Abstract—Vehicle handling and stability plays an important role

in the vehicle design. With the improvement in engine power and better road conditions it influences vehicle safety at high-speeds and also make steering more comfortable. Handling performance evaluation is highly subjective and is usually assessed by various standard test procedures. While field testing is time consuming and costly computer aided simulation is fast, less expensive and is also repeatable. On the other hand, it requires precise model building and experimental validation. The paper focuses on model building of suspension system, hardpoint optimization and experimental correlation of a single-seater electric vehicle Toyota-COMS.

Keywords—Experimental Testing & Correlation, Hardpoint Optimization, K&C (Kinematic and Compliance), Suspension design

Abbreviations— CAD: Computer Aided design, CMA: Caster

moment arm, DOE: Design of experiments, DOF: Degrees of freedom, SAC: Steering axis offset at wheel center, LCA: Lower Control Arm, SSC: Steady state cornering, TB: Twist Beam, F value: Fisher distribution value

I. INTRODUCTION USPENSION K&C characteristics play a decisive role in vehicle ride & handling characteristics. First independent

vehicle suspension has appeared a century ago. Thus, enough research has been done on modeling and simulation of suspension systems[1-5], but there are very few papers that discuss the design optimization of suspension system[6-10]. There are some papers which study the effect of suspension kinematics on the vehicle handling characteristics[11-12]. The efforts to design and optimize suspension for quantitative ride and handling characteristics have to be better channelized.

The first step in designing a suspension, is to determine the mounting positions of various suspension components called “Hardpoints” by kinematic analysis. The design parameters such as camber, caster, toe, king pin inclination, scrub and caster moment arm govern the controllability of a vehicle,

Husain Kanchwala is a Research scholar in Indian Institute of Technology, Kanpur-208016, India (Mobile : +34-631365195; e-mail: husaink@iitk.ac.in)

Wu Nan is a Research scholar in Graduate school of Information, Production and Systems, Waseda University, Kitakyushu, Fukuoka Prefecture, Japan (e-mail: wunan@ruri.waseda.jp).

Harutoshi Ogai is a Professor in Graduate school of Information, Production and Systems, Waseda University, Kitakyushu, Fukuoka Prefecture, Japan (e-mail: ogai@waseda.jp).

while the design parameters like roll center height, roll stiffness, ride camber, ride steer, suspension rate, and percentage anti-dive/ anti-squat are responsible for the stability and ride comfort. Generally, the parameters which influence the ride comfort contradict with the ones which affects handling. Therefore, a trade-off between ride and handling is established.

All of the above discussed parameters are highly dependent on the position of various hardpoints. The conclusion is some hardpoints depends heavily on some suspension characteristics, thus, we need to identify them and look for a set of hardpoints which meet the design criteria.

II. OVERVIEW The vehicle used for this study is been equipped with

MacPherson strut front suspension and twist beam rear suspension. Various hardpoints have been modified in the existing vehicle. The modified vehicle has been tested against various test procedures and the test results have been correlated with that of the mathematical model. For developing the half car mathematical models ADAMS Car and for developing full car model CarSim is been used.

MSC ADAMS[13] is a mechanical system dynamics simulation tool widely used by suspension designers in automotive industry. It is a virtual prototype software which includes various interfaces for modelling, equation solving, optimization, simulation and visualizing aids. It also enables user to import rigid body models from different CAD softwares and flexible bodies from packages like MSC Nastran

CarSim[14] developed by Mechanical Simulation Corp., is used to simulate the vehicle dynamic behavior. It can communicate very easily with softwares such as Matlab and Simulink in order to run a co-simulation. Models typically run more than ten times faster than real time on a 3 GHz PC, enabling real-time (RT) testing with hardware in the loop (HIL) systems.

The K&C characteristics of the optimized model from ADAMS have been exported to CarSim where they are used to build the full car model as it is easy to integrate vehicle S-function block with the controller model developed in Simulink (As the vehicle would be equipped with autonomous driving set-up which is not been discussed in this paper).

