A STUDY OF THE PROPERTIES THAT INFLUENCE VEHICLE ROLLOVER PROPENSITY Except where reference is made to work of others, the work described in this thesis is my own or was done in collaboration with my advisory committee. This thesis does not include proprietary or classified information Randall John Whitehead Certificate of Approval: George T. Flowers Professor Mechanical Engineering David M. Bevly, Chair Assistant Professor Mechanical Engineering Nels Madsen Associate Professor Mechanical Engineering Stephen L. McFarland Dean Graduate School A STUDY OF THE PROPERTIES THAT INFLUENCE VEHICLE ROLLOVER PROPENSITY Randall John Whitehead A Thesis Submitted to the Graduate Faculty of Auburn University in Partial Fulfillment of the Requirements for the Degree of Masters of Science Auburn, Alabama December 16, 2005 iii A STUDY OF THE PROPERTIES THAT INFLUENCE VEHICLE ROLLOVER PROPENSITY Randall John Whitehead Permission is granted to Auburn University to make copies of this thesis at its discretion, upon the request of individuals or institutions and at their expense. The author reserves all publication rights. Signature of the Author Date of Graduation iv THESIS ABSTRACT A STUDY OF THE PROPERTIES THAT INFLUENCE VEHICLE ROLLOVER PROPENSITY Randall John Whitehead Master of Science, December 16, 2005 (B.M.E., Auburn University, 2003) 180 Typed Pages Directed by David M. Bevly In this thesis, a vehicle?s load condition is varied to investigate its impact on roll stability and a stability threshold is derived empirically using vehicle simulation. A vehicle model is developed and simulated using MATLAB. Experiments performed by the National Highway Transportation and Safety Administration (NHTSA), are used to validate the simulation. Data from these experiments is also used to validate a stability threshold developed from the simulation. Scaled passenger vehicles in conjunction with computer simulation have proven to be a valuable tool in determining rollover propensity. The stability threshold is also validated by scaled vehicle experiments. This is made possible with the lower cost and increased safety of using a scaled vehicle versus full size passenger vehicles. A simple electronic stability control (ESC) is then developed to keep the scaled vehicle within the v stability threshold. The ESC is tested using varying vehicle properties with a constant vehicle model to see how these property changes affect the ESC?s effectiveness to prevent rollover. The ESC is then implemented with an Intelligent Vehicle Model (IVM) which updates the controller?s vehicle model as vehicle properties such as loading conditions change. It is shown that an IVM greatly increases the success of ESC in keeping the vehicle in the stability region. vi ACKNOWLEGEMENTS This thesis is dedicated to my Dad and Mom, Larry and Judy Whitehead, whose support throughout the past six years allowed this research to be conducted. Special thanks to my faulty advisor Dave Bevly. You guided me through this process and spurred me on to completion. Thank you for the opportunity to further my education. To the Auburn University GAVLAB, thank you for the technical advice, physical data, and good laughs. To Rob Daily, lab manager, your MATLAB expertise and wealth of knowledge on all other technical matters was invaluable to my studies. To Christof Hamm, the ?token? lab EE, without his knowledge of circuits and computer software, I would have had to learn things the ?hard way.? To Will Travis, thanks for your hard work on our research together. To Matt Heffernan, the ?intellectual? discussions about vehicle dynamics and racing aroused my curiosity and motivated me to search for answers. To the rest of the Wilmore gang, Evan Gartley, Ren Flennigan, and Paul Pearson, thanks for the company and long nights cramming for exams and final projects. Thanks to the ?Dungeon? inmates who fought valiantly beside me against the swarms of fruit flies. Thanks to Dr. George Flowers and Dr. Nels Madsen, ?Mad Dog,? for being on my thesis defense committee. Dr. Flowers thanks for revelation that Diet Coke is great, vii even at 8 a.m. classes. Dr. Madsen thanks for three seasons of Hitchcock League baseball, and enlightening me to the power of the FBD ( 0??F )! Thanks to Undergrad research assistants for staying up late with me and on weekends to get projects complete. Thanks to Taylor Owens for all the engineering jokes and teaming up with me in classes and triathlons. Thanks to the Auburn Formula SAE Team for getting me started and interested in vehicle dynamics. Thanks to Bradley Kirkland for being a good friend and roommate. Thanks to all my other friends who gave me great memories in my years at Auburn University. Thanks to my LORD, Jesus Christ, who taught me to trust and follow. There is peace in knowing that no matter how bad I mess things up, He never gives up on me. That is where I find hope. He directed all my paths. Proverbs 3:5-6. viii Style of Journal Used: ASME Journal of Dynamic Systems, Measurement, and Control Computer Software Used: Microsoft Word 2003 ix TABLE OF CONTENTS LIST OF FIGURES...????????????????????????? xi LIST OF TABLES.?????????????????????????? xv 1. INTRODUCTION 1.1 Motivation????????????????????????.. 1 1.2 Background and Literature Review.??...??.????????..... 2 1.3 Purpose of Thesis and Contribution????????...?????? 8 1.4 Outline of Thesis?????????????????????? 9 2. VEHICLE MODEL 2.1 Introduction?...??.???????????????????? 11 2.2 Bicycle Model?????...?????????????????. 12 2.3 Roll Model?????....??????????????????. 15 2.4 Lateral Weight Transfer and Normal Wheel Loads..?..??????... 20 2.5 The Pacejka Tire Model: The ?Magic? Formula?????????? 22 2.6 Garage?????????????????????????? 28 2.7 Summary of Model Assumptions?.??????????????. 29 2.8 Summary and Conclusion??????????????????.. 30 3. VEHICLE MODEL VALIDATION 3.1 Introduction?...??.???????????????????? 32 3.2 Steady State Analysis versus Transient Analysis?????????. 33 x 3.3 NHTSA Phase IV Comparison????????????????. 42 3.4 Auburn University GAVLAB Blazer?????????????? 56 3.5 Lateral Weight Transfer Simulation??????????????.. 59 3.6 Summary and Conclusion??????????????????.. 64 4. ANALYSIS OF VEHICLE ROLLOVER STABILITY THRESHOLD 4.1 Introduction????????.....??...??.?????????... 65 4.2 NHTSA Phase IV Loading Conditions?????????????.. 66 4.3 Stability Limit Development?????????????????. 72 4.4 Tire Property Variation???????????????????.. 80 4.5 Summary and Conclusion??????????????????.. 83 5. ELECTRONIC STABILITY CONTROL DEVELOPMENT 5.1 Introduction????????.....??...??.?????????... 86 5.2 Electronic Stability Control??...??????????????? 87 5.3 Intelligent Vehicle Model ??...???????????????.. 90 5.4 Summary and Conclusion??????????????????.. 95 6. IMPLEMENTATION ON A SCALED VEHICLE 6.1 Introduction????????.....??...?.????????.... 97 6.2 Motivation for using Scaled Vehicles???????????? 98 6.3 Scaled Vehicle Description???????..??????.??.. 99 6.4 Data Acquisition???????????.???????..?... 104 6.5 Simulation and Experiment Vehicle Dynamic Comparison???? 108 6.6 Stability Limit Development??????????????...?. 110 6.7 Stability Limit Validation??????????...?????...... 115 6.8 Electronic Stability Control.??..????????.???.??. 118 xi 6.9 Intelligent Vehicle Model????????????.????.... 120 6.10 Summary and Conclusion????????????????? 123 7. CONCLUSIONS 7.1 Summary????????????????????????.... 125 7.2 Recommendation for Future Work??????????????? 127 REFERENCES??...?????????????????????.??? 130 APPENDICES?...???????????????????????.?? 134 A Vehicle Garage??????????????????????. 135 B Scaled Vehicle Properties??????????????????. 138 C Vehicle Simulation????????????????????... 140 xii LIST OF FIGURES 1.1 Static Stability Factor?????????????????????... 3 1.2 Fishhook Maneuver??????????????????????.. 4 1.3 NHTSA?s Rollover Propensity Star Rating System??????????. 5 2.1 SAE Vehicle Coordinate System?????????????????. 11 2.2 Bicycle Model Free Body Diagram????????????????. 13 2.3 Sprung Mass Roll Free Body Diagram??????????????? 16 2.4 Un-sprung Mass Roll Free Body Diagram?????????????... 16 2.5 Linear vs. Non-Linear Tire Model with a Normal Force of 5 kN????? 24 2.6 Non-Linear Tire Model with Varying Normal Forces?????????. 25 2.7 The Friction Circle??????????????????????... 27 2.8 2001 Chevrolet Blazer?????????????????????.. 29 3.1 Maneuver A, S.S. Roll Model??????????????????.. 34 3.2 Maneuver A, Transient Roll Model????????????????. 35 3.3 Maneuver B, S.S. Roll Model??????????????????.. 37 3.4 Maneuver B, Transient Roll Model????????????????.. 38 3.5 Zoom-In View of Test 2 Transient Model Wheel Load Response????.. 41 3.6 Steer Angle Input for the Fishhook 1a Maneuver??????????? 43 3.7 J-Turn Maneuver Roll Dynamics for Experimental and Simulation Data?... 45 3.8 J-Turn Maneuver Yaw Dynamics for Experimental and Simulation Data?.. 46 3.9 Auburn University GAVLAB Blazer Center of Gravity drawn in the Nominal Configuration????????????????????? 47 3.10 Nominal Blazer Dynamics for a Fishhook 1a Maneuver????????. 48 3.11 Auburn University GAVLAB Blazer Center of Gravity drawn in the RRR Configuration????????????????????????... 49 3.12 RRR Configuration Blazer Dynamics for a Fishhook 1a Maneuver 51 3.13 Auburn University GAVLAB Blazer Center of Gravity drawn in the RMB xiii Configuration????????????????????????... 52 3.14 RMB Configuration Blazer Dynamics for a Fishhook 1b Maneuver???.. 53 3.15 RMB Configuration Blazer Wheel Loads for a Fishhook 1b Maneuver??. 55 3.16 Auburn University NCAT Test Track Facility???????????? 56 3.17 Lane Change Maneuver with the Auburn University GAVLAB Blazer??. 57 3.18 Lane Change Maneuver Yaw and Roll Angle Data for the GAVLAB Blazer 58 3.19 Roll FBD Un-Sprung Mass.............................................................................. 60 3.20 Maneuver B, Transient Roll, Front Lateral Weight Transfer??????... 62 3.21 Step Maneuver with Roll Center Height equal to Un-Sprung Mass CG Height???????????????????????????.. 63 4.1 SIS Constant with Weight Split Variation?????????????... 68 4.2 TWL Velocity for Weight Split Variation?????????????... 69 4.3 Vehicle Understeer Gradient for a variation of Weight Splits??????. 70 4.4 TWL Velocity for Various CG Heights??????????????... 71 4.5 Velocity and Yaw Dynamics at Rollover for CG Height Variation???? 74 4.6 Roll and Side Slip States at Rollover for CG Height Variation?????... 75 4.7 Roll Rate for CG Height Variation????????????????.. 76 4.8 Lateral Tire Force at Rollover for CG Height Variation????????.. 77 4.9 Velocity and Yaw Dynamics at Rollover for Weight Split Variation???.. 78 4.10 Roll and Side Slip Thresholds for Weight Split Variation???????... 79 4.11 Tire Curves for Variation of Peak Lateral Tire Force (Fz =5 kN)????? 80 4.12 Tire Curves for Variation of Tire Cornering Stiffness (Fz =5 kN)????... 82 5.1 ESC Simulated on the Nominal Blazer Configuration?????????. 89 5.2 ESC with and without IVM for CG Height Variation?????????. 91 5.3 Effect on Dynamic Behavior due to CG Height Move without IVM???.. 92 5.4 Stability Limits for Weight Split Variation?????????????.. 95 6.1 First Generation Scale Vehicle Test bed??????????????.. 100 6.2 Second Generation Scale Vehicle Test bed?????????????.. 102 6.3 Vehicle Control Schematic???????????????????... 104 6.4 First Generation IMU?????????????????????... 105 6.5 Wireless Data Acquisition Package????????????????. 106 xiv 6.6 Second Generation Data Acquisition PCB Board. IMU and Microprocessor shown on top, GPS Receiver and Radio Modem shown on bottom???? 107 6.7 First Generation Scaled Vehicle Experimental versus Simulation Data??.. 109 6.8 Fishhook Steering Profile for Stability Limit Development??????? 111 6.9 Stability Limits for CG Height Variation??????????????. 113 6.10 Stability Limits for the Weight Split Variation???????????? 114 6.11 CG Height Variation Experiment Stability Limit??????????? 117 6.12 Experimental Stability Limit for Variations in the CG Height?????? 118 6.13 Fishhook Maneuver with and without ESC enabled?????????? 119 6.14 Results for the Fishhook 1a with 60:40 Weight Split?????????.. 121 6.15 Results for the Fishhook 1a with 40:60 Weight Split?????????.. 122 xv LIST OF TABLES 2.1 Linear Tire Model vs. Non-Linear Pacejka Tire Model????????... 24 2.2 Pacejka Parameters??????????????????????... 26 3.1 Comparison of Models for Maneuver A??????????????.. 36 3.2 Comparison of Models for Maneuver B??????????????.. 39 3.3 Nominal vs. RRR Blazer Configuration??????????????.. 49 3.4 Nominal vs. RMB Blazer Configuration??????????????.. 52 4.1 NHTSA Phase IV Vehicle Load Configurations [Forkenbrock, 2002]??... 66 4.2 Blazer TWL Velocity in Fishhook 1b Maneuver [Forkenbrock, 2002]??... 66 4.3 Peak Lateral Tire Force Variation Results?????????????... 81 4.4 Tire Cornering Stiffness Variation Results?????????????.. 83 5.1 ESC with and without IVM for CG Height Change?????????? 93 5.2 Blazer Stability Threshold for Two Weight Splits??????????.. 93 6.1 IMU Specifications??????????????????????.. 108 6.2 ESC with and without IVM???????????????????.. 123 1 CHAPTER 1 INTRODUCTION 1.1 Motivation Over the past decade, the occurrence of rollover incidents in vehicle accidents has received much attention due to the increase in Sport Utility Vehicle (SUV) miles driven on the roadways and litigation claiming that engineers are to blame for designing unsafe vehicles. The National Highway Traffic Safety Administration (NHTSA) reported in 2002 that 3% of all light passenger vehicle crashes on United States roads involve rollover, yet rollover accidents are responsible for 1/3 of all passenger vehicle occupant fatalities [Hilton, 2003]. The statistics reveal that SUVs rollover fatalities are more prevalent than in cars; 59% of total fatalities in SUV accidents occurred in rollover crashes, while rollovers accounted for only 23% of total fatalities in car accidents [Ponticel, 2003]. There are many differences in the vehicle properties of SUVs versus those of passenger cars. Understanding how these vehicle properties affect rollover is important when designing a safe vehicle. 10,376 fatalities and 229,000 injuries were caused by crashes that involved rollover in 2003 alone [FARS, 2005]. Along with lives, there is an economic cost that can be decreased by designing vehicles that are less prone to rollover. Motor vehicle crashes in the United States cost an estimated $231 billion in economic costs [NHTSA, 2002]. It is easy to see that a reduction in the number of 2 vehicle rollovers can have a positive impact on the economy. By utilizing research to understand how vehicle properties and vehicle dynamics affect rollover, both human life and economic dollars can be saved. 1.2 Background and Literature Review The vehicle dynamics and properties that affect the occurrence of vehicle rollover have been studied for many years. Early vehicle rollover studies consisted of static tests to determine a vehicle?s rollover propensity [Allen, 1990]. Statistical studies of vehicle rollover accidents were also performed to correlate vehicle properties with rollover propensity [Klein, 1992]. The most notable of these parameters are the vehicle center of gravity (CG) height and track width. In fact, they are the sole parameters that determine the Static Stability Factor (SSF). The rollover studies that used static tests proposed the SSF as a measure to quantify vehicle rollover propensity. As stated before, the SSF is a static measure of two variables: the average of the vehicle?s front and rear track widths and the vehicle?s CG height, as shown in Equation (1.1). HeightCG 2 hTrack WidtSSF ?= 1.1 An illustration of the SSF taken from NHTSA?s website is shown in Figure 1.1. Note that an increase in SSF corresponds to a decrease in rollover propensity. The SSF is a very accurate measure of a vehicle?s static rollover stability during tripped and non- 3 tripped rollover. However, static rollover tests neglect the transient dynamics and tire dynamics that are involved in the abrupt changes in velocity and steer angle that come before crashes [Chrstos, 1992]. It is noted that suspension characteristics during dynamics tests are important when determining a vehicle?s rollover stability and since tires are non-linear, the tires? lateral forces saturate during extreme maneuvers [Hac, 2002]. Figure 1.1. Static Stability Factor [NHTSA, 2005] In the 1990?s, the National Highway Traffic Safety Administration?s (NHTSA) vehicle safety star rating system used only the SSF to determine a vehicle?s rollover propensity. However, due to the high fatality rate of rollover crashes, Congress passed the ?Transportation Recall, Enhancement, Accountability and Documentation Act,? (TREAD), in November of 2000. TREAD charged NHTSA to conduct dynamic rollover resistance rating tests, which NHTSA in turn made part of its New Car Assessment Program (NCAP). For the purposes of a dynamic rollover resistance rating test, NHTSA selected the Fishhook steering maneuver as a primary candidate to use in conducting 4 rollover experiments. An illustration of the Fishhook maneuver, taken from NHTSA?s website is shown in Figure 1.2. Figure 1.2. Fishhook Maneuver [NHTSA, 2005] For vehicle dynamic testing, the rollover propensity of a vehicle is determined from the highest speed for which the vehicle can complete the selected maneuver without achieving two-wheel lift. Since the vehicle testing is conducted on-road, the results are more repeatable and give more control over the test environment than off-road tripped tests [Forkenbrock, 2003]. The evaluation procedure is only meant to test vehicles for on-road, un-tripped, rollover propensity. Although this only accounts for a small percentage of rollover crashes, the results are still a valuable measure of overall rollover stability for relative comparison of vehicles [Viano, 2003]. In 2004, NHTSA added the dynamic test to its star rating system for rollover propensity. Figure 1.3 shows NHTSA?s 5 chart for determining a vehicle?s rollover propensity star rating. It uses both the SSF and dynamic test results in order to determine rollover propensity. However, because the majority of rollover events are tripped and the CG height and track width are the vehicle properties that have the greatest effect on tripped rollover propensity, the new rating system placed more weight on the SSF [Cooperrider, 1990]. Figure 1.3. NHTSA?s Rollover Propensity Star Rating System [NHTSA, 2005] Along with developing a rollover propensity rating system, the Department of Transportation (DOT) has taken other action to make vehicles safer. In 2004, all new vehicles were mandated by the DOT to have a tire pressure monitor in each tire and a tire pressure warning light visible to the driver. The light acts as an indicator to the driver that the tire with low pressure is a hazard and needs to be repaired. This mandate was a 6 result of the occurrence of multiple incidents in which tires with low pressure blew-out on highways and resulted in rollover fatalities. Electronic Stability Control (ESC) is the current buzz technology in automotive industry, and rightly so. NHTSA has published reports stating that ESC reduces single vehicle crashes by 67% and fatal crashes by 64%. The 10,376 rollover related fatalities in 2003 and the 8,476 single vehicle rollover fatalities reveal the potential that ESC has to save lives [FARS, 2005]. The ESC is designed and implemented in order to help the driver maintain vehicle stability, while tracking the` driver?s desired path. The first generation of ESC uses a lateral accelerometer and a yaw rate gyroscope (gryo) to monitor the yaw dynamics and use the vehicle?s antilock braking system to brake the wheels independently to minimize understeer and oversteer [van Zanten, 2000]. ESC?s effectiveness is so impressive that Daimler-Chrysler, Ford, and General Motors have announced that ESC, once only an option on SUVs, will be a standard feature in all of their 2007 SUV models [Voelcker, 2005]. Since the first generation of ESC systems were so successful, continuous efforts to improve ESC are ongoing [Hac, 2004]. The accurate detection of vehicle rollover and vehicle state estimation are current topics being studied [Johansson, 2004]. The ability to precisely estimate vehicle states such as side slip angle and slip angle rate greatly increases the ESC?s ability to detect instability and prevent rollover [Bevly, 2001; Ryu, 2004; Nishio, 2001]. Many researchers are putting emphasis on the design and implementation of new actuators, such as active suspension and steer-by-wire, to improve ESC performance [Wilde, 2005; Huang, 2004; Vilaplana, 2004]. And even more ESC research is being conducted on improving control algorithms by optimizing the use of the 7 actuators and sensors in conjunction with vehicle state estimation [Carlson, 2003; Plumlee, 2004; Schubert, 2004]. Vehicle modeling of rollover and vehicle dynamics that influence rollover has been studied much in recent years in order to better understand and prevent rollover [Nalecz, 1993]. LaGrange?s method, an energy method, works well with impact modeling such as tripping [Ginsberg, 1998]. Tripped rollover scenarios are most commonly modeled using LaGrange?s method, although researchers have used both Lagrange?s and Newton?s method to develop models for untripped rollover events [Garrott, 1992; Day, 2000;]. Newton?s method works in modeling non-impact dynamics like a vehicle?s yaw dynamics as seen in the bicycle model [Baumann, 2004]. Scaled vehicles are being used as valuable research tools to investigate vehicle yaw dynamics and test ESC algorithms [Travis, 2004; Brennan, 2003]. The rollover testing of full-size vehicles is an expensive and dangerous endeavor, unlike scaled vehicles where the vehicles are cheaper and experiments are safer [Yih, 2000]. If results from scaled vehicles tested in a controlled environment can be related to the dynamic behavior of full-size vehicles, then the use of scaled vehicles can be an effective means of investigating rollover [Brennan, 2001]. Brennan and Alleyne developed the Illinois Roadway Simulator for design and evaluation of yaw dynamics controllers [Brennan, 1998]. The Illinois Roadway Simulator research and scaled vehicle experiments at Auburn University have shown that scaled vehicle test beds are successful in capturing the vehicle?s dynamics and can be used to develop ESC technologies for full size passenger vehicles [Whitehead, 2004; Brennan, 2004]. 