STRUCTURAL OPTIMIZATION OF SAE BAJA CAR FRAME

Submitted by
Akshay Murkute – Torsional analysis and system integration
Anket Hirachand Marathe – Roll-over impact analysis
Ivan Rodrigues – Side impact analysis
Karthik Ramagiri – Front impact analysis
Kaushik Munusamy Kabilan – Modal analysis

MAE 598- 2016

Final Report

ABSTRACT
Multi tubular space frames, often referred to as roll-cage acts as a structural embody for various
types of automotive vehicles. The frame has to conform to the rules specified by SAE. The frame is
subjected to the various dynamic and static loads encountered during front, side, roll over, torsional
impact. The frame should be stiff enough to react against all the loads acting on it with apt strength
to weight ratio. To meet these criterions it is important to consider various parameters involved in
the design of a roll-cage, right from the material to be used up to the forces and impacts that it
might encounter. Through this study, we aimed to design, analyse and optimize a roll-cage so as to
achieve the target of apt strength to weight ratio. Factor of safety of 2 was set which gave the
conservative design. CAD Modelling Software packages and Optimization package in ANSYS
Workbench 16.0 was used for the study. The geometric characteristics of inner and outer radius
were set as parameters. Initially, static structural analysis was performed to check for the members
under high stresses. The model was first optimized for front impact due to its severity in the ATV
operation. The front optimized model was updated in the geometry and then the side impact
optimization study was performed for system integration. The behaviour of combined design
variables for front and side was studied to check for weight saving. Roll over optimization study was
performed on the side optimized geometry. To account for the couple acting on the frame while
negotiating bumps, the model was optimized for torsional loads. Model was then optimized for
modal analysis such that the natural frequencies fall out of the engine idling frequency of 28 Hz. The
iterative process of refinement of candidate points and updating the geometry were carried out in
all the optimizations. Flowchart of the loops and the model development was followed for all the
iterations. The above designed chassis is stiffer and stronger than the preliminary design. The final
optimized mass after the system was 60.789 kg and maximum combined stress was 227.01 MPa. The
trade-off analysis combined with topology optimization study can give us the better approximations
of the possible optimized weight. The results hence concluded were based on conservative design
approach. If the stress constraints are relaxed, further reduction in mass can be achieved.

1
 Table of contents

1 Abstract 1
1.1 Rules and constraints 5
1.2 Optimization formulation 6
1.3 Assumptions 6
1.4 Meshing and optimization methods 7
1.5 Model 9
2 The Frontal impact Analysis 11
2.1 Loading Conditions 11
2.2 Static Structural Analysis 12
2.3 Optimization Results 13
3 The Side Impact Analysis 15
3.1 Model 15
3.2 Side Impact Loading Conditions 15
3.3 Optimization 16
4 The Roll-over Impact 19
4.1 Model 19
4.2 Roll-Over Impact Loading Conditions 19
4.3 Optimization 21
5 The Modal Analysis 23
5.1 Mathematical Model 24
5.2 Boundary Conditions 25
5.3 Results: 25
6 Torsional Loading Analysis 29
6.1 Validation 29
6.2 Analytical solution 30
6.3 Loading and Boundary Condition: 30
7 System Integration Study 32
7.1 Combined Loading Condition 33
7.2 Optimization results 36
7.3 Validation 38
8 Conclusion 39
9 Reference 40

2
 List of figures

Fig 1.1: Meshed Model

Fig 1.2: CATIA model
Fig 1.3: Cross section having dimensions Ro and Ri
Fig 1.4: ANSYS Design Modeller
Fig 2.1: Loading and Boundary Conditions for the frontal impact Loading
Fig 2.2: Stress induced in the frame in frontal impact loading
Fig 2.3: Plot of Candidate points and the corresponding parameter values
Fig 2.4: Mass Convergence
Fig 2.5: Stress Convergence
Fig 3.1: Loading and Boundary Condition for Side Impact Loading
Fig 3.2: Initial stresses induced in the chassis due to side impact loading
Fig 3.3: Convergence of the mass of the chassis
Fig 3.4: Convergence of the Maximum combined stress in the chassis
Fig 3.5: Plot of the parameters of the study corresponding to the two candidates
Fig 3.6: Side Impact Loading for the optimized model
Fig 4.1: Loading and Boundary condition for Roll-over impact loading
Fig 4.2: Initial stresses induced in the chassis due to roll-over impact loading
Fig 4.3: Plot of the three candidate points with their corresponding parameters
Fig 4.4: Mass Convergence
Fig 4.5: Stress Convergence
Fig 5.1: Boundary condition
Fig 5.2: List of frequencies w.r.t mode shapes
Fig 5.3: First mode shape of vibration
Fig 5.4: Second mode shape of vibration
Fig 5.5: Third mode shape of vibration
Fig 5.6: Fourth mode shape of vibration
Fig 5.7: Fifth mode shape of vibration
Fig 5.8: Sixth mode shape of vibration
Fig 6.1: Torsional Loading and Boundary Condition
Fig 6.2: Maximum Combined Stresses in the chassis under torsional loading
Fig 6.3: Mass Convergence
Fig 7.1: Combined Loading and Boundary conditions
Fig 7.2: Stresses induced in the frame due to the combined loading
Fig 7.3: Initial Design (CATIA)
Fig 7.4: Initial Design (ANSYS)
Fig 7.5: Flowchart showing scheme of iterative loop
Fig 7.6: The eight common design variables
Fig 7.7: Mass Convergence
Fig 7.8: Bending stiffness vs thickness graph

