Published Project Reports are written primarily for the Customer rather than for a general audience and are published with the Customer s approval.

Transcription

1 TRL Limited PUBLISHED PROJECT REPORT PPR112 VERTEC FINAL PROJECT REPORT FOR THE DfT Version: 1. by M Dodd (TRL Limited) Prepared for: Project Record: VERTEC PPAD 9/33/118 Client: Department for Transport (DfT), Transport, Technology & Standards Division, TTS6 (Lawrence Thatcher) Copyright TRL Limited June 26 This report has been prepared for the Department for Transport (DfT), Transport, Technology & Standards Division (TTS6). The views expressed are those of the author(s) and not necessarily those of the DfT. Published Project Reports are written primarily for the Customer rather than for a general audience and are published with the Customer s approval. Project Manager Approvals Quality Reviewed

2 This report has been produced by TRL Limited, under/as part of a Contract placed by the DfT. Any views expressed are not necessarily those of the DfT. TRL is committed to optimising energy efficiency, reducing waste and promoting recycling and re-use. In support of these environmental goals, this report has been printed on recycled paper, comprising 1% postconsumer waste, manufactured using a TCF (totally chlorine free) process.

3 CONTENTS Executive summary i 1 Introduction 1 2 Outputs and objectives 2 3 Summary of work packages WP1: Definition of modelling environment and reference vehicles, components and manoeuvres WP2: Tyre-pavement interaction assessment and model implementation WP3: Reference tests with passenger cars and control systems WP4: reference tests on HGVs WP5: Development and validation of passenger car vehicle model WP6: Development and validation of HGV vehicle model and of driving simulator WP7: Ranking of the most dangerous situations and new guidelines for the design of safer products WP8: Exploitation and dissemination List of VERTEC deliverables 31 4 Conclusions 31 Acknowledgements 33 References 33 Appendix A. Tests - Outdoor vehicle passenger car Tyre Technology Expo Paper 35 TRL Limited PPR 112

4

5 Published Project Report Version: 1. Executive summary TRL was part of a consortium working on an EC 5 th Framework research project called VERTEC (VEhicle, Road, Tyre, and Electronic Control Systems Interaction). The aim of the project was to increase vehicle primary safety by developing a fully integrated model for the simulation of the road-tyre-vehicle-driver system in the most potentially dangerous situations. This model was also to be the base for the development of an upgraded driving simulator. Special focus was drawn to the most advanced Vehicle Electronic Control Systems and the representation of both passenger cars and Heavy Good Vehicles (HGVs). The main outputs of the project were the integrated simulation environment and an improved driving simulator for cars and trucks. The purpose of the outputs was to allow the partners of the consortium to detect and rank the most dangerous driving situations in order to define and supply guidelines for the design of safer roads, vehicles, tyres and electronic devices. Full scale reference tests using a passenger car and an HGV provided a database of experimental data to help design, develop and validate the best possible models for the vehicle, tyre, driver and road surface. Simulation environments for both a passenger car and an HGV were successfully created. For the passenger car, two modelling environments were developed. The first was a very accurate vehicle dynamic simulation using a multi-body approach with a co-simulation between ADAMS and Matlab/Simulink. The second model was a simplified real-time equivalent developed in Matlab/Simulink. The HGV vehicle model was built using the multi-body environments in ADAMS. The passenger car Simulink model was adapted and modified for real time use in the VTI driving simulator. It was planned to run the model with the vehicle dynamic stability (ABS, ASR & VSC) controller models developed in this project. However, the source code for these blocks could not be provided and so the simulator was run without the control subsystems. The simulator model was validated for a number of test subjects, with different levels of skill, performing a double lane change manoeuvre. The significant difference in performance found between inexperienced and experienced drivers indicated that future driver models should include a parameter to represent the skill level of the driver. An analysis of HGV accidents was used to identify the most dangerous situations for each specific road type. The accident data supplied by TRL used a method which combined data from the STATS 19 database as well as the England Trunk Road Database. This meant that it was possible to identify the exact section of the trunk road network where each accident occurred. This was a different method of analysis and its use should be considered in future projects that include accident analysis. TRL delivered the largest amount of data among the partners, including over 2, HGV accidents, most of which occurred on rural highways and primary roads. As well as providing a large evidence base that meant that the results of the analysis remained relevant to the UK. Problems with the development and validation of the models caused delays in delivering the working models and so the programmes of simulations to detect and rank the most dangerous driving situations were not completed by the official end date of the project (December 25). To date, the programme of simulations for the passenger car have been started but not yet completed. Approximately 25 simulations have been run and the results analysed to identify test configurations that resulted in a dangerous situation. As there were insufficient data from the simulations it was not possible to define detailed guidelines for the safer design of passenger cars and HGVs. Guidelines for the design of electronic control systems showed that the tyre-characteristic had the strongest influence on the overall vehicle performance. The application of the model for providing variable posted speed limits as a function of the actual handling condition seemed to be the most promising as the model could be implemented on the road side and could potentially provide a regular refreshed update of the posted limit. TRL Limited i PPR 112

6 Published Project Report Version: 1. This report has been prepared for the UK Department for Transport (DfT) to summarise the findings of the VERTEC project from its start (1 st December 22) until its completion date (3 th November 25). TRL Limited ii PPR 112

