Model Based Suspension Calibration for Hybrid Ride and Handling Recovery

THESIS

Presented in Partial Fulfillment of the Requirements for the Degree Master of Science in the Graduate School of The Ohio State University

By

Matthew Joseph Organiscak, B.S.

Graduate Program in Mechanical Engineering

The Ohio State University

2014

Thesis Committee:

Dr. Giorgio Rizzoni, Advisor

Dr. Shawn Midlam-Mohler

Dr. Jeffrey Chrstos

Copyright by

Matthew Joseph Organiscak

2014

ABSTRACT

Automotive manufacturers spend many years designing, developing, and manufacturing each new model to the best of their ability. The push to shorten the time of a design cycle is motivated by reducing development costs and creating a more competitive advantage.

Model based design and computer simulations have an increasing presence in the automotive industry for this reason. In the automotive industry, current ride and handling tuning methods are subjective in nature. There are few, if any, objective evaluations of the vehicle ride and handling performance. EcoCAR 2 is a three year collegiate design competition, in which 15 teams compete to develop a vehicle with lower petroleum consumption and fewer emissions. The teams begin the task with a 2013 Chevrolet

Malibu and are challenged with maintaining consumer acceptability. The stock vehicle has been modified greatly by removal of the stock powertrain and battery system. Nearly

900 lbs of batteries and supporting components have been added to the trunk of the car along with an electric motor and single speed transmission. This change in vehicle mass has led to issues with poor ride and handling performance. Model based calibration of the suspension dampers can be seen as a method to recover some of the lost performance.

A CarSIM Model of the vehicle was developed after numerous measurements involving a vehicle inertial measurement facility as well as a kinematics and compliance test from a

ii suspension parameter measurement device. The CarSIM model was validated using experimental testing data. The vehicle was subjected to several different tests including a steady state handling test, transient handling test, and a ride test. To understand how different damper curve parameters would affect the performance of the vehicle, a design of experiments was developed using CarSIM to obtain the outputs. The seven objective metrics based on passenger comfort were used to create seven response surface equations based on the results of the DOE. Once the response surface equations are validated and considered to be accurate enough to be useful, a minimization optimization was performed to determine what inputs will return the lowest output value from each response surface. The individual response surface equation minimization showed that it was able to decrease uncomfortable accelerations and pitch motions while maintaining a vehicle that is easy to drive through transient handling maneuvers.

iii

DEDICATION

To my loving parents who have dedicated most of their lives to me, let all my

accomplishments also be yours

iv

ACKNOWLEDGEMENTS

I would like to thank, Dr. Shawn Midlam-Mohler, Dr. Rizzoni, and Dr. Jeff Christos for their guidance and patience throughout this project and other EcoCAR related projects throughout my career at The Ohio State University. Their support through these projects was crucial to the success of many students like myself at the Center for Automotive

Research.

I would like to thank David Emerling for setting me up with many industry sponsors that have made this project possible.

I would like to thank everyone on the EcoCAR 2 team for their camaraderie, support, and drive to compete at a high level

I would like to thank Gary Heydinger, Anmol Sidhu, Jonathan Coyle, and Hank Jebode at

SEA Limited for their help in measuring the EcoCAR 2 vehicle. This is an enormous task and few students are able to have such strong resources dedicated to them

I would like to thank Tom Mallin and Craig Jennings at Tenneco for their support and professional knowledge on the subject of dampers and damper tuning.

v I would like to thank Nick Stucky, Pat Majors, Seewoo Lee, Tim Donley and Adam

Homan for their support of the EcoCAR 2 program and the data measured for the sole purpose of my project.

I would like to thank Mike Neal, Mike Leduk and Ed Argalas for providing me with professional inside on ride and handling tuning as well as important vehicle data used in my model development

I would like to thank my friends from the racing world: Shawn MacNealy, Patrick Moro, and David Stacy for providing me with help when I needed it the most.

Lastly, I would like thank my friends and family for their patience and support during the most stressful times in my life.

vi

VITA

June 2008 ...... Dublin Coffman High School

December 2012 ...... B.S. Mechanical Engineering,

The Ohio State University

January 2013 to Present ...... Graduate Research Associate, Department

of Mechanical Engineering,

The Ohio State University

PUBLICATIONS

K. Bovee, A. Hyde, M. Yard, M. Organiscak, E. Gallo, A. Huster, A. Garcia, M. Yatsko, J. Ward, S. Midlam-Mohler and G. Rizzoni, "Fabrication of a Parallel-Series PHEV for the EcoCAR 2 Competition," SAE, 2013-01-2491.

K. Bovee, A. Hyde, M. Yard, T. Trippel, M. Organiscak, A. Garcia, E. Gallo, M. Hornak, A. Palmer, J. Hendricks, S. Midlam-Mohler and G. Rizzoni, "Design of a Parallel- Series PHEV for the EcoCAR 2 Competition," SAE, 2012-01-1762.

FIELDS OF STUDY

Major Field: Mechanical Engineering

vii

TABLE OF CONTENTS

ABSTRACT ...... II

DEDICATION...... IV

ACKNOWLEDGEMENTS ...... V

VITA...... VII

TABLE OF CONTENTS ...... VIII

LIST OF TABLES ...... XIII

LIST OF FIGURES ...... XIV

CHAPTER 1: INTRODUCTION ...... 1

1.1 Introduction ...... 1

1.2 EcoCAR 2 Competition ...... 1

1.3 EcoCAR 2 Architecture ...... 4

viii 1.4 Motivation ...... 5

CHAPTER 2: LITERATURE REVIEW AND BACKGROUND...... 10

2.1 Introduction ...... 10

2.1.1 Un-damped Ride Theory...... 11

2.1.2 Damped Ride Theory ...... 12

2.2 Subjective Handling Tuning ...... 13

2.3 Objective Handling Tuning Combined Approach ...... 15

2.4 Objective Ride Quality ...... 15

2.5 Simulated Objective Transient Handling ...... 16

CHAPTER 3: VEHICLE MEASUREMENT AND CHARATERIZATION ...... 17

3.1 Purpose of Vehicle Measurement ...... 17

3.2 Vehicle Inertial Measurement ...... 17

3.3 Suspension Parameter Measurement ...... 19

3.4 Auxiliary Wheel Rate Measurement ...... 21

3.5 Tire Measurement ...... 25

3.6 CarSIM ...... 27

ix 3.7 Vehicle Testing ...... 28

3.7.1 Steady State Turn Testing ...... 32

3.7.2 Transient Testing ...... 34

3.7.3 Ride Testing ...... 35

3.8 Damper Tuning ...... 37

CHAPTER 4: INITIAL VEHICLE ASSEMENT ...... 41

4.1 Baseline Vehicle Assessment ...... 41

4.2 Competition and Post-Competition Vehicle Assessment...... 42

4.3 Pre-Competition Stability Testing ...... 43

4.4 Possibilities for Tuning ...... 53

4.5 Difficulties in Tuning ...... 54

CHAPTER 5: RIDE AND HANDLING SIMULATION ...... 56

5.1 Model Development ...... 56

5.1.1 ...... 58

5.1.2 Rigid Sprung Mass ...... 59

5.1.3 ...... 60

5.1.4 Front and Rear Suspension Kinematics ...... 61

5.1.5 Front and Rear Suspension Compliance ...... 62 x 5.1.6 Aerodynamics, Brakes and Powertrain ...... 64

5.2 Model Validation ...... 64

5.2.1 Validation Tests ...... 64

5.2.2 Possible Reasons for Differences in Simulated Results ...... 74

CHAPTER 6: REFINEMENT...... 76

6.1 Development of Objectives ...... 76

6.2 Un-Damped Refinement ...... 77

6.3 Overall Simulation Process ...... 82

6.4 Inputs ...... 83

6.5 Response Surface Equations ...... 88

6.6 Response Surface Validation ...... 90

6.7 Optimization ...... 105

6.8 Results ...... 107

CHAPTER 7: CONCLUSIONS AND FUTURE WORK ...... 119

7.1 Conclusions ...... 119

7.2 Future Work ...... 120

xi BIBLIOGRAPHY ...... 121

APPENDIX: LIST OF SYMBOLS AND ABBREVIATIONS ...... 123

APPENDIX: FIGURES ...... 124

xii

LIST OF TABLES

Table 1: Vehicle Technical Specifications for the EcoCAR 2 Vehicle ...... 3

Table 2: Comparison of the Vehicle Mass From OEM to Competition ...... 41

Table 3: Understeer Gradient for Stock and Modified Car...... 44

Table 4: Metrics Chosen to Objectively Evaluate the Vehicle Performance ...... 77

Table 5: Response Surface Equation Validation Results ...... 91

Table 6: Individual RSE Output Minimization Results ...... 108

Table 7: Results of Overall Optimization ...... 109

xiii

LIST OF FIGURES

Figure 1: EcoCAR 2 Vehicle Architecture ...... 5

Figure 2: EcoCAR Vehicle Development General Timeline ...... 7

Figure 3: Ride and Handling Role in the EcoCAR Competition ...... 8

Figure 4: EcoCAR 2 Year 3 Competition Dynamic Event Points ...... 9

Figure 5: Corner Car Model [1] ...... 11

Figure 6: Triangle Steer Input [6] ...... 14

Figure 7: Ramp Steer Input [6] ...... 14

Figure 8: Drawing of Inertial Measurement in Pitch Mode [8] ...... 18

Figure 9: The EcoCAR 2 Vehicle on the Vehicle Inertial Measurement Table ...... 18

Figure 10: Vehicle on the Suspension Parameter Measurement Device ...... 20

Figure 11: EcoCAR Rear Suspension Control Arm ...... 22

Figure 12: EcoCAR Rubber Suspension Bushing ...... 23

xiv Figure 13: Wheel Rate Measurement Set Up ...... 24

Figure 14: Wheel Travel Measurement ...... 24

Figure 15: Example of Cornering Stiffness (Fy) Tire Data ...... 25

Figure 16: Example of a Tire Testing Machine (Souce: MTS) ...... 26

Figure 17: CarSIM Animation of Vehicle During Steady State Turn ...... 27

Figure 18: Diagram Showing Vehicle Measurements Compared to Vehicle Performance

Measurements ...... 28

Figure 19: VBOX Instrumentation in the EcoCAR 2 Vehicle ...... 30

Figure 20: Strut System Used to Hold the Acclerometer in the Vehicle ...... 31

Figure 21: Path Followed for Steady State Turn ...... 33

Figure 22: Double Lane Change Layout Used for Testing ...... 34

Figure 23: The EcoCAR Vehicle Driving the Double Lane Change Test ...... 35

Figure 24: The EcoCAR 2 Vehicle Driving Over a the Large Bump Ride Profile ...... 37