8 1.3 Purpose of Thesis and Contribution It is well known that center of gravity height and track width, the two parameters that make up the SSF, are the major vehicle parameters that contribute to rollover propensity. In this thesis, the effects of other vehicle properties, such as longitudinal weight distribution, are evaluated to determine its influence on rollover propensity, while holding the SSF constant. The purpose of this work is to analyze vehicle rollover utilizing vehicle simulation and vehicle experimental data. A non-linear vehicle model is created in order to study how vehicle properties affect rollover propensity. A detailed development and description of a vehicle simulation, including validation testing, and a discussion of the results for several parametric variation studies are provided. An instrumented scaled vehicle is used for the first time to study vehicle rollover and relate it to full scale passenger vehicles. The scaled vehicle is used to acquire vehicle dynamics data leading to rollover and to evaluate a stability threshold created by simulation. The vehicle simulation correlates the scaled vehicle with a full scale passenger vehicle in order to validate the use of scaled vehicles for studying rollover. The vehicle simulation is also used to develop a stability threshold which is a function of the vehicle?s loading condition. This work also develops a method to increase the effectiveness of ESC. Although ESC?s ability to prevent single vehicle rollover is already high, there is room for even more improvement by using an Intelligent Vehicle Model (IVM). The IVM utilizes information about the vehicle to update the ESC?s vehicle model and controller limits with a new stability threshold as the vehicle?s loading condition changes. This thesis uses simulation in conjunction with an ESC on a scaled vehicle test bed to determine the 9 effectiveness of the ESC as vehicle properties change. The simulation is used to determine the vehicle?s stability threshold at different vehicle properties. The ESC with the IVM utilizes the property changes to the vehicle model and the controller gains are adjusted to keep the vehicle within the stability threshold. This is different from a regular ESC where the vehicle model does not change, and one stability threshold is used at all times. Vehicle loading conditions are the most common way vehicle properties are varied. This thesis focuses on the influence of two of those properties, CG height and weight split, and compares the effectiveness of ESC with and without the use of an IVM. 1.4 Outline of Thesis The purpose of this study is to evaluate and determine the parameters that influence or change vehicle rollover propensity using both experimental data and simulation. The first step was to derive a vehicle model for the simulation. The vehicle model was developed using Newton?s laws of motion. The free body diagrams, equations, and assumptions are shown in Chapter 2 [Newton, 1687]. The computer simulation uses a transient roll and yaw dynamic model to capture the dynamics of the vehicle. The vehicle model is verified via comparison with experimental data from a passenger vehicle in Chapter 3. The vehicle properties? values used are given in Appendix A. 10 The loading conditions that effect vehicle rollover are evaluated in Chapter 4. The vehicle model developed in Chapters 2 and 3 is used to create a stability threshold for variations of these loading conditions. In Chapter 5, an ESC that uses throttle and steering control to maintain vehicle stability is developed. The ESC is evaluated in simulation and a method to improve its effectiveness is tested. The IVM, which updates the ESC?s vehicle model and stability threshold for variations in vehicle load configuration, is developed in this chapter and is shown to increase the ESC?s effectiveness. Finally, to aid in studying rollover propensity, this research uses an instrumented scaled vehicle test bed to validate the vehicle model, the stability threshold, the ESC and the IVM that were developed in Chapters 2-5. The scaled vehicle is instrumented with a six degree of freedom (DOF) inertial measurement unit (IMU) and a global positioning system (GPS). The scaled test bed is also equipped with a CG relocator which allows the vehicle?s loading condition to be varied in the longitudinal and vertical axes. The detailed analysis of rollover using a 1/10th scaled vehicle is given in Chapter 6. Appendix B contains the values of the scaled vehicle properties and a description of how the roll mass moment of inertia was quantified. 11 CHAPTER 2 VEHICLE MODEL 2.1 Introduction In this chapter, the model of a typical four-wheel passenger vehicle is developed. The developed vehicle model has three degrees of freedom (DOF) and was developed by deriving the equations of motion (EOM) using the free body diagrams (FBD) of the sprung and un-sprung masses. The vehicle coordinate system can be seen in Figure 2.1. The primary dynamics of concern in this thesis are the yaw and roll motions. Figure 2.1. SAE Vehicle Coordinate System [Milliken, 1995] 12 The pitch dynamics (which cause longitudinal weight transfer) are neglected since longitudinal accelerations were kept small in the experiments considered in this thesis. The model, which is implemented using the MATLAB programming language, was developed to be extremely flexible. Individual vehicle properties such as center of gravity location and suspension setup are easily changed. Additionally, the simulation can utilize either use the transient yaw equations with either the steady state roll dynamic equations or the transient roll dynamic equations. The difference in these two models is apparent when analyzing the vehicle dynamics of transient maneuvers. However, in steady state maneuvers, the difference is small, as is shown in Chapter 3. 2.2 Bicycle Model The transient yaw equations are derived from the ?bicycle model? free body diagram shown in Figure 2.2. The 2-wheel bicycle model is the most commonly known version; however, for the purposes of this research, a 4-wheel bicycle model is used so that lateral weight transfer can be included in the yaw dynamics. The 4-wheel bicycle model, assumes that the slip angles are symmetric about the x-axis of the vehicle, which is valid at high speeds and zero Ackerman Effect [Gillespie, 1992]. 13 Figure 2.2. Bicycle Model, Yaw FBD The tire forces and slip angles in Figure 2.2 are drawn using a positive sign convention. In actuality, the tire slip angles as shown would produce a negative lateral tire force as shown in Figure 2.2. Summation of moments about the center of gravity yields the yaw acceleration found in Equations (2.1 and 2.2). [ ]mNrIM zz ???=? & 2.1 ( )[ ] [ ]2/cos1 sradaFbFIr yfyr z ???+???= ?& 2.2 where: a = Length between the CG and front tire patch V = Vehicle velocity vector b = Length between the CG and rear tire patch Vf = Front tire velocity vector ? = Steer angle Vr = Rear tire velocity vector Fy = Tire lateral forces Vx = Vehicle velocity in the x-axis r = Yaw rate Vy = Vehicle velocity in the y-axis 14 The lateral dynamics are shown in Equations (2.3 and 2.4) by summing forces in the y-axis direction. ( )( ) [ ]NrVVMF yty ???+?=? ?cos& 2.3 ( ) ( ) ( ) [ ] 2/coscos smrV M FFV t yfyr y ???? ?+= ??& 2.4 The tire slip angles are calculated as the angle between the velocity vector of the tire and the direction that the tire centerline is pointing. Equation (2.5) represents the front tire slip angle, which includes the steer angle, ? , since the vehicle being modeled is front wheel steer. Equation (2.6) corresponds to the rear tire slip angle. [ ]radV arV x y f ???? ? ??? ? ? ??? ? ??? ? ?+= ? ?? 1tan 2.5 [ ]radV brV x y r ???? ? ??? ? ??= ?1tan? 2.6 Similarly, the vehicle side slip angle, shown in Equation (2.7), is the angle between the vehicle velocity vector and the direction the vehicle centerline is heading. The side slip rate, shown in Equation (2.8), is an important dynamic property in stability control systems. A large spike in the ?& value is an indicator of yaw instability since it is the rate at which the vehicle?s velocity vector and heading are changing [Ryu, 2004]. Additionally, the side slip rate is more accurately estimated than side slip angle which current technology. [ ]radVVy ??? ? ? ??? ?= ?1sin? 2.7 15 ( ) ( ) [ ]srad dt nn /1 ?? ? ?? ? ? ?+= ???& 2.8 The lateral acceleration, shown in Equation (2.9), is comprised of yV& and the component of centripetal acceleration, shown in Equation (2.10), that is perpendicular to the vehicle velocity vector. ( )( ) [ ]2/cos smaVa cenyy ??+= ?& 2.9 [ ]2/ smVracen ??= 2.10 The vehicle?s longitudinal and lateral velocities are defined by Equations (2.11 and 2.12). ( ) [ ]smVVx /cos ?= ? 2.11 ( ) [ ]smVVy /sin ?= ? 2.12 2.3 Roll Model The roll equations are derived by separating the sprung and un-sprung masses in the y-z plane, as shown in Figures 2.3 and 2.4, and applying Newton?s second law (as formulated for rigid bodies). ?Inside? and ?Outside? represent the sides of the car that correspond to the inside and outside of a turn. In the figures shown below, the front of the vehicle is pointing into the page resulting in a positive roll angle towards the outside of the vehicle. 16 Figure 2.3. Roll FBD Sprung Mass Figure 2.4. Roll FBD Un-Sprung Mass Fyi Fzi Fzo Ry Rz rc CGm Fbi + Fki Fbo + Fko mg Inside Outside Fyo S/2 trk Marb ? CGM Mg F bi + Fki Fbo + Fko Inside Outside Ry Rz rc d1 Marb 17 The forces, moment, and lengths on Figures 2.3 and 2.4 are defined as: CGM = Sprung mass center of gravity Marb = Anti-roll bar moment CGm = Un-sprung mass center of gravity ? = Roll angle d1 = Length between the rc and CGM rc = Roll center Fb = Damper force (o ? outside, i ? inside) Ry = Reaction force in the y-axis Fk = Spring force (o ? outside, i ? inside) Rz = Reaction force in the z-axis Fy = Tire lateral force S = Length between the springs and dampers Fz = Tire normal force trk = Track width The steady state roll model can be derived, by setting the acceleration and velocity dynamic states to zero. With this simplification, roll angle is analyzed as a linear function of lateral acceleration. This assumes that the total roll stiffness is linear and d1 is constant due to the small angle linearization. Equation (2.13) is the linearized roll angle equation. [ ]radgadWk dW y tt t ss ??? ?=? ? 1 1 2.13 where: ay = Lateral acceleration k?t = Total roll stiffness g = Gravitational constant Wt = Total sprung weight 18 The transient roll model is more complex than the steady state version since it is non-linear and includes the acceleration and velocity states. The transient roll model is very useful in that it incorporates the transient dynamics into the lateral weight transfer equations which monitor the normal forces on the wheels. Also, roll rate is a vehicle state of much importance in anti-rollover stability control systems. With the roll EOM, the roll rate magnitude can be compared with rollover incidents to develop limits for a new stability control systems. The roll equation of motion is found in Equations (2.14 and 2.15) by summing the moments about the x-axis. [ ]mNIM xx ????=? && 2.14 ( ) ( )[ ] [ ]211 /)cos()sin(1 sraddRdRRDMRSMI yz x ????+???+????? ? ? ??? ?=?&& 2.15 In the above equation, ?&& is the roll acceleration, RSM is the torque from the roll stiffness, and RDM is the torque from the roll damping. The roll stiffness (RS) and roll stiffness moment (RSM) values are calculated in Equations (2.16 and 2.17). ( ) ( )[ ] [ ]radmNkkSkSkRS arbrarbfkrsrkfsf /5.05.0 22 ??++??+??= 2.16 ( ) ( )[ ][ ] [ ]mNkkSkSkRSM arbrarbfkrsrkfsf ????++??+??= 22 5.05.0 2.17 Additionally, the roll damping (RD) and roll damping moment (RDM) are represented by Equations (2.18 and 2.19) respectively. ( ) ( )[ ] ( )[ ]sradmNSbSbRD brrbff //5.05.0 22 ????+??= 2.18 ( ) ( )[ ][ ] [ ]mNSbSbRDM brrbff ??????+??= &22 5.05.0 2.19 19 where: bf = Front shock damping ksr = Rear spring stiffness br = Rear shock damping Sbf = Length between front shocks karbf = Front anti-roll bar stiffness Sbr = Length between rear shocks karbr = Rear anti-roll bar stiffness Skf = Length between front springs ksf = Front spring stiffness Skr = Length between rear springs The x and y axes reaction Forces at the roll center are given by Equations (2.20 and 2.21). Where M is the sprung mass. [ ]NaMR yy ??= 2.20 [ ]NgMRz ??= 2.21 It is important to note that the roll centers are assumed to be stationary to simplify the model. Also note that the spring and damper forces are assumed to be symmetric about the x-axis. This is due to the complexity of calculating the suspension kinematics during simulation. By not calculating the suspension kinematics during maneuvers, the simulation runtime is decreased. By comparison with experimental data, it is revealed in following chapters that this simplification is valid. In order to simplify the lateral weight transfer calculation, the un-sprung mass roll dynamics are neglected. The un-sprung mass transient dynamics are neglected because rollover detection is possible by using the less complex steady state equations. The transient dynamics for the un-sprung mass are needed to determine the height of wheel lift. However, for this study, only the instant that the wheel is lifted is required to signal a rollover event. 20 2.4 Lateral Weight Transfer and Normal Wheel Loads The lateral weight transfer is derived from the un-sprung mass FBD. The lateral weight transfer is measured as the difference between the inside and outside tire normal forces as shown in Equation (2.22). [ ] [ ]NFFdF zizoz ??= 2.22 To accurately determine the normal force on each wheel, the FBD of the un- sprung mass must be configured using both front and rear components. However, only the steady state roll dynamics of the un-sprung mass are modeled since the simulation is only concerned with rollover occurrences. Once wheel lift is detected, (which is when the transient roll dynamics of the un-sprung mass would be critical), the vehicle simulation is terminated and declared a rollover event. Therefore, the lateral weight transfer of the front and rear axles are found using Equations (2.23 and 2.24), respectively. ( )[ ] [ ]NhFhhRFSFSMtrkdF cgmyfcgmrcfyfbfbfkfkfarbf f zf ??+??+?+?+?= 2 2.23 ( )[ ] [ ]NhFhhRFSFSMtrkdF cgmyrcgmrcryrbrbrkrkrarbr r zr ??+??+?+?+?= 2 2.24 where: Fbf = Front damper force Ryf = Reaction force on the front Fbr = Rear damper force Ryr = Reaction force on the rear Fkf = Front spring force Sbf = Length between front shocks 21 Fkr = Rear spring force Sbr = Length between rear shocks Fyf = Front axle lateral force Skf = Length between front springs Fyr = Rear axle lateral force Skr = Length between rear springs hcgm = Un-sprung mass CG height trkf = Front track width hrcf = Front roll center height trkr = Rear track width hrcr = Rear roll center height Since the suspension kinematics of the vehicle?s modeled are hard to calculate, and require additional time for simulation, the roll center height is assumed remain in the stationary static position. In order to obtain the front and rear reaction forces in the y- axis, the weight split of the vehicle must be incorporated. This allows the magnitude of the reaction force to be distributed proportionally to the mass on the front (Mf) and rear (Mr) axles as shown in Equations (2.25 and 2.26). [ ]NMMRR fyyf ??= 2.25 [ ]NMMRR ryyr ??= 2.26 The normal forces of each tire are then calculated using the initial static force on each axle and the lateral weight transfer of each axle. Wf and Wr are the static forces on the front and rear axle respectively. The front and rear inside normal tire forces, Fzfi and Fzri, are shown in Equations (2.27 and 2.28), while the front and rear outside normal tire, Fzfo and Fzro, forces are shown in Equations (2.29 and 2.30). [ ]NdFWF zffzfi ?? ? ? ?? ? ?= 2 2.27 22 [ ]NdFWF zrrzri ?? ? ? ?? ? ?= 2 2.28 [ ]NdFWF zffzfo ?? ? ? ?? ? += 2 2.29 [ ]NdFWF zrrzro ?? ? ? ?? ? += 2 2.30 2.5 The Pacejka Tire Model: The ?Magic Formula? The normal forces of the tires are not only used to determine vehicle rollover; they must be used in order to determine the lateral force of each tire. Through the years, many tire studies have been conducted resulting in various tire models. As early as 1925, a researcher named Broulhiet discovered the concept of side slip and its role in producing lateral forces in a pneumatic tire [Broulhiet, 1925]. Fromm also contributed to the early studies of tire side slip and yaw [Fromm, 1954]. In the ?50s and ?60s, the ?friction circle? was developed and experiments were conducted in order to understand its effects on handling [Radt, 1960; Ellis, 1963; Radt, 1963; Morrison, 1967]. In 1970, Dugoff, in conjunction with the Highway Safety Research Institute at the University of Michigan, published a study in which tire experiments were conducted resulting in an analytical tire model [Dugoff, 1969; Dugoff, 1970]. A lead engineer in tire research throughout the ?60s, Hans Pacejka, published the ?Magic Formula? in 1987 for use in determining the lateral forces for a pneumatic passenger vehicle tire in vehicle simulations [Pacejka, 23 1987]. In the 1987 study, Pacejka and his colleagues recorded tire data on vehicles and developed an equation that models the empirical data of the tire?s dynamics quite well. In 1989, Pacejka published an even more refined and detailed ?Magic Formula? [Pacejka, 1989]. This tire model is a highly non-linear model where lateral force is a function of the tire slip angle and tire normal force. Additionally, the tire model contains parameters such as peak tire lateral force, cornering stiffness, and tire model curvature which can be modified to capture the effects for different tires. Not only does the ?Magic Formula? model the tire?s lateral dynamics, it contains the tire?s longitudinal dynamics as well. The vehicle model in this research uses the Pacejka model because of its accuracy and ease of use. Figure 2.5 shows the difference between the linear and non-linear ?Magic Formula? tire model. The linear model has a constant tire cornering stiffness, C?, with no saturation, while the Pacejka tire model has a non-constant C? as the tire curve transitions into the non-linear region. In the non-linear Pacejka tire model, once the peak force is reached, the lateral tire force decreases with increasing tire slip angle. This is called tire saturation. 24 Figure 2.5. Linear vs. Non-Linear Pacejka Tire Model with a Normal Force of 5 kN Note that the linear tire model approximates the non-linear Pacejka model for small slip angles; however, Table 2.2 shows that this approximation rapidly loses validity as tire slip angle increases. Table 2.1. Linear Tire Model vs. Non-Linear Pacejka Tire Model Tire Slip Angle, ?, [deg] 2.5? 5.0? 7.5? Lateral Force difference between Linear and Non-Linear Tire Models 5.6% 37.1% 88.4% 25 The peak tire lateral force is a function of tire slip angle and normal load. Note that increasing normal load reaches a point where it no longer produces significant gains in lateral force. This relationship is seen in Figure 2.6, which shows that both the C? and peak force change with changing normal load. Figure 2.6. Non-Linear Tire Model with Varying Normal Forces The inputs of the Pacejka tire model are tire slip angle and tire normal force, and the output is tire lateral force. The non-linear Pacejka lateral tire force is calculated using the Equation (2.31). ( )( )????= BaCDFy tansin 2.31 Equations (2.32 to 2.36) show the constants that are found in Equation (2.31). Equation (2.32) is known as the shape factor, which is independent of the tire normal force. 26 3.1=C 2.32 Equation 2.29 represents the peak factor which determines the peak lateral force of the tire curve. zz FaFaD ?+?= 2 2 1 2.33 Equation 2.30 controls the curvature of the tire curve, thus it is called the curvature factor. 87 2 6 aFaFaE zz +?+?= 2.34 The cornering stiffness, ?C , factor is found in equation 2.31. Where the BCD variable is effectively ?C . ( )( )zFaaaaBCD ???= 543 tansin 2.35 Equation 2.32 is developed to be used later in equation 2.33. ( ) ( )??? ??? ? ?? ? ?+??= Ba B EE tan1 2.36 The Pacejka parameters used in the model are those found in Pacejka, 1987, and can be seen in Table 2.2. Note that these parameters assume that the dimensional unit for the tire slip angle is degrees and for the tire normal force is kilo-Newton. The output of the Pacejka model, lateral tire force, has units of Newtons. Table 2.2 Pacejka Parameters a0 = 0 a5 = 0.208 a1 = -22.1 a6 = 0 a2 = 1011 a7 = -0.354 a3 = 1078 a8 = 0.707 a4 = 1.82 27 Steps were taken to simplify the non-linear tire model in this vehicle model. In Figure 2.8, the radius of the friction circle represents the maximum magnitude of force that the tire can produce at any one time. If the vehicle is braking or accelerating (longitudinal axis), the maximum lateral force decreases from the non-braking, non- accelerating scenario. The longitudinal effects of braking and acceleration were neglected since the majority of the experiments were conducted at a constant longitudinal velocity. Figure 2.7. The Friction Circle Only the lateral forces of the tire model are considered in this model, and some of those are left out as well. For instance, all forces that come from camber thrust and all moments from the self aligning torque are neglected. However, following chapters will Longitudinal Force Lateral Force Braking Forward Acceleration Maximum Total Force Produced by the Tire 28 show how the model accurately captures a passenger vehicle?s dynamics without these tire parameters. 2.6 Garage The simulation uses a MATLAB m-file entitled ?Garage.? This appropriately entitled garage contains multiple vehicles and their vehicle properties. The 2001 Chevy Blazer in the 4-door, 2 wheel drive package that is loaded into to the garage contains the most accurate vehicle property data. Other vehicles that are in the garage include modified Blazers that match NHTSA Phase IV experiments along with a scaled vehicle. The accuracy of these vehicle properties is critical for the vehicle simulation. A list containing the value and description of each vehicle property needed for the vehicle simulation. The vehicle properties used for each vehicle are found in Appendix A. With the information in Appendix A and an accurate tire model, the vehicle simulation accurately reproduces the yaw and roll dynamics of passenger vehicles as will be shown in Chapter 3. 29 Figure 2.8. 2001 Chevrolet Blazer 2.7 Summary of Model Assumptions Throughout the derivation of any model, assumptions are sometimes made in order to simplify the simulation. This is necessary due to computer processor power limitations, as assumptions and linearizations shorten the runtime of simulations. Some assumptions must be made due to unknown vehicle characteristics and properties such as suspension kinematics. The bicycle model in this study was modified in order to include weight transfer and a non-linear tire model. It is effectively a four wheel bicycle model because it 30 includes weight transfer. The simplification that the tire slip angles are equal on the inside and outside of the front and rear axles is still assumed. This assumption is valid at highway traveling speeds [Gillespie, 1992]. In the roll dynamics derivation, both the front and rear roll centers are assumed to be constant throughout the suspension?s range of motion. This approximation neglects the suspension kinematics. Therefore the simulation?s runtime is decreased. The un-sprung mass of the roll free body diagram seen in Figure 2.4 is assumed to be in steady state. The main information needed from the un-sprung mass is the tire?s normal load, and the steady state equations provide this information. Just as the previous assumption, the purpose of this simplification is to decrease the simulation run time. This model focuses on the roll and yaw dynamics alone, and neglects the longitudinal weight transfer effects of pitch. This assumption was made in order to simply the model. Neglecting the pitch dynamics is made possible by keeping the longitudinal acceleration and braking to a minimum in the simulation. Via simulation comparison with experimental data, it is shown in the following chapters that these assumptions are valid. 