FRICTION POTENTIAL AND SAFETY : PREDICTION OF

Transcription

1 BRITE EURAM PROJECT BRPR CT ND INTERNATIONAL COLLOQUIUM ON VEHICLE TYRE ROAD INTERACTION FRICTION POTENTIAL AND SAFETY : PREDICTION OF HANDLING BEHAVIOR FLORENCE, FEBRUARY 23 rd 2001 Title : Correlations between trailer/vehicle braking tests data and testing conditions Authors: Diego Donadio, Diego Speziari - Pirelli Tyres ABSTRACT An analysis of correlations showing the effects of the tyre tread compound on the performance of the tyre in braking has been conducted, focusing on specific parameters and matching their ranking with the results of over 800 trailer tests. In this context, the variation of performances with the imposed testing conditions have been verified. All the presented data have been provided within the context of Vert European research project. Florence February 23 rd 2001 Paper pag. 1

2 INTRODUCTION A huge amount of braking tests on trailers has been performed within the VERT European project. In some cases these tests represented the 100% longitudinal part of a more complete test, including also lateral and combined characterisation of the tire. However, only longitudinal behaviours have been taken in account in this work. Correlations have been traced between the results of the tests and the three subsequent groups of data: - A set of structural lumped parameters of the tire; - A set of laboratory properties of the rubber compound of the tire; - Test conditions In this paper, correlations between the performance indexes and the third group of variables ( testing conditions ) are presented. Moreover, CETE has performed several braking tests on a Xantia, with and without the activation of ABS system: they have been taken in account also. A general scheme of the situation is represented below. Florence February 23 rd 2001 Paper pag. 2

3 AVAILABLE TEST DATA Each one of the tests is characterised by several factors, as follows: - The testing device (depending on the partner who made each test). - The surface nature. The surfaces on which the tests have been carried out are: - Dry asphalt - Wet asphalt - Snow - Ice However, each one of the general categories above includes two of more different particular types of ground surface. The difference is often valuable in the resulting data, since they depend, for example, on the asphalt texture or the snow compactness. - The tire characteristics, particulary: - The tire general size and type (including manufacturer, summer/winter, etc.) - The tire tread depth - For the Tread Compound group, the compound of the tread layer. - The water film thickness misured on the ground. - The speed (generally included between 50 ad 100 Km/h). - The vertical load on the tire. It has been considered as an internal parameter of each test, assuming more than one value; i.e., each test has been conducted with different loads (tipically, three) in order to define also the dependence of the tire behaviour on the normal force. In this work, all the performance indexes are normalized on a 4000N vertical load. Other parameters that have not been used to define and identify the test but, generally, could have an influence are: - Inflating pressure. It is strictly related to the tire type, and has not been considered as test condition variable; although a narrow number of tests have been performed with an inflating pressure slightly different from the reference value. - Air temperature and humidity. These data have been recorded during the tests, but not yet statistically compared with the testing outputs (in some cases of incoherence, they have been considered, but without reaching any satisfying explanations). The next tables show in more detail the combinations of the testing conditions. Florence February 23 rd 2001 Paper pag. 3

4 Activity of each partner Type general types and distribution to partners Florence February 23 rd 2001 Paper pag. 4

5 Tyre Tyre Code Tread Compound code 175 / 65 R 14 Summer 175 S AD 175 / 65 R 14 Winter 175 W / 65 R 15 Summer 185 S AA 185 / 65 R 15 Winter 185 W WN (100Hz only) 195 / 65 R 15 Summer 195 S AA 195 / 65 R 15 Winter 195 W / 45 R 17 Summer 225 S / 45 R 17 Winter 225 W WN (100Hz only) AIPC (smooth) AIPC / 65 R15 CO / 65 R15 COAA AA 195 / 65 R15 CO / 65 R15 CO / 65 R15 COAB AB 195 / 65 R15 COAC AC 175 / 65 R14 Studless 175 WL WD 175 / 65 R14 Stud 175 WS WE 195 / 65 R15 Studless 195 WL WF 195 / 65 R15 Stud 195 WS WG Tyres sizes and codes Combined ad Long + Lateral tests performed Florence February 23 rd 2001 Paper pag. 5

