Tire Characteristics Sensitivity Study Master’s thesis in Automotive Engineering FOAD MOHAMMADI Department of Applied Mechanics Vehicle Dynamics Group Vehicle Engineering and Autonomous Systems CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden 2012 Master’s thesis 2012:34 MASTER’S THESIS IN AUTOMOTIVE ENGINEERING Tire Characteristics Sensitivity Study FOAD MOHAMMADI Department of Applied Mechanics Vehicle Dynamics Group Vehicle Engineering and Autonomous Systems CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden 2012 Tire Characteristics Sensitivity Study FOAD MOHAMMADI c© FOAD MOHAMMADI, 2012 Master’s thesis 2012:34 ISSN 1652-8557 Department of Applied Mechanics Vehicle Dynamics Group Vehicle Engineering and Autonomous Systems Chalmers University of Technology SE-412 96 Gothenburg Sweden Telephone: +46 (0)31-772 1000 Cover: Tire kicker: A metaphorical representation of the engineer’s view on the vast world of tires, From: The Tire as a Vehicle Component, Dr. Gerald R. Potts, Tire technology Expo 2012, Cologne, Germany Chalmers Reproservice Gothenburg, Sweden 2012 Tire Characteristics Sensitivity Study Master’s thesis in Automotive Engineering FOAD MOHAMMADI Department of Applied Mechanics Vehicle Dynamics Group Vehicle Engineering and Autonomous Systems Chalmers University of Technology Abstract Tires are the main contact between a vehicle and the road. Tire modelling for full vehicle simulations is a challenge due to tires’ non-linear behaviour and strong coupling between their different degrees of freedom. Several approaches for tire modelling with regard to handling are available. Empirical models such as Magic Formula fitted to test data are the most widely used. The aim of this thesis is to find an alternative tire model with relatively few parameters that are physically meaningful and hence suitable for parametric sensitivity studies. The model should also be representative of tire behaviour by an acceptable fit to tire test data. A survey of applications of tire models and suitable parameters for sensitivity analysis has been performed within Volvo cars with focus on handling. Simultaneously a literature study has been done to identify and categorize available tire models. TMeasy was chosen to be used in the parametric sensitivity studies. This semi-physical tire model fits to test data relatively well and is based on fewer parameters compared to the Magic Formula. The main reason for the selection of TMeasy is that the parameters describing the model, represent tire characteristics that were identified in the application survey as target parameters for sensitivity analysis. The effect of tire parameters on vehicle handling characteristics have been investigated by implementing TMeasy in Adams/Car and performing full vehicle simulations. Keywords: Tire models, TMeasy, Sensitivity analysis, Vehicle dynamics, Handling i ii Preface This thesis work has been conducted as a partial requirement for the Master of Science degree in “Automotive Engineering” at Chalmers University of Technology, Gothenburg, Sweden in cooperation with Volvo Car Corporation. All the stages of the project were performed at CAE vehicle dynamics team, CAE & Objective testing, Volvo Cars Research & Development during January – June 2012. I would like to acknowledge and thank my supervisor at Volvo Cars, Sergio da Silva for his guidance and support and my examiner at Chalmers, Mathias Lidberg for sharing his knowledge and expertise. Special thanks to all the engineers that participated in the application survey, for it would have been simply impossible without their help. I also would like to thank all the staff of the CAE vehicle dynamics group for their inputs and unconditional help throughout this project. Gothenburg, June 2012 Foad Mohammadi iii iv “Automobiles and trucks are machines for using tires.” - Maurice Olley, 1947 v vi Contents Abstract i Preface iii Contents vii 1 Introduction 1 1.1 Motivation and background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2 Targeting Tire Modelling Applications 3 2.1 Use-cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 2.2 Tire parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2.3 Application targeting conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 3 Review of Tire Models 7 3.1 Magic Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 3.1.1 Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 3.1.2 Force generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 3.2 TMeasy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 3.2.1 Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 3.2.2 Force generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 3.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 4 Model Verification 19 4.1 Comparison between TMeasy and Magic Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 4.2 Straight line braking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 4.3 Constant radius cornering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 4.4 Single lane change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 5 Sensitivity Analysis 29 5.1 Effect of maximum longitudinal force on stopping distance . . . . . . . . . . . . . . . . . . . . . . . 29 5.2 Effect of maximum lateral force and cornering stiffness on linear range understeer gradient . . . . 31 5.3 Effect of maximum lateral force and cornering stiffness on yaw rate settling time . . . . . . . . . . 33 6 Conclusions and Recommendations 35 6.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 6.2 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 A Appendix: other tire models 37 B Appendix: Adams/Tire module axis system 41 vii viii 1 Introduction 1.1 Motivation and background In 1887, John Boyd Dunlop made the first practical pneumatic tire from rubber for his son’s bicycle. Dunlop’s patent was declared invalid due to a prior work done on rubber made tires. He was credited with “realizing rubber could withstand the wear and tear of being a tire while retaining its resilience.” That was the beginning of a new field of research and a new chapter in the automobile history. Until the early 1920’s extensive experimentation was done by tire companies to increase the durability of the rubber by adding different materials to it such as wood, leather and finally steel cords. Pneumatic tires started to be used in aircraft and automotive industry. This started a new era in vehicle design from ride and handling perspectives. The importance of tires became more and more clear to engineers and researchers as automotive industry evolved. Specially during the past three decades enormous progress has been made in chassis engineering and drivability. Today’s vehicle are significantly safer and more comfortable compared to their predecessors. The chassis and suspension design and tuning has been improved by implementation of new analysis techniques and computer simulations. Evolution of electronic control systems has also provided new opportunities in chassis control, among which the most well known are the anti-lock brakes, active suspension, and traction control systems. In general, the product maturity increased in the automotive industry despite the short development times. This has brought new challenges for the engineers leading to more and more reliance on computer simulations and analysis of the components and systems. Being the only contact between the road and the vehicle, the importance of tire modelling became more evident. As depicted in Figure 1.1.1, the effect of tires on overall vehicle operational characteristics is significant. Chapter 1 Introduction 1.1 Motivation and background During the past twenty years, enormous progress has been made in chassis engineering and drive- ability. Today’s vehicles are significantly safer and at the same time more comfortable than cars were in the past. The reasons for this progress can be found in the areas of the various electronic systems and the precise design and tuning by using modern analysis and simulation methods. This results in an increase of product quality and shorter development times. In this context, the significance of modelling and simulation of the tyre behaviour becomes increasingly important. As the link between the vehicle and the road, the tyre ultimately determines the driving characteristics that can be realised and is an important factor for the ride comfort. An overview of the influence of the road, tyres and vehicle on operational characteristics is depicted in figure 1.1. In this figure it can be seen that the tyre affects many different aspects of the vehicle behaviour. Figure 1.1: Influence of the road, tyre and vehicle on the driving behaviour [6]. Due to the advances in simulation techniques of vehicle development and engineering, the mod- elling of the tyre is of special importance. Thereby, not only the reliability of quantitative results but also the extension to higher frequency ranges is becoming a necessity. A detailed description of differ- ent tyre models and the validation of measurements combined with specific capability tests to show 1 Figure 1.1.1: Influence of road, vehicle and tires on overall operational characteristics [3] Still, after 125 years from their introduction, tires are treated as a black box. Tire behaviour is heavily dependent on its structure, construction and rubber decomposition. The later appears to be the most complex aspect. For many years, tire rubber compounding was driven by experimentation. The relation between a specified measure of a specific tire compound and its effect on the road forces is very hard to derive, if not impossible. Material and assembly variations occur during manufacturing and curing leading to the fact that even in the same batch of production, there are not two tires that are exactly the same. The other challenge in modelling tires is their non-linear behaviour and strong coupling between different degrees of freedom. There are several parameters that effect tire behaviour. These parameters in most of the cases have mutual influence on each other. Extensive research has been done to identify these relationships and isolate their effects. The rise of precision electronics and measurement systems enabled tire and vehicle engineers to have a 1 deeper insight into tire modelling by measuring tire force and moment characteristics using test rigs and test vehicles. These tests, apart from being extremely costly, are entitled to problems such as repeatability, tire wear, road condition variances and measurement errors. Furthermore, these tests are done for only a limited number of working conditions. This necessitates the need for a model of some sort to interpolate or extrapolate the force and moment characteristics for other working conditions. On the other hand, for tire testing it is required that an actual tire is built and measured in a testing facility. Due to short development times in automotive industry, prototype tire availability for testing is very costly and sometimes impossible. Another consequence of these short development times was more reliance on computer modelling and simulations rather than prototype testing. These were the driving forces for the introduction of tire models. Today, understanding the tire behaviour and modelling techniques is vital for product development within the automotive industry. This is required to avoid costly prototypes and replace them with computer simulation models. It is needed to assist the vehicle-tire co-development process in which tire suppliers and vehicle manufacturers work in harmony to deliver a good overall system (the vehicle) rather than separately developed subsystems in isolation. The knowledge gained from studying tire behaviour in handling can assist the development of suspension and chassis systems leading to improved vehicle handling characteristics and resulting in more driving pleasure and safety for the end costumer. 