Buku Referensi Anas 4 - Matrix Stiffness Method

Structural Analysis IV Chapter 4 – Matrix Stiffness Method
Chapter 4 - Matrix Stiffness Method
4.1 Introduction ......................................................................................................... 3

4.1.1 Background .................................................................................................... 3
4.1.2 Basic Concepts............................................................................................... 4
4.1.3 Computer Programs to Support Learning...................................................... 6
4.2 Basic Approach .................................................................................................. 10
4.2.1 Individual Element ...................................................................................... 10
4.2.2 Assemblies of Elements............................................................................... 12
4.2.3 Example 1 .................................................................................................... 14
4.2.4 General Methodology .................................................................................. 20
4.2.5 Member contribution to global stiffness matrix .......................................... 22
4.2.6 Interpretation of Stiffness Matrix ................................................................ 27
4.2.7 Restricting a Matrix – Imposing Restraints ................................................. 29
4.3 Plane Trusses ..................................................................................................... 32
4.3.1 Introduction.................................................................................................. 32
4.3.2 Truss Element Stiffness Matrix ................................................................... 35
4.3.3 Element Forces ............................................................................................ 40
4.3.4 Example 2: Basic Truss ............................................................................... 43
4.3.5 Example 3: Adding Members ...................................................................... 52
4.3.6 Example 4: Using Symmetry....................................................................... 56
4.3.7 Self-Strained Structures ............................................................................... 59
4.3.8 Example 5 – Truss with Differential Temperature ...................................... 63
4.3.9 Example 6 – Truss with Loads & Self Strains ............................................ 69
1
4.3.10 Problems ................................................................................................... 74

4.4 Beams .................................................................................................................. 77
4.4.1 Beam Element Stiffness Matrix................................................................... 77
4.4.2 Beam Element Loading ............................................................................... 82
4.4.3 Example 7 – Simple Two-Span Beam......................................................... 84
4.4.4 Example 8 – Non-Prismatic Beam .............................................................. 88
4.4.5 Problems ...................................................................................................... 92
4.5 Plane Frames ...................................................................................................... 95
4.5.1 Plane Frame Element Stiffness Matrix ........................................................ 95
4.5.2 Example 9 – Simple Plane Frame ............................................................. 104
4.5.3 Example 10 –Plane Frame Using Symmetry............................................. 109
4.5.4 Problems .................................................................................................... 115
4.6 Appendix .......................................................................................................... 120
4.6.1 Plane Truss Element Stiffness Matrix in Global Coordinates................... 120
4.6.2 Coordinate Transformations ...................................................................... 129
4.6.3 Past Exam Questions ................................................................................. 137
4.7 References ........................................................................................................ 148
Rev. 1
2
4.1 Introduction
4.1.1 Background
The matrix stiffness method is the basis of almost all commercial structural analysis
programs. It is a specific case of the more general finite element method, and was in
part responsible for the development of the finite element method. An understanding
of the underlying theory, limitations and means of application of the method is
therefore essential so that the user of analysis software is not just operating a ‘black
box’. Such users must be able to understand any errors in the modelling of structures
which usually come as obtuse warnings such as ‘zero pivot’ or ‘determinant zero:
structure unstable: aborting’. Understanding the basics presented herein should
hopefully lead to more fruitful use of the available software.
3
4.1.2 Basic Concepts
Node
The more general name for a connection between adjacent members is termed a node.
For trusses and frames the terms joint and node are interchangeable. For more
complex structures (e.g. plates), they are not.
Element
For trusses and frames element means the same as member. For more complex
structures this is not the case.
Degree of Freedom
The number of possible directions that displacements or forces at a node can exist in
is termed a degree of freedom (dof). Some examples are:
• Plane truss: has 2 degrees of freedom at each node: translation/forces in the x and y
directions.
• Beams: have 2 degrees of freedom per node: vertical displacement/forces and
rotation/moment.
• Plane Frame: has 3 degrees of freedom at each node: the translations/forces similar
to a plane truss and in addition, the rotation or moment at the joint.
• Space Truss: a truss in three dimensions has 3 degrees of freedom: translation or
forces along each axis in space.
• Space Frame: has 6 degrees of freedom at each node: translation/forces along each
axis, and rotation/moments about each axis.
4
Thus a plane truss with 10 joints has 20 degrees of freedom. A plane frame with two
members will have three joints (one common to both members) and thus 9 degrees of
freedom in total.
Local and Global

Forces, displacements and stiffness matrices are often derived and defined for an axis
system local to the member. However there will exist an overall, or global, axis
system for the structure as a whole. We must therefore transform forces,
displacements etc from the local coordinate system into the global coordinate system.
5
4.1.3 Computer Programs to Support Learning
Matlab Truss Analysis Program
Description
To support the ideas developed here we will introduce some Matlab scripts at each
point to demonstrate how the theory described can be implemented for computer
calculation. This collection of scripts will build into a program that can analyse pin-
jointed trusses. The scripts will only demonstrate the calculations process, and do not
have any graphical user interface facilities. This keeps the calculation process
unencumbered by extra code. (In fact probably 90+% of code in commercial
programs is for the graphical user interface and not for the actual calculations
process.) Of course, this is not to say that graphical displays of results are
unimportant; gross mistakes in data entry can sometimes only be found with careful
examination of the graphical display of the input data.
The scripts that are developed in these notes are written to explain the underlying
concepts, and not to illustrate best programming practice. The code could actually be
a lot more efficient computationally, but this would be at cost to the clarity of
calculation. In fact, a full finite element analysis program can be implemented in
under 50 lines (Alberty et al, 1999)!
It is necessary to use a scripting language like Matlab, rather than a spreadsheet

program (like MS Excel) since the number of members and member connectivity can
change from structure to structure.
The program will be able to analyse plane pin-jointed-trusses subject to nodal loads
only. It will not deal with member prestress, support stiffness or lack of fits: it is quite
rudimentary on purpose.
6
Use
To use the program, download it from the course website (www.colincaprani.com).
Extract the files to a folder and change the current Matlab directory to that folder.
After preparing the data (as will be explained later), execute the following statement
at the command line:
>> [D F R] = AnalyzeTruss(nData,eData)
This assumes that the nodal data is stored in the matrix nData, and the element data
matrix is stored in eData – these names are arbitrary. Entering the required data into
Matlab will also be explained later.
7
TrussMaster
To provide a bridge between the obvious workings of the Matlab program and the
more ‘black box’ nature of commercial programs, TrussMaster has been developed.
This program can analyse plane trusses of any size. It has a front end that illustrates a
fairly rudimentary commercial program interface, coupled with a back end (and some
dialogs) that expose the calculations the program carries out. In this manner it is easy
to link the hand calculations of the examples with the computer output, strengthening
the link between theory and practice of the method.
TrussMaster is available on the college computers. The help file can be downloaded
from the course website (www.colincaprani.com). It will be used in some labs.
8
LinPro
LinPro is very useful as a study aid for this topic: for example, right click on a
member and select “Stiffness Matrix” to see the stiffness matrix for any member. The
latest version (2.7.3) has a very useful “Study Mode”, which exposes the structure
and member stiffness matrices to the user. A user familiar with the underlying theory
can then use the program for more advanced purposes, such as spring supports, for
example.
You can download LinPro from www.line.co.ba.
9
4.2 Basic Approach
4.2.1 Individual Element

We consider here the most basic form of stiffness analysis. We represent a structural
member by a spring which has a node (or connection) at each end. We also consider
that it can only move in the x-direction. Thus it only has 1 DOF per node. At each of
its nodes, it can have a force and a displacement (again both in the x-direction):
Notice that we have drawn the force and displacement vector arrows in the positive x-
direction. Matrix analysis requires us to be very strict in our sign conventions.
Using the basic relationship that force is equal to stiffness times displacement, we can
determine the force at node 1 as:
F1 = k ( net displacement at 1)
Thus:
F1 = k ( u1 − u2 ) = ku1 − ku2 (4.2.1)
Similarly for node 2:
10
k ( u2 − u1 ) =
F2 = −ku1 + ku2 (4.2.2)
We can write equations (4.2.1) and (4.2.2) in matrix form to get the element stiffness
matrix for a 1-DOF axial element:
 F1   k −k   u1 
 =  
k  u2 
(4.2.3)
 F2   − k
And using matrix notation, we write:
{F } = [k ]{u }
e e
(4.2.4)
Here:
• {F } is the element force vector;
e
• [k ] is the element stiffness matrix;
• {u } is the element displacement vector.

e
It should be clear that the element stiffness matrix is of crucial importance – it links
nodal forces to nodal displacements; it encapsulates how the element behaves under
load.
The derivation of the element stiffness matrix for different types of elements is
probably the most awkward part of the matrix stiffness method. However, this does
not pose as a major disadvantage since we only have a few types of elements to
derive, and once derived they are readily available for use in any problem.
11
4.2.2 Assemblies of Elements

Real structures are made up of assemblies of elements, thus we must determine how
to connect the stiffness matrices of individual elements to form an overall (or global)
stiffness matrix for the structure.
Consider the following simple structure:
Note that the individual elements have different stiffnesses, k1 and k2 . Thus we can
write the force displacement relationships for both elements as:
 F1   k1 −k1   u1 
 =  
k1  u2 
(4.2.5)
 F2   − k1
 F2   k2 −k2  u2 
 =  
k2  u3 
(4.2.6)
 F3   − k2
We can expand these equations so that they encompass all the nodes in the structure:
 F1   k1 −k1 0   u1 
   −k  
 F2  = k1 0  u2  (4.2.7)
 1 
F  0 0  u3 
 3  0
 F1  0 0 0   u1 
  0 k  
=
 F2  −k2  u2  (4.2.8)
 2 
F  k2  u3 
 3 0 −k2
12
We can add equations (4.2.7) and (4.2.8) to determine the total of both the forces and
displacements at each node in the structure:
 F1   k1 −k1 0   u1 
    
 F2  =−k k1 + k2 −k2  u2  (4.2.9)
 1 
F   0 −k2 k2  u3 
 3 
As can be seen from this equation, by adding, we have the total stiffness at each node,
with contributions as appropriate by each member. In particular node 2, where the
members meet, has total stiffness k1 + k2 . We can re-write this equation as:
{F} = [K ]{u} (4.2.10)
In which:
• {F} is the force vector for the structure;
• [ K ] is the global stiffness matrix for the structure;
• {u} is the displacement vector for the structure.
13
4.2.3 Example 1
Problem
The following axially-loaded structure has loads applied as shown:
The individual member properties are:
Member Length (m) Area (mm2) Material, E (kN/mm2)

1 0.28 400 70
2 0.1 200 100
3 0.1 70 200
Find the displacements of the connections and the forces in each member.
14
Solution
Our first step is to model the structure with elements and nodes, as shown:
Calculate the spring stiffnesses for each member:
 EA  70 ⋅ 400
=
k1  =  = 100 × 103 kN/m (4.2.11)
 L 1 0.28
 EA  100 ⋅ 200
=
k2  =  = 200 × 103 kN/m (4.2.12)
 L 2 0.1
 EA  200 ⋅ 70
=
k3  =  = 140 × 103 kN/m (4.2.13)
 L 3 0.1
Next we calculate the individual element stiffness matrices:
 F1  3 
100 −100   u1 
  = 10    (4.2.14)
 F2   −100 100  u2 
 F2  3 
200 −200  u2 
  = 10    (4.2.15)
 F3   −200 200  u3 
15
 F3  3 
140 −140  u3 
  = 10    (4.2.16)
 F4   −140 140  u4 
We expand and add the element stiffness matrices to get:
 F1   100 −100 0 0   u1 
   −100 0  u2 
 F2  3  (100 + 200 ) −200
 
  = 10 (4.2.17)
 F3   0 −200 ( 200 + 140 ) −140  u3 
 F4   
 0 0 −140 140  u4 
Notice how each member contributes to the global stiffness matrix:
Node 1 Node 2 Node 3 Node 4
Node 1 0 0
Node 2 0
Node 3 0
Node 4 0 0
Notice also that where the member stiffness matrices overlap in the global stiffness
matrix that the components (or entries) are added. Also notice that zeros are entered
where there is no connection between nodes, e.g. node 1 to node 3.
16
We cannot yet solve equation (4.2.17) as we have not introduced the restraints of the
structure: the supports at nodes 1 and 4. We must modify equation (4.2.17) in such a
way that we will obtain the known results for the displacements at nodes 1 and 4.
Thus:
0 1 0 0 0   u1 
F 
 2
0
3  (100 + 200 ) −200 0  u2 
  = 10   (4.2.18)
 F3  0 −200 ( 200 + 140 ) 0   u3 
 0   
0 0 0 1  u4 
What we have done here is to ‘restrict’ the matrix: we have introduced a 1 on the
diagonal of the node number, and set all other entries on the corresponding row and
column to zero. We have entered the known displacement as the corresponding entry
in force vector (zero). Thus when we now solve we will obtain u=
1
u=
4
0.
For the remaining two equations, we have:
 F2  3 
300 −200  u2 
  = 10    (4.2.19)
 F3   −200 340  u3 
And so:
u2  1 1 340 200  −50  1  3 