6 Distribution of braking tests available (darker green: wet) Synthesis outputs of the tests Total: 834 For each test, following the procedure described in the next paragrafh we obtained a set of n braking force versus longitudinal slippage curves (with n=number of vertical loads imposed); the performance obtained has been condensed in four parameters: - Peak value of braking normalised force (µ MAX ) - The corresponding value of longitudinal slippage (x PEAK ) - The slope of the curve in the linear field (K) - The braking force at locked wheel (SLIP). These are the data used to perform the correlations, crossing them with tire dinamic structural parameters and tread compound laboratory parameters (see subsequent paragraph). The xpeak value does not appear in this report, because it is always strictly inversely related to te slope (K), hence their correlations with other parameters are always quite similar. Florence February 23 rd 2001 Paper pag. 6

7 Synthetic outputs generating procedure The steps followed to generate the results to be analysed, starting from the raw data provided by partners, were: - Putting together (when necessary) and arranging the files in.xls format - Importing them in an Access Database - Running queries in order to fit some fields with standard values related to tyre type and conditions - Fitting all the data with Magic Formulae and obtaining Pacejka coefficients - Checking raw data quality and fitting reliability - Automatically generating the test output values of interest (Curve slope in linear field, maximum value, slip value, maximum/slip ratio, peak slippage value). Once available the Pacejka coefficients for each test, it is possible to calculate the longitudinal force for each condition of longitudinal slippage and vertical load, but the interpolation is reliable only in the range defined by the minimum and maximum load applied during the test. All the sintesis parameters calculated are referred to vertical loads of 3500N and 4000N, included in most ranges. All the correlations have finally been done uniquely referring to 4000N load, that all the indexes are normalized with. Peak, slide, slope values Florence February 23 rd 2001 Paper pag. 7

8 Structural lumped tyre parameters. The picture illustrates a two-dimensions lumped parameters model which is able to represent the tire in-plane behaviour in a frequency range up to about 100Hz. It is modeled as two rigid parts (hub and ring), linked by equivalent springs and damper; the transient nature of contact longitudinal force is given by a 1 st order differential equation, suitable for the range of small slippage values 1. The values of the lumped parameters relative to some of the Pirelli tires tested had been previously identified and then have been included among the correlations data. Actually, the reliability of these values is not fully ensured; this is the reason why their correlations are not often mentioned inside this work. 1 Cfr. Techniques for determining the paramet ers of a two-dimensional tire model for the study of the ride comfort, G. Matrascia, Tire science and technology, 1997, Vol. 25, Number 3 Florence February 23 rd 2001 Paper pag. 8

9 TRAILER TESTS ON WET ASPHALT The major number of tests in this task have been conducted on wet asphalt; the testing condition were very various, because of the change of: - Speed - Water depth - Tread depth generating definitely different results. Consequently, in order to analyse the results, each time a complex multi-variable analysis has been done, based on the optimisation of a linear model taking in account the influences of a set of parameters, which are: - Clas s parameters: - The tyre size (only in the case of the whole data group, see paragraph 2.2.1) - Testing device - Surface type - Numerical parameters: - Tyre tread depth - Speed - Water depth - Studied parameters (structural; compound) Florence February 23 rd 2001 Paper pag. 9

10 The next table resumes the different combinations of tests performed on wet asphalt: Device Detailed Surface Water_Depth NRows CETE Asphalt CETE CETE Asphalt CETE CETE Asphalt CETE DA Asphalt Vizzola 1 19 DA Asphalt Vizzola 3 1 DA Misto Esterno Vizzola 1 1 PIRELLI Asphalt Vizzola 3 22 PIRELLI Asphalt Vizzola 7 24 PIRELLI Asphalt Vizzola 1 51 VTI Asphalt VTI 1 92 VTI Asphalt VTI 3 46 VTI Asphalt VTI 7 59 Total: 724 Velocities: 20, 30, 40, 50, 60, 80, 90, 100 Km/h Tread Depths: 0, 2, 4 mm. The multi-variable model for wet group has been optimised with the following specifications: Object variable: Effect variables: µ MAX ; K (slope); Slide - Speed - Water depth - Surface - Device - Tyre size - Tread depth Results are showed above. - m MAX value - Florence February 23 rd 2001 Paper pag. 10