7 Published Project Report Version: 1. 1 Introduction TRL was part of a consortium working on an EC 5 th Framework research project called VERTEC (VEhicle, Road, Tyre, and Electronic Control Systems Interaction). The other partners involved in the project were: Pirelli Pneumatici SpA (Tyre manufacturer-italy. Project Co-ordinator) Nokian Tyres (Winter Tyre manufacturer-finland) Porsche (Vehicle manufacturer-germany) CRF (Fiat vehicle research centre-italy) TRW (Electronic control system manufacturer-germany) CETE (Road safety organisation-france) VTI (Road safety organisation-sweden) HUT (University-Finland) UNIFI (University-Italy) Volvo (Passenger Car and Heavy Good Vehicle manufacturer-sweden) The aim of the project was to increase vehicle primary safety by developing a fully integrated model for the simulation of the road-tyre-vehicle-driver system in the most potentially dangerous situations. This model was also to be the base for the development of an upgraded driving simulator. Special focus was drawn to the most advanced Vehicle Electronic Control Systems and the representation of both passenger cars and Heavy Goods Vehicles (HGVs). TRL s involvement in this project was co-funded by the EC (5%), the Department for Transport (25%), and the Highways Agency (25%). TRL involvement is summarised in Table 1. Table 1: TRL effort in VERTEC WP Task Name MM Effort Original Amended Project management 2 2 Definition of modelling environment and reference 1 vehicles, componenet and manoeuvres Reference tests with passenger car & control 3 systems Reference tests on HGVs 5 1 Development and validation of HGV model and Driving 6 Simulator 2 2 Ranking of the most dangerous situations and new 7 guidelines for design of safer products 4 7 Originally TRL was scheduled to assist in the reference tests on HGVs in WP4 by carrying out some of the tests and as well as helping to define the test specification and analysing the results. As it was not cost effective for Volvo to ship the HGV to TRL it was decided that TRL would only assist with the analysis of the results. The unused man-months (MM) were transferred to WP6 to assist VTI in the implementation of their existing HGV model into the Matlab-Simulink environment. Again not all of the man-months were used in this task so TRL shifted the remaining time into WP7 (Tk7.a) to assist in defining the simulations which were carried out in order to rank the most dangerous situations. In addition TRL was able to assist the other partners by carrying out some of the simulations in the Simulink environment. TRL Limited 1 PPR 112

8 Published Project Report Version: 1. This report has been prepared for the UK Department for Transport (DfT) to summarise the findings of the VERTEC project from its start (1 st December 22) until its completion date (3 th November 25). Some problems were encountered during the development and validation of the computer models which resulted in the project not being completed on time. At the time of writing this report project partners were still working to resolve the problems and to complete the programme of work. 2 Outputs and objectives The main outputs of the project were the following tools: Integrated Simulation Environment suitable for transport safety investigation as well as time-to-market reduction, allowing a detailed representation of cars and trucks, electronic control systems, human driver and tyre-road interaction. Improved Driving Simulator for cars and trucks using a full vehicle cabin with special motion systems reproducing the dynamics combined with screens reproducing the environment. Such a device can simulate in a very realistic way any driving task of both heavy and light vehicles in any environmental condition, thereby enabling safety investigations without any risk for the drivers. The exploitation of such tools would allow the partners of the consortium to achieve the following direct goals: Detection and Ranking of the most dangerous situations. In particular, such an investigation focuses on the worst combination of environment, weather, road conditions and the interaction between driver and electronic control systems in the case of both cars and trucks. Definition of Guidelines for improving the design of vehicles, roads (including their maintenance), tyres and electronic control systems from the point of view of primary safety. These guidelines are expected to contribute to the improvement of safety of all of the abovementioned products, launched on the market after two years from the end of the project. 3 Summary of work packages The project comprised of nine work packages as shown in Figure 1. WP (Project Management) and WP8 (Exploitation and Dissemination) are not shown in the diagram. WP2: Tyre/Pavement interaction assessment and model implementation WP1: Definition of modelling environment and reference vehicles, components and manoeuvres WP3: Reference tests with passenger cars and control systems WP5: Development and validation of passenger car vehicle model WP4: Reference tests on HGVs WP6: Development and validation of HGV vehicle model and of driving simulator WP7: Ranking of the most dangerous situations and new guidelines for the design of safer products Figure 1: VERTEC work packages TRL Limited 2 PPR 112

9 Published Project Report Version: WP1: Definition of modelling environment and reference vehicles, components and manoeuvres This work package was split into four sub-tasks. The first sub-task concerned the set-up and implementation of the VERTEC website ( to facilitate data exchange and communication between partners and for the exploitation of the project results. The reference manoeuvres and test conditions to be used in WP3 and WP4 were defined in the second sub-task. A Lancia Lybra 2.4 JTD, provided by CRF, was used as the reference passenger car (Figure 2). Figure 2: VERTEC reference car - Lancia Lybra 2.4 JTD Table 2 shows the minimum set of parameters recorded during each test. Table 2: Passenger car recorded parameters during reference tests Pure lateral Pure longitudinal Combined (brake in turn) 1 Steering wheel angle x x x 2 Lateral acceleration x x 3 Yaw rate x x 4 Side slip angle x x 5 Roll rate x x 6 Throttle x x x 7 Velocity x x x 8 Longitudinal acceleration x x x 9 Pitch rate x x 1 Front left wheel velocity x x x 11 Front right wheel velocity x x x 12 Rear left wheel velocity x x x 13 Rear right wheel velocity x x x 14 Front left pressure x x x 15 Front right pressure x x x 16 Rear left pressure x x x 17 Rear right pressure x x x 18 Pressure at pump x x 19 Brake pedal displacement x x 2 Brake pedal force x x number of channels The programme of reference tests was shared between the project partners and Table 3 shows the different manoeuvres that were carried out in WP3. TRL Limited 3 PPR 112

10 Published Project Report Version: 1. Table 3: Test matrix for reference tests with passenger car Conditions Manoeuvres Dry Wet Snow/Ice (1) (2) (1) (2) (1) (2) Steady state at constant radius (ISO 4138) X X X X X X Step steer (ISO 741) X X X X ISO lane change (ISO 3888) X X X X X X Braking in a curve (ISO 7975) X X X X X X Power off in a curve (5m) X X Random Steer Input (ISO 741 and ISO/TR 8726) X X Limit braking (V = 9km/h) X a X a X a X a X X Acceleration in a straight line X X X X X X Traction in a curve X X X X X X Change of µ in a curve (R=1m) X X Pedal force calibration X (1). Open/closed loop manoeuvre - active system OFF (2). Open loop manoeuvre - active system ON (except ISO Lane Change w hich is close d loop - active system ON) (a). tests done on different surfaces and on longitudinal/transverse split surfaces Tests were carried out using four different types of summer tyre, supplied by Pirelli, and three different types of winter tyre, one of which was a studded tyre, supplied by Nokian Tyres (Figure 1). Both new and worn tyres were tested. The worn passenger tyres had a tread depth of 3mm. Figure 3: Reference passenger car test tyres: Nokian WR 25/6R15, Pirelli P7 25/6R15 The HGV reference tests were carried out with a typical European 4t vehicle combination consisting of a Volvo FH12 4x2 tractor unit and a three-axle semi trailer loaded with iron blocks in load frames. For safety reasons the vehicle was equipped with out-riggers. The drive axle of the tractor had dual tyres and all other axles had singles, as shown in Figure 4. Figure 4: VERTEC reference HGV Volvo FH12 TRL Limited 4 PPR 112