3
 List of Tables

Table 1.1: Material Properties

Table 2.1: Optimized cross sectional dimensions
Table 3.1: Comparison between Initial model and Optimized model
Table 4.1: Optimized cross sectional dimensions
Table:6.1 Torsional analysis candidate points
Table 7.1: Loop candidate points front
Table 7.2: Loop candidate points side

4
1.1 Rules and constraints:

According to SAE International 2016 Collegiate Design Series, Baja SAE® rules, under
section B8.3.12 Roll Cage and Bracing Materials “(B) A steel shape with bending
stiffness and bending strength exceeding that of circular steel tubing with an outside
diameter of 25mm (1 in.) and a wall thickness of 3 mm (0.120 in.).The wall thickness
must be at least 1.57 mm (0.062 in.) and the carbon content must be at least 0.18%,
regardless of material or section size. The bending stiffness and bending strength must
be calculated about a neutral axis that gives the minimum values.”

Bending stiffness = 𝐸 ∗ 𝐼

Where,

E = Modulus of elasticity

I = Second moment of area for the structural cross section

Bending strength is given as follows:

𝑆𝑦 ∗ 𝐼
𝑀=
𝑐

Where,

Sy = Yield strength

c = Distance from neutral axis to extreme fiber.

5
1.2 Optimization Formulation:

Objective : Minimize Mass

𝑀𝑎𝑠𝑠 = 𝛴𝑓(𝑅𝑜𝑛 , 𝑅𝑖𝑛 )

𝑅𝑜𝑛 : 𝑂𝑢𝑡𝑒𝑟 𝑅𝑎𝑑𝑖𝑢𝑠 𝑜𝑓 𝑡ℎ𝑒 𝑛𝑡ℎ 𝑚𝑒𝑚𝑏𝑒𝑟

𝑅𝑖𝑛 : 𝐼𝑛𝑛𝑒𝑟 𝑅𝑎𝑑𝑖𝑢𝑠 𝑜𝑓 𝑡ℎ𝑒 𝑛𝑡ℎ 𝑚𝑒𝑚𝑏𝑒𝑟

Subject To :

i) Yield Strength= 460 MPa

ii) Factor of Safety >= 2

iii) Inner radius (Ri)

Lower Bound = 7 mm

Upper Bound = 12 mm

iv) Outer radius (Ro)

Lower Bound = 12.7 mm

Upper Bound = 16.7 mm

1.3 Assumptions:

1) The material considered is AISI 4130 chromoly steel with yield strength of 460
MPa.

2) Tubes in same plane have same cross section considering fabrication and
assembly feasibility

3) Cross members were placed from the available reference model

4) Tubes symmetric about XZ plane have same cross section

5) Gross Vehicle Mass was assumed to be 280 kg

6) Since the front impact is instantaneous, the time of impact was assumed to be
0.15s

6
1.4 Meshing and Optimization Methods

Fig 1.1: Meshed Model

Hex Dominant mesh was used using BEAM189 elements.

BEAM189 3-node 3-D beam

The BEAM189 element is suitable for analyzing slender to moderately stubby/thick

beam structures. The element is based on Timoshenko beam theory which includes
shear-deformation effects. The element provides options for unrestrained warping
and restrained warping of cross-sections.

The element is a quadratic three-node beam element in 3-D. With default settings, six
degrees of freedom occur at each node; these include translations in the x, y, and z
directions and rotations about the x, y, and z directions. An optional seventh degree
of freedom (warping magnitude) is available. The element is well-suited for linear,
large rotation, and/or large-strain nonlinear applications.

Elasticity, plasticity, creep and other nonlinear material models are supported. A cross-
section associated with this element type can be a built-up section referencing more
than one material. Added mass, hydrodynamic added mass and loading, and
buoyant loading are available.

7
Design of Experiment Methodology

The Latin Hypercube sampling method was used to create the design points in the
DOE.

Latin hypercube sampling (LHS) is a statistical method for generating a sample of

plausible collections of parameter values from a multidimensional distribution. The
sampling method is often used to construct computer experiments.
In the context of statistical sampling, a square grid containing sample positions is
a Latin square if (and only if) there is only one sample in each row and each column.
A Latin hypercube is the generalization of this concept to an arbitrary number of
dimensions, whereby each sample is the only one in each axis
aligned hyperplane containing it.

When sampling a function of N variables, the range of each variable is divided

into M equally probable intervals. M sample points are then placed to satisfy the Latin
hypercube requirements; note that this forces the number of divisions, M, to be equal
for each variable. Also note that this sampling scheme does not require more samples
for more dimensions (variables); this independence is one of the main advantages of
this sampling scheme. Another advantage is that random samples can be taken one
at a time, remembering which samples were taken so far. [5]

Response Surface Optimization Methodology

Since the objective function of mass was continuously differentiable the NLPQL
methodology was utilized to perform the response surface opitmization.