Figure 25: Diagram Showing the Internal Layout of a Twin Tube Damper (source

Tenneco) ...... 38

Figure 26: Force vs Velocity Damper Dynomometer Curve (source Tenneco) ...... 39

xv Figure 27: EcoCAR Team Rebuilding a Damper ...... 40

Figure 28: 35mph Stock Car Driving Line ...... 45

Figure 29: 35mph Stock Car ...... 46

Figure 30: 55mph Stock Car Driving Line ...... 47

Figure 31: 55mph Stock Car ...... 48

Figure 32: 35mph Modified Car Driving Line ...... 49

Figure 33: 35mph Modified Car ...... 50

Figure 34: 55mph Modified Car Driving Line ...... 51

Figure 35: 55mph Modified Car ...... 52

Figure 36: Diagram Showing the Max Potential of Damper Curve Adjustment (source

Tenneco) ...... 54

Figure 37: CarSIM Main Screen for GUI ...... 56

Figure 38: The Vehicle Model Screen in the CarSIM GUI ...... 57

Figure 39: Tire Data Screen in the CarSIM GUI ...... 59

Figure 40: Rigid Sprung Mass Section of the CarSIM GUI ...... 60

Figure 41: Steering Section of the CarSIM GUI ...... 61

xvi Figure 42: Front Kinematics Section of the CarSIM GUI ...... 62

Figure 43: Front Compliance Section of the CarSIM GUI ...... 63

Figure 44: Lateral Acceleration 35mph Steady State Turn Validation ...... 65

Figure 45: Yaw Rate 35mph Steady State Turn Validation ...... 66

Figure 46: Lateral Acceleration 55mph Steady State Turn Validation ...... 67

Figure 47: Yaw Rate 55mph Steady State Turn Validation ...... 68

Figure 48: Lateral Acceleration 50mph Double Lane Change Validation ...... 69

Figure 49: Yaw Rate 50mph Double Lane Change Validation ...... 70

Figure 50: Large Bump Profile used for Ride Validation...... 71

Figure 51: Filtered Large Bump Profile used for Ride Validation ...... 72

Figure 52: Pitch Rate 40mph Ride Bump Validation ...... 73

Figure 53: Vertical Acceleration 40mph Ride Bump Validation ...... 74

Figure 54: Rear Ride Frequency Calculation ...... 81

Figure 55: Diagram for Overall Refinement Process ...... 83

Figure 56: Damper Curve Variable Breakdown ...... 84

Figure 57: Damper Curve Variable Adjustment Ranges ...... 85 xvii Figure 58: 3D Point Distribution for three of the 12 input factors ...... 86

Figure 59: All Front Damper Curves Simulated ...... 87

Figure 60: All Rear Damper Curves Simulated ...... 88

Figure 61: Response Surface Fitting for Maximum Vertical Acceleration at the Beginning

of the Ride Bump ...... 89

Figure 62: Response Surface Fitting for Maximum Lateral Acceleration during the

Double Lane Change Test ...... 90

Figure 63: A Comparison of Max Lateral Acceleration using the Outputs from 15 Sets of

Input Factors ...... 92

Figure 64: The Full Output of Lateral Acceleration from CarSIM for Run 10 ...... 93

Figure 65: A Comparison of Max Roll Angle using the Outputs from 15 Sets of Input

Factors ...... 94

Figure 66: The Full Output of Roll Angle from CarSIM for Run 10 ...... 95

Figure 67: A Comparison of Max Steering Wheel Rate using the Outputs from 15 Sets of

Input Factors ...... 96

Figure 68: The Full Output of Steering Wheel Rate from CarSIM for Run 10 ...... 97

xviii Figure 69: A Comparison of Max Pitch Angle at the Beginning of the Bump using the

Outputs from 15 Sets of Input Factors ...... 98

Figure 70: The Full Output of Pitch Angle at the Beginning of the Bump from CarSIM

for Run 10 ...... 99

Figure 71: A Comparison of Max Pitch Angle at the End of the Bump using the Outputs

from 15 Sets of Input Factors...... 100

Figure 72: The Full Output of Pitch Angle at the End of the Bump from CarSIM for Run

10...... 101

Figure 73: A Comparison of Max Vertical Acceleration at the Beginning of the Bump

using the Outputs from 15 Sets of Input Factors ...... 102

Figure 74: The Full Output of Vertical Acceleration at the Beginning of the Bump from

CarSIM for Run 10 ...... 103

Figure 75 A Comparison of Max Vertical Acceleration at the End of the Bump using the

Outputs from 15 Sets of Input Factors ...... 104

Figure 76: The Full Output of Vertical Acceleration at the End of the Bump from CarSIM

for Run 10 ...... 105

Figure 77: Damper Curves Calculated from the Optimization Minimization ...... 111

xix Figure 78: Lateral Acceleration Output from CarSIM for the Optimized Damper Curves

...... 112

Figure 79: Roll Angle Output from CarSIM for the Optimized Damper Curves ...... 113

Figure 80: Steering Wheel Rate Output from CarSIM for the Optimized Damper Curves

...... 114

Figure 81: Vertical Acceleration for the Beginning of the Bump Output from CarSIM for

the Optimized Damper Curves...... 115

Figure 82: Vertical Acceleration for the End of the Bump Output from CarSIM for the

Optimized Damper Curves ...... 116

Figure 83: Pitch Angle for the Beginning of the Bump Output from CarSIM for the

Optimized Damper Curves ...... 117

Figure 84: Pitch Angle for the End of the Bump Output from CarSIM for the Optimized

Damper Curves ...... 118

Figure 85: Baseline Vehicle VIMF Data ...... 124

Figure 86: Modified Vehicle VIMF Data ...... 125

Figure 87: Competition DCA Results ...... 126

xx

CHAPTER 1: INTRODUCTION

1.1 Introduction

Automobiles require many complex systems working in harmony to effectively deliver a pleasant experience to the driver and passengers. Automotive manufacturers spend many years designing, developing, and manufacturing each new model to the best of their ability. The push to shorten the time of a design cycle is motivated by reducing development costs and creating a more competitive advantage. Model based design and computer simulations have an increasing presence in the automotive industry for this reason. Production phases that are early in the design cycle are being phased out in favor of computer simulation in order to save resources. This can currently be seen in vehicle control and powertrain development. A good example of this trend is on the EcoCAR 2 development process.

1.2 EcoCAR 2 Competition

EcoCAR 2 is a three year collegiate competition that challenges university students to re- engineer a 2013 Chevrolet Malibu to become more fuel efficient, reduce emissions, and maintain consumer acceptability. This competition has numerous industry sponsors

1 including General Motors, the Department of Energy, and many more. Teams explore numerous architectures that utilize alternative fuels and hybrid electric systems.

Each year in the competition represents a different phase of the design cycle. In year one, the teams do all of the design of the vehicle on paper without an actual car. In year two, the vehicle is built, but not expected to be a final vehicle. Year three gives the teams time to fine tune their to a 99% buyoff stage. During each of the three years, the teams focus on developing and meeting metrics including performance, energy efficiency, and usability.

2 Table 1: Vehicle Technical Specifications for the EcoCAR 2 Vehicle

Production Competition OSU Parallel- Specification 2013 Malibu Design Target Series PHEV ECOCAR COMPETITION REQUIREMENTS Acceleration 0-60 (s) 8.2 sec 11.5 sec 11.2 s Acceleration 50-70 (s) 8.0 sec 10 sec 5.4 s 43.7 m [143.4 Braking 60-0 54.8 m [180 ft] 43.7 m (143.4 ft) ft] Highway Gradeability @ 10+% @ 60 3.5% @ 60 mph 3.5% @ 60 mph 20 min mph

Cargo Capacity 16.3 ft3 7 ft3 7 ft3 Passenger Capacity 5 ≥ 4 5 Mass 1,589 kg ≤ 2078 kg 2,000 kg (4,410 lb) Starting Time ≤ 2 sec < 15 sec < 2 sec Ground Clearance 155 mm >127 mm >127 mm 736 km (457 Range ≥ 322 km [200 mi] 415 km (258 mi) mi) (CAFÉ) ECOCAR COMPETITION TARGETS Charge Depleting Range* N/A ** 72.0 km (44.7 mi) Charge Depleting Fuel N/A ** 0 lge/100km Consumption* 7.52 Charge Sustaining Fuel N/A ** (lge/100km(ge) Consumption* [670.0 Wh/km] 8.83 (lge/100 7.12 (lge/100 km) 2.61 (lge/100 km) UF-Weighted Fuel Energy km) Consumption* [787 Wh/km] [634 Wh/km] [232.2 Wh/km] UF-Weighted AC Electric N/A ** 192.7 (Wh/km) Energy Consumption* UF-Weighted Total 787 (Wh/km) 634 (Wh/km) 424.8 (Wh/km) Energy Consumption* UF-Weighted WTW 774 (Wh Petroleum Energy (PE) 624 (Wh PE/km) 79.9 (Wh PE/km) PE/km) Use* UF-Weighted WTW GHG 253 (g 204 (g GHG/km) 185.5 (g GHG/km) Emissions* GHG/km) Criteria Emissions Tier II Bin 5 Tier II Bin 5 < Tier II Bin 5

* Evaluated using EcoCAR 2 combined “4-cycle” weighting method, which weights US06, 505 and HWFET to reflect 43% city driving, 57% highway driving

** There is no competition design target, but EcoCAR teams are expected to report their predictions in these categories for their VTS.

3

1.3 EcoCAR 2 Architecture

The Ohio State EcoCAR 2 vehicle is a series-parallel plug in electric hybrid. This design utilizes E85 as its primary range extending fuel and an 18.9kW-hr battery pack for all electric use. The E85 engine is a 1.8L Honda compressed natural gas engine that has been converted to run on E85. Mated to the engine is a 6-speed manual transmission that has been automated by the team to operate as an automatic transmission. An 80kW electric machine is coupled to the input shaft of the transmission allowing the input of the transmission to speed match to the wheel speed during shifts. This machine also drives the vehicle in parallel and all electric modes. The front machine can also be used to generate from the engine in series mode. The rear axle is driven by a second 80kW electric machine that utilizes a single speed transmission to drive the wheels.

4

Figure 1: EcoCAR 2 Vehicle Architecture

1.4 Motivation

In the automotive industry, current ride and handling tuning methods are subjective in nature. Engineers drive the vehicles during numerous stages of production, and manually adjust different suspension components in order to set the car to feel a specific way when driven. There are few, if any, objective evaluations of the vehicle ride and handling performance. The dampers specifically may require 100 to 200 rebuilds during the development cycle of a vehicle. Each change requires all four dampers to be removed from the vehicle, disassembled, re-assembled and installed back on to the vehicle.

5 A tool for quick damper calibration has the capability to reduce development time, and increase repeatability on vehicle ride and handling metrics. This tool could create a quick starting point for the suspension tuning and reduce the overall number of suspension rebuilds required as well as eliminate the need for tuning at early stages of development.

In the case of the EcoCAR 2 development process, this could change the way that the team develops vehicle overall.

Currently, the mechanical, electrical, and software teams develop models and simulations of their respective portions of the vehicle in year one before any physical vehicle development has taken place. In year two, the team may not finish the vehicle build with enough time to test and tune before it competes. Year three is often the only time that chassis tuning is considered, and at this point it is only a last minute fix to make sure that the car drives safely. The vehicles are currently developed without much consideration to , and chassis tuning is saved until the final year.

6

Figure 2: EcoCAR Vehicle Development General Timeline

A chassis tuning tool for ride and handling could be useful, by allowing the EcoCAR team to introduce a vehicle dynamics model early in the development process. This would allow for the original vehicle design to be more influenced by vehicle performance and driver comfort goals. This tool would also give the team a starting point when tuning chassis components and reduce development time.