11 Published Project Report Version: 1. The reference manoeuvres for the HGV included: Single lane change Entering curve with increasing curvature Obstacle avoidance Braking in a straight line (on dry and wet surfaces) Pirelli supplied the tyres for the reference manoeuvres as shown in Figure 5. Similarly to the passenger car, the tyres were tested in a new and worn condition. The worn HGV tyres had a groove depth of 6 mm (trailer 8 mm). Figure 5: Reference HGV test tyres: Pirelli FH55 315/8R22.5 (tractor front), ST35 315/8R22.5 (tractor rear), TH65 385/65R22.5 (trailer) The modelling environment was defined in the third sub-task. For the passenger car, two modelling environments were developed, as shown in Figure 6. Two models were required because of two counter-acting requirements. The first requirement was a very accurate vehicle dynamic simulation, even if a slow computational time was a consequence; a multi-body modelling approach with a cosimulation between ADAMS and Matlab/Simulink was chosen for this purpose. The second requirement was that the vehicle model must be used in a driver simulator, so it had to be hardwarein-the-loop capable, i.e. the model must be executed in real time; a simplified equivalent model was chosen for this purpose. For this purpose Pirelli developed a 14 DOF vehicle Simulink-model. The HGV vehicle model was built using the multi-body environments using some simplifications and assumptions related to flexible body. Details on the design of the model are described in section Note: Red blocks are subsystems developed in ADAMS. Blue blocks are subsystems developed in Simulink. Figure 6: VERTEC simulation environments TRL Limited 5 PPR 112

12 Published Project Report Version: 1. The final sub-task of this work package was an extensive literature review carried out among different countries to help the identification of potentially dangerous situations when HGVs are involved. The literature review database was made available to the Consortium through the VERTEC web site and contains a total of 72 papers. The literature review showed that accident data are characterised by many parameters and that different countries or institutes have their own way to collect data and organise accident data. Therefore it was agreed that the only way to complete an extensive HGV accident analysis involving data from many European countries, was to create a common database (VERTEC HGV Accident DB). The HGV accident data was collected from the UK, France, Finland and Italy using a common format worksheet to make the collected data as homogeneous as possible despite discrepancies in the information available in different databases. For the data analysis a set of key variables were identified (severity of the accident, road type, HGV type, road geometry, dry or wet pavement, accident mode etc) and the data was then split into different scenarios each of which was represented by a given road type and severity (all accidents, accidents with injuries or fatal and only fatal accidents). Table 4 shows the number of accidents available in the VERTEC database for each scenario. Table 4: Number of HGV accidents recorded in the VERTEC database Accident Type All Injuries & Fatal Fatal Fatal in USA Road Types Rural Highway Urban Highway Primary Road Secondary Road Slip-roads 7,566 accidents 233 accidents 498.2km Unknown length [Italy] (Italy) 1542 accidents 738 accidents 5,69 accidents 23 accidents 58 accidents km 214km 2957km 212 Unkown length [1,356 Italy] [183 France] [114 France] [France] [44Italy] [119 France] [555 UK] [5,495 UK] [1 France] [13,927 UK] [4 UK] 651 accidents 14 accidents 698 accidents 121 accidents 4 accidents 4,214km + Finland 214km 2,957km + Finland 212km + Finland Unknown length [13 Italy] [4 France] [2 France] [3 France] [Italy] [8 France] [1 UK] [451 UK] [118 Finland] [56 UK] 227 Finland] [7 Finland] 1,618 accidents 59 accidents 1,533 accidents 251 accidents [US FARS] [US FARS] [US FARS] [US FARS] - TRL extracted a subset of accidents from the STATS 19 database, each involving at least one HGV, which took place in the period 1994 to 21. Due to the nature of the information required for this task, it was necessary to establish a correlation between the STATS19 data and the England Trunk Road Database. In this way, it was possible to identify the exact section of the trunk road network where each accident featured in STATS 19. Consequently, information about the radius of curvature of the road and average traffic density could be related to each accident. This was a different method of analysis and its use should be considered in future projects that include accident analysis. TRL delivered the largest amount of data among the partners, including over 2, HGV accident cases, most of which occurred on rural highways and primary roads. As well as providing a large evidence base this meant that the results of the analysis remained relevant to the UK. Each different variable was analysed in terms of percentage distributions in a given scenario, and for each scenario the HGV accident rates were considered. The analysis of the influence of key variables on HGV accidents highlighted that tractor semi-trailer combinations were most frequently involved in accidents resulting in a fatality or serious injury. For each specific road type, the most dangerous situations were identified were: TRL Limited 6 PPR 112

13 Published Project Report Version: 1. For rural highways: a front to rear accident involving a tractor-semi trailer with another road user on a curve with a radius of 5m - 1m on dry or wet surfaces. For urban highways: a side-side collision involving a tractor-semi trailer and another road user in sharp curves (with radii lower than 5m) in dry condition. For primary roads: a front to side accident involving a tractor-semi trailer and another road user on wet surfaces. Although small radii bends seemed to be more critical, larger radii lead to higher probabilities of fatal events so it was recommended that the two conditions of R<3m and R>5m should both be investigated. For secondary roads: a front to side accident involving a single unit truck or a tractor semi-trailer with another user vehicle on dry surfaces on curves with radii greater than 3m. The comparison between different types of roads highlighted that accident rates based on HGV traffic (the number of accidents per million heavy-good vehicles km) seemed to be a more realistic indicator than the standard accident rate (the number of accidents per million vehicle km) because it gave a direct relation between accidents and roads without being influenced by the traffic composition. The comparison between different types of roads showed that primary roads, if compared with rural highways, have an accident rate (HGV) more than 3 times greater for accidents where someone was killed or seriously injured (KSI) and almost 5 times greater for fatal accidents, while if compared with urban highways they have an accident rate about 1.7 times greater for KSI accidents. 3.2 WP2: Tyre-pavement interaction assessment and model implementation In order to use the tyre model for simulation purposes a very accurate tyre characterisation was needed. Empirical tests on using the same type of tyres and in similar conditions were performed to gain accurate and reliable simulation results. Indoor tests were carried out on three different pieces of equipment, designed to test the tyres lateral and longitudinal behaviour under different wheel loads, wheel angles and road conditions (Figure 7). Figure 7: Cleat test machine (left), MTS flat track machine (middle), VTI tyre test facility (right) Outdoor tests on the passenger car tyres were also carried out using Nokian s Friction measurement vehicle and VTI s BV12 test vehicle (Figure 8). TRL Limited 7 PPR 112