Non-Linear Programming by Quadratic Lagrangian (NLPQL) is a sequential quadratic

programming (SQP) method which solves problems with smooth continuously
differentiable objective function and constraints. The algorithm uses a quadratic
approximation of the Lagrangian function and a linearization of the constraints. To
generate a search direction a quadratic sub-problem is formulated and solved. The
line search can be performed with respect to two alternative merit functions, and the
Hessian approximation is updated by a modified BFGS formula. [6]

8
1.5 Model
Modelling of the SAE BAJA Chassis

The Chassis was initially modelled in CATIA as a surface model considering the
minimum design requirements following the SAE BAJA rules.

Fig 1.2: CATIA model

Fig 1.3: Cross section having dimensions Ro and Ri

9
 Fig 1.4: ANSYS Design Modeller

Since we couldn’t define the parameters in the model designed in CATIA, we had to
do the modelling again in ANSYS. The points of intersection of members were
extracted into a point file. This point file was imported into ANSYS Design Modeler and
the points were connected by lines to generate line bodies. Each line body is assigned
with a cross section and the Outer radius and Inner radius were given the initial values
of 12.7mm and 11.13mm respectively. The Outer radius and Inner radius were made
as parameters in each cross section. Finally all the line bodies are made into a single
part. The line bodies act as beams.

Material properties

From a large selection of alloys of steel, AISI 4130 Chromoly steel was chosen to be the
suitable material because of its less weight per meter length when compared to AISI
1018.

AISI 4130 Chromoloy steel was created as a new material in the Engineering data in
ANSYS. Isotropic Elasticity property was used to define the properties of the material
such as Young’s modulus, poisson’s ratio, bulk and shear modulus. The material
properties assigned are as follows:-

Density g/cm^3 Young's Modulus MPa Poisson's Ratio Bulk Modulus MPa Shear Modulus MPa
7.85 2.e+005 0.285 1.5504e+005 77821
Table 1.1: Material Properties

10
2. The Frontal impact Analysis:

The frame is made of steel members. The steel members are usually connected in such
a way that the members undergo pure compression or tension. The integrity of the
frame is such that on impact the roll cage is prevented from failure. The frame forms
the major part of the cockpit. The maximum impact force that a car experiences can
be that from frontal impact. Hence the cockpit is structurally an important part for the
driver.

The Load required for frontal impact is obtained by creating a scenario where the car
is moving at a top speed of 65 kmph undergoing a head on collision with rigid body.
The mass of the car including the driver is assumed to be 280kg. The various loads are
calculated using basic mechanics

Calculations

Gross Mass of the car, 𝑚 = 280 𝑘𝑔

m
Velocity of the car, v = 65 kmph = 18 s
(𝑀𝑎𝑥 𝑉𝑒𝑙𝑜𝑐𝑖𝑡𝑦)

Time of impact, 𝑡 = 0.15 𝑠 (𝑇𝑖𝑚𝑒 𝑓𝑟𝑜𝑚 𝑡𝑜𝑝 𝑠𝑝𝑒𝑒𝑑 𝑡𝑜 𝑓𝑢𝑙𝑙 𝑠𝑡𝑜𝑝)

𝑣𝑓 −𝑣𝑖 0−18
Acceleration, 𝑎= = = 120 𝑚/𝑠 2
𝑡 0.15

Force, 𝐹 = 𝑚 ∗ 𝑎 = 280 ∗ 120 = 33.33𝑘𝑁

The above calculated force is approximately equal to a force of 12g. The

deceleration is taken as a worst case scenario for human-body. This complies with the
maximum limit of deceleration a human body can withstand before passing out i.e
9g.

2.1 Loading Conditions

The geometry of the frame is modelled on ANSYS design modeller.

A uniformly distributed load of 33.33 kN in the positive x direction is applied uniformly

on the 7 members in the front most y-z plane of the frame. The displacement of 4
points on the fire-wall, right behind the driver is constrained in all degrees of freedom:
Translation in X-Y-Z and rotation about X-Y-Z.

11
 Fig 2.1: Loading and Boundary Conditions for the frontal impact Loading

2.2 Static Structural Analysis:

The model is made with minimum requirements in accordance with SAE rule book i.e.
1 inch outer diameter and 1.57mm thickness. On applying the load to the model a
good 10 members have stress value more than 230Mpa. Inner Radius and Outer
Radius of these members were selected as parameters. A total of 20 parameters were
used for Design of experiments. A Method of Latin Hyper Cube Sampling was selected
for creating the design points. A total of 534 Design points were created for Design
Optimization.

Fig 2.2: Stress induced in the frame in frontal impact loading

12
2.3 Optimization Results:

NLPQL optimization method is followed for optimization. The following were the
candidate points chosen and the corresponding mass and stress are 53.88 kg and
188.98 MPa respectively provided by the response surface:

Fig 2.3: Plot of Candidate points and the corresponding parameter values

Parameter Ri, Ro - CircularTube# Ri, mm Ro, mm

P54, P55 - CircularTube1 9.903788263 14.30017032
P56, P57 - CircularTube2 9.90588924 14.29800747
P58, P59 - CircularTube3 9.825920845 14.37716541
P64, P65 - CircularTube6 9.669965749 14.53161714
P66, P67 - CircularTube7 9.670029443 14.53164087
P69, P68 - CircularTube8 9.666101054 14.53578958
P70, P71 - CircularTube9 9.683794126 14.51800322
P76, P77 - CircularTube12 9.705170176 14.49687747
P94, P95 - CircularTube21 9.714685724 14.48744424
P102, P103 - CircularTube25 9.633941816 14.56737746

Mass, kg 53.88002227
Maximum Stress, MPa 188.9801829
Table 2.1: Optimized cross sectional dimensions

The above values of inner and outer radius are the candidate points. As seen from the
graph the value lies between the bounds and show that the bounds chosen are in
fact valid.