2.8 Summary and Conclusion The derivation and intricacies of the vehicle model used in this study have been shown in this chapter. The governing equations of yaw and roll were derived using 31 Newton?s method which included summing forces and moments on the sprung and un- sprung mass free body diagrams. The free body diagrams include the bicycle model for the yaw dynamics and sprung and un-sprung mass free body diagrams for the roll dynamics and lateral weight transfer. The ?Magic Formula? tire model, developed by Hans Pacejka, was used for the vehicle model and shown. The non-linear tire model is a function of normal force and tire slip angle. It was chosen for its accuracy and ease of implementation in the vehicle simulation. Finally, the assumptions and simplification used in the model?s derivation were shown and explained. This model is validated with experimental data in the next chapter. 32 CHAPTER 3 VEHICLE MODEL VALIDATION 3.1 Introduction This chapter contains the validation of the vehicle model developed in the previous chapter. First, a comparison of the steady state roll model with the transient roll model is conducted. The purpose of this comparison is to see if the steady state response of the transient roll model matches the predicted steady state roll. After verifying that the steady state response is accurate for the transient roll model, the full transient dynamics response is validated. This validation is done by comparing data from the NHTSA Phase IV research with the transient roll model using a 2001 Chevy Blazer. The transient roll model is further validated using an experimental Blazer test vehicle at Auburn University. Once the validity of the transient model is shown, the lateral weight transfer dynamics of the transient roll model are explored and evaluated. The analysis in this chapter confirms the validity of the transient roll model such that the model can be used to conduct experimental simulations in later chapters. 33 3.2 Steady State Roll Analysis versus Transient Roll Analysis A simple simulation was developed to determine the validity of the transient roll model that was developed in Chapter 2. The rationale behind this first test is to check the accuracy of the transient roll model by comparing it with the steady state roll analysis after the vehicle has reached steady state. The steady state roll model (SSRM) is well known and documented by previous research such that it can be considered as the ?true? measurement of the vehicle states [Gillespie, 1992; Dixson, 1998]. Both vehicle models use the same transient yaw dynamics and vehicle properties. The only difference is how the roll dynamics are calculated. In order to correctly compare the transient roll model with the SSRM, the steady state values of the vehicle states are evaluated and compared at the end of the maneuver. For this test, a step steer at constant velocity is used as the maneuver to excite the vehicle. The 2001 Blazer in the nominal configuration is the vehicle use for this analysis. The maneuver consists of a constant velocity held at 20 MPH and a step steer of 5 degrees that is filtered at 1.5 Hz. This maneuver is labeled as maneuver A for clarity in this chapter. The 1.5 Hz filter is implemented using a second order Butterworth filter. This filter is employed in order to better model a vehicle?s true steering input. The simulation is updated every 0.001 second or 1000 Hz. Figure 3.1 shows the response of the Blazer using the SSRM. 34 Figure 3.1. Maneuver A, S.S. Roll Model Figure 3.2 uses the transient roll model with the same vehicle and maneuver as Figure 3.1. The difference between these models can be seen by comparing the roll angle and wheel loads within these figures. The transient roll model in Figure 3.2 contains an oscillation while the SSRM in Figure 3.1 does not. The reason for this is that the transient roll model derives the roll angle via the roll equation of motion which includes roll acceleration and damping, as was seen Equation (2.15), while the SSRM assumes roll acceleration and damping are zero thus making the SSRM roll angle a function of lateral acceleration, as was seen in Equation (2.13). The wheel loads are derived by calculating the lateral weight transfer on the front and rear axles, which was shown by Equation (2.22) in Chapter 2. The lateral weight transfer is calculated as a function of roll angle in 35 the SSRM and as a function of the roll dynamics in the transient roll model as seen in Equations (2.23 and 2.24). Since the roll dynamics are oscillatory in the transient roll model, the wheel loads are also oscillatory. Figure 3.2. Maneuver A, Transient Roll Model As stated earlier, in order to compare the different roll models, the dynamic states are compared once the simulations have reached steady state for each model. The steady state values are measured 10 seconds after the maneuver was started. The difference in the states are compared by computing the percent difference for the two models using Equation (3.1) where the SSRM measurement is used as the truth measurement. 36 100% Roll SS RollTransient - Roll SS DifferencePercent ?= 3.1 Table 3.1 contains the steady state results for maneuver A. There is no wheel lift during this maneuver, and the steady state dynamic values for the two models are approximately the same as seen by the low percent difference between the two. Table 3.1. Comparison of Models for Maneuver A Dynamic State S.S. Roll Trans. Roll % Difference Units Lateral Acceleration .2554118 .2554118 0.0 [g] Roll Angle 1.419507 1.419058 .0316307 [deg] Side Slip 2.278848 2.278849 .0000439 [deg] Yaw Rate 16.09942 16.09941 .0000621 [deg/s] Front Weight Transfer 526.9914 526.3616 .1195085 [N] Rear Weight Transfer 1,421.585 1,421.087 .0350313 [N] Front Lateral Force 2,650.778 2,650.778 0.0 [N] Rear Lateral Force 2,137.870 2,137.869 .0000468 [N] Roll Rate 5.557720 5.55596 .0316676 [deg/g] For the next test, the step steer of 5 degrees filtered at 1.5 Hz is used again. However, in this simulation (Maneuver B), the velocity is increased to 35 MPH. In this input scenario, the inside rear tire in both models lifts and remains un-weighted as the states are measured at the 10 second time interval. Figures 3.3 and 3.4 contain the response of the SSRM and the transient roll model, respectively, during maneuver B. In Figures 3.3 and 3.4, the rear inside wheel load is indicated by the dotted line. It is clear that the wheel load for this tire goes to zero, which indicated wheel lift. 37 Figure 3.3. Maneuver B, S.S. Roll Model 38 Figure 3.4. Maneuver B, Transient Roll Model The comparison of the models? steady state dynamic states for maneuver B is recorded in Table 3.2. 39 Table 3.2. Comparison of Models for Maneuver B Dynamic State SS Roll Trans. Roll % Difference Units Lateral Acceleration .8498721 .8499465 .008754 [g] Roll Angle 4.723351 4.707426 .337154 [deg] Side Slip -54.79752 -54.74284 .099785 [deg] Yaw Rate 56.05783 56.03894 .033697 [deg/s] Front Weight Transfer 1,751.426 1,728.569 1.30505 [N] Rear Weight Transfer 4,185.143 4,185.143 0 [N] Front Lateral Force 8,894.704 8,895.999 .014559 [N] Rear Lateral Force 7,040.835 7,040.957 .001733 [N] Roll Rate 5.55772 5.538496 .345897 [deg/g] Rear Inner Wheel Lift time .373 .214 42.6273 [sec] By examining Table 3.2, the greatest discrepancy between the two models is the time at which the rear inner wheel lifts. However, even during this maneuver, the percent difference between the two models is small enough that the models are still very comparable. The SSRM takes an additional 0.159 seconds than the transient roll model to lift the inside rear wheel. This is explained by investigating the rear lateral weight transfer equation which dictates the rear wheel loads. Equation (3.2) describes the general weight transfer, while Equation (3.3) describes the full weight transfer for the rear axle. [ ] [ ]NFFdF zizoz ??= 3.2 ( )[ ] [ ]NhFhhRFSFSMtrkdF cgmyrcgmrcryrbrbrkrkrarbr r zr ??+??+?+?+?= 2 3.3 40 where: Fbr Rear damper force hrcr Rear roll center height Fkr Rear spring force Marbr Rear anti-roll bar moment Fyr Rear axle lateral force Ryr Reaction force on the rear Fzi Axle inside tire wheel load Sbr Length between rear shocks Fzo Axle outside tire wheel load Skr Length between rear springs hcgm Un-sprung mass CG height trkr Rear track width The lateral weight transfer equation, Equation (3.2), is calculated as the difference between inside and outside wheel loads on each axle. An increase in the rear, lateral weight transfer, dFzr, causes a decrease in wheel load on the rear inside tire. For this simulation, the maximum value of dFzr is one half of the weight on the rear axle, and at this maximum value, the rear inside wheel load is zero. The lateral weight transfer equation is identical for both roll models; however, the rear damper force (Fbr) is zero in the SSRM, while it has a value in the transient roll model. The rear damper force is a function of roll velocity and the longitudinal distance between the CG and where the damper connects to the sprung mass (Sbr). brS and Fbr are the variables in roll damping moment (RDM), as seen in Equation (2.19). The RDM value is positive when the roll angle velocity is positive. Note that increasing roll angle towards the outside of the vehicle is positive. For the step maneuver, the dFzr increases faster in the transient roll model than in the SSRM due to the positive roll angle velocity. This causes the transient roll model to produce a wheel lift before the SSRM. 41 It is also important to note that while the transient roll model is the first to produce wheel lift, the wheel that lifted first, touches down again between time .515 and .543 seconds as seen in Figure 3.5. During this time, the roll angle velocity is negative which causes the RDM to be negative and results in the dFzr decreasing from its maximum, thus causing touch down. Touch down does not occur with the SSRM because the roll dynamics are neglected. Therefore, once wheel lift occurs in this model, it does not regain normal wheel load. Figure 3.5. Zoom-In View of Test 2 Transient Model Wheel Load Response With the minimal percent difference between the two models in the two maneuvers found in Tables 3.1 and 3.2, the transient roll is considered accurate at steady 42 state for the dynamic states analyzed. Although, this analysis of the SSRM and the transient roll model is successful in capturing the steady state response of the transient roll model, further analysis of the transient model is needed to verify the full transient response of the model. 3.3 NHTSA Phase IV Comparison Due to the increasing fatality rate caused by rollover crashes (especially in SUVs) Congress charged NHTSA to conduct dynamic rollover resistance rating tests. For the purposes of a dynamic rollover resistance rating test, NHTSA selected the Fishhook steering maneuver as a primary candidate, which was refined in the Phase IV of the TREAD act investigation. During the TREAD act, NHTSA took exhaustive data measurements of the vehicle dynamics from the vehicles they tested. This data was made available to the general public and is used in this chapter to further validate the vehicle model. In order to validate the vehicle model described in Chapter 2, simulation results were compared to the NHTSA Phase IV experimental data for the Fishhook 1a maneuver (also known as the Fixed Timing Fishhook). The Fishhook 1a maneuver uses a steering input consisting of an initial steer followed by a counter steer at a set entrance velocity. The velocity profile of this maneuver is characterized by the vehicle reaching a desired steady state speed, known as the entrance speed, and coasting through the rest of the 43 maneuver once the initial steer is begun. The steer angle input for the Fishhook 1a maneuver is shown in Figure 3.6. The steer angle initially starts at zero degrees and then commanded to steer angle ?A? at a rate of 720 degrees per second at the hand wheel. In the Fishhook 1a maneuver, the steer angle ?A? is held constant for 0.250 seconds then a counter steer to ?-A? at the same rate occurs. The steer angle ?-A? is held constant for 3 seconds, after which it returns to zero, completing the maneuver. The value ?A? is specific to each vehicle configuration, and is defined by multiplying 6.5 by the steer angle of the handwheel at which the vehicle experiences 0.3 g of lateral acceleration in the Slowly Increasing Steer (SIS) maneuver. The SIS maneuver is performed at a constant velocity of 50 mph with a continually increasing steer input of 13.5 degrees per second at the hand wheel. Figure 3.6. Steer Angle Input for the Fishhook 1a Maneuver 44 To accurately simulate the vehicle, there are three parameters of the Blazer that must be quantified. These parameters are the suspension stiffness, damping, and the front-to-rear roll stiffness ratio. However, ranges for these values are known. Therefore a method to approximate a value for each of these properties without ?tuning? the properties for a specific maneuver was used. NHTSA evaluated the Blazer using multiple maneuvers such as the J-Turn, Fishhook 1a, and Fishhook 1b. Since the maneuver of most dynamic interest, the Fishhook 1a, is used later for model validation, the J-Turn maneuver is compared with the simulation in order to back out the approximate unknown properties. The frequency and damping of the suspension?s roll dynamics are seen in the roll data in Figure 3.7. With this information, the suspension roll stiffness and roll damping are varied to match the roll angle frequency and damping. Since the differences between the front and rear roll stiffness and roll damping are not known at this point, they are kept equal to one another. 45 Time [sec] Figure 3.7. J-Turn Maneuver Roll Dynamics for Experimental and Simulation Data With the suspension roll stiffness and roll damping approximated, there still exists an error between the experiment and simulation yaw data. The simulation showed tire saturation while the experiment did not. In order to minimize this error, the vehicle?s front to rear roll stiffness was changed to match the yaw rate. However, the total roll stiffness was held constant to retain the roll dynamic response. Changing the front to rear roll stiffness changes the understeer gradient which dictates the maximum yaw rate a vehicle can attain for a given maneuver. The yaw dynamics for the experimental data and the simulation are shown in Figure 3.8. With the roll stiffness, roll damping, and 46 front to rear roll stiffness ratio approximated for the J-Turn maneuver, the comparison of the simulation with the experimental data for a Fishhook maneuver can be performed with confidence that the parameters are not ?tuned? specifically for the Fishhook maneuver. Figure 3.8. J-Turn Maneuver Yaw Dynamics for Experimental and Simulation Data 47 The vehicle used for comparison in this study is a 2001 Chevy Blazer 4x2. NHTSA?s Phase IV experiments recorded data for the Blazer in three different configurations: Nominal, Reduced Rollover Resistance (RRR), and Rear Mounted Ballast (RMB). The Nominal configuration is the Blazer equipped with a driver, data acquisition, and outriggers on board. It has a weight distribution of 55:45 (front to rear), CG height of 26.3 inches, and a track width of 56.89 inches. These parameters result in a static stability factor (SSF) of 1.048 (Equation 1.1). Complete property values of each vehicle can be found in Appendix A. Figure 3.7 shows the Nominal configuration center of gravity drawn on the Auburn University GAVLAB Blazer. Figure 3.9. Auburn University GAVLAB Blazer Center of Gravity drawn in the Nominal Configuration 48 Figure 3.10 shows the Nominal Blazer experimental data and simulation data during a Fishhook 1a maneuver. Both the yaw and roll dynamics of the transient yaw and roll model closely match to dynamic states measured in the actual vehicle. Figure 3.10. Nominal Blazer Dynamics for a Fishhook 1a Maneuver 49 The RRR configuration is the vehicle in the Nominal configuration with a 181 lbs roof ballast added on the top of the vehicle and shown in Figure 3.11. This serves to raise the CG vertically by 5% from the Nominal configuration to 27.6 inches. This results in a new SSF of 0.989, while maintaining the same vehicle longitudinal weight split as the Nominal configuration. Figure 3.11. Auburn University GAVLAB Blazer Center of Gravity drawn in the RRR Configuration In order to conduct this simulation, the CG height, roll inertia, and yaw rate were the only parameters that were changed in the garage when converting the Nominal Blazer to the RRR Blazer. The changes of these properties were recorded in the NHTSA study and are shown in Table 3.3. Table 3.3. Nominal vs. RRR Blazer Configuration [Forkenbrock, 2002] Weight (lbs) CG Height (inches) Roll Inertia (ft-lbs-sec2) Yaw Inertia (ft-lbs-sec2) Pitch Inertia (ft-lbs-sec2) Nom. Blazer 4154 26.3 520 2765 2573 RRR Blazer 4335 27.6 579 2766 2637 50 The roll inertia for the RRR configuration is 11.3 percent higher than the Nominal configuration. However, since the placement of the roof ballast is near the CG in the x and y axes, the yaw inertia changes by less than 0.04 percent. Though pitch inertia is neglected in vehicle model, thus not needed, it is interesting to note the roof mounted ballast causes it to increase by 2.49 percent. The increase in the pitch inertia will cause an increase in understeer during heavy breaking with the Blazer. In Figure 3.12, the experimental and simulated RRR Blazer configuration data is compared. Again, the vehicle model follows the actual RRR Blazer Fishhook 1a experiment data. The largest discrepancy between the simulation and experimental data is in the roll angle. This is most likely due to wheel lift of the inside tires. 51 Figure 3.12. RRR Configuration Blazer Dynamics for a Fishhook 1a Maneuver The RMB configuration is the vehicle in the Nominal configuration with weight added to the rear of the vehicle, which moves the CG longitudinally by 10.5 inches toward the rear, but retains the same SSF. This center of gravity shift is illustrated in Figure 3.13. The RMB has a front to rear weight distribution of 44:56, but maintains the SSF of 1.048. For the Nominal and RRR configurations, experimental data for a 52 Fishhook 1a maneuver is available. However, for the RMB configuration, only experimental data for a Fishhook 1b maneuver is available. The Fishhook 1b maneuver is similar to the Fishhook 1a, but does have some subtle differences [Forkenbrock, 2002; NHTSA, 2002]. Figure 3.13. Auburn University GAVLAB Blazer Center of Gravity drawn in the RMB Configuration Table 3.4 reveals the change in vehicle properties between the Nominal and RMB Blazer configurations. Table 3.4. Nominal vs. RMB Blazer Configuration [Forkenbrock, 2002] Weight Split Front:Rear CG Height (inches) Roll Inertia (ft-lbs-sec2) Yaw Inertia (ft-lbs-sec2) Pitch Inertia (ft-lbs-sec2) Nom. Blazer 55:45 26.3 520 2765 2573 RMB Blazer 44:56 26.1 567.9 3604 3368.3 53 Figure 3.14. RMB Configuration Blazer Dynamics for a Fishhook 1b Maneuver Figure 3.14 shows experimental and simulation results from Fishhook 1b for the RMB configuration. The largest discrepancy between the simulation results and the experiment data is again in the roll angle. The experimental data reveals that two-wheel lift occurs. This is seen in the roll angle measurement along with the NHTSA notes for 54 this experiment. However, the simulation result predicts only one-wheel lift with the second wheel approaching wheel lift. This is seen from the normal forces for each tire, as shown in Figure 3.15. This difference is most likely due to limitations of the simulation model. Some of the model?s assumptions discussed in Chapter 2 may not be valid for all scenarios the vehicle encounters. For example, the roll stiffness is more than likely not linear in the real Blazer. Also, the assumption of the un-sprung mass being in steady state is no longer accurate after one wheel lift occurs for the roll dynamics (yet it appears to remain valid for the yaw dynamics). Also note that this discrepancy is small and it occurs only in the Fishhook 1b RMB data. 55 Figure 3.15. RMB Configuration Blazer Wheel Loads for a Fishhook 1b Maneuver 56 3.4 Auburn University GAVLAB Blazer The GPS and Vehicle Dynamics Lab in the Department of Mechanical Engineering at Auburn has performed maneuvers with an instrumented Chevrolet Blazer. These tests were conducted at Auburn?s NCAT facility, shown in Figure 3.16. Figure 3.16. Auburn University NCAT Test Track Facility The experimental data was again compared with simulation results using the Nominal configuration of the Blazer. In these tests, the velocity and steer angle were recorded and used as inputs in the simulation. The test track limits maneuvers to either single or double lane-change maneuvers. Figure 3.17 shows a comparison of the Auburn experimental data with results from the simulation model during lane-change maneuvers. 57 Time (sec) Figure 3.17. Lane Change Maneuver Data for the GAVLAB Blazer Though the yaw dynamics of the experiment and simulation data closely match each other, differences between the simulation and experiment roll angles can be clearly seen in Figure 3.18. This is caused by a change in the bank angle of 4 degrees during the lane change maneuver, commonly known as the road crown (which aids in water drainage). 58 Figure 3.18. Lane Change Maneuver Yaw and Roll Angle Data for the GAVLAB Blazer The parameters used in this simulation are the ones used for the Nominal Blazer which were tuned with the NHTSA experimental data for the J-Turn maneuver. The similarity of the dynamic behavior in the simulation and the various experiments provide much confidence in the vehicle model. 59 3.5 Lateral Weight Transfer Simulation In the some of the simulation maneuvers, the vehicle exhibits an initial counter weight transfer, analogous to a non-minimum phase system. In other words, during a maneuver, the weight transfer is initially in opposite direction of steady state weight transfer as seen in Figure 3.5. The steady state values for the dynamic states reveal that the outside wheel loads are higher than the inside wheel loads, as seen in Figure 3.4. However, at the beginning of a step maneuver, the front inside wheel load increases and the front outside decreases, as seen in Figure 3.5. Closer analysis shows that during the first 0.1685 seconds of the maneuver the weight transfer on the front axle is negative. This increases the inside wheel load. For the normal force on the inside wheel to increase during a turn is counter intuitive. In steady state, the normal force on an axle?s inside wheel decreases and the normal force on the outside wheel of the same axle increases as expected. In order to understand this phenomenon, the weight transfer equation must be further examined. The weight transfer equation for the front axle and rear axle are the same. However, vehicle parameters for the front and rear axles are different. The equation of the front weight transfer from Chapter 2 is shown again here as Equation (3.4). ( )[ ] [ ]NhFhhRFSFSMtrkdF cgmyfcgmrcfyfbfbfkfkfarbf f zf ??