14 Published Project Report Version: 1. Figure 8: Nokian friction measurement vehicle (left) & VTI s BV12 tyre test vehicle (right) The test method for outdoor HGV tyre tests was modified from the common Linked Vehicle Method (LVM) tests used by Nokian Tyres for development testing of heavy vehicle tyres. In the LVM tests, the test vehicle with test tyres was connected with a link and a force measurement device to a heavier ballast vehicle (Figure 9). Figure 9: LVM method for testing HGV tyres The results of different methods, the Pirelli Flat Track and VTI BV12 measurement vehicle, correlated quite well. Some examples, shown in Figure 1, illustrate that the shapes of the curves were quite close, although the highest longitudinal force generation of BV12 occurred at a smaller slip angle than for the Pirelli Flat Track. This may have been caused by the differences in surface roughness. Figure 1: Flat Track vs. BV12 (left) & lateral force v. slip percentage (right) The results also showed that the friction of the ice and snow surfaces varied quite significantly from day to day, mainly due to environmental conditions such as sunshine, wind and temperature. Examples of this are shown in Figure 11. TRL Limited 8 PPR 112

15 Published Project Report Version: 1. Figure 11: Examples of day to day variation in friction due to environmental conditions On winter surfaces, the vehicle speed was not found to have a major influence on friction behaviour of the tyre. Therefore the tyre data obtained from the low speed test could be used in the modelling and analysis of high-speed vehicle manoeuvres. The passenger car tyres on ice surface produced the highest forces at very low slip; typically the peak lateral force on ice was between 2 and 5 of slip angle. On the high friction surfaces the peak forces were generated at higher wheel slip. For the tyres tested on asphalt the typical slip angle for highest lateral force was around 6. The longitudinal force maximum was achieved at 1% wheel slip. The HGV tyres were tested both on low friction (ice) and high friction (steel bar) surfaces using the VTI tyre test facility. In Figure 12 an example of the longitudinal braking characteristics of the tyres is shown. It is clear that worn tyres generally exhibit higher friction values than the new ones. This increase of the friction could stem from the artificially worn tyre treads resulting in greater ice adhesion, compared to normally worn treads. It was also apparent that the locked wheel friction level was quite a lot lower than the peak friction. This was due to water build-up between tyre and ice during braking. Uneven artificial wear caused some of the worn tyres to exhibit a lateral friction force at zero slip angles. Such nonlinear behaviour for very small slip angles was a result of the relatively fast steering process during the test, which induced relaxation effects in the data. VTI tyre test facitlity: Braking on rough ice (-5 deg C) 8kN VTI tyre test facitlity: Cornering on smooth ice (-5 deg C) 2kN FH55 New TH65 New ST35 New FH55 Worn TH65 Worn ST35 Worn Longitudinal Friction coeff FH55 New TH65 New ST35 New FH55 Worn TH65 Worn ST35 Worn Lateral Friction coeff Slip % Slip angle [deg] Figure 12: Example HGV tyre characteristics, longitudinal braking (left) & cornering (right) The highest forces produced by the HGV tyres on icy surfaces occurred at smaller slip angles than the passenger car tyres. Typically the slip angle at the maximum lateral force was around 2 on ice surface. For the high friction tests of HGV tyres, the lateral force peak was around 1, and the longitudinal force peak was between 1% and 2% of wheel slip. TRL Limited 9 PPR 112

16 Published Project Report Version: 1. The chosen tyre/road interaction model was the widely used Pacejka s Magic Formulae Tyre (MF- Tyre) model (Delft Tyre, 21). It consists of a set of mathematical formulae which reproduce the steady-state tyre behaviour by relating tyre/road contact forces (Fx, Fy) and moments (Mz, My, Mx) to longitudinal slip ratio (M), lateral slip angle (N), inclination angle (O) and vertical load (Fz). MF-Tyre empirical formulae are based on coefficients which are valid for a particular tyre/road interaction. Therefore the set of coefficients had to be identified from curve fitting of the experimental data, an example is shown in Figure 13. Both the steady-state condition (stationary or slowly changing tyre slip, vertical load and camber angle) and transient condition (fast changing of tyre slip, vertical load or camber angle) were characterised. Figure 13: Curve fitting for pure lateral slip for different vertical loads Once the MF-Tyre model was integrated in the vehicle model (in WP5), simulation of ABS braking manoeuvres showed unexpected large wheel oscillations, caused by the tyre transient model overestimating the relaxation length at non-zero slip. By introducing a slip decreasing relaxation length coefficient into the vehicle model, the ABS braking simulation was improved as shown in Figure Vehicle speed FL Wheel speed 35 3 Vehicle speed FL Wheel speed [m/s] 15 [m/s] [s] [s] Figure 14: Comparison of ABS braking manoeuvre with and without the transient model TRL Limited 1 PPR 112