The graphs below show the convergence of mass and stress. It can be seen from the
graph that minimum mass is 53.88 kg and the corresponding maximum stress is 188.98
MPa. The algorithm further tries to reduce the mass by converging the solution
towards a maximum stress of 230 MPa. This candidate point is further updated on 5

13
refinement loops. When the design points at the end of the refinement loops are
updated in static structural analysis, the maximum stress obtained is 257MPa. The error
obtained is a result of limited number of design points updated in Design of
Experiments. Further with more design points for the 20 parameters a solution closer to
the optimum value can be obtained. Hence the above solution of 53.88 kg is the
optimum mass.

Fig 2.4: Mass Convergence

Fig 2.5: Stress Convergence

14
3. The Side Impact Analysis

The side impact loading analysis of the chassis of the Baja car is carried out to ensure
the safety of the driver for any collisions on the side with another vehicle. In the side
impact test the chassis is aligned sideways relative to the oncoming vehicle. In reality
though, some of the impact energy will be taken up by the suspensions and the
wheels, but we do not consider that in this simulation to make it a worst case scenario
of the impact.

3.1 Model

The optimized model of the chassis for the front impact loading condition is imported
to be analysed for the side impact loading condition.

3.2 Side Impact Loading Conditions

The load required for the side impact is obtained by creating a scenario where one
car is moving towards the other’s side at a top speed of 65 kmph. The mass of the car
including the driver is assumed to be 280kg. The loads to be applied to the chassis are
calculated as follows [3]:

Calculations:

Gross Mass of the car, 𝑚 = 280 𝑘𝑔

m
Velocity of the car, v = 65 kmph ≈ 17.86 s
(𝑀𝑎𝑥 𝑉𝑒𝑙𝑜𝑐𝑖𝑡𝑦)

Time of impact, 𝑡 = 0.30 𝑠 (𝑇𝑖𝑚𝑒 𝑓𝑟𝑜𝑚 𝑡𝑜𝑝 𝑠𝑝𝑒𝑒𝑑 𝑡𝑜 𝑓𝑢𝑙𝑙 𝑠𝑡𝑜𝑝)

𝑣𝑓 −𝑣𝑖 0−17.86
Acceleration, 𝑎= 𝑡
= 0.30
≈ 59.53 𝑚/𝑠 2

Force, 𝐹 = 𝑚 ∗ 𝑎 = 280 ∗ 59.53 ≈ 16.66𝑘𝑁

A uniformly distributed load of 16.66 kN in the positive y-direction is applied on carefully

selected eight members. These members are selected by choosing those members
lying in the x-z plane at the most negative position in the y-direction of the frame.

The 6 points at the opposite extreme (lying in the x-z plane at the most positive position
in the y-direction) of the frame are constrained in all the six degrees of freedom (3
DOF translation and 3 DOF rotation). The loading is done so to consider the worst case
scenario of side impact loading.

15
 Fig 3.1: Loading and Boundary Condition for Side Impact Loading

Application of the calculated load of 16.66kN on the specified members of the

chassis resulted in the following:-

Fig 3.2: Initial stresses induced in the chassis due to side impact loading

It was seen that this loading of the model resulted in 2 members of the frame having
the maximum combined stress value to be greater than the allowable stress limit of
230Mpa. These two members are the firewall and the cross member under the driver
closest to the firewall.

3.3 Optimization

Both the inner and outer radii of the two members were set as parameters resulting in
a total of 4 parameters in this subsystem of the study. The Latin Hyper Cube Sampling
methodology was employed for creating the design points required for the Design of
Experiments study which generated a total of 25 design points corresponding to the
4 parameters within the specified ranges.

16
The determination of the maximum combined stress values and the weight of the
chassis corresponding to each of the 25 design points as part of the DOE was followed
by the generation of the Response Surface to obtain the optimal response.

Non-Linear Programming by Quadratic Lagrangian (NLPQL) was the optimization

method followed for the optimization study. The allowable convergence limit was set
at 0.0001% and the maximum number of iterations was set at 40. The objective
function minimization of the mass along with the allowable stress constraint was input
and also the parameter relationship specifying the minimum thickness of each
member.

Fig 3.3: Convergence of the mass of the chassis

Fig 3.4: Convergence of the Maximum combined stress in the chassis

The two plots above show the convergence of the mass and the maximum combined
stress values as the number of iterations are increased, where the convergence is

17
achieved at the 21st iteration requiring 92 evaluations generating two candidate
points.