Model Building, Hardpoint Optimization & Experimental Correlation of a Single Seater

EV- Toyota COMS Husain Kanchwala, Wu Nan, and Harutoshi Ogai

S

2nd International Conference on Research in Science, Engineering and Technology (ICRSET’2014), March 21-22, 2014 Dubai (UAE)

http://dx.doi.org/10.15242/IIE.E0314507 66

III. PROCEDURE Firstly, for preparing half car models, suspension links are

modeled as rigid bodies, while, twist beam is modeled as a flexible body imported from MSC Nastran. This is done in order to ensure that the compliance characteristics of the mathematical model closely resemble the actual vehicle.

Secondly, design guidelines and acceptance criteria of vehicle suspension K&C characteristics are proposed. Thirdly, key hardpoints which have great influence on K&C characteristics are obtained by performing sensitivity analysis. Hardpoint optimization study is been done on the selected set of hardpoints to meet the desired objective function.

Finally, experimental correlation study is been performed based on certain standard test procedures like ISO lane change, Constant radius cornering, etc. to validate the simulation results. The detailed procedure is shown below.

Fig. 1 Schematic Procedure of design and development

IV. MODEL DEVELOPMENT Accurate hardpoint measurement is extremely essential as

they influence vehicle dynamic characteristics. This data is highly confidential and is generally measured using a Coordinate Measuring Machine (CMM). In our case we lack this facility so manual measurements are made. The vehicle is jacked up and existing hardpoints are measured precisely using vernier caliper, plum bob, angle gauge and shims. Repeated measurements were taken to remove human measurement errors. The adjoining picture shows the vehicle on four jacks and half car front and rear suspension models respectively.

Fig. 2 Vehicle on jacks and half car front & rear Suspension models

For the purpose of having model validation twist beam modal analysis is been done to ensure that the fundamental frequency of the model been developed is far apart the road undulation frequency. Also higher modes have been extracted to verify that they fall well apart from wheel hop frequency. The mode shapes and the frequencies obtained were verified by the design data provided by Toyota.

Fig. 3 Twist Beam Modal Analysis

For performing optimization study a rigid body model of the twist beam suspension with torsional flexibility is been developed to reduce computational time and effort. TB is replaced by two rigid links connected by torsion spring whose stiffness curve is computed from Patran TB model

Fig. 4 Twist Beam Rigid body model with torsion spring

2nd International Conference on Research in Science, Engineering and Technology (ICRSET’2014), March 21-22, 2014 Dubai (UAE)

http://dx.doi.org/10.15242/IIE.E0314507 67

Fig. 5 Torsion Spring stiffness curve obtained from Patran

The front and rear suspension rigid body models were verified using the model verification checklist as shown below.

TABLE I

FRONT AND REAR SUSPENSION MODEL CHECKLIST

V. KINEMATICS AND COMPLIANCE STUDY For obtaining K&C characteristics a parallel and opposite

wheel travel study of 50 mm for both front & rear suspensions is been performed and the model is judged based on the following acceptance criteria.

TABLE II

ACCEPTANCE CRITERIA: FRONT SUSPENSION

No1 Camber 0to-0.5 deg 8 Ride Steer -6to-10deg/m 2 Caster 3 to 5 deg 9 Ride camber -12to-18deg/m

3 Toe 0-0.3 deg Toe-out 10 Roll Centre 110-150 mm

4 KPI 8 to 13 deg 11 Anti-Dive 20 to 35 %5 Scrub -15to30 mm 12 Anti-Lift 7 to 12 %6 CMA 20to30 mm 13 Susp Rate 28-38 N/mm7 SAC < 3 mm 14 Roll stiffness 1.2-2e3 Nm/deg

Criteria ValueNo Criteria Value

TABLE III

ACCEPTANCE CRITERIA: REAR SUSPENSION

1 Camber 0 to-0.5 deg 5 Anti-Dive 20 to 35 %2 Ride Steer -6to-10 deg/m 6 Anti-Lift 7 to 12 %3 Ride camber -12to-18 d/m 7 Susp Rate 30-45 N/mm4 Roll centre 110-150 mm 8 Roll stiffness .5-1.2e3Nm/d

Criteria ValueNo Criteria ValueNo

The hardpoints of the existing vehicle do not satisfy all of the above mentioned criteria. In order to meet these design specifications hardpoint optimization study is been performed using Monte Carlo simulation in ADAMS-Insight which is a DOE and optimization tool.