1.2 Scope At the start of the vehicle development there is none or very little data of the final tire available for modelling purposes. Furthermore, due to the aforementioned challenges it is hard to isolate the effect of tires in vehicle dynamic testing. To better understand the effect of tire parameters on vehicle handling, sensitivity studies can be performed by varying a certain tire parameter and investigating the effect on a certain vehicle dynamic handling characteristic. Therefore there is a need for a tire model that is able to change the tire physical characteristics separately and be relatively easy to use in vehicle dynamics simulations in terms of computational effort and user-friendliness. Empirical models such as Magic Formula has been used for many years in vehicle dynamic handling simulations and incorporates some capabilities for such sensitivity studies. Although the large number of parameters and scaling factors makes it somewhat difficult to use. So there is a need for a simpler tire model that has physically meaningful parameters that are easy to change. The requirements for such a model are set based on an application survey done in this work. In the survey the aim is to find the application of tire models within Volvo Cars Corporation (VCC) with a focus on vehicle dynamic handling applications. The goal of this thesis work is to develop a method to study the effect of tire characteristics on vehicle handling through sensitivity analysis. 2 2 Targeting Tire Modelling Applications So far the importance of understanding the effect of tire characteristics on vehicle handling have been mentioned. Sensitivity analysis is to be used to gain knowledge about tire characteristics influence on vehicle handling. Therefore it is essential to identify the cases in which the effect of tires are to be studied. To select the cases for the sensitivity study, an application targeting survey is performed to identify the following: • Use-cases: In what handling scenarios are these sensitivity studies applicable? Example: Effect of tires in straight line braking performance of the vehicle. • Use-case metrics: What indexes are used for those use-cases? Example: Stopping distance in meters in the straight line braking scenario. • Tire parameters: For a particular use-case, the effect of which tire characteristics are needed to be investigated? These characteristics are represented by tire parameters. Example: Effect of maximum longitudinal force on the stopping distance in the straight line braking scenario. The needed information was gathered by interviewing concept, development, calculation and test engineers as well as technical specialists and attribute leaders that work with tires, handling and steering. Moreover, to better understand the vehicle-tire co-development process in VCC, further investigation has been done on VCC-tire supplier relationship, tire selection procedure from start to the end of a project and last but not the least, tire models and modelling methods currently used in VCC. 2.1 Use-cases In order to identify scenarios for the sensitivity analysis, various handling and steering tests and the corresponding simulations are investigated. The aim is to identify the cases in which the tires would play a role in determining the performance of the vehicle in that particular scenario. The performance is measured by one or more quantity. This quantity is called the use-case metric and is used to assess the effect of tire parameters on the handling characteristics of the vehicle in the sensitivity analysis for each particular use-case. Table 2.1.1 presents the results of the application targeting survey. Note that some of the use-cases studied and their corresponding metrics are omitted in this version of the report due to confidentiality issues. Straight line braking test is used to assess the steady-state braking performance of the vehicle as well as transient and steady-state pitch gradients of the vehicle. Stopping distance measured in meters is the metric used to describe vehicle’s performance in this test. Tires’ longitudinal force characteristics such as maximum force and stiffness can influence the metric in this test. Constant radius cornering is used to characterize the steady-state turning performance of the vehicle. This test includes the full driving spectrum from normal to limit cornering situations by covering a range of lateral accelerations in a constant radius skidpad setup. Linear range understeer gradient measured in deg/g is the metric used to describe the vehicle cornering characteristics in this test. Tires’ lateral characteristics such as maximum force and cornering stiffness are known to influence vehicle’s performance in this test. The difference between these parameters for the front and rear tires will have an even greater effect on the metric. The main objective of the Yaw stability test is to determine the transient behaviour of the vehicle by the means of a steering step input and observing the yaw overshoot as well as the maximum vehicle side slip angle. Tire’s lateral force characteristics such as cornering stiffness, maximum force and corresponding slip angle can affect the metric. Since this test involves a transient state, tire’s relaxation is also important in determining the yaw stability of the vehicle. Power on in a corner mimics a real-world driving scenario where a driver tends to take a 90 degree turn under full acceleration (e.g. drive onto a main road from a side road with fast approaching traffic). This test is used to objectively determine a vehicle’s directional response properties when a combined throttle and steer input is applied. Turning capability (deg) is used as a metric to identify vehicle’s capability to stay on the road. Tires’ combined force generation and the lateral and longitudinal force characteristics can influence the turning capability of the vehicle. Brake in a turn test is used to objectively determine a vehicle’s directional response properties when brakes are applied in a turn. The test is performed for two cases: one with a moderate lateral acceleration and 3 Use-case Name Metric Tire Parameters Straight line braking Stopping distance (m) Maximum longitudinal force, Longitudinal stiff- ness, Shape (for ABS modulation). Constant radius cornering Linear range understeer gradient (deg/g) Maximum lateral force, Cornering stiffness. Yaw stability Maximum side slip angle (deg) Maximum lateral force, Slip angle at maximum force, Cornering stiffness, tire relaxation. Power on in a corner Turning capability (deg) Maximum combined lateral force, Slip angle at maximum combined lateral force, Corner- ing stiffness, Maximum combined longitudinal force, Slip angle at maximum combined longi- tudinal force, Longitudinal stiffness. Brake in a turn Max. side slip rear delta (deg) Maximum combined lateral force, Slip angle at maximum combined lateral force, Corner- ing stiffness, Maximum combined longitudinal force, Slip angle at maximum combined longi- tudinal force, Longitudinal stiffness. Single lane change Yaw rate settling time (s) Maximum combined lateral force, Slip angle at maximum combined lateral force, Corner- ing stiffness, Maximum combined longitudi- nal force, Slip angle at maximum combined longitudinal force, Longitudinal stiffness, tire relaxation. Table 2.1.1: Application targeting survey results one with a high lateral acceleration. Maximum side slip angle difference in rear is the metric that is used to assess the vehicle performance in a sensitivity study for this particular use-case. Just as the power on cornering test, tires’ lateral and longitudinal force characteristics as well as combined force generation will affect the metric. Single lane change test is used to objectively determine a vehicle’s transient response behaviour (yaw stability and response) under closely controlled test conditions similar to lane change manoeuvres in real traffic. Yaw settling time is used as a metric to measure the yaw stability of the vehicle in this test. Since the test is done in a transient combined slip situation, tires’ lateral and longitudinal steady-state force characteristics, combined force generation and transient behaviours such as tire relaxation are known to influence the yaw rate settling time of the vehicle. 2.2 Tire parameters Tire force and moment characteristics can be represented by one or more parameters. Therefore these parameters have physical significance in a sense that changing them results in changes in tire force and moment characteristics. A summary of tire parameters identified in the application targeting survey is presented in Table 2.2.1. Tire parameters are grouped based on their effect on tires’ lateral, longitudinal and self-aligning characteristics. Some other tire parameters are also considered that can not be grouped into these categories such as vertical load sensitivity (i.e. effect of vertical force on force and moment characteristics) and rolling radius. 2.3 Application targeting conclusion These parameters together with the study done on the objective handling and steering test procedures suggest that the ideal model should be simple and user-friendly and run based on the physically meaningful parameters mentioned. It should have a good representation of tire force and moment characteristics in pure and combined 4 Longitudinal Lateral Self-aligning torque Maximum force Maximum force Maximum torque Slip ratio at max. force Slip angle at max. force Slip angle at max. torque Shape of curve Shape of curve Shape of curve Maximum combined force Maximum combined force Slip ratio at max. combined force Slip angle at max. combined force Other Longitudinal stiffness Cornering stiffness Vertical load sensitivity Rolling resistance Transient response Rolling radius Table 2.2.1: Application targeting survey: Tire Parameters Summary slip situations. Also it should have the possibility of modelling camber and relaxation effects. Since all the tests are done on dry and flat surfaces, the model does not need to take effects such as belt deformations and enveloping properties into account. Concerning the sensitivity study, it should be easy to change the parameters in the model and hence change the tire force and moment characteristics. Thereby one can change a single tire parameter and perform full-vehicle simulations and compare the results with the original tire. This way it is possible to investigate the sensitivity of a metric in a particular use-case to a certain tire parameter. 5 6 3 Review of Tire Models During the past 50 years different approaches for modelling tires were explored. Semi-empirical tire models such as Magic Formula that fit to tire test data were developed to represent tires in vehicle dynamic simulations. By the improvement of computational power and simulation capabilities, complex physical tire models were developed to predict force and moment characteristics of the tire based on its physical characteristics and construction. While the later are widely used for ride, comfort and durability purposes, semi-empirical tire models such as Magic Formula based on tire force and moment measurements are the most common for vehicle dynamics handling simulations. There are also models that lay in between these two extremes: Tire models that use similarity method by manipulating the basic characteristics of the tire, and models based on simple mechanical representations of the tire structure. Figure 3.0.1 roughly summarizes various consequences that arise by taking each approach in tire modelling. 2.5. TIRE MODELS (INTRODUCTORY DISCUSSION) Several types of mathematical models of the tire have been developed during the past half-century; each type for a specific purpose. Different levels of accuracy and complexity may be introduced in the various categories of utilization. This often involves entirely different ways of approach. Figure 2.11 degree of fit number of full scale tests complexity of formulations effort insight in tyre behaviour number of special experiments approach more empirical theoretical from experimental data only using similarity method through simple physical model through complex physical model fitting full scale tyre test data by regression techniques distorting, rescaling and combining basic characteristics using simple mechanical representation, possibly closed form solution describing tyre in greater detail, computer simulation, finite element method e.g. Magic Formula e.g. Brush model e.g. RMOD-K. FTire FIGURE 2.11 Four categories of possible types of approach to develop a tire model. Exercise 2.3 Partial Differential Equations with Longitudinal Slip Included Establish the differential equations for the sliding velocities similar to the Eqns (2.58, 2.59) but nowwith longitudinal slip k included (a and 4 remain small). Note that in that case Vr s Vc and that Vr may be expressed in terms of Vcx (zVc) and k. Also, find the partial differential equations governing the deflections in the adhesion zone similar to the Eqns (2.60, 2.61). 