+??+?+?+?= 2 3.4 60 During the initial 0.1685 seconds of the step maneuver, the roll angle and velocity are positive and therefore the moments from the anti-roll bar ( arbfM ), springs ( kfkf FS ? ) and dampers ( bfbf FS ? ) are all positive. By investigating the un-sprung mass free body diagram in Figure 3.19, it is seen that the tire lateral forces are also positive for this left steer, step steer maneuver. Figure 3.19. Roll FBD Un-Sprung Mass The only variable remaining in the weight transfer equation that could be negative and cause dFzf to be negative is the term ( )cgmrcfyf hhR ?? . For the step maneuver, the lateral reaction force is always positive. Therefore, the variables causing the initial counter weight transfer are the un-sprung mass CG height and the roll center height. When the roll center height is below the un-sprung mass CG height, the moment arm becomes negative and is multiplied by a positive lateral reaction force, thus causing the Fyi Fzi Fzo Ry Rz rc CGm Fbi + Fki Fbo + Fko mg Inside Outside Fyo S/2 trk Marb 61 dFzf to be negative at the beginning of this maneuver. On the Blazer vehicle model, the front roll center height is below the un-sprung mass CG height and the rear un-sprung mass height is equal to the un-sprung mass height. Because the rear weight transfer is never negative, note that the inside rear wheel load is never greater than the outside rear in Figure 3.5. This is due to the fact that the un-sprung mass CG height and the roll center height are the same; the moment arm of the ( )cgmrcfyf hhR ?? term is zero. To clarify which variable or variables is causing the weight transfer to be negative, three terms of the front lateral weight transfer are analyzed separately. These terms include the total front weight transfer, the front weight transfer without the ( )cgmrcfyf f hhRtrk ???2 term, and the ( )cgmrcfyf f hhRtrk ???2 term. Figure 3.16 shows these terms during a step maneuver. 62 Figure 3.20. Maneuver B, Transient Roll, Front Lateral Weight Transfer It can see in Figure 3.20 that the dFzf is negative before time 0.1685 seconds. This is because the Ryf component is greater than the other terms of Equation (3.4). This narrows the possible vehicle properties that cause the counter weight transfer down to two: the roll center height and the un-sprung mass CG height. To further explore the counter weight transfer on the front axle, a simulation of a step steer input is performed with both front and rear roll center heights equal to the un- sprung mass CG height. The response is shown in Figure 3.21. 63 Figure 3.21. Step Maneuver with Roll Center Height equal to Un-Sprung Mass CG Height It is shown that the counter weight transfer does not exist when the roll center height is equal to the un-sprung mass CG height. Also note the convention of the negative tire slip angles producing positive lateral tire forces. The conclusion is that a roll center height below the un-sprung mass CG height creates an initial counter weight transfer. 64 3.6 Summary and Conclusion In this chapter, the accuracy and validity of the transient roll model has been demonstrated. The transient roll model was shown to produce the same steady state results as the steady state roll model. Experimental vehicle data acquired from NHTSA?s Phase IV rollover research was used to determine unknown vehicle properties. Additional data from the Phase IV testing was then used to validate that the transient roll model matches the transient dynamics of the experimental data. It has also been shown that the vehicle model accurately captures the dynamics of the vehicle when properties are changed, as shown by the RRR and RMB Blazer configurations. Experimental data produced by the Auburn University GAVLAB was used to provide further validation that the vehicle model matches experimental data for a lane change maneuver. An initial counter weight transfer was seen in the wheel load data during the simulation maneuvers. This phenomenon was examined and ultimately produced more confidence in the simulation?s accuracy. This wide range of validation of the model developed in Chapter 2 allows for confidence in the vehicle dynamics analysis in the following chapters. 65 CHAPTER 4 ANALYSIS OF VEHICLE ROLLOVER STABILITY LIMIT 4.1 Introduction This chapter investigates the effect of various vehicle parameters on rollover propensity using computer simulation. The vehicle model used in the simulation was developed in Chapter 2. The computer simulation?s accuracy was verified in Chapter 3 by comparing the simulation data to experimental data from NHTSA?s Phase IV testing on rollover of passenger vehicles. In this chapter, the vehicle model is subjected to a specific steering input defined by NHTSA, called the Fishhook 1a. An analysis of the vehicle loading condition and its influence on rollover propensity is conducted using the validated vehicle simulation. 66 4.2 NHTSA Phase IV Analysis As discussed previously, NHTSA performed the Phase IV of TREAD rollover analysis in 2002. Chapter 3 showed data for a 2001 Blazer in various configurations that were tested, and this data was used to validate the vehicle simulation. In the NHTSA tests, the vehicle?s loading condition was changed to three specific locations. The Nominal, Reduced Rollover Resistance (RRR), and Rear Mounted Ballast (RMB) configurations represent these loading condition changes, which are shown in Table 4.1. Table 4.1. NHTSA Phase IV Vehicle Load Configurations [Forkenbrock, 2002] Nominal RRR RMB CG Height [inches] 26.3 27.6 26.1 Weight Split Front : Rear 55:45 55:45 44:56 NHTSA performed a Fishhook 1b maneuver with the Blazer in each of these configurations. To determine vehicle rollover, the Blazer was equipped with outriggers to allow the vehicle to lift both wheels of one side of the vehicle. Entrance speeds of the maneuver were increased until two-wheel-lift (TWL) was detected. TWL was defined as both wheels on one side of the vehicle being two inches off the ground. Table 4.2 contains the NHTSA Phase IV data which shows the Blazer lowest entrance speed at which TWL was detected. Table 4.2. Blazer TWL Velocity in Fishhook 1b Maneuver [Forkenbrock, 2002] Nominal RRR RMB TWL Velocity [mph] 40.1 36.2 34.9 67 The results of this experiment show that the Blazer?s tendency to rollover increases as the CG height is increased. The RRR configuration has a CG height 1.3 inches higher than the Nominal configuration and experiences TWL at 36.2 mph while the Nominal configuration requires 3.9 mph more velocity before TWL is achieved. This follows intuition and the results of the SSF explained previously. The RMB configuration, like the RRR, increases the Blazer?s tendency to rollover. However, the CG height of the RMB is 0.2 inches lower than the Nominal CG height. The physical difference between these two configurations causing the TWL velocity variation is the weight split. The Nominal and RMB have approximately the same SSF, and weight splits of 55:45 and 44:56, respectively. NHTSA performed a Fishhook 1b maneuver on both of these configurations, and the Nominal configuration experienced TWL at 40.2 mph, while the RMB had TWL at 34.9 mph. In order to better quantify the difference between the Nominal and RMB configurations, the vehicle model is used to simulate and duplicate NHTSA?s Phase IV data for the Blazer. The vehicle?s weight split and CG height are varied to assess their affect on rollover propensity. The front to rear weight split is varied from 30:70 to 70:30, while holding the SSF constant. The Nominal configuration of the Blazer is also modified so that the roll center, suspension stiffness and damping are the same on the front and rear. With these changes to the suspension, the effects of changing the weight split are isolated from other factors. The parameters varied in this simulation test are the lengths ?a? and ?b?, while their sum, the wheelbase, is held constant. The Slowly Increasing Steer (SIS) maneuver and constant are used and adopted from the NHTSA Fishhook 1a test. The change in weight distribution caused the SIS constant to also 68 change for each different weight split. For each parametric variation of the vehicle load condition, a new SIS constant is determined as described in Chapter 3. The SIS constant varied from 0.675 to 1.827 degrees at the tire for the splits of 30:70 to 70:30 respectively as shown in Figure 4.1. 0.50 0.75 1.00 1.25 1.50 1.75 2.00 30 40 50 60 70 Percent Weight on the Rear Axle SI S C on sta nt [d eg ] Figure 4.1. SIS Constant Change for Weight Split Variation The dynamic test used to determine the effects of these property changes is the Fishhook 1a maneuver. The Fishhook 1a is a highly repeatable maneuver as cited by NHTSA?s Phase IV research and is an easily programmed open loop input (unlike its the Fishhook 1b, which requires roll velocity data in a closed loop feedback control) [Forkenbrock, 2003]. As in the NHTSA study, rollover velocity is calculated by determining the vehicle speed at which the steering maneuver causes TWL. As previously described, NHTSA determines TWL as both wheels on one side of the vehicle being lifted 2 inches off the 69 ground. However, TWL for the simulation is defined as the instant that the normal forces on both tires on one side of the vehicle go to zero. This definition is due to the fact that the simulation is not able to measure the height of each wheel after wheel lift since the un-sprung mass dynamics are not modeled. The simulation is still accurate until the point where wheel lift occurs. The difference in the definition of TWL between the simulation and experiment results in a discrepancy in the velocity that TWL occurs. Since the simulation rollover is defined as the instant that TWL occurs and not after both wheels have lifted 2 inches, the rollover velocity for the simulation is lower than the NHTSA experiment data. Figure 4.2 shows the simulation of TWL velocity across a range of vehicle weight splits. NHTSA?s Nominal and RMB configurations for the Blazer are also shown in Figure 4.2. 25 30 35 40 45 50 30 40 50 60 70 Percent Weight on the Rear Axle TW L V elo cit y [ mp h] Simulation NHTSA Nominal Configuration NHTSA RMB Configuration Figure 4.2. TWL Velocity for Weight Split Variation 70 Figure 4.3 shows the corresponding understeer curve. Inspection of Figures 4.2 and 4.3 reveals a correlation between understeer and rollover propensity. As the Blazer?s weight is shifted toward the rear, the vehicle begins to oversteer as well as roll at a lower velocity than when the weight is shifted toward the front axle. Also note that as the weight is shifted toward the front axle, the corresponding SIS increases causing the Fishhook 1a to become a more severe maneuver. However, the TWL velocity continues to increase as the weight is moved to the front axle. NHTSA?s Phase IV experiments reveal this same correlation between weight split and rollover propensity, as seen in Figure 4.2. Figure 4.3. Vehicle Understeer Gradient for a Variation of Weight Splits Another interesting observation from the simulation data was seen for two vehicle configurations with a SSF equal to 1.048. One vehicle configuration experiences TWL at 71 28.6 mph and the other experiences TWL at 45.2 mph. These speeds were recorded for various weight distributions of 30:70 through 70:30 front to rear, respectively. Although the SSF is considered the most important measurement of a vehicle?s rollover propensity, two vehicles with the same SSF may experience a difference in rollover velocity for the same maneuver. This clearly demonstrates that there are other vehicle parameters that play an important role with regard to rollover propensity besides center of gravity height and track width (SSF parameters). The simulation was also configured to vary the CG height on the Blazer to assess its effect on rollover propensity. The Blazer configuration used was the Nominal case, and the only parameter varied was the CG height. This results in a constant SIS of 1.268 degrees measured at the tire. Figure 4.4 shows the corresponding Blazer simulation results. Also shown in the Figure are the Nominal and RRR TWL velocities. 25 35 45 55 65 75 5 10 15 20 25 30 35 40 Center of Gravity Height [inch] TW L V elo cit y [ mp h] Simulation Data NHTSA Nominal Configuration NHTSA RRR Configuration Figure 4.4. TWL Velocity for Various CG Heights 72 With the track width is held constant, the SSF varies from 3.467 to 0.2774 with change in CG height. The Nominal configuration has a SSF of 1.048, which corresponds to a CG height of 16.76 inches. The TWL velocity for the Nominal configuration as determined by the simulation is 38.8 mph. NHTSA documented TWL for the same configuration and maneuver as being 40.2 mph. Again, the difference in the TWL velocities is caused by the different definition of TWL between the simulation and experiment. The observed trend of the data is not surprising since it is well known that a vehicle will roll at a lower velocity if the CG is raised or the SSF is decreased. 4.3 Stability Limit Development From Section 4.2, it can be seen that a vehicle?s rollover threshold varies not only with CG height, but also the CG longitudinal location, otherwise known as the vehicle?s weight split. In this section, vehicle states other than rollover velocity are measured to further develop an understanding of how the vehicle?s dynamic states at rollover vary with the vehicle?s load condition. Vehicle simulations are again used to determine the stability threshold for different vehicle load conditions. This is done by independently varying the vehicle?s center of gravity height and weight split. The maximum value of the each of the following states is recorded when rollover is detected: maneuver entrance velocity, lateral acceleration, yaw rate, roll angle, roll rate, side slip angle, and side slip rate. For the general Fishhook maneuver, rollover is consistently measured slightly after the maximum 73 counter steer. The Fishhook maneuver used is different than the NHTSA Fishhook 1a because the maneuver?s steer angle magnitude is independent of the SIS constant described in Chapter 3. Therefore, the maneuver is unchanged for variations of vehicle properties, thus allowing for a more concise comparison between the property variations. For these simulations, a Fishhook steering profile was chosen with a maximum steer angle of 5 degrees and a maximum steering rate of 40 deg/s, both measured at the tires. This profile was filtered by a second order 0.75 Hz Butterworth filter to remove the abrupt changes in the simulation?s steer angle. The first property varied was the CG height. The Nominal Blazer configuration was used as the baseline. The CG height was varied between 65 and 160 percent of the baseline CG height. The rollover velocity, lateral acceleration and yaw rate are shown in Figure 4.5. As shown in the previous section, the vehicle will roll at lower velocities as the CG height is raised. Also, note that as the CG height is raised, the lateral acceleration and yaw rate that the vehicle can achieve before experiencing a rollover event is decreased. These results confirm those of the SSF, which states that the vehicle is more prone to rollover as the CG height is increased. 74 Figure 4.5. Velocity and Yaw Dynamics at Rollover for CG Height Variation The roll angle and roll rate along with the side slip angle and the side slip rate are important states when determining rollover [Hac, 2004]. Currently, Ford utilizes the roll rate dynamic state in its electronic stability control system for SUVs, and much research is ongoing to use the side slip rate state in vehicle stability control systems. The following states at rollover are shown in Figure 4.6: roll angle, roll rate, side slip, and side slip rate. 75 Figure 4.6. Roll and Side Slip States at Rollover for CG Height Variation As the CG height is increased, the roll angle and roll rate at rollover increase. This increase in roll angle and roll rate is caused by the increase in the vertical distance (d1) between the roll center axis and the CG height as the CG height is raised. The angle of roll per lateral acceleration is known as the Roll Rate property (not to be confused with the roll rate [deg/s] state), as seen in Equation (4.1) [Gillespie, 1992] and Figure 4.7 shows the change in Roll Rate as the CG height is varied. Roll Rate ? ? ? ?? ??? ??+? ???= gdgMk dgM t deg180 1 1 pi? 4.1 76 Figure 4.7. Roll Rate for CG Height Variation The lateral force on the front and rear axles for the CG height variation is shown in Figure 4.8. It can be seen that as the CG height increases, the lateral force at vehicle rollover is decreased. 77 Figure 4.8. Lateral Tire Force at Rollover for CG Height Variation The second property which was varied on the Nominal Blazer is the vehicle?s weight split. The vehicle weight split was varied from a weight ratio of 70:30 (front axle to rear axle) to a ratio of 30:70. The same maneuver and vehicle from the CG height variation simulation were used. Figure 4.9 shows the velocity, lateral acceleration, and yaw rate at rollover for the weight split variation simulations. 78 Figure 4.9. Velocity and Yaw Dynamics at Rollover for Weight Split Variation As noted in Section 4.2, the maximum speed the Blazer can safely negotiate the Fishhook maneuver without rollover decreases as the weight is shifted towards the rear axle. However, the yaw rate at which the vehicle rolls increases as the weight is moved toward the rear axle [Gillespie, 1992]. This is explained by the change in the vehicle?s weight transfer causing limit oversteer as the weight split is shifted rearward. As weight is shifted toward the rear axle, the weight transfer on the rear axle increases and saturates the rear tires faster than the front tires and this causes limit oversteer [Milliken, 1995]. If the vehicle?s velocity is increased, the limit oversteer will cause the yaw rate to approach infinity. The vehicle?s understeer gradient is also decreased as the weight is shifted to the rear as shown previously in Figure 4.3. For the same steering input and velocity, a vehicle with a lower understeer gradient will have a greater yaw rate than one with a higher understeer gradient [Gillespie, 1992]. 79 The maximum lateral acceleration that Blazer can achieve before rollover also increases as the weight split is moved rearward. The lateral acceleration at rollover as the weight is shifted to the rear mimics the yaw rate at rollover. Although Equation (4.2) is only true at steady state, it does explain why the lateral acceleration at rollover trend follows the yaw rate at rollover trend. 2rVa y ?= 4.2 Figure 4.10 shows the maximum roll angle, roll rate, side slip angle, and side slip rate achieved at rollover. The rollover roll angle remains nearly constant as weight split is varied. However, the rollover roll angle rate decreases as the vehicle?s weight is shifted to the rear axle. The side slip rate varies from 0 to 0.5 deg/s at rollover yet the side slip angle decreases at the weight split is shifted toward the rear. Figure 4.10. Roll and Side Slip Thresholds for Weight Split Variation 80 4.4 Tire Property Variation As noted in Chapter 2, the non-linear, ?Magic Formula? tire model can be easily modified. In this section, the tire?s cornering stiffness and peak lateral force are varied to see how they affect rollover propensity. Again, a Fishhook maneuver of 5 degrees is used to study this effect. To examine how peak lateral tire force influences rollover, three different tires were used. The three different tires are modeled in Figure 4.11 with the tire normal force set at 5 kN. As seen in Figure 4.11, the cornering stiffness is held constant at 2400 N/deg for each tire, but each tire has a different peak lateral force. Figure 4.11. Tire Curves for Variation of Peak Lateral Tire Force (Fz =5 kN) 81 The limits of each state at rollover for the three different tires models during the Fishhook maneuver are given in Table 4.3. It can be seen that the tire with the lower peak lateral tire force has a higher rollover velocity for this maneuver. Additionally, the tire with the lowest peak lateral force requires a higher velocity to produce the required lateral force for rollover on each axle than the other tires. Table 4.3. Peak Lateral Tire Force Variation Results Peak Lateral Forces Dynamic States at Rollover 6000 [N] 4500 [N] 3000 [N] Velocity [mph] 26.0 26.0 26.5 Yaw Rate [deg/s] 18.256 18.204 17.8839 Lateral Acceleration [g] 0.3512 0.3509 0.3463 Roll Angle [deg] 2.191 2.178 2.131 Roll Rate [deg/s] 16.268 16.104 15.026 Side Slip Angle [deg] 1.449 1.443 1.344 Side Slip Rate [deg/s] 6.485 6.356 5.907 The tire cornering stiffness was also varied to determine its effect on rollover propensity. Four different cornering stiffnesses are evaluated. The tire curve for each tire with a normal load of 5 kN is shown in Figure 4.12. The peak lateral tire force was held constant at 4500 N, as seen in Figure 4.12. 82 Figure 4.12. Tire Curves for Variation of Tire Cornering Stiffness (Fz =5 kN) The limits of each state at rollover for the tire models with the tire cornering stiffness variation are seen in Table 4.4. The tire with the lowest cornering stiffness requires the highest velocity in the Fishhook maneuver to rollover. The rollover velocity is higher in the tire with the lowest cornering stiffness because a higher velocity is required to produce the same lateral force that the other tire models rollover. 83 Table 4.4. Tire Cornering Stiffness Variation Results Tire Cornering Stiffness [N/deg] Dynamic States at Rollover 3000 2400 1500 1000 Velocity [mph] 25.25 26.0 28.0 29.25 Yaw Rate [deg/s] 18.574 18.214 17.50 16.604 Lateral Acceleration [g] 0.3651 0.3511 0.3245 0.2934 Roll Angle [deg] 2.282 2.180 1.985 1.773 Roll Rate [deg/s] 18.801 16.154 11.523 8.1247 Side Slip Angle [deg] 1.710 1.444 0.9541 1.381 Side Slip Rate [deg/s] 7.214 6.369 4.769 5.472 4.5 Summary and Conclusion In this chapter, simulation data along with NHTSA experimental data were used to evaluate how two vehicle properties, CG height and weight split, affect a vehicle?s rollover limit. The simulations were also used to explore how the cornering stiffness and peak lateral force of the tire affect a vehicle?s rollover propensity. It was shown that the SSF is not the sole determinate of a vehicle?s rollover propensity, and that there are vehicle properties other than SSF, such as weight distribution and tire properties, that can significantly influence rollover propensity. It was also shown that there is a correlation between the understeer gradient and rollover propensity. As the vehicle understeer increases, the TWL velocity tends to increase. Finally, rollover stability limits were developed in this chapter. The stability threshold consisted of many vehicle dynamics states: velocity, lateral acceleration, yaw 84 rate, roll angle, roll angle rate, side slip angle, and side slip rate. Those were the states used to express the rollover stability limits. These stability limits will be used with an ESC in the following chapters. 