11 The lower are the blue indexes on the right side, the higher and stronger is the influence of the relative variable on the predicted variable (in this case, µ MAX ). Therefore, the influences of speed, water and tyre tread depths, tyre type, type of surface are strong and clear, as showed in the next diagrams. The device influence is less clear. The warning LostDFs means that the number of possible values of the variable is too high relative to the number of data and other variables entered in the model, making less clear the influence of the same variable. The lower is the range of values (numerical or not) assumed by each variable, the higher is the probability to find a good relation. The next table shows the specific influence of each parameter ( Estimate coloumn), with a numer that must be considered as a multiplication coefficient in the case of numeric variables (speed, depths) and a value associated with each possible value assumed by the variable in the case of category variables (such as Tyre type, device, surface). Here the considered output is the µmax=maximum_force/load. Response: Mu RSquare Observations (or Sum Wgts) 717 Parameter Estimates Term Estimate Std Error t Ratio Prob> t Intercept Biased <.0001 Speed <.0001 Water Depth <.0001 Tread Depth <.0001 Tyre[175S-COAC] Tyre[185S-COAC] Tyre[185W-COAC] Tyre[195S-COAC] Tyre[195W-COAC] <.0001 Tyre[225S-COAC] Tyre[AIPC-COAC] Tyre[CO97-COAC] Tyre[CO98-COAC] Tyre[CO99-COAC] Tyre[COAA-COAC] Tyre[COAB-COAC] Device[CE-VT] Biased Device[DA -VT] Biased <.0001 Device[PI-VT] Biased Surface[BB-ME] Biased <.0001 Surface[DC-ME] Biased Surface[ES -ME] Zeroed 0 0?? Surface[F1 -ME] Biased <.0001 Surface[F3 -ME] Zeroed 0 0?? Florence February 23 rd 2001 Paper pag. 11

12 Speed Water Depth Speed Leverage Prob>F: < Tread Depth Water Depth Leverage Prob>F: < Tyre Tread Depth Leverage Prob>F: < Device Prob>F: < Surface Prob>F: Prob>F: < Florence February 23 rd 2001 Paper pag. 12

13 Remarks - Note that tread and water thicknessess in the diagrams shown above are not the few, discrete values really used in the tests, but they appear as an amount of continously changing values. This happens because the showed diagrams actually are of legerage type (Sall, 1990) 2. - The main effect of reduction of peak force value with the increase of water depth and speed and with the reduction of the tire tread depth, is fully noticeable and is in the correct directions. The effect of the tread thickness appears to be stronger than the water depth one. - Tyre different properties effect also is evident; the worse behaviour of winter sizes on wet is acceptable, while more surprising is the collocation of 225/45 R17 Summer tyres among more normal sizes. The same ranking visible in the leverage plot can be retrieved in the table of parameter estimates. - Similar models have been calculated for the subsequent parameters: - K - Slide - Peak/Slide in the subsequent, the most relevant results only will be shown. 2 in a leverage plot, the horizontal position of each point can be shifted with reference to the true value assumed by the sample (for numeric variables), since it is representative of the influence of the single data. The shift is influenced by the other effects in the model. In the case of category variable, the horizontal position is determined by the resulting ranking, and could make visible different groups of data with a very different behaviour. However, the more the region in which the line is traced is narrow, and the slope of this line is high, the more the effect is consistent. For more informations about leverage plots, see JMP3.1 Statistics and Graphics Guide, page 153. Florence February 23 rd 2001 Paper pag. 13

14 - initial slope value - Response: K Summary of Fit RSquare RSquare Adj Root Mean Square Error Mean of Response Observations (or Sum Wgts) 712 Parameter Estimates Term Estimate Std Error t Ratio Prob> t Intercept Biased <.0001 Speed <.0001 Water Depth <.0001 Tread Depth Tyre[175S-COAC] <.0001 Tyre[185S-COAC] Tyre[185W-COAC] Tyre[195S-COAC] Tyre[195W-COAC] <.0001 Tyre[225S-COAC] Tyre[AIPC-COAC] <.0001 Tyre[CO97-COAC] Tyre[CO98-COAC] Tyre[CO99-COAC] Tyre[COAA-COAC] Tyre[COAB-COAC] Device[CE-VT] Biased Device[DA -VT] Biased Device[PI-VT] Biased Surface[BB-ME] Biased Surface[DC-ME] Biased Surface[ES-ME] Zeroed 0 0?? Surface[F1-ME] Biased <.0001 Surface[F3-ME] Zeroed 0 0?? Effect Test Source Nparm DF Sum of Squares F Ratio Prob>F Speed e <.0001 Water Depth e <.0001 Tread Depth e Tyre e <.0001 Device LostDFs Surface e <.0001 LostDFs Florence February 23 rd 2001 Paper pag. 14