 = ⋅ =   × 10 m
−3
  
u3  10 ( 300 )( 340 ) − ( −200 )( −200 )  200 300  100  62 20 
3
(4.2.20)
0.048
=  mm
 0.322 
To find the forces in the bars, we can now use the member stiffness matrices, since
we know the end displacements:
17
Member 1
 F1  3 
100 −100   0  −4.8
=  10  =    × 10 −3
  (4.2.21)
 F2   −100 100  0.048  4.8 
Thus Member 1 has a tension of 4.8 kN, since the directions of the member forces are
interpreted by our sign convention:
Also note that it is in equilibrium (as we might expect).
Member 2
 F2  3 
200 −200  0.048 −54.8
=  10  =    × 10
−3
  (4.2.22)
 F3   −200 200  0.322   54.8 
Member 2 thus has tension of 54.8 kN.
Member 3
 F3  3 
140 −140  0.322   45.08 
=  10  =    × 10 −3
  (4.2.23)
 F4   −140 140   0  −45.08
Thus Member 3 has a compression of 45.08 kN applied to it.
18
Problem
Find the displacements of the connections and the forces in each member for the
following structure:
Ans. 0.22 mm, 2.11 mm
19
4.2.4 General Methodology
Steps
The general steps in Matrix Stiffness Method are:
1. Calculate the member stiffness matrices
2. Assemble the global stiffness matrix
3. Restrict the global stiffness matrix and force vector
4. Solve for the unknown displacements
5. Determine member forces from the known displacements and member stiffness
matrices
6. Determine the reactions knowing member end forces.
20
Matlab Program - Implementation

These steps are implemented in the Matlab Program as follows:
function [D F R] = AnalyzeTruss(nData,eData)
% This function analyzes the truss defined by nData and eData:
% nData = [x, y, xLoad, yLoad, xRestraint, yRestraint]
% eData = [iNode, jNode, E, A];
kg = AssembleTrussK(nData, eData); % Assemble global stiffness matrix

fv = AssembleForceVector(nData); % And the force vector
[kgr fv] = Restrict(kg, fv, nData); % Impose restraints
D = fv/kgr; % Solve for displacements
F = ElementForces(nData,eData,D); % Get the element forces
R = D*kg; % Get the reactions
The output from the function AnalyzeTruss is:

• D: vector of nodal deflections;
• F: vector of element forces;
• R: vector of nodal forces (indicating the reactions and applied loads).
The input data required (nData and eData) will be explained later.
21
4.2.5 Member contribution to global stiffness matrix

Consider a member, ij, which links node i to node j. Its member stiffness matrix will
be:
Node i Node j
Node i k11ij k12ij
Node j k21ij k22ij
Its entries must then contribute to the corresponding entries in the global stiffness
matrix:
… Node i … Node j …
… … … … … …
Node i … k11ij … k12ij …
… … … … … …
Node j … k21ij … k22ij …
… … … … … …
If we now consider another member, jl, which links node j to node l. Its member
stiffness matrix will be:
22
Node j Node l
Node j k11jl k12jl
Node l k21jl k22jl
And now the global stiffness matrix becomes:
… Node i … Node j … Node l …
… … … … … … … …
Node i … k11ij … k12ij … … …
… … … … … … … …
k22ij +
Node j … k21ij … … k12jl …
k11jl
… … … … … … … …
Node l … … k21lj … k22jl …
… … … … … … … …
In the above, the identifiers k11 etc are sub-matrices of dimension:
ndof × ndof
where ndof refers to the number of degrees of freedom that each node has.
23
Matlab Program – Element Contribution

Considering trusses, we have 2 degrees of freedom (DOFs) per node, the x direction
and the y direction. Thus, for a truss with nn number of nodes, there are 2nn DOFs in
total. The x-DOF for any node i is thus located at 2i-1 and the y-DOF at 2i.
Consider a truss member connecting nodes i and j. To add the 4×4 truss element
stiffness matrix into the truss global stiffness matrix, we see that each row adds into
the following matrix columns:
2i-1 2i 2j-1 2j
The rows in the global stiffness matrix corresponding to the rows of the element
stiffness matrix are:
1. Row 1: Adds to row 2i-1 of the global stiffness matrix;
2. Row 2: Adds to row 2i;
3. Row 3: adds to row 2j-1;
4. Row 4: adds to row 2j.
Note of course that the column and row entries occur in the same order.
These rules are implemented for our Truss Analysis Program as follows:
function kg = AddElement(iEle,eData,ke,kg)
% This function adds member iEle stiffness matrix ke to the global
% stiffness matrix kg.
% What nodes does the element connect to?

iNode = eData(iEle,1);
jNode = eData(iEle,2);
% The DOFs in kg to enter the properties into

DOFs = [2*iNode-1 2*iNode 2*jNode-1 2*jNode];
% For each row of ke

for i = 1:4
% Add the row to the correct entries in kg
kg(DOFs(i),DOFs) = kg(DOFs(i),DOFs) + ke(i,:);
end
24
Matlab Program – Global Stiffness Matrix Assembly

The function that assembles the truss global stiffness matrix for the truss is as
follows:
function kg = AssembleTrussK(nData, eData)

% This function assembles the global stiffness matrix for a truss from the
% joint and member data matrices
% How many nodes and elements are there?

[ne ~] = size(eData);
[nn ~] = size(nData);
% Set up a blank global stiffness matrix

kg = zeros(2*nn,2*nn);
% For each element

for i = 1:ne
E = eData(i,3); % Get its E and A
A = eData(i,4);
[L c s] = TrussElementGeom(i,nData,eData); % Geometric Properties
ke = TrussElementK(E,A,L,c,s); % Stiffness matrix
kg = AddElement(i,eData,ke,kg); % Enter it into kg
end
Note that we have not yet covered the calculation of the truss element stiffness
matrix. However, the point here is to see that each element stiffness matrix is
calculated and then added to the global stiffness matrix.
25
Matlab Program – Force Vector

Examine again the overall equation (4.2.10) to be solved:
{F} = [K ]{u}
We now have the global stiffness matrix, we aim to calculate the deflections thus we
need to have a force vector representing the applied nodal loads. Again remember
that each node as two DOFs (x- and y-loads). The code for the force vector is thus:
function f = AssembleForceVector(nData)
% This function assembles the force vector
% How may nodes are there?

% Set up a blank force vector

f = zeros(1,2*nn);
% For each node

for i = 1:nn
f(2*i - 1) = nData(i, 3); % x-load into x-DOF
f(2*i) = nData(i, 4); % y-load into y-DOF
end
26
4.2.6 Interpretation of Stiffness Matrix

It is useful to understand what each term in a stiffness matrix represents. If we
consider a simple example structure:
We saw that the global stiffness matrix for this is:
 K11 K12 K13   k1 −k1 0 

K
K= K 22 K 23  = −k k1 + k2 −k2 
 21   1 
 K 31 K 32 K 33   0 −k2 k2 
If we imagine that all nodes are fixed against displacement except for node 2, then we
have the following:
27
From our general equation:
 F1   K11 K12 K13  0   K12 

  =    
= F2  K 21 K 22 K 23  1   K 22  (4.2.24)
 
F   K 31 K 32 K 33  0  K 
 3  32 
Thus:
 F1   K12   −k1 
     
 F=
2  K 22=
 k1 + k2  (4.2.25)
F  K   −k 
 3  32   2 
These forces are illustrated in the above diagram, along with a free-body diagram of
node 2.
Thus we see that each column in a stiffness matrix represents the forces required to
maintain equilibrium when the column’s DOF has been given a unit displacement.
This provides a very useful way to derive member stiffness matrices.
28
4.2.7 Restricting a Matrix – Imposing Restraints

In Example 1 we solved the structure by applying the known supports into the global
stiffness matrix. We did this because otherwise the system is unsolvable; technically
the determinant of the stiffness matrix is zero. This mathematically represents the fact
that until we apply boundary conditions, the structure is floating in space.
To impose known displacements (i.e. supports) on the structure equations we modify

the global stiffness matrix and the force vector so that we get back the zero
displacement result we know.
Considering our two-element example again, if node 1 is supported, u1 = 0 . Consider

the system equation:
 F1   K11 K12 K13   u1 

    
 F2  =  K 21 K 22 K 23  u2  (4.2.26)

F  K K 33  u3 
 3   31 K 32
Therefore to obtain u1 = 0 from this, we change K and F as follows:
 0  1 0 0   u1 
    
 F2  = 0 K 22 K 23  u2  (4.2.27)

 F  0 K K 33  u3 
 3  32
Now when we solve for u1 we will get the answer we want: u1 = 0 . In fact, since we
now do not need this first equation, we could just consider the remaining equations:
 F2   K 22 K 23  u2 
 =  
K 33  u3 
(4.2.28)
 F3   K 32
29
And these are perfectly solvable.
Thus to summarize:
To impose a support condition at degree of freedom i:

1. Make the force vector element of DOF i zero;
2. Make the i column and row entries of the stiffness matrix all zero;
3. Make the diagonal entry ( i, i ) of the stiffness matrix 1.
30
Matlab Program – Imposing Restraints

To implement these rules for our Truss Analysis Program, we will first create of
vector which tells us whether or not a DOF is restrained. This vector will have a zero
if the DOF is not restrained, and a 1 if it is.
Once we have this vector of restraints, we can go through each DOF and modify the
force vector and global stiffness matrix as described before. The implementation of
this is as follows:
function [kg f] = Restrict(kg, f, nData)

% This function imposes the restraints on the global stiffness matrix and
% the force vector
% How may nodes are there?

% Store each restrained DOF in a vector

RestrainedDOFs = zeros(2*nn,1);
% For each node, store if there is a restraint

for i = 1:nn
% x-direction
if nData(i,5) ~= 0 % if there is a non-zero entry (i.e. supported)
RestrainedDOFs(2*i-1) = 1;
end
% y-direction
if nData(i,6) ~= 0 % if there is a support
RestrainedDOFs(2*i) = 1;
end
end
% for each DOF

for i = 1:2*nn
if RestrainedDOFs(i) == 1 % if it is restrained
f(i) = 0; % Ensure force zero at this DOF
kg(i,:) = 0; % make entire row zero
kg(:,i) = 0; % make entire column zero
kg(i,i) = 1; % put 1 on the diagonal
end
end
31
4.3 Plane Trusses
4.3.1 Introduction
Trusses are assemblies of members whose actions can be linked directly to that of the
simple spring studied already:
EA
k= (4.3.1)
L
There is one main difference, however: truss members may be oriented at any angle
in the xy coordinate system (Cartesian) plane:
Thus we must account for the coordinate transformations from the local member axis
system to the global axis system.
32
Matlab Program – Data Preparation

In the following sections we will put the final pieces of code together for our Truss
Analysis Program. At this point we must identify what information is required as
input to the program, and in what format it will be delivered.
The node data is stored in a matrix nData. Each node of the truss is represented by a
row of data. In the row, we put the following information in consecutive order in
columns:
1. x-coordinate;
2. y-coordinate;
3. x-load: 0 or the value of load;
4. y-load: 0 or the value of load;
5. x-restraint: 0 if unrestrained, any other number if restrained;
6. y-restraint: 0 if unrestrained, any other number if restrained.
The element data is stored in a matrix called eData. Each element has a row of data
and for each element the information stored in the columns in order is:
1. i-Node number: the node number at the start of the element;
2. j-Node number: the other node the element connects to;
3. E: the Modulus of Elasticity of the element material;
4. A: the element area;
We will prepare input data matrices in the above formats for some of the examples
that follow so that the concepts are clear. In doing so we keep the units consistent:
• Dimensions are in m;
• Forces in kN
• Elastic modulus is in kN/mm2;
• Area is mm2.
33
Matlab Program – Data Entry

To enter the required data, one way is:
1. Create a new variable in the workspace (click on New Variable);
2. Name it eData for example;
3. Double click on the new variable to open the Matlab Variable Editor;
4. Enter the necessary input data (can paste in from MS Excel, or type in);
5. Repeat for the nodal data.
34
4.3.2 Truss Element Stiffness Matrix

For many element types it is very difficult to express the element stiffness matrix in
global coordinates. However, this is not so for truss elements. Firstly we note that the
local axis system element stiffness matrix is given by equation (4.2.3):
 k −k   1 −1
=[k ] = k   (4.3.2)
 −k k   −1 1 
Next, introducing equation (4.3.1), we have:
EA  1 −1
[k ] =
L  −1 1 
(4.3.3)
However, this equation was written for a 1-dimensional element. Expanding this to a
two-dimensional axis system is straightforward since there are no y-axis values:
1 0 −1 0  ← xi
 0  ← yi
EA  0 0 0
[k ] =   (4.3.4)
L −1 0 1 0 ← x j
 