17 Published Project Report Version: WP3: Reference tests with passenger cars and control systems The Lancia Lybra was tested in the conditions described in section 3.1 (WP1). This included reference tests on dry, wet, snowy and icy surfaces. The purpose of these reference tests was to provide a database of experimental data to help design, develop and validate the best possible models for the vehicle, tyre, driver and road surface. TRL completed a programme of full vehicle reference tests in dry and wet conditions on the TRL Research Track. TRL also performed analysis of this data and, as sub task leader for Tk3.b, was responsible for the completion of deliverable R3.2 (Reference tests on wet, snowy and icy road). The steady-state constant radius tests showed that the vehicle approached saturation of the front axle at a lower lateral acceleration on wet stone mastic asphalt (SMA) surface than in the dry. This was because the coefficient of friction for the wet surface is lower than that for the dry. On snow and ice the limit of lateral acceleration was very low and was reached suddenly. There was no significant difference between ESP-on and ESP-off tests because the ESP was usually active after the vehicle had started sliding and was already in an un-recoverable spin. In terms of vehicle dynamics, the ESP showed its ability in reducing the side slip angle and consequently the rear axle lateral slip during the step steer manoeuvre. The ESP was able to influence the yaw rate oscillations, helping in stabilising them more easily, as shown in Figure Step Steer - VEL = 1km/h - SWA = 11deg Normalised Value Time (sec) Steering Wheel Angle (SWA/12) [deg] - ESP Off Slip Angle (SA/2) [deg] - ESP Off Yaw Rate (YR/3) [deg/s] - ESP On Yaw Rate (YR/3) [deg/s] - ESP Off Steering Wheel Angle (SWA/12) [deg] - ESP On Slip Angle (SA/2) [deg] - ESP On Figure 15: Step Steer Manoeuvre at 1 km/h ESP-on and ESP-off For the ISO lane change manoeuvre, the ESP allowed the driver to steer later and with lower angles, resulting in a smaller side slip angle and yaw rate as shown in Figure 16. TRL Limited 11 PPR 112

18 Published Project Report Version: 1. SteeringWheel Angle [deg] Yaw Rate [deg/s] Steering Wheel Angle (SWA) [deg] Yaw Rate (YR) [deg] Hydraulic Brake Pressure [bar] Front Left Caliper (PRSX) Front Right Caliper (PFDX) Rear Left Caliper (PRSX) Rear Right Caliper (PRDX) Time (sec) Figure 16: ESP Intervention ISO Lane Change Manoeuvre at 9 km/h For the brake in a turn tests the ratio between the maximum value and initial value of the yaw rate and side slip angle were plotted against the mean value of the longitudinal deceleration during the braking manoeuvre. The analysis did not reveal any significant difference between the ESP-on and ESPoff configurations. Brake in a turn tests were also carried out in a split-µ condition. CRF simulated this condition using a water controlled depth pool. The results shown in Figure 17 show how the ESP configuration was easier to control and consequently how the manoeuvres were more repeatable. Figure 17: Comparison between ESP-on and ESP-off - Split-µ brake in a turn The random steer input manoeuvres were carried out to investigate the vehicles response to the varying frequency of the steering wheel angle actuation. At higher lateral accelerations the intervention of the ESP had the effect of reducing the resonant peak of the yaw rate/steering wheel angle transfer functions (Figure 18) TRL Limited 12 PPR 112

19 Published Project Report Version: 1. Figure 18: Transfer function - yaw rate/steering wheel angle vs. frequency (Hz) Straight line braking tests were carried out in wet and dry conditions on a very thin asphalt concrete (VTAC) surface, a polished cement concrete (PCC) surface and on an asphalt concrete (AC) surface. The results from these surfaces showed that the stopping distances for the new tyres were between 2% and 22% lower than the corresponding distance with the worn tyres. The benefit of studded tyres were seen from the tests on rough ice as the tyre had good friction even with higher slip values, making it easier for the ABS to control the wheel speed. Acceleration tests were carried out on low friction surfaces (snow and ice) with full throttle in first gear with the ESP switched on. With the ASR switched on the system limited the speed of the engine so that the speed difference to the rear tyres did not get too large. With the ASR switched off, the engine ran at maximum along with the front wheels, as shown by the increase in wheel slip in Figure 19. In theory the test vehicle should have a greater acceleration with the ASR switched on. However, the tests with the ASR switched on gave similar results to the tests with the ASR switched off. One explanation for this could be that snow got packed into the grooves of the tyres which reduced the friction. In ASR-off, the tyres were rotating very fast, which may have cleaned snow out from tyre grooves. 2 Wheel Slip [%] Front Left Wheel (FSX) [%] Front Right Wheel (FDX) [%] Speed [km/h] Hydraulic Pressure [bar] Vehicle Speed (VEL) [km/h] 2 Front Left Wheel (VFSX) [km/h] Front Right Wheel (VFDX) [km/h] 1 Rear Left Wheel (VRSX) [km/h] Rear Right Wheel (VRDX) [km/h] Master Cylinder (MCP) [bar] 1 Front Left Caliper (PFSX) [bar] Front Right Caliper (PFDX) [bar] 5 Rear Left Caliper (PRSX) [bar] Rear Right Caliper (PRDX) [bar] Time (Sec) Figure 19: Acceleration in a straight line on low friction surface ASR-off TRL Limited 13 PPR 112

20 Published Project Report Version: 1. All of the data collected in WP3 were compiled into a common format and saved into a database of results so that they could be easily accessed and used during the validation of the computer models. 3.4 WP4: reference tests on HGVs Like the passenger car test, reference tests on dry, wet, snowy and icy surfaces were carried out using the reference HGV described in section 3.1. The purpose of these reference tests was again to provide a database of experimental data to help design, develop and validate the best possible models for the vehicle, tyre, driver and road surface. As described in section 1, TRL performed analysis of some of the data from the HGV reference tests and contributed to the deliverable for subtask Tk4.b. Steady-state characterisation tests were carried out using a constant radius test according to ISO (23). The speed was incrementally increased up to the point of rollover. From the data the slip angle gradients of all the axles were calculated. As shown in Figure 2, the gradients all increased with acceleration, particularly the front axle, which explains the strongly increasing under steer of the vehicle for increasing lateral acceleration. Figure 2: Slip angle gradients of reference HGV The roll angle behaviour of the chassis to the axles and of the trailer to the ground was found to be linear, and the frame torsion between the axles was estimated to be.7deg/g. The transient response of the HGV was found to be non-linear, especially at low lateral accelerations. The response of the vehicle was characterised using random steer tests, single sine-wave tests, continuous sinusoidal tests, and swept sine steer tests. A comparison between the results of the random steer input tests with new and worn tyres (5mm of tread) showed a significant change to the transient behaviour of the vehicle. Using the worn tyres resulted in a larger bandwidth of the frequency response which gave a more constant gain in a larger frequency range (Figure 21). TRL Limited 14 PPR 112