In the EcoCAR 2 competition, non-energy efficiency and emissions related points can be broken down into drive comfort and performance. This best describes how vehicle ride and handling fits into the competition. Handling is evaluated in autocross and max lateral

7 acceleration events, while ride is evaluated in a dynamic consumer acceptability event.

Together these events make up 95/520 points of dynamic events. While this is not the majority of the dynamic event points, it is still a portion large enough to make a difference in the outcome of the competition.

Figure 3: Ride and Handling Role in the EcoCAR Competition

8

Figure 4: EcoCAR 2 Year 3 Competition Dynamic Event Points

An analytical method for computing chassis component specifications has a potential to be useful to both the automotive industry as well as the EcoCAR 2 Team. By objectifying ride and handling metrics, chassis system tuning could become more consistent. This tool saves time and money during the development process, as well as contributing to a more well-rounded vehicle overall.

9

CHAPTER 2: LITERATURE REVIEW AND BACKGROUND

2.1 Introduction

Vehicle suspension systems are comprised of suspension springs to support and suspend the weight of the vehicle and dampers to control the motion of the wheels. If a vehicle only had springs, and no dampers, a vehicle would theoretically bounce forever after hitting a bump. Additional components include anti-roll bars, jounce bumpers, and suspension control arms. These components are typically tuned subjectively in the automotive industry as ride and handling tuning is largely based on driver and passenger perception. Objective metrics can be developed to describe subjective feelings and evaluated objectively.

Ride and Handling is a generic term used to describe vehicle performance while cornering and driving on roads with noticeable surface bumps or elevation changes.

Handling relates to the ability of the vehicle to maneuver while cornering and the way the driver interacts with the car while cornering. Ride relates to the ability of the vehicle to separate variations in the road surface from the driver and passenger. The road inputs that affect ride are typically bumps that are large enough to cause visual movement in the vehicle. Road inputs that are higher frequency and cause more sound and vibration

10 related issues are referred to as noise [1]. In general, lower frequency road inputs affect ride, while higher frequencies affect noise, vibration, and harshness, known as NVH.

2.1.1 Un-damped Ride Theory

Vehicle suspensions are designed to isolate the road roughness from the sprung mass of the vehicle. Main suspension springs and dampers separate the sprung mass from the un- sprung mass. The tire itself can be thought of as a spring, and this separates the un- sprung mass from the road surface input. When the springs of the system are summed up, it can be combined into an overall suspension stiffness, which is called ride rate.

Figure 5: Corner Car Model [1]

11 When the damper is taken out of the equation, the sprung mass is separated by the total suspension stiffness from the road. Input frequencies from the road will cause the mass to resonate at a specific frequency. Both the front axle and the rear axle of the vehicle have different un-damped natural frequencies. For passenger cars, these frequencies are between 1-1.5 Hz (cycles/sec) [1]. The higher the un-damped natural frequency, the faster the system will resonate. Racing cars which do not see high amplitude road inputs typically have higher un-damped natural frequencies. For passenger cars, the un-damped natural frequency of the rear axle is typically higher than the front by about 20%. This reduces pitching of the sprung mass by allowing the rear axle to “catch up” to the front axle. A pitching motion of the sprung mass is considered to be less bearable by passengers than a straight up and down heaving motion [1].

2.1.2 Damped Ride Theory

Suspension dampers play multiple roles in a vehicle suspension. While the springs of the suspension allow for the vehicle to absorb impacts, little energy is actually dissipated during a bump and rebound event. If a vehicle was outfitted with only springs and no dampers, the vehicle would theoretically oscillate forever after hitting a bump. Dampers are used to dissipate energy from the suspension while it is in compression and rebound.

Dampers are also used to control the wheel motion of the vehicle during handling maneuvers [1]. Dissipating energy stored in a vehicle suspension during multiple cornering maneuvers is important to control the weight of the sprung mass as it rolls and pitches. The multiple roles that the dampers play in the vehicle suspension make

12 achieving optimal performance more difficult than the tuning of other suspension components.

2.2 Subjective Handling Tuning

Currently, the automotive industry tunes damper characteristics in a subjective manner.

This means that the engineer, who is evaluating the vehicle, is trying to perceive the ride and handling of the vehicle the same way that a customer would perceive it. The engineer then adjusts the dampers appropriately to meet the metrics that they use their own body to feel.

Numerous tests are used to subjectively evaluate performance based on numerous metrics. In a method chosen by engineers at General Motors, transient handling metrics can be broken down into agility, stability, precision, and roll support [6]. Three tests are used to evaluate these metrics. A low amplitude triangle pulse steer, a high amplitude triangle pulse steer, and a ramped step steer. The triangle input means that the steering wheel angle is applied at a constant rate to a certain angle amplitude, and then reduced at a constant rate back to zero. This is used to judge vehicle agility and lag between steering input and vehicle response. At larger amplitudes, stability is evaluated by understanding how the rear axle tracks with respect to the front axle [6].

13

Figure 6: Triangle Steer Input [6]

Figure 7: Ramp Steer Input [6]

The ramp steer is a constant rate steering wheel input that is used to mimic entry into a corner. This is used to evaluate all of the metrics in harmony [6].

14 2.3 Objective Handling Tuning Combined Approach

Correlating subjective handling metrics to objective handling metrics can be a difficult task. Engineers at MIRA in Nuneaton, England assembled a test with eight drivers and two cars. A 49 question evaluation form was developed for the drivers to rate the vehicle performance in various maneuvers. The questions were tied to objective metrics such as hand wheel angle, vehicle response times, and vehicle roll angle. There were three tests on which the drivers focused: steady state steering, a J-turn test, and a steering wheel impulse. The goal is to determine how the rating system of the drivers correlated to the measured performance of the vehicles. This showed how a driver would rate a car based on its measured outputs. Many overall metrics were linked to subjective feelings, but it is difficult to narrow down subjective feelings to specific aspects of handling [3].

2.4 Objective Ride Quality

Objective ride quality is the quality of comfort that a person experiences in an automobile when traversing non-flat roads [4]. Engineers at The University of Texas at Austin utilized guidelines from ISO Standard 2631 which define how ride can be measured objectively. In this study, accelerometers were attached at the seat of the vehicle where the passengers would feel the inputs. They measured vertical and longitudinal accelerations that would be felt by the passengers. Acceleration vs. time data was converted to acceleration vs. frequency data and divided into 1/3 octave bands defined by the ISO standard. This is used to find the vector sum of the band. These can be compared to the ISO reduced comfort curves [4]. 15 Two vehicles were tested around Austin, Texas with known profiles. Seven tests were performed in increments of 10 mph from 30 to 80 mph. An extra test was performed at

50 mph to prove consistency. It was determined that magnitude of inputs and length of time that inputs are felt by the passenger were considered to lead to the most discomfort

[4]. While the test was used for other comparison purposes, it shows a valid method for objective ride comfort evaluation.

2.5 Simulated Objective Transient Handling

Objective handling evaluation of a vehicle design using a simulation is important as it allows the vehicle to be evaluated before it exists as a physical prototype. Engineers at the University of Pretoria used a simulation to evaluate a vehicle during a double lane change maneuver. A vehicle model, lateral tire force model, and a driver model were developed for use in the simulation. Steer angle, yaw rate, and roll rate were used to compare the experimental tests to the simulated tests. Similar trends here validate the model as usable for testing and gathering data. Vehicle component such as tire parameters, suspension stiffness, damper characteristics, and vehicle inertia were varied and then compared to the original results to understand how they affect the performance of the vehicle [5]. This testing is useful, because it provides an example of how a vehicle was evaluated objectively and useful results were obtained. The closed loop driver model replaced a real life driver and improved the repeatability of the testing.

16

CHAPTER 3: VEHICLE MEASUREMENT AND CHARATERIZATION

3.1 Purpose of Vehicle Measurement

In order to generate a computer model of a vehicle, many characteristics of the vehicle must be known. The model used in this project required well over 110 vehicle parameters at a minimum to be defined in order for CarSIM to be able to run a simulation. Numerous measurements were taken from individual components and systems of components in order to characterize the vehicle for the CarSIM model. The major measurements from the vehicle involve mass and inertial measurements from the vehicle sprung mass, kinematics and compliance tests for the front and rear suspension, individual suspension compliance tests, and tire testing.

3.2 Vehicle Inertial Measurement

In order to measure the total vehicle mass and sprung mass component of the vehicle, the

Vehicle Inertial Measurement Facility (VIMF) was used at SEA Limited. This device measures 4 inertias that are useful in describing the vehicle when it yaws, pitches, and rolls. The machine collects this data by attaching the vehicle at the appropriate ride height to a large platform. The platform is then swung is various configurations to put the vehicle into yaw, pitch, and roll. This movement is used to calculate the inertias. 17 Another important output of the VIMF Test is the center of gravity height of the total vehicle. [7]

Figure 8: Drawing of Inertial Measurement in Pitch Mode [8]

Figure 9: The EcoCAR 2 Vehicle on the Vehicle Inertial Measurement Table

18

The ability to calculate center of gravity height is used to estimate the ratio of sprung to center of gravity of the sprung mass. This test is performed by measuring the total center of gravity height of the vehicle at different trim heights. A change in trim height would cause the sprung mass to increase in height, while most of the unsprung mass will stay stationary. The difference is used to calculate the ratio between sprung and unsprung mass. This is useful, because measurements of unsprung mass are fairly difficult and time consuming. These measurements typically require vehicle disassembly. Unsprung mass weight, sprung mass weight, and sprung mass center of gravity height are required inputs for the CarSIM model. [8]

3.3 Suspension Parameter Measurement

In order to understand how the wheels of the vehicle move when the vehicle is under a wide array of conditions, a kinematics and compliance test was completed on the vehicle.

The Suspension Parameter Measurement Device (SPMD) utilizes several different test scenarios to determine the kinematics of the wheels and the compliance of the suspension components at all four wheels. The machine operates in three different modes.

19

Figure 10: Vehicle on the Suspension Parameter Measurement Device

The first mode measures the actual force at the tire patch and movement of each wheel during suspension compression and rebound. This is accomplished by attaching the body of the vehicle to four large hydraulic cylinders that lift and lower the vehicle body in both heave and roll modes. Scales with added low friction pads measure the force output from each wheel due to the suspension. As the cylinders push and pull the vehicle body through the motions the scales, along with body mounted string potentiometers measure the combined suspension stiffness. At each wheel, there are three linear potentiometers and several string potentiometers that measure wheel motion longitudinally, laterally and vertically as well as wheel camber and toe [9].

20 An additional operating mode utilizes a pneumatic cylinder to pull and push on the low friction pads beneath the tires longitudinally and laterally. This test measures the compliance in the suspension components, bushings, and tires. This information will tell the model how the suspension will act when influenced by certain loads [9].

The last mode measures the steering system of the vehicle. The steering ratio from the steering wheel to the wheel is measured throughout the full travel. The compliance in the steering system is also measured with a large electric ball screw actuator that applies a load through a long lever to one of the front wheels while the steering wheel is held straight [9]. This is important to understand how the wheels will move when the steering wheel is turned.

