy optimization problem composition provided in midas NFX can
be classified as Table 5.11.1 depending on objective functions.
Table 5.11.1 Types of problem composition based on objective functions
Production
Design
Objective Functions Related Analysis Conditions
Constraints
(common)
Static compliance Volume fraction · Linear static analysis

(minimum)
Dynamic compliance Volume fraction · Frequency response

· Molding
(minimum) analysis
direction
Volume ratio (minimum) Displacement/Stress · Linear static analysis
· Symmetrical
· Frequency response
condition
analysis
(1-3-axis
(displacement
symmetry)
constraint)
Average eigenvalue Volume fraction · Modal analysis

(maximum)
• Static compliance
Static compliance being a function of element density is expressed in the form of
global deformation energy as shown in Equation (5.11.2).
229
Analysis Manual
N
cs  f Tu  uTKu   ueT ke  xe  ue (5.11.2)
e 1
f : Load vector
u , ue : Global & element displacement vectors
K , ke : Global & element stiffness matrices
The sensitivity of static compliance is expressed as Equation (5.11.3) applying the

adjoint method to Equation (5.11.2).
cs f T K
2 u  uT u (5.11.3)
xe xe xe
The sensitivity for the loads included in Equation (5.11.3) exists for the loads in the
form of body force, which varies with the element density, such as gravity and
rotational inertia force.
• Dynamic compliance
Dynamic compliance defines the compliance for complex number responses such
as frequency responses, which is expressed in the form of the magnitude of a
complex number as shown in Equation (5.11.4).
cd  f Tu (5.11.4)
The sensitivity of dynamic compliance uses the adjoint method and introduces the
conjugate complex number expression as Equation (5.11.5).
cd  f Tu  f T S  
  T 2 u  uT u  (5.11.5)
xe  f u  xe xe  

S : Dynamic effective stiffness:  2M  iC  K , refer to (5.4.13)
230
Algorithm
f ,u : Conjugate complex numbers of f and u respectively

  : Real number part of complex number
• Average eigenvalues
The average value of eigenvalues in a design mode uses the reciprocal formulation
11
of eigenvalues as Equation (5.11.6).
1
 m w
   i  (5.11.6)
 i 1 i 
 : Average eigenvalue
i : i th eigenvalue
wi : Weight factor for i th eigenvalue
m : Number of design modes
The sensitivity of the average eigenvalue can be expressed using the sensitivity of
eigenvalues as Equation (5.11.7).
 m
w   
  2  2i  i  (5.11.7)
xe i 1 i  xe 
The sensitivity of eigenvalue for each mode in Equation (5.11.7) is expressed in

terms of eigenvector, element matrix and mass matrix of the corresponding mode
as shown in Equation (5.11.8) if it is not a duplicate eigenvalue.
i  K M 
 iT   i i (5.11.8)
xe 
 ex xe 
11
Ma Z.-D, Kikuchi N. and Cheng H.-C., “Topological design for vibrating structures,” Comput. Methods. Appl. Mechs. Eng., Vol. 121,
1995
231
Analysis Manual
• Volume fraction
Volume fraction represents the ratio of the volume at the state of finalized material
distribution to the volume of the entire design domain. In the case of
homogeneous material distribution, the mass ratio and the weight ratio are the
same. Since the design variable of toplogy optimization is the density of elements,
the sensitivity of volume fraction can be expressed as the ratio of the original
volume of the corresponding element to the total volume of the design domain.
 N 
   ve xe V0 
 V  x  / V0 
  e 1   ve (5.11.9)
xe xe V0
V0 : Total volume of the design domain
V : Total volume of the design domain reflecting the

element density
ve : Volume of the element, e
5.11.2 Material Interpolation Scheme
Material interpolation schemes are introduced to correct mass, which is dependent

on the physical stiffness or element density toplogy of actual material when the
density of an element retains a value in between 0 and 1 during the process of
toplogy optimization. midas NFX selectively uses the two methods shown in Table
5.11.2.
232
Algorithm
Table 5.11.2 Material interpolation schemes

SIMP RAMP12
Solid Isotropic Material with Rational Approximation of
Penalization Material Properties
xe
Interpolation equation ke  xe   xep ke0 ke  xe   ke 0
1  q 1  xe 
ke ke 1 q
Differential equation  pxep 1ke 0  ke 0
xe xe 1  q 1  xe  2
 