15 Whole-Model Test K Predicted Speed Water Depth Speed Leverage Water Depth Leverage Prob>F :< Tread Depth Prob>F :< Tyre Tread Depth Leverage Tyre Leverage Prob>F :< Prob>F :< Florence February 23 rd 2001 Paper pag. 15

16 Device Surface Device Leverage Prob>F :<0.1885, LostDFs Surface Leverage Prob>F :<0.0001, LostDFs Note that the most relevant effects on braking behaviour slope at small slippages are the changes in surface and tyre type. Speed and water depth effects are much smaller, but in the correct directions. Particularly, the next plot shows the leverage effect of each one of the surfaces: BB CETE DC VTI ES CETE F1 Vizzola (Pirelli) F3 Vizzola (Pirelli) ME Vizzola (Pirelli) Application of devices on surfaces However, the stiffness is generally the most difficult parameter to correctly identify, since it is strongly affected by possibile errors and precision loss in the global slippage measurement, which can be critical especially for lower values. In addition to that, it must be considered that in most cases each surface has been tested with the respective device, depending on the owner (see table above), and therefore the influence of the surface could actually hide the behaviour of the device. But the effect test and the leverage plots (see previous pages) show a non clear correlation with the variety of device levels. This seems to confirm the real influence of the surface texture. These considerations are done within the multi-variable model, i.e. taking in account the effect of speed and water and tread depth. Otherwise, it should be difficult to get to any conclusions. Florence February 23 rd 2001 Paper pag. 16

17 Response: Slide Summary of Fit RSquare RSquare Adj Root Mean Square Error Mean of Response Observations (or Sum Wgts) slide value - Parameter Estimates Term Estimate Std Error t Ratio Prob> t Intercept Biased <.0001 Speed <.0001 Water Depth Tread Depth <.0001 Tyre[175S-COAC] Tyre[185S-COAC] Tyre[185W-COAC] Tyre[195S-COAC] Tyre[195W-COAC] Tyre[225S-COAC] Tyre[AIPC-COAC] Tyre[CO97-COAC] Tyre[CO98-COAC] Tyre[CO99-COAC] Tyre[COAA-COAC] Tyre[COAB-COAC] Device[CE-VT] Biased Device[DA -VT] Biased Device[PI-VT] Biased <.0001 Surface[BB-ME] Biased <.0001 Surface[DC-ME] Biased Surface[ES-ME] Zeroed 0 0?? Surface[F1-ME] Biased <.0001 Surface[F3-ME] Zeroed 0 0?? Effect Test Source Nparm DF Sum of Squares F Ratio Prob>F Speed <.0001 Water Depth Tread Depth <.0001 Tyre <.0001 Device <.0001 LostDFs Surface <.0001 LostDFs Florence February 23 rd 2001 Paper pag. 17

18 Whole-Model Test Slide Predicted 4000 Speed 4000 Water Depth Speed Leverage Prob>F :< Tread Depth Water Depth Leverage Prob>F :< Tyre Tread Depth Leverage Prob>F :< Tyre Leverage Prob>F :< Florence February 23 rd 2001 Paper pag. 18

19 4000 Device 4000 Surface Device Leverage Prob>F :< Surface Leverage Prob>F :< Florence February 23 rd 2001 Paper pag. 19

20 - peak/slide ratio- Response: Peak/Slide Summary of Fit Rsquare RSquare Adj Root Mean Square Error Mean of Response Observations (or Sum Wgts) 710 Parameter Estimates Term Estimate Std Error t Ratio Prob> t Intercept Biased <.0001 Speed <.0001 Water Depth <.0001 Tread Depth <.0001 Tyre[175S-COAC] Tyre[185S-COAC] Tyre[185W-COAC] Tyre[195S-COAC] Tyre[195W-COAC] Tyre[225S-COAC] Tyre[AIPC-COAC] <.0001 Tyre[CO97-COAC] Tyre[CO98-COAC] Tyre[CO99-COAC] Tyre[COAA-COAC] Tyre[COAB-COAC] Device[CE-VT] Biased Device[DA -VT] Biased Device[PI-VT] Biased <.0001 Surface[BB-ME] Biased Surface[DC-ME] Biased Surface[ES-ME] Zeroed 0 0?? Surface[F1-ME] Biased <.0001 Surface[F3-ME] Zeroed 0 0?? Effect Test Source Nparm DF Sum of Squares F Ratio Prob>F Speed <.0001 Water Depth <.0001 Tread Depth <.0001 Tyre <.0001 Device <.0001 LostDFs Surface <.0001 LostDFs Florence February 23 rd 2001 Paper pag. 20