21 Published Project Report Version: 1. Figure 21: Frequency response of random steer tests with new (left) and worn (right) tyres The steady state characterisation tests identified the rollover threshold of the reference HGV to be at 4.2m/s² on a 45m radius circle. This result correlated well with the tilt-table test that was also carried out. In this test the tilt angle was determined to be 23. This corresponds to a lateral acceleration of 4.2m/s². The reference manoeuvres were carried out to investigate the behaviour of the vehicle in realistic closed-loop manoeuvres with respect to lateral and roll stability and to longitudinal performance showed that the dynamic rollover limit was higher than the static one in all the tested manoeuvres. The single lane change manoeuvre was found to require a steering input very similar to the single sine steer test, except for the steering corrections after the first period. Interestingly Figure 22 shows that although rollover did not occur in any of the single lane changes, the static rollover threshold was exceeded. Figure 22: Single lane change manoeuvre The obstacle avoidance tests did result in a rollover in the second turn when passing an obstacle offset by 4m from the path (Figure 23, left), however when the obstacle was offset by 3m (Figure 23, right) the vehicle did not rollover even though the static rollover threshold had been exceeded. TRL Limited 15 PPR 112

22 Published Project Report Version: 1. Figure 23: Obstacle avoidance tests, with rollover (left) and without rollover (right) The reducing radius test, where the vehicle entered a curve with a linearly increasing curvature showed that rollover occurred at approximately 5m/s², well above the static rollover limit. The deliverables did not offer an explanation as to why rollover occurred at a higher lateral acceleration. Straight line braking tests were carried out on high friction asphalt in wet and dry conditions. The results showed that there was a 9% increase in stopping distance and a 12% reduction in mean deceleration for the tests on a wet surface when compared with the results on the dry track. Tests showed an equal braking performance with new and worn tyres. A straight line braking test was also carried out with the trailer brakes disconnected. The results showed an increase of 85% in stopping distance and a reduction of 55% in the mean deceleration of the vehicle. This highlights the impact of incompatibility between a tractor and trailer. From an initial speed of 7km/h, with the tractor and trailer brakes working, the vehicle came to a stop in 36.7m. Without the trailer brakes operational the vehicle would still have been travelling at 45km/h (28mile/h) at this point. The results of the straight line braking tests on a low friction surface showed significant variability as a result of friction changes on the icy surface from one day to the next. It was recommended that such tyre characterisation tests should be completed in a single day to minimise the variation or that the friction should be measured in direct comparison to the test. 3.5 WP5: Development and validation of passenger car vehicle model The objective of the first sub task in this work package was to develop the subsystem model, namely: Vehicle Brake system & electronic control system Tyre/road interaction Road properties Track & driver The reference model was designed in ADAMS-Car using the standard components to build up the car s subsystems. The dynamic response of the virtual vehicle was compared with experimental test data. As shown in Figure 24, the passive vehicle showed a good correlation between simulation (blue) and experiment (red). TRL Limited 16 PPR 112

23 Published Project Report Version: 1. Figure 24: Comparison between passive car model and experimental data After the vehicle model was set up in the ADAMS-environment, the parameters were transferred to the real-time-capable vehicle model in VDSIM. To verify this model a validation against the ADAMS model was performed. Figure 25 shows results for a typical manoeuvre (ISO Lane Change at 8 km/h) and the good correlation between the simulation models. Figure 25: ADAMS VDSIM vehicle model validation The base-line parameters for the brake and electronic control system models were chosen according to the geometry of the components (booster size/ valve diameters in the HCU/ calliper size/ brake rotor diameter). The basic functionalities of the hydraulic unit with integrated ECU (ABS/TC/VSC/EBD) were checked by feeding in open loop signals to the control valves (similar to a specified release test TRL Limited 17 PPR 112

24 Published Project Report Version: 1. on a hydraulic rig). Figure 26 shows the good correlation between the simulation results (blue lines) and the test rig measurements (green lines). Figure 26: Brake system and electronic control system model validation The main work in tyre modelling was concentrated on the widely used Pacejka-approach. After setting up a model following the MF-Pacejka equations the tyre-test-results were analysed to identify the adequate MF-model-parameters. The curve fitting of the stationary tyre characteristics and the parameter identification turned out well. Unfortunately it was not possible to get numerically stable results whenever an ABS/ECS-intervention was active; therefore the start of the validation phase was postponed until the system was running in stable conditions. Analysis with different tyre models showed that the instabilities were created mainly by the numerical procedure used to simulate the tyre relaxation. The long relaxation time of the tyre made the communication between the tyre friction forces and the brake friction forces so slow that the wheels were able to speed up to higher velocities than the car. A solution was implemented by switching off the transient part in the Pacejka-tool and activating a new relaxation calculation (Figure 27). Figure 27: Improvement of tyre model during ABS interventions The road-property sub-model, already available for the ADAMS-environment from the preceding VERT-project, was transferred to the Simulink-environment. Some minor modifications have been performed to increase the flexibility of the model with respect to a more general track description. The VERTEC driver model was developed translating Mitschke s transfer functions (Mitschke 199) from frequency space into real space, e.g. a hold-back time in the transfer function is represented as a preview time of the driver. The parameters for the model were either directly transferred from Mitschke or they were evaluated using the results of the VERT-project. The implementation of the driver/track-model into the VDSIM environment was delayed due to the problems with the tyre model described above and so this task did not start until the middle of 25. It took nearly six months of work to stabilise the interface, i.e. the information exchange between driver and track, and by the official end date of the project the model was able to perform only some closed loop manoeuvres. TRL Limited 18 PPR 112