Related Analysis Linear static Modal, frequency response
Penalty coefficient p  3.0 ~ 4.0 q  5.0 ~ 6.0
1.0
0.8
Young's modulus ratio
0.6
p=1.0 or q=0.0
p=3.0
0.4 q=5.0
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
Material density, x e
Figure 5.11.1 Density dependent stiffness ratios for material interpolation

models
12
M. Stolpe and K. Svanberg, “An alternative interpolation scheme for minimum compliance topology optimization,” Structural and
Multidisciplinary Optimization, Vol. 22, 2001
233
Analysis Manual
5.11.3 Optimality Criteria
In an optimization problem defined in Equation (5.11.1), optimality criteria (OC)

include constraints together with Lagrange multipliers in the objective function thus
composing it into an unconstrained optimization problem. The criteria then find
the Karush-Kuhn-Tucker (KKT) conditions in which the differential value of the
newly constituted objective function becomes 0.
c M h
  i i  0 (5.11.10)
xe i 1 xe
i : Lagrange multiplier
In case only one constraint exists, Bendsø e proposed an empirical design variable
update method13 using the optimality criteria as follows:
max  xmin , xe  m  if xe Be  max  xmin , xe  m 



xe new   xe Be if max  xmin , xe  m   xe Be  min 1, xe  m  (5.11.11)

min 1, xe  m 
 if min 1, xe  m   xe Be
 : Numerical damping, 0    0.5
m : Update limit length, 0  m  0.65
xenew : Updated density
c h
Be :     1 at optimality 
xe xe
13
Bendsø e, M.P., Optimization of structural topology, shape and material, Springer, Berlin, Heidelberg, New York, 1995
234
Load/Constraint
6. Load/Constraint
6.1 Loads
The types of static loads that can be used in midas NFX are force, inertia force,
displacement and thermal expansion, which are summarized in Table 6.1.1.
Table 6.1.1 Types of loads that can be used in midas NFX
Types Range of Applications
Nodal force Nodes
Pressure load Two- and three-dimensional elements
Bar element loads Bar elements
Gravity All elements with mass
Rotational forces All elements with mass
Specified displacements Nodes
Temperature loads Nodes, Bar elements, Shell elements
Preloads Bar elements
• Nodal loads
A nodal load is the most basic type of load, which allows the user to input three
force components and three moment components at each node. The directions of
the components can be defined in any specific coordinate system.
• Pressure loads
A pressure load is entered in the form of a distributed load on the faces or edges of
elements. It can be used for two-dimensional or three-dimensional elements. Its
application directions can be in the axis directions of a specific coordinate system,
in a specific vector direction and the vertical direction. Figure 6.1.1 shows some
examples of pressure loads applied to various elements.
235
Analysis Manual
Figure 6.1.1 Pressure loads applied on various elements
• Gravity
Gravity is used in modeling self-weight or inertial force of structures, and it can be
applied to any elements having masses. A different magnitude of gravity can be
applied depending on the location, or it can be applied in the directions relative to
any specific coordinate system.
• Rotational inertial forces

Rotational inertial force is used to model the inertial forces resulting from centrifugal
forces and angular accelerations generated from angular velocity where a structure
rotates about a specific axis.
ω, α
r
αr
z
y m
F
x
ω  (ω  r)
Figure 6.1.2 Defining rotational inertia force and the direction