21 5.0 Whole-Model Test Peak/Slide Predicted 5.0 Speed 5.0 Water Depth Speed Leverage Prob>F :< Tread Depth Water Depth Leverage Prob>F :< Tyre Tread Depth Leverage Prob>F :< Tyre Leverage Prob>F :< Florence February 23 rd 2001 Paper pag. 21

22 5.0 Device 5.0 Surface Device Leverage Prob>F :<0.0001, LostDFs Surface Leverage Prob>F :<0.0001, LostDFs Florence February 23 rd 2001 Paper pag. 22

23 It is interesting to observe the general influence of speed on the peak/slide ratio (first plot of previous page). With increasing speed, the ratio tends to increase, emphasising the gain due to the use of an hypothetical ABS system, as explained in the next slide: This consideration is comfirmed by the vehicle tests performed with and without use of ABS system, on dry and wet asphalt (see relative section). Below, the found general effects of wet testing conditions on performance parameters are summarised: Florence February 23 rd 2001 Paper pag. 23

24 TRAILER TESTS ON SNOW The distribution of braking tests available on snow is the following: 175WL 175WS 195WL 195WS 175WL 175WS 195WL 195WS 175WL 175WS 195WL 195WS Device Surface Tread Depth: always 8 mm ALL SIZES : MULTI-VAR MODEL Whole-Model Test WL 175WS 195WL 195WS Mu Predicted R 2 = Florence February 23 rd 2001 Paper pag. 24

25 Response: Mu Summary of Fit Rsquare RSquare Adj Root Mean Square Error Mean of Response Observations (or Sum Wgts) 46 Parameter Estimates Term Estimate Std Error t Ratio Prob> t Intercept <.0001 Tyre[175M-195WS] Tyre[175S-195WS] Tyre[175W-195WS] Tyre[185W-195WS] <.0001 Tyre[195S-195WS] <.0001 Tyre[195W-195WS] <.0001 Tyre[225S-195WS] <.0001 Tyre[225W-195WS] Tyre[CO97-195WS] Tyre[CO98-195WS] Tyre[CO99-195WS] <.0001 Tyre[COAA-195WS] Tyre[COAB-195WS] <.0001 Tyre[COAC-195WS] Tyre[175WL-195WS] Tyre[175WS-195WS] Tyre[195WL-195WS] <.0001 Surf_cod[LN-TS] Surf_cod[M2-TS] Device[NO -VT] Speed Effect Test Source Nparm DF Sum of Squares F Ratio Prob>F Tyre <.0001 Surf_code Device Speed Tyre Surf_code Tyre Leverage Prob>F : < Surf_code Leverage Prob>F : Florence February 23 rd 2001 Paper pag. 25

26 0.45 Speed Speed Leverage Prob>F : The effect of speed is nearly null. The very high value of R 2 is mostly due to the two very distinct groups of data winter and summer tyres. It is then convenient to split them in two groups: summer and winter tyres. Summer TYRES : ALL SIZES Response: Mu Summary of Fit Rsquare RSquare Adj Root Mean Square Error Mean of Response Observations (or Sum Wgts) 24 Parameter Estimates Term Estimate Std Error t Ratio Prob> t Intercept <.0001 Tyre[175M-COAC] Tyre[175S-COAC] Tyre[195S-COAC] Tyre[225S-COAC] Tyre[CO97-COAC] Tyre[CO98-COAC] Tyre[CO99-COAC] Tyre[COAA-COAC] Tyre[COAB-COAC] Surf_cod[LN-TS] Surf_cod[M2-TS] Speed Effect Test Source Nparm DF Sum of Squares F Ratio Prob>F Tyre Surf_code Speed Florence February 23 rd 2001 Paper pag. 26

27 Whole-Model Test Tyre Mu Predicted Tyre Leverage Surf_code Speed Speed Leverage Remarks: - Differences among tyres are quite strong; - The three surfaces appear as discernible; - The speed influence is null. WINTER TYRES - peak value - Response: Mu Summary of Fit RSquare RSquare Adj Root Mean Square Error Mean of Response Observations (or Sum Wgts) 22 Florence February 23 rd 2001 Paper pag. 27