86 CHAPTER 5 ELECTRONIC STABILITY CONTROL DEVELOPMENT 5.1 Introduction As was stated in Chapter 1, vehicle rollover accidents have received much attention in the past decade. There is high demand for SUVs (vehicles with low SSFs) in the automobile market despite their high rollover propensity. The automotive engineering world?s solution to making these vehicles safer is electronic stability control (ESC). As was shown in the previous chapters, the foundation of ESC begins with an understanding of how vehicle dynamics and vehicle properties affect rollover. These principles are used to develop the stability controller in this chapter. The first generation ESCs implemented yaw control by limiting yaw rates with differential braking [Tseng, 1999]. These ESCs used yaw rate gyros, wheel speed sensors, and lateral accelerometers and the pre-existing anti-lock braking actuator. Although differential braking is still the most common actuator used in ESCs, others such as active suspensions and steer-by-wire, are being developed to enter the market [Vilaplana, 2004; Amberkar, 2004]. Although yaw control has been shown to reduce rollover, Ford utilizes an ESC that couples the yaw and roll dynamics for control and therefore equips their SUVs with a roll rate gyro [van Zanten, 2000]. 87 In this chapter, a simple ESC is developed and implemented on a Blazer in simulation. It uses steer angle and braking to control the vehicle?s stability and to track the driver?s intended path. 5.2 Electronic Stability Control The ESC developed and used in this thesis is a proportional and integral (PI) yaw controller. It uses the vehicle?s brakes to control velocity and controls the steer angle of the front tires. Controller limits are given from the stability threshold derived in Chapter 4. The ESC takes these limits into account and adjusts the steer angle and velocity to keep the vehicle within the stability region. In order to implement the P-I controller, the vehicle model must first be linearized. Placing the linearized equation of the bicycle model into the state space form is shown in Equation (5.1). [ ]??? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?+? ? ? ?? ?? ? ? ? ? ? ? ? ? ? ? ? ? ? ?? ????? =? ? ? ?? ? z f f zz I Ca VMT C r VI C I C VMT C VMT C r 32 2 21 1 & & 5.1 Equations (5.2 ? 5.4) provide the definitions of the constants C1, C2, and C3. raaf CCC +=1 5.2 88 araf CbCaC ???=2 5.3 araf CbCaC ?+?= 22 3 5.4 The tire cornering stiffnesses fC? and rC? represent the total axle cornering stiffness of the front and rear axles, respectively. The axle cornering stiffness is found by adding the cornering stiffness of each tire for that axle. The yaw dynamics are controlled with a PI controller. The transfer function for the PI controller is given in Equation (5.5). s ksk Input Output ip +?= 5.5 Where: kp = Proportional Gain ki = Integral Gain The controller gains were chosen by imputing the desired response frequency of 1 Hz and desired damping of 0.707 into the MATLAB ?place? command. The ?place? command sets the controller?s eigenvalues to achieve the desired output. The proportional gain was 1.0, and the integral gain was -0.5. The Nominal Blazer configuration is simulated with and without the ESC activated. The Fishhook 1a steering profile is once again used. The maneuver velocity was set at 35 mph, yet the vehicle can only negotiate this maneuver with a maximum speed of 32 mph. The ESC response is shown in Figure 5.1. The ESC effectively keeps 89 the vehicle within the stability threshold. It is also seen that as the velocity is decreased, the steer angle is decreased in order to limit the driver the maximum yaw rate to the rollover limit yaw rate. Figure 5.1. ESC Simulated on the Nominal Blazer Configuration 90 5.3 Intelligent Vehicle Model Although statistics were shown in Chapter 1 that ESC?s ability to prevent single vehicle rollover is high, there is room for improvement by using an intelligent vehicle model (IVM) that modifies the ESC?s vehicle model and controller limits. Vehicle payload variation is the most common vehicle property varied. The ESC with the IVM is different from a regular ESC where the vehicle model does not change, and only one stability limit is known. The purpose of the IVM is to update the ESC controller limits as the vehicle?s load condition changes. The IVM utilizes the stability limit for a range of load conditions to know the vehicle?s changing rollover limit. Although not developed in this thesis, sensors and estimation can be used to inform the IVM of weight split and CG height changes. Simulations were conducted to determine how the IVM could serve to increase the effectiveness of ESC in reducing vehicle rollover. Simply adding a full load of passengers to the Blazer can change the CG height by as much as 10%. With this information, the Nominal Blazer was simulated driving through a Fishhook maneuver with a maximum steer angle of 5 degrees at 24 mph. The Nominal Blazer can navigate this maneuver without crossing the stability threshold. However, the Blazer with the driver and four passengers has a CG height 10% higher than the Nominal configuration. This raises the CG height from 0.663 to 0.73 meters. The Blazer with the passengers can only negotiate the Fishhook maneuver at 22 mph. In Figure 5.2, simulation results are seen of the Blazer with the passengers implemented with ESC. Two configurations are 91 shown, the Blazer with and without the IVM. The stability threshold for the Blazer with the passengers is also shown. Figure 5.2. ESC with and without IVM for CG Height Variation The Blazer with the IVM updates the stability threshold from the Nominal configuration of 0.33 g for lateral acceleration and 17.5 deg/s of yaw rate to the new configuration that has the CG height 10% higher. The stability threshold for the loaded condition is 0.295 g for lateral acceleration and 17.0 deg/s for yaw rate. It is seen in Figure 5.2 that without the IVM, the vehicle never activates its ESC because the nominal stability threshold is never breached. Therefore, the vehicle without IVM rolls as seen in Figure 5.3. 92 Figure 5.3. Effect on Dynamic Behavior due to CG Height Move without IVM However, the vehicle with IVM updates the ESC limit for yaw rate and lateral acceleration which causes the ESC to activate. With the IVM, the vehicle never exceeds the stability threshold and does not roll over. Table 5.1 contains the stability thresholds for both vehicle configurations and the maximum yaw rate and lateral acceleration achieved during the maneuver. 93 Table 5.1. ESC with and without IVM for CG Height Change CG Height 0.663 [m] 0.73 [m] 0.73 [m] IVM No No Yes ESC Enabled Enabled Engaged Rollover Incident No Yes No Maneuver Maximum Lateral Acceleration 0.32g 0.32g 0.295g Maneuver Maximum Yaw Rate 17.3 deg/sec 17.3 deg/sec 17.0 deg/sec Stability Threshold Lateral Acceleration 0.33g 0.295g 0.295g Stability Threshold Yaw Rate 17.5 deg/sec 17.5 deg/sec 17.0 deg/sec Utilizing the IVM during CG Height changes appears to have merit, yet there is difficulty in using the IVM during weight split variations. Table 5.2 shows the stability threshold for the Blazer at two different weight splits, 60:40 and 40:60. Table 5.2. Blazer Stability Threshold for Two Weight Splits Weight Split Front:Rear Dynamic States at Rollover 60:40 40:60 Velocity [mph] 30.9 [mph] 24.4 [mph] Yaw Rate [deg/s] 19.2 [deg/s] 26.1 [deg/s] Lateral Acceleration [g] 0.46 [g] 0.51 [g] Roll Angle [deg] 8.10 [deg] 6.84 [deg] Roll Rate [deg/s] 1.4 [deg/s] 1.5 [deg/s] Side Slip Angle [deg] -0.05 [deg] 0.2 [deg] Side Slip Rate [deg/s] 0.5 [deg/s] 0.3 [deg/s] Assuming that an ESC is implemented for the 60:40 weight split, it will use a stability threshold of 19.2 deg/s for yaw rate and 0.46 g of lateral acceleration. The maximum velocity that the 60:40 weight split can safely navigate the Fishhook maneuver is 30.9 mph. After a change in weight split to 40:60, the maximum safe velocity that the vehicle can achieve is lowered to 24.4 mph. However, the maximum yaw rate is 94 increased to 26.1 deg/s and the maximum lateral acceleration is increased to 0.51 g. As was shown in Chapter 4, the vehicle?s understeer gradient decreases as the weight split is shifted rearward. This causes the vehicle to oversteer more as the weight is shifted rearward. For a oversteer vehicle, the maximum lateral acceleration and yaw rate increase for the same maneuver velocity as an understeer vehicle. The use of an IVM in this case of changing weight split is redundant. The ESC limit is set at 19.2 deg/s for yaw rate and 0.46 g for lateral acceleration to start out with. Since a shift of the weight split toward the rear only increases the maximum yaw rate and lateral acceleration, as shown in Figure 5.4, the initial controller limit will suffice to keep the vehicle stable. However, if the weight split is shifted towards the front axle of the vehicle, the rollover limit decreases and the IVM is effective. In this case, the IVM would update the ESC?s stability limit and cause the ESC to engage before rollover occurred. Also, note that as the weight split deviated from the nominal configuration by being shifted to the rear, the nominal controller limit becomes more conservative. As the controller?s limit becomes more conservative, the vehicle?s response potential is not meet. By utilizing the IVM, the controller limit can be updated to avoid the ESC limit from becoming ultra conservative, and causes the vehicle?s response to be maximized. The rollover limits for the weight split variation is shown in Figure 5.4. 95 Figure 5.4. Stability Limits for Weight Split Variation 5.4 Summary and Conclusion In this chapter, a simple ESC was developed and implemented in simulation. The vehicle?s stability threshold for changes in load conditions was used in coordination with the ESC to increase the ESC?s ability to prevent rollover. The use of the stability threshold to change the ESC?s limits is called an Intelligent Vehicle Model since it updates changes to the vehicle?s parameters and stability limit. The significance of the IVM was seen in simulation when the vehicle CG location was modified. The IVM 96 proved to be valuable for changes in CG height. However, the IVM was shown to be redundant in cases where the vehicle?s weight split is shifted rear of the nominal configuration, but useful in keeping the controller limit from being ultra conservative. 97 CHAPTER 6 IMPLEMENTATION ON A SCALED VEHICLE 6.1 Introduction Rollover testing using full-size vehicles is an expensive and somewhat dangerous endeavor. However, if dynamic behavior from the scaled vehicle tested in a controlled environment can predict the dynamic behavior of full-size vehicles, then scaled vehicles offer an alternative safe approach to investigate rollover. In this chapter, an analysis of vehicle rollover is performed on a scaled vehicle. Details of the scaled vehicle and its modifications are given, including development and implementation of a wireless IMU for providing critical vehicle measurements. Simulation data and experimental data of the scaled vehicle are compared to validate the dynamic similarity of the scaled vehicle with a passenger vehicle. Stability limits for the scaled vehicle with variation of the CG height and weight split are then derived using the vehicle simulation. Finally, a stability control system is developed for the scaled vehicle and the stability threshold is used in conjunction with an intelligent vehicle model to increase the stability controller?s effectiveness. 98 6.2 Motivation for using Scaled Vehicles Scaled vehicles have proven to be reliable test beds for a variety of applications such as yaw controller development and tire testing [Brennan, 2000; Burns, 2002; Hallowell, 2003]. Scaled vehicles also provide several advantages when testing and designing. For example, costs associated with a scale model vehicle are significantly lower and modifications are easier to make on a scale vehicle than a full-scale passenger vehicle. Most radio control cars are purchased for less than five hundred dollars, while it is difficult to purchase the cheapest of new passenger vehicle for fewer than ten thousand dollars. The cost of modification and maintenance is also much less expensive on a scaled vehicle than a passenger vehicle. The testing area required for scaled vehicles is much smaller, allowing the testing environment to be more accurately controlled. Compared with the Honda owned vehicle test facility in East Liberty, Ohio, where many vehicle tests are conducted, a scaled test center is much easier to acquire and maintain. Simple, low-cost road simulations have been built for testing scaled vehicles [Brennan 1998]. And pushing the vehicle to its limits in order to observe what happens in the non- linear regions of the vehicle is much safer with a scale vehicle than with a full-sized vehicle [Yih, 2000]. 99 6.3 Scaled Vehicle Description A 1:10 scale radio controlled (RC) car was used as a test bed in order to validate the rollover experiments done in the simulation. The vehicle was modified in order to allow the CG location, spring stiffness, and roll center height to be easily modified. Changing the CG location provides the opportunity to view dynamics occurring at different weight splits. It also provides a means to adjust the distance between CG height and roll center height, which is a crucial parameter when assessing steady state roll [Hac, 2002]. There are two generations of 1/10th scale vehicles which were used to validate simulation results. The first generation test vehicle was configured with a rear wheel drive and front wheel steer configuration to better emulate many of today?s RWD SUVs. Figure 6.1 shows the first generation scaled vehicle. 100 Figure 6.1. First Generation Scale Vehicle Test bed Other notable aspects of the vehicle?s setup include off-road tires, a CG relocator, and IMU board. Scaled vehicle street tires have a relatively high cornering stiffness in comparison with full size vehicles. Therefore, in order to compensate for this effect, knobby tires are used to help scale the cornering stiffness as suggested in [Brennan, 1999]. The steering servo used was a Futaba 3003 which has a response time of 240 deg/s. The CG relocator was created to move the CG in both the longitudinal and vertical directions. It also served as a roll cage that protects the vehicle?s servos, motor and IMU. The first generation test bed was primarily used to record data during fishhook maneuvers and to study how changing the loading condition effects rollover. During the research, several drawbacks were noticed with this test bed. A tennis court was used as a 101 test area for experiments, which provided a small area for testing. Therefore a hard acceleration was needed in order to reach the maneuver speed, which caused major wheel spin on the two-wheel drive vehicle and led to rapid tire wear. A lack of repeatability and measurement of the vehicle?s true velocity were some other faults of the first generation test bed. During testing, radio disturbances would disrupt a maneuver causing the scaled vehicle to frequently veer off its desired course and collide with surrounding objects during experiments, which made testing long and tedious. Additionally, the same vehicle configuration would not consistently cause rollover at the same programmed velocity. Other limitations included a weak roll cage, which was easily bent during rollover events, and a CG relocator design that raised the center of gravity too high when weight split changes were needed. A second generation scaled vehicle was built to compete in NHTSA?s first annual Enhanced Safety Vehicle Competition (ESV). Despite the vehicle being designed with the intention to show a scaled safety technology for the ESV competition it also became a vehicle dynamics and ESC test bed to study rollover. This second generation scaled vehicle is shown in Figure 6.2. 102 Figure 6.2. Second Generation Scale Vehicle Test bed The second generation test bed was also a 1/10th scale vehicle. However, this vehicle was equipped with four-wheel-drive. The four-wheel-drive allowed rear wheel spin to be eliminated during acceleration before entering a maneuver. This vehicle was equipped with a more elaborate CG relocator, which enables the weight split to be changed while retaining a low center of gravity. A 0.3 kg weight is positioned fore or aft to change weight split, and risen up or down along four pieces of ??aluminum all-thread to change the CG height. The weight split can also be changed by adding weights to rods that are bolted into aluminum blocks attached fore and aft of the front and rear axles. The roll cage is stronger than the previous generation in order to better protect the data acquisition system. The black box in Figure 6.2 contains the wireless data acquisition 103 system which enabled improved telemetry. It was both small, and allowed wireless feedback control to be implemented on the test bed. The wireless feedback control system is implemented by sending driver inputs via a steering wheel and pedal assembly. Similar to the ESC described in Chapter 5, the ESC is preset with the yaw rate at the stability limit and decides if the driver?s inputs are outside the vehicle?s stability region. An inertial navigation system monitors the vehicle states and informs the ESC through a wireless connection. If the driver is within the stability threshold, the ESC simply passes the driver?s commands to the radio transmitter. If the driver?s inputs bring the vehicle to the stability threshold, then the controller modifies the inputs accordingly to keep the vehicle from rolling over while maintaining the driver?s intended path as close as possible. The inputs of the driver or the adjusted inputs from the ESC are sent to the radio transmitter which sends commands to the vehicle actuators. The wireless feedback control schematic is shown in Figure 6.3. 104 Figure 6.3. Vehicle Control Schematic 6.4 Data Acquisition The first generation data acquisition system consisted of an inertial sensing module and was constructed to meet the basic dimensional requirements of fitting on a scaled vehicle. The system contained two gyroscopes capable of measuring 150 deg/s at a bandwidth of 40 Hz, one two-axis accelerometer capable of measuring ?2g at 50Hz, a GPS receiver, and a Rabbit microprocessor. The gyroscopes were oriented to obtain roll rate and yaw rate on the scale car. The accelerometer was placed in the horizontal plane to obtain longitudinal and lateral accelerations. The GPS unit provided vehicle position, Wireless Communication Driver Inputs Vehicle Actuators Base Station INS ESC Transmitter On Board Systems 105 velocity, and course measurements, and when used in conjunction with accelerometers and gyroscopes, can also provide measurements of vehicle sideslip [Bevly, 2001]. The Rabbit was set up to record the inertial sensors at 100 Hz and the GPS receiver at 1 Hz for a duration of 45 seconds at a time. After each 40 second sample period, the data must be uploaded to a computer and the unit must be reset. This first generation data acquisition system is shown in Figure 6.4. Figure 6.4. First Generation IMU The second generation data acquisition system consists of an inertial measurement unit (IMU) that uses the same accelerometers and gyros as the first generation IMU. The IMU board that is mounted on the second generation scaled vehicle and includes a six- axis accelerometer/gyroscope package (3 accelerometers and 3 gyros), a GPS receiver, a wireless transmitter, a Rabbit 2000 microprocessor, and power management components. The wireless transmitter is a radio modem. It is used to collect inertial and position data and send it to the base station where the ESC analyzes the vehicle?s dynamics. The components are mounted on a printed circuit board and placed in a protective box which is then mounted to the vehicle. The second generation data acquisition package can be seen in Figures 6.5 and 6.6. 106 Figure 6.5. Wireless Data Acquisition Package 107 Figure 6.6. Second Generation Data Acquisition PCB Board. IMU and Microprocessor shown on top, GPS Receiver and Radio Modem shown on bottom All data sampling is controlled by the Rabbit 2000 microprocessor which uses C code to log and transmit data. The microprocessor begins by collecting GPS and IMU data to be sent to the base station. It then prepares a packet of data to be sent through the wireless transmitter/receiver (MaxStream Xcite 9) which sends the data to the base 108 station. This data packet can be either literal data received from the GPS and IMU units or raw data parsed from their messages and processed onboard. The data is received by a wireless receiver plugged into the base station and is further processed by the ESC to determine the vehicle?s stability. The dimensions of the INS box are 6.75? x 4.25? x 2.5?. Table 6.1 includes the specifications of the IMU. 6.5 Scaled Vehicle Simulation and Experiment Comparison Simulations of the scaled vehicle and scaled vehicle experiments were used to compare scaled vehicle dynamics with passenger vehicle dynamics. The simulations and experiments were also used to verify the load dependant stability limits that were discussed in Chapter 4. Vehicle simulations were compared with real experimental data in order to determine if scaled vehicle dynamics correlate with passenger vehicle dynamics. If the vehicle model derived for the passenger vehicle (which was verified in Chapter 3) also matches experiment data for the scaled vehicle, then scaled vehicles are valid tools for Table 6.1 IMU Specifications Component IMU605 - Accelerometer IMU605 ? Gyroscope RCB-LJ? GPS Update Rate (Hz) Up to 256 (default 60) Up to 256 (default 60) 4 Range +/- 2g +/- 150 deg/sec n/a Output format Digital UART Digital UART Digital UART 109 researching the vehicle dynamics that effect passenger vehicle rollover. Figure 6.7 shows experimental and simulation data, which has not been filtered, of the first generation scaled vehicle during a maneuver. Figure 6.7. First Generation Scaled Vehicle Experimental versus Simulation Data Note that the same governing equations used for the Blazer?s dynamics also accurately capture the dynamics of the scaled vehicle. 110 6.6 Stability Limit Development The stability limit developed for the Blazer in Chapter 4 was also developed for the scaled vehicle and tested with experiments in order to assess its validity. The simulation was limited to developing stability thresholds by changing only one vehicle property at a time. Therefore, two separate stability thresholds were developed for the load condition: CG height variation and weight split variation. Vehicle simulations were used to determine the stability threshold for different vehicle loading conditions. This was done by varying individual parameters such as the CG height, and recording the velocity, lateral acceleration, yaw rate, and roll angle at which rollover is detected. For these simulation experiments, a Fishhook steering profile was chosen with a maximum steer angle of 5 degrees and a maximum steering rate of 40 degrees per second, both measured at the wheels. This profile, shown in Figure 6.8, was filtered using a 2nd order Butterworth filter with a 1 Hz bandwidth. The velocity profile for the maneuver was set to a constant in order to simplify the results. 