of the applied force
236
Load/Constraint
As shown in Figure 6.1.2, the rotational inertia force of the node that rotates about
the axis, R , is expressed as follows:
F  m(ω  (ω  r)  α  r) (6.1.1)
ω : Rotational angular velocity vector
r : Location vector from the rotating axis
α : Rotational angular acceleration vector
• Forced displacements
A forced displacement is specified at a specific node when the location of the node
is known after its deformation has taken place. Because a forced displacement
creates a deformation in the structure, it is categorized as a load, but it generates
constraint forces, which is the characteristic of a boundary condition.
If u A represents all the degrees of freedom of a problem to be analyzed, the
following shows how the degrees of freedom are distinguished between the ones
assigned with forced displacements and the ones without.
u 
uA   F  (6.1.2)
u S 
uF : Degrees of freedom without assigned forced displacements
uS : Degrees of freedom with assigned forced displacements
Using the same principle, stiffness matrices can be distinguished and expressed as
follows:
K K FS  u F  f F 
K AAu A   FF  f  
K SS  u S  A f S 
(6.1.3)
 K SF
Since u S is an already determined value in the above equation, the second row of
the matrix equation does not retain any meaning. By using u S and rearranging
the first row, the load by a forced displacement can be calculated as follows:
237
Analysis Manual
K FF uF  f A  K FS uS (6.1.4)
• Temperature loads
A temperature load is expressed in terms of T , which represents the difference
between the initial and final temperatures. The temperature load generates a
strain, ε  α(Tm )T , based on the coefficient of thermal expansion, which is
determined by the material temperature, Tm . As such, the initial and the final
temperatures must be defined for each structure. Table 6.1.2 shows the means to
define a temperature.
Table 6.1.2 Types of temperatures that can be defined on structures
Basic temperatures All nodes at which temperatures are not defined
Nodal temperatures Nodes
Bar, membrane, plane strain, shell elements

Element temperatures
Temperature gradient: bar, shell elements
The temperatures shown in Table 6.1.2 may be used for both initial and final
temperatures. The reference temperature, which is defined in material properties,
is used if the initial temperature is not specified.
• Preloads
A preload represents a member force to which an element is subjected at the initial
stage of analysis. Bolt connection forces are an example. Preloads are analyzed
separately from other loads, and the analysis procedures are as follows:
For example, if P is expressed to represent the magnitude of the preload that is
being applied on a bar element, it is considered as the element’s internal force, and
the analysis for the preload is then performed.
Ku  fint (6.1.5)
fint : Element’s internal force by the preload, P
238
Load/Constraint
By ignoring the element’s stiffness under a preload, the magnitude of the preload is
accurately delivered to adjoining elements. The analysis results are considered as
initial stresses and applied to the analysis along with general loads.
Ku  f  fint (σ) (6.1.6)
fint (σ) : Element’s internal force by an initial stress
239
Analysis Manual
6.2 Constraint Conditions
Constraint conditions are classified into single-point constraints and multi-point

constraints. A single-point constraint means that a constraint condition is applied
to a single node. A multi-point constraint means that a number of nodal degrees
of freedom are constrained to maintain certain relationships between the degrees of
freedom.
• Single-point constraints
A single-point constraint selectively fixes the degrees of freedom of an individual
node, which results in the removal of the corresponding degrees of freedom. The
single-point constraints are typically used at supports where no displacements are
expected or to specify symmetrical conditions. Single-point constraints are also
used to eliminate degrees of freedom, which need to be excluded from the analysis.
This method is equivalent to removing singularities in the stiffness matrix. It is
especially important to properly set the directions to which the constraint conditions
need be applied. For example, the point at which two rod elements meet, as in
the following exercise, can have a degree of freedom only in the axial direction.
No other degrees of freedom must be included in the analysis. The other
directions must be appropriately constrained to eliminate singularities in the
stiffness matrix. But appropriate directions of constraints cannot be determined if
the coordinate axes are defined in the x-y coordinate system as shown in the figure.
Therefore, a new nodal coordinate system needs to be defined, which consists of
the element’s axial direction and the vector, n, which is in the perpendicular
direction. The n direction is then constrained, which will result in proper
constraints for the analysis.
240
Load/Constraint
y
n
ROD 2
ROD 1 x
Figure 6.2.1 Example of the usage of a nodal displacement coordinate system
• Multi-point constraints
A multi-point constraint applies constraint conditions using linear relationship
equations between a number of nodal degrees of freedom. The general form of
the linear relationship equations is as follows:
R ju j  0 (6.2.1)
Rj : Coefficient of linear relationship equation
uj : Degrees of freedom related to constraint condition
When a number of multi-point constraint conditions are involved, the equations can
be expressed in a matrix form as,
R M uM  0 (6.2.2)
In order to apply the above multi-point constraint conditions to a system of equation