0 0 0 0 ← y j
Next, using the general element stiffness transformation equation (See the Appendix):
[k ] = [T] [k ][T]
T
(4.3.5)
And noting the transformation matrix for a plane truss element from the Appendix:
35
 cos α sin α 0 0 
T 0   − sin α cos α 0 0 
T= P  =  (4.3.6)
0 TP   0 0 cos α sin α 
 
 0 0 − sin α cos α 
We have:
−1
 cos α sin α 0 0  1 0 −1 0 
 0  0 0 0 0
EA  − sin α cos α 0
[k ]   ⋅
L  0 0 cos α sin α   −1 0 1 0
   
 0 0 − sin α cos α   0 0 0 0
(4.3.7)
 − sin α cos α 0 0 
 
 0 0 cos α sin α 
 
 0 0 − sin α cos α 
Carrying out the multiplication gives:
 cos 2 α cos α sin α − cos 2 α − cos α sin α 

 
EA  cos α sin α sin 2 α − cos α sin α − sin 2 α 
k= (4.3.8)
L  − cos 2 α − cos α sin α cos 2 α cos α sin α 
 
 − cos α sin α − sin 2 α cos α sin α sin 2 α 
If we examine the nodal sub-matrices and write c ≡ cos α , s ≡ sin α :
 c2 cs −c 2 −cs 
 
EA  cs s 2 −cs − s 2 
[k ] =  2 (4.3.9)
L −c −cs c 2 cs 
 
 −cs − s
2
cs s2 
36
Labelling the nodal sub-matrices as:
k11 k12 
[k ] =   (4.3.10)
k21 k22 
Then we see that the sub-matrices are of dimension 2 × 2 (No. DOF × No. DOF) and
are:
EA  c cs 
2
k11 =   (4.3.11)
L cs s 2 
And also note:
k11 = k22 = -k12 = -k21 (4.3.12)
Therefore, we need only evaluate a single nodal sub-matrix ( k11 ) in order to find the
total element stiffness matrix in global coordinates.
37
Matlab Program – Element Stiffness Matrix

Calculating the element stiffness matrix for our Truss Analysis Program is easy. The
only complexity is extracting the relevant data from the input node and element data
matrices. Rather than try determine the angle that the truss member is at (remember
we only have the nodal coordinates), we can calculate cos α and sin α directly (e.g.
adjacent/hypotenuse). Further, the element length can be found using Pythagoras,
given the nodal coordinates. These element properties are found in the script below:
function [L c s] = TrussElementGeom(iEle,nData,eData);
% This function returns the element length

% What are the coordinates of these nodes?

iNodeX = nData(iNode,1);
iNodeY = nData(iNode,2);
jNodeX = nData(jNode,1);
jNodeY = nData(jNode,2);
% Use Pythagoras to work out the member length

L = sqrt((jNodeX - iNodeX)^ 2 + (jNodeY - iNodeY)^ 2);
% Cos is adjacent over hyp, sin is opp over hyp

c = (jNodeX - iNodeX)/L;
s = (jNodeY - iNodeY)/L;
The E and A values for each element are directly found from the input data element
E = eData(i,3); % Get its E and A

A = eData(i,4);
Thus, with all the relevant data assembled, we can calculate the truss element
stiffness matrix. In the following Matlab function, note that we make use of the fact
that each nodal sub-matrix can be determined from the nodal sub-matrix k11 :
38
function k = TrussElementK(E,A,L,c,s)
% This function returns the stiffness matrix for a truss element
k11 = [ c^2 c*s;

c*s s^2];
k = (E*A/L) * [ k11 -k11;

-k11 k11];
39
4.3.3 Element Forces

The forces applied to a member’s ends are got from the element equation:
{F } = [k ]{u }
e e
(4.3.13)
Expanding this in terms of nodal equations we have:
 Fi  k11 k12   δi 
 =  δ  (4.3.14)
F
 j  k21 k22  j
Thus we know:
Fj = k21 ⋅ δi + k22 ⋅ δ j (4.3.15)
From which we could determine the member’s axial force. However, for truss
members, we can determine a simple expression to use if we consider the change in
length in terms of the member end displacements:
∆Lx = δ jx − δ ix (4.3.16)
∆Ly = δ jy − δ iy (4.3.17)
And using the coordinate transforms idea:
∆L = ∆Lx cos α + ∆Ly sin α (4.3.18)
Also we know that the member force is related to the member elongation by:
40
EA
=
F ⋅ ∆L (4.3.19)
L
Thus we have:
EA
=
F ⋅  ∆Lx cos α + ∆Ly sin α  (4.3.20)
L 
And introducing equations (4.3.16) and (4.3.17) gives:
EA δ jx − δ ix 
=
F ⋅ [ cos α sin α ]   (4.3.21)
L δ jy − δ iy 
A positive result from this means tension and negative compression.
41
Matlab Program – Element Force

Once the element nodal deflections are known, the element forces are found as
described above. Most of the programming effort is dedicated to extracting the nodal
deflections that are relevant for the particular member under consideration:
function F = TrussElementForce(nData, eData, d, iEle)

% This function returns the element force for iEle given the global
% displacement vector, d, and the node and element data matrices.

% Get the element properties

E = eData(iEle,3); % Get its E and A
A = eData(iEle,4);
[L c s] = TrussElementGeom(iEle,nData,eData); % Geometric Properties
dix = d(2*iNode-1); % x-displacement at node i

diy = d(2*iNode); % y-displacement at node i
djx = d(2*jNode-1); % x-displacement at node j
djy = d(2*jNode); % y-displacement at node j
F = (E*A/L) * (c*(djx-dix) + s*(djy-diy));
Note also that the way the program is written assumes that tension is positive and
compression is negative. We also want to return all of the element forces, so we use
the function just described to calculate all the truss elements’ forces:
function F = ElementForces(nData,eData,d)
% This function returns a vector of the element forces
% How many elements are there?

[ne ~] = size(eData);
% Set up a blank element force vector

F = zeros(ne,1);
% For each element

for i = 1:ne
% Get its force and enter into vector
F(i) = TrussElementForce(nData, eData, d, i);
end
42
4.3.4 Example 2: Basic Truss
Problem
Analyse the following truss using the stiffness matrix method.
Note that:
• E = 200 kN/mm 2 ;
• The reference area is A = 100mm 2 .
43
Solution
STEP 1: Determine the member stiffness matrices:
Member 12
The angle this member makes to the global axis system and the relevant values are:
1 1
c ≡ cos α= cos 45=

⇒ c 2=
2 2
1 1 1
s ≡ sin α= sin 45=

⇒ s=
2
⇒ cs=
2 2 2
Therefore:
 EA   c cs  200 ⋅ 100 2 0.5 0.5

2
=k1112 =   2 0.5 0.5

 L 12 cs s  10 2  
Thus:
0.5 0.5
k1112 = 103   (4.3.22)
0.5 0.5
Notice that the matrix is symmetrical as it should be.
44
Member 23
The angle this member makes to the global axis system and the relevant values are:
1 1
c ≡ cos α = cos315 = ⇒ c2 =
2 2
1 1 1
s ≡ sin α =
sin 315 =
− ⇒ s2 = ⇒ cs =
−
2 2 2
Therefore:
 EA   c cs  200 ⋅ 100 2  0.5 −0.5

2
=k1123 =   2  −0.5 0.5 

 L  23 cs s  10 2  
Thus:
 0.5 −0.5
k1123 = 103   (4.3.23)
 −0.5 0.5 
Again the matrix is symmetrical.
45
STEP 2: Assemble the global stiffness matrix
For 3 nodes, the unrestricted global stiffness matrix will look as follows:
 K 11 K 12 K 13  ← Node 1
=K K 21 K 22 K 23  ← Node 2 (4.3.24)
 
 K 31 K 32 K 33  ← Node 3
Note that each of the sub-matrices is a 2×2 matrix, e.g.:
 k11 xx k12 xy  ← Node 1 x

K 11 = 
k22 yy  ← Node 1 y
(4.3.25)
 k21 yx
The member stiffness nodal sub-matrices contribute to the global stiffness nodal sub-
matrices as follows:
 K 11 K 12 K 13  k1112 k1212 0 
=K = K 21 K 22 K 23  k21 k2212 + k1123 k1223  (4.3.26)
   12 
 K 31 K 32 K 33   0 k2123 k2223 
Expanding this out and filling in the relevant entries from equations (4.3.22) and
(4.3.23) whilst using equation (4.3.12) gives:
46
 0.5 0.5 −0.5 −0.5 0 0 

 0.5 0.5 −0.5 −0.5 0 0 
 
 −0.5 −0.5 1 0 −0.5 0.5 
K = 103   (4.3.27)
 −0.5 −0.5 0 1 0.5 −0.5
 0 0 −0.5 0.5 0.5 −0.5
 
 0 0 0.5 −0.5 −0.5 0.5 
STEP 3: Write the solution equation in full
{F} = [K ]{δ} (4.3.28)
Thus, keeping the nodal sub-matrices identifiable for clarity:
 R1 x   0.5 0.5 −0.5 −0.5 0 0   δ1 x 

 R   0.5 0.5 −0.5 −0.5  δ 
  0 0
1 y
   1y 
 0   −0.5 −0.5 1 0 −0.5 0.5  δ 2 x 
  = 10   
3
(4.3.29)
 −100   −0.5 −0.5 0 1 0.5 −0.5  δ 2 y 
 R3 x   0 0 −0.5 0.5 0.5 −0.5 δ 3 x 
    
 R3 y   0 0 0.5 −0.5 −0.5 0.5  δ 3 y 
In which we have noted:

• R1x is the reaction at node 1 in the x-direction (and similarly for the others);
• The force at node 2 is 0 in the x-direction and -100 kN (downwards) in the y-
direction.
47
STEP 4: Restrict the equation.

Now we impose the boundary conditions on the problem. We know:
• δ=
1x
δ=
1y
0 since node 1 is pinned;
• δ=
3x
δ=
3y
0 again, since node 3 is pinned.
Thus equation (4.3.29) becomes:
 0  1 0 0 0 0 0   δ1 x 
 0  0  
1 0 0 0 0  δ 1 y 
   
 0  0 0 1 0 0 0  δ 2 x 
  = 10   
3
(4.3.30)
−100  0 0 0 1 0 0  δ 2 y 
 0  0 0 0 0 1 0  δ 3 x 
    
 0  0 0 0 0 0 1  δ 3 y 
Since both DOFs are restricted for nodes 1 and 3, we can thus write the remaining
equations for node 2:
 0  3 
1 0  δ 2 x 
  = 10    (4.3.31)
−100  0 1  δ 2 y 
STEP 5: Solve the system

The y-direction is thus the only active equation:
103 δ 2 y
−100 = (4.3.32)
Thus:
δ2y =
−0.1 m =
100 mm ↓ (4.3.33)
48
STEP 6: Determine the member forces

For truss member’s we outlined a simple method encompassed in equation (4.3.21).
In applying this to Member 12 we note:
• δ=
1x
δ=
1y
0 since it is a support;
• δ 2 x = 0 by solution;
• δ 2 y = −0.1 again by solution.
Thus:
EA δ jx − δ ix 
=
F ⋅ [ cos α sin α ]  
L δ jy − δ iy 
 1 1  0 −0  100
F=
103    =
− =
−50 2 kN (4.3.34)
 2 2  −0.1 − 0  2
And so Member 12 is in compression, as may be expected. For Member 23 we

similarly have:
 1 1  0 −0  100
F=
103  −   =
− =
−50 2 kN (4.3.35)
 2 2  0 − ( −0.1)  2
And again Member 23 is in compression. Further, since the structure is symmetrical

and is symmetrically loaded, it makes sense that Member’s 12 and 23 have the same
force.
STEP 7: Determine the reactions

To determine the remaining unknown forces we can use the basic equation now that
all displacements are known:
49
 R1 x   0.5 0.5 −0.5 −0.5 0 0  0 

 R   0.5 0.5 −0.5 −0.5
 1y  0 0  0 
  
 0   −0.5 −0.5 1 0 −0.5 0.5   0 
  = 10    
3
(4.3.36)
 −100   −0.5 −0.5 0 1 0.5 −0.5  −0.1
 R3 x   0 0 −0.5 0.5 0.5 −0.5  0 
    
 R3 y   0 0 0.5 −0.5 −0.5 0.5  0 
Thus we have:
 0 
[ −0.5 −0.5]   =
R1 x = +50kN (4.3.37)
 −0.1
 0 
[ −0.5 −0.5]   =
R1 y = +50kN (4.3.38)
−0.1
 0 
[ −0.5 0.5]   =
R3 x = −50kN (4.3.39)
−0.1
 0 
[0.5 −0.5]   =
R3 y = +50kN (4.3.40)
 −0.1
Again note that the sign indicates the direction along the global coordinate system.
We can now plot the full solution:
50
Matlab Program – First Use

All necessary functions have been explained. The main function is given on page 21.
This also gives the single line of code that finds the reactions. The input data for the
example truss just given is:
Node Data Element Data
x y Fx Fy Rx Ry Node i Node j E A
0 0 0 0 1 1 1 2 200 70.71
10 10 0 -100 0 0 2 3 200 70.71
20 0 0 0 1 1
And the results from the program are:
Node DOF D R Element F