VI. HARDPOINT OPTIMIZATION In McPherson Strut Front Suspension we had 7 Hardpoints

thus, 21 coordinates (x, y, z coordinates for each hardpoint). It was not possible to alter all the coordinates because of various constraints.

The coordinates chosen for study are: 1. LCA_Rear_z: We have u- hole mounting bracket in the chassis so the lca_rear mounting point can be shifted up/down. 2. LCA_Outer_x, y, z: These coordinates can be easily altered by spring ride height and ball joint adjustment. 3. Tie_Rod_inner_x, z: These can be shifted by moving the steering rack fore/aft or up/down. Note that length of existing rack cannot be changed, so, y coordinate is kept unchanged. 4. Tie_Rod_outer_x, y, z: This point can be freely moved by loosening or tightening the Jam Nut and by outer ball.

Fig. 6 Front Suspension with suggested modifications (encircled)

As for the Front Suspension in Rear Suspension also only twist beam to chassis hardpoint x and z coordinates can be easily shifted and they have been considered as design variables. The suggested modification is shown herewith.

Fig. 7 Rear Suspension with suggested modifications (encircled)

2nd International Conference on Research in Science, Engineering and Technology (ICRSET’2014), March 21-22, 2014 Dubai (UAE)

http://dx.doi.org/10.15242/IIE.E0314507 68

Quadratic model fitting procedure is been adopted under Monte Carlo simulation scheme. Initial values of variables are sampled at random from input probability distributions. Each set of samples is called an iteration, and resulting outcome of each iteration is recorded. These iterations are done a no. of times, and the result is a probability distribution. Model reliability is judged on the basis of various model fit estimates. This method is advantageous as it automatically eradicates non-sensitive design variables by assigning them a low weight in the final model, thus, there is no need to perform sensitivity analysis separately. Also it is quick and converges rapidly.

Fig. 8 Monte Carlo method implementation in ADAMS Insight

The chart below shows the fit estimates for Toe angle as one of the design variable for front suspension. Such fit estimates were calculated for each of the design variable. It has been verified for each of the design variable that it has a high F value and low P value so that the model fitting is reliable. Note that 128 runs were made, 1 run was for the first trial. The model has 9 dof as we have 9 design variables, therefore, the dof in which error can propagate is 118.

TABLE IV

MODEL FIT ESTIMATES FOR TOE ANGLE

VII. RESULTS The set of optimized hardpoints obtained after Monte Carlo

optimization study are shown below.

TABLE V FRONT SUSPENSION: OPTIMIZED DESIGN OBJECTIVES & HARDPOINTS

TABLE VI REAR SUSPENSION: OPTIMIZED DESIGN OBJECTIVES & HARDPOINTS

VIII. EXPERIMENTAL TESTING & CORRELATION The mathematical model of the vehicle developed has to be

validated with the experimental results. In order to test various vehicle characteristics it has been equipped with two hemispherical GPS sensors, one inertial measurement unit, and one accelerometer below driver’s seat (H point) to measure linear accelerations in all three directions. Below figures shows the measurement and data acquisition system from Onosokki and the vehicle equipped with various sensors.

2nd International Conference on Research in Science, Engineering and Technology (ICRSET’2014), March 21-22, 2014 Dubai (UAE)

http://dx.doi.org/10.15242/IIE.E0314507 69

Fig. 9 OnoSokki Measurement and Data Acquisition System[15]

Fig. 10 Vehicle Equipped with Sensors

For validating the mathematical model and to assess whether the modified vehicle shows better handling characteristics it is tested against some standard test procedures.Various tracks for vehicle testing are shown below,

Fig. 11 Acceleration, DLC, CRC and Slalom test tracks prepared in Graduate School of IPS Waseda University (Clockwise left to right)

The acceptance criteria used is a general design guideline for single seater hatchback category of vehicles in Automotive OEMs. If the vehicle passes the criteria then it was indicated by green color, if it falls 10 percent above the set limit it was indicated by yellow and if it still overshoots then it was assigned a red color. This means that there is a need for improvement in that particular characteristic. It is also been ensured that the percentage correlation is above 90 percent by repetitive testing and model updating (for instance, the torsional flexibility of frame was not modelled previously and it was updated during correlation study).