solver, u M is separated into independent degrees of freedom (DOF) and
dependent degrees of freedom, and then the dependent degrees of freedom are
eliminated from the system of equations. First, the degrees of freedom associated
with the multi-DOF constraint conditions are classified into independent DOF, u I ,
and dependent DOF, u D , as follows:
241
Analysis Manual
u 
uM   I  , R M  R I RD  (6.2.3)
u D 
Using the above equation, Equation (6.2.2) can be expressed as,
R I u I  R Du D  0 (6.2.4)
If the inverse matrix of R D exists in the given equation, the relationship equation
between the independent DOF and dependent DOF can be rearranged as follows:
uD  RD1R I uI  GuI (6.2.5)
Using this equation, the dependent DOF, u D , can be eliminated from the system of
equations, which consist of the entire model.
Because the range of applications for multi-point constraints is very extensive,

multi-point constraints can be utilized in various cases as follows:
► When relative motion between two nodes needs to be simulated
► When hinges or sliding joints need to be simulated
► When the adjoining parts of different elements of inconsistent numbers of
degrees of freedom need to be connected
► When loads need to be distributed and subsequently applied
► When constraint conditions need to be applied in directions that do not coincide
with the displacement coordinate system assigned to a node
Constraints between the degrees of freedom generated in rigid body/interpolation

elements correspond to a type of multi-point constraints. Nevertheless, the use of
rigid body/interpolation elements is more convenient for the types of behaviors
expressed by such elements.
• Automatic single-point constraints

midas NFX has an automatic single-point constraint function, which automatically
242
Load/Constraint
detects the singularity of stiffness matrix (single-point singularity) at the node level
and automatically assigns constraint conditions. If this function is used, a 3x3
stiffness matrix, which comprises displacements or rotations at the node level, is
analyzed, and constraint conditions are created in the directions close to the zero
stiffness.
When automatic constraints are used in the model consisted of rod elements as
shown in Figure 6.2.1, a constraint condition is automatically created in the n
direction, which does not retain any stiffness component, even if the previously
explained single-DOF constraint is not defined. In such a case, it is crucial to
define a nodal displacement coordinate system, which includes the element’s axial
direction and its perpendicular vector, n , in the same manner as before.
• Calculating constraint forces

Constraint forces act on the degrees of freedom to which constraint conditions are
assigned. Once the solution has been obtained, the constraint forces for single-
point and multi-point constraints can be calculated as below. The single-point
constraint force, f S , and the multi-point constraint force, f M , must satisfy the
following equation of equilibrium:
fint  fext  fS  fM (6.2.6)

fext : External load vector
: Internal force vector (  BT σd )

fint
The above equation is now separated into the components of independent degrees
of freedom ( , I ) and dependent degrees of freedom ( , D ) and then a condition is
imposed not to apply single-point constraints and multi-point constraints to the
same degrees of freedom.

 fint, I 
 
 fext , I 
 f S , I  
 fM , I 

     (6.2.7)
fint, D 
  
fext , D   0   f M , D 

243
Analysis Manual
The constraint forces on the dependent degrees of freedom, f M , D , can be

calculated accordingly.
fM , D  fint, D  fext , D (6.2.8)
Also, the constraint forces on the independent degrees of freedom, f M , I , satisfy

the following relationship using Equation (6.2.5):
fM , I  GT fM , D (6.2.9)
Once the multi-point constraint forces are thus obtained, the single-point constraint
forces, f S , I , can be finally calculated from Equation (6.2.7).
244
Load/Constraint
6.3 Thermal Loads/Boundary Conditions
Thermal loads and boundary conditions for heat transfer analysis used in midas
NFX include temperature boundary conditions assigned to nodes, heat flux applied
to lines and surfaces, convection and radiation, which are summarized in Table
6.3.1. Figure 6.3.1 shows the load conditions and boundary conditions that can be
considered for heat transfer analysis in midas NFX.
Table 6.3.1 Thermal loads/boundary conditions that can be used in midas NFX
Prescribed nodal temperature Nodes
Heat generation 1-, 2-, 3-dimensional elements
Heat flux Nodes, 1-, 2-, 3-dimensional elements
Convection Nodes, 1-, 2-, 3-dimensional elements
Radiation Nodes, 1-, 2-, 3-dimensional elements
Cavity radiation 2-, 3-dimensional elements
Flux
Convection
Cavity radiation
Heat generated internally
Radiation
Insulated
Prescribed temperature
Figure 6.3.1 Conceptual diagram of thermal loads/boundary conditions
245
Analysis Manual
• Prescribed nodal temperatures