3.4 Auxiliary Wheel Rate Measurement

There are several contributors to the spring rate of the suspension when measured at the wheel. The major contributor is the suspension spring which is mounted inside of the suspension linkage. Other smaller components of the wheel rate are the bushings inside of the control arms and the gas pressure in the damper. This force occurs because the bushing rubber deflects when the suspension articulates into jounce or rebound from the position that it was installed. This deflection of the rubber causes a spring force. For the purpose of this project, the bushing contribution must be measured in order to separate the main coil spring rate from the total wheel rate. The suspension motion ratio was

21 calculated by comparing the wheel rate to the spring rate. The bushing rate can be up to

25% of the total wheel rate (info from Mike Neal). This information must be calculated in order to understand what the effect of the coil spring may be on the total wheel rate.

While the suspension springs are simple to measure out of the car with a spring rating machine, the bushing rate is not simple to measure outside of the vehicle. Measurement was done directly at the wheel with a scale and a floor jack. This is similar to a wheel rate measurement, except the spring was removed. This measurement then shows the contribution of the bushings as well as the gas from the damper.

Figure 11: EcoCAR Rear Suspension Control Arm

22

Figure 12: EcoCAR Rubber Suspension Bushing

Figure 12 shows a close up of the rubber bushing in a control arm of the rear suspension of the EcoCAR 2 vehicle. When the suspension compresses, the rubber inside of the bushing deflects as the inside of the bushing twists independently of the outside of the bushing. This causes a rotational resistive spring force about the bushing.

23

Figure 13: Wheel Rate Measurement Set Up

Figure 14: Wheel Travel Measurement 24

3.5 Tire Measurement

In order to build a vehicle model, measurements from the tires used on the vehicle must be gathered for numerous conditions to which the tire may be subjected. During a tire test, a tire will be pressed onto a large roller or belt and spun at different vertical loads, slip angles, and camber angles. The force and moment outputs from the tire are then measured and recorded. These tests are designed to mimic conditions that a tire may see during actual driving. Tire measurement was completed by Cooper Tire, the manufacturer of the tire for the EcoCAR 2 vehicle.

Figure 15: Example of Cornering Stiffness (Fy) Tire Data 25

Figure 16: Example of a Tire Testing Machine (Souce: MTS)

Figure 16 shows an example of a tire being tested on a force and moment tire testing machine. Figure 15 shows example data for lateral force that a tire may output under certain slip angle conditions during multiple wheel vertical loads. The actual tire lateral force is extrapolated from the data.

26 3.6 CarSIM

CarSIM is software used to simulate the behavior of numerous types of highway vehicles.

Vehicle models are assembled with hundreds of measurements from a vehicle and testing procedures are developed to test these models. Vehicle inputs for a testing procedure range from driver acceleration, braking and steering inputs, to roadway surface information. CarSIM software utilizes a simple GUI, simple output data plotting tool, and animates the test procedure. It allows for outputs to excel and Matlab software which can be further used for post processing. Figure 17 shows the animation function of

CarSIM which allows the user to visually see the vehicle as it drives through the test.

This gives the user a visual check to make sure that the test is running properly.

Figure 17: CarSIM Animation of Vehicle During Steady State Turn

27

3.7 Vehicle Testing

While the physical vehicle is measured in great detail for the development of a model, the vehicle performance must also be measured. It is important to have an understanding of how the measured parameters of the vehicle will perform under certain conditions. While vehicle and component measurements can be considered an input, performance measurements are the respective outputs to those inputs. These performance outputs are useful for comparing the experimentally measured vehicle to the modeled vehicle in order to evaluate the accuracy of the model.

Figure 18: Diagram Showing Vehicle Measurements Compared to Vehicle Performance

Measurements

28 The vehicle was subjected to several different tests including a steady state handling test, transient handling test, and a ride test. These tests were performed with the EcoCAR vehicle at the Transportation Research Center in their Vehicle Dynamics Area. Each test was performed with a VBOX measurement system. This is a high accuracy GPS based data logging system that is capable of measuring numerous vehicle performance metrics.

In this case, vehicle speed, driver line, three linear accelerations, and 3 rotational speeds were measured and used to describe the vehicle performance. The accelerations and rotational speeds were measured using an IMU04 unit which is an extension of the

VBOX system. This unit was mounted above the center console arm rest for convince. It was attached to the vehicle by a strut system that held the unit in place by placing force between the vehicle ceiling and the vehicle floor. The GPS receiver was mounted to the roof of the vehicle above the IMU.

29

Figure 19: VBOX Instrumentation in the EcoCAR 2 Vehicle

30

Figure 20: Strut System Used to Hold the Acclerometer in the Vehicle

The Transportation Research Center (TRC) is an independent automotive proving ground vehicle for research and development. The vehicle dynamics area (VDA) used is a 50 acre asphalt pad is a multipurpose space for numerous types of dynamic vehicle testing.

In the case of the EcoCAR 2 vehicle, the VDA was able to be used for all of the handling testing. The profile roads at TRC were used for evaluation of vehicle ride. These roads provide a number of repeatable input frequencies into the suspension.

31 3.7.1 Steady State Turn Testing

Steady state turn testing evaluates the behavior of the vehicle under a turning condition where the lateral movement of the sprung vehicle mass is nearly static. This is a constant speed test in which the driver slowly increases the steering wheel angle until the vehicle does not produce any additional lateral acceleration. The driver may describe this feeling as excessive understeer or oversteer. In the case of the EcoCAR 2 vehicle, the steering wheel angle was increased until the car began to understeer greatly. This ensures that the vehicle has had a chance to build up as much lateral acceleration as possible.

Steady state turning is mostly a test of the performance of the vehicles tires in this steady state roll configuration. In theory, the dampers will not play any role in the behavior of the vehicle in this test, assuming that the vehicle is on a completely flat surface and the driver turns in slowly. For the purpose of this project, the objective is to obtain max lateral acceleration and understeer gradient data.

32 GPS Path Measured 35mph Steady State Turn 30

25

20

15

10 Y Distance (m)

5

0

-5 0 20 40 60 80 100 120 X Distance (m)

Figure 21: Path Followed for Steady State Turn

Figure 21 shows the actual vehicle path during a steady state turn. This data is calculated from GPS data. Notice that the vehicle gets up to speed in a straight line and then the turn is initiated. The radius is constantly decreasing for a short time until the vehicle begins to understeer heavily and the vehicle is steered back to straight.

33 3.7.2 Transient Testing

Transient vehicle testing is designed to evaluate the behavior of the vehicle in a non- steady state handling maneuver. The test should have constantly changing conditions during the test. The double lane change maneuver was chosen for the transient testing procedure because it is easy to repeat and easy to drive. This test involves three stations of cones, each representing a lane. The vehicle is driven at a constant speed through the first lane, forced to turn into the second lane, and then turn back into the third lane This procedure is also called ISO 3888-2. The exact test procedure from ISO 3888-2 was not used, but the general idea was maintained for the testing. The lanes were slightly modified in width to allow for the test to be easier to maneuver, and the speed was held nearly constant. [10]

Figure 22: Double Lane Change Layout Used for Testing

34

Figure 23: The EcoCAR Vehicle Driving the Double Lane Change Test

3.7.3 Ride Testing

Ride evaluates how the sprung mass of the vehicle performs when the vehicle is subjected to frequency inputs at the wheels. At TRC, the vehicle was driven over a few different ride profiles to try to understand what sort of input shape and frequency would excite the sprung mass the most. The more the sprung mass of the vehicle moves when

35 excited by the ride profile, the easier it will be to see trends in the accelerometer IMU data. Unlike the handling tests, the ride tests were performed in a straight line.

The objective of the ride testing was to understand how the driver and passengers would feel in the vehicle when moving over various ride profiles. Objective characteristics that are useful include sprung mass, vertical acceleration, pitch velocity and number of oscillations after the event.

The profile that was chosen was approximately a trapezoid shape that was nearly 78m long. The top measures about 50m long and 0.3m tall. The profile ramps up evenly on both sides to the 0.3m tall flat top. The tests were run at a constant speed. Speeds were slowly increased starting at 20mph and ending at 40mph when the vehicle was showing a great amount of pitch and heave.

36

Figure 24: The EcoCAR 2 Vehicle Driving Over a the Large Bump Ride Profile

3.8 Damper Tuning

Vehicle dampers are a large tuning tool for the ride and handling of a vehicle. Passive dampers are typically oil filled cylinders that utilize a piston that moves through the oil with suspension movement. The velocity of the piston produces force at the damper, controlling the wheel motion. Current passive dampers used on production vehicles are either monotube or twin tube design. They are gas charged to keep the oil that is displaced by the movement of the piston from cavitating and reducing the damping performance.

37

Figure 25: Diagram Showing the Internal Layout of a Twin Tube Damper (source

Tenneco)

Currently, automotive manufacturers tune dampers subjectively along with other tunable suspension components. Each time a damper needs to be adjusted, it must be removed from the vehicle, disassembled, and rebuilt. The process of one damper rebuild could take up to 20 minutes. This must be repeated for each damper on the vehicle. The damping force is adjusted by thin metal shims on either the piston or additional valves added between the main cylinder and the gas reservoir.

38

Figure 26: Force vs Velocity Damper Dynomometer Curve (source Tenneco)

Figure 26 shows a force vs. velocity curve. This is the information that is used to describe the output of a damper at specific velocities. This information is measured using a damper dynamometer.

39

Figure 27: EcoCAR Team Rebuilding a Damper

Figure 27 shows the EcoCAR team rebuilding a damper to set it to factory settings. The work must be done is a clean environment to prevent any bit of debris from getting inside of the damper. Any obstruction can cause serious effects on the performance of the damper. This adds to the complexity of rebuilding dampers.

40

CHAPTER 4: INITIAL VEHICLE ASSEMENT

4.1 Baseline Vehicle Assessment

The team was given a 2013 Chevorlet Malibu Eco as a base vehicle for the OSU

EcoCAR 2 vehicle. The vehicle began as a new car as if it was purchased at a dealership.

This means that all of the OEM vehicle attributes were unmodified. The OEM 2013

Chevy Malibu weighs 3580 lb with a front to rear weight distribution of 59/41.

Table 2: Comparison of the Vehicle Mass From OEM to Competition

OEM Competition Difference Vehicle Vehicle Front Axle (lb) 2121.9 1980 -141.9 Rear Axle (lb) 1458.5 2514.6 1056.1 F/R % 59.26 44.05 15.211465 Total (lb) 3580.4 4494.6 914.2 CG Height (in) 22.31 21.92 -0.39

Along with the testing of the EcoCAR vehicle at TRC, a second OEM comparison vehicle was also tested in the same maneuvers. This testing provides the team with data

41 that can be used to compare the modified vehicle to the OEM vehicle, before and after tuning.

4.2 Competition and Post-Competition Vehicle Assessment

The stock vehicle has been modified greatly by removal of the stock powertrain and battery system. It has been replaced with the new OSU designed front hybrid powertrain and matching electronics. Nearly 900 lbs of batteries and supporting components have been added to the trunk of the car along with an electric motor and single speed transmission. The rear suspension has been replaced with a heavier Buick Lacrosse all- wheel drive capable suspension that has been modified to house the rear transmission and electric motor. The un-sprung mass has been lightened with lightweight tires and rims along with lighter brakes on the front axle.

During the dynamic consumer acceptability event, the vehicle was subjectively rated at a

6.5/10 which is considered a fair rating. While the front of the car was said to feel fine, the rear was said to be under damped. The vehicle also hit the jounce stops abruptly.