111 Figure 6.8. Fishhook Steering Profile for Stability Limit Development The second generation scaled vehicle was used to determine the stability limit since it is the vehicle that was used in the experiments. For a given configuration, the test maneuver was repeated with the velocity being increased by 0.1 MPH increments until rollover occurred. NHTSA defines rollover to be the point at which both tires on one side of the vehicle are lifted by two inches [Forkenbrock, 2003]. For this simulation experiment, TWL over the duration of 0.4 seconds was declared to be a rollover event since the simulation cannot determine the height of tire lift. For the simulation, TWL is defined as the instance that wheel loads on one side of the vehicle go to zero. The stability limit was defined as the lateral acceleration, yaw rate, and roll angle at which a rollover event occurs. This stability limit was calculated for a variety of different vehicle properties [Whitehead, 2004]. 112 The first scaled vehicle experiment varied the CG height with a neutral suspension setup (i.e. the weight split was set to 50:50, and the front and rear springs were the same stiffness). Figure 6.9 shows the stability threshold for the vehicle test bed as a function of CG height. The vehicle requires a higher lateral acceleration to rollover as the CG height is lowered, which follows intuition and the results of the static stability factor. The yaw rate required for rollover in this maneuver also increases as the CG height is lowered. Once the CG height is below a certain height, the vehicle no longer experiences rollover but begins to slide. This is due to the fact that at low CG heights, the moment created by the un-sprung mass height and the tire lateral force does not provide enough moment to roll the vehicle before the tires saturate. The vehicle no longer rolls over because of a decrease in the moment in the lateral weight transfer equation caused by the spring and damper loads, and their distance from the vehicle?s centerline, S. This decrease in the spring and damper loads is caused by the decrease of the moment caused by the lateral acceleration of the sprung mass and the distance between the roll center and the CG height. The springs and dampers transfer this moment to the wheels, and since this moment is decreased by the lowering of the CG height, the lateral weight transfer is decreased as well. 113 Figure 6.9. Stability Limits for CG Height Variation Note that there is a difference of 0.375 g in the lateral acceleration rollover limit. The vehicle requires 0.472g of lateral acceleration to roll at a CG height of 0.1m, while .097g will roll the vehicle with a CG height of 0.35m. There is also a difference of 40.3 deg/sec in the yaw rate limit for rollover. The rollover yaw rate threshold is 71.8 deg/sec and 23.0 deg/sec for the CG heights of 0.1m and 0.35m respectively. In order to develop the stability threshold for the longitudinal CG variation in the simulation, the vehicle?s weight split was varied from 30:70 to 70:30, front to rear axle weights, respectively, and the CG height was held constant at 0.15 meters. The results of this simulation are shown in Figure 6.10. 114 Figure 6.10. Stability Limits for the Weight Split Variation Both the yaw rate and lateral acceleration stability limits decrease as weight is shifted from 30:70 to 36:64. The stability limits for the 30:70 weight split are 0.183 g and 44.8 deg/s. The rollover limits of the scaled vehicle for the weight split simulation vary by 0.186 g for lateral acceleration and 15.6 deg/s for yaw rate. Stability limits are 0.309g and 58.2 deg/s for a 70:30 weight split, and 0.166 g and 42.6 deg/s for a 36:64 weight split. A higher lateral acceleration and yaw rate is required for vehicle rollover as the weight is shifted towards the front axle. It is interesting to note that the vehicle understeer gradient also increases as the weight split is shifted to the front. Therefore, a generalization can be made that an understeer vehicle has a higher stability threshold than an oversteer vehicle, although this trend is not always followed. 115 The change in the stability threshold during these experiments shows how important vehicle loading data is to the ESC. For example, the stability limit changes from 0.270 g to 0.183 g and 54.6 deg/s to 44.9 deg/s as the vehicle?s weight split is changed 60:40 to 40:60. Therefore, if the ESC model is not updated it will use 0.270 g and 54.6 deg/s as the stability limit, which are .087g and 9.7 deg/s more than the actual stability limit. This may not seem like much, but the stability threshold is actually off by 32.22% for lateral acceleration and 17.77% for the yaw rate. Knowing the stability threshold as a function of loading condition gives the ESC better ability to keep the vehicle stable by taking out errors in the controller?s reference limits. 6.7 Stability Limit Validation To validate the stability limits derived from the simulations, experiments which varied the properties center of gravity height and weight split were conducted with the scaled vehicle test bed. The first generation scale vehicle test bed operating on a concrete tennis court was used for these experiments. This environment was preferred over a parking lot due to the smoothness of the surface, and lack of gravel, and minimal grade. Lack of gravel is important because gravel is perceived by the scaled vehicle test bed as a basketball sized rock. The grade of the test environment used a constant value of 2 degrees. The size of the test area was small for high speed maneuvers, and radio frequency disturbances frequently disrupted experiments. 116 The Fishhook maneuver used in the simulation was also used in the experimental tests. The vehicle was driven by a computer in order to make the maneuver accurate and repeatable. The computer was programmed to accelerate the vehicle for four seconds in order to reach the desired velocity before the maneuver was initiated. Unlike the NHTSA study, where the vehicle begins to coast once the maneuver is entered, the velocity for the experiment was held constant throughout the maneuver. Figure 6.11 shows the experiment results of the CG height variation for the stability threshold, which follows the static stability factor and the simulation results. The CG height was varied from 0.128 to 0.195 meters which corresponds a SSF variation of 0.4564 to 0.6138. The results of the experiment closely match the simulation. For a center of gravity height of 0.16 meters, the simulation shows the vehicle rolling over at a velocity of 6.25 MPH while the rollover velocity of the experiment was recorded at approximately 6.5 MPH. The proximity of these results provides validity gives value to the developed stability limits. Additionally, it is important to note that the same trend is evident in both experiment and simulation. 117 5.5 6 6.5 7 7.5 8 8.5 9 9.5 0.12 0.14 0.16 0.18 0.2 CG Height (m) Ro llo ve r V elo cit y ( MP H) Figure 6.11. Experimental Stability Limit for Variations in the CG Height The second set of experiments varied the longitudinal weight distribution. The first generation test bed was again used at the tennis court to perform this test. The maneuver and computer controls were kept the same as in the previous CG height variation experiments. Figure 6.12 shows the experimental stability limit for rollover. It can be seen that the experimental results validate the simulation shown previously in Figure 6.10. Although weight split variation in the simulation and experiment follow the same derived trend, their rollover velocities do not match. This is due to the center of gravity height being higher in the simulation than the experiment. With the CG height setup in the experiment, the vehicle will not roll at weight splits favored to the front because of tire saturation. Qualitatively, the experimental results do follow the simulation trend which acknowledges that the vehicle is less prone to rollover as the weight split is shifted towards the front of the vehicle. 118 9 9.5 10 10.5 11 11.5 12 12.5 13 13.5 14 40 50 60 70 80 90 Percent Total Weight on Rear Axle Ro llo ve r V elo cit y ( MP H) Figure 6.12. Weight Split Variation Experiment Stability Limit 6.8 Electronic Stability Control As described in Chapter 5, the ESC is implemented as a PI controller via MATLAB simulation. The ESC?s purpose is to keep the vehicle within the stability threshold. The stability limit is defined as the lateral acceleration and yaw rate at which the vehicle will experience TWL during a fishhook maneuver. The ESC monitors these two states and when the stability threshold is approached, the controller adjusts steer angle and velocity to maintain stability. When the ESC enabled vehicle is below the 119 stability limits, the ESC does not modify with the driver?s inputs. However, once the driver?s inputs exceed the stability limit, the ESC adjusts the steer angle and velocity to limit the yaw rate to the yaw rate stability limit and limit the lateral acceleration to the lateral acceleration stability limit. Since the scaled vehicle used was an off-the-shelf radio control vehicle, the only actuators available to the ESC on the scaled vehicle testbed were the steering servo and velocity control unit which brakes and accelerates the rear wheels. Figure 6.13 shows a simulated scaled vehicle with and without the ESC enabled. The maneuver that the vehicle is attempting to negotiate is a Fishhook maneuver at a maximum steer angle of 5 degrees with a maximum steer rate of 40 degrees per second at the tires. The driver?s desired velocity is 8 MPH. Figure 6.13. Results for the Fishhook Maneuver with and without ESC enabled 120 In order to maintain stability in this maneuver, the ESC adjusts the steering angle and actuates the brakes in order to not exceed the lateral acceleration and yaw rate limit. However, the ESC still tries to follow the driver?s desired path by calculating the desired radius of the turn. The lateral acceleration stability threshold for this particular vehicle setup is 0.25g and the yaw rate stability threshold is 53.6 deg/sec. Figure 6.13 reveals how the ESC keeps the vehicle within the stability threshold when it would otherwise become unstable. 6.9 Intelligent Vehicle Model (IVM) As a vehicle?s payload changes, the Intelligent Vehicle Model (IVM) gives the controller knowledge of the change in the stability limit and adjusts the controller limits accordingly. Sensors, in conjunction with vehicle parameter estimation techniques would constantly monitor the vehicle weight split and CG height. The stability limit for the load condition would then be used as the controller?s limit in the ESC. In order to see how the IVM can improve the ESC effectiveness, a simulation was performed with the scaled vehicle model using a 60:40 and 40:60 weight split (front to rear axle weights, respectively). The CG height and static stability factor were held constant for this property variation. Again, a Fishhook maneuver with a maximum steer angle of 5 deg and maximum steer rate of 40 deg/s was used for the simulation. The 121 vehicle longitudinal velocity was set to 6 MPH and held constant throughout the maneuver. In the initial configuration, the scaled vehicle had a weight split of 60:40 front to rear. The stability limit for this load condition is 0.270 g for lateral acceleration and 53.6 deg/sec for yaw rate. The ESC vehicle model is initially programmed to use these state limits in order to maintain stability. Since the vehicle does not exceed this limit during the maneuver, the ESC is never activated. It can be seen from the wheel loads in Figure 6.14 that the vehicle does not roll, but one wheel lift does occur. Figure 6.14. Results for the Fishhook 1a with 60:40 Weight Split 122 In the second configuration, the weight split is modified to 40:60. This new load condition results in a new stability limit of 0.183 g for lateral acceleration and 44.9 deg/s for yaw rate. The simulation was performed assuming the ESC is not equipped with the IVM. Therefore, the vehicle model still uses 0.270 g and 53.6 deg/s as the rollover stability limit. The same maneuver used for the 60:40 configuration causes this vehicle configuration to exceed the stability limit, causing rollover. Since the vehicle model is not updated, the ESC never intervenes and vehicle rollover occurs as shown in Figure 6.15. Figure 6.15. Results for the Fishhook 1a with 40:60 Weight Split 123 However, if the vehicle was equipped with the IVM along with the ESC, then the ESC would have known the correct stability limits. Therefore, the ESC could use the updated stability limit for the modified weight split. This would have enabled the ESC to correctly prevent the rollover as shown in Figure 6.13. A comparative analysis of these configurations is given in Table 6.2. Table 6.2. ESC with and without IVM Weight Split Front:Rear 60:40 40:60 40:60 IVM No No Yes ESC Engaged Disengaged Engaged Rollover Incident No Yes No Maneuver Maximum Lateral Acceleration 0.25g 0.25g 0.18g Maneuver Maximum Yaw Rate 53.6 deg/sec 53.6 deg/sec 44.0 deg/sec Stability Threshold Lateral Acceleration 0.270g 0.183g 0.183g Stability Threshold Yaw Rate 54.6 deg/sec 44.9 deg/sec 44.9 deg/sec 6.10 Summary and Conclusion In this chapter, scaled vehicles have been shown to be valuable in evaluating vehicle loading conditions and their correlation with rollover. It has also been shown that scaled vehicles are valuable tools for studying ESC systems. Section 6.6 quantified a stability threshold in simulation for a scaled vehicle, and proposed the use of an IVM system. The IVM updates the stability threshold in the vehicle model for an ESC system 124 as the vehicle?s payload changes. Section 6.7 verified the stability threshold with scaled vehicle experiments. Section 6.9 provided a scenario in which implementing an IVM would prevent rollover, while an ESC without IVM would not. Furthermore, this chapter has shown that scaled vehicles can be used to validate simulation results as well as provide valid information on the behavior of full scale passenger vehicles. 125 CHAPTER 7 CONCLUSION 7.1 Summary An investigation of rollover propensity using simulation and experimental data has been presented. A vehicle model was developed and validated for studying rollover propensity. This vehicle model was then use to explore the changes that vehicle parameters have on rollover propensity. Stability thresholds for a range of vehicle load conditions were developed as well. An ESC was developed for mitigating rollover in simulation. Also, an IVM was developed and implemented with the ESC in order to improve ESC?s effectiveness in preventing rollover. Finally, a scaled vehicle was used to validate the research. A summary of the research highlights in each chapter is provided below followed by a discussion of recommendations for future research. In Chapter 2, a vehicle model was created to capture the major dynamics that govern vehicle rollover. Both non-linear yaw and roll dynamics were used to reproduce the transient dynamics in these axes. Newton?s method, using FBDs, were used to develop the governing equations of motion. The non-linear, Pacejka, ?Magic Formula? tire model was used to model the tires. The diagrams and equations were detailed in the order in which they were used to develop the vehicle model. 126 Chapter 3 contains the evaluation and validation of the vehicle model developed in Chapter 2. In Section 2.2, the steady state results of the steady state roll equations were compared with those of the transient roll equations. The transient roll equations yielded the same steady state dynamic state values as the steady state roll equations. This validated the steady state dynamic states of the transient roll equations used in the vehicle model. Section 2.3 contained NHTSA Phase IV data which validated the transient dynamics of the yaw and roll equations in the vehicle model by matching with the simulation data. Experimental results using a transient maneuver with a Blazer at Auburn University provided further validation of the transient vehicle model. Finally, a discussion on the effect of roll centers below the un-sprung mass on weight transfer was given. In Chapter 4, the correlation between vehicle load condition and rollover propensity was investigated. Analysis of the various load configurations of the NHTSA Blazer led to the development of a vehicle rollover stability threshold. The rollover stability threshold was derived as a function of various vehicle dynamic states such as lateral acceleration, yaw rate, and roll rate. A simple ESC was developed in Chapter 5, and it is simulated on a Blazer. An IVM was developed to aid the ESC in preventing rollover. It utilized the stability threshold created in Chapter 4 with knowledge of vehicle properties changes to update the controller?s vehicle model and state limits. It was shown that the IVM could improve ESC effectiveness during CG height changes. However, the IVM proved to be redundant as the vehicle?s weight split was shifted to the rear axle. 127 In Chapter 6, the experimental test bed developed in this study allowed the robustness of ESC to be evaluated via changing vehicle properties such as weight split, center of gravity height, and suspension setup. These scenarios simulate changes in vehicle loading and suspension modifications, which are relevant in today?s world of transporting children to soccer practice in SUVs, low rider hot rods, and modified four wheel drive vehicles. The scaled vehicle was used as a test bed to validate the stability limits created in simulation, explore vehicle properties that influence rollover, and investigate ESC algorithms. It is equipped with telemetry which consists of GPs and INS sensors. This chapter also quantified a rollover stability limit for the scaled vehicle and utilized the IVM, which updates the stability threshold in the vehicle model of an ESC. This chapter developed a scenario demonstrating that the IVM would aid in preventing rollover. The significance of the Intelligent Vehicle Model can be seen from the exercises in Section 6.9. IVM is a tool that ESC can use to improve its ability to prevent vehicle rollover. 7.2 Recommendations for Future Work Results form this thesis should extended to derive the stability limit as a function of more vehicle properties. The vehicle properties that should be studied next include front and rear roll center heights as well as front and rear roll stiffness and damping. The 128 vehicle states that should be monitored in the stability threshold include yaw rate, lateral acceleration, roll angle, roll angle rate, side slip rate, and side slip angle. The effect of tire properties should be further investigated for a wider range of paths and maneuvers, not just for steering profiles. This will provide a more accurate analysis of how a tire?s cornering stiffness and peak lateral force effect vehicle rollover. A higher DOF model, such as Adams or CARSIM, could also be used to derive the stability limits. The higher DOF models will include the suspension kinematics and dynamics of the un-sprung mass. These results should be compared with the results of the work presented in this thesis. Since the ESC and IVM were only tested in simulation, further research could be conducted via experiments. Experimental data verifying the effectiveness that IVM has on preventing rollover is worth acquiring. Either a passenger vehicle or scaled vehicle could be used. Though the research presented in this thesis used the Fishhook maneuver, other maneuvers such as the ISO-3888 lane change maneuvers could be used. This maneuver requires a driver to follow a marked path instead of a steer profile. This would provide further validation of the ESC and IVM developed in this thesis. Methods to inform the IVM of vehicle property changes should be explored. String potentiometers can be tuned to measure normal wheel loads. Accelerometers from an on-board IMU could be used with the wheel loads to determine the vehicle?s weight split in a static case. Methods such as using a Kalman Filter to estimate the weight split should also be developed. A method to accurately estimate the CG height should also be derived and implemented. These methods to determine the vehicle load condition status could be implemented in simulation and experiments. 129 The scaled vehicle research in this thesis provided a relatively inexpensive means of acquiring real vehicle data. While this test bed provided experimental data, it could be improved. Because of the size of the 1:10th scale vehicle issues such as implementing sensors and finding a suitable test environment were never fully solved. For instance, a pebble to a full size vehicle is magnified 10 times for the 1:10th scale vehicle, and a seemingly smooth surface for a passenger vehicle is rough to a scaled vehicle. For future work, a larger scale vehicle (1:4th) should be used than the one used in this study. A larger vehicle, though more expensive, would be easier to instrument and might ultimately produce more accurate results. 130 REFERENCES [1] Allen, R. W., Szostak,H.T, Rosenthal, T. J. and Klyde,D. H. 1990. ?Field Testing and Computer Simulation Analysis of Ground Vehicle Dynamic Stability.? SAE Paper No. 900127. [2] Amberkar, S., et. al., ?A Control System Methodology for Steer by Wire Systems.? 2004. SAE Paper No. 2004-01-1106. [3] Baumann, F.W., et. al., ?Effects Causing Untripped Rollover of Light Passenger Vehicles in Evasive Maneuvers.? 2004. SAE Paper No. 2004-01-1057. [4] Bevly, D. M., Ryu, J., Sheridan, R., Gerdes, J. C, "Integrating INS Sensors with GPS Velocity Measurements for Continuous Estimation of Vehicle Side- Slip and Tire Cornering Stiffness," Proceedings of the 2001 American Control Conference, Vol.1, June 2001, pp.25-30. [5] Brennan, S., et. al., ?The Illinois Roadway Simulator ? A Hardware-in-the-Loop Testbed Vehicle Dynamics and Control.? Proceedings of the American Control Conference. June, 1998. [6] Brennan, S. and Alleyne, A. ?Using Scale Testbed: Controller Design and Evaluation.? IEEE Control Systems Magazine, vol 21, 2001, pp. 15-26. [7] Brennan, S. and Alleye, A. ?Modeling and Control Issues Associated With Scaled Vehicles,? 1999, University of Illinois at Urbana-Champaign. [8] Brennan, S. and Alleyne, ?The Illinois Roadway Simulator: A Mechatronic Testbed for Vehicle Dynamics and Control.? IEEE Transactions on Mechatronics. Vol. 5, No. 4, pp. 349-359, December 2000. [9] Brennan, S., ?Similarity Conditions for Comparing Closed-Loop Vehicle Roll and Pitch Dynamics.? Proceeding of the 2004 American Control Conference. Boston, MA. June 30 ? July 2, 2004. [10] Broulhiet, G., ?The Suspension of the Automobile Steering Mechanism: Shimmy and Tramp?, Bull Soc. Ing. Civ. Fr. 78, pp. 540-554, July 1925. [11] Burns, Stephen R., O?Brien, Richard T., Piepmeier, Jenelle A., ?Steering Controller Design Using Scale-Model Vehicles,? 