Prescribed nodal temperatures are used to assign uniform known temperatures to
a model. In the same manner as forced displacements of structural analysis, the
temperature degrees of freedom of nodes to which temperature conditions are
assigned are eliminated from the global degrees of freedom and influence the load
vectors.
• Heat generation
Heat generation is used to simulate the quantity of heat generated inside a solid.
By inputting the rate of heat generation per unit volume, r , the effects of heat
generation inside elements can be obtained.
• Heat flux
Heat flux expresses power per unit area or energy per unit time and unit area. In
midas NFX, heat flux can be applied to nodes, faces and edges of elements.
When heat flux is assigned to a node, the area is determined through an additional
area factor value. When heat flux is applied to an edge, the area is determined by
using the thickness information of the edge or the area factor provided by the user.
• Convection
In midas NFX, natural convection conditions caused by the difference between the
ambient temperature and the surface temperature can be applied to nodes,
boundary faces and edges of elements. The amount of heat exchange or heat
flux caused by convection is quite empirical. midas NFX provides two different
forms of heat flux relationship equations by convection.
q  h(T * )(T  TA )c 1

(6.3.1)
q  h(T * )(T c  TAc )
T : Surface temperature
TA : Ambient temperature
c : Exponent
h : Surface convection coefficient
246
Load/Constraint
The surface convection coefficient can be expressed as a function of the surface

temperature or ambient temperature. When convection is assigned to a node, the
area is determined through an additional area factor. When convection is
assigned to an edge, the area is determined by using the thickness information of
the edge or the area factor provided by the user.
• Radiation
Heat exchange by radiation takes place when a difference between the surface and
ambient temperatures exists. midas NFX can apply heat exchange conditions
caused by radiation. The heat flux from radiation is expressed as,
q   F ( T 4  TA4 ) (6.3.2)
 : Stephan-Boltzmann constant
 : Emissivity
 : Absorptivity
F : Radiation view factor
It must be cautioned that the equation of the radiation heat transfer relationship is
expressed relative to absolute temperatures.
• Cavity Radiation
Since heat exchanges occur in interaction between individual surfaces in radiation
analysis of the interior of a cavity consisted of a number of surfaces, it takes a form
different from the radiation heat exchange with the ambient temperature, which has
been introduced previously. The heat flux per unit area, which is transferred to the
ith face by cavity radiation, is expressed as follows:
 i
qi 
Ai j
  j j FijCkj 1 Tj 4  Ti 4 
(6.3.3)
1
Cij  [( ij  (1   i ) Fij )] (no summation)
Ai
247
Analysis Manual
i : Emissivity of the i
th
face
Fij : Shape factor of the i th face with respect to the j th
face
th
Ai : Area of the i face
 ij : Kronecker delta
dAi ni
i
Rij nj
j
dAj
Figure 6.3.2 Geometric shape and radiation shape factor
A radiation shape factor shows the level of radiation heat exchange generated
between two faces and is geometrically defined as follows:
1 cos i cos  j
Fij 
Ai 
Ai Aj  Rij 2
dAi dAj (6.3.4)
The above integration is valid only if a visibility relationship between two points of
both faces is maintained. Parts that do not maintain such relationships are
excluded from the integration. A radiation blockage corresponds to such a case
(Figure 6.3.3) in which no radiation heat reaches certain faces due to the blockage
by a third object. In case of an enclosed cavity, the sum of the radiation shape
factors of all other faces relative to the given ith face produces 1, which is a
248
Load/Constraint
measure of accuracy in calculations for radiation shape factors.
F
j
ij 1 (6.3.5)
In case of an open cavity, the sum from the above equation becomes smaller than
1 in which case radiation heat is transferred to the ambient air. midas NFX
provides a function to automatically calculate radiation shape factors for specific
three-dimensional shapes.
G Cavity
Blockage
Blocked from G
Figure 6.3.3 Example of blocked radiation heat trasnfer
249
Analysis Manual
6.4 Singularity Errors
When a singularity error of a stiffness matrix occurs, no unique solution exists, and
it suggests that errors exist in the finite element model. Singularity errors can be
classified into single point singularity errors, which can be determined at a single
node, and mechanism form of singularity errors, which can be detected by
evaluating the global stiffness.
• Single point singularity errors