81Chapter | 2 Basic Tire Modeling Considerations Figure 3.0.1: Four categories of possible types of approach to develop a tire model [25] In this chapter two tire models are represented: Magic Formula which is used as the reference tire model and TMeasy which is the model selected for the sensitivity analysis. Appendix A includes all the other tire models that did not stand out to be used in this work, but were reviewed for the sake of comparison and categorization. Throughout this thesis slip ratio (longitudinal slip) is defined as [25]: sx = −Vcx − reΩ Vcx (3.0.1) In which Vcx is the longitudinal velocity of the wheel center, re is the effective rolling radius and Ω is the rotational velocity of the wheel. Slip angle is defined as: sy = − tan−1 ( Vcy Vcx ) (3.0.2) where Vcy and Vcx are lateral and longitudinal velocities of the wheel center, respectively. 7 3.1 Magic Formula 3.1.1 Description Developed in 1987 by Egbert Bakker and Lars Nyborg from Volvo Cars and Hans B. Pacejka from Delft University of Technology, this semi-empirical tire model is the most commonly used in vehicle dynamics studies. Several versions of this model has been published as a result of this cooperation as in [6], [7],and [26]. TNO , a spin-off company from TU-Delft, and Pacejka developed this model even further with the commercial name of “MF-tyre”. Several versions of this model were released (For more information of different releases from TNO refer to the backward compatibility graph in [21, p. 14]). MF-Tyre is able to model steady-state as well as transient behaviours up to 8 Hz on smooth surfaces which makes it suitable for use in handling and steering analyses. Magic Formula is named as such because of its unique formulation that can fit any test data regardless of physical properties of the tire through a set of mathematical parameters. 3.1.2 Force generation The general form of the formula can be written as bellow: y = D sin[C arctan{(1− E)Bx+ E arctan(Bx)}] (3.1.1) and Y (X) = y(x) + SV (3.1.2) x = X + SH (3.1.3) Y(x) is the output and it could either be defined as lateral force Fy = y(sy) in which the input X becomes the slip angle sy or it could be used for the longitudinal force Fx = y(sx) in which X becomes the slip ratio sx. The remaining coefficients of the Magic Formula are described as B stiffness factor C shape factor D peak value E curvature factor SH horizontal shift SV vertical shift The basic format of the Magic Formula in pure slip including camber effects is described here in detail. For the complete formulation of the model refer to Chapter 4.3 of [25]. Tire properties can be changed by the use of “user scaling factors”. Table 3.1.1 shows 13 of these scaling factors that are used to change tire characteristics for the pure slip condition. The normalized change in vertical load is defined as: dfz = Fz − F ′zo F ′zo (3.1.4) in which Fz is the vertical load and F ′zo is the adapted nominal load which is roughly approximated by: F ′zo = λFzoFzo (3.1.5) in which Fzo is the nominal rated load. 8 Scaling factor Description λFzo nominal (rated) load λµx,y peak friction coefficient λµV with slip speed Vs decaying friction λKxsx brake slip stiffness λKysy cornering stiffness λCx,y shape factor λEx,y curvature factor λHx,y horizontal shift λV x,y vertical shift λKyγ camber force stiffness λKzγ camber torque stiffness λt pneumatic trail (effecting aligning torque stiffness) λMr residual torque Table 3.1.1: Magic formula user scaling factors for pure slip Longitudinal force (pure longitudinal slip) Fxo = Dx sin[Cx arctan{Bxsxx − Ex(Bxsxx − arctan(Bxsxx))}] + SV x (3.1.6) sxx = sx + SHx (3.1.7) Cx = pCx1 · λCx (> 0) (3.1.8) Dx = µx · Fz · ζ1 (> 0) (3.1.9) µx = (pDx1 + pDx2dfz) · λ∗µx (> 1) (3.1.10) Ex = (pEx1 + pEx2dfz + pEx3df 2 z ) · {1− pEx4sgn(sxx)} · λEx (≤ 1) (3.1.11) Kxsx = Fz · (pKx1 + pKx2dfz) · exp(pKx3dfz) · λKxsx (= BxCxDx = ∂Fxo/∂sxx at sxx = 0) (= CFsx) (3.1.12) Bx = Kxsx/(CxDx + εx) (3.1.13) SHx = (pHx1 + pHx2dfz) · λHx (3.1.14) SV x = Fz · (pV x1 + pV x2dfz) · {|Vcx|/(εV x + |Vcx|)} · λV x · λ′µx · ζ1 (3.1.15) Lateral force (pure lateral slip) Fyo = Dy sin[Cy arctan{Bysyy − Ey(Bysyy − arctan(Bysyy))}] + SV y (3.1.16) syy = s∗y + SHy (3.1.17) Cy = pCy1 · λCy (> 0) (3.1.18) Dy = µy · Fz · ζ2 (3.1.19) µy = (pDy1 + pDy2dfz) · (1− pDy3γ∗2) · λ∗µy (> 0) (3.1.20) Ey = (pEy1 + pEy2dfz) · {1− (pEy3 + pEy4γ ∗)sgn(syy)} · λEy (≤ 1) (3.1.21) Kysyo = pKy1F ′ zo sin[2 arctan{Fz/(pKy2F ′zo)}] · λKysy (= ByCyDy = ∂Fyo/∂syy at syy = γ = 0) (= CFsy ) (3.1.22) Kysy = Kysyo · (1− pKy3γ∗2) · ζ3 (3.1.23) By = Kysy/(CyDy + εy) (3.1.24) SHy = (pHy1 + pHy2dfz) · λHy + pHy3γ ∗ · λKyγ · ζ0 + ζ4 − 1 (3.1.25) SV y = Fz · {(pV y1 + pV y2dfz) · λV y + (pV y3 + pV y4dfz)γ ∗ · λKyγ} · λ′µy · ζ2 (3.1.26) Kyγo = {pHy3Kysyo + Fz(pV y3 + pV y4dfz)}λKyγ (=∼ ∂Fyo/∂γ at syy = γ = 0) (= CFγ) (3.1.27) 9 Self-aligning torque (pure lateral slip) Mzo = M ′zo +Mzro (3.1.28) M ′zo = −to · Fyo (3.1.29) to = t(syt) = Dt cos[Ct arctan{Btsyt − Et(Btsyt − arctan(Btsyt))}] · cos′ sy (3.1.30) syt = s∗y + SHt (3.1.31) SHt = qHz1 + qHz2dfz + (qHz3 + qHz4dfz)γ ∗ (3.1.32) Mzro = Mzr(syr) = Dr cos[Cr arctan(Brsyr)] (3.1.33) syr = s∗y + SHf (= syf ) (3.1.34) SHf = SHy + SV y/K ′ ysy (3.1.35) K ′ysy = Kysy + εK (3.1.36) Bt = (qBz1 + qBz2dfz + qBz3df 2 z ) · (1 + qBz5|γ∗|+ qBz6γ ∗2) · λKysy/λ∗µy (> 0) (3.1.37) Ct = qCz1 (> 0) (3.1.38) Dto = Fz · (Ro/F ′zo) · (qDz1 + qDz2dfz) · λt · sgnVcx (3.1.39) Dt = Dto · (1 + qDz3|γ∗|+ qDz4γ ∗2) · ζ5 (3.1.40) Et = (qEz1 + qEz2dfz + qEz3df 2 z ){1 + (qEz4 + qEz5γ ∗) 2 π arctan(BtCtsyt)} (≤ 1) (3.1.41) Br = (qBz9 · λKy/λ∗µy + qBz10ByCy) · ζ6 (3.1.42) Cr = ζ7 (3.1.43) Dr = FzRo{(qDz6 + qDz7dfz)λMrζ2 + (qDz8 + qDz9dfz)γ ∗λKzγζ0} cos′ sy · λ∗µysgnVcx + ζ8 − 1 (3.1.44) Kzαo = DtoKysyo (=∼ −∂Mzo/∂syy at syy = γ = 0) (= CMsy ) (3.1.45) Kzγo = FzRo(qDz8 + qDz9dfz)λKzλ −DtoKyγo (=∼ ∂Mzo/∂γ at sy = γ = 0) (= CMγ) (3.1.46) (3.1.47) For the longitudinal force in pure slip the Magic Formula coefficients, , coefficients B, C, D, E, SH and SV are defined by 14 constants as presented in Table 3.1.2. For the lateral force in pure slip, coefficients B, C, D, E, SH and SV can be described by 18 constants as the result of the fitting process, Table 3.1.3. As shown in Table 3.1.4, number of constants used to describe the aligning torque in pure slip conditions is 25. 10 Constant Description pCx1 Shape factor Cfx for longitudinal force pDx1 Longitudinal friction Mux at Fznom pDx2 Variation of friction Mux with load pEx1 Longitudinal curvature Efx at Fznom pEx2 Variation of curvature Efx with load pEx3 Variation of curvature Efx with load squared pEx4 Factor in curvature Efx while driving pKx1 Longitudinal slip stiffness Kfx/Fz at Fznom pKx2 Variation of slip stiffness Kfx/Fz with load pKx3 Exponent in slip stiffness Kfx/Fz with load pHx1 Horizontal shift Shx at Fznom pHx2 Variation of shift Shx with load pV x1 Vertical shift Svx/Fz at Fznom pV x2 Variation of shift Svx/Fz with load Table 3.1.2: Magic formula constants for longitudinal force in pure slip Constant Description pCy1 Shape factor Cfy for lateral force pDy1 Lateral friction Muy pDy2 Variation of friction Muy with load pDy3 Variation of friction Muy with squared inclination pEy1 Lateral curvature Efy at Fznom pEy2 Variation of curvature Efy with load pEy3 Inclination dependency of curvature Efy pEy4 Variation of curvature Efy with inclination pKy1 Maximum value of stiffness Kfy/Fznom pKy2 Load at which Kfy reaches maximum value pKy3 Variation of Kfy/Fznom with inclination pHy1 Horizontal shift Shy at Fznom pHy2 Variation of shift Shy with load pHy3 Variation of shift Shy with inclination pV y1 Vertical shift Svy/Fz at Fznom pV y2 Variation of shift Svy/Fz with load pV y3 Variation of shift Svy/Fz with inclination pV y4 Variation of shift Svy/Fz with inclination and load Table 3.1.3: Magic formula constants for lateral force in pure slip 11 Constant Description pBz1 Trail slope factor for trail Bpt at Fznom pBz2 Variation of slope Bpt with load pBz3 Variation of slope Bpt with load squared pBz4 Variation of slope Bpt with inclination pBz5 Variation of slope Bpt with absolute inclination pBz9 Slope factor Br of residual moment Mzr pBz10 Slope factor Br of residual moment Mzr pCz1 Shape factor Cpt for pneumatic trail pDz1 Peak trail Dpt = Dpt*(Fz/Fznom*R0) pDz2 Variation of peak Dpt with load pDz3 Variation of peak Dpt with inclination pDz4 Variation of peak Dpt with inclination squared pDz6 Peak residual moment Dmr = Dmr/(Fz*R0) pDz7 Variation of peak factor Dmr with load pDz8 Variation of peak factor Dmr with inclination pDz9 Variation of Dmr with inclination and load pEz1 Trail curvature Ept at Fznom pEz2 Variation of curvature Ept with load pEz3 Variation of curvature Ept with load squared pEz4 Variation of curvature Ept with sign of Alpha-t pEz5 Variation of Ept with inclination and sign Alpha-t pHz1 Trail horizontal shift Sht at Fznom pHz2 Variation of shift Sht with load pHz3 Variation of shift Sht with inclination pHz4 Variation of shift Sht with inclination and load Table 3.1.4: Magic formula constants for self-aligning torque in pure slip 12 3.2 TMeasy 3.2.1 Description TMeasy was first conceived as a tire model for vehicle dynamic simulations of agricultural vehicles. It has been successfully used in real-time applications with cars and heavy vehicles in programs such as veDYNA and SIMPACK [18]. This semi-empirical tire model is developed to assist in situations where there is none or insufficient tire data available. The idea is that in these occasions an engineering estimation can be done to interpolate or extrapolate the properties from a similar tire. The aim of TMeasy is to give useful tire forces from little information with model parameters that have concrete physical meaning. TMeasy model construction is presented here briefly. 3.2.2 Force generation 6 W. HIRSCHBERG ET AL. 2.5 Wheel Load The vertical tyre force Fz can be calculated as a function of the normal tyre deflection 4z = eTn 4z and the deflection velocity 4ż = eTn 4ṙ Fz = Fz(4z, 4ż) , (8) where the restriction Fz ≥ 0 holds. 2.6 Longitudinal Force and Longitudinal Slip The longitudinal slip sx = vx − rD Ω rD |Ω| . (9) is defined by the non-dimensional relation between the sliding velocity in longi- tudinal direction vGx = vx − rD Ω and the average transport velocity rD |Ω|. Fx x M x G dFx 0 sxsxsx M G F F adhesion full sliding adhesion/ sliding Figure 3: Typical Longitudinal Force Characteristics The typical graph of the longitudinal force Fx as a function of the longitu- dinal slip sx can be defined by the parameters initial inclination (longitudinal stiffness) dF 0 x , location sMx and magnitude of the maximum FM x , start of full sliding sGx and the sliding force FG x , Fig. 3. Curves without peaks can be treated by setting the parameter maximum force FM equal to the sliding force FG. The remaining parameter sM may be used for fitting the curve shape between the adhesion and the sliding region. 