25 Published Project Report Version: 1. In parallel the driver model was tested using available manoeuvres and on artificial tracks. The model was improved and a global set of parameters was evaluated reproducing the main results of the preceding VERT-project. Some of the open-loop manoeuvres had to be carried out in a partially closed loop manner. For example in an open loop steady state cornering manoeuvre, the steering wheel angle for a test would ideally be prescribed and remain constant throughout the test. However, it was proposed that after the vehicle had reached the desired speed, the driver model should try to keep the lane without any braking intervention (i.e. a steering only driver). Due to the delays in developing, integrating and validating the model this solution was necessary as there was not enough time to carry out multiple simulations for each configuration to determine the required steering wheel angle for a particular speed and radius of bend. The aim of the second sub-task was to merge all the sub-system models developed in the task Tk5.a, with the tyre model of task Tk2.b in order to assemble a full passenger car model. The reference vehicle model of the passenger car was made available in ADAMS-Car. The model was able to simulate open loop manoeuvres in conditions of constant friction with the active system switched off, as well as variable conditions of friction with the active system switched on using co-simulation technique. A standard post processing was built to analyse data directly inside the Matlab-Simulink environment. The suspension was validated in the case of a parallel vertical wheel travel input. The experimental data showed a hysteretic behaviour, due to the real behaviour of the rubber elements. This was not reproduced by the multi-body suspension model; nevertheless the comparison between experimental and simulated curves (Figure 33) showed that they had a comparable slope across a wide input range, suggesting that the suspension model was good enough to be used in the whole vehicle model. Figure 28: Parallel wheel travel on front suspension The first stage of the full vehicle model validation concerned the behaviour of the model in a passive state, i.e. with the control systems switched off. Some of the results, for example the step steer tests on a high friction surface shown in Figure 29, showed a good fit between the simulation (blue line) and the experimental data (red line) in both the initial transient phase and in the following steady-state phase. TRL Limited 19 PPR 112

26 Published Project Report Version: 1. Figure 29: Step steer manoeuvre passive car on a high friction surface A swept sine steer manoeuvre was also used to study the vehicle dynamic behaviour. Figure 3 shows an acceptable comparison with respect to the experimental data. The only noticeable difference was the slightly higher resonance of yaw rate and side slip angle versus steering wheel angle Gain Yaw Gain rate PSIP-DVOL St. [1/s] Wh. Angle Gain BETA-DVOL [ ] Gain Sideslip Angle St. Wh. Angle Figure 3: Swept sine steer manoeuvre on a high friction surface The second stage of the full vehicle validation concerned the behaviour of the vehicle with the control systems switched on. The control systems fitted on the reference vehicle were the anti-lock brake system (ABS), the vehicle dynamic stability control (VSC) and the traction control system (ASR). The behaviour of the ABS was checked using a straight line braking manoeuvre. The results showed that the front and rear brake pressures in the model had a different distribution to the experimental data (Figure 31). Figure 31: Comparison of front and rear brake pressures limit braking high friction Similar tests on a low friction (ice) surface showed negligible differences between the experimental data and the simulation. There was a difference in the longitudinal deceleration however this was caused by an initial offset in the experimental signal. TRL Limited 2 PPR 112

27 Published Project Report Version: 1. The behaviour of the ASR was checked using a power-on manoeuvre. After a 6% increase in the tyre parameter LMUX, the simulation showed a good correlation with the experimental result showing correct intervention of the ASR in the model (Figure 32). Figure 32: Power on manoeuvre before (left) and after (right) modification to LMUX parameter A step steer manoeuvre on high friction was used to check the VSC control system. The model behaviour was found to be more stable compared to the real vehicle. One potential solution was to make the simulation more similar to the experimental data by modifying the balance of the axles at the limit, specifically to reduce the overall adherence of the rear axle with respect to the front one, as shown in Figure 33. However, since the validation of the suspension and the tyre fitting of the experimental data were satisfactory it was hard to find a genuine physical reason for such a modification of the rear axle. Therefore it was regarded to be more theoretically correct to keep the original identified vehicle data. Figure 33: Effect of modifying adherence of rear axle Finally, the brake in a turn manoeuvre was considered to assess the behaviour with the ABS and VSC. The model behaved reasonably compared with the experimental data, although the increased stability of the model resulted in lower wheel slip as shown in Figure 34. Figure 34: Front wheel velocities during brake in a turn manoeuvre TRL Limited 21 PPR 112

28 Published Project Report Version: 1. The increased stability of the simulations meant that if the vehicle over-steered in a simulation then it could confidently be predicted to do the same in real life. However, if the simulation showed no signs of instability then there would be a degree of uncertainty as to whether or not it would become unstable on the road. 3.6 WP6: Development and validation of HGV vehicle model and of driving simulator Tk6.a: Development of HGV vehicle model The HGV vehicle model was built using multi-body environments and some simplifications and assumptions in the following way. The frame was divided into two rigid longitudinal arm links with two rigid transversal arms linked with spherical joint and bushings in order to have a warp degree of freedom. The cabin was considered to be suspended and the mass inertia, front and rear kinematics suspension were modelled. The kinematics of the front suspension was modelled assuming the bushings were as rigid as possible. The leaf spring was modelled with several parallel standard beams in order to have the possibility of an asymmetric moment of inertia. The anti roll bar was modelled with standard beam elements. The rear suspension kinematics was developed considering bushings as rigid as possible. The air springs have linear characteristics. For the steering system a rigid model was developed with no booster and two concentrated springs to take into account steering elasticity. The engine was modelled with mass and inertia and was attached to the frame with bushings. The driveline was modelled as a conceptual driveline with rotational inertia, final drive ratio and gear box ratio. The brake system was modelled in terms of disc radius, calliper piston radius and friction coefficient in order to receive a pressure value on each calliper from the ABS and VSC Matlab- Simulink model. The trailer frame was modelled in two parts with the front and rear connected with a revolute joint. A torsion spring was modelled between the two parts of the trailer body in order to have a torsion degree of freedom. The kinematics of the trailer suspension was modelled considering bushings as rigid as possible except the one for the connection between the longitudinal arm and frame. The air spring has linear characteristics depending on the static load. The connection between tractor and trailer, the fifth wheel and the frame was modelled through two transversal bushings in order to have quasi-free pitch movement of the fifth wheel. The connection between the trailer and the tractor was modelled through a non linear bushing to simulate the free play at the connection. Tyre fitting was needed in order to identify the correct tyre characteristics from the experimental data measured on test bench. Only dry conditions were considered for this exercise as the level of acceleration and the full weight condition were not expected to be significantly different in wet conditions. Figure 35 shows that the fitting was more successful on high friction (left) than on low friction (right) although both conditions gave satisfactory results. Figure 35: Tyre fitting on high (left) and low friction surfaces (right) TRL Limited 22 PPR 112