2002, IEEE Paper No. 0- 7803-7339-1/02 [12] Carlson, C. R., and Gerdes, J. C., ?Optimal Rollover Prevention with Steer by Wire and Differential.? Proceedings of IMECE 2003. [13] Christos, J. P. and Guenther, D. A. ?The Measurement of Static Rollover Metrics,? 1992, SAE Paper No. 920582 [14] Cooperrider, N. K., Hammound, S. A. and Thomas, T.M. ?Testing and Analysis of Vehicle Rollover Behavior,? 2000, SAE Paper No. 900366 [15] Day, T. D., and Garvey, J. T., ?Applications and Limitations of 3-Dimensional Vehicle Rollover Simulation.? 2000. SAE Paper No. 2000-01-0852. [16] D?Entremont, K. L., Zhengyu, L. and Nalecz, A. G. ?An Investigation into 131 Dynamic Measures of Vehicle Rollover Propensity,? 1993, SAE Paper No. 930831 [17] Dixson, J. C., ?Tires, Suspension and Handeling? Society of Automotive Engineers. Warrendale, PA. ISBN: 1-56091-831-4. [18] Dugoff, H., Fancher, P.S., Segel, L., ?Tire Performance Characteristics Affecting Vehicle Response to Steering and Braking Control Inputs.? Report for Tire System Section, Office of Vehicle Systems Research. National Bureau of Standards, Washington, D.C. August, 1969 [19] Dugoff, H., Fancher, P.S., Segel, L., ?An Analysis of Tire Traction Properties and their influence on Vehicle Dynamic Performance.? SAE Paper 700377, 1970 [20] Eger, R., et. al., ?Rollover Simulation Based on a Nonlinear Model.? 1998. SAE Paper No. 980208. [21] Ellis, J. R., ?The Dynamics of Vehicles During Braking.? Symposium on the Control of Vehicles, Institution of Mechanical Engineers, June 1963 [22] Fatality Analysis Reporting System (FARS) Web-Based Encyclopedia. National Center for Statistics and Analysis, National Highway Traffic Safety Administration, U.S. Department of Transportation. http://www- fars.nhtsa.dot.gov/. [23] Forkenbrock, G.J., Garrott, W.R., Heitz, M. and O?Hara, B.C. NHTSA. 2002. ?A Comprehensive Experimental Examination of Test Maneuvers that May Induce On-Road, Untripped, Light Vehicle Rollover ? Phase IV of NHTSA?s Light Vehicle Rollover Research Program? (October). Report No. DOT HS 809 513. [24] Forkenbrock, G. J., Garrott, W.R., Heitz M. and O?Hara, B.C. 2003. ?An Experimental Examination of J-Turn and Fishhook Maneuvers that May Induce On-Road, Untripped, Light Vehicle Rollover.? SAE Paper No. 2003-01-1008. [25] Fromm, H., ?Brief Report on the History of the Theory of Shimmy?, NACA TM 1395, pp. 181, 1954. [26] Garrott, W. R., and Heydinger, G. J. 1992. An Investigation, Via Simulation, of Vehicle Characteristics that Contribute to Steering Maneuver Induced Rollover. SAE Paper No. 920585. [27] Gillespie, T. D. 1992. ?Fundamentals of Vehicle Dynamics.? Society of Automotive Engineers. Warrendale, PA. ISBN: 1-56091-199-9. [28] Ginsberg, J. H., ?Advance Engineering Dynamics.? 2nd ed. Cambridge University Press, 1998. ISBN: 0-521-47021-8. [29] Hac, A. 2002. Rollover Stability Index Including Effects of Suspension Design. SAE Paper No. 2002-01-0965. [30] Hac, A., et. al., ?Detection of Vehicle Rollover.? 2004. SAE Paper No. 2004-01- 1757. [31] Hallowell, Stephen J. and Ray, Laura R. ?All Wheel Driving Using Independent Torque Control of Each Wheel, 2003,? IEEE Paper No. 0-7803-7896-2/03 [32] Hilton, J., and Shankar, U., ?Motor Vehicle Traffic Injury and Fatality Estimates ? 2002 Early Assessment,? Report No. DOT HS 809 586, May 2003, p. 15 [33] Huang, J., et. al., ?Control Oriented Modeling for Enhanced Yaw Stability and 132 Vehicle Steerability.? Proceeding of the 2004 American Control Conference. Boston, MA. June 30 ? July 2, 2004. [34] Johansson, B., and Gafvert, M., ?Untripped SUV Rollover Detection and Prevention.? 43rd IEEE Conference on Decision and Control. December 14 ? 17, 2004. [35] Klein, T. M., ?A Statistical Analysis of Vehicle Rollover Propensity and Vehicle Stability.? 1992. SAE Paper No. 920584. [36] Milliken, D. L. and Milliken, W. F. 1995. Race Car Vehicle Dynamics. Society of Automotive Engineers. Warrendale, PA. ISBN: 1-56091-526-9. [37] Morrison, W. R. B., ?On the Lateral Dynamics of an Automobile in Accelerated Motions.? Mechanical and Chemical Engineering Transactions, The Institution of Engineers, Australia, November 1967 [38] Newton, I., ?Philosophi? Naturalis Principia Mathematica.? Londini : Jussu Societatis Regiae ac Typis Josephi Streater. Prostat apud plures bibliopolas, 1687. [39] NHTSA. Phase IV Rollover Testing. 2002. George Washington Transportation Research Institute. [40] NHTSA 2005: SafeCar.gov, Rollover. 18, Oct. 2005, < http://www.nhtsa.dot.gov/cars/testing/ncap/Rollover/Index.htm>. [41] Nishio, A., et. al., ?Development of Vehicle Stability Control System Based on Vehicle Sideslip Angle Estimation.? 2001. SAE Paper No. 2001-01-0137. [42] Pacejka, H. B., Bakker, E., and Nyborg, L., ?Tyre Modelling for Use in Vehicle Dynamics Studies.? SAE Paper No. 870421, 1987 [43] Pacejka, H. B., Bakker, E., and Lidner L., ?A New Tire Model with an Application in Vehicle Dynamics Studies.? SAE Paper No. 890087, 1989. [44] Plumlee, J. H., et. al. ?Control of a Ground Vehicle Using Quadratic Programming Based Control Allocation Techniques.? 2004. [45] Ponticel, P. 2003. ?Dynamic Testing Rollover on the Way,? Automotive Engineering International (November 2003): 26-28. [46] Radt, H. S., and Milliken, W. F., ?Motions of Skidding Automobiles.? Paper 205- A, Presented at SAE Summer Meeting, Chicago, June 1960 [47] Radt, H. S. and Pacejka, H. B., ?Analysis of the Steady-State Turning of an Automobile.? Symposium on the Control of Vehicles, Institution of Mechanical Engineers, June 1963 [48] Ryu, J., et. al., ?Vehicle Sideslip and Roll Parameter Estimation using GPS.? Design Division Stanford University. [49] Schubert, P. J., et. al. ?Electronic and Algorithms for Rollover Sensing.? 2004. SAE Paper No. 2004-01-0343. [50] Travis, W., Whitehead, R., Bevly, D. M., and Flowers G. T. ?Using Scaled Vehicles to Investigate the Influence of Various Properties on Rollover Propensity.? Proceedings of the 2004 American Controls Conference, Boston, 30 June ? 2 July, 2004. [51] Tseng, H. E., et.al., ?Technical Challenges in the Development of Vehicle Stability Control System.? Proceedings of the 1999 IEEE International Conference on Control Applications. August 22-27, 1999. 133 [52] U.S. Department of Transportation, National Highway Traffic Safety Administration, The Economic Impact of Motor Vehicle Crashes 2000 (Washington, DC: 2002), also available at http://www.nhtsa.dot.gov/people/economic, as of December 2002. [53] van Zanten, A.T., ?Bosch ESP Systems: 5 Years of Experience.? 2000. SAE Paper No. 2000-01-1633. [54] Viano, D. C. and Chantal, P. ?Case Study of Vehicle Maneuvers Leading To Rollovers: Need for a Vehicle Test Simulating Off-Road Excursions, Recovery and Handling,? 2003 SAE Paper No. 2003-01-0169. [55] Vilaplana, M.A., et. al., ?Control of Sideslip and Yaw Rate in Cars Equipped with 4-Wheel Steer-by-Wire.? 2004. SAE Paper No. 2004-01-2076. [56] Voelcker, John. ?Top Ten Tech Cars: A Year of Stability.? IEEE Spectrum. March 2005: 22-30. [57] Whitehead, R., Travis, W., Bevly, D. M., Flowers, G. T., ?A Study of the Effect of Various Vehicle Properties on Rollover Propensity,? 2004, SAE Paper No. 2004-01-2094. [58] Wilde, J. R., et. al. ?Experimental Evaluation of Fishhook Maneuver Performance of a Kinetic Suspension System.? 2005. SAE Paper No. 2005-01-0392. [59] Yih, Paul. ?Radio Controlled Car Model as a Vehicle Dynamics Test Bed,? Mechanical Engineering Department, Stanford University, Sept. 2000 134 APPENDICES 135 APPENDIX A VEHICLE PROPERTIES A.1 Introduction Appendix A.2 contains a list and description of the vehicle properties used in the vehicle model developed in Chapter 2. Appendix A.3 contains the values of the vehicle properties for each of the Blazer configurations used in the NHTSA Phase IV experiments. 136 A.2 Garage Legend Name Description Units MT Vehicle Total Mass kg M Vehicle Sprung Mass kg m Vehicle Un-Sprung Mass kg g Gravitational Constant m/s2 Wt Total Vehicle Weight N Iz Vehicle Yaw Inertia N-m-s2 Ix Vehicle Roll Inertia N-m-s2 WB Vehicle Wheelbase m a Distance from the CG to the front axle m b Distance from the CG to the rear axle m hcgM CG height of Sprung Mass m hcgm CG height of Un-Sprung Mass m hrcf Front Roll Center height m hrcr Rear Roll Center height m trkf Front vehicle track m trkr Rear vehicle track M Skf Distance between front springs M Skr Distance between rear springs M Sbf Distance between front dampers M Sbr Distance between rear dampers M Karbf Front Stabilizer, (anti-roll bar) Stiffness N-m/rad Karbr Rear Stabilizer, (anti-roll bar) Stiffness N-m/rad Ksf Front Spring Stiffness N/m Ksr Rear Spring Stiffness N/m Bf Front Shock Damping Coefficient N-s/m Br Rear Shock Damping Coefficient N-s/m 137 A.3 Garage Vehicles and Parameters Parameter Vehicle Description Nominal Blazer RRR Blazer RMB Blazer MT 1907 kg 1988.78 kg 2237.71 kg M 1525 kg 1606.78 kg 1855.71 kg m 382 kg 382 kg 382 kg g 9.81 m/s2 9.81 m/s2 9.81 m/s2 Wt 18707.67 N 19509.93 N 21951.94 N Iz 3833.31 N*m*s2 4094.98 N*m*s2 N*m*s2 Ix 734.04 N*m*s2 813.49 N*m*s2 N*m*s2 WB 2.718 m 2.718 m 2.718 m a 1.216 m 1.216 m 1.522 m b 1.502 m 1.502 m 1.196 m hcgM 0.6629 m 0.6960 m 0.6614 m hcgm 0.35 m 0.35 m 0.35 m hrcf -0.1 m -0.1 m -0.1 m hrcr 0.35 m 0.35 m 0.35 m trkf 1.445 m 1.445 m 1.445 m trkr 1.405 m 1.405 m 1.405 m Skf 0.7747 m 0.7747 m 0.7747 m Skr 0.9906 m 0.9906 m 0.9906 m Sbf 0.7747 m 0.7747 m 0.7747 m Sbr 0.7620 m 0.7620 m 0.7620 m Karbf 700 N/deg 700 N/deg 700 N/deg Karbr 400 N/deg 400 N/deg 400 N/deg Ksf 75000 N/m 75000 N/m 75000 N/m Ksr 70000 N/m 70000 N/m 70000 N/m Bf 5000 N*s/m 5000 N*s/m 5000 N*s/m Br 4000 N*s/m 4000 N*s/m 4000 N*s/m 138 APPENDIX B SCALED VEHICLE PROPERTIES B.1 Introduction Appendix B contains the parameters of the scaled vehicle. These parameters are used in the vehicle simulation and come from the scaled vehicle that was used in the experiments. 139 B.2 Scaled Vehicle Parameters Parameter Description Value Units MT Vehicle Total Mass 3.20 kg M Vehicle Sprung Mass 2. kg m Vehicle Un-Sprung Mass 0.276 kg g Gravitational Constant 9.81 m/s2 Wt Total Vehicle Weight 31.392 N Iz Vehicle Yaw Inertia 0.20 N-m-s2 Ix Vehicle Roll Inertia 0.10 N-m-s2 WB Vehicle Wheelbase 0.2556 m a Distance from the CG to the front axle 0.1374 m b Distance from the CG to the rear axle 0.1182 m hcgM CG height of Sprung Mass 0.1565 m hcgm CG height of Un-Sprung Mass 0.020 m hrcf Front Roll Center height 0.010 m hrcr Rear Roll Center height 0.0127 m trkf Front vehicle track 0.1683 m trkr Rear vehicle track 0.1746 M Skf Distance between front springs 0.0492 M Skr Distance between rear springs 0.0762 M Sbf Distance between front dampers 0.0492 M Sbr Distance between rear dampers 0.0762 M Ksf Front Spring Stiffness 300 kN/m Ksr Rear Spring Stiffness 300 kN/m Bf Front Shock Damping Coefficient 100 N-s/m Br Rear Shock Damping Coefficient 100 N-s/m 140 APPENDIX C VEHICLE SIMULATION C.1 Introduction There are 15 MATLAB functions and scripts that make up the vehicle simulation. As stated in Chapter 2, the vehicle simulation is 3 DOF including the yaw and roll transient dynamics. The trapezoidal method is used to solve the integrals. The time step must be set at .001 seconds to achieve good results when using this method. The various pieces of MATLAB code are shown in the following sections of Appendix C. C.2 Vehicle Simulation Layout The simulation beings with the function ?main.m.? The vehicle, maneuver, and velocity profile are each chosen by the user in this function. The property variation can also be chosen from the ?main? function. The ?Garage.m? contains vehicle properties for various vehicles. The steering profile is produced in the ?steer_profile.m? function. ?Velocity.m? produces the longitudinal velocity profile for the maneuver. The ?property_changer.m? allows vehicle properties to be varied until two-wheel-lift is detected in the ?two_wheel_lift_detector? script. The simulation switch script contains 141 the numerical integration to solve the equations of motion and yield the vehicle?s dynamic states. ?RubberandString.m? is called from the simulation switch when a value for tire lateral force is required. Various tire models are found in this function such as the non-linear, Pacejka ?Magic Formula,? and the linear tire model. C.3 Main.m % Vehicle Simulation close all clear all clc % Property you wish to vary % 0 - No Changes % 100 - Nominal Vehicle Condition looking for wheel lift % 1 - Weight Split % 2 - CG Height prop=1; % Simulation type % '100' - Steady State Roll % '300' - Transient Roll sim=300; % Vehicle Model car='2001 Blazer Nom'; % Maneuver Profile % step - Step Steer % 1a - Fishhook maneuver='1a'; % Max steer angle [deg] A=5; max_steer=A; % Velocity Profile % step - Step Profile % ramp a - Contant ramp 142 % ramp - Ramp at a given acceleration % coast - Coast Down vel_profile='step'; if prop == 0 | 100 ZZ=0; speed_i=15; % [MPH] Velocity of No Change Option 0 else ZZ=1; speed_i=15; % [MPH] Initial Velocity of Property Changer Option 1:10 end % Wheel Lift Detector % 1 1-Wheel-Lift % 2 2-Wheel-Lift wheel_lift_detector=2; lift_time=1; %ts*sec - value being the duratin of wheel lift ie. value at 1000 = 1 sec wheel lift % Tire Model % Linear % Pacejka % Dugoff tire='Pacejka'; % Simulation Time endtime=8; % 8 seconds is minimum time for steer profile 1a ts=0.001; % Sample Time NN=endtime/ts; % # of calculations % Loading Vehicle Data [MT,M,m,g,Wt,Iz,Ix,zeta,a,b,h_cg_M,h_cg_m,h_rc_f,h_rc_r,trk_f,trk_r,... Sk_f,Sk_r,Sb_f,Sb_r,Sarb_f,Sarb_r,ks_f,ks_r,... karb_f,karb_r,Ca_f,Ca_r,SIS_const,steer_ratio,WB,b_f,... b_r] = garage(car); disp('Vehicle Parameters Loaded...') % Loading Steering Profile del=steer_profile(maneuver,max_steer,SIS_const,endtime,ts,steer_ratio); disp('Steering Profile Loaded...') % Roll Stiffness kphi_f = karb_f+0.5*ks_f*Sk_f^2; % Front Roll Stiffness [N*m*rad] kphi_r = karb_r+0.5*ks_r*Sk_r^2; % Rear Roll Stiffness [N*m*rad] kphi_t = kphi_f+kphi_r; % Total Roll Stiffness % Calculating the roll center at the CG 143 [h_rc_cg]=roll_center_height(WB,a,h_rc_f,h_rc_r); % Property Changer Property_Changer % Initial Conditions initial_variables % Property Variation Loop for qq = 1:QQ flag=0; two_wheel_lift_i(1:NN)=0; two_wheel_lift_o(1:NN)=0; t_w_l_i(1:NN)=0; t_w_l_o(1:NN)=0; w_l_fi(1:NN)=0; w_l_fo(1:NN)=0; w_l_ri(1:NN)=0; w_l_ro(1:NN)=0; zz=0; while zz <= ZZ while flag == 0 % Velocity speed=speed_i+(zz/10); %(zz increments 1 each step, adjust the denominator for mesh density) % Ramp Acceleration Note: Must be an even number accel=2; % Loading Velocity Profile vel=velocity(vel_profile,speed,accel,endtime,ts,MT); % Simulation Script Simulation_Switch % Roll Rate Roll_Rate(qq)=(((-M*g*d1(qq))/(-kphi_t+M*g*d1(qq))))*180/pi; %deg roll/g % Index Counter zz=zz+1; if prop == 0 flag = 1; end % (if statement) end % Flag while if flag == 1 break end % (Flag break) end % (Velocity while) end % (Property while) 144 C.4 Garage.m % Garage Function % This function contains properties for the various cars in our "garage", % which the main m-file calls for. (Basically you get to pick which ride you want for each simulation) % Input a Vehicle name: % RC % Neutral Blazer % 2001 Blazer Nom % 2001 Blazer RRR % 2001 Blazer RMB % Output: % MT % kg % mass of entire vehicle % M % kg % Sprung Mass % m % kg % Unsprung mass % g % m/s^2 % Local Gravity % Iz % N-m-sec^2 % Mass Moment of inertia about z-axis % Ix % N-m-sec^2 % Mass Moment of inertia about x-axis % zeta % Assumed damping coefficient % a % m % x-axis Distance from CG to front tire patch % b % m % x-axis Distance from CG to rear tire patch % h_cg_M % m % CG height in z-axis of Sprung Mass % h_cg_m % m % CG height of Un-Sprung mass % h_rc_f % m % Front Roll Center height in z-axis % h_rc_r % m % Rear Roll Center height in z-axis % trk_f % m % Front vehicle track % trk_r % m % Rear vehicle track % Sk_fM % m % Distance between front springs attachment points % on Sprung Mass % Sk_fm % m % Distance between front springs attachment points % on Un-Sprung Mass % Sk_rM % m % Distance between rear springs % attachment points on Sprung Mass % Sk_rm % m % Distance between rear springs % attachment points on Un-Sprung Mass % Sb_fM; % m % Distance between front dampers attachment % points on Sprung Mass % Sb_fm; % m % Distance between front dampers attachment % points on Un-Sprung Mass % Sb_rM; % m % Distance between rear dampers attachment % points on Sprung Mass % Sb_rm; % m % Distance between rear dampers attachment % points on Un-Sprung Mass % Sarb_f % m % Distance between front Anti-RBs 145 % Sarb_r % m % Distance between rear Anti-RBs % k_f % N/m % Stiffness of front spring (Individual Front Spring) % k_r % N/m % Stiffness of rear spring (Individual Rear Spring) % k_stabf % N*m/rad % Front Stabilizer Stiffness Guess % k_stabr % N*m/rad % Rear Stabilizer Stiffness % Ca_f % Front Axle Corning Stiffness % Ca_r % Rear Axle Corning Stiffness % steer_ratio=18; % handwheel/steer angle % SIS_const % Steer angle (at the handwheel) at which .3g of lateral % acceleration is achieved on the SkidPad at 50 MPH function [MT,M,m,g,Wt,Iz,Ix,zeta,a,b,h_cg_M,h_cg_m,h_rc_f,h_rc_r,trk_f,trk_r,... Sk_f,Sk_r,Sb_f,Sb_r,Sarb_f,Sarb_r,ks_f,ks_r,karb_f,karb_r,... Ca_f,Ca_r,SIS_const,steer_ratio,WB,b_f,b_r] = garage(car) switch car; case 'Neutral Blazer' % NHSTA Nominal Loading Configuration MT=1907.16; % kg % mass of entire vehicle M=MT*.8; % kg % Sprung Mass m=MT*.2; % kg % Unsprung mass g=9.81; % m/s^2 % Local Gravity Wt=MT*g; % Total weight of vehicle (N) Iz=3833.31; %N-m-sec^2 % Mass Moment of inertia about z-axis Ix=734.04; %N-m-sec^2 % Mass Moment of inertia about x-axis WB=2.7180; a=WB*(.5); % m % x-axis Distance from CG to front tire patch b=WB-a; % m % x-axis Distance from CG to rear tire patch h_cg_M=.6629; %.6629 % m % CG height in z-axis of Sprung Mass h_cg_m=.3; %m % CG height of Un-Sprung mass h_rc_f=.1; % m % Front Roll Center height in z-axis h_rc_r=.1; % m % Rear Roll Center height in z-axis trk_f=1.445; % m % Front vehicle track trk_r=trk_f; % m % Rear vehicle track Sk_f=0.7747; % m % Distance between front springs Sk_r=Sk_f; % m % Distance between rear springs Sb_f=Sk_f; % m % Distance between front dampers Sb_r=Sb_f; % m % Distance between rear dampers Sarb_f=1.0287; % m % Distance between front Anti-RBs Sarb_r=Sarb_f; % m % Distance between rear Anti-RBs karb_f=0; % N*m/rad % Front Stablizer Stiffness karb_r=0; % N*m/rad % Rear Stablizer Stiffness ks_f=60000; % N/m % Stiffness of front springs ks_r=40000; % N/m % Stiffness of rear springs b_f=5000; % N/m/s % Front dampers b_r=b_f; % N/m/s % Rear dampers Ca_f=1500; % N/deg_slip % Front tire Cornering Stiffness 146 Ca_r=Ca_f; % N/deg_slip % Rear tire Cornering Stiffness steer_ratio=18; % handwheel/steer angle SIS_const=1.268*steer_ratio; %0.3282; % Phase IV results = 0.3159 % degrees at handwheel % Slowly Increasing Steer simulation where vehicle experiences 0.3g at this steer angle case '2001 Blazer Nom' % NHSTA Nominal Loading Configuration MT=1907.16; % kg % mass of entire vehicle M=MT*.8; % kg % Sprung Mass m=MT*.2; % kg % Unsprung mass g=9.81; % m/s^2 % Local Gravity Wt=MT*g; % Total weight of vehicle (N) Iz=3833.31; % N-m-sec^2 % Mass Moment of inertia about z-axis Ix=734.04; % N-m-sec^2 % Mass Moment of inertia about x-axis a=1.216; % m % x-axis Distance from CG to front tire patch b=1.502; % m % x-axis Distance from CG to rear tire patch WB=a+b; % m % Wheelbase h_cg_M=.6629; % m % CG height in z-axis of Sprung Mass h_cg_m=.4; % m % CG height of Un-Sprung mass h_rc_f=-0.1; % m % Front Roll Center height in z-axis h_rc_r=0.35; % m % Rear Roll Center height in z-axis trk_f=1.445; % m % Front vehicle track trk_r=1.405; % m % Rear vehicle track Sk_f=0.7747; % m % Distance between front springs Sk_r=0.9906; % m % Distance between rear springs Sb_f=Sk_f; % m % Distance between front dampers Sb_r=.762; % m % Distance between rear dampers Sarb_f=1.0287; % m % Distance between front Anti-RBs Sarb_r=.6731; % m % Distance between frear Anti-RBs karb_f=0; % N*m/rad % Front Stablizer Stiffness karb_r=0; % N*m/rad % Rear Stablizer Stiffness ks_f=75000; % N/m % Stiffness of front springs ks_r=75000; % N/m % Stiffness of rear springs b_f=1000; % N/m/s % Front dampers b_r=1000; % N/m/s % Rear dampers Ca_f=1500; % N/deg_slip % Front tire Cornering Stiffness Ca_r=Ca_f; % N/deg_slip % Rear tire Cornering Stiffness steer_ratio=18; % handwheel/steer angle SIS_const=1.268*steer_ratio; %0.3282; % Phase IV results = 0.3159 % degrees at handwheel ... % Slowly Increasing Steer simulation where vehicle experiences 0.3g at this steer angle case '2001 Blazer RRR' % NHTSA Reduced Rollover Resistance Roof Ballast MT=1988.78; % kg % mass of entire vehicle 4WD=(1963.95 kg) 147 M=MT*.8; % kg % Sprung Mass m=MT*.2; % kg % Unsprung mass g=9.81; % m/s^2 % Local Gravity Iz=4094.98*0.95; % N-m-sec^2 % Mass Moment of inertia about z-axis Ix=813.49*.75; % N-m-sec^2 % Mass Moment of inertia about x-axis zeta=.7; % Assumed damping coefficient a=1.216; % m % x-axis Distance from CG to front tire patch b=1.502; % m % x-axis Distance from CG to rear tire patch WB=a+b; % m % Wheelbase h_cg_M=0.6960; % m % CG height in z-axis of Sprung Mass h_cg_m=.4; % m % CG height of Un-Sprung mass h_rc_f=-.1; % m % Front Roll Center height in z-axis h_rc_r=0.35; % m % Rear Roll Center height in z-axis trk_f=1.445; % m % Front vehicle track trk_r=1.405; % m % Rear vehicle track Sk_f=0.7747; % m % Distance between front springs Sk_r=0.9906; % m % Distance between rear springs Sb_f=Sk_f; % m % Distance between front dampers Sb_r=.762; % m % Distance between rear dampers Sarb_f=1.0287; % m % Distance between front Anti-RBs Sarb_r=.6731; % m % Distance between frear Anti-RBs Farb_f=700; % N % Force/2 exerted by ARB per degree. Creates a moment around roll center for Sprung and Un-Sprung mass Farb_r=400; % N % it is directly related to Roll stiffness created by the ARBs k_stabf=0; % N*m/rad % Front Stablizer Stiffness * Not used in Transient Simulation k_stabr=0; % N*m/rad % Rear Stablizer Stiffness * Not used in Transient Simulation k_f=75000; % N/m % Stiffness of front springs Estimate of hand calculation k_r=70000; % N/m % Stiffness of rear springs b_f=5000; % N/m/s % Front dampers b_r=4000; % N/m/s % Rear dampers Ca_f=2500; % N/deg % Front Axle Corning Stiffness Ca_r=Ca_f; % Rear Axle Corning Stiffness steer_ratio=18; % handwheel/steer angle SIS_const=1.268*steer_ratio; %0.3282; % Phase IV results = 0.3159 % degrees at handwheel ... % Slowly Increasing Steer simulation where vehicle experiences 0.3g at this steer angle case '2001 Blazer RMB' % NHTSA Rear-Mounted Basslast Configuration MT=2237.71; % kg % mass of entire vehicle 4WD=(1963.95 kg) M=MT*.9; % kg % Sprung Mass m=MT*.1; % kg % Unsprung mass g=9.81; % m/s^2 % Local Gravity 148 Iz=4886.38*0.95; % N-m-sec^2 % Mass Moment of inertia about z-axis Ix=769.97*0.75; % Mass Moment of inertia about x-axis a=1.522; % m % x-axis Distance from CG to front tire patch b=1.196; % m % x-axis Distance from CG to rear tire patch h_cg_M=0.6614; % m % CG height in z-axis of Sprung Mass h_cg_m=.4; % m % CG height of Un-Sprung mass h_rc_f=-.15; % m % Front Roll Center height in z-axis h_rc_r=0.3; % m % Rear Roll Center height in z-axis trk_f=1.445; % m % Front vehicle track trk_r=1.405; % m % Rear vehicle track Sk_f=0.7747; % m % Distance between