Two candidate points were generated:-

Fig 3.5: Plot of the parameters of the study corresponding to the two candidates

The best candidate point chosen resulted in the optimized values as follows:-

CT_4-Ro CT_4-Ri CT_22-Ri CT_22-Ro Mass Stress

Initial 12.7 11.13 11.13 12.7 53.88 555.68

Optimized 15.05 12 11.447 13.017 58.048 227.76

Table 3.1: Comparison between Initial model and Optimized model

Fig 3.6: Side Impact Loading for the optimized model

18
The solution from the response surface for the above values of input cross sectional
dimensions was a weight of 58.048kg and a maximum combined stress of 230 MPa,
which has a 0.98% error from the verified value for the same input cross sectional
dimensions. Further iterative refinement of the response surface will reduce the error
closer to zero resulting in an accurate response surface approximation and
correspondingly optimized cross sectional dimensions for the side impact loading.

4. The Roll-over Impact

The SAE Baja car is designed for off-road rough terrain. The chances of the car rolling
over is high when it encounters hills or valleys. Therefore the car has to be designed
taking care of the safety of the driver, addressing all the possible situations of danger.
The Roll-over Impact analysis is performed to design the roll-cage for maximum safety
during the roll-over incidents.

4.1 Model

The optimized parameters obtained from front impact and side impact analysis are
used as inputs for the roll-over impact loading analysis.

4.2 Roll-Over Impact Loading Conditions

Fig 4.1: Loading and Boundary condition for Roll-over impact loading

19
The load required for the roll-over impact loading analysis is obtained by using 25% of
the load used for the frontal impact analysis as follows[2]:

Calculations:

Force, 𝐹 = 0.25 ∗ 𝐹𝑟𝑜𝑛𝑡𝑎𝑙 𝑖𝑚𝑝𝑎𝑐𝑡 𝑙𝑜𝑎𝑑 = 0.25 ∗ 16.66𝑘𝑁 ≈ 8.33kN

The uniformly distributed load of 8.33 kN is applied in the negative Z-direction uniformly
on the top 4 members of the frame.

The frame is constrained along the 4 vertices at the bottom on the XY-plane in all the
six degrees of freedom (3 DOF translation and 3 DOF rotation) by the use of fixed
supports. This setup virtually simulates the roll-over impact loading condition
encountered in a real world situation on the track.

Applying the calculated load of 8.33kN on the specified members of the chassis
resulted in the following:-

Fig 4.2: Initial stresses induced in the chassis due to roll-over impact loading

It was found that the stresses induced in the two roll-hoops was 237.62 MPa which was
higher than the allowable stress limit. The chassis therefore had to be optimized to
have stresses within this loading condition. The inner and outer radii of the roll-hoops
were selected as the parameters for the optimization study.

20
4.3 Optimization

The two parameters selected above were used for the Design of experiments. The
Latin Hyper Cube Sampling method was selected for creating the design points,
which resulted in 10 Design points being created during the DOE process.

The Response Surface to obtain the optimal response was created based on the input
from the DOE using the Standard response surface- 2nd order polynomial type method.

A response point is chosen by default by the software. This response point is used as
the starting point for the optimization process. The optimization setup of the roll-over
impact analysis is similar to front and side impact analysis, where minimization of the
mass is the objective of the study and stress limit is the constraint. The parameter
relationship governing the minimum thickness cross section criteria was applied during
the response surface optimization step. The Nonlinear Programming by Quadratic
Lagrangian (NLPQL) method with allowable convergence of 0.0001% and a
maximum of 40 iterations is used for the optimization process.

This response surface Optimization process converged in 3 iterations correspondingly

employing 8 evaluations providing the converged results of the mass of the chassis
and the maximum combined stress induced in the chassis.

The optimization study generated 3 candidate points, out of which the best
candidate point was selected. Comparing the Maximum combined stress values and
mass of the chassis of this candidate point from the Response Surface and that from
Static Structural analysis gave an error of 22.71% which suggested more scope of
optimization towards getting better results. Therefore, the best candidate point from
the obtained result was chosen again as a refinement point to get a better
approximation of the response surface for the current study.

The three candidate points obtained after refinement of the response surface are
explained below:-

21
 Fig 4.3: Plot of the three candidate points with their corresponding parameters

The above graph shows placement of the candidate points between the specified
bounds of variation. We see that all the candidate points are within the bounds and
are giving valid results of optimization.

The following was the candidate points chosen:-

Parameter Ri, Ro - Mass(kg) Max Combined

CircularTube# Ri, mm Ro, mm Stress(MPa)
P84 – P85 11.3208 12.8908 58.101 228.37
CircularTube16_Plane
Table 4.1: Optimized cross sectional dimensions

Fig 4.4: Mass Convergence

22
The above graph represents the mass convergence with increasing number of
iterations of optimization study. The mass converged to 58.101 kg. We see that the
mass is minimized by the decreasing value of mass with the increase in number of
iterations.

Fig 4.5: Stress Convergence

The above graph shows the stress convergence with the increasing number of
iterations of the optimization study. We see that the stress converges to 228.37 MPa
for the optimized value of mass.

5. The Modal Analysis

Our objective here was to build a dynamically stable chassis to withstand all kind of
terrain during its mobility. The BAJA car, being an off-road racing vehicle
experiences severe uneven loading. We had to optimize the frame so that it can
withstand all static and dynamic loads as well as being light in weight.

When the natural frequency of vibration a frame equals the excitation frequency of
forced vibration, there occurs a phenomena of resonance which causes severe
deflections to the structure. These excessive vibrations and resonance causes lot of
failure to the structure/frame due to the harsh conditions in which the vehicle is
driven. Here, we are finding the natural frequency of the frame under its self-weight.
This frequency should be well above the range of any excitation frequency caused
by external factors. Generally, the natural frequency of the model can be
calculated using the following equation:

23
 𝑘
𝑊𝑛 = √( )
𝑚

where k is the mass and m stands for stiffness.