2.7 Lateral Slip, Lateral Force and Self Aligning Torque Similar to the longitudinal slip sx, Eq. (9), the lateral slip can be defined by sy = vGy rD |Ω| , (10) where the sliding velocity in lateral direction is given by vGy = vy (11) Figure 3.2.1: Typical longitudinal force characteristics As shown in Figure 3.2.1, a typical longitudinal force Fx as a function of longitudinal slip sx can be described by the following parameters[18]: • Initial inclination (longitudinal stiffness) dF 0 x • Maximum longitudinal force FMx • Longitudinal slip at maximum force sMx • Sliding force FGx • Longitudinal slip at sliding force sGx If the curve does not have a peak then FMx can be set equal to FGx . In that case sMx can be used to distinguish between adhesion and sliding region. The acting point of the resultant lateral force depends on the distribution of the force over the contact area length L (Figure 3.2.2). For small slip angles, when all the tread particles stick to the road, an almost linear distribution of force on the contact area length appears. In such a case the resultant lateral force is exerted on a point behind the center of the contact area. At average slip angle values the particles at the end of the contact area start to slide. That moves the acting point of the resulting lateral force forwards, sometimes even before the contact area center point. And finally at high slip angles only a small portion of tread particles at the front of the contact area stick to the ground. In these extreme cases the resulting force is applied at the center of the contact area. The self-aligning torque Ms is then obtained by multiplying the resulting lateral force Fy by the dynmic tire offset or pneumatic trail n. Ms = −nFy (3.2.1) Lateral force Fy and the dynamics tire offset are both functions of the lateral slip sy as shown in Figure 3.2.3. The parameters describing the lateral force are: • Initial inclination (cornering stiffness) dF 0 y 13 USER-APPROPRIATE TYRE-MODELLING 7 and the lateral component of the contact point velocity vy follows from Eq. (7). As long as the tread particles stick to the road (small amounts of slip), an almost linear distribution of the forces along the contact area length L appears. At average slip values the particles at the end of the contact area start sliding, and at high slip values only the parts at the beginning of the contact area stick to the road, Fig. 4. L a d h e si o n F y small slip values La d h e si o n F y sl id in g moderate slip values L fu ll sl id in g F y large slip values n F = k F sy ** y F = F f ( s )y * y F = Fy Gz z Figure 4: Principle of the Lateral Force Distribution at Different Slip Values The distribution of the lateral forces over the contact area length also defines the acting point of the resulting lateral force. At small slip values the working point lies behind the centre of the contact area (contact point P). With rising slip values, it moves forward, sometimes even before the centre of the contact area. At extreme slip values, when practically all particles are sliding, the resulting force is applied at the centre of the contact area. The resulting lateral force Fy with the dynamic tyre offset or pneumatic trail n as a lever generates the self aligning torque MS = −nFy . (12) The lateral force Fy as well as the dynamic tyre offset are functions of the lat- eral slip sy. Typical plots of these quantities are shown in Fig. 5. Characteristic parameters for the lateral force graph are initial inclination (cornering stiffness) dF 0 y , location sMy and magnitude of the maximum FM y , begin of full sliding sGy , and the sliding force FG y . The dynamic tyre offset has been normalized by the length of the contact area L. The initial value (n/L)0 as well as the slip values s0y and sGy characterize the graph sufficiently. 2.8 Generalized Tyre Characteristics The longitudinal force as a function of the longitudinal slip Fx = Fx(sx) and the lateral force depending on the lateral slip Fy = Fy(sy) can be defined by their characteristic parameters initial inclination dF 0 x , dF 0 y , location sMx , sMy and magnitude of the maximum FM x , FM y as well as sliding limit sGx , s G y and sliding force FG x , FG y , Fig. 6. Figure 3.2.2: Lateral force distribution over the contact area for different slip values 8 W. HIRSCHBERG ET AL. Fy y M y G dFy 0 sysysy M G F F adhesion adhesion/ sliding full sliding adhesion adhesion/sliding n/L 0 sy sy Gsy 0 (n/L) adhesion adhesion/sliding M sy sy Gsy 0 S full sliding full sliding Figure 5: Typical Plot of Lateral Force, Tyre Offset and Self Aligning Torque When experimental tyre values are missing, the model parameters can be pragmatically estimated by adjustment of the data of similar tyre types. Fur- thermore, due to their physical significance, the parameters can subsequently be improved by means of comparisons between the simulation and vehicle testing results as far as they are available. During general driving situations, e.g. acceleration or deceleration in curves, longitudinal sx and lateral slip sy appear simultaneously. The combination of the more or less differing longitudinal and lateral tyre forces requires a normal- ization process, cf. [4], [3]. One way to perform the normalization is described in the following. Generalized Slip The longitudinal slip sx and the lateral slip sy can vectorally be added to a generalized slip s = √( sx ŝx )2 + ( sy ŝy )2 , (13) where the slips sx and sy were normalized in order to perform their similar weighting in s. For normalizing, the normation factors ŝx and ŝy are calculated from the location of the maxima sMx , sMy the maximum values FM x , FM y and the initial inclinations dF 0 x , dF 0 x . Generalized Force Similar to the graphs of the longitudinal and lateral forces the graph of the generalized tyre force is defined by the characteristic parameters dF 0, sM , FM , sG and FG. The parameters are calculated from the corresponding values of Figure 3.2.3: Typical plots of lateral force, dynamic offset and self-aligning torque • Maximum lateral force FMy • Lateral slip at maximum force sMy • Sliding force FGy • Lateral slip at sliding force sGy The dynamic tire offset is normalized by the length of the contact area L. The initial value n/L0 together with slip values s0y and sGy describe the dynamic tire offset as a function of lateral slip. TMeasy simulates the tire behaviour in combined slip situations (braking or accelerating in a curve, for instance) by generalizing the tire characteristics through a normalization process (see Figure 3.2.4). The longitudinal slip sx and lateral slip sy can be generalized by vector addition as: s = √( sx ŝx )2 + ( sy ŝy )2 (3.2.2) For normalizing, the normation factors ŝx and ŝx are calculated based on the location of the maxima sMx , sMy , the maximum values FMx , FMy and the initial inclinations dF 0 x , dF 0 y . ŝx = FMx dF 0 x and ŝy = FMy dF 0 y (3.2.3) The generalized tire parameters are then calculated with the corresponding values of the longitudinal and 14 2 Modeling Concept When choosing small values vN > 0 the singularity at rD |Ω| = 0 is avoided. In addition, the generalised slip points then into the direction of the sliding velocity for a locked wheel. In normal driving situations, where rD |Ω| � vN holds, the differences between the primary and the modified slips are hardly noticeable. Fy sx s sy S ϕ FS M FM dF0 F(s) dF S y Fy Fy M Ssy Msy 0 Fy sy dFx 0 Fx M Fx SFx sx M sx S sx Fx s s Figure 5: Generalised tyre characteristics The graph F = F (s) of the generalised tyre force can be defined by the characteristic param- eters dF 0, sM , FM , sS and F S, Fig. 5. These parameters are calculated from the corresponding values of the longitudinal and lateral force characteristics. An elliptic function grants a smooth transition from the characteristic curve of longitudinal to the curve of lateral forces in the range of ϕ = 0 to ϕ = 90◦. The longitudinal and the lateral forces follow then from the according projections in longitudinal Fx = F cosϕ = F sNx s = F s sNx = f sNx (20) and lateral direction Fy = F sinϕ = F sNy s = F s sNy = f sNy , (21) where f = F/s describes the global derivative of the generalised tyre force characteristics. The generalised tyre force characteristics F = F (s) is now approximated in intervals by appropriate functions, Fig. 6. In the first interval a rational fraction is used which is defined by the initial inclination dF 0 and the location sM and the magnitude FM of the maximum tyre force. Then, the generalised tyre force characteristics is smoothly continued by two parabolas until it finally reaches the sliding area, were the generalised tyre force is simply approximated by a straight line. 2.4 Self Aligning Torque The distribution of the lateral forces over the contact patch length also defines the point of application of the resulting lateral force. At small slip values this point lies behind the center 7 Figure 3.2.4: Generalized tire characteristics [19] lateral tire parameters and normalization factors. dF 0 = √( dF 0 x ŝx cosϕ )2 + ( dF 0 y ŝy sinϕ )2 , sM = √( sMx ŝx cosϕ )2 + ( sMy ŝy sinϕ )2 , FM = √( FMx cosϕ )2 + ( FMy sinϕ )2 , (3.2.4) sG = √( sGx ŝx cosϕ )2 + ( sGy ŝy sinϕ )2 , FG = √( FGx cosϕ )2 + ( FGy sinϕ )2 The angular functions cosϕ = sx/ŝx s and sinϕ = sy/ŝy s (3.2.5) grant a smooth transition from characteristic curve of longitudinal to the curve of lateral forces in the range of ϕ = 0 to ϕ = 90◦. The function F = F (G) can be described in intervals by a broken rational function, a cubic polynomial and a constant FG. F (s) =    sMdF 0 σ 1 + σ ( σ + F 0 sM FM − 2 ) , σ = s sM , 0 ≤ s ≤ sM FM − (FM − FG)σ2(3− 2σ) , σ = s− sM sG − sM , sM < s ≤ sG FG , s > sG (3.2.6) When defining the curve parameters, one just has to make sure that the condition dF 0 ≥ 2 FM sM is satisfied, otherwise the function will have a turning point in the interval 0 ≤ s ≤ sM . By projecting the generalized force in longitudinal and lateral directions, the corresponding forces can be obtained: Fx = F cosϕ and Fy = F sinϕ (3.2.7) 15 Fz = Fznom Longitudinal force Fx Lateral force Fy Self-aligning torque Mz dF 0 x Initial slope dF 0 y Initial slope (n/L)0 Norm. pneumatic trail FMx Max force FMy Max force s0y sy trail changes sign sMx sx where fx(sx) = FMx sMy sy where fy(sy) = FMy sGy sy trail tends to zero FGx Sliding force FGy Sliding force sGx sx where fx(sx) = FGx sGy sy where fy(sy) = FGy Fz = 2× Fznom Longitudinal force Fx Lateral force Fy Self-aligning torque Mz dF 0 x Initial slope dF 0 y Initial slope (n/L)0 Norm. pneumatic trail FMx Max force FMy Max force s0y sy trail changes sign sMx sx where fx(sx) = FMx sMy sy where fy(sy) = FMy sGy sy trail tends to zero FGx Sliding force FGy Sliding force sGx sx where fx(sx) = FGx sGy sy where fy(sy) = FGy Table 3.2.1: TMeasy tire parameters The self-aligning torque can be calculated by Equation 3.2.2 via the dynamic tire offset. The dynamic offset is approximated as function of the lateral slip by a line and a cubic polynomial: n L =    (n/L)0(1− |sy|/s0y) , |sy| ≤ s0y −(n/L)0 |sy| − s0y s0y ( sGy − |sy| sGy − s0y )2 , s0y < |sy| ≤ sGy 0 , |sy| > sGy (3.2.8) The value of (n/L)0 can be approximated. For small lateral slip values, sy ≈ 0, the distribution of lateral force over the contact area length can be approximated by a triangle, see Figure 3.2.2. The point that the resultant lateral force is applied then given by n(Fz → 0 , sy = 0) = 1 6 L (3.2.9) This value can only serve as a reference point. The uneven distribution of pressure in longitudinal direction of the contact area results in the change of the deflection profile and the dynamic tire offset. The list of parameters that are needed to model a tire in TMeasy are shown in Table 3.2.1. Note that tire parameters for only two normal loads are needed (Nominal normal load and 2 times the nominal load). Tire force and moment characteristics for other normal loads are calculated based on quadratic functions. Forces and moments due to camber are calculated separately and then added to the tire force and moment values. These effects are discussed in detail in [18]. 