front springs Sk_r=0.9906; % m % Distance between rear springs Sb_f=Sk_f; % m % Distance between front dampers Sb_r=.762; % m % Distance between rear dampers Sarb_f=1.0287; % m % Distance between front Anti-RBs Sarb_r=.6731; % m % Distance between frear Anti-RBs Farb_f=700; % N % Force/2 exerted by ARB per degree. Creates a moment around roll center for Sprung and Un-Sprung mass Farb_r=400; % N % it is directly related to Roll stiffness created by the ARBs k_stabf=0; % N*m/rad % Front Stablizer Stiffness * Not used in Transient Simulation k_stabr=0; % N*m/rad % Rear Stablizer Stiffness * Not used in Transient Simulation k_f=60000; % N/m % Stiffness of front springs Estimate of hand calculation k_r=60000; % N/m % Stiffness of rear springs b_f=5000; % N/m/s % Front dampers b_r=4000; % N/m/s % Rear dampers Ca_f=2500; % N/deg % Front Axle Corning Stiffness Ca_r=Ca_f; % Rear Axle Corning Stiffness steer_ratio=18; % handwheel/steer angle SIS_const=1.268*steer_ratio; %0.3282; % Phase IV results = 0.3159 % degrees at handwheel ... % Slowly Increasing Steer simulation where vehicle experiences 0.3g at this steer angle case 'RC' %Weight Distribution weight_front=0.809; %kg Mass on front of vehicle weight_rear=0.941; %kg Mass on rear of vehicle weight_left=0.899; %kg Mass on left of vehicle weight_right=0.854; %kg Mass on right of vehicle weight_split=[0.4623 0.5377]; %Front-Rear Weight Split %Total Mass=1.65 wheelbase=0.2556; %m Wheelbase 149 %Unsprung Mass Components wheel_mass=0.034; %kg Mass of wheel x4 = .136 upper_arm_mass=0.001; %kg Mass of upper arm x4=.004 lower_arm_mass=0.005; %kg Mass of lower A arm x4=.02 axle_mass=0.005; %kg Mass of Axle x4=.02 spindle_mass=0.024; %kg Mass of spindle x4=.096 MT=3.20; % kg % mass of entire vehicle m=.276; % kg % Un-sprung mass M=MT-m; % kg % Sprung Mass g=9.81; % m/s^2 % Local Gravity Wt=MT*g; % Total weight of vehicle (N) Iz=.20; % N-m-sec^2 % Mass Moment of inertia about z-axis (yaw) Ix=.10; % N-m-sec^2 % Mass Moment of inertia about x-axis (roll) Iy=.2; % N-m-sec^2 % Inertia about y-axis (pitch) a=0.1374; % m % x-axis Distance from CG to front tire patch b=0.1182; % m % x-axis Distance from CG to rear tire patch WB=a+b; h_cg_M=.1565; % m % CG height (in z-axis) of Sprung Mass h_cg_m=.02; % m % CG height of Un-Sprung mass h_rc_f=0.010; % m % Front Roll Center height in z-axis h_rc_r=0.0127; % m % Rear Roll Center height in z-axis trk_f=0.1683; % m % Front vehicle track trk_r=0.1746; % m % Rear vehicle track Sk_fM=0.0492; % m % Distance between front spring attachment points on sprung mass Sk_fm=0.1016; % m % Distance between front spring attachment points on un- sprung mass Sk_f=(Sk_fM+Sk_fm)/2; Sk_rM=0.0762; % m % Distance between rear springs attachment points on Sprung Mass Sk_rm=0.1206; % m % Distance between rear springs attachment points on Un- Sprung Mass Sk_r=(Sk_rM+Sk_rm)/2; Sb_fM=0.0492; % m % Distance between front dampers attachment points on Sprung Mass Sb_fm=0.1016; % m % Distance between front dampers attachment points on Un-Sprung Mass Sb_f=(Sb_fM+Sb_fm)/2; Sb_rM=0.0762; % m % Distance between rear dampers attachment points on Sprung Mass Sb_rm=0.1206; % m % Distance between rear dampers attachment points on Un-Sprung Mass Sb_r=(Sb_rM+Sb_rm)/2; karb_f=0; 150 karb_r=0; Sarb_f=0; % m % Distance between front Anti-Roll Bars Sarb_r=0; % m % Distance between rear Anti-RBs ks_f=150*100; % N/m % Stiffness of front spring (Individual Front Spring) ks_r=150*100; % N/m % Stiffness of rear spring (Individual Rear Spring) b_f=100; b_r=100; k_stabf=0; % N*m/rad % Front Stabilizer Stiffness k_stabr=0; % N*m/rad % Rear Stabilizer Stiffness Ca_f=100; % N/rad % Front Axle Corning Stiffness Ca_r=100; % N/rad % Rear Axle Corning Stiffness steer_ratio=1; SIS_const=2.787*steer_ratio; otherwise disp('We do not have the car you request in the garage. See "help garage" and pick a car in the garage, or start walking.') end C.5 Initial Conditions % Initializing Variables % Initial Conditions a_f(1:NN)=0; a_r(1:NN)=0; r(1:NN)=0; r_dt(1:NN)=0; Vy(1:NN)=0; Vy_dt(1:NN)=0; B(1:NN)=0; B_dt(1:NN)=0; t(1:NN)=0; ay(1:NN)=0; phiM(1:NN)=0; phiM_ddt(1:NN)=0; phiM_dt(1:NN)=0; Fy_fl(1:NN)=0; Fy_fr(1:NN)=0; Fy_rl(1:NN)=0; Fy_rr(1:NN)=0; Fy_total(1:NN)=0; ay(1:NN)=0; 151 Fz_fi(1:NN)=0; Fz_fo(1:NN)=0; Fz_ri(1:NN)=0; Fz_ro(1:NN)=0; d1(1:NN)=0; d1_dt(1:NN)=0; d1_ddt(1:NN)=0; y(1:NN)=0; y_dt(1:NN)=0; y_ddt(1:NN)=0; z_b_r(1:NN)=0; z_b_r_dt(1:NN)=0; z_b_r_ddt(1:NN)=0; z_b_f(1:NN)=0; z_b_f_dt(1:NN)=0; z_b_f_ddt(1:NN)=0; z_k_r(1:NN)=0; z_k_r_dt(1:NN)=0; z_k_r_ddt(1:NN)=0; z_k_f(1:NN)=0; z_k_f_dt(1:NN)=0; z_k_f_ddt(1:NN)=0; dFz_f(1:NN)=0; dFz_r(1:NN)=0; Fy_f(1:NN)=0; Fy_fi(1:NN)=0; Fy_fo(1:NN)=0; Fy_r(1:NN)=0; Fy_ri(1:NN)=0; Fy_ro(1:NN)=0; Fy_t(1:NN)=0; Vx(1:NN)=0; a_cen(1:NN)=0; two_wheel_lift_o(1:NN)=0.125; two_wheel_lift_i(1:NN)=0.150; Marb_f(1:NN)=0; Marb_r(1:NN)=0; Fb_f(1:NN)=0; Fb_r(1:NN)=0; Fk_f(1:NN)=0; Fk_r(1:NN)=0; Ry(1:NN)=0; Ry_f(1:NN)=0; Ry_r(1:NN)=0; Rz(1:NN)=0; RS(1:NN)=0; 152 RD(1:NN)=0; fo_lift(1:NN)=0; fi_lift(1:NN)=0; ro_lift(1:NN)=0; ri_lift(1:NN)=0; two_wheel_lift_i(1:NN)=0; two_wheel_lift_o(1:NN)=0; t_w_l_i(1:NN)=0; t_w_l_o(1:NN)=0; w_l_fi(1:NN)=0; w_l_fo(1:NN)=0; w_l_ri(1:NN)=0; w_l_ro(1:NN)=0; roll_ay(QQ)=0; roll_r(QQ)=0; roll_V(QQ)=0; roll_B_dt(QQ)=0; roll_B(QQ)=0; roll_phiM(QQ)=0; roll_phiM_dt(QQ)=0; roll_a_f(QQ)=0; roll_a_r(QQ)=0; roll_Fy_f(QQ)=0; roll_Fy_r(QQ)=0; roll_Fy_t(QQ)=0; roll_dFz_f(QQ)=0; roll_dFz_r(QQ)=0; roll_del(QQ)=0; roll_t(QQ)=0; Fz_fo_lift = 0; Fz_ro_lift = 0; Fz_fi_lift = 0; Fz_ri_lift = 0; counter=0; C.6 roll_center_height.m function[h_rc_cg]=roll_center_height(WB,a,h_rc_f,h_rc_r) if h_rc_f < h_rc_r; theta=atan((h_rc_r-h_rc_f)/WB); x=a*tan(theta); h_rc_cg=x+h_rc_f; else if h_rc_f > h_rc_r; 153 theta=atan((h_rc_f-h_rc_r)/WB); x=a*tan(theta); h_rc_cg=x+h_rc_r; else if h_rc_f == h_rc_r; h_rc_cg=h_rc_f; end end end C.7 Velocity.m function [vel]=velocity(profile,speed,accel,endtime,ts,MT); NN = endtime/ts; milephour2meterpsec = 1606.344/3600; % Conversion from mph to meter/s vel(1:NN) = 0*milephour2meterpsec; % initializing the velocity vector switch profile case 'step' vel(1:NN)=speed*milephour2meterpsec; case 'ramp a' % constantly increasing ramp for k=1:NN-1 vel(k+1)=(speed/NN)+vel(k); end vel(1:NN)=vel(1:NN)*milephour2meterpsec; case 'ramp' % starting the vehicle from standstill to designated velocity % Both Maximum speed and acceleration are user defined period_1 = 5; % [sec] index_1 = period_1/ts; period_2 = (1/accel)*speed+period_1; index_2 = period_2/ts; for k=index_1:index_2 vel(k) = vel(k-1)+ts*accel; end vel(index_2+1:NN)=vel(index_2); vel(1:NN)=vel(1:NN)*milephour2meterpsec; case 'coast' % Coast Down drag=0.85; F_fric=600; t_coast(NN)=0; vel_dt(NN)=0; start_time=2/ts; vel(1:start_time)=speed; for cc=start_time:NN 154 vel_dt(cc+1)= -drag/MT*vel(cc)^2-F_fric/MT; vel(cc+1)=vel(cc)+ts*0.5*(vel_dt(cc+1)+vel_dt(cc)); t_coast(cc+1)=t_coast(cc)+ts; end vel=Velocity_Smoothing(vel); vel=vel*milephour2meterpsec; end C.8 Velocity_Smoothing.m function[vel_out]=Velocity_Smoothing(vel_in) % Velocity Profile Smoothing % This file takes out the sharp changes in velocity wn=100; [B,A]=butter(2,wn/1000*2*pi); vel_out=filter(B,A,vel_in); C.9 Steer_Profile.m function [del_out]=steer_profile(maneuver,max_steer,SIS_const,endtime,ts,steer_ratio); NN=endtime/ts; del(1:NN)=0; % initializing del switch maneuver % Slowly Increasing Steer (SIS) Maneuver case 'SIS' ramptime=30; % sec for n=1:NN del(n+1)=(max_steer*pi/180)/(ramptime/ts)+del(n); end % Step Maneuvers case 'step' del(1:NN)=max_steer*pi/180; % Radians % Ramp Maneuvers case 'ramp' ramptime=5; % the time it takes to ramp linearly from zero to the set steer angle del(1)=0; t(1)=0; for n=1:NN; if n <= ramptime/ts del(n+1)=(max_steer*pi/180)/(ramptime/ts)+del(n); 155 end if n > 5/ts del(n+1)=del(n); end end % Sinusoidal Maneuvers case 'sine' del(1)=0; t(1)=0; for n=1:NN wn=n*180/pi/20000; del(n+1)=(max_steer*pi/180)*sin(wn); end % Fishhook 1a case '1a' period_1=2; index_1=period_1/ts; A=chop(max_steer*pi/180,4); steer_rate=chop(40*pi/180,2); period_2=chop(A/steer_rate+period_1,4); index_2=period_2/ts; for k = index_1:index_2 del(k+1)=steer_rate*ts+del(k); end period_3=.25+period_2; index_3=period_3/ts; for kk=index_2:index_3 del(kk+1)=del(kk); end period_4=chop(2*A/steer_rate+period_3,4); index_4=period_4/ts; for k=index_3:index_4 del(k+1)=del(k)-steer_rate*ts; end period_5=3+period_4; index_5=period_5/ts; for k=index_4:index_5 del(k+1)=del(k); end period_6=chop(A/steer_rate+period_5,4); index_6=period_6/ts; for k=index_5:index_6 del(k+1)=steer_rate*ts+del(k); end del(index_6+1:NN)=0; end 156 [del_out] = Steering_Smoothing(del); C.10 Steering_Smoothing.m function[del_out]=Steer_Smoothing(del_in) % Steering Profile Smoothing % This file takes out the sharp changes in steering angle wn=.5; [B,A]=butter(2,wn/1000*2*pi); del_out=filter(B,A,del_in); C.11 RubberandString.m function [Fy]=rubberandstring(Ca,tire,alpha,Fz) alpha=-alpha*180/pi; % There is a sign change here do to my sign convention on my FBD Fz=Fz/1000; % Conversion from N to kN switch tire case 'Pacejka' % The stadard Pacejka "Magic Tire Model" found is SAE 870421 % Pacejka Parameters a0=0; % 0 a1=-10.1; % -22.1 a2=1011; % 1011 a3=1578; % 1078 a4=1.82; % 1.82 a5=0.208; % 0.208 a6=0; % 0 a7=-0.352; % -0.354 a8=.707; % 0.707 % Pacejka Equations if Fz <= 0; Fy = 0; else C = 1.30; % Shape Factor % 1.30 D = a1*Fz^2+a2*Fz; % Peak Factor E = a6*Fz^2+a7*Fz+a8; % -1.5 % Curvature Factor BCD= a3*sin(a4*atan(a5*Fz)); % Ca % Cornering Stiffness Parameter (Calpha) 157 B=BCD/(C*D); % Stiffness Factor phi = (1-E)*alpha+(E/B)*atan(B*alpha); Fy = D*sin(C*atan(B*phi)); mu=Fy/Fz/1000; end case 'Linear' % This is a simple linear tire model. Not realistic, but it is % used to debug code, and to see what is happening in the simulation Ca=2500.0; % N/deg_slip Cornering Stiffness per tire Fy=Ca*alpha; case 'Dugoff No Fz' % This is a Dugoff type model, however Fz is set at 5 kN Ca=1002.5; % N/deg_slip Cornering Stiffness per tire Fy_Max=5000; % Newtons saturation_alpha=Fy_Max/Ca; % degress Saturation tire slip angle if alpha <= saturation_alpha Fy=Ca*alpha; end if alpha >= -saturation_alpha Fy=Ca*alpha; end if alpha > saturation_alpha Fy=Fy_Max; end if alpha < -saturation_alpha Fy=-Fy_Max; end case 'Dugoff' % This is a variation of Dugoff's tire model. Fy_Max is a funtion % of Fz and the coefficent of friction. % SAE 700377 mu=1.02; Fy_Max=Fz*mu*1000; Ca=1002.5; saturation_alpha=Fy_Max/Ca; if alpha <= saturation_alpha Fy=Ca*alpha; end if alpha >= -saturation_alpha Fy=Ca*alpha; end if alpha > saturation_alpha Fy=Fy_Max; 158 end if alpha < -saturation_alpha Fy=-Fy_Max; end end C.12 Property Changer % Property Changer Switch switch prop case 0 % No Changes [a,b,Wf,Wr,M_f,M_r,m_f,m_r,h_cg_M,Iz,Ix,d1,h1,Ca_f,Ca_r] =No_Change(a,b,g,m,WB,MT,Wt,h_cg_M,Iz,Ix,M,h_rc_cg,Ca_f,Ca_r); QQ=length(a); case 100 % No Changes but looking for wheel lift [a,b,Wf,Wr,M_f,M_r,m_f,m_r,h_cg_M,Iz,Ix,d1,h1,Ca_f,Ca_r] =No_Change(a,b,g,m,WB,MT,Wt,h_cg_M,Iz,Ix,M,h_rc_cg,Ca_f,Ca_r); QQ=length(a); case 1 % Weight Split Change [a,b,Wf,Wr,M_f,M_r,m_f,m_r,h_cg_M,Iz,Ix,d1,h1,Ca_f,Ca_r] =Weight_Split_Changer(a,b,g,m,WB,MT,Wt,h_cg_M,Iz,Ix,M,h_rc_cg,Ca_f,Ca_r); QQ=length(a); case 2 % CG Height Change [h_cg_M,a,b,Wf,Wr,M_f,M_r,m_f,m_r,Ix,Iz,h_cg_Percent,d1,h1,Ca_f,Ca_r] =CG_Height_Changer(h_cg_M,a,b,g,m,WB,MT,Wt,M,Ix,Iz,h_rc_cg,Ca_f,Ca_r); QQ=length(h_cg_M); end C.13 No Change function[a,b,Wf,Wr,M_f,M_r,m_f,m_r,h_cg_M,Iz,Ix,d1,h1,Ca_f,Ca_r]=No_Change(a_ol d,b_old,g,m,WB,MT,Wt,h_cg,Iz_old,Ix_old,M,h_rc_cg,Ca_f_old,Ca_r_old); % This function causes no change to the initial vehicle properties qq=1; h_cg_M(qq)=h_cg; a(qq)=a_old; b(qq)=b_old; 159 h1(qq)=h_cg_M(qq)-h_rc_cg; % Negative number means that the CG is below the roll center d1(qq)=h1(qq); Ix(qq)=Ix_old; Iz(qq)=Iz_old; d1(qq)=h1(qq); Ca_f(qq)=Ca_f_old; Ca_r(qq)=Ca_r_old; % Weight ratio Wr(qq)=a(qq)*MT*g/(a(qq)+b(qq)); % Total weight on rear tires (N) Wf(qq)=MT*g-Wr(qq); % Total weight on front tires (N) M_f(qq)=Wf(qq)/g; % Sprung Mass on front axle (kg) M_r(qq)=Wr(qq)/g; % Sprung Mass on rear axle (kg) % I assume that the un-sprung mass has same weight split as the sprung mass m_f(qq)=m*Wf(qq)/Wt; % Un-Sprung Mass on front axle (kg) m_r(qq)=m*Wr(qq)/Wt; % Un-Sprung Mass on rear axle (kg) C.14 Weight_Split_Changer.m function [a,b,Wf,Wr,M_f,M_r,m_f,m_r,h_cg_M,Iz,Ix,d1,h1,Ca_f,Ca_r]=Weight_Split_Changer(a _old,b,g,m,WB,MT,Wt,h_cg,Iz_old,Ix_old,M,h_rc_cg,Ca_f_old,Ca_r_old); % This function changes vehicle weight split % It makes vectors of the variable it is outputting % Weight Split Changer QQ=80; qq=1; for qq = 1:QQ a(qq)=.28*WB+WB*(qq/160); b(qq)=WB-a(qq); h_cg_M(qq)=h_cg; Iz(qq)=Iz_old; Ix(qq)=Ix_old; h1(qq)=h_cg_M(qq)-h_rc_cg; d1(qq)=h1(qq); Ca_f(qq)=Ca_f_old; Ca_r(qq)=Ca_r_old; % Weight ratio Wr(qq)=a(qq)*MT*g/(a(qq)+b(qq)); % Total weight on rear tires (N) Wf(qq)=MT*g-Wr(qq); % Total weight on front tires (N) M_f(qq)=Wf(qq)/g; % Sprung Mass on front axle (kg) M_r(qq)=Wr(qq)/g; % Sprung Mass on rear axle (kg) % I assume that the un-sprung mass has same weight split as the sprung mass 160 m_f(qq)=m*Wf(qq)/Wt; % Un-Sprung Mass on front axle (kg) m_r(qq)=m*Wr(qq)/Wt; % Un-Sprung Mass on rear axle (kg) end C.15 CG_Height_Changer.m function [h_cg_M,a,b,Wf,Wr,M_f,M_r,m_f,m_r,Ix,Iz,h_cg_Percent,d1,h1,Ca_f,Ca_r]=CG_Heigh t_Changer(h_cg_M_old,a_old,b_old,g,m,WB,MT,Wt,M,Ix_old,Iz_old,h_rc_cg,Ca_f_old, Ca_r_old) % This function changes the CG height QQ=40; for qq=1:QQ h_cg_M(qq)=((.60)+(qq/2)/(QQ/2))*h_cg_M_old; h_cg_Percent(qq)=h_cg_M(qq)/h_cg_M_old*100; a(qq)=a_old; b(qq)=b_old; h1(qq)=h_cg_M(qq)-h_rc_cg; d1(qq)=h1(qq); Ix(qq)=Ix_old; Iz(qq)=Iz_old; Ca_f(qq)=Ca_f_old; Ca_r(qq)=Ca_r_old; % Weight ratio Wr(qq)=a(qq)*MT*g/(a(qq)+b(qq)); % Total weight on rear tires (N) Wf(qq)=MT*g-Wr(qq); % Total weight on front tires (N) M_f(qq)=Wf(qq)/g; % Sprung Mass on front axle (kg) M_r(qq)=Wr(qq)/g; % Sprung Mass on rear axle (kg) % I assume that the un-sprung mass has same weight split as the sprung mass m_f(qq)=m*Wf(qq)/Wt; % Un-Sprung Mass on front axle (kg) m_r(qq)=m*Wr(qq)/Wt; % Un-Sprung Mass on rear axle (kg) end C.16 Simulation Switch % Simulation Switch switch sim case 100 n=1; while n <= NN d1=h1; 161 % Velocity in x-axis Vx(n)=vel(n)*cos(B(n)); % m/s % Tire Slip Angles a_f(n)=atan((Vy(n)+a(qq)*r(n))/Vx(n))-del(n); % tire slip angle front (rad) a_r(n)=atan((Vy(n)-b(qq)*r(n))/Vx(n)); % tire slip angle rear (rad) % Spring and Damper Displacement and Velocity z_k_f(n) = (Sk_f/2)*sin(phiM(n)); % Front Spring Displacement [m] z_k_r(n) = (Sk_r/2)*sin(phiM(n)); % Rear Spring Displacement [m] z_b_f_dt(n) = phiM_dt(n)*(Sb_f/2)*cos(phiM(n)); % Front Damper Displacement [m/s] z_b_r_dt(n) = phiM_dt(n)*(Sb_r/2)*cos(phiM(n)); % Rear Damper Displacement [m/s] % Spring Forces (Individual springs) Fk_f(n) = ks_f*z_k_f(n); % Front Individual Spring Force [N] Fk_r(n) = ks_r*z_k_r(n); % Rear Individual Spring Force [N] % Damper Forces (Individual dampers) Fb_f(n) = 0; % Front Indiviual Damper Force [N] Zero for SS Fb_r(n) = 0; % Rear Individual Damper Force [N] Zero for SS % Anti-Roll Bar Moment Marb_f(n) = phiM(n)*karb_f; % [N*m] Marb_r(n) = phiM(n)*karb_r; % [N*m] % Reaction Forces Ry(n) = ay(n)*M; %(ay(n)+y_ddt(n))*M; % [N] Ry_f(n) = Ry(n)*M_f(qq)/M; % Ry allocated to the Front [N] Ry_r(n) = Ry(n)*M_r(qq)/M; % Ry allocated to the Rear [N] Rz(n) = M*g; % [N] % Weight Transfer % Fzo-Fzi dFz_f(n)=(2/trk_f)*(Marb_f(n)+Sk_f*Fk_f(n)+Sb_f*Fb_f(n)+... Ry_f(n)*(h_rc_f-h_cg_m)+Fy_f(n)*h_cg_m); % Front Lateral Weight Transfer dFz_r(n)=(2/trk_r)*(Marb_r(n)+Sk_r*Fk_r(n)+Sb_r*Fb_r(n)+... Ry_r(n)*(h_rc_r-h_cg_m)+Fy_r(n)*h_cg_m); % Rear Lateral Weight Transfer % This is a cap for the dFz so you don't have more mass transfer than is physically possible if abs(dFz_f(n)) >= Wf(qq)/2 dFz_f(n)=sign(dFz_f(n))*Wf(qq)/2; end if abs(dFz_r(n)) >= Wr(qq)/2 dFz_r(n)=sign(dFz_r(n))*Wr(qq)/2; end % Calculating the new Fz value at each wheel Fz_ri(n)=Wr(qq)/2-dFz_r(n); % Force in the z-axis on rear inner tire [N] Fz_ro(n)=Wr(qq)/2+dFz_r(n); % Force in the z-axis on rear outer tire [N] Fz_fi(n)=Wf(qq)/2-dFz_f(n); % Force in the z-axis on front inner tire [N] 162 Fz_fo(n)=Wf(qq)/2+dFz_f(n); % Force in the z-axis on front outer tire [N] switch wheel_lift_detector case 1 % Individual Wheel-Lift detector One_Wheel_Lift_Detector case 2 % Two Wheel Lift Detector Two_Wheel_Lift_Detector end % Lateral Forces from Non-Linear Pacejka Tire Model % Send rubberandstring.m slip angles in radians and Fz in N Fy_fo(n+1)=rubberandstring(Ca_f(qq),tire,a_f(n),Fz_fo(n)); % Lateral Force Front inner (N) Fy_fi(n+1)=rubberandstring(Ca_f(qq),tire,a_f(n),Fz_fi(n)); % Lateral Force Front outer [N] Fy_ro(n+1)=rubberandstring(Ca_r(qq),tire,a_r(n),Fz_ro(n)); % Lateral Force Rear inner (N) Fy_ri(n+1)=rubberandstring(Ca_r(qq),tire,a_r(n),Fz_ri(n)); % Lateral Force Rear outer (N) Fy_f(n)=Fy_fo(n)+Fy_fi(n); Fy_r(n)=Fy_ro(n)+Fy_ri(n); Fy_t(n)=Fy_f(n)+Fy_r(n); % Yaw Dynamics r_dt(n+1) = (-b(qq)*Fy_r(n)+a(qq)*Fy_f(n)*cos(del(n)))/Iz(qq); % Yaw Accel. [rad/sec^2] r(n+1)=r(n)+0.5*ts*(r_dt(n+1)+r_dt(n)); % Yaw Rate [rad/sec] Vy_dt(n+1)= (Fy_r(n)+Fy_f(n)*cos(del(n)))/MT-vel(n)*r(n)*cos(B(n)); Vy(n+1)=Vy(n)+ts*0.5*(Vy_dt(n+1)+Vy_dt(n)); B_dt(n)=asin((Vy_dt(n)/vel(n))); if Vx(n) == 0 B(n+1)=0; else B(n+1)=atan((Vy(n+1)/Vx(n))); % Vehicle Sideslip [rad/sec] end a_cen(n)=r(n)*vel(n); % Centripital Acceleration [m/sec^2] ay(n+1)=Vy_dt(n+1)+a_cen(n)*cos(B(n+1)); % Lateral Acceleration [m/sec^2] % Steady State Roll phiM(n+1)=((M*g*d1(qq))/(kphi_t-M*g*d1(qq)))*(ay(n+1)/g); % Roll Angle [rad] % Time Counter t(n+1)=t(n)+ts; % Time [sec] % Index Counter n=n+1; end case 300 163 n=1; d1=h1; % Velocity in x-axis Vx(n)=vel(n)*cos(B(n)); % m/s % Tire Slip Angles a_f(n)=atan((Vy(n)+a(qq)*r(n))/Vx(n))-del(n); % tire slip angle front (rad) a_r(n)=atan((Vy(n)-b(qq)*r(n))/Vx(n)); % tire slip angle rear (rad) % Spring and Damper Displacement and Velocity z_k_f(n) = (Sk_f/2)*sin(phiM(n)); % Front Spring Displacement [m] z_k_r(n) = (Sk_r/2)*sin(phiM(n)); % Rear Spring Displacement [m] z_b_f_dt(n) = phiM_dt(n)*(Sb_f/2)*cos(phiM(n)); % Front Damper Displacement [m/s] z_b_r_dt(n) = phiM_dt(n)*(Sb_r/2)*cos(phiM(n)); % Rear Damper Displacement [m/s] % Spring Forces (Individual springs) Fk_f(n) = ks_f*z_k_f(n); % Front Individual Spring Force [N] Fk_r(n) = ks_r*z_k_r(n); % Rear Individual Spring Force [N] % Damper Forces (Individual dampers) Fb_f(n) = b_f*z_b_f_dt(n); % Front Indiviual Damper Force [N] Fb_r(n) = b_r*z_b_r_dt(n); % Rear Individual Damper Force [N] % Anti-Roll Bar Moment Marb_f(n) = phiM(n)*karb_f; % [N*m] Marb_r(n) = phiM(n)*karb_r; % [N*m] % Reaction Forces Ry(n) = ay(n)*M; %(ay(n)+y_ddt(n))*M; % [N] Ry_f(n) = Ry(n)*M_f(qq)/M; % Ry allocated to the Front [N] Ry_r(n) = Ry(n)*M_r(qq)/M; % Ry allocated to the Rear [N] Rz(n) = M*g; % [N] % Weight Transfer % Fzo-Fzi dFz_f(n)=(2/trk_f)*(Marb_f(n)+Sk_f*Fk_f(n)+Sb_f*Fb_f(n)+... Ry_f(n)*(h_rc_f-h_cg_m)+Fy_f(n)*h_cg_m); % Front Lateral Weight Transfer dFz_r(n)=(2/trk_r)*(Marb_r(n)+Sk_r*Fk_r(n)+Sb_r*Fb_r(n)+... Ry_r(n)*(h_rc_r-h_cg_m)+Fy_r(n)*h_cg_m); % Rear Lateral Weight Transfer % This is a cap for the dFz so you don't have more mass transfer than is physically possible if abs(dFz_f(n)) >= Wf(qq)/2 dFz_f(n)=sign(dFz_f(n))*Wf(qq)/2; end if abs(dFz_r(n)) >= Wr(qq)/2 dFz_r(n)=sign(dFz_r(n))*Wr(qq)/2; end % Calculating the new Fz value at each wheel 164 Fz_ri(n)=Wr(qq)/2-dFz_r(n); % Force in the z-axis on rear inner tire [N] Fz_ro(n)=Wr(qq)/2+dFz_r(n); % Force in the z-axis on rear outer tire [N] Fz_fi(n)=Wf(qq)/2-dFz_f(n); % Force in the z-axis on front inner tire [N] Fz_fo(n)=Wf(qq)/2+dFz_f(n); % Force in the z-axis on front outer tire [N] switch wheel_lift_detector case 1 % Individual Wheel-Lift detector One_Wheel_Lift_Detector case 2 % Two Wheel Lift Detector Two_Wheel_Lift_Detector end % Lateral Forces from Non-Linear Pacejka Tire Model % Send rubberandstring.m slip angles in radians and Fz in N Fy_fo(n+1)=rubberandstring(Ca_f(qq),tire,a_f(n),Fz_fo(n)); % Lateral Force Front inner (N) Fy_fi(n+1)=rubberandstring(Ca_f(qq),tire,a_f(n),Fz_fi(n)); % Lateral Force Front outer [N] Fy_ro(n+1)=rubberandstring(Ca_r(qq),tire,a_r(n),Fz_ro(n)); % Lateral Force Rear inner (N) Fy_ri(n+1)=rubberandstring(Ca_r(qq),tire,a_r(n),Fz_ri(n)); % Lateral Force Rear outer (N) Fy_f(n)=Fy_fo(n)+Fy_fi(n); Fy_r(n)=Fy_ro(n)+Fy_ri(n); Fy_t(n)=Fy_f(n)+Fy_r(n); % Yaw Dynamics r_dt(n+1) = (-b(qq)*Fy_r(n)+a(qq)*Fy_f(n)*cos(del(n)))/Iz(qq); % Yaw Accel. [rad/sec^2] r(n+1)=r(n)+0.5*ts*(r_dt(n+1)+r_dt(n)); % Yaw Rate [rad/sec] Vy_dt(n+1)= (Fy_r(n)+Fy_f(n)*cos(del(n)))/MT-vel(n)*r(n)*cos(B(n)); Vy(n+1)=Vy(n)+ts*0.5*(Vy_dt(n+1)+Vy_dt(n)); % Side Slip Angle (Beta) B(n+1)=asin((Vy(n)/vel(n))); % Vehicle Sideslip [rad] B_dt(n+1)=(B(n+1)-B(n))/ts; % Vehicle Sideslip rate [rad/sec] % Lateral Acceleration a_cen(n)=r(n)*vel(n); % Centripital Acceleration [m/sec^2] ay(n+1)=Vy_dt(n+1)+a_cen(n)*cos(B(n+1)); % Lateral Acceleration [m/sec^2] ay_test(n)=vel(n)^2/R; % Roll Stiffness RS=kphi_t; RSM(n)=RS*phiM(n); % Roll Damping RD(n)=(0.5*b_f*Sb_f^2*cos(phiM(n))+0.5*b_r*Sb_r^2*cos(phiM(n))); RDM(n)=RD(n)*phiM_dt(n); % Roll Dynamics 165 phiM_ddt(n+1)=(1/Ix(qq))*(-RSM(n)- RDM(n)+Rz(n)*d1(qq)*sin(phiM(n))+Ry(n)*d1(qq)*cos(phiM(n))); phiM_dt(n+1)=phiM_dt(n)+ts*0.5*(phiM_ddt(n+1)+phiM_ddt(n)); % Roll Velocity [rad/s] phiM(n+1)=phiM(n)+ts*0.5*(phiM_dt(n+1)+phiM_dt(n)); % Roll Angle [rad] % Time Counter t(n+1)=t(n)+ts; % Time [sec] % Index Counter n=n+1; end end C.17 Two-Wheel-Lift Detector % Script of Two-Wheel-Lift Detector % Detects wheel lift of both 'inside' or 'outside' tires % Depending on the maneuver the 'outside' tires may have wheel lift. This % does not actually mean that the outside tires lift. The terms outside % and inside were used instead of left and right for ease in debugging step % inputs. % Two wheel-lift detector of outside tires if Fz_fo(n) <= 0 & Fz_ro(n) <= 0 counter=1+counter; fo_lift(n) = 1; ro_lift(n) = 1; roll_fo_lift(qq) = 1; roll_ro_lift(qq) = 1; two_wheel_lift_o(n)=1; roll_ay(qq)=max(ay(1:n)); roll_r(qq)=(r(n)); roll_V(qq)=(vel(n)); roll_B_dt(qq)=(B_dt(n)); roll_B(qq)=(B(n)); roll_phiM(qq)=(phiM(n)); roll_phiM_dt(qq)=(phiM_dt(n)); roll_a_f(qq)=(a_f(n)); roll_a_r(qq)=(a_r(n)); roll_Fy_f(qq)=(Fy_f(n)); roll_Fy_r(qq)=(Fy_r(n)); roll_Fy_t(qq)=(Fy_t(n)); roll_dFz_f(qq)=(dFz_f(n)); roll_dFz_r(qq)=(dFz_r(n)); roll_del(qq)=del(n); 166 roll_t(qq)=t(n); end t_w_l_o(n+1)=two_wheel_lift_o(n)+t_w_l_o(n); % Two wheel lift detector of inside tires if Fz_fi(n) <= 0 & Fz_ri(n) <= 0 counter=1+counter; fi_lift(n) = 1; ri_lift(n) = 1; roll_fi_lift(qq) = 1; roll_ri_lift(qq) = 1; two_wheel_lift_i(n)=1; roll_ay(qq)=max(ay(1:n)); roll_r(qq)=(r(n)); roll_V(qq)=(vel(n)); roll_B_dt(qq)=(B_dt(n)); roll_B(qq)=(B(n)); roll_phiM(qq)=(phiM(n)); roll_phiM_dt(qq)=(phiM_dt(n)); roll_a_f(qq)=(a_f(n)); roll_a_r(qq)=(a_r(n)); roll_Fy_f(qq)=(Fy_f(n)); roll_Fy_r(qq)=(Fy_r(n)); roll_Fy_t(qq)=(Fy_t(n)); roll_dFz_f(qq)=(dFz_f(n)); roll_dFz_r(qq)=(dFz_r(n)); roll_del(qq)=del(n); roll_t(qq)=t(n); end t_w_l_i(n+1)=two_wheel_lift_i(n)+t_w_l_i(n); if prop == 0 % Do Nothing else if prop == 100 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 if t_w_l_i(n+1) > lift_time flag=1; break end if t_w_l_o(n+1) > lift_time flag=1; break end end,end