The major contributing factor for the forced vibration is the engine, which is mounted
on the back of the vehicle. The vibrating frequency of the engine is within the range
of 15Hz-25Hz as in most cases of a single cylinder 310cc, 10HP engine. Therefore, we
have to model the chassis such that its natural frequency at various mode shapes is
well above this range. By obtaining frequencies as mentioned above we will be
avoiding resonance and thus creating a dynamically stable model.

5.1 Mathematical Model

The equations of motion of a multi DOF system is generally automatically created by

FEM software that we are using here, but for analytical purposes it can be derived
using the Lagrange’s equation as shown below [1]:

𝑑 𝜕𝑇 𝜕𝑇 𝜕𝑉
( )− + = 𝑄𝑖
𝑑𝑡 𝜕𝑞𝑖̇ 𝜕𝑞𝑖 𝜕𝑞𝑖

where T is the Kinetic energy, V is the potential energy and Q is the generalized
degrees of freedom.

In the modal analysis, the dissipative (non-conservative) forces are zero. The system is
taken to be conservative and on the RHS of the equation we have vector of zeros as
shown below:

[𝑚]. 𝑞̈⃗ + [𝑘]. 𝑞⃗ = 0

⃗⃗

The above equation describes the undamped free vibration of the chassis that we
are considering. The characteristic determinant of the corresponding eigenvalue
problem is given by Equation:

∆= [𝛼 [𝐼] − [𝐷]] = 0

Here, [D] is the dynamic matrix of the system, 𝛼 is the eigenvalue vector and the
identity matrix. Once the eigenvalues are known, the mode shapes or eigenvectors
for each mode “i” can be calculated using the equation below:

⃗⃗𝑖 = 0
[𝛼𝑖 [𝐼] − [𝐷]]. 𝑄 ⃗⃗

24
5.2 Boundary Conditions:

The chassis frame is fixed at the suspension points in order to know about the various
mode shapes of the upper body structure. Fixed Supports are given at suspension
points since the wheels and suspension are mounted to the axle thereby restricting
the DOF of lower base to zero.

Fig 5.1: Boundary condition

5.3 Results:

Fig 5.2: List of frequencies w.r.t mode shapes

25
The mode shapes that were obtained from the analysis are as follows:

Fig 5.3: First mode shape of vibration

The First mode shape explains about the longitudinal vibrations caused on the upper
part of the structure with a natural frequency of 31.819 Hz which is lowest natural
frequency.

Fig 5.4: Second mode shape of vibration

In this mode shape also the upper part of the frame experiences longitudinal
vibrations with a frequency of 37.087 Hz.

26
 Fig 5.5: Third mode shape of vibration

In this mode shape, the rear part of the chassis vibrates with a natural frequency of
41.304 Hz.

Fig 5.6: Fourth mode shape of vibration

In this mode shape, the upper frame, firewall and the beams connecting the rear end
vibrate with a frequency of 54.038 Hz

27
 Fig 5.7: Fifth mode shape of vibration

In this mode shape the entire upper frame and beams connecting them vibrate with
a natural frequency of 67.071 Hz.

Fig 5.8: Sixth mode shape of vibration

In the sixth mode shape the frontal structure vibrates with a frequency of 89.199 Hz
and this is the highest frequency experienced by this model under self-weight and
gravity.

28
The frequency modes that determine its dynamic characteristics are in the range of
31Hz- 89Hz and less than 100Hz frequency mark.

Once the accessories like seats, engine, suspensions and steering systems are added
to the vehicle the mass of the vehicle increases thereby increasing the natural
frequency. Thus we find that the natural frequency range is well above the excitation
frequency range.

This implies that the model is completely stable and safe for uneven loading
conditions.

6. Torsional Loading Analysis:

The frame should be stiff enough to sustain dynamic suspension loads. When the
vehicle is negotiating the bump there might be a case of alternating bumps on left
and right wheels. Considering the left wheel is having the upward travel (jounce) and
the right wheel is having the downward travel (rebound) the spring forces will act in
the opposite direction composing a couple on front of the vehicle. This couple tries to
produce the torsional stress in the frame. For the worst case scenario the diagonally
opposite wheels are having the opposite wheel travel i.e. front right wheel is having
the vertically upward travel and at same time rear left wheel is having the vertically
downward travel producing a couple diagonally. This couple is responsible for the
torsional stresses in the vehicle.

Torsional Stiffness of SAE Baja car has to be greater than 3.5 kNm/degree. Hence the
optimization problem becomes:

OPTIMIZATION FORMULATION:

Objective : Minimize Mass

Mass = f (Ro, Ri)
Subject To :
i) Torsional Stiffness > 3.5 KNm/degree (SAE Requirement)

6.1 VALIDATION:

To check whether the optimize design satisfies the torsional strength requirement,
analytical study was performed. The maximum deformation was at the rear
suspension mount which is correct on the lines of weight distribution.