3.3 Summary Different types of tire models were reviewed in this chapter. The use of semi-empirical tire models is generally preferred instead of more complex tire models in handling simulations due to their higher accuracy in representing tire force and moment characteristics. Meanwhile the complex tire models are widely used in ride, comfort, durability and misuse scenarios. Among the semi-empirical tire models a simple model was needed that is representative of the tire behaviour and characteristics. That means it should fit well to tire test data and incorporate effects such as camber, relaxation and combined slip. At the same time it should provide the possibility of changing the tire parameters presented in the application survey (Chapter 2). A MF-Tyre tire property file for use in Adams/Car has more than 130 parameters for model description. The large number of parameters makes Magic Formula rather difficult to use for sensitivity studies. Moreover, not all of these parameters have physically explicit meaning and can not even be measured on a real tire (parameters such as curvature factor pEx4 or exponent pKx3). Tire parameters realized in the application 16 targeting in most cases relate to more than one constant in the magic formula model. For example for changing the maximum longitudinal force for a particular camber angle and vertical load case, two parameters (pDx1 and pDx1) and 2 user scaling factors (λµx and λFzo) have to be adjusted. Magic formula’s accuracy comes with the cost of increased number of parameters and complexity in formulation. Therefore it is desired to find an alternative tire model with less parameters that can be related to the tire parameters realized in the application survey. However, because of its high accuracy in representing tire characteristics, Magic Formula is used as the reference tire model in this project. It was shown that the parameters used in TMeasy are in most cases the same as the parameters realized in the application survey results (page 4). Also it is fairly easy to change these parameters and all of them have physical meaning. When test data is missing, model parameters can be pragmatically estimated by adjustment of the data from a similar tire. Later in the project when there is more data available, these parameters can be easily derived and replace the estimated values in the model. TMeasy model presented in the literature however has limitations. The modelling of the self-aligning torque is inaccurate due to simplified contact area length and resultant force application point assumptions. The model is unable to capture asymmetries in tire force and moment behaviour. Also TMeasy is not able to model tire relaxation effects. Despite these shortcomings, considering the direct relation of model parameters to the results of the application survey, its simplicity and good longitudinal and lateral force representation, TMeasy appears to be the perfect compromise suitable for sensitivity studies. Figure 3.3.1 categorizes the models reviewed based on the approaches defined in Figure 3.0.1. Note that this is a rough categorization and some tire models can be considered to be in between two or three of these categories as they utilize various approaches to model specific tire characteristics. Figure 3.3.1: Tire models categorization 17 Requirement MFtyre TMeasy MF-Swift TreadSim Fiala Dynamic Friction UniTire Hankook RMOD-K FTire & CDtire Tire Force and moment representation pure slip X X X X X X X X X X Tire Force and mo- ment representation combined slip X X X X X X X X X Camber effects X X X X X X X X X Relaxation effects X X X X X X X Turn slip effect X X X X X X Number of parameters (scaled: 5 highest, 1 lowest) 4 2 4 4 1 4 3 3 5 5 Correlation to tire pa- rameters realized in the application target- ing survey (scaled: 5 highest, 1 lowest) 3 5 1 1 4 1 1 1 1 1 Table 3.3.1: Tire model requirements summary 18 4 Model Verification Upon selecting TMeasy as the suitable tire model for the sensitivity studies, it is necessary to verify the model. For this purpose TMeasy is verified against a MFtyre model (Magic Formula model by TNO). Full vehicle simulations are performed in three specific use-cases with TMeasy and MFtyre implementations in Adams/Car and their influence on the results and particularly the use-case metrics are investigated. The vehicle used in these simulations is the Adams/Car demo vehicle [1]. The simulation is set up in Adams/Car and done on the demo vehicle, Figure 4.0.1, once with the MFtyre model and once with TMeasy tire model. Note that for model verification and sensitivity studies the ISO coordinate system is used for tire force and moments, since it is standard in Adams/Tire module [2] and conforms with TYDEX tire testing and modelling standards [24]. The ISO coordinates and TYDEX C and W coordinate systems are reviewed in Appendix B. Figure 4.0.1: Adams/Car Demo vehicle and the TYDEX-C axis system 4.1 Comparison between TMeasy and Magic Formula First, a TMeasy model is created based on the MFtyre model data for a generic tire called “tire 1”. This is done by plotting the longitudinal, lateral and self-aligning torque curves for the MFtyre model; and then measuring the TMeasy parameters on the curves and implementing them in the TMeasy tire property file. The TMeasy parameters for tire 1 are shown in Table 4.1.1. Figures 4.1.1, 4.1.2 and 4.1.3 show the comparison between MFtyre and TMeasy longitudinal, lateral and self-aligning torque characteristics, respectively, for a nominal vertical load of 3000N and zero camber angle. Figure 4.1.1 shows that TMeasy matches MFtyre reletively well with 2.9% maximum relative error in longitudinal force. The error is due to the fact that TMeasy, as expected, is unable to capture the asymmetric force behaviour seen at negative slip ratios for this tire. In lateral force characteristics depicted in Figure 4.1.2, TMeasy matches MFtyre with maximum relative error of 3.7%. The tire asymmetric behaviour seen in positive slip angles are not realized by TMeasy. 19 Longitudinal force Fx Lateral force Fy Fz = 3 kN Fz = 6 kN Fz = 3 kN Fz = 6 kN dF 0 x = 82.2 kN dF 0 x = 236.2 kN dF 0 y = 53.7 kN dF 0 y = 95 kN FMx = 3.57 kN FMx = 6.57 kN FMy = 3.32 kN FMy = 6.08 kN sMx = 0.160 sMx = 0.100 sMy = 0.197 sMy = 0.196 FGx = 3.29 kN FGx = 6.01 kN FGy = 3.26 kN FGy = 5.83 kN sGx = 0.700 sGx = 0.500 sGy = 0.291 sGy = 0.349 Self-aligning torque Mz (n/L)0 = 0.17 (n/L)0 = 0.25 s0y = 0.190 s0y = 0.180 sGy = 0.400 sGy = 0.350 Table 4.1.1: Tire 1 TMeasy tire parameters Figure 4.1.1: Longitudinal force characteristics Figure 4.1.3 shows TMeasy’s disadvantage in simulating self-aligning torque behaviour for moderate to high slip angles. This is due to the rather simplistic assumptions made in modelling of the lateral force distribution and the resultant force application point described in section 3.2. Figure 4.1.4 shows the combined slip characteristics for TMeasy and MFtyre. Minor differences seen in the size of the friction ellipses are due to several reasons. First is the asymmetric tire behaviour not realized by TMeasy. The other reason is in the different approaches used in MFtyre and TMeasy for combined force calculations. While MFtyre fits some parameters to the test data obtained from combined slip tire tests, TMeasy relies on the data from pure slip tests and generates a general force curve based on normalizing the lateral and longitudinal forces. For higher vertical loads this difference increases where MFtyre “squares out” the elliptical shape resulting in higher lateral forces in the presence of a high longitudinal force. This phenomena is known to be MFtyre’s disadvantage in modelling combined situations for very high vertical loads or on wet surfaces. As mentioned in section 3.2, the current formulation of TMeasy does not include relaxation effects. Figure 4.1.5 compares TMeasy and MFtyre responses to a step input slip angle of 5 degrees in magnitude starting at t = 0.5 sec. It can be seen that MFtyre models the transient behaviour by including relaxation effects, while TMeasy simply switches the formula used to calculate the force (Equation 3.2.6). Straight-line braking, Constant radius cornering and Single lane change simulations are performed to ensure TMeasy result correlation with the same simulations with MFtyre. These simulations and their results are 20 Figure 4.1.2: Lateral force characteristics discussed in the following sections. Figure 4.1.3: Self-aligning torque characteristics 21 Figure 4.1.4: Combined slip characteristics Figure 4.1.5: Transient behaviour characteristics 22 4.2 Straight line braking The aim of this test is to measure the stopping distance of the vehicle in meters. While the vehicle is travelling on a straight line with 150 km/h, full ABS braking is performed (120 N on the brake pedal) with steering wheel fixed to 0 deg. The stopping distance is the distance travelled while the vehicle decelerates with a constant rate from 100 km/h to 2 km/h. It is measured from 100 km/h so that the dynamics of braking such as pitch angle changes are removed. It is measured to 2 km/h to remove the discrepancies in sensors due to standstill and pitch angle recovery. This test is selected to demonstrate the pure steady-state longitudinal force generation in TMeasy and to ensure its correlation to MFtyre results. The resulting metrics are shown in Table 4.2.1. Tire model Stopping distance (m) ∆ MFtyre 34.4 TMeasy 34.8 -0.4 m (-1.2%) Table 4.2.1: TMeasy vs. MFtyre straight line braking results Both cases reach 100 km/h at t = 1.23 sec. Figure 4.2.1 illustrates the chassis longitudinal accelerations and longitudinal forces on the front left and rear left tires for both cases (longitudinal forces on the right side tires have the same behaviour and are not showed here for simplicity). It is evident that TMeasy can easily capture the steady state pure longitudinal behaviour of the tire. There are however minor differences in the transient part (before t = 1.23 sec). These are due to the differences in rolling radius calculations between TMeasy and MFtyre. The rolling (dynamic) radius formulation in TMeasy is very simplified based on an approximate assumption as a first guess. This influences the normal load calculation which in turn effects the interpolation algorithms used to calculate the longitudinal force on the tire. Comparing the results and considering the minimal error in calculation of the metric (stopping distance normally varies with a larger error when this test is done on a prototype vehicle several times), it can be concluded that the TMeasy model correlates well with the MFtyre model in longitudinal force characteristics. Figure 4.2.1: TMeasy vs. MFtyre straight line braking results: (top) longitudinal acceleration (middle) longitudinal force on the left front tire (bottom) longitudinal force on the left rear tire 23 4.3 Constant radius cornering The aim of this test is to asses the pure steady-state lateral force characteristics of the TMeasy tire model and ensure its reliability by comparing it to MFtyre. The test is done on a skid-pad track with a radius of 40m for both right and left turns. The vehicle is then driven such that it covers a range of lateral accelerations from 0.1 to 1 g by slowly increasing the velocity. The linear range understeer gradient (deg/g) is the metric used for this particular use-case. (linear range = up to 0.4 g lateral acceleration) The simulations are performed in Adams/Car with the demo vehicle and the results are shown in Table 4.3.1. Tire model Understeer gradient (deg/g) ∆ MFtyre (left turn) 1.73 TMeasy (left turn) 1.69 -0.04 deg/g (-2.3%) MFtyre (right turn) 1.71 TMeasy (right turn) 1.69 -0.02 deg/g (-1.2%) Table 4.3.1: TMeasy vs. MFtyre constant radius cornering results The difference seen in the right and left hand understeer gradient for MFtyre, although small, is due to tire asymmetric behaviour. This difference is not realized by the TMeasy model. Figure 4.3.1 shows the relation between the front wheel angle and lateral acceleration for both left and right turns for TMeasy and MFtyre. The “equivalent” front wheel angle (bicycle model front wheel) is obtained by dividing the steering wheel angle by the steering gear ratio. The slope of any curve in the linear range is the understeer gradient. (Note that the values on the horizontal and vertical axes are absolute values for presentation purposes.) Figure 4.3.1: TMeasy vs. MFtyre constant radius cornering understeer behaviour comparison In general it can be said that TMeasy correlates to MFtyre in the linear range. However MFtyre yields a higher progressiveness for the curve in higher lateral accelerations. Comparing left to right hand turns there is not much difference between the two TMeasy curves while MFtyre shows that the vehicle is more understeer in a right hand turn compared to a left hand turn for high lateral accelerations. This as mentioned before is due to tire asymmetric behaviour. 