28 Parameter Estimates Term Estimate Std Error t Ratio Prob> t Intercept <.0001 Tyre[175W-195WS] Tyre[185W-195WS] Tyre[195W-195WS] Tyre[225W-195WS] Tyre[175WL-195WS] Tyre[175WS-195WS] Tyre[195WL-195WS] Device[NO-VT] Surf_cod[LN-TS] Surf_cod[M2-TS] Speed Effect Test Source Nparm DF Sum of Squares F Ratio Prob>F Tyre Device Surf_code Speed Whole-Model Test WL 175WS 195WL 195WS Mu Predicted Correlations with compound parameters haven t been made (only Pirelli 185W and 225W data available, for three tests). Florence February 23 rd 2001 Paper pag. 28

29 Tyre 0.44 Device Tyre Leverage Device Leverage Prob>F: Surf_code 0.44 Prob>F: Speed Speed Leverage unlikprob>f: Prob>F: Remarks: - The dependence on tyre type is quite clear; but, removing the blue cross in the right upper corner of first plot (185W, snow Mella2, by Nokian, 30 Km/h), general R 2 decreases to 0.883, and tyre leverage probability F increases to : Florence February 23 rd 2001 Paper pag. 29

30 0.44 Tyre Tyre Leverage - Devices are undiscernible; - Surfaces influence is good; this is mainly due to the difference of behaviour between Mella2 snow and the surfaces Taso and Lento, as shown in the plot. Unlikely, the ranking is different from the one obtained for the only -summer-group (see relative section). - The speed influence can be considered as null. - slide value - Response: Slide Summary of Fit RSquare RSquare Adj Root Mean Square Error Mean of Response Observations (or Sum Wgts) 22 Parameter Estimates Term Estimate Std Error t Ratio Prob> t Intercept <.0001 Tyre[175W-195WS] Tyre[185W-195WS] Tyre[195W-195WS] Tyre[225W-195WS] Tyre[175WL-195WS] Tyre[175WS-195WS] Tyre[195WL-195WS] Device[NO-VT] Surf_cod[LN-TS] <.0001 Surf_cod[M2-TS] <.0001 Speed Effect Test Source Nparm DF Sum of Squares F Ratio Prob>F Tyre Device Surf_code <.0001 Speed Florence February 23 rd 2001 Paper pag. 30

31 1700 Whole-Model Test WL 175WS 195WL 195WS Slide Predicted Tyre 1700 Device Tyre Leverage Device Leverage Prob>F: Surf_code Prob>F: Speed Speed Leverage Prob>F: < Prob>F: Florence February 23 rd 2001 Paper pag. 31

32 Remarks: - Clear dependence on tyre; - Device effect is not meaningful; - Very strong surface effect: the same than for µ; - Very weak effect of speed. CONCLUSIONS ABOUT SNOW General conclusions about braking tests on snow are the following: - The dependence on the surface and, obviously, on the tyre type is always very relevant; - The device effect is never clear: this means that there is no evidence of systematic errors on the measurements; TRAILER TESTS ON ICE ALL SIZES Tyre Sizes: Devices: Surfaces: Speeds: 175WL 175WS 195WL 195WS Total 48 Tread Depth: always 8 mm - peak value Response: Mu Summary of Fit Rsquare RSquare Adj Root Mean Square Error Mean of Response Observations (or Sum Wgts) 48 Florence February 23 rd 2001 Paper pag. 32

33 Parameter Estimates Term Estimate Std Error t Ratio Prob> t Intercept <.0001 Surf_cod[M1-T3] <.0001 Surf_cod[SR-T3] Surf_cod[T1 -T3] Device[NO -VT] Speed Tyre[175M-195WS] Tyre[175S-195WS] Tyre[175W-195WS] Tyre[185S-195WS] Tyre[185W-195WS] Tyre[195S-195WS] Tyre[195W-195WS] Tyre[225S-195WS] <.0001 Tyre[225W-195WS] Tyre[CO97-195WS] Tyre[CO98-195WS] Tyre[CO99-195WS] <.0001 Tyre[COAA-195WS] Tyre[COAB-195WS] Tyre[COAC-195WS] Tyre[175WL-195WS] Tyre[175WS-195WS] Tyre[195WL-195WS] <.0001 Effect Test Source Nparm DF Sum of Squares F Ratio Prob>F Surf_code <.0001 Device Speed Tyre < Whole-Model Test Mu Predicted R 2 = Florence February 23 rd 2001 Paper pag. 33