Single point singularity errors occur when the characteristics of elements are not
accurately understood, and incompatible connectivity is used. For example,
singularities will occur when elements such as rod and membrane elements, which
do not retain stiffness in specific directions, are used. Or defining the stiffness of a
spring element only in specific directions will result in singularities. In such cases,
singularity errors can be eliminated by properly using single point constraints in
order to obtain a normal solution. Single-point singularity errors can be
determined at the node level without having to decompose the stiffness matrix.
Such singularities can be thus eliminated by using the automatic single point
constraint function described in Section 6.2.
• Mechanism singularity errors

Singularity errors in a mechanism form exhibit in relation to two or more nodes.
Especially such errors often occur if constraint conditions are improperly assigned.
u1 u2
Figure 6.4.1 Elastic link without any constraint conditions
250
Load/Constraint
In the case of the system in the above figure, the equilibrium equation can be
k k   u1   p1 
 k   
k  u2   p2 
(6.4.1)

This stiffness matrix does not have an inverse matrix and will result in numerous
solutions if P1   P2 . The eigenvalues of this stiffness matrix include 0, and the
corresponding eigenvector will lead to a structure generating displacements without
deformation energy. For a general structure, the stiffness matrix retains six
eigenvalues of 0 corresponding to a rigid-body motion when no constraint
conditions are assigned at all, which will result in inability to calculate a unique
solution due to the singularity errors.
The presence of mechanism singularity errors can be found out during the
decomposition of a stiffness matrix in which values close to 0 in the diagonal terms
are considered singularity errors. If a singularity error is identified during the
process of decomposition, the program execution is terminated, or the calculations
can still proceed by adding small stiffness to the diagonal terms. The means to
add stiffness to the diagonal terms produces the same results as adding spring
elements to a finite element model.
251
Analysis Manual
6.5 Nonlinearity of Loads
Nonlinearity of a system often stems from the nonlinearity of materials (plasticity,

hyperelasticity, etc.) or geometric nonlinearity. In addition, nonlinear analysis of a
system is required at times based on nonlinearity of loads. For example, the
directions of loads change with the displacements of the structure, or even worse,
the magnitudes of loads depend on the behaviors of the structure. When carrying
out a geometric nonlinear analysis, midas NFX enables the user to reflect the
follower load effects in which the load directions change with the displacements of
the structure.
• Follower loads
If the direction of a nodal force is determined by the relative locations of two distinct
points, its magnitude and direction are as follows:
( xk   x j  )
f   Fn  F (6.5.1)
xk  x j
j, k : Nodes which determine the load direction
: Load component ( x, y, z )

If f  is differentiated with respect to x j , the following can be written:
f   ( x  x )( x  x )  
 F 
k j k j
 (6.5.2)
x j  xk  x j
3
xk  x j 
 
The derivative with respect to xk can be also calculated by the same process,
and consequently, an asymmetric stiffness matrix is constituted. Thus, the
stiffness due to the nonlinearity of loads is called load stiffness, and it is common to
produce matrices of very significant asymmetry.
252
Load/Constraint
A pressure load acting perpendicular to a specific face changes its direction with
the change in the face direction. Figure 6.5.1 show the effects of large
displacements between the case of forces acting perpendicular to a face and the
case of forces acting in a fixed direction.
Normal pressure Pressure in specified direction
Figure 6.5.1 Change in direction of a pressure load under large deformation
The process of calculating load stiffness caused by a pressure load acting

perpendicular to element faces is shown below.1 The nodal load at the I th
node of
an element is integrated using the pressure, p( , )  p N ( , ) . J J
x x I
f I   p( , )  N ( , )d d (6.5.3)
 