24 4.4 Single lane change This use-case is selected to asses two different aspects of tire modelling. First is to compare TMeasy and MFtyre in a combined slip situation. Second is to show one of TMeasy model limitations which is replicating the tire transient behaviour. This test is done by giving a sine input to the steering wheel of the vehicle travelling at 80 km/h, Figure 4.4.1. The speed is kept constant throughout the test by applying throttle if needed. The metric is the settling time of the chassis yaw rate after the steering sine input is finished and its value is zero again ( t = 3 sec in the figure). The settling time is defined as the time it takes for chassis yaw rate to settle to a value within ±1% of its final value: ts = t(ṙ→ṙfinal±1%) − tsteering end (4.4.1) Figure 4.4.1: TMeasy vs. MFtyre single lance change test steering input Table 4.4.1 compares the results obtained for two tire models. Figure 4.4.2 shows the yaw rate vs. time for the simulations done with TMeasy and MFtyre. The settling time for the TMeasy model is less than the one for the MFtyre model. The reason for that is that the MFtyre model needs time to build up the lateral force when the slip angle changes (relaxation effect) while the TMeasy model generates a the new lateral force for a changed slip angle instantly (Figure 4.1.5). Thus if we assume at t = 3 sec the slip angle on all tires become zero, it takes more time for MFtyre to remove the lateral force from the tires compared to TMeasy in which lateral forces become zero instantly. Tire model Settling time (s) ∆ MFtyre 0.725 TMeasy 0.505 -0.220 s (-30.3%) Table 4.4.1: TMeasy vs. MFtyre single lane change results When the steering input starts to deviate from 0 in the beginning of the sine input, just after t = 1 sec, a minor delay can be seen from the MFtyre model. This results in a slight relative error which accumulates throughout the steering input phase. Another source of this relative error, particularly seen at the peak values, is due to the different rolling radius calculation methods used in these models and its effect on the normal load on the tire which in turn effects the longitudinal and lateral force values generated. 25 Figure 4.4.3 summarizes the effect of the errors mentioned from another perspective, by observing the chassis roll angle. It can be seen that the vehicle running with the TMeasy tire model rolls less than the one running with MFtyre. This means that the vehicle with TMeasy model has lower lateral weight transfer, hence higher effective axle cornering stiffness. This causes the vehicle to stabilize faster compared the vehicle with the MFtyre tire model. 26 Figure 4.4.2: TMeasy vs. MFtyre single lance change test chassis yaw rate Figure 4.4.3: TMeasy vs. MFtyre single lance change test chassis roll angle 27 28 5 Sensitivity Analysis The aim of the sensitivity analysis is to investigate the effect of the tire parameters selected in the application survey on the metrics defined for each particular use-case. Three use-cases are presented here: straight line braking with stopping distance as the metric, constant radius cornering with linear range understeer gradient as the metric and single lane change with yaw rate settling time as the metric. This chapter strives to demonstrate the working method for a sensitivity study. Investigating the results of the analysis to draw conclusions and provide guidelines for tire development are outside the scope of this thesis work. The reference tire is “tire 1” modelled by TMeasy. The tire is hence called “tmeasy 1” with parameters defined in Table 3.2.1 (page 16). In each simulation a single tire parameter is changed and the results are compared to the simulations performed with the reference tire. The aim is to find out how much the metric changes in a particular use-case due to X % change in a single tire parameter. In some cases changing one tire parameter necessitates a change in other tire parameters due to physical or mathematical relations between them. For example changing the cornering stiffness without shifting the slip angle at maximum force, will affect the shape of the curve as well, which in some cases might not be a good representation of an actual lateral force behaviour for a tire. 5.1 Effect of maximum longitudinal force on stopping distance For the straight line braking test only the effect of maximum longitudinal force on stopping distance is required. Two tires are made based on the reference tire. The maximum longitudinal force for the tire “tmeasy 1aa” is 10% more than the reference tire, while the same force for tire “tmeasy 1ab” is 10% less than the reference tire. As full ABS braking (120 N on pedal) is performed special care was taken to ensure the shape of the curve is preserved for all three tires. This was done by changing the sliding force along with the maximum force. These changes are done for both nominal vertical load and 2 × nominal vertical load. Figure 5.1.1 compares the longitudinal force characteristics of these three tires. Table 5.1.1 compares the metric stopping distance for these three tire models. Tire model Changes made Stopping distance (m) ∆ tmeasy 1 Reference 34.8 tmeasy 1aa 10% increase in FMx and FGx 32.5 -2.3 m (-6.6%) tmeasy 1ab 10% decrease in FMx and FGx 38 +3.2 m (+9.2%) Table 5.1.1: Straight line braking sensitivity analysis results As shown in Figure 5.1.2, increasing the maximum longitudinal force have increased the steady-state longitudinal deceleration rate of the vehicle resulting in a decrease in the stopping distance. While decreasing the maximum longitudinal force decreases the steady-state deceleration rate of the vehicle and hence the vehicle should travel more to get from 100 to 2km/h. 29 Figure 5.1.1: Longitudinal force vs. slip ratio characteristics for tires tmeasy 1, tmeasy 1aa and tmeasy 1ab Figure 5.1.2: Longitudinal acceleration sensitivity for straight line braking events with tires tmeasy 1, tmeasy 1aa and tmeasy 1ab 30 5.2 Effect of maximum lateral force and cornering stiffness on linear range understeer gradient For the constant radius cornering test two tires are introduced. Tire “tmeasy 1ba” has 10% increase in maximum lateral force compared to the reference tire. Tire “tmeasy 1bb” has 30% decrease in cornering stiffness compared to the reference tire. These changes are done for both nominal vertical load and 2 × nominal vertical load. Figure 5.2.1 shows the lateral force characteristic for the tires used in this study. In case of tmeasy 1bb the slip angle at maximum lateral force is slightly increased to make a smooth transition in the curve just before the peak. Figure 5.2.1: Lateral force vs. slip angle characteristics for tires tmeasy 1, tmeasy 1ba and tmeasy 1bb Figure 5.2.2 compares the metric understeer gradient for all three cases. Table 5.2.1 shows the linear range understeer gradient for the simulations performed with these tires. It can be seen that the tire with less cornering stiffness (tmeasy 1bb) is more understeer in the linear range but becomes less understeer compared to the reference tire in high lateral accelerations due to the shift in the slip angle at maximum lateral force. Increasing the maximum lateral force (tmeasy 1ba) have caused the vehicle to be less understeer for all the lateral accelerations. Tire model Changes made Understeer gradient (deg/g) ∆ tmeasy 1 Reference 1.69 tmeasy 1ba 10% increase in FMy and FGy 1.66 -0.03 deg/g (-1.8%) tmeasy 1bb 30% decrease in dF 0 y 1.85 +0.19 deg/g (+11.4%) Table 5.2.1: Constant radius cornering sensitivity analysis results 31 Figure 5.2.2: Understeer gradient for constant radius cornering events with tires tmeasy 1, tmeasy 1ba and tmeasy 1bb 32 5.3 Effect of maximum lateral force and cornering stiffness on yaw rate settling time For the single lane change the same tires used in constant radius cornering test are utilized. Table 5.3.1 shows the yaw rate settling time for all three cases and their difference. Figure 5.3.2 shows the yaw rate response of the vehicle in all three cases just after the sine steering input is finished at t = 3 sec. It can be deduced that the settling time have increased for both tires tmeasy 1ba and tmeasy 1bb, compared to the reference case. The tire with the higher lateral force capability (tmeasy 1bb) results in a lower yaw rate during the whole manoeuvre, Figure 5.3.1. Tire model Changes made Settling time (s) ∆ tmeasy 1 Reference 0.505 tmeasy 1ba 10% increase in FMy and FGy 0.905 +0.400 s (+79.2%) tmeasy 1bb 30% decrease in dF 0 y 1.2 +0.695 s (+137%) Table 5.3.1: Single lane change sensitivity analysis results Figure 5.3.1: Yaw rate for single lane change events with tires tmeasy 1, tmeasy 1ba and tmeasy 1bb Figure 5.3.3 shows these differences from the tire’s perspective. The lateral forces on the left rear tire are selected to be presented because of the weight transfer to the left at the end of the manoeuvre and also in rear because the differences are more evident. It can be seen that tmeasy 1ba generates more force than the reference tire. tmeasy 1bb with less cornering stiffness generates the same lateral force as the reference tire but with a slower response which is due to the shift of the slip angle at maximum force. The effect of this shift can be seen at the end of the manoeuvre, where for the same slip angle the tire with the lower cornering stiffness generates less lateral force. In section 4.4 it was stated that TMeasy does not incorporate the tire relaxation effects and hence it does not correlate with the results from MFtyre simulations. Based on that, special care should be taken in using the actual values generated by TMeasy for this special use-case where the metric is heavily influenced by the tire relaxation effects. However TMeasy can still be used to identify the trends and the direction of changes in these kind of sensitivity studies (with transient effect). This is particularly of importance as this shows that, despite the shortcomings mentioned with the model, the results can still be used to draw conclusions on what tire parameters should change and in what direction, in order to achieve a desired value for a use-case metric. 33 Figure 5.3.2: Yaw rate settling after t = 3 sec for single lane change events with tires tmeasy 1, tmeasy 1ba and tmeasy 1bb Figure 5.3.3: Lateral force on the left rear tire for single lane change events with tires tmeasy 1, tmeasy 1ba and tmeasy 1bb 34 6 Conclusions and Recommendations 6.1 Conclusion The aim of this work was defined as “Developing a method to study the effect of tire characteristics on vehicle handling through sensitivity analysis”. For that purpose an application survey was done within Volvo Cars to identify different use-cases in handling field of vehicle dynamics. In each use-case a metric was defined as an index. A metric is used to describe the vehicle characteristics in a particular use-case or scenario. Also in each use-case, certain tire parameters were identified to be used in sensitivity studies. A literature study was done to identify and categorize available tire models and to select one that is suitable for sensitivity studies based on the results of the application survey. TMeasy was chosen because it represents tire force and moment behaviour relatively well. Another reason for selecting TMeasy is that the model parameters have physical significance and are the same as the tire parameters realized in the application survey for sensitivity analysis. TMeasy, however, has certain limitations such as inaccuracy in self-aligning torque modelling, exclusion of relaxation effects and inability to capture tire asymmetric force and moment generation. TMeasy was then verified against MFtyre in three particular use-cases. The use-cases were selected to assess different aspects of tire modelling as well as showing the limitations of TMeasy. In general, it was shown that the results of the metrics generated with TMeasy model correlates well the ones generated with MFtyre. Sensitivity analysis was done in three use-cases in which the effect of different tire parameters where investi- gated. The physical significance of TMeasy model parameters