This evaluation validates the idea that the rear axle should be fairly under damped due to the added rear weight. It also validates that the front axle should feel similar to the OEM, as the weight over the front axle is similar to the OEM vehicle

42 4.3 Pre-Competition Stability Testing

An additional requirement for the EcoCAR competition is that the modified vehicle is safe to drive. This is determined by the ability of the car to pass a test called On Road

Safety Evaluation (ORSE). This test is comprised of manuvers involving hard accelerations, hard braking, double lane changes, and slaloms at low and high speeds. In order to guarantee that the EcoCAR vehicle would pass ORSE, the vehicle was tested at

TRC along with a base vehicle for comparison.

In order to determine if the vehicle is stable, the vehicle must be prone to understeer in a similar fashion to the stock vehicle. In this test a stock vehicle was driven through a constant speed, reducing radius turn until it was clear that the vehicle was not gaining lateral acceleration. This is characterized by either overwhelming understeer, or an unstable spin, in the case of oversteer. [11] The modified vehicle was then driven through the same test. Both vehicles were outfitted with VBOX instrumentation along with the 6 axis IMU. Equation (1) was used to calculate Understeer Gradient:

( )

(1) ( )

( )

43 ( )

( )

( )

Table 3 Shows that the vehicle exceeds the understeer gradient calculated from the experimental test results. This is acceptable from a safety standpoint.

Table 3: Understeer Gradient for Stock and Modified Car

Stock Car Modified Car Speed 35mph 55mph 35mph 55mph Understeer Gradient 6.5 2 9.2 5.3

Figure 28, Figure 30, Figure 32, and Figure 34 show the actual vehicle path followed calculated from GPS data during testing. Figure 29, Figure 31, Figure 33, and Figure 35 show a comparison of lateral acceleration to wheel toe angle. Data points have been added to show when the vehicle is determined to be unable to develop extra lateral acceleration with more steering wheel input.

44 Steady State Steering - Stock Car - 35mph - Driver Line -20

-30

-40

-50

Y (meter) -60

-70

-80

-90 60 65 70 75 80 85 90 95 100 105 X (meter)

Figure 28: 35mph Stock Car Driving Line

45 Steady State Steering - Stock Car - 35mph - Speed and Steering 40

30 Speed CAN [MPH] Speed VBOX [MPH] 20 Steering Toe Angle [deg] X: 3.34 Y: 8.926 MPH , Deg 10

0 0 1 2 3 4 5 6 7 Time (s) YawRate and Lateral Accleration 150

100 X: 3.34 50 Y: 100.6

YawRate [deg/s] 0 deg/s , Gx100 Lat Accel [Gx100] -50 0 1 2 3 4 5 6 7 Time (s)

Figure 29: 35mph Stock Car

46 Steady State Steering - Stock Car - 55mph - Driver Line 210

200

190

180

Y (meter) 170

160

150

140 80 90 100 110 120 130 140 150 160 170 180 X (meter)

Figure 30: 55mph Stock Car Driving Line

47 Steady State Steering - Stock Car - 55mph - Speed and Steering 60

Speed CAN [MPH] 40 Speed VBOX [MPH] Steering Toe Angle [deg] 20

MPH , Deg X: 2.47 Y: 4.708

0 0 1 2 3 4 5 6 Time (s) YawRate and Lateral Accleration 150 X: 2.47 Y: 108.9 100

50

0 deg/s , Gx100 YawRate [deg/s] Lat Accel [Gx100] -50 0 1 2 3 4 5 6 Time (s)

Figure 31: 55mph Stock Car

48 Steady State Steering - Modified Car - 35mph - Driver Line 65

60

55

50

45

40 Y (meter)

35

30

25

20 -40 -35 -30 -25 -20 -15 -10 -5 0 5 10 X (meter)

Figure 32: 35mph Modified Car Driving Line

49 Steady State Steering - Modified Car - 35mph - Speed and Steering 40

30 Speed CAN [MPH] 20 Speed VBOX [MPH] Steering Toe Angle [deg]

MPH , Deg 10

0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 Time (s) YawRate and Lateral Accleration 100

50

0 YawRate [deg/s]

deg/s , Gx100 Lat Accel [Gx100]

-50 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 Time (s)

Figure 33: 35mph Modified Car

50 Steady State Steering - Modified Car - 55mph - Driver Line 230

220

210

200

Y (meter) 190

180

170

160 80 100 120 140 160 180 200 220 X (meter)

Figure 34: 55mph Modified Car Driving Line

51 Steady State Steering - Modified Car - 55mph - Speed and Steering 60

40

X: 2.91 20 X: 2.07 Y: 4.614 Y: 7.157 Speed CAN [MPH]

MPH , Deg Speed VBOX [MPH] 0 Steering Toe Angle [deg] -20 0 1 2 3 4 5 6 7 Time (s) YawRate and Lateral Accleration 150 X: 2.07 Y: 88.42 100 X: 2.91 50 Y: 98.07

0 deg/s , Gx100 YawRate [deg/s] Lat Accel [Gx100] -50 0 1 2 3 4 5 6 7 Time (s)

Figure 35: 55mph Modified Car

On average, both the OEM and the modified vehicle show lateral accelerations above 0.9

G. In nearly all of the tests the lateral acceleration begins to plateau around 4 degrees of steering toe angle and has fully plateaued by 8 degrees of steering toe angle.

52 4.4 Possibilities for Tuning

A vehicle chassis is made up of many components that exist to provide a certain experience for the driver and passengers of the vehicle. This means that there are many options for tuning when looking to change the performance of the chassis. Typical tunable suspension components are springs, sway bars, dampers, jounce bumpers, control arm bushings, and tires.

For the EcoCAR vehicle, many of these tunable components require extra engineering design time, expensive parts, and long lead times in order to replace them. Components like tires and control arm bushings are too difficult to modify as a small engineering team. For this reason, damper tuning was chosen as the main adjustable component of the chassis for refinement.

Figure 36 shows the compression side of a typical damper dynamometer curve. This diagram shows the extreme adjustability potential of the damper performance. Typically a damper will have a low speed and high speed slope that is separated by a blow off point. In this case, the vehicle will not see extremely high speed wheel movements due to the nature of the competition. Because of this, the extremely high speed section will not affect the refinement.

53

Figure 36: Diagram Showing the Max Potential of Damper Curve Adjustment (source

Tenneco)

4.5 Difficulties in Tuning

Additional issues caused by the development process create difficulties when it comes to tuning the EcoCAR 2 chassis for ride and handling. The EcoCAR vehicle is constantly changing with the addition and subtraction of components that may or may not be planned. This requires an accurate log of vehicle changes. For example, if the VIMF test is run with the vehicle at an early stage in the development process, then there will most 54 likely be a great deal of changes to the mass of the vehicle which may render the VIMF data less accurate. This can be an issue when the testing costs time and money and needs to be re-done.

There is variation in all manufactured components. This means that no two dampers will ever measure the same when placed on a dynamometer due to their physical construction.

This is true for all of the components on a vehicle. Variations like this can make tuning difficult if the exact performance of each component is not known.

55

CHAPTER 5: RIDE AND HANDLING SIMULATION

5.1 Model Development

Building the vehicle model is a much simpler process than measuring and characterizing the vehicle. The physical measurements will be entered into CarSIM using the CarSIM

GUI. Once the model of the vehicle is thought to be complete, the validation process will begin.

Figure 37: CarSIM Main Screen for GUI 56

Figure 37 shows the main CarSIM screen. The notable items in this screen are access to the vehicle model and the procedure model under Test Specifications. To the right of this menu, the Run Control section initiates a run and controls additional output files to

Matlab and Excel. On the far right, post processing allows the user to animate the run into a video and create plots to view the analytical results. Figure 38 shows the vehicle model screen which is broken down into a sprung mass aerodynamics, brake, powertrain, steering, front and rear suspension model.

Figure 38: The Vehicle Model Screen in the CarSIM GUI

57 In most of the inputs, there are options to make the data more or less complex. Often a look up table of data can be exchanged for a simple scalar that provides a linear relationship. Because of this, CarSIM allows the development of many models of different accuracies.

5.1.1 Tires

There are numerous ways to enter tire data into CarSIM. In the case of EcoCAR 2, the force and moment data was delivered in the form of tables of data showing tire reaction forces during certain loads, slip angles, and camber angles. Because of this, the data was entered in the default tabular form. In the top left corner on Figure 39, a few tire parameters are required. It is important that the effective rolling radius and spring rate parameters are entered here.

The shear forces and moments section requires tables of longitudinal and lateral forces as well as aligning and overturning moment with respect to tire slip angle under a varying normal force. The inputs in the center of the screen require tire lateral force, aligning moment and overturning moment with respect to camber and varying normal forces.

Lastly, the dynamic properties section requires a few easily calculated scalars to be entered.

58

Figure 39: Tire Data Screen in the CarSIM GUI

5.1.2 Rigid Sprung Mass

The rigid sprung mass screen in the CarSIM GUI contains information that describes the sprung mass as a whole. This information includes location of the center of gravity, total sprung mass weight, height of the sprung mass from the center of the wheels, and sprung mass inertias. This information is all found using the VIMF and no additional measurement is required to complete this section of the CarSIM model. Figure 40 shows the sprung mass screen from the CarSIM GUI.

59

Figure 40: Rigid Sprung Mass Section of the CarSIM GUI

5.1.3 Steering

Most of the information in the steering section of the CarSIM GUI is difficult information to measure. For the EcoCAR vehicle, the only added information was the steering ratio look up table. Two tables were used. One represented the steering rack ratio, while the other represented the steering linkage ratio. This information was obtained using the

SPMD. The rest of the data that remains was left from the D-Class sedan template that is built into CarSIM. This is thought to be accurate enough for these purposes. Figure 41 shows the steering screen in the CarSIM GUI.

60

Figure 41: Steering Section of the CarSIM GUI

5.1.4 Front and Rear Suspension Kinematics

Suspension kinematics information describes the motion of the wheels during compression and rebound. Un-sprung mass, wheel center distances, and static toe and camber angles are the scalars that have been updated to the EcoCAR vehicle measurements. The longitudinal and lateral wheel movements are look up tables that contain data on the motion of the wheel during suspension travel. The wheel camber and toe look up tables have information on how the wheels camber and toe during wheel travel. Figure 42 shows the front kinematics screen from the CarSIM GUI. This is

61 similar to the screen for the rear axle. Nearly all of this data was measured using the

SPMD.

Figure 42: Front Kinematics Section of the CarSIM GUI

5.1.5 Front and Rear Suspension Compliance

The suspension compliance information contains data on how the wheel will react when forces are applies in numerous ways to the tire patch. This data is measured using the

SPMD. The left side of the screen, shown in Figure 43 is reserved for springs, dampers

62 and jounce bumpers. There are look up tables or simple scalar inputs depending on the depth of the measurement. In the case of the EcoCAR 2 vehicle, it was easiest to use the overall rate at the wheel instead of trying to separate out the jounce bumper from the spring. On the right of Figure 43 there is space for auxiliary roll moment, or sway bar data, and general suspension component compliance. This data is difficult to obtain without the use of the SPMD or a different kinematic and compliance test machine. The

SPMD does not calculate all of the compliances that CarSIM has the ability to compute.