The test procedures are given below.

VIII (A). ACCELERATION AND BRAKING

Acceptance Criteria: 1) Maximum Vertical Acceleration < 0.5 g 2) Max Pitch Rate < 15 deg/sec 3) Maximum pitch angle < 3.5 deg 4) Maximum H-point vertical change< 0 to 25mm

Boundary condition: Speed Profile

Results:

2nd International Conference on Research in Science, Engineering and Technology (ICRSET’2014), March 21-22, 2014 Dubai (UAE)

http://dx.doi.org/10.15242/IIE.E0314507 70

VIII (B). DOUBLE LANE CHANGE (DLC)

Acceptance Criteria: 1) Peak Lateral Acceleration < 0.5 g 2) Peak Roll angle < 3.5 deg 3) Peak Pitch angle < 0.7 deg 4) Yaw Angle < 20 deg 5) Peak Pitch Rate < 10 deg/sec 6) Maximum vehicle velocity achieved > 30 kmph

Boundary conditions: 1) Initial Steer =0 deg 2) Maximum steer = 60 deg 3) Max Velocity= till max lateral acceleration is achieved 4) Track details according to ISO 3888

Results:

VIII (C). STEADY STATE CIRCULAR (SSC) TEST

Background: SSC is performed to determine the understeer gradient (K). There are two methods to measure K, Constant Radius and Constant Speed Method. We have used the first method in which vehicle undergo a constant radius turn and steering angle is measured with respect to lateral acceleration.

Fig 11 Constant Radius Method to determine K [16]

Acceptance Criteria: 1) Peak lateral acceleration < 0.6 g. 2) Susp roll gradient (Roll angle/ay)<4 deg/g 3) Body slip gradient (slip angle/ay)<5.5 deg/g 4) Pitch gradient (Pitch angle/ay)< 0.5 deg/g 5) Understeer gradient @0.5g< -30 deg/g 6) Understeer gradient @0.1g< -15 deg/g

Boundary conditions: 1) Initial velocity = 2m/sec. 2) Final velocity = Till vehicle reaches max lateral

acceleration,ay (constant ax of 0.1m/sec2) 3) Constant Radius tracks of 10,20,30 m respectively to find

understeer gradient variation with track radius

Results: Results of all the above acceptance criteria are shown only for 20 m radius test. Moreover, variation of K is shown for tests performed at all radius and at various levels of lateral acceleration. The vehicle shows understeer characteristics as indicated by the values of understeer gradient.

2nd International Conference on Research in Science, Engineering and Technology (ICRSET’2014), March 21-22, 2014 Dubai (UAE)

http://dx.doi.org/10.15242/IIE.E0314507 71

VIII (D). SLALOM OR SINUSOIDAL STEER Acceptance Criteria: 1) Peak Lateral Acc < 0.8 g 2) Peak Roll angle < 3.5 deg 3) Yaw Rate < 50 deg/sec 4) Max Roll rate < 15 deg/sec 5) Peak Sideslip Angle < 3 deg 6) Maximum Front Outside Tire Load (N) < 1.6 * static load

Boundary conditions: 1) Initial Steer =0deg 2) Maximum steering =60 deg. 3) Velocity= until max lateral acceleration level is achieved. 4) Track details: Seven cones are placed in a line with

distance between two cones as six times the wheel base 5) Vehicle path following check by position sensors.

Results:

IX. CONCLUSION AND FUTURE WORK Through this study the process of designing a suspension

assembly is been addressed based on the calculation of

suspension hardpoints by kinematic simulation. The data generated by the optimization software provides design insight and directions for the production process.

At first, a MacPherson Strut independent front suspension then Twist beam rear suspension is optimized by using virtual simulation. Finally, the correlation study is been performed to validate the model with experimental testing.

So far we have looked upon passive control by optimizing the hardpoints so that the vehicle handling is better and it exhibits a smooth ride. As the extension of this work we will now focus upon building an active control for vehicle ride and handling dynamics. The vehicle would be equipped by an autonomous drive set-up designed for the old aged people mainly concentrated around Kitakyushu in Fukuoka Prefecture, Japan. For this purpose the vehicle would be equipped with steering, accelerator and brake controller and efforts would be made to make ride more comfortable.