made it easy to change the tire characteristics and study their effects on vehicle handling. Even in the use-case where the metric was influenced by relaxation effects, it was deduced that the method can still be applied with the current implementation of TMeasy to identify which tire parameters can be used and in what direction the changes can be made. As a conceptual model TMeasy can be used when there is little or no test data available. In such a case tire parameters can be pragmatically estimated from another tire, thanks to the parameter set used in TMeasy model. Sensitivity studies can be performed to analytically investigate the effect of tire characteristics on handling attributes of the vehicle. In general, all the chassis systems in a road vehicle are developed based on tire characteristics. Particularly suspension systems are designed around the tire properties such that they maximize the performance of the tire. This has always been done by a “black box” approach where tire is assumed to be unchangeable and the whole suspension design and geometry is altered to reach the targets set for different attributes. Fulfilling those numerous ride, handling, steering, durability, NVH and packaging targets always results in set of compromises depending on the importance and priority of each area and the corresponding attributes. So in some cases an attribute is sacrificed to achieve a target in another. Engineers have been successfully able to optimize the vehicle for certain road and tire properties. Although it is not possible to make any changes on the road conditions, tire development is somewhat under the influence of car manufacturers. Having few physically meaningful parameters in the tire model makes it possible to perform optimization on the vehicle-tire combination rather than optimising each subsystem in isolation. Therefore having the possibility of changing tire parameters can open new opportunities for chassis and suspension design, specially in cases where there has to be a trade-off in the chassis design to accommodate a target for a particular attribute. However, driving the tire development based on analytical methods like this has been unsuccessful in the past. Technically it is hard for the manufacturer to know the exact force and moment characteristic of a tire in advance. Although approximative methods are used, the actual force and moment characteristics are obtained by physically testing a tire after its production. This is very costly and in some cases impossible due to short lead times. Therefore, in the automotive industry, the tire selection process is mostly driven by subjectively testing the tires on a prototype vehicle. The method developed can be used to support this selection process. In most cases the macro-scale measures used in subjective testing can be related to physical tire parameters. Utilizing this method, it is possible to change those parameters analytically and investigate the effects on the metrics. This ultimately helps in reducing the testing hours needed for tire selection by providing an opportunity to do a pre-selection analytically using simulations. Another way this method can help in tire selection process and reducing testing time is by enabling the engineers to set targets for tire development using aforementioned vehicle-tire optimizations. 35 6.2 Recommendations To increase the confidence in the method, certain tasks should be performed. Within the scope of this work, the TMeasy model was verified using the MFtyre model. It was assumed that MFtyre represents the tire force and moment characteristics precisely. This assumption, however, is not entirely correct. MFtyre has some limitations as well, for example in the combined force generation under high loads or on low friction surfaces. Also the quality of data used for obtaining the MFtyre for tire 1 is unknown. Therefore, it is recommended that the simulations performed with TMeasy be compared with physical objective tests, to be able to validate the tire model. It goes without saying that the vehicle model used in Adams/Car has to be validated beforehand. This can be done on any previous test and the corresponding validated Adams/Car model and hence it does not require any separate physical tests to be performed. The presence of a ”good quality” test data for the tires is also vital for building accurate TMeasy tire models for the sake of validation. The number of sensitivity study cases and tire parameters involved in this work were limited, due to time constraints. To improve the confidence in the method, it is highly recommended that more use-cases and tire parameters be involved in further analyses. Regarding TMeasy, the behaviour of self-aligning torque could be improved by introducing more mature assumptions for lateral force distribution over the contact area length and calculation of the application point for the resultant force. Relaxation effects can also be included by introducing lagged longitudinal and lateral slip quantities, or by using a simple stretched string model, as is currently used in MFtyre for simulating tire relaxation effects. Also some scaling parameters could be used to model tire asymmetrical force and moment behaviour. In the sensitivity studies performed, the tire parameters used were the ones that are known to have an effect on the use-case metrics. The next step would be investigating the influence of other tire parameters such as vertical stiffness, unloaded radius and tire width, on vehicle handling. Ultimately, it is desirable to use the method in a suspension-tire optimization process to further investigate the possibilities it might yield in chassis and suspension design as well as tire development. 36 A Appendix: other tire models MF-SWIFT Short Wavelength Intermediate Frequency Tire model is constructed using the magic formula (MF) tire model plus a rigid ring which represents the tire belt. It is made to work in higher frequencies than the MF model, typically between 60-100 Hz as long as the wavelength is relatively short (>20 cm). It also include gyroscopic effects. The model is described in [25, Ch. 9] and [28]. It has evolved since then by TNO under the commercial name of MF-Swift. Different versions with added features have been released, refer to the backward compatibility graph in [21, p. 14]). MF-SWIFT is mostly used in ride and comfort applications. TreadSim This physical tire model was originally developed by Pacejka and presented in section 3.3 of [25]. The calculation of the forces and moments is based on the time simulation of the deformation history of one tread element while moving through the contact zone. In [29] Uil argues that a weak point of TreadSim is that the peak lateral friction coefficient for high vertical loads is generally too high and its dependency on inflation pressure does not match the behaviour found in tire F&M Measurements. Poorly modelled contact patch shape and pressure distribution are recognised as the cause for that. There are also some assumptions that are not realistic such as realization of carcass deformation of approximately 0.5 m. Although TreadSim is a physical tire model, the tire properties need to be determined from a tire measurement data set. Dynamic tire friction model Is a physical brush model based on LuGre friction modelling presented in [10]. The LuGre friction model was developed by Department of Automatic Control at Lund University (Sweden) and Laboratoire d’Automatique de Grenoble (France).It describes a dynamic force phenomena when frictional surfaces are sliding on each other. In the formulation for tires LuGre assumes that the friction surface is consisted of bristles that their movements are described by differential equations. Duer has feature developed the model by describing the dynamics of longitudinal lateral friction forces as well as self-aligning moment [11], [9] and [8]. The model parameters are highly dependent on the tire-road condition. Apart from being hard to measure, most of the parameters do not represent physical characteristics of the tire. These being mostly friction model parameters that are hard to relate to tire characteristics. CDtire CDtire was originally developed for comfort and durability purposes. The CDtire model family consists of: • CDtire 20: a rigid ring model with (global) viscous-elastic side-wall, geometrically parametrized normal contact and partial differential equation based tangential contact, • CDtire 30: a flexible belt, rod-type in-plane model with (local) viscous-elastic side-wall and brush type contact, • CDtire 40: a flexible belt, shell-type 3D model with (local) viscous-elastic side-wall and brush type contact, [5]. In terms of construction and applications this model is very similar to FTire. For some results of model verification, see [4]. RMOD-K RMOD-K is the name of a family of tire models that mostly reside on the “complex physical models” category. In its most sophisticated form (flexible belt) this detailed finite element model describes the actual shape of tire structure and can cover road undulation frequencies up to 100 Hz. The belt is attached to the rim with a side wall model that considers the effect of pressurized air in the tire. The belt can have more than one layer. These layers can interact with each other (frictional forces, etc.). The friction model includes both sliding and adhesion functions and also realizes the effect of temperature and contact pressure. The complexity can 37 be reduced based on application. In other words, the model can switch to simplified representations or run in hybrid mode. RMOD-K is mainly designed for ride, comfort and durability applications and it has been implemented in ADAMS [23]. For a complete list of literature (mostly in German) refer to [22]. Hankook tire model This semi-physical tire model is developed by Hankook tire Co. Ltd R&D center [14]. The steady-state behaviour is modelled through use of analytical formulation based on physical characteristics of the tire. Three forces and three application points are determined this way. For the transient behaviour different slip sweep rates are used to determine the transient tire characteristics. Although some of the parameters represent physical characteristics, the analytical formulation is mostly polynomial form with mathematical fitting parameters. In the literature it is stated that these parameters are to be experimentally derived but it is not clear how and from what kind of tests (slip sweeps supposedly). Fiala Fiala tire model is the standard tire model in all Adams/Tire modules [2]. It is a physical model, where the carcass is simulated as a beam with an elastic foundation for modelling the deformations in the lateral direction. The carcass-road contact is modelled by elastic brush elements. It is assumed that the contact patch has a rectangular shape with a uniform pressure distribution across the footprint. Based on these assumptions and neglecting the influence of camber angle, the expressions for steady-state slip conditions can be analytically derived. Based on these slip conditions the lateral and longitudinal forces are derived. The nine inputs needed for the Fiala tire model are very simple and can be found in Table A.0.1 . Parameter Description Mt Mass of tire Vertical damping Vertical damping coefficient Vertical stiffness Vertical tire stiffness CSLIP Partial derivative of longitudinal force (Fx) with re- spect to longitudinal slip ratio at zero longitudinal slip CALPHA Partial derivative of lateral force (Fy) with respect to slip angle at zero slip angle UMIN Coefficient of friction with full slip (slip ratio = 1) UMAX Coefficient of friction at zero slip Rolling resistance Rolling resistance coefficient Relax length X Relaxation length in X Relax length Y Relaxation length in Y Table A.0.1: Fiala tire model parameters In general the Fiala tire model is extremely simple and easy to use. But it is not suitable for sensitivity studies for several reasons. The main reason is that the representation of tire behaviour is extremely simplified. Important phenomena such as camber effects are not considered. The other reason is that not all the