Figure 43: Front Compliance Section of the CarSIM GUI

63 5.1.6 Aerodynamics, Brakes and Powertrain

Due to the constant speed nature of the tests, aerodynamics, brake and powertrain were left un-changed from the D-Class vehicle model that CarSIM has built in. Measuring these parameters for the model was thought to yield little additional accuracy for the effort required. The Rear drive-torque ratio was set to 0.5 as the EcoCAR 2 vehicle typically utilizes an equal torque split between the front and rear axles in all electric operation. All of the ride and handling testing was performed in all electric mode.

5.2 Model Validation

5.2.1 Validation Tests

Open loop testing was performed on the vehicle post-competition vehicle which was similar to the vehicle that was on the kinematics and suspension testing machine. Driver steering wheel angle and vehicle speed are used as inputs to the model. The outputs of the model are then compared to the experimental data from the actual vehicle.

5.2.1.1 Steady State Steering

Steady state steering was performed at two speeds, 35mph and 55mph. Once the vehicle was up to the desired testing speed, the steering wheel angle was slowly increased until the lateral acceleration was not increasing with steering wheel angle.

64 Lateral Acceleration Measured VS CarSIM 35mph Steady State Turn 1.2 VBox Data CarSIM Output at CG 1 CarSIM Output at VBox Location

0.8

0.6

0.4

Lateral AccelerationLateral (G) 0.2

0

-0.2 0 1 2 3 4 5 6 7 8 9 10 Time (sec)

Figure 44: Lateral Acceleration 35mph Steady State Turn Validation

65 Yaw Rate Measured VS CarSIM 35mph Steady State Turn 40 VBox Data 35 CarSIM Output

30

25

20

15 YawRate (deg/s) YawRate 10

5

0

-5 0 1 2 3 4 5 6 7 8 9 10 Time (sec)

Figure 45: Yaw Rate 35mph Steady State Turn Validation

66 Lateral Acceleration Measured VS CarSIM 55mph Steady State Turn 1.2 VBox Data CarSIM Output at CG 1 CarSIM Output at VBox Location

0.8

0.6

0.4

Lateral AccelerationLateral (G) 0.2

0

-0.2 0 2 4 6 8 10 12 14 Time (sec)

Figure 46: Lateral Acceleration 55mph Steady State Turn Validation

67 Yaw Rate Measured VS CarSIM 55mph Steady State Turn 25 VBox Data CarSIM Output 20

15

10 YawRate (deg/s) YawRate 5

0

-5 0 2 4 6 8 10 12 14 Time (sec)

Figure 47: Yaw Rate 55mph Steady State Turn Validation

While the results from the testing at 55mph are similar to those at 35mph, there is much more noise in the data due to the sliding of the tires across the road surface. The sliding is much more noticeable at 55mph than at 35 mph, and it causes a high frequency

“chopping” feeling at the front axle of the vehicle. This input appears in the data when the lateral acceleration plateaus and the car under steers greatly.

68 The CarSIM results follow similar trends to the results from the experimental testing. It appears that the CarSIM output does not reach as high of a value for lateral acceleration and yaw rate as the experimental data.

5.2.1.2 Double Lane Change

Lateral Acceleration Measured VS CarSIM 50 mph Double Lane Change 1.5 VBox Data CarSIM Output at CG CarSIM Output at VBox Location 1

0.5

0 Lateral AccelerationLateral (G)

-0.5

-1 -2 0 2 4 6 8 10 12 Time (sec)

Figure 48: Lateral Acceleration 50mph Double Lane Change Validation

69 Yaw Rate Measured VS CarSIM 50 mph Double Lane Change 25 VBox Data 20 CarSIM Output

15

10

5

0

-5 YawRate (deg/s) YawRate

-10

-15

-20

-25 -2 0 2 4 6 8 10 12 Time (sec)

Figure 49: Yaw Rate 50mph Double Lane Change Validation

5.2.1.3 Ride

The ride profile geometry was provided by TRC in order to show how the model vehicle performs when driven over the bump that was used to evaluate the vehicles experimentally. It was discovered that a simple trapezoid shape was not effective as a model for the geometry of the profile. The validation results were far from close. This led to the use of the actual measured profile data for the test. The file was opened and manipulated using OpenCRG 1.0.6. Data for the profile was taken every 5mm down the

70 length of the roadway. This created a rather choppy curve which was filtered with a floating average filter in order to reduce the small variances which could cause issues with test results.

Un-filtered Bump Geometry 0.3

0.25

0.2

0.15

0.1 Height (m) Height 0.05

0

-0.05

-0.1 50 100 150 200 250 300 350 Distance (m)

Figure 50: Large Bump Profile used for Ride Validation

71 Bump Geometry 0.35

0.3

0.25

0.2

0.15 Height (m) Height 0.1

0.05

0

-0.05 50 100 150 200 250 300 350 Distance (m)

Figure 51: Filtered Large Bump Profile used for Ride Validation

When the CarSIM results and the experimental testing data are overlapped it can be scene that the two share a trend in amplitude as well as frequency.

72 Pitch Rate Measured VS CarSIM Ride Large Bump 30 VBox Data 25 CarSIM Output

20

15

10

5

0 PitchRate (deg/s) -5

-10

-15

-20 0 5 10 15 20 25 Time (sec)

Figure 52: Pitch Rate 40mph Ride Bump Validation

73 Vertical Acceleration Measured VS CarSIM Ride Large Bump 1 VBox Data 0.8 CarSIM Output at CG CarSIM Output at VBox Location 0.6

0.4

0.2

0

-0.2 Vertical Acceleration (G)

-0.4

-0.6

-0.8 0 5 10 15 20 25 Time (sec)

Figure 53: Vertical Acceleration 40mph Ride Bump Validation

5.2.2 Possible Reasons for Differences in Simulated Results

A model of a physical system is always an estimation to some extent. This means that it is impossible for the model to fully represent the physical system. A model always has room for improvement. In general, there are several systems there are very difficult to model on a vehicle. Tires are the most difficult major system. The tire testing done for the EcoCAR 2 vehicle was done on a belt system with a specific surface that will not match the surface at the TRC Vehicle Dynamics Area. This especially causes problems when the tire begins to slide as the performance becomes very dependent on the surface 74 properties of the asphalt. In addition to this, there are many environmental factors such as temperature that change how some components perform. This is especially true with tires and dampers. These factors are out of control of the EcoCAR team during testing, and the CarSIM model does not have a temperature input for each component. This is an example of a small way that the model could improve.

In some situations, data doesn’t exist to the required inputs of the model and CarSIM must extrapolate to find the data which leaves room for error. This is the case for the high load on the outside tire during steady state and double lane change testing. This load exceeds the data that was entered for that tire.

In the case of this project, several parts of the model were assumed to be close to the average data that CarSIM has for the vehicle class similar to the Chevrolet Malibu. This is data that is difficult to measure and doesn’t have a huge impact on the way the tests are run. An example of this is the aerodynamics of the vehicle. Because the tests were run at a constant speed, general aerodynamic properties were used. Measuring the vehicle for actual aerodynamics data would be an improvement, but in this case was not worth it to the team to accomplish.

75

CHAPTER 6: REFINEMENT

6.1 Development of Objectives

The first step of the refinement of the ride and handling of the EcoCAR vehicle is to select metrics by which the vehicle will be judged. The main purpose of this project is to develop a method for model based damper calibration. The metrics that are chosen will only be used to prove the concept of the process used. For this reason, the metrics will be chosen based on driver comfort and perception of the vehicle. The simplest comfort metrics to consider are the planar accelerations of the passengers as well as the pitch and roll motions that the passengers can feel. Another useful metric is driver work. This can be determined in the transient test using driver steering wheel velocity. If the driver model must move the steering wheel faster, then the driver is working harder to keep the vehicle on its path.

76 Table 4: Metrics Chosen to Objectively Evaluate the Vehicle Performance

Max Lateral Acceleration (G)

Double Lane Change Max Roll Angle (deg)

Max Steering Wheel Rate (deg/s)

Max Vertical Acceleration at Beginning of Bump (G)

Max Vertical Acceleration End of Bump (G) Ride Max Pitch At Beginning of Bump (Deg)

Max Pitch At End of Bump (Deg)

The ride metrics were split between the beginning and the end of the bump to separate the effects of the vehicle going in the upwards direction and the vehicle dropping off of the bump.

6.2 Un-Damped Refinement

Before tuning for the selected metrics, the un-damped natural frequency must be returned to near stock. With the added weight of the components added to the EcoCAR 2 vehicle, the ground clearance and the natural frequency of the rear axle suffered. The trim height and ride frequency were too low. In order to remedy this situation, new springs were designed for the rear of the vehicle to replace the OEM springs. The objective of the new

77 springs was to set the rear ride frequency of the vehicle to the OEM un-damped ride frequency as well as set the trim height to an acceptable level.

This task requires measurement of any components that contribute to the spring rate of the suspension at the wheels of the OEM and modified vehicle. First a rear corner of the

OEM vehicle was characterized based off of its spring rate, auxiliary bushing and gas spring rate, motion ratio and tire vertical stiffness. This allows for the calculation of the ride rate, which leads us to calculate natural frequency with the addition of mass.

Because the ride frequency for the EcoCAR must match the base vehicle, the ride frequency must be calculated first for the OEM vehicle. First, to calculate ride rate, the wheel rate must first be calculated. In the case of the EcoCAR 2 vehicle, spring force at trim height data was given to the team from the OEM vehicle. Using the weight of the vehicle on the rear wheels and a measurement of the auxiliary wheel rate, a motion ratio can be calculated.

𝐹 𝐹 ∗ 𝑀 𝑅 + 𝐹 (2)

𝑥 𝑀 𝑅 (3) 𝑥

78 𝐹 𝑥 ∗ (4)

𝑥 𝑥 𝑥 ∗ ∗ ∗ ( ) + (5) 𝑥 𝑥 𝑥

∗ (𝑀 𝑅 ) + (6)

Once a spring rate, motion ratio, and auxiliary wheel rate can be measured or calculated, the rate at the wheels can be determined using equation (7). The ride rate can then be found using the spring rate at the wheel along with the vertical spring rate of the tire and equation (8). Equation (9) is then used to calculate the natural frequency of the rear axle. The OEM rear suspension natural frequency was found to be 1.45 Hz.

∗ 𝑅𝑅 (10) +

𝑅𝑅 𝑅 𝑅 ( )

𝑅 ( )

79

𝑅 ( )

𝑅𝑅 √ (11) 𝑀

𝐹 ( )

𝑀 𝑀 ( )

Once a target natural frequency can be determined, the equations can be used in reverse to decide on a spring rate that will give the rear suspension of the EcoCAR 2 vehicle a natural frequency of 1.45 Hz.

80 Spring Rate Relative to Ride Frequency 1.6 Frequency vs Spring Rate Ideal Frequency 1.5

X: 117 1.4 Y: 1.45

1.3

1.2 Ride Frequency Ride

1.1

1

0.9 30 40 50 60 70 80 90 100 110 120 130 Spring Rate

Figure 54: Rear Spring Ride Frequency Calculation

Using Matlab, the natural frequency of spring rates 35 N/mm to 130 N/mm were calculated and the rate that matched 1.45 Hz was 117 N/mm. Once the rate is calculated, the spring length is calculated by comparing the compressed length of the new spring at trim height to the compressed length of the original spring at trim height. The compressed length of the new spring should be set to match the compressed length of the old spring.