1 x 0 50 1 -70.71
y 0 50 2 -70.71
2 x 0 0
y -0.100 -100
3 x 0 -50
y 0 50
These results, of course, correspond to those found by hand.
The importance of the graphical display of the results should also be noted: there
could have been clear mistakes made in the preparation of the input data that would
not reveal themselves unless the physical interpretation of the results is appreciate by
drawing the deflected shape, the member forces, and the directions of the reactions.
51 Dr. C. Caprani
4.3.5 Example 3: Adding Members
Problem
Analyse the truss of Example 2 but with the following member 14 added:
Solution
With the addition of node 4 we now know that the nodal sub-matrices global stiffness
equation will be 4×4 with the fully expanded matrix being 16×16. Rather than
determine every entry in this, let’s restrict it now and only determine the values we
will actually use. Since nodes 1, 3 and 4 are pinned, all their DOFs are fully restricted
out. The restricted equation thus becomes:
{F } = [K22]{δ }
2 2
(4.3.41)
Next we must identify the contributions from each member:

• We already know the contributions of Members 12 and 23 from Example 2.
• The contribution of Member 24 is to nodes 2 and 4. Since node 4 is restricted, we
only have the contribution k1124 to K22 .
Thus K22 becomes:
52 Dr. C. Caprani
K22 = k2212 + k1123 + k1124 (4.3.42)
Next determine k1124 : this member makes an angle of 270° to the global axis system
giving:
c ≡ cos α = cos 270 = 0 ⇒ c2 = 0

s ≡ sin α =
sin 270 =
−1 ⇒ s 2 =
1 ⇒ cs =
0
Therefore:
 EA   c cs  200 ⋅ 100 0 0  0 0
2
k1124=    2
=  = 2 × 103  
 L  24 cs s  10 0 1  0 1 
Thus:
0 0 
k1124 = 103   (4.3.43)
0 2 
Hence the global restricted stiffness matrix becomes:
1 0  3 
0 0 3 
1 0
K22 = 103   + 10 0 2  = 10 0 3 (4.3.44)
0 1     
Writing the restricted equation, we have:
 0  3 
1 0  δ 2 x 
  = 10 0 3  δ  (4.3.45)
−100     2y 
53 Dr. C. Caprani
From which we find the only equation
103 ( 3δ 2 y )
−100 = (4.3.46)
Thus:
δ2y =
−0.033 m =
33.3 mm ↓ (4.3.47)
The member forces are:
 1 1  0 −0 
F12 = 103    = −23.6kN (4.3.48)
 2 2  −0.033 − 0 
 1 1  0−0 
F23 =
103  −  =−23.6kN (4.3.49)
 2 2  0 − ( −0.033) 
 0−0 
F24 = 103 [ 0 2]   = −66.6kN (4.3.50)
−0.033 − 0 
Thus we have the following solution:
54 Dr. C. Caprani
Matlab Program – Input/Output

The input data for this example is:
Node Data Element Data

x y Fx Fy Rx Ry Node i Node j E A
0 0 0 0 1 1 1 2 200 70.71
10 10 0 -100 0 0 2 3 200 70.71
20 0 0 0 1 1 2 4 200 100
10 0 0 0 1 1
The results are:
Node DOF D R Element F

1 x 0 16.66 1 -23.57
y 0 16.66 2 -23.57
2 x 0 0 3 66.66
y -0.033 -100
3 x 0 -16.66
y 0 16.66
4 x 0 0
y 0 66.66
55 Dr. C. Caprani
4.3.6 Example 4: Using Symmetry
Problem
Analyse the truss of Example 3 taking advantage of any symmetry:
Solution
Looking at the structure it is clear that by splitting the structure down the middle
along member 24 that we will have two equal halves:
Notice that we have changed the following:
56 Dr. C. Caprani
• The load is halved since it is now equally shared amongst two halves;
• Similarly the area of member 24 is halved.
We now analyse this new truss as usual. However, we can make use of some previous
results. For Member 12:
0.5 0.5
k1112 = 103   (4.3.51)
0.5 0.5
And for Member 24
1  3 0 0   3 
0 0
= =10  
2  0 2   0 1 
k1124 10 (4.3.52)
 
Since the area is halved from that of Example 3, its stiffnesses are halved.
In restricting we note that the only possible displacement is node 2 in the y-direction.
However, we will keep using the node 2 sub-matrices until the last moment:
0.5 0  3 
0 0 3 
0.5 0 
= 103 
K22  + 10 =
0 1  10  0 1.5 (4.3.53)
 0 0.5    
Thus:
 0  3 
0.5 0  δ 2 x 
  = 10  0 1.5 δ  (4.3.54)
−50     2y 
And now imposing the boundary condition δ 2 x = 0 :
57 Dr. C. Caprani
10 [1.5]{δ }
{−50} = 3
2y
(4.3.55)
From which we solve for the displacement:
10 [1.5]{δ }
{−50} = 3
2y
(4.3.56)
δ2y =
−0.033 m =
33.3 mm ↓
This (of course) is the same result we obtained in Example 3. For the member forces
we have:
 1 1  0 −0 
F12 = 103    = −23.6kN (4.3.57)
 2 2  −0.033 − 0 
 0−0 
103 [ 0 −1] 
F24 = =−33.3kN (4.3.58)
 0 − ( −0.033 ) 
Member 12 has the same force as per Example 3 as is expected.
It might appear that Member 24 has an erroneous force result. It must be remembered
that this is the force in the half-member (brought about since we are using symmetry).
Therefore the force in the full member is 2 × 33.3 =
66.6 kN as per Example 3.
58 Dr. C. Caprani
4.3.7 Self-Strained Structures
Introduction
A self-strained structure is one where strains are induced by sources other than
externally applied loads. The two main examples are temperature difference and lack
of fit of a member. For example consider the effect if member 13 in the following
structure was too long and had to be ‘squeezed’ into place:
It should be intuitively obvious that to ‘squeeze’ the member into place a

compressive force was required to shorten it to the required length:
59 Dr. C. Caprani
Once the member has been put in place, the source of the ‘squeezing’ is removed.
Since the member wants to spring back to its original length, it pushes on its joints:
In this way members 12 and 14 will now go into tension whilst member 13 will
remain in compression, but a smaller compression than when it was ‘squeezed’ into
place since joint 1 will deflect to the right some amount.
In a similar way to lack of fit, examined above, if member 13 had been subject to a
temperature increase it would try to elongate. However this elongation is restrained
by the other members inducing them into tension and member 13 into some
compression.
60 Dr. C. Caprani
Lack of Fit
We consider a member with original length of LO that is required to be of length
LReq'd . Thus a change in length of ∆L must be applied:
LReq'd= LO + ∆L (4.3.59)
Thus:
• ∆L is positive: the member is too short and must be lengthened to get into place;
• ∆L is negative, it is too long and must be shortened to get into place.
Thus we must apply a force to the member that will cause a change in length of ∆L .
From basic mechanics:
FL
∆L = O (4.3.60)
EA
Thus the force required is:
∆L
=
F EA ⋅ (4.3.61)
LO
From the above sign convention for ∆L :

• F is positive when the member must be put into tension to get it in place;
• F is negative when the member must be put into compression to get it in place.
Lastly, remember to apply the member force in opposite direction to the member’s
nodes.
61 Dr. C. Caprani
Temperature Change
We consider a member that is subject to a differential (i.e. different to the rest of the
structure) temperature change of ∆T degrees Celsius. Also we must know the
coefficient of linear thermal expansion, α , for the material. This is the change in
length, per unit length, per unit change in temperature:
 ∆L 
α ≡  
C (4.3.62)
 LO 
Thus the thermal strain induced in the member is:
ε T= α ∆T (4.3.63)
And so the change in length is:
∆L= α LO ∆T (4.3.64)
Also, since σ = Eε , we find the force in the member:
FT σ=
= T
A EAε T (4.3.65)
So finally, from equation (4.3.63), the force required to suppress the temperature
change is:
FT EAα ∆T
= (4.3.66)
Once again, apply this force in the opposite direction to the member’s nodes.
62 Dr. C. Caprani
4.3.8 Example 5 – Truss with Differential Temperature
Problem
Member 13 of the following truss is subject to a temperature change of +100 °C.
Calculate the deflections of node 1 and the final forces in the members.
Take: α = 2 × 10−5 C −1 ; EA= 2 × 104 kN ; the area of member 12 as 2A; the area of
member 13 as A; and, the area of member 14 as A√2.
63 Dr. C. Caprani
Solution
First we must recognize that there are two stages to the actions in the members:
• Stage I: all displacements are suppressed and only the temperature force in
member 13 is allowed for;
• Stage II: displacements are allowed and the actions of the temperature force in
member 13 upon the rest of the structure are analyzed for.
The final result is then the summation of these two stages:
The force induced in member 13 when displacements are suppressed is:
FT EAα ∆T
=
( 2 104 )( 2 × 10−5 ) ( +100 )
=× (4.3.67)
= 40kN
64 Dr. C. Caprani
Stage I
All displacements are suppressed. Thus:
=δ1x 0;=δ1 y 0 (4.3.68)
=F12I 0;=F13I F=
T
; F14I 0 (4.3.69)
Stage II
Displacements are allowed occur and thus we must analyse the truss. Using the
matrix stiffness method, and recognizing that only joint 1 can displace, we have:
{F } = [K11]{δ }
1 1
(4.3.70)
Also, since we cleverly chose the node numbers, the member contributions are just:
K11 = k1112 + k1113 + k1114 (4.3.71)
• Member 12:
1 1
c ≡ cos α =
cos120 =− ⇒ c2 =
2 4
3 3 3
s ≡ sin α =
sin120 = ⇒ s 2 = ⇒ cs =
−
2 4 4
 1 3
 − 
 EA   c cs  2 × 10 ( 2 )  4
2 4
4 
=k1112 =   2
 L  23 cs s  2.0  3 3 
− 
 4 4 
65 Dr. C. Caprani
• Member 13:
c ≡ cos α =
cos180 =−1 ⇒ c 2 =
1
s ≡ sin α= sin180= 0 ⇒ s 2= 0 ⇒ cs= 0
 EA   c cs  2 × 10 1 0 
2 4
=k1113 =   2
 L  23 cs s  1.0 0 0 
• Member 14:
1 1
c ≡ cos α =
cos 225 =
− ⇒ c2 =
2 2
−1 1 1
s ≡ sin α= sin 225= ⇒ s 2= ⇒ cs=
2 2 2
=
 EA   c cs  2 × 10
k1114 =
2 4
2 0.5 0.5 ( )
  2  
 L  23 cs s  1.0 2 0.5 0.5
Thus from equation (4.3.71) we have:
2 × 104  7 2 − 3
K11 =   (4.3.72)
4  2 − 3 5 
From the diagram for Stage II, we can see that the force applied to joint 1 is acting to
the right and so is positive. Thus:
66 Dr. C. Caprani
40  2 × 104  7 2 − 3  δ1 x 

 =    (4.3.73)
0 4  2 − 3 5  δ1 y 
Solve this to get:
δ1 x  4 1  5 3 − 2  40 
 =   
(
δ1 y  2 × 10 7 ⋅ 5 − 2 − 3   3 − 2 )
4 2
7   0 
  (4.3.74)
 1.15 
= 10−3  m
−0.06 
Using equation (4.3.21) we can now find the member forces for Stage II:
2 × 104 ( 2 )  1 3   0 − (1.15 )  −3
F = II
⋅ −  10 =+12.54 kN (4.3.75)
2  0 − ( −0.06 ) 
12
2.0  2
2 × 104  0 − (1.15 )  −3
F13II = ⋅ [ −1 0]  10 =+23.0 kN (4.3.76)
1.0  0 − ( −0.06 ) 
F =
2 × 104 2  1
⋅ −
( ) −
1   0 − (1.15 )  −3
 10 =+15.4 kN
II
(4.3.77)
2  0 − ( −0.06 ) 
14
1.0 2  2
67 Dr. C. Caprani
Final
The final member forces are the superposition of Stage I and Stage II forces:
F12 =F12I + F12II =+

0 12.54 =
12.54 kN (4.3.78)
F13 =F13I + F13II =−40 + 23.0 =−17.0 kN (4.3.79)
F14 =F14I + F14II =+

0 15.4 =
15.4 kN (4.3.80)
Thus the final result is:
68 Dr. C. Caprani
4.3.9 Example 6 – Truss with Loads & Self Strains
Problem
Analyse the same truss as Example 5, allowing for the following additional load
sources:
• 80 kN acting horizontally to the left at node 1;
• 100 kN acting vertically downwards at node 1;
• Member 14 is 5√2 mm too short upon arrival on site.
All as shown below:
69 Dr. C. Caprani
Solution
Again we will separate the actions into Stage I and Stage II scenarios.
Stage I
Displacements are suppressed and as a result the only sources of forces are self-
straining forces:
The forces and displacements for Stage I are thus:
FT = 40kN as before,
∆L 5 2 × 10−3
F∆L =
EA 2 ( 2 × 104 )
= =
100 2 kN (4.3.81)
L 2
=δ1x 0;=δ1 y 0 (4.3.82)
F12I =
0; F13I =
−40; F14I =
100 2 (4.3.83)
70 Dr. C. Caprani
Stage II
In this stage displacements are allowed and the forces in the self-strained members
are now applied to the joints, in addition to any external loads. Thus we have:
Clearly we need to resolve the forces at node 1 into net vertical and horizontal forces:
Since the members have not changed from Example 5, we can use the same stiffness
matrix. Therefore we have :
−140  2 × 104  7 2 − 3  δ1 x 