29
6.2 ANALYTICAL SOLUTION:

F = 3333 N
L = Distance between diagonally opposite suspension mounts=490mm
D = Vertical deformation in suspension mounts
Θ = Angular deformation
Tan(θ)= D/(L/2)
Torsional Stiffness = (F x L) / θ.
D=1.252 mm

Torsional Stiffness= 4.8 KNm/degree

Thus optimized design satisfies the torsional stiffness requirement.

6.3 Loading and Boundary Condition:

Fig 6.1: Torsional Loading and Boundary Condition

Front suspension point Force along positive Z-direction = 2.4 KN

Force at the diagonally opposite suspension point at the Rear along the positive Z-
direction = 3.3 KN
The Diagonally Opposite Suspension Mountings are fixed in all six DOF.

30
 Fig 6.2: Maximum Combined Stresses in the chassis under torsional loading

The design points from optimized Roll over impact were updated in the model for
torsional impact. The static structural analysis was run considering the couples. The
analysis showed that the longitudinal members below the driver seat were under high
stresses. The cross sections of these members were set as design variables. The
optimization study was performed.

Fig 6.3: Mass Convergence

31
 Initial Initial
Parameter Ri, Ro – CircularTube# Ri,mm R0,mm Ri, mm Ro, mm
P1, P2 – CircularTube13 11.13 12.7 12 14.835
P12, P13 – CircularTube5 11.13 12.7 12 15.351
P26, P27 – CircularTube11 11.13 12.7 12 14.555
Table:6.1 Torsional analysis candidate points

The final optimized mass after torsional impact was 60.879 kgs and maximum stress
was 227.01 Mpa.

7. System Integration Study

An All-terrain vehicle is subjected to various dynamic and static loading such as Front,
side and roll over impact along with various mode shapes. After the individual
subsystems are optimized then it is necessary to integrate them together and come
up with an overall system optimum. Since using All-in-one approach was too
complicated given the large number of design variables, decomposition method was
used. Local variables for individual subsystem were identified. This was followed by
definition of the master optimization formulation which accounted for common
design variables among the subsystems. The iterative loop was then performed
wherein the master problem was solved with respect to common design variables
(individual local variables fixed) and the subproblems were solved with respect to the
local variables (common variables fixed). Trade off analysis between the Maximum
stress and the mass of the chassis was performed.

OVERALL SYSTEM OPTIMIZATION FORMULATION:

Objective : Minimize Overall Mass
Mass = f (Ro, Ri)
Subject To :
Combined Loading: Front + Side + Roll Over + Torsional
i) Yield Strength= 460 MPa
ii) Factor of Safety >= 2
iii) Inner radius (Ri)
Lower Bound = 7 mm
Upper Bound = 12 mm Bounds for “8”Combined
iv) Outer radius (Ro) Design Variables
Lower Bound = 12.7 mm
Upper Bound = 16.7 mm

32
Since the weight is the function of the geometry, it depends on all the design variables
combined from individual subsystems. We have 54 tubular members, taking both the
inner and outer radii of each member gives us a total of 108 design variables. Due to
limitation on the available optimization packages concerning the maximum number
of parameters that can be used for the DOE (20 for Design of Engineering Experiments
for Latin Hypercube Sampling), we had to make certain assumptions.
Keeping in mind the objective of each subsystem, the following assumptions were
made:
7.1 Combined Loading Condition
To start with system optimization we analyzed the frame subjected to combined
loading.

Fig 7.1: Combined Loading and Boundary conditions

This combined loading resulted in the following stresses being generated in the
chassis.

Fig 7.2: Stresses induced in the frame due to the combined loading

33
 Fig 7.3: Initial Design (CATIA)

Fig 7.4: Initial Design (ANSYS)

The analysis helped us in identifying the members which were under high stress. Hence
we decided the common design variables for overall system optimization. The lateral
member below rack mounting, firewall members, front roll over hoop were the
common members. This gave 8 common design variables. The flow chart to carry out
the optimization loop is as below:

34
 CAD Model Development

Static Structural Analysis

Front

Side

Roll Over

Modal

Torsional

Fig 7.5: Flowchart showing scheme of iterative loop

Loop 2:
Front impact analysis was performed and the design was optimized satisfying all the
constraints. Then the optimized values from the front impact were updated on the
geometry. These updated model was used for side impact optimization. As the lateral
members for rack mounting were common members, they were used as combined
design variables (shown in yellow) for front and side impact. Now, the optimization
loop was run and design was optimized for side impact. The common design variables
were updated to the front optimized design to check for the trade-off.

35
 Parameter Ri, Ro - Initial Initial
CircularTube# Ri,mm R0,mm Ri, mm Ro, mm Loop 1 Ri Loop 1 R0
P54, P55 - CircularTube1 11.13 12.7 9.903788 14.30017 9.99212 14.20847
P56, P57 - CircularTube2 11.13 12.7 9.905889 14.29801 9.78244 14.40183
P58, P59 - CircularTube3 11.13 12.7 9.825921 14.37717 9.80074 14.38194
P64, P65 - CircularTube6 11.13 12.7 9.669966 14.53162 9.67211 14.53275
P66, P67 - CircularTube7 11.13 12.7 9.670029 14.53164 9.69047 14.53275
P69, P68 - CircularTube8 11.13 12.7 9.666101 14.53579 9.77251 14.45948
P70, P71 - CircularTube9 11.13 12.7 9.683794 14.518 9.68453 14.43281
P76, P77 - CircularTube12 11.13 12.7 9.70517 14.49688 9.74652 14.48741
P94, P95 - CircularTube21 11.13 12.7 9.714686 14.48744 9.73145 14.47652
P102, P103 - CircularTube25 11.13 12.7 9.633942 14.56738 9.64513 14.39474

Table 7.1: Loop candidate points front

As it is clear from the table, the cross sections of the members 8, 9, 12, 25 which are
connecting to the common variables would change. Circular tube 2 and 3 are
common design variables. Thus the corresponding mass would also change. After
updating the design variable, the optimized mass for the front design is reduced by
1.237 kgs, whereas the stress value has increased by 14 Mpa. This is a trade-off.