29 Published Project Report Version: 1. The main interest of the validation was the full vehicle lateral behaviour. The MSC.ADAMS driver was designed for passenger cars, not car and trailer combinations, therefore only open-loop tests were planned. If the correct vehicle mass was used as a driver parameter then the simulated driver was not able to complete the tests because the driver did not turn steering wheel angle enough and so the vehicle cut the corner of its intended path. If the vehicle mass parameter was reduced then the driver was able to stay inside the single lane change track but the driver was still unable to complete the obstacle avoidance test with those parameters. Several parameter combinations were tried but the driver was not able to succeed, because of the trailer behaviour. Figure 36 shows an example of the level of validation obtained in the lateral dynamics tests on high friction. Figure 36: Single lane change manoeuvre on a high friction surface Figure 37 compared the experimental test data (red) of an open-loop single lane change manoeuvre on low friction with the simulation data on rough (blue) and smooth ice (black). The steering wheel angle from the experimental test data was used as an input for the simulation. The oscillation of the lateral acceleration was from the tractor cabin as this is where the accelerations and yaw rate were measured from. Figure 37: Comparison of single lane change manoeuvre on low friction The lateral acceleration correlation was good, but the differences in slip angle and yaw rate caused deviation in the path driven by the model. Figure 38 shows the path of closed-loop and open-loop simulations. The blue and red curves show that the open-loop simulation model did not complete the single lane change track. At the beginning the vehicle is following the right path, but then suddenly turns to another direction. That is why it was decided to do closed loop simulations. TRL Limited 23 PPR 112

30 Published Project Report Version: 1. Figure 38: Path of tractor body during single lane change manoeuvre The simulation model with rough ice tyre parameters showed a very good correlation with the experimental test data, except for tractor side slip angle that was consistently greater in the model. The VTI driving simulator (Figure 39) is a moving base system with a linear motion unit driven by a steel belt instead of a chain. The passenger car Simulink model was adapted and modified for real time use in the VTI simulator. It was planned to run the model with the vehicle dynamic stability (ABS, ASR & VSC) controller models developed by TRW however as TRW were unable to provide the source code for these block, VTI was not able to create the real time application. Therefore the simulator was run without the control subsystems. Figure 39: VTI driving simulator Overall the model worked as expected although there were some stability problems at low velocities (Figure 4). These were caused by the way the wheel rotational velocities and transient longitudinal slip were calculated resulting in oscillating tyre forces. The problem was overcome by using a different method to calculate the velocities and slip values. TRL Limited 24 PPR 112

31 Published Project Report Version: 1. Figure 4: Oscillations in lateral acceleration when starting from zero speed The simulator model was validated using a number of test subjects, with different levels of skill, to perform a double lane change manoeuvre. Figure 41 shows how an inexperienced driver (blue) lost control of the car, which resulted in a spin and hitting three cones. Such a difference in performance clearly indicated that future driver models should include a parameter to represent the skill level of the driver. Nm 5 fp1 fp5 Steering wheel torque m/s distance [m] 5 fp1 fp5 Lateral acceleration distance [m] 5 Lateral position hits()fp: 5(3): m distance [m] Figure 41: Double lane change manoeuvre performed in VTI driving simulator The structure and code of the original VTI coded HGV model was kept unchanged as far as possible to avoid errors when converting to Matlab-Simulink. The tyre forces and wheel dynamic calculations were newly developed S-function blocks added to the original model. The implemented tyre model was updated to Magic Formula version 5.2 with the exception of the transient slip equations, which were not used for the wheel dynamics calculations (wheel rotational speeds) to avoid problems especially at low speeds. TRL Limited 25 PPR 112

32 Published Project Report Version: 1. The transient response shown in Figure 42 is for one sinusoidal period with.5 Hz on the steering wheel. The validation using the simulator model shows good similarity between the experimental test data (left) and simulation (right) especially considering the fact at the time the tyre data was uncertain and had to be approximated from past experience. Figure 42: single lane change manoeuvre for tractor and semi-trailer combination 3.7 WP7: Ranking of the most dangerous situations and new guidelines for the design of safer products. The first sub-task had two objectives; to define the criteria for a dangerous situation and then through a programme of simulations to determine the most dangerous situations. The simulations with HGVs were carried out in ADAM'S and the simulations with the Lancia Lybra were carried out in the VDSIM environment (Matlab Simulink model). TRL was heavily involved in defining the matrices of simulations for the passenger car and HGV models. TRL assisted in identifying the problems with the VDSIM model. When defining a test matrix of simulations it was agreed to carry out the simulations in a number of stages. The first stage was a small selection of simulations (approximately 25) using only one of each of the supplementary parameters (rainfall, crossfall, macrotexture and slope) at a small number of speeds, radii, friction coefficients (SFC) and section combinations. The purpose of the Stage I simulations was to identify the areas in which the dangerous situations were most likely to occur for each of the manoeuvres. From the statistical study completed in task Tk1.b, which detected eight scenarios of accidents depending on the type of road and severity (fatal or injured), three manoeuvres were proposed for Stage 1: Negotiating a curve Braking in straight line Avoidance manoeuvre in straight line It was agreed to complete the Stage I simulations with the ABS system activated and the VSC disabled. This was to highlight which tests were "safe/unsafe" for the standard car so that only the "unsafe" conditions were investigated in the subsequent stages. This approach would also allow the level of intervention of the ESP system to be identified. Table 5 details the five criteria used to analyse the results of the Stage I simulations and to determine which configurations were likely to result in a dangerous situation. The first criterion was linked to the vehicle s position on the road. For each simulation, the position of the vehicle was tested to determine if the vehicle deviated out of its lane during the manoeuvre. The second criterion was used to evaluate the difficulty of following an ideal trajectory. For each simulation, the ideal lateral acceleration (V²/R with V: speed of the vehicle and R: radius of the bend) was compared with the TRL Limited 26 PPR 112