 =    (4.3.84)
−200  4  2 − 3 5  δ1 y 
71 Dr. C. Caprani
Solve this:
δ1 x  4 1  5 3 − 2  −140 
  =   
( )
δ1 y  2 × 10 7 ⋅ 5 − 2 − 3   3 − 2 7  −200 
4 2
  (4.3.85)
−3.7 
= 10−3  m
 −7.8 
Using equation (4.3.21) we can find the member forces for Stage II:
2 × 104 ( 2 )  1 3  0 − ( −3.7 )  −3
F =
II
⋅ −  10 =+98.1 kN (4.3.86)
2  0 − ( −7.8 ) 
12
2.0  2
2 × 104 0 − ( −3.7 )  −3
F =
II
⋅ [ −1 0]  10 =−74.0 kN (4.3.87)
13
1.0  0 − ( −7.8 ) 
F =
2 × 104 2  1
⋅ −
( ) −
1  0 − ( −3.7 )  −3
 10 =−162.6 kN
II
(4.3.88)
2  0 − ( −7.8 ) 
14
1.0 2  2
72 Dr. C. Caprani
Final
As before, the final member forces are the Stage I and Stage II forces:
F12 =F12I + F12II =+

0 98.1 =98.1 kN (4.3.89)
F13 =F13I + F13II =−40 − 74.0 =−114.0 kN (4.3.90)
F14 =+
F14I F14II =
100 2 − 162.6 =
−21.6 kN (4.3.91)
Thus the final result is:
73 Dr. C. Caprani
4.3.10 Problems
Problem 1
Determine the displacements of joint 1 and the member forces for the following truss.
Take EA= 2 × 104 kN .
Problem 2
Determine the displacements of joint 1 and the member forces for the following truss.
Take EA= 2 × 104 kN , the area of both members is A√2.
Ans. δ1 x = +5 mm , δ1 y = 0
74 Dr. C. Caprani
Problem 3
Using any pertinent results from Problem 2, determine the area of member 14 such
that the horizontal displacement of node 1 is half what is was prior to the installation
of member 14. Determine also the force in member 14. Take EA= 2 × 104 kN ,
Ans. A14 = A , F14 = −50 kN
Problem 4
Determine the displacements of the joints and the member forces for the following
truss. Take EA= 2 × 104 kN , the area of all members is A√2, except for member 24
which has an area of A.
75 Dr. C. Caprani
Problem 5
truss. Take EA= 2 × 104 kN , the area of the members is as shown.
Problem 6
truss. Take EA= 2 × 104 kN , the area of all members is A, except for member 14
which has an area of A√2.
76 Dr. C. Caprani
4.4 Beams
4.4.1 Beam Element Stiffness Matrix

To derive the beam element stiffness matrix, we recall some results obtained
previously, summarized here:
Next we must adopt strict local element sign convention and node identification:
77 Dr. C. Caprani
Anti-clockwise moments and rotations (i.e. from the x-axis to the y-axis) are positive
and upwards forces are positive.
Thus for a vertical displacement of ∆ at node i, now labelled δ iy , we have the
following ‘force’ vector:
 12 EI 
 L3 
 Fiy   6 EI 
M   
 i   L2 
 = δ iy (4.4.1)
F
 jy  − 12 EI 
 M j   L  3
 6 EI 
 
 L2 
Similarly, applying the same deflection, but at node j, δ jy , gives:
 12 EI 
− L3 
 Fiy   6 EI 
M   − 2 
 i  L 
 = δ jy (4.4.2)
F
  jy 12 EI 
 M j   L3 
 6 EI 
− 2 
 L 
Next, applying a rotation to node i, θ i , gives:
78 Dr. C. Caprani
 6 EI 
 L2 
 Fiy   4 EI 
M   
 i  L 
 = θ i (4.4.3)
F
  −
jy 6 EI 
 M j   L2 
 2 EI 
 
 L 
And a rotation to node j, θ j , gives:
 6 EI 
 L2 
 Fiy   2 EI 
M   
 i  L 
 = θ j (4.4.4)
F
  −
jy 6 EI 
 M j   L2 
 4 EI 
 
 L 
Since all of these displacement could happen together, using superposition we thus
have the total force vector as:
 12 EI   6 EI   12 EI   6 EI 
 L 3  L 2
−
 L  3  L2 
 Fiy         
M   6 EI   4 EI   − 6 EI   2 EI 
 i  L2   L   L2   L 
 =  δ +
 iy  θ +
 i  δ jy +  θ j (4.4.5)
 Fjy  − 12 EI  − 6 EI   12 EI  − 6 EI 
 M j   L3   L2   L3   L2 
 6 EI   2 EI   6 EI   4 EI 
     −   
 L2   L   L2   L 
Writing this as a matrix equation, we have:
79 Dr. C. Caprani
 12 EI 6 EI 12 EI 6 EI 
 L3 −
L2 L3 L2 
 Fiy   6 EI 
2 EI  δ iy 
M  
4 EI 6 EI
− 2  
 i   L2 L L L   θi 
 =   (4.4.6)
 Fjy   − 12 EI 6 EI 12 EI 6 EI  δ jy 
− 2 − 2 
 M j   L3 L L3 L   θ j 
 6 EI 2 EI 6 EI 4 EI 
 − 2 
 L2 L L L 
This is in the typical form:
{F } = [k ]{u }
e e
(4.4.7)
And so the beam element stiffness matrix is given by:
 12 EI 6 EI 12 EI 6 EI 
 L3 −
L2 L3 L2 
 
 6 EI 4 EI 6 EI
− 2
2 EI 
 L2 L 
[k ] =  L L (4.4.8)
12 EI 6 EI 12 EI 6 EI 
− 3 − 2 − 2 
 L L L3 L 
 6 EI 2 EI 6 EI 4 EI 
 − 2 
 L2 L L L 
Next we note a special case where the vertical displacements of the beam nodes are
prevented and only rotations of the beam ends is allowed. In this case, all terms
relating to the translation DOFs are removed giving us the reduced stiffness matrix
for a beam on rigid vertical supports:
80 Dr. C. Caprani
 4 EI 2 EI 
 L 
[k ] =  L  (4.4.9)
 2 EI 4 EI 
 L L 
As we did for trusses, we will often write these equations in terms of nodal sub-
matrices as:
 Fi  k11 k12   δi 
 =   (4.4.10)
Fj  k21 k22  δ j 
81 Dr. C. Caprani
4.4.2 Beam Element Loading
Applied Loads
Beam loads are different to truss loads since they can be located anywhere along the
element, not only at the nodes – termed intermodal loading Beams can also have
loads applied to the nodes – nodal loading. We deal with these two kinds of loads as
follows:
• Nodal loads: apply the load to the joint as usual;

• Inter-nodal loads: apply the equivalent concentrated loads to the joints (these are
just fixed end moment reactions to the load, with the direction reversed).
If a member’s nodes are locked against rotation, the member end forces due to inter-
nodal loading will just be the fixed end moment and force reaction vector we are
familiar with {FF } . If a member also displaces, the total member end forces are:
} {F } + [k ]{δ}
{F=Tot F
(4.4.11)
Thus the general stiffness equation becomes:
{F} = [K ]{δ} (4.4.12)
Where {F} is now the vector of net nodal loads:
=
Net Nodal Load Nodal Load − ∑ Fixed End Reactions (4.4.13)
82 Dr. C. Caprani
Lastly, we must note that inter-nodal loads on adjacent members will result in
multiple loads on a node. Thus we must take the algebraic sum of the forces/moments
on each node in our analysis, bearing in mind the sign convention.
As an example, the equivalent nodal loads for a UDL applied to a beam element are:
Member End Forces

After the deformations of the beam are known, we can use the element stiffness
matrices to recover the end forces/moments on each element due to both
deformations and the inter-nodal loading directly from equation (4.4.11).
83 Dr. C. Caprani
4.4.3 Example 7 – Simple Two-Span Beam
Problem
For the following beam, find the rotations of joints 2 and 3 and the bending moment
diagram. Take EI = 6 × 103 kNm 2 .
Solution
First we write the general equation in terms of nodal sub-matrices:
 F1   K 11 K 12 K 13   δ1 
    
F2  = K 21 K 22 K 23  δ 2  (4.4.14)

F   K K 33  δ3 
 3   31 K 32
Next we note that the only possible displacements are the rotations of joints 2 and 3.
Thus we can restrict the equation by eliminating joint 1 as follows:
 F1   K 11 K 12 K 13   δ1 
    
F2  = K 21 K 22 K 23  δ 2  (4.4.15)

F   K K 33  δ3 
 3   31 K 32
To give:
84 Dr. C. Caprani
F2  K 22 K 23  δ 2 
 =  
K 33  δ3 
(4.4.16)
F3   K 32
Since this beam is on rigid vertical supports, we can use the beam stiffness matrix
given by equation (4.4.9). Thus we are left with two equations:
 M 2   k22 k23  θ 2 
 =  
k33  θ 3 
(4.4.17)
 M 3   k32
The member contributions to each of these terms are:

• k22 = Term 22 of Member 12 + Term 11 of Member 23 ;
• k23 = Term 12 of Member 23 ;

• k33 = Term 22 of Member 23
Thus, for Member 12 we have:
 4 EI 2 EI   4( 6) 2 (6) 
 L  
L  4 2
=[ ]12 
k =  10 3
 =
6 6 
103   (4.4.18)
 2 EI 4 EI   2 ( 6) 4 ( 6)  2 4
 L L   6 6 
And for Member 23:
 4( 6) 2 (6) 
  4 2
=[k ]23 10=
3
 6 6 
103   (4.4.19)
 2( 6) 4 (6)  2 4
 6 6 
85 Dr. C. Caprani
Thus the global stiffness equation is:
M 2  3 
8 2  θ 2 
  = 10    (4.4.20)
M 3   2 4  θ 3 
To find the moments to apply to the nodes, we determine the fixed-end moments
caused by the loads on each members. Only Member 23 has load, and its fixed end
moments are:
Our sign convention is anti-clockwise positive. Thus the moments to apply to the
joints become (refer to equation (4.4.13)):
 30  3 
8 2  θ 2 
−  =
10  2 4  θ  (4.4.21)
−30    3
Solving the equation:
θ 2  1 1  4 −2  −30  −3 
−90 14 
  =     10   rads (4.4.22)
θ 3  10 ( 8 ⋅ 4 − 2 ⋅ 2 )  −2 8   30 
3
150 14 
Since we know that anti-clockwise is positive, we can draw the displaced shape (in
mrads):
86 Dr. C. Caprani
Using the member stiffness matrices we can recover the bending moments at the end
of each member, now that the rotations are known, from equation (4.4.11):
 M 1  0  3 
4 2  −3  0   −12.9 
   = + 10  2 4 10 −90 14  =
  kNm (4.4.23)
 2  
M 0      −25.7 
 M 2   30  3 
4 2  −3 −90 14  +25.7 
 =   + 10  10 150 14  =  kNm (4.4.24)
 M 3  −30  2 4    0 
Thus the final BMD can be drawn as:
87 Dr. C. Caprani
4.4.4 Example 8 – Non-Prismatic Beam
Problem
For the following beam, find the vertical deflection of joint 2 and the bending
moment diagram. Take EI= 12 × 103 kNm 2 .
Solution
First we write the general equation in terms of nodal sub-matrices:
 F1   K 11 K 12 K 13   δ1 
    
F2  = K 21 K 22 K 23  δ 2  (4.4.25)

F   K K 33  δ3 
 3   31 K 32
Next we note that the only possible displacements are those of joint 2. Thus we can
restrict the equation to:
{F } = [K ]{δ }
2 22 2
(4.4.26)
The member contributions to K 22 are:

• Sub-matrix k22 of member 12;
• Sub-matrix k11 of member 23.
88 Dr. C. Caprani
That is:
[K 22 ]
= [k22 ] + [k11 ]
12 23
(4.4.27)
For member 12, we have, from equation (4.4.8):
 12 EI 6 EI   12 ⋅ 12 6 ⋅ 12 
 L3 − 2   − 2 
L 4 3
4 3 
2.25 −4.5
=k2212 =  10  =
3
 10   (4.4.28)
− 6 EI 4 EI  − 6 ⋅ 12 4 ⋅ 12   −4.5 12 
 L2 L   
 4 2
4  
And for member 23:
12 EI 6 EI  12 ⋅ 24 6 ⋅ 24 
 L3 L2  103  = 43 42  103  4.5 9 
=k1123 =     9 24  (4.4.29)
 6 EI 4 EI   6 ⋅ 24 4 ⋅ 24   
 L2 L  
 4 2
4  
Since the load is a directly applied nodal load we can now write equation (4.4.26),
using equations (4.4.27), (4.4.28), and (4.4.29), as:
 F2 y  −100  3 
6.75 4.5 δ 2 y 
=
   =  10  4.5 36   θ  (4.4.30)
M 2   0    2 
Solving:
δ 2 y  1  36 −4.5 −100  −3 
−16.16 
  10 =
−3
    (4.4.31)
( 6.75 ⋅ 36 − 4.5 ⋅ 4.5)  −4.5 6.75   0 
10
 θ2   2.02 
89 Dr. C. Caprani
Thus we have a downwards (negative) displacement of 16.2 mm and an

anticlockwise rotation of 2.02 mrads at joint 2, as shown:
Next we recover the element end forces. For member 12, from equation (4.4.6) we
have:
 F1 y   2.25 4.5 −2.25 4.5   0   45.5 

M   4.5 −4.5 6  −3  0   
 1 3 
12
10   84.8 
  10 =    (4.4.32)
 F2 y
 −2.25 −4.5 2.25 −4.5   −16.16   −45.5 
 M 2    
 4.5 6 −4.5 12   2.02   97.0 
And for member 23:
 F2 y   4.5 9 −4.5 9  −16.16   −54.5 

M   9 24 −9 12  −3  2.02   −97.0 
 2 3  10   
  10 =    (4.4.33)
 F3 y   −4.5 −9 4.5 −9   0   54.5 
 
 M 3   9 12 −9 24   0  −121.2 
Thus the member end forces are:
90 Dr. C. Caprani
As can be seen, the load is split between the two members in a way that depends on
their relative stiffness.
The total solution is thus:
91 Dr. C. Caprani
4.4.5 Problems
Problem 1
Determine the bending moment diagram and rotation of joint 2. Take
EI= 10 × 103 kNm 2 .
Problem 2
Determine the bending moment diagram and the rotations of joints 1 and 2. Take
EI= 20 × 103 kNm 2 .
92 Dr. C. Caprani
Problem 3
Determine the bending moment diagram and the displacements of joints 2 and 3.
Take EI= 20 × 103 kNm 2 .
Problem 4
Determine the bending moment diagram and the vertical displacement under the 100
kN point load. Take EI= 10 × 103 kNm 2 .
93 Dr. C. Caprani
Problem 5
EI= 20 × 103 kNm 2 .
Problem 6
Determine the bending moment diagram and the rotations of all joints. Take
EI= 40 × 103 kNm 2 . You may use Excel or Matlab to perform some of the numerical
calculations. Check your member stiffness and global stiffness matrices with LinPro,
and your final results. Identify and explain discrepancies. Verify with LUSAS.
94 Dr. C. Caprani
4.5 Plane Frames
4.5.1 Plane Frame Element Stiffness Matrix

A plane frame element is similar to a beam element except for some differences:
• The presence of axial forces;
• The member may be oriented at any angle in the global axis system;
• The inter-nodal loads may be applied in the local or global coordinates.
These points are illustrated in the following:
Lastly, an easy way to deal with inter-nodal point loads ( PG , PL ) is to introduce a

node under the point load (splitting the member in two), then it is no longer inter-
nodal and so no transformations or equivalent load analysis is required. The downside
to this is that the number of equations increases (which is only really a problem for
analysis by hand).
95 Dr. C. Caprani
Axial Forces
To include axial forces, we can simply expand the beam element stiffness matrix to
allow for the extra degree of freedom of x-displacement at each node in the member
local coordinates. Thus expanding equation (4.4.8) to allow for the extra DOFs gives:
 X 11 0 0 X 14 0 0 
 12 EI 6 EI 12 EI 6 EI 
 0 0 − 3 
 L3 L2 L L2 
 6 EI 4 EI 6 EI 2 EI 
 0 0 − 2 
L2
[k ] =  L L L  (4.5.1)
X 41 0 0 X 44 0 0 
 
 0 12 EI 6 EI 12 EI 6 EI 
− 3 − 2 0 − 2
 L L L3 L 
 6 EI 2 EI 6 EI 4 EI 
 0 0 − 2 
 L2 L L L 
However, these terms that account for axial force are simply those of a plane truss
element in its local coordinate system:
EA  1 −1
[k ] =
L  −1 1 
(4.5.2)
96 Dr. C. Caprani
 EA EA 
 L 0 0 − 0 0 
L
 
 0 12 EI 6 EI
0 −
12 EI 6 EI 
 L3 L2 L3 L2 
 6 EI 4 EI 6 EI 2 EI 
 0 0 − 2 
[k ] =  L2 L L L 

(4.5.3)
EA EA
− 0 0 0 0 
 L L 
 12 EI 6 EI 12 EI 6 EI 
 0 − 3 − 0 − 2 
L L2 L3 L
 
 0 6 EI 2 EI
0
6 EI
− 2
4 EI 
 L2 L L L 
This is the stiffness matrix for a plane frame element in its local coordinate system
and can also be written in terms of nodal sub-matrices as:
k11 k12 
[k ] =   (4.5.4)
k21 k22 
Where the nodal sub-matrices are as delineated in equation (4.5.3).
Note that if axial forces are neglected, we can just use the regular beam element
stiffness matrix instead, though coordinate transformation may be required.
97 Dr. C. Caprani
Transformation to Global Coordinates

From the Appendix, the plane frame element stiffness matrix in global coordinates is:
K e  = [ T] [k ][ T]
T
(4.5.5)
As a consequence, note that we do not need to perform the transformation when:

1. The member local axis and global axis system coincide;
2. The only unrestrained DOFs are rotations/moments.
Again from the Appendix, the transformation matrix for a plane frame element is:
 cos α sin α 0 0 0 0
 − sin α cos α 0 0 0 0
 
 0 0 1 0 0 0
T=  (4.5.6)
 0 0 0 cos α sin α 0
 0 0 0 − sin α cos α 0
 
 0 0 0 0 0 1
98 Dr. C. Caprani
Inter-nodal Loads
In plane frames, loads can be applied in the global axis system, or the local axis
system. For example, if we consider a member representing a roof beam, we can have
the following laods:
• Case 1: Gravity loads representing the weight of the roof itself;
• Case 2: Horizontal loads representing a horizontal wind;
• Case 3: Net pressure loads caused by outside wind and inside pressures.
Case 1 Case 2 Case 3
Most structural analysis software will allow you to choose the axis system of your
loads. However, in order to deal with these loads for simple hand analysis we must
know how it works and so we consider each case separately.
In the following the member local axis system has a prime (e.g. x’) and the global
axis system does not (e.g. x).
99 Dr. C. Caprani
Case 1: Vertically Applied Loads

In this case we can consider an equivalent beam which is the projection of the load
onto a horizontal beam of length LX :
Since the resulting nodal forces and moments are in the global axis system no further
work is required.
100 Dr. C. Caprani

Case 2: Horizontally Applied Loads:

Similarly to vertically applied loads, we can consider the horizontal projection of load
onto an equivalent member of length LY .
Again the resulting nodal loads are in the global axis system and do not require any
modification.
101 Dr. C. Caprani

Case 3: Loads Applied in Local Member Axis System

In this case there is no need for an equivalent beam and the fixed-fixed reactions are
worked out as normal:
However, there is a complication here since the reactions are now not all in the global
axis system. Thus the forces (not moments) must be transformed from the local axis
to the global axis system. Thus there is a simple case:
If axial forces are neglected, only moments are relevant and so no transformations are
required.
For generality though we can use the transformations given in the Appendix:
{F} = [T] {F '}

T
(4.5.7)
Writing this out in full for clarity, we have:
102 Dr. C. Caprani

 Fixij  cos α − sin α 0 0 0 0   Fixij' 

 ij    
 Fiy   sin α cos α 0 0 0 0   Fiyij' 

 M iij   0 0 1 0 0 0   M iij 
 ij  =    (4.5.8)
 Fjx   0 0 0 cos α − sin α 0   Fjxij ' 
 Fjyij   0 0 0 sin α cos α 0   Fjyij ' 
 ij    
 M j   0 0 0 0 0 1   M ijj 
103 Dr. C. Caprani

4.5.2 Example 9 – Simple Plane Frame
Problem
For the following frame, determine the rotation of the joints and the bending moment
diagram. Neglect axial deformations. Take EI = 1 × 103 kNm 2 .
Solution
The fact that we can neglect axial deformation makes this problem much simpler. As
a consequence, the only possible displacements are the rotations of joints 1 and 2.
Since node 3 is fully restricted out, we have the following partially-restricted set of
equations in terms of nodal sub-matrices:
 F1  K11 K12  δ1 

 =   (4.5.9)
F2  K21 K22  δ 2 
If we expand this further, we will be able to restrict out all but the rotational DOFs:
104 Dr. C. Caprani

 ⋅  ⋅ ⋅ ⋅ ⋅ ⋅ ⋅  ⋅ 
 ⋅  ⋅ ⋅ ⋅ ⋅ ⋅ ⋅  ⋅ 
    
 ∑ M 1  ⋅ ⋅ k33 ⋅ ⋅ k36  θ1 
 =   (4.5.10)
 ⋅  ⋅ ⋅ ⋅ ⋅ ⋅ ⋅  ⋅ 
 ⋅  ⋅ ⋅ ⋅ ⋅ ⋅ ⋅  ⋅ 
    
∑ M 2  ⋅ ⋅ k63 ⋅ ⋅ k66  θ 2 


• k66 Term 66 of Member 12 + Term 33 of Member 23 .
• Member 12:
Looking at equation (4.5.3):
 4 EI  4 ⋅ 103
Term 33=   = = 4 × 103 (4.5.11)
 L 12 1
 2 EI  2 ⋅ 103
Term 36=   = = 2 × 103 (4.5.12)
 L 12 1
 2 EI  2 ⋅ 103
Term 63=   = = 2 × 103 (4.5.13)
 L 12 1
 4 EI  4 ⋅ 103
Term 66=   = = 4 × 103 (4.5.14)
 L 12 1
• Member 23:
Again, from equation (4.5.3):
105 Dr. C. Caprani

 4 EI  4 ⋅ 103
Term 33=   = = 4 × 103 (4.5.15)
 L  23 1
Thus the system equation becomes:
 ∑ M1  3 
4 2  θ1 
  = 10   θ  (4.5.16)
 ∑ M 2   2 8  2
Next we must find the net moments applied to each node. There are no directly
applied nodal moment loads, so the ‘force’ vector is, from equation (4.4.13):
{F} = −{F }F
(4.5.17)
• Member 12 Moments:
wL2 12 ⋅ 12
M =12
1
= = +1 kNm
12 12 (4.5.18)
wL2 12 ⋅ 12
M 212 =
− =
− =−1 kNm
12 12
• Member 23 Moments:
106 Dr. C. Caprani

PL 16 ⋅ 1
M 223 = = = +2 kNm
8 8
(4.5.19)
PL 16 ⋅ 1
M 323 =
− =
− =
−2 kNm
8 8
Thus the net nodal loads become:
∑ M   +1  −1
{F} =
− {F } =
− 1
=− =  kNm (4.5.20)
∑ M − +  −1
F
2  1 2
And so equation (4.5.16) is thus:
−1 3 
4 2  θ1 
  = 10  2 8  θ  (4.5.21)
−1   2
Which is solved to get:
θ1  1 1  8 −2  −1 − 3 14 
=
   =     × 10 rads
−3
(4.5.22)
θ 2  10 ( 4 ⋅ 8 − 2 ⋅ 2 )  −2 4  −1  −1 14 
3
The negative results indicate both rotations are clockwise.
107 Dr. C. Caprani

Lastly, we must find the member end forces. Since we only need to draw the bending
moment diagram so we need only consider the terms of the member stiffness matrix
relating to the moments/rotations (similar to equation (4.4.9)). Also, we must account
for the equivalent nodal loads as per equation (4.4.11):
• Member 12:
 M 112  +1 3 
4 2  − 3 14  −3  0 
 12  =   + 10   10 =  kNm (4.5.23)
 M 2  −1  2 4   −1 14  −12 7 
• Member 23:
 M 223  +2  3 
4 2  −1 14  −3 +12 7 
 23  =   + 10   10 =  kNm (4.5.24)
 M 3  −2  2 4  0  −17 7 
108 Dr. C. Caprani

4.5.3 Example 10 –Plane Frame Using Symmetry
Problem
For the following frame, determine the rotation of the joints, the displacement under
the 8 kN point load and the bending moment diagram. Neglect axial deformations.
Take EI = 1 × 103 kNm 2 .
Solution
Again, the fact that we can neglect axial deformation makes this problem much
simpler. Since the structure is symmetrical and it is symmetrically loaded, it will not
sway. Further, because of this symmetry, we can adopt the following model for
analysis:
109 Dr. C. Caprani