81 6.3 Overall Simulation Process

To understand how different damper curve parameters would affect the performance of the vehicle, a design of experiments was developed using CarSIM to obtain the outputs.

The seven metrics were used to create seven response surface equations based on the results of the DOE. The response surface equations were then used to complete an optimization which minimized the values of the response surface equations and returned the damper curves that were considered to be most optimal.

82 The

Figure 55: Diagram for Overall Refinement Process

6.4 Inputs

The inputs to the DOE simulation were damper curves. The damper curves were broken down into a high and low speed damping zone for both compression and rebound. The location of the blow off point and the high speed slope were chosen as separate variables

83 to describe the damper curve. Compression and rebound are made up each of three variables, making one entire damper curve a combination of six independent variables.

Because the front dampers and the rear dampers are typically different front to back, but similar side to side, the damping of a full car can be described with 12 factors.

Figure 56: Damper Curve Variable Breakdown

84 The model type chosen for the DOE was quadratic with interactions based on experience from an advisor. This requires the total number of 12 factor sets to be 91 based on equation (12). 182 sets were generated as an extra factor of safety.

+ ∗ + ∑ (13)

𝐹

Figure 57: Damper Curve Variable Adjustment Ranges 85

The limits of the generated variables were developed using engineering judgment and experience from team sponsors at Tenneco in Monroe, MI. Figure 57 shows the range of the limits used for the generation of damper curves. The D-optimal 3 layer space filling

Matlab function was used for the input data generation. This is shown in Figure 58 in an easy to visualize 3D plot. This plot only contains spacing for three of the 12 factors as a

12 dimensional plot is difficult to visualize.

Figure 58: 3D Point Distribution for three of the 12 input factors

Figure 59 and Figure 60 show the all of the damper curves for the front and the rear of the vehicle that were generated for the DOE.

86 Damper Curves to be Tested - Front Axle 5000

4000

3000

2000

1000

0 Force Force (N) -1000

-2000

-3000

-4000

-5000 -2000 -1500 -1000 -500 0 500 1000 1500 2000 Velocity (mm/s)

Figure 59: All Front Damper Curves Simulated

87 Damper Curves to be Tested - Rear Axle 5000

4000

3000

2000

1000

0 Force Force (N) -1000

-2000

-3000

-4000

-5000 -2000 -1500 -1000 -500 0 500 1000 1500 2000 Velocity (mm/s)

Figure 60: All Rear Damper Curves Simulated

6.5 Response Surface Equations

Response surface equations represent the seven metrics chosen for the optimization. The

182 sets of 12 factors were run through CarSIM for both the open loop ride and closed loop double lane change test. The seven metrics were gathered for each test and a quadratic fit was calculated to create response surface equations for each metric. The

88 response surfaces will be used to predict the output value that would have come out of the

CarSIM simulation.

Figure 61: Response Surface Fitting for Maximum Vertical Acceleration at the Beginning

of the Ride Bump

Figure 61 and Figure 62 show plots that represent the quality of the response surface fit.

RSE predicted value is plotted against the actual simulation value. The closer the points are to the black line, the better the quality of the RSE fit.

89

Figure 62: Response Surface Fitting for Maximum Lateral Acceleration during the

Double Lane Change Test

6.6 Response Surface Validation

As the response surfaces are estimated data fits of actual simulation data, and their accuracy must be validated to ensure that their output values are useful. For this, 15 random input sets of factors were chosen to compare the output results from the response surface equation (RSE) to the CarSIM model. The maximum and minimum difference in results for each of the response surfaces are listed in Table 5. This shows how close the

RSE output could be to the actual CarSIM output, as well as how far off it could be.

90 Table 5: Response Surface Equation Validation Results

Percent of Typical RSE Value Test Type Response Surface Output Difference Value for Test 10 Max Lateral Acceleration Max 0.0054 0.62% 0.88091 (G) Min 0.0002 0.02% Double Max 0.0285 0.68% Lane MaxRollAngle (deg) 4.1645 Min 0.0025 Change 0.06% Max 3.4642 1.34% MaxStWhRate (deg/s) 258.14 Min 0.2699 0.10% Max 0.0233 3.87% Max Vertical Acceleration 0.60345 at Beginning of Bump (G) Min 0.0010 0.17% Max 0.0482 6.06% Max Vertical Acceleration 0.7954 End of Bump (G) Min 0.0007 0.09% Ride Max Pitch At Beginning of Max 0.0359 1.21% 2.956 Bump (Deg) Min 0.0009 0.03% Max Pitch At End of Bump Max 0.2345 8.07% 2.9044 (Deg) Min 0.0055 0.19%

also shows the values for test number 10 of the 15 verification tests. Figure 64, Figure

66, Figure 68, Figure 70, Figure 72, Figure 74, and Figure 76 show the actual CarSIM output from the trials with a data curser to compare to the value from Table 5. This data is useful in showing that the reponse surface equations can be used to actually predict how certain damper curve inputs could cause the vehicle to respond.

91 Maximum Lateral Acceleration for DLC RSE Verification 0.94

CarSIM Output 0.93 RSE Output

0.92

0.91

0.9

Output Output Value (G) 0.89

0.88

0.87

0.86 0 5 10 15 Test Number

Figure 63: A Comparison of Max Lateral Acceleration using the Outputs from 15 Sets of

Input Factors

92 Lateral Acceleration at During DLC for Verification Run 10 1

0.8 X: 237 Y: 0.8783 0.6

0.4

0.2

0

-0.2

Lateral AccelerationLateral (G) -0.4

-0.6

-0.8

-1 0 50 100 150 200 250 300 350 Data Points

Figure 64: The Full Output of Lateral Acceleration from CarSIM for Run 10

93 Maximum Roll Angle RSE Verification 4.4

CarSIM Output 4.35 RSE Output

4.3

4.25

4.2 Output Output Value (deg) 4.15

4.1

4.05 0 5 10 15 Test Number

Figure 65: A Comparison of Max Roll Angle using the Outputs from 15 Sets of Input

Factors

94 Roll Angle During DLC for Verification Run 10 5 X: 236 Y: 4.153 4

3

2

1

0

-1 Roll Angle Roll (deg)

-2

-3

-4

-5 0 50 100 150 200 250 300 350 Data Points

Figure 66: The Full Output of Roll Angle from CarSIM for Run 10

95 Maximum Steering Wheel Rate RSE Verification 268 CarSIM Output RSE Output 266

264

262

260 Output Output Value (deg/sec)

258

256 0 5 10 15 Test Number

Figure 67: A Comparison of Max Steering Wheel Rate using the Outputs from 15 Sets of

Input Factors

96 Steering Rate During DLC for Verification Run 10 300 X: 197 Y: 262

250

200

150

100

50

0

Steering Rate (deg/s) Rate Steering -50

-100

-150

-200 0 50 100 150 200 250 300 350 Data Points

Figure 68: The Full Output of Steering Wheel Rate from CarSIM for Run 10

97 Maximum Pitch Angle at Beginning of Bump RSE Verification 3 CarSIM Output RSE Output

2.95

2.9

2.85 Output Output Value (deg)

2.8

2.75 0 5 10 15 Test Number

Figure 69: A Comparison of Max Pitch Angle at the Beginning of the Bump using the

Outputs from 15 Sets of Input Factors

98 Pitch at Beginning of Bump for Verification Run 10 0.5

0

-0.5

-1

-1.5 Pitch (deg)

-2

-2.5 X: 450 Y: -2.894

-3 0 100 200 300 400 500 600 Data Points

Figure 70: The Full Output of Pitch Angle at the Beginning of the Bump from CarSIM

for Run 10

99 Maximum Pitch Angle at End of Bump RSE Verification 3.5 CarSIM Output 3.4 RSE Output

3.3

3.2

3.1

3 Output Output Value (deg)

2.9

2.8

2.7 0 5 10 15 Test Number

Figure 71: A Comparison of Max Pitch Angle at the End of the Bump using the Outputs

from 15 Sets of Input Factors

100 Pitch at End of Bump for Verification Run 10 3

X: 65 2.5 Y: 2.881

2

1.5

1

0.5 Pitch (deg)

0

-0.5

-1

-1.5 0 50 100 150 200 250 300 350 400 Data Points

Figure 72: The Full Output of Pitch Angle at the End of the Bump from CarSIM for Run

10

101 Maximum Vertical Acceleration at Beginning of Bump RSE Verification 0.68 CarSIM Output 0.66 RSE Output

0.64

0.62

0.6

0.58

Output Output Value (deg) 0.56

0.54

0.52

0.5 0 5 10 15 Test Number

Figure 73: A Comparison of Max Vertical Acceleration at the Beginning of the Bump

using the Outputs from 15 Sets of Input Factors

102 Vertical Acceleration at Beginning of Bump for Verification Run 10 0.6

X: 448 Y: 0.5594 0.4

0.2

0

-0.2

Vertical Acceleration (G) -0.4 X: 459 Y: -0.5593

-0.6

-0.8 0 100 200 300 400 500 600 Data Points

Figure 74: The Full Output of Vertical Acceleration at the Beginning of the Bump from

CarSIM for Run 10

103 Maximum Vertical Acceleration for at End of Bump RSE Verification 0.9 CarSIM Output RSE Output

0.85

0.8

0.75 Output Output Value (G)

0.7

0.65 0 5 10 15 Test Number

Figure 75 A Comparison of Max Vertical Acceleration at the End of the Bump using the

Outputs from 15 Sets of Input Factors

104 Vertical Acceleration at End of Bump for Verification Run 10 0.8 X: 72 Y: 0.7902 0.6

0.4

0.2

0

Vertical Acceleration (G) -0.2

-0.4

-0.6 0 50 100 150 200 250 300 350 400 Data Points

Figure 76: The Full Output of Vertical Acceleration at the End of the Bump from CarSIM

for Run 10

6.7 Optimization

Once the response surface equations are validated and considered to be accurate enough to be useful, a minimization optimization can be performed to determine which inputs will return the lowest output value from each response surface. For example, to reduce the amount of vertical acceleration that the driver of the vehicle feels, the damper curve

105 inputs that can obtain the lowest vertical acceleration in the ride profile test would be the most sought after. The minimization allows us to do this for each RSE. Table 6 shows the results of the individual minimizations.

The optimization of individual response surface equations can be helpful for understanding trends when comparing inputs to outputs, but not useful for developing a balance between all of the vehicle metrics. For this, a minimization optimization can be run with all of the response surface equations in mind. This requires a function that dictates the importance of each RSE and can place a penalty on certain output values.

This is called a cost function, or fitness function. This function combines a normalized output of each RSE along with a penalty and weight coefficient. The penalty coefficient can be used to forbid certain output values by adding a weight to those values. The weight coefficient applies a general weight to each RSE output value. The result of the function is an output from 0 to 1.