ACKNOWLEDGMENT The Authors would like to thank the Management of

Graduate School of IPS, Waseda University for providing the financial support and allowing us to use the lab facilities of the university. We would appreciate Onosokki Technical support team who helped us a lot regarding the use of measurement and acquisition system. We would also like to extend our thanks to Toyota Body Company for making the vehicular confidential data available to us specially tire related data which are quite difficult to obtain and its determination requires a lot of experimental effort.

REFERENCES [1] Cronin, D.L., 1981. “MacPherson strut kinematics”. Mechanism and

Machine Theory, 01/1981,16(6): 631-644. [2] M´antaras, D.A., P. Luque and C. Vera, 2004. “Development and

validation of a three-dimensional kinematic model for the McPherson steering and suspension mechanisms”. Mechanism and Machine Theory, 39(6): 603-619 http://dx.doi.org/10.1016/j.mechmachtheory.2003.12.006

[3] Attia, H.A., 2001. “Dynamic modelling of the double wishbone motor-vehicle suspension system”. Eur. J. Mech., 21(1): 167-174.

[4] Li, X., Y. Zhang, L. Zhou and M. Peng, 2003. “Characteristic analysis and calculation of wishbone independent suspension with torsion bar”. Journal of Automobile Eng., 25(1): 15-19

[5] Jia Zuo, Ying Shi Tang, Wei Cai Wang, Jian Chao Fu. “Optimization of Hardpoints of Independent Air Suspension Based on ADAMS”. Applied Mechanics and Materials (Volumes 184 - 185): 360-363

[6] Erik Wendeberg. “Using optimization to auto-correlate suspension characteristics to K&C measurements”. Master’s Thesis 2013:48, Department of Applied Mechanics, Chalmers University of Technology, Göteborg, Sweden 2013.

[7] J. S. Hwang, et al. “Kinematic design of a double wishbone type front suspension mechanism using multi-objective optimization”, 5th Australasian Congress on Applied Mechanics, ACAM 2007, Brisbane, Australia

[8] Simionescu, P. A. and Beale, D., 2002, “Synthesis and Analysis of the Five-Link Rear Suspension System used in Automobile”, Mechanism and Machine Theory, Vol. 32, pp. 815~232. http://dx.doi.org/10.1016/S0094-114X(02)00037-X

[9] Bian, X.L., B.A. Song and W. Becker, 2004. “The optimisation design of the McPherson strut and steering mechanism for automobiles”. Forschungim Ingenieurwesen, 68(1): 60-65.

2nd International Conference on Research in Science, Engineering and Technology (ICRSET’2014), March 21-22, 2014 Dubai (UAE)

http://dx.doi.org/10.15242/IIE.E0314507 72

[10] Habibi, H., K.H. Shirazi and M. Shishesaz, 2007. “Roll steer minimization of McPherson-strut suspension system using genetic algorithm method”. Mechanism & Machine Theory, 43(1): 57-67. http://dx.doi.org/10.1016/j.mechmachtheory.2007.01.004

[11] Makita, M., 1999. “An application of suspension kinematics for intermediate level vehicle handling simulation”. Society of Automobile Eng. Japan, 20(4): 471-477

[12] Lee, S.H., U.K. Lee and C.S. Han, 2001. Enhancement of vehicle handling characteristics by suspension kinematic control. Proceedings of Institution of Mechanical Engineers, 215: 197-216. http://dx.doi.org/10.1243/0954407011525584

[13] www.mscsoftware.com/product/adams [14] www.carsim.com/products/carsim/ [15] www.onosokki.co.jp [16] Thomas Gillespie, Fundamentals of Vehicle Dynamics [17] William Milliken & Douglas Milliken, Race Car Vehicle Dynamics [18] J C Dixon, Suspension Analysis and Computational Geometry

2nd International Conference on Research in Science, Engineering and Technology (ICRSET’2014), March 21-22, 2014 Dubai (UAE)

http://dx.doi.org/10.15242/IIE.E0314507 73