7.2 Optimization results:

Fig 7.6: The eight common design variables

36
 Fig 7.7: Mass Convergence

Now the loop is continued for the roll over impact. The optimized points from side
impact were updated in the model. The firewall members were common design
variables between side and roll over impact. The optimization run was carried out for
roll over impact. As it is clear from the table, the cross sections of the members 16, 17
which are connecting to the common variables would change. Circular tube 4 are
common design variables. Thus the corresponding mass would also change. Now the
mass reduced to 57.24 kg but the stress increases. The following results were obtained
for the change in design variables. The optimized mass from side impact was initially
58.09 before iteration.

Parameter Ri,Ro - Initial Initial

CircularTube# Ri,mm R0,mm Ri, mm Ro, mm Loop 1 Ri Loop 1 R0
P60, P61 - CircularTube4 11.13 12.7 12 15.05 11.34 14.96
P56, P57 - CircularTube2 9.9058 14.29801 9.78244 14.40183 9.8245 14.0214
P58, P59 - CircularTube3 9.8259 14.37717 9.80074 14.38194 9.928 14.128
P84, P85 - CircularTube16 11.13 12.7 11.13 12.7 11.28 12.66
P86, P87 - CircularTube17 11.13 12.7 11.13 12.7 11.28 12.66
P96, P97 - CircularTube22 11.13 12.7 11.447 13.017 11.48 12.94

Table 7.2: Loop candidate points side

To continue the iteration, the geometry was updated after roll over optimization.
Since there were no common design variables for modal and torsional impact, they
were optimized independently. There is trade off analysis between mass and stress

37
which needs to be performed before finalizing on any of the optimized variables. Final
optimized mass of the vehicle 60.789 and the stress 227.01 Mpa.

7.3 Validation:

To check if the selected cross sections meets the bending strength and stiffness
criterion.
The graphs of bending strength and bending stiffness were plotted vs wall thickness
for the selected candidate points. Candidate point corresponding to parameter 57
was chosen, since it is the common design variable for front and side impact. This
candidate point plays important role in trade off analysis. (Figure 1)

Candidate ‘57’

Fig 7.8: Bending stiffness vs thickness graph

Bending Strength =S*I/c

Bending Stiffness = E*I

I = Second Moment of Inertia = π * (Ro4 – Ri4)/4

From the graph we can see that the optimized any random selected candidate
points meets the stiffness and the strength requirements.
Note:
In Loop 2, combined design variables for front, side and roll-over impacts were
considered. The above optimum solutions were obtained at the end of loop 2. Due to
computational limitations we limited the number of iterations to 2.

38
8. Conclusion

The designed chassis is analyzed for front, side, roll-over and torsion impacts. The
optimized parameters from each analysis is carried over to the next analysis following
the order of front, side, roll-over and torsion impact respectively. Since the optimized
parameters are integrated at each subsystem, the whole chassis is optimized at the
completion of the chain as it reaches the torsion analysis. The combined variable play
a vital role in the optimization study to meet the objective of minimized weight. We
see that the optimized weight of the chassis at the end of first iteration is found to be
60.789kg and the corresponding stress was found to be 227.01Mpa. The optimized
values of parameters obtained at the end of the first iteration of the system integration
loop all fall between the specified ranges. The stress obtained at the end of first
iteration is 227.01Mpa which is less than 230Mpa which was set as the constraint on
the Maximum combined stress, from which we can say that the model obtained after
torsion is safe. Since the stress at this point is 227.01Mpa, anymore decrease in weight
will lead to increase in stress which will then not meet the constraint of 230Mpa of
maximum combined stress. Therefore we can say that the model obtained after
torsion is both optimized for weight and is considered safe with a factor of safety of 2.
However on relaxing the factor of safety, a chassis with much lesser weight can be
obtained, but doing this will only compromise the safety of the driver which is of the
prime importance during the designing of the car. Hence tradeoff analysis combined
with optimization study can give us better approximation of the possible optimized
weight.

39
9. REFERENCES

1. A Vibration Analysis Of Vehicle Frame by Yogendra S. Rajput et.al

2. Static Analysis and Optimisation of Chassis and Suspension of an All-Terrain

Vehicle by Thanneru Raghu Krishna Prasad et.al

3. Stress Analysis of Roll cage for an All Terrain Vehicle by Mahesh .P. Sharma et.al

4. BAJA SAE Series Rules , www.sae.org/students/mbrules.pdf

5. https://en.wikipedia.org/wiki/Latin_hypercube_sampling

6. http://www.ai7.uni-bayreuth.de/nlpqlp.htm

40