Notice two things from this model:

• we have renumber the joints – there is no need to retain the old numbering system;
• The remaining DOFs are θ 2 and δ 3 y - we can restrict all other DOFs. Thus in terms
of nodal sub-matrices we immediately have:
F2  K11 K12  δ 2 

 =   (4.5.25)
F3  K21 K22  δ3 
And expanding this further, we restrict out all other restrained DOFs:
 ⋅  ⋅ ⋅ ⋅ ⋅ ⋅ ⋅  ⋅ 
 ⋅  ⋅ ⋅ ⋅ ⋅ ⋅ ⋅  ⋅ 
    
∑ M 2  ⋅ ⋅ k66 ⋅ k68 ⋅  θ 2 
 =   (4.5.26)
 ⋅  ⋅ ⋅ ⋅ ⋅ ⋅ ⋅  ⋅ 
 ∑ F3 y  ⋅ ⋅ k86 ⋅ k88 ⋅ δ 3 y 
    
 ⋅  ⋅ ⋅ ⋅ ⋅ ⋅ ⋅  ⋅ 

• k66 Term 66 of Member 12 + Term 33 of Member 23 ;
• k66 = Term 55 of Member 23 .
Transformation of the member stiffness matrices is not required. Member 12 only has
a rotational DOF and Member 23’s local member coordinate system is parallel to the
global axis coordinate system.
110 Dr. C. Caprani

• Member 12:
From equation (4.5.3):
 4 EI  4 ⋅ 103
Term 66=   = = 4 × 103 (4.5.27)
 L 12 1
• Member 23:
Again, from equation (4.5.3):
 4 EI  4 ⋅ 103
Term 33=   = = 4 × 103 (4.5.28)
 L  23 1
 6 EI  6 ⋅ 10 3
Term 35 = − 2  =− 2 =−6 × 103 (4.5.29)

 L  23 1
 6 EI  6 ⋅ 10 3
Term 53 = − 2  =− 2 =−6 × 103 (4.5.30)

 L  23 1
 12 EI  12 ⋅ 103
Term 55=  3  = = 12 × 103 (4.5.31)
 L  23
3
1
Thus the system equation becomes:
∑ M 2  3 
8 −6   θ 2 
  = 10    (4.5.32)
 ∑ F3 y   −6 12  δ 3 y 
The 4 kN point load is directly applied to node 3 so this causes no difficulty. The
equivalent nodal loads for the UDL are:
111 Dr. C. Caprani

wL2 12 ⋅ 12
M =
12
1
= = +1 kNm
12 12 (4.5.33)
wL2 12 ⋅ 12
M 212 =
− =
− =−1 kNm
12 12
Notice that we do not need to find the vertical reaction forces as there is no sway of
the frame and we are neglecting axial deformation.
The nodal load vector, from equation (4.4.13) is thus:
 0  −1  +1
{∑ F} ={F} − {F } =  −   =  (4.5.34)
−4 −4
F
 0     
And so equation (4.5.32) is thus:
 +1 3 
8 −6   θ 2 
  = 10    (4.5.35)
−4   −6 12  δ 3 y 
Which is solved to get:
 θ2  1 1 12 6   +1  −0.2  rads

=
  = 
 6 8  −4    × 10 −3
(4.5.36)
δ
 3 y 10 3

8 ⋅ 12 − ( −6 )( −6 ) 
      −0.43  m
112 Dr. C. Caprani

The negative results indicate the rotation is clockwise and the displacement
downwards, as may be expected:
Lastly then we find the bending moments. For member 12 only the terms relating to
bending moments are relevant.
• Member 12:
 M 112  +1 3 
4 2   0  −3 +0.6 
 12  =   + 10   10 =
  kNm (4.5.37)
 M 2  −1  2 4  −0.2   −1.8
However, for member 23, the downwards deflection also causes moments and so the
relevant DOFs are rotation of node i and vertical movement of node j (as calculated
earlier). It is easier to see this if we write the member equation in full:
• Member 23:
 ⋅  ⋅ ⋅ ⋅ ⋅ ⋅ ⋅  ⋅ 
 ⋅  ⋅ ⋅ ⋅ ⋅ ⋅ ⋅  ⋅ 
    
M 2 
23
⋅ ⋅ 4 ⋅ −6 ⋅  −0.2 
=  10 
3
  × 10
−3
(4.5.38)
 ⋅  ⋅ ⋅ ⋅ ⋅ ⋅ ⋅  ⋅ 
 ⋅  ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ −0.43
 23    
M 3  ⋅ ⋅ 2 ⋅ −6 ⋅  ⋅ 
113 Dr. C. Caprani

Thus:
 M 223   +1.8 
 23  =   kNm (4.5.39)
 M 3   +2.2 
And so the BMD is:
114 Dr. C. Caprani

4.5.4 Problems
Problem 1
Determine the bending moment diagram and the rotation of joint 2. Take
EI= 10 × 103 kNm 2 and neglect axial deformations.
(Ans. θ 2 = 5 6 mrads )
Problem 2
EI= 20 × 103 kNm 2 and neglect axial deformations.
115 Dr. C. Caprani

Problem 3
Determine the bending moment diagram. Take EI= 20 × 103 kNm 2 and neglect axial
deformations.
Problem 4
Determine the bending moment diagram, the rotation of joint 2, and the horizontal
displacements of joints 2 and 3. Take EI= 10 × 103 kNm 2 and neglect axial
deformations.
(Ans. θ 2 =
−11.33 mrads; δ 2 x =
44.0mm )
116 Dr. C. Caprani

Problem 5
Determine the bending moment diagram. Take EI= 20 × 103 kNm 2 and neglect axial
deformations.
Problem 6
Determine the bending moment diagram and the vertical displacement of joint 3.
Take EI= 40 × 103 kNm 2 and neglect axial deformations.
117 Dr. C. Caprani

Problem 7
Determine the bending moment diagram, the rotation of joint 2, and the vertical
displacement under the 80 kN point load. Take EI= 10 × 103 kNm 2 and neglect axial
deformations.
(Ans. θ 2 =
−1.071 mrads; δ y =
−1.93mm )
Problem 8
For the frame of Problem 1, determine the bending moment diagram and the rotation
and vertical displacement of joint 2 if member 24 has EA= 10 × 103 kN . Neglect axial
deformation in the other members.
(Ans. θ 2 = 0.833 mrads; δ y = −0.01mm )
118 Dr. C. Caprani

Problem 9
Determine the bending moment diagram for the prismatic portal frame. Take
EI= 20 × 103 kNm 2 and neglect axial deformations. You may use Excel or Matlab to
perform some of the numerical calculations. Check your member stiffness and global
stiffness matrices with LinPro, and your final results. Identify and explain
discrepancies. Verify with LUSAS.
119 Dr. C. Caprani

4.6 Appendix
4.6.1 Plane Truss Element Stiffness Matrix in Global Coordinates
Compatibility Conditions
Firstly we indentify the conditions of compatibility of a truss element nodal
deflections and the member elongation. We use the following notation for the
deflections at each node of the truss:
If we now consider the deflected position of the truss member, we have:
120 Dr. C. Caprani

Obviously the change in length of the truss will be related to the difference between
the nodal deflections. Hence, we define the changes in movements such that an
elongation gives positive changes:
∆δ x = δ jx − δ ix ∆δ y = δ jy − δ iy
Moving the deflected position of node i back to its original location gives:
Looking more closely at the triangle of displacements at node j, and remembering

that we are assuming small deflections—which in this case means the deflected
position of the member is still at a rotation of θ . Hence we have:
121 Dr. C. Caprani

And so the elongation is given by:
e = ∆δ x cosθ + ∆δ y sin θ
(4.6.1)
= (δ jx − δ ix ) cosθ + (δ jy − δ iy ) sin θ
Now multiply out and re-order to get:
e = −δ ix cosθ + −δ iy sin θ + δ jx cosθ + δ jy sin θ

(4.6.2)
If we define a direction vector, α , and a displacement vector, δ , as:
− cosθ  δ ix 
 − sin θ   
  δ iy 
=α =  δ   (4.6.3)
 cos θ  δ jx 
 sin θ  δ jy 
Then, from (4.6.2) and (4.6.3), we can say:
122 Dr. C. Caprani

e = αtδ (4.6.4)
Thus we have related the end displacements to the elongation of the member which
therefore maintain compatibility of displacement.
123 Dr. C. Caprani

Virtual Work for Element Forces

Looking at the forces acting on the nodes of the bar element, we have:
This is a force system in equilibrium—the external nodal loading is in equilibrium

with the internal bar force, N. If we consider a pattern of compatible displacements
such as the following:
124 Dr. C. Caprani

We can apply virtual work to this:
δW = 0
δ WE = δ WI
And we have:
Set of forces in
equilibrium
Set of compatible
displacements
Substituting in our notations for the bar element:
eN = Fixδ ix + Fiyδ iy + Fjxδ jx + Fjyδ jy (4.6.5)
If we define the force vector, F , as:
 Fix 
F 
 iy 
F=  (4.6.6)
 Fjx 
 Fjy 
Then we can write (4.6.5) as:
F t δ = eN (4.6.7)
125 Dr. C. Caprani

If we use (4.6.4) we how have:
F t δ = α t δN (4.6.8)
Post-multiply both sides by δ −1 , and noting that N is a scalar, gives:
Ft = αt N
F = αN (4.6.9)
Expanding this out gives:
 Fix  − N cosθ 
   
 Fiy   − N sin θ 
 =  (4.6.10)
 Fjx   N cosθ 
 Fjy   N sin θ 
Which are the equations of equilibrium of the bar element:
126 Dr. C. Caprani

Relating Forces to Displacements

Lastly, in order to relate the end forces to the element nodal displacements, we note
from the constitutive law:
EA
=
N ⋅e (4.6.11)
L
And so from (4.6.9) we have:
EA
F=α e (4.6.12)
L
And using equation (4.6.4) gives:
EA t
F=α αδ (4.6.13)
L
EA t
Hence the term α α relates force to displacement and is called the stiffness
L
matrix, k , which is evaluated by multiplying out terms:
EA t
k =αα
L
− cosθ 
 − sin θ  (4.6.14)
EA  
=  [ − cosθ − sin θ cosθ sin θ ]
L  cosθ 
 sin θ 
And multiplying this out gives:
127 Dr. C. Caprani

 cos 2 θ cosθ sin θ − cos 2 θ − cosθ sin θ 

 
EA  cosθ sin θ sin 2 θ − cosθ sin θ − sin 2 θ 
k= (4.6.15)
L  − cos 2 θ − cosθ sin θ cos 2 θ cosθ sin θ 
 
 − cosθ sin θ − sin 2 θ cosθ sin θ sin 2 θ 
And for clarity, we write out the final equation in matrix form and in full:
F = kδ (4.6.16)
 Fix   cos 2 θ cosθ sin θ − cos 2 θ − cosθ sin θ  δ ix 

F    
 iy  EA  cosθ sin θ sin 2 θ − cosθ sin θ − sin 2 θ  δ iy 
 =   (4.6.17)
 Fjx  L  − cos θ
2
− cosθ sin θ cos 2 θ cosθ sin θ  δ jx 
 Fjy   
 − cosθ sin θ − sin 2 θ cosθ sin θ sin 2 θ  δ jy 
So for example, the stiffness that relates a horizontal force at node j to the horizontal
displacement at node j is:
 EA 
Fjx =  cos 2 θ  δ jx
 L 
And other relationships can be found similarly.
128 Dr. C. Caprani

4.6.2 Coordinate Transformations
Point Transformation
We consider the transformation of a single point P from one coordinate axis system
xy to another x’y’:
From the diagram, observe:
OC = x ' coordinate of P
(4.6.18)
PC = y ' coordinate of P
Also:
OB = x coordinate of P
(4.6.19)
PB = y coordinate of P
129 Dr. C. Caprani

Next we can say:
= OA + AC
OC (4.6.20)
= PD − CD
PC (4.6.21)
Introducing the relevant coordinates:
= =
OA OA cos α x cos α (4.6.22)
= BD
AC = PB sin
= α y sin α (4.6.23)
OC= x=' x cos α + y sin α (4.6.24)
Next we have:
= =
PD PB cos α y cos α (4.6.25)
= AB
CD α x sin α
= OB sin= (4.6.26)
PC= y=' y cos α − x sin α (4.6.27)
Writing equations (4.6.24) and (4.6.27) together:
=x ' x cos α + y sin α (4.6.28)
130 Dr. C. Caprani

− x sin α + y cos α
y' = (4.6.29)
And now in matrix form gives:
 x '   cos α sin α   x 

 =  
cos α   y 
(4.6.30)
 y '  − sin α
Often we write:
c ≡ cos α
(4.6.31)
s ≡ sin α
To give:
 x '  c s   x 
 =   (4.6.32)
 y '  − s c   y 
Lastly, if we generically name the two coordinate systems as q and q’, we then have
in matrix form:
{q'} = [T ]{q}
N
(4.6.33)
Where [ TN ] is the nodal transformation matrix given by:
 cos α sin α 
TN = 
cos α 
(4.6.34)
 − sin α
131 Dr. C. Caprani