𝑀 (14)

𝐹 ∗ 𝐹 ∗ ̅ + 𝐹 ∗ 𝐹 ∗ ̅ + 𝐹 ∗ 𝐹 ∗ ̅ + 𝐹 ∗ 𝐹 ∗ ̅ + 𝐹 ∗ 𝐹 ∗ ̅

𝐹 𝐹

𝐹 𝐹

̅ ̅ ̅ ̅ ̅ 𝑅

106

If the cost function can appropriately choose input parameters that give the best output result, then the majority of the difficulty is in selecting the weighting factors that determine the importance of each RSE. In the case of the EcoCAR competition, the point structure for the dynamic events could be used to dictate the weighting factors as it would be better to bias the vehicle performance towards the events that have the most point potential.

6.8 Results

The individual response surface equation minimization showed that it was able to return results near the bottom of the physical range for nearly all seven response surface equations. Table 6 shows the output value of each minimization. These values can be compared to the output physical limits to understand their success. Not all of the values were lower than some of the randomized inputs used during the DOE simulations with

CarSIM. This shows that there may be some local maximums and minimums in the RSE fit. While these results prove that the process can be effective, the fit could be improved for better results.

107 Table 6: Individual RSE Output Minimization Results

Output Response Surface Equation Value

Max Lateral Acceleration (G) 0.8659

Max Roll Angle (deg) 3.731

Max Steering Wheel Rate (deg/s) 235.5

Max Vertical Acceleration at Beginning of Bump (G) 0.4612

Max Vertical Acceleration End of Bump (G) 0.6344

Max Pitch At Beginning of Bump (Deg) 2.716

Max Pitch At End of Bump (Deg) 2.017

The optimization of the combined response surface equations shows a notable minimization of each RSE output across the board. The weighting coefficients for the cost function were chosen based on the EcoCAR 2 point structure which values ride comfort above handling comfort. For this reason, the ride events were weighted higher than the handling events. This can be seen in Table 7 along with the results of the optimization. Once the optimal damper curves were selected, the ride and double lane change events were run using the optimized curves and the seven outputs from those simulations are shown in Table 7. These outputs are compared to the possible limits of the outputs determined by the range of results from the initial CarSIM simulations. If the 108 output value from CarSIM was the same as the lowest value in the simulated range then the percent value between max and min would be 0%. If the output value from CarSIM was the same as the highest value in the simulated range then the percent value between max and min would be 100%. Nearly all seven values are under 50% and average around

23% of the physical range.

Table 7: Results of Overall Optimization

Percent of Physical CarSIM Test Response Surface Optimizer Value Output Output Type Output Weight Between Max Limits Value and Min Max Lateral Max 0.97 5% 0.855 17.9% Acceleration (G) Min 0.83 Double Max 4.70 Lane Max Roll Angle (deg) 10% 3.72 2.0% Min 3.70 Change Max Steering Wheel Max 268.00 5% 245.3 15.9% Rate (deg/s) Min 241.00 Max Vertical Max 0.77 Acceleration at 20% 0.4949 16.6% Min 0.44 Beginning of Bump (G) Max Vertical Max 0.95 Acceleration End of 20% 0.72902 40.3% Min 0.58 Ride Bump (G) Max Pitch At Beginning Max 3.18 20% 2.8242 20.9% of Bump (Deg) Min 2.73 Max Pitch At End of Max 3.75 20% 2.8392 50.8% Bump (Deg) Min 1.90

109 The cost function weights did not show a direct correlation with the percent value between min and max number. The handling outputs on average were minimized lower than the ride outputs while their percentage was set to be lower. The weights on the cost function do not affect the results as dramatically as expected. This could be due to a large minimum in the cost function that does not change much regardless of weighting factors. Despite this problem, the optimizer was still able to choose a two damper curves that caused the vehicle to show positive results in the simulated testing through CarSIM.

In the case of EcoCAR, the engineers could be confident that these damper curves would achieve relatively good results without having to do any physical testing with the vehicle.

Figure 77: Shows the damper curves that have been chosen by the optimizer.

110 Final Damper Results 8000 Front Damper Curve 6000 Rear Damper Curve

4000

2000

Force Force (N) 0

-2000

-4000

-6000 -2000 -1500 -1000 -500 0 500 1000 1500 2000 Velocity (mm/s)

Figure 77: Damper Curves Calculated from the Optimization Minimization

111 Maximum Lateral Acceleration in DLC 1

0.8 X: 234 Y: 0.8541 0.6

0.4

0.2

0

-0.2

Lateral AccelerationLateral (G) -0.4

-0.6 X: 160 Y: -0.8462 -0.8

-1 0 50 100 150 200 250 300 350 Data Points

Figure 78: Lateral Acceleration Output from CarSIM for the Optimized Damper Curves

112 Maximum Roll Angle in DLC 4 X: 236 Y: 3.413

3

2

1

0

Roll Angle Roll (deg) -1

-2

-3 X: 174 Y: -3.72

-4 0 50 100 150 200 250 300 350 Data Points

Figure 79: Roll Angle Output from CarSIM for the Optimized Damper Curves

113 Maximum Steering Wheel Angle Rate in DLC 250 X: 197 200 Y: 245.3

150

100

50

0

-50 Steering Wheel Rate (deg/s) Rate Wheel Steering -100

-150

-200 0 50 100 150 200 250 300 350 Data Points

Figure 80: Steering Wheel Rate Output from CarSIM for the Optimized Damper Curves

114 Maximum Vertical Acceleration at Beginning of Bump 0.5 X: 448 0.4 Y: 0.4949

0.3

0.2

0.1

0

-0.1

Vertical Acceleration (G) -0.2

-0.3

X: 458 -0.4 Y: -0.4718

-0.5 0 100 200 300 400 500 600 Data Points

Figure 81: Vertical Acceleration for the Beginning of the Bump Output from CarSIM for

the Optimized Damper Curves

115 Maximum Vertical Acceleration at End of Bump 1

X: 73 0.8 Y: 0.729

0.6

0.4

0.2

Vertical Acceleration (G) 0

-0.2

-0.4 0 50 100 150 200 250 300 350 400 Data Points

Figure 82: Vertical Acceleration for the End of the Bump Output from CarSIM for the

Optimized Damper Curves

116 Maximum Pitch Angle at Beginning of Bump 0.5

0

-0.5

-1

-1.5 Pitch Angle (deg)

-2

-2.5 X: 450 Y: -2.824

-3 0 100 200 300 400 500 600 Data Points

Figure 83: Pitch Angle for the Beginning of the Bump Output from CarSIM for the

Optimized Damper Curves

117 Maximum Pitch Angle at End of Bump 3

X: 63 2.5 Y: 2.839

2

1.5

1

0.5 Pitch Angle (deg)

0

-0.5

-1 0 50 100 150 200 250 300 350 400 Data Points

Figure 84: Pitch Angle for the End of the Bump Output from CarSIM for the Optimized

Damper Curves

118

CHAPTER 7: CONCLUSIONS AND FUTURE WORK

7.1 Conclusions

In the push to shorten the development time of vehicles, model based calibration of vehicle dampers can improve the chassis tuning process of a vehicle. For the EcoCAR team, model based damper calibration means that ride and handling considerations would begin in the first year of the competition and mature with the rest of the vehicle systems.

The vehicle can even be tuned without having a working vehicle.

This process involves measuring and characterizing the EcoCAR Vehicle and deciding on what components are feasible to adjust, then developing a vehicle model and validating the vehicle model. Finally the process involves generating a design of experiments, creating response surface equations, and developing an optimizer that predicts the best chassis attributes.

The process of objectively tuning vehicle dampers has been proven to produce useful results. The steps that make up the process can be repeated for EcoCAR teams in the future. The metrics used for this proof of concept were fairly basic metrics that could be nearly any part of the chassis. The requirements for each EcoCAR competition will dictate what objective tuning metrics should be used. These requirements will also 119 dictate the way that the cost function is weighted. The next challenge is to understand how to ask the optimizer the right question, to get the most desirable answer.

7.2 Future Work

Many aspects of model based damper calibration can be improved. The development of new objective metrics and cost function weights that are developed specifically for a future EcoCAR competition would be a large and extremely important project. This project shows that the process of developing a model of the vehicle can be useful. To further this project, metrics could be developed to more accurately represent what to adjust in order to get the most points to win the competition.

Additional tuning metrics should be taken into account for tuning future vehicles. If vehicle dynamics will be included in early design considerations, then more time will allow for more possibilities in tuning. Jounce bumpers and sway bars could be added to the damper tuning possibilities. Instead of just tuning for driver and passenger comfort, tuning could improve vehicle performance as well.

The process of model based damper calibration could be used to simulate semi-active dampers. This is especially advantageous, because the semi-active damper algorithm could be tuned using simulation, before ever being driven on the road. The possibilities of model based chassis calibration are endless.

120

BIBLIOGRAPHY

[1] T. D. Gillespie, Fundamentals of Vehicle Dynamics. Warrendale, PA: Society of Automotive Engineers Inc, 2014.

[2] A. C. Nuti, “Objective metric x subjetive evaluation,” SAE Technical Paper, 2003.

[3] D. A. Crolla, D. C. Chen, J. P. Whitehead, and C. J. Alstead, “Vehicle handling assessment using a combined subjective-objective approach,” SAE Technical Paper, 1998.

[4] T. L. Brown, S. T. Mear, N. E. Moore, S. M. Kannapan, K. M. Marshek, J. Cuderman, and K. Efatpenah, “An experimental procedure for estimating ride quality for passive and semi-active suspension automobiles,” SAE Technical Paper, 1992.

[5] A. F. Naude and J. L. Steyn, “Objective evaluation of the simulated handling characteristics of a vehicle in a double lane change manoeuvre,” SAE Technical Paper, 1993.

[6] I. A. Badiru and M. W. Neal, “Use of DFSS Principles to Develop an Objective Method to Assess Transient Vehicle Dynamics,” SAE International, Warrendale, PA, 2013-01-0708, Apr. 2013.

[7] N. J. Durisek, K. J. Granat, G. J. Heydinger, and D. A. Guenther, “Repeatability and Bias Study on the Vehicle Inertia Measurement Facility (VIMF),” SAE Technical Paper, 2009.

[8] R. A. Bixel, G. J. Heydinger, N. J. Durisek, D. A. Guenther, and S. J. Novak, “Developments in Vehicle Center of Gravity and Inertial Parameter Estimation and Measurement,” SAE Technical Paper, 1995.

[9] D. A. Coovert, H. F. Chen, and D. A. Guenther, “Design and Operation of a New- Type Suspension Parameter Measurement Device,” SAE Technical Paper, 1992.

121 [10] C. Constant, “The test (or VDA test),” Car Engineer. .

[11] “Vehicle Dynamics Terminology,” SAE International Surface Vehicle Recommended Practice, J670_200801, Jan. 2008.

122

APPENDIX A: LIST OF SYMBOLS AND ABBREVIATIONS

CAD Computer Aided Design

DCA Dynamic Consumer Acceprability

DOE Department of Energy

E85 85% ethanol and 15% gasoline fuel by volume

ISO Internation Organization for Standardization

NVH Noise, Vibrations, and Harshness

OEM Original Equipment Manufacturer

PHEV Plug-in Hybrid Electric Vehicle

RSE Response Surface Equations

SPMD Suspension Parameter Mesaurement Device

TRC Transportation Research Center

VIMF Vehicle Inertial Measurement Facility

123

APPENDIX B: FIGURES

Figure 85: Baseline Vehicle VIMF Data

124

Figure 86: Modified Vehicle VIMF Data

125

Figure 87: Competition DCA Results

126