The geometric stiffness, K IJ , can be calculated by differentiating the above

equation with respect to node locations.
f I  f I f I f I 
K IJ    (6.5.4)
x J  x1J x2J x3J 
Each row that constitutes the stiffness can be obtained from the differential of x ,
which changes depending on the displacements in Equation (6.5.3).
1
Hibbit, H.D., “Some follower forces and load stiffness”, International Journal for Numerical Methods in Engineering, Vol. 4, 19
79
253
Analysis Manual
f I   x  x x   x   I
x Jj   x Jj      x Jj   
 p       N d d (6.5.5)
In addition, loads, gravity and rotational inertia forces of bar elements retain the
effects of follower loads. Also, whether or not to consider the effects of changing
each load following the displacements can be selected. Table 6.5.1 summarizes
the types of loads to which follower load effects can be applied in midas NFX.
Table 6.5.1 Loads that can consider the effects of follower loads in midas NFX
Types Range of Applications/Methods
Use load stiffness in case the direction is

Nodal loads
defined by two nodes.
Use load stiffness in case loads are acting

Pressure loads
perpendicular to the faces of elements.
Use load stiffness in case the direction is

Bar element loads
defined in reference to ECS.
Gravity Do not use load stiffness.
Rotational inertia forces Use load stiffness
254

Midas NFX Analysis Manual

Uploaded by

Copyright:

Available Formats

Midas NFX Analysis Manual

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Midas NFX Analysis Manual

Uploaded by

Copyright:

Available Formats

What types of loads can MIDAS NFX analyze with follower load effects?

What types of loads can MIDAS NFX analyze with follower load effects?

What are some examples of how load directions can change with structural displacement?

What are some examples of how load directions can change with structural displacement?

Analysis Manual

Modeling, Integrated Design & Analysis Software

MIDAS Information Technology Co., Ltd.

AutoCAD is a registered trademark of Autodesk, Inc.

Table 1.1 Analysis Functions of midas NFX

Linear static structural analysis

such analyses as prestressed analysis or thermal stress analysis.

► Define Analysis Model (nodes, elements, mesh, loads, boundary conditions)

1.1 Unit System

In order to define an analysis model, relavant information such as the sizes of

Table 1.1.1 Units used in English/SI Systems

Physical Quantity English SI

1.2 File System

Table 1.2.1 Main Files of midas NFX

File Name Format Description

Table 1.2.2 Files generated during Analysis

File Name Time of Generation/Description

InputName.DASM#.bin Generated in all analyses, data related to finite

Table 1.3.1 Matrix Notation

4th order tensor C

u  u1e1  u 2e2  u3e3 (1.3.1)

In the component notation, the subscript i represents a covariant base vector or a

A  Aijeie j , C  C ijkl eie jek el (1.3.3)

2.1 Nodal Points and Degrees of Freedom

For example, 3 displacement and 3 rotational degrees of freedom are assigned to

Figure 2.1.1 Displacement and rotational degrees of freedom in a

In general, each degree of freedom is expressed using the following symbols:

2.2 Coordinate System

midas NFX uses the following types of coordinate systems:

Figure 2.2.1 Rectangular and cylindrical coordinate systems

specify the directions of an anisotropic material. A specific coordinate system is

Table 2.2.1 Types of coordinate systems used in midas NFX

Types of Coordinate Systems Description

A single standard, rectangular coordinate

NDCS: Nodal Displacement Rectangular/cylindrical coordinate systems,

A rectangular coordinate system, which is

Rectangular/cylindrical coordinate systems,

ERCS: Element Result Coordinate Rectangular/cylindrical coordinate systems for

EFCS: Element Formulation A coordinate system used for finite element

Figure 2.2.2 Various coordinate systems used in midas NFX

2.3 Representation of Finite Rotation

Geometric nonlinear analysis, which includes finite rotation, requires a variable

Figure 2.3.1 Direction and magnitude of a rotation vector

In finite rotation, it must be cautioned that a compound rotation, which consists of

final rotational value, θ* , retains the property of θ*  θ  θ . Also, the

q  (q0 , q)  (cos( θ / 2), sin( θ / 2)e) (2.3.1)

The product of two quaternions can be calculated as follows:

q*  q q  (q0q0  q  q, q0q  q0q  q  q) (2.3.2)

Initial configuration Final configuration

Figure 2.3.2 Examples of compound rotations, which invalidate the

• Rigid Body/Interpolation Element

Table 3.1 Classification based on the characteristics of elements

Element Type Characteristics

3.1 Finite Element Formulation

 Gext   (T  u)σ   εT (Dε  σ)   σT (u  ε)d  (3.1.1)

 Gext : Virtual work due to an external force