Force/Displacement Transformation
Forces and moments can be oriented in the local member axis system or in the global
structure axis system. In general we will need to transform the forces and
displacements of both nodes, thus we write:
 Fi '  TN 0   Fi 
 =  
TN  Fj 
(4.6.35)
F j '  0
And finally we can write:
{F '} = [T]{F} (4.6.36)
Where:
TN 0
[ T] =  0 TN 
(4.6.37)

Similarly for deflections:
{δ '} =[T]{δ} (4.6.38)
A very useful property of the transformation matrix (not derived here) is that it is
orthogonal. This means that its transpose is equal to its inverse:
[ T] = [ T]
T −1
(4.6.39)
132 Dr. C. Caprani

Thus when either a force or displacement is known for the local axis system, it can be
found in the global axis system as follows:
{F} = [T] {F '}

T
(4.6.40)
{δ} = [T] {δ '}

T
(4.6.41)
133 Dr. C. Caprani

Transformations for Plane Truss Element

For a plane truss member, there will be x and y components of force at each of its
nodes. Using the transformation for a point, we therefore have:
 Fx '   cos α sin α   Fx 

 =  
cos α   Fy 
(4.6.42)
 Fy '   − sin α
And so for a truss element, we have directly from equation (4.6.34):
 cos α sin α 
TN = 
cos α 
(4.6.43)
 − sin α
And so, from equation (4.6.37),
 − sin α cos α 0 0 
[T] =   (4.6.44)
0 0 cos α sin α 
 
 0 0 − sin α cos α 
For clarity, we write the transformation out in full:
 Fix '   cos α sin α 0 0   Fix 

    
 Fiy '   − sin α cos α 0 0   Fiy 
 =   (4.6.45)
 Fjx '   0 0 cos α sin α   Fjx 
 Fjy '   0 0 − sin α

cos α   Fjy 
134 Dr. C. Caprani

Transformations for Plane Frame Element

Based on the DOF transformation matrix for a plane truss member (in terms of
forces), we can determine the transformation matrix for a plane frame node quite
easily:
 Fxe   cos α sin α 0   Fx 

 e  − sin α  
 Fy  = cos α 0   Fy  (4.6.46)
 
 e  0 1   M 
M  0
This is because a moment remains a moment in the plane. So for a single node, and
both nodes, we have, respectively:
{F '} = [T ]{F}N
(4.6.47)
 Fi '  TN 0   Fi 
 =  
TN  Fj 
(4.6.48)
F j '  0
Thus, we can now write the final transformation matrix for a plane frame element as:
 cos α sin α 0 0 0 0
 − sin α cos α 0 0 0 0
 
 0 0 1 0 0 0
T=  (4.6.49)
 0 0 0 cos α sin α 0
 0 0 0 − sin α cos α 0
 
 0 0 0 0 0 1
135 Dr. C. Caprani

Element Stiffness Matrix Transformation

Using the general expression for a single element:
F e = K eδ e (4.6.50)
Regardless of member type or the number of dimensions, we will always have some
coordinate transform from local to global coordinates such that:
F e = TF (4.6.51)
δe = Tδ (4.6.52)
Hence from equation (4.6.50) we can write:
TF = K e Tδ (4.6.53)
And so the force-displacement relationship in the global axis system is:
F = T−1K eT  δ (4.6.54)
The term in brackets can now be referred to as the element stiffness matrix in global
coordinates. Thus, using equation (4.6.39), we write:
K Ge = TT K eL T (4.6.55)
136 Dr. C. Caprani

4.6.3 Past Exam Questions
Summer 2001
137 Dr. C. Caprani

Summer 2002
138 Dr. C. Caprani

Summer 2004
139 Dr. C. Caprani

Summer 2006
140 Dr. C. Caprani

Sample Paper 2006/7

1. (a) Using the stiffness method, determine the displacement of the joints of the pin-jointed truss shown in Fig.
Q1(a), under the load as shown.
(10 marks)
 




 
FIG. Q1(a)
(b) Members 15 and 16 are added to the truss of Fig. 1(a) to form the truss shown in Fig. Q1(b). However,
member 16 is found to be 15 mm too long and is forced into place. The same load of 100 kN is again to be
applied. Using the stiffness method, determine the displacement of the joints and the force in member 16.
(15 marks)
Take EA = 2×104 kN and the cross sectional areas of the members as:
Members 12, 13, and 16: 3A;

Diagonal Members 14 and 15: 3√2A.
  

 


 
FIG. Q1(b)
Ans. (a) 25 kN; 75 kN; 25√2 kN;

(b) 50√2 kN; 156.1 kN; 60.4 kN; -100√2 kN; -50√2 kN
141 Dr. C. Caprani

Semester 1 2006/7
1. Using the stiffness method, determine the displacement of the joints and the forces in the members of the pin-
jointed truss shown in Fig. Q1, allowing for:
(i) The 100 kN vertical load as shown, and;
(ii) A lack of fit of member 12, which was found to be 5 mm too short upon arrival at site, and
which was then forced into place.
Take EA = 2×104 kN and the cross sectional areas of the members as:
• Members 12: 3A;
• Members 13 and 14: 3√2A.
(25 marks)
Ans. 50 kN; -75√2 kN; -25√2 kN.
142 Dr. C. Caprani

Semester 1 Repeat 2006/7

1. Using the stiffness method, determine the displacement of the joints and the forces in the members of the pin-
jointed truss shown in Fig. Q1, allowing for:
(ii) The 100 kN vertical load as shown, and;
(ii) A lack of fit of member 12, which was found to be 10√2 mm too short upon arrival at site, and
which was then forced into place.
Take EA = 2×104 kN and the cross sectional areas of all members as 3√2A.
(25 marks)



 
 
FIG. Q1
Ans. 125√2 kN; -50√2 kN; -75√2 kN.
143 Dr. C. Caprani

Semester 1 2007/8
QUESTION 1
Using the stiffness method, determine the displacement of the joints and the forces in the members of the pin-jointed
truss shown in Fig. Q1, allowing for:
(i) The 100 kN load as shown, and;
(ii) A lack of fit of member 13, which was found to be 4 mm too short upon arrival at site, and which was then
forced into place;
(iii) A temperature rise of 20 ˚C in member 24.
Note:
= 125 × 103 kN and the coefficient of thermal expansion α =
Take EA 2 × 10-5 °C-1 .
(25 marks)
  

 

FIG. Q1
Ans. -55.7 kN; +69.7 kN; -55.3 kN.
144 Dr. C. Caprani

Semester 1 2008/9
QUESTION 1
Using the stiffness method, for the continuous beam shown in Fig. Q1, do the following:
(i) determine the displacement of the joints;
(ii) draw the bending moment diagram;
(iii) determine the reactions.
Note:
Take EI= 10 × 103 kNm 2 .
(25 marks)
 
 
  
 
   
FIG. Q1
Ans. 98.7 kNm; 102.6 kNm; 60.9 kNm.
145 Dr. C. Caprani

Semester 1 2009/10
QUESTION 1
Using the stiffness method, for the frame shown in Fig. Q1, do the following:
(i) determine the vertical displacement at the centre of the middle span;
(ii) draw the bending moment diagram;
(iii) determine the reactions.
Note:
Take EI= 10 × 103 kNm 2 .
(25 marks)
  
 

 
    
FIG. Q1
Ans. -11.88 mm
146 Dr. C. Caprani

Semester 1 2010/11
QUESTION 1
Using the stiffness method, for the truss shown in Fig. Q1:
(a) Determine:
(i) The displacement of the joints;
(ii) The forces in the members;
(iii) The deflected shape of the structure.

(15 marks)
(b) Determine the lack of fit of member 23, which would result in no horizontal displacement of joint 2 under the
100 kN load shown.
(10 marks)
Note:
• Take E = 200 kN/mm 2 for all members.
• Area for member 12 is A = 400 10 mm 2 .
• Area for member 23 is A = 960 mm 2 .
• Area for member 13 is A = 640 2 mm 2 .


  
 
FIG. Q1
Ans. 68.7 kN, 34.6 kN, 49.3 kN; 5.21 mm.
147 Dr. C. Caprani

4.7 References
• Alberty, J., Carstensen, C. and Funken, S.A. (1999), ‘Remarks around 50 lines of
Matlab: short finite element implementation’, Numerical Algorithms, 20, pp. 117-
137, available at: web address.
• Brown, D.K. (1990), An Introduction to the Finite Element Method using Basic
Programs, 2nd Edn., Taylor and Francis, London.
• Carroll, W.F. (1999), A Primer for Finite Elements in Elastic Structures, John
Wiley & Sons, New York.
• Coates, R.C., Coutie, M.G. and Kong, F.K. (1987), Structural Analysis, Chapman
and Hall.
• Davies, G.A.O. (1982), Virtual Work in Structural Analysis, John Wiley & Sons.
• Desai, C.S. and Abel, J.F. (1972), Introduction to the Finite Element Method: A
Numerical Method for Engineering Analysis, Van Nostrand Reinhold, New York.
• Ghali, A. and Neville, A.M. (1997), Structural Analysis – A unified classical and
matrix approach, 4th edn., E&FN Spon, London.
• McGuire, W., Gallagher, R.H. and Ziemian, R.D. (2000), Matrix Structural
Analysis, 2nd Edn., John Wiley & Sons.
• Meek, J.L. (1991), Computer Methods in Structural Analysis, 2nd Edn., E&FN
Spon.
• Przemieniecki, J.S. (1968), Theory of Matrix Structural Analysis, McGraw-Hill,
New York.
• Sack, R.L. (1989), Matrix Structural Analysis, Waveland Press, Prospect Heights,
Illinois, US.
• Thompson, F., and Haywood, G.G. (1986), Structural Analysis Using Virtual
Work, Chapman and Hall.
• Weaver, W. and Gere, J.M (1990), Matrix Analysis of Framed Structures, 3rd
Edn., Van Nostrand Reinhold, New York.
148 Dr. C. Caprani

Buku Referensi Anas 4 - Matrix Stiffness Method

Uploaded by

Copyright:

Available Formats

Buku Referensi Anas 4 - Matrix Stiffness Method

Uploaded by

Document Information

Original Description:

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

Buku Referensi Anas 4 - Matrix Stiffness Method

Uploaded by

Copyright:

Available Formats

Structural Analysis IV Chapter 4 – Matrix Stiffness Method

Chapter 4 - Matrix Stiffness Method

4.1 Introduction ......................................................................................................... 3

4.3.10 Problems ................................................................................................... 74

4.1.2 Basic Concepts

Local and Global

4.1.3 Computer Programs to Support Learning

Matlab Truss Analysis Program

It is necessary to use a scripting language like Matlab, rather than a spreadsheet

You can download LinPro from www.line.co.ba.

4.2 Basic Approach

4.2.1 Individual Element

F1 = k ( u1 − u2 ) = ku1 − ku2 (4.2.1)

Similarly for node 2:

And using matrix notation, we write:

• [k ] is the element stiffness matrix;

• {u } is the element displacement vector.

4.2.2 Assemblies of Elements

Consider the following simple structure:

{F} = [K ]{u} (4.2.10)

• {u} is the displacement vector for the structure.

The individual member properties are:

Member Length (m) Area (mm2) Material, E (kN/mm2)

Calculate the spring stiffnesses for each member:

Next we calculate the individual element stiffness matrices:

We expand and add the element stiffness matrices to get:

Notice how each member contributes to the global stiffness matrix:

Node 1 Node 2 Node 3 Node 4

For the remaining two equations, we have:

u2  1 1 340 200  −50  1  3 

Also note that it is in equilibrium (as we might expect).

Member 2 thus has tension of 54.8 kN.

Thus Member 3 has a compression of 45.08 kN applied to it.

Ans. 0.22 mm, 2.11 mm

4.2.4 General Methodology

Matlab Program - Implementation

kg = AssembleTrussK(nData, eData); % Assemble global stiffness matrix

The output from the function AnalyzeTruss is:

4.2.5 Member contribution to global stiffness matrix

Node i k11ij k12ij

Node j k21ij k22ij

Node i … k11ij … k12ij …

Node j … k21ij … k22ij …

Node j k11jl k12jl

Node l k21jl k22jl

And now the global stiffness matrix becomes:

… Node i … Node j … Node l …

Node i … k11ij … k12ij … … …

Node l … … k21lj … k22jl …

In the above, the identifiers k11 etc are sub-matrices of dimension:

Matlab Program – Element Contribution

% What nodes does the element connect to?

% The DOFs in kg to enter the properties into

% For each row of ke

Matlab Program – Global Stiffness Matrix Assembly

function kg = AssembleTrussK(nData, eData)

% How many nodes and elements are there?

% Set up a blank global stiffness matrix

% For each element

k = (EA/L) [ k11 -k11;

F = (EA/L) (c(djx-dix) + s(djy-diy));