parameters have physical meaning. For example UMAX is a theoretical value and can not be obtained in reality because there is always some slip in the contact patch (i.e. zero slip ratio is practically impossible to achieve). Due to these shortcomings the use of Fiala tire model in full vehicle simulations for sensitivity studies is rather questionable. UniTire UniTire is a semi-physical unified non-linear steady-state tire model made by Guo [20]. It works based on fitting a mathematical function to the test data in order to obtain the tire resultant force. This resultant force can then be divided to lateral and longitudinal forces. These forces are dependent on their respective friction coefficients that are calculated by a revised version of Savkoor’s formula [27]. This formula is based on few friction parameters and slip velocities of the contact patch in lateral and longitudinal directions. This enables 38 UniTire to model the speed dependency of the tire forces and moments, as it is presented in [17]. In this paper, Guo argues that the effect of speed on the steady-state forces and moments generated by the tire are not negligible. These differences are due to the friction variation as a function of the sliding speed. However, the friction parameters describing the UniTire model are hard to obtain from conventional tire tests. The rest of the parameters are mathematical fits and do not have physical meaning. Also it is extremely hard to connect these parameters to the tire characteristics realized in the application survey for sensitivity studies. The current version of the model does not cover the transient tire behaviour. Therefore UniTire is not suitable for parametric sensitivity studies. FTire FTire (Flexible Ring Tire Model) is designed as a “2.5-dimensional” non-linear vibration model. Although mostly used for ride and comfort purposes, it can be used in handling simulations as well. This model is reviewed here since it is implemented in Adams/Car and it is an example of the complex theoretical tire models. FTire is considered as a discrete element model and is a compromise between the computationally heavy finite element models and the simple pure in-plane models [15]. In Ftire the belt is represented by a slim ring that can be displaced and bent in any direction relative to the rim. As shown in Figure A.0.1, the flexible belt is represented by a number (50 to 200) of segments elastically attached to the wheel rim and to each other. Each segment is consisted of a number of tread blocks that have radial, tangential and lateral non-linear stiffness and damping properties. The radial deflection of these tread elements is dependent on the road profile and orientation of the belt elements. Sliding coeffi- cient and sliding velocity on the ground determine the tangential and lateral deflections. The six components of the forces acting at the wheel center are calculated by integrating the forces in the elastic foundation of the belt. Figure A.0.1: FTire belt segment degrees of freedom: (a) Translation, (b) Torsion and (c) Lateral bending, [16] Because of this construction FTire can be used with both short and long wavelength obstacles cov- ering frequencies as high as 120 Hz. It can also cover complete stand-still without any model switch. And due to its high accuracy can be used in complex situations such as ABS braking on an uneven road. Here a summary of the parameters needed for a FTire model is presented: • Rolling circumference under normal running conditions • Rim diameter • Tread width • Mass of the tire • One out of: – Portion of tire mass that moves with the belt, OR – Tire radial stiffness • Increase of overall radial stiffness at high speed with respect to the radial stiffness at stand-still and the wheel speed at which this dynamic stiffening reaches half of its value. 39 • Natural frequencies and respective damping moduli of first, second and fourth vibration modes of the inflated but unloaded tire with fixed rim. • One out of: – Natural frequency of the fifth mode,OR – Belt in-plane bending stiffness • One out of: – Natural frequency of the sixth mode,OR – Belt out-of-plane bending stiffness • Profile height • Rubber height over steel belt for zero profile height • Tread rubber stiffness • Percentage of contact area with respect to overall footprint area • Tread rubber damping and elastic moduli • Moment of inertia of the rim and other rotating parts • Maximum and sliding friction coefficients A detailed description of these parameters can be found in [12]. As it can bee seen the number of parameters that are needed for FTire to run is relatively large. These parameters are obtained from special measurements and test procedures described in detail in [13]. As mentioned before FTire is designed mostly for ride and comfort applications. The FTire parameters describe the structural properties of the tire and are not directly related to its handling characteristics. As an example it is hard, if not impossible, to change these parameters to reach a certain level of lateral force. The reason for this is that most of these parameters are completely interrelated to each other. For instance changing the profile height could change the natural frequencies or other parameters for different vibration modes. FTire, although presented as a compromise between the simple in-plane models and finite element models, is still computation heavy and time consuming specifically if it is compared to empirical tire models. Therefore FTire is not suitable for the sensitivity studies in the handling area. 40 B Appendix: Adams/Tire module axis sys- tem 27Learning Adams/Tire About Axis Systems and Sign Conventions ISO Coordinate System Wheel heading along road ISO W:X This is not the same as the direction in which the wheel is traveling. If the tire reverses its direction, the axis system flips 180 degrees about the z'-axis. Direction to the left along the road ISO W:Y The direction to the left along ground as viewed from behind a forward rolling tire. Expressed as right-hand orthogonal to the definitions of x' and z' (such as y = Z x X). Z-coordinate ISO W:Z Perpendicular to the road in the neighborhood of the origin of the tire axis system in a positive (downward) direction. (If the road is flat and in the x-y plane, this is negative global z.) Variable name and abbreviation: Description: Figure B.0.1: ISO coordinate system ISO-C (TYDEX C) Axis System The TYDEX STI specifies the use of the ISO-C axis system for calculating translational and rotational velocities, and for outputting the force and torque at the tire hub. The properties of the ISO-C axis system are: • The origin of the ISO-C axis system lies at the wheel center. • The + x-axis is parallel to the road and lies in the wheel plane. • The + y-axis is normal to the wheel plane and, therefore, parallel to the wheel’s spin axis. • The + z-axis lies in the wheel plane and is perpendicular to x and y (such as z = x × y). ISO-W (TYDEX W) Contact-Patch Axis System The properties of the ISO-W (TYDEX W) axis system are: • The origin of the ISO-W contact-patch system lies in the local road plane at the tire contact point. • The + x-axis lies in the local road plane along the intersection of the wheel plane and the local road plane. • The + z-axis is perpendicular (normal) to the local road plane and points upward. • The + y-axis lies in the local road plane and is perpendicular to the + x-axis and + z-axis (such as y = z × x). 41 Adams/Tire About Axis Systems and Sign Conventions 24 TYDEX-C Axis System Used in Adams/Tire ISO-W (TYDEX W) Contact-Patch Axis System The properties of the ISO-W (TYDEX W) axis system are: • The origin of the ISO-W contact-patch system lies in the local road plane at the tire contact point. • The + x-axis lies in the local road plane along the intersection of the wheel plane and the local road plane. • The + z-axis is perpendicular (normal) to the local road plane and points upward. • The + y-axis lies in the local road plane and is perpendicular to the + x-axis and + z-axis (such as y = z x x). Figure B.0.2: TYDEX-C axis system used in Adams/Tire 25Learning Adams/Tire About Axis Systems and Sign Conventions TYDEX W-Axis System Used in Adams/Tire Road Reference Marker Axis System The road reference marker axis system is the underlying coordinate system that Adams/Tire uses internally. For example, the tire translational displacement and local road normal for a three-dimensional road are expressed in the axis system of the road reference marker. The properties of the reference marker axis system are: • The GFORCE reference marker defines the axis system. • The + z-axis points upward. About Tire Kinematic and Force Outputs Adams/Tire calculates the kinematic quantities of slip angle, inclination angle, and longitudinal slip. These are based on the location, orientation, and velocity of the tire relative to the road. In turn, Adams/Tire calculates the forces and moments of the tire using the tire kinematics as inputs to the tire mode you select. Sign Conventions for Tire Outputs The table below, Conventions for Naming Variables, and the figure, ISO Coordinate System, show the sign conventions for tire kinematic and force outputs. Figure B.0.3: TYDEX-W axis system used in Adams/Tire 42 References [1] Adams/Car. User Manual 2010. Version MD Adams 2010. MSC, 2010. [2] Adams/Tire. User Manual 2010. Version MD Adams 2010. MSC, 2010. [3] D. Ammon. “Vehicle dynamics analysis tasks and related tyre simulation challenges”. In: Vehicle System Dynamics 43:S1 (2005), pp. 30–47. [4] M. Baecker, A. Gallrein, and H. Haga. “A Tire Model for Very Large Tire Deformations and its Application in Very Severe Events”. In: J. Mater. Manuf. 3(1) (2010), pp. 142–151. [5] M. Baecker, A. Gallrein, and H. Haga. “Simulating Very Large Tire Deformations with CDTire”. In: J. Passeng. Cars - Mech. Syst. 2(1) (2009), pp. 765–771. [6] E. Bakker, L. Nyborg, and H. B. Pacejka. “Tire modelling for use in vehicle dynamics studies”. In: SAE (1988). [7] E. Bakker, H. B. Pacejka, and L. Lidner. “A new tire model with an application in vehicle dynamics studies”. In: SAE (1989). [8] J. Deur, J. Asgari, and D. Hrovat. “A 3D brush-type dynamic tire friction model”. In: Vehicle System Dynamics 42:S3 (2004), pp. 133–173. [9] J. Deur, J. Asgari, and D. Hrovat. “A dynamic tyre friction model for combined longitudinal and lateral motion”. In: 2001 ASME International Mechanical Engineering Congress and Exposition, Vol. 2. 2001. [10] C. C. deWit et al. “A new model for control of systems with friction”. In: IEEE Transactions On Automatic Control 40 (No. 3) (1995), pp. 419–425. [11] J. Duer. “Modeling and analysis of longitudinal tire dynamics based on the lugre friction model”. In: 3rd IFAC Workshop Advances in Automotive Control. 2001. [12] FTire - Flexible ring tire model. Modelization and parameter specification. Cosin Scientific Software. [13] FTire Parametrization. Proposed measurements and data processing procedures. Cosin Scientific Software. [14] G. Gim, Y. Choi, and S. Kim. “A semi-physical tire model for a vehicle dynamics analysis of handling and braking”. In: Vehicle System Dynamics: International Journal of Vehicle Mechanics and Mobility 45:S1 (2007), pp. 169–190. [15] M. Gipser. “FTire, a new fast tire model for ride comfort simulations”. In: (). [16] M. Gipser. “The FTire tire model family”. In: (). [17] K. Guo et al. “A study on speed-dependent tyre-road friction and its effect on the force and the moment”. In: Vehicle System Dynamics 43:S1 (2005), pp. 329–340. [18] W. Hirschberg, G. Rill, and H. Weinfurter. “User-Appropriate Tyre-Modelling for Vehicle Dynamics in Standard and Limit Situations”. In: Vehicle System Dynamics 38:S2 (2002), pp. 103–125. [19] W. Hirschberg, G. Rilly, and H. Weinfurterz. “Tyre Model TMeasy”. In: CCG Seminar TV 4.08 Tyre Models in Vehicle Dynamics: Theory and Application. 2010. [20] D. Lu and K. Guo. “UniTire steady state model: overview and applications”. In: 3rd international conference on advanced computer control (ICACC 2011). [21] MF-Tyre & MF-Swift. User Manual 2008. Version 6.1. TNO, 2008. [22] C. Oertel. RMOD-K 7 literature. •. url: http://www.rmod-k.com/index.php?opti