34 0.25 Surf_code 0.25 Device Surf_code Leverage Prob>F: < Device Leverage Prob>F: Speed 0.25 Tyre Speed Leverage Prob > F : Tyre Leverage Prob>F: < Remarks: - Good surface effect; - Very good Tyre effect. Winter tyres performance is higher, but the difference with summer is not so strong as in the case of snow; - Speed and device influences are null. Florence February 23 rd 2001 Paper pag. 34

35 - slide value - Response: Slide Summary of Fit RSquare RSquare Adj Root Mean Square Error Mean of Response Observations (or Sum Wgts) 48 Parameter Estimates Term Estimate Std Error t Ratio Prob> t Intercept <.0001 Surf_cod[M1-T3] <.0001 Surf_cod[SR-T3] Surf_cod[T1-T3] Device[NO -VT] Speed Tyre[175M -195WS] Tyre[175S-195WS] Tyre[175W-195WS] Tyre[185S-195WS] Tyre[185W-195WS] Tyre[195S-195WS] Tyre[195W-195WS] <.0001 Tyre[225S-195WS] <.0001 Tyre[225W-195WS] Tyre[CO97-195WS] <.0001 Tyre[CO98-195WS] Tyre[CO99-195WS] <.0001 Tyre[COAA-195WS] Tyre[COAB-195WS] Tyre[COAC-195WS] Tyre[175WL-195WS] Tyre[175WS-195WS] <.0001 Tyre[195WL-195WS] <.0001 Effect Test Source Nparm DF Sum of Squares F Ratio Prob>F Surf_code <.0001 Device Speed Tyre < Whole-Model Test Slide Predicted R 2 = Florence February 23 rd 2001 Paper pag. 35

36 800 Surf_code 800 Device M Surf_code Leverage Prob > F: < Device Leverage Prob > F: Speed 800 Tyre Speed Leverage Tyre Leverage Remarks: Prob > F: Prob > F: Good surface influence (particularly, M1 is separated and weak ); - Device influence is not feasible; - Speed seems to decrease the slide force value; - There is a very strong and sensible influence of tyre type. Florence February 23 rd 2001 Paper pag. 36

37 CONCLUSIONS ABOUT ICE - The specific device does not influences the test in a dramatic way; - A strong dependence on surface occur; BRAKING TESTS ON VEHICLE WITH/WITHOUT ABS Always within VERT project, braking tests with a vehicle have been performed by CETE on wet and dry asphalt, with and without the use of ABS system, braking at a speed of 80Km/h. The tyres tested are the same six 195/65 R15 Summer tyres used in trailer tests to investigate the tread compound effect (plus a Winter tyre). Several plots were traced, for each possible combination of test on vehicle related test on trailer, wet surface. In most cases, data distribution was not significant at all. The few interesting cases are showed below, where distances are reported in percentage, with reference to AA compound ABS on, separately for wet and dry: Distance% By Peak value Distance% By Peak value Picco Trailer: Pirelli, 1 mm water depth, 50 Km/h; Vehicle: ABS on Distance% By Peak value Picco Trailer: VTI, 1mm water depth, 100 Km/h; Vehicle: ABS on Distance% By Slide Picco Trailer: VTI, 7 mm water, 100Km/h; Vehicle: ABS on Slide Trailer: Pirelli, 1 mm water, 50 Km/h; Vehicle: ABS off Florence February 23 rd 2001 Paper pag. 37

38 Distance% By Slide Slide Trailer: Pirelli, 1mm, 90 Km/h; Vehicle: ABS off Some of the conclusions of CETE find a comfirmation in the trailer braking tests database. Particularly: - In CETE report one of the comments is: The greater efficiency of ABS on a wet road surface compared to a dry surface (30% increase in stopping distance without ABS on a wet surface compared to 11% on a dry road surface). This point might be retrieved on the curves longitudinal force slippage generated with trailer tests. In effect, let see the statistical distribution, on wet and on dry tests, of the peak/slide ratio: - Wet surface Florence February 23 rd 2001 Paper pag. 38

39 - Dry surface The peak of distribution in tests on wet is shifted towards higher values than the dry ones. This comfirms the greater increase of performance of ABS on wet. - ABS performance increase with respect to normal braking has been valued by CETE as higher in a intermediate range of velocities. This is not in discord with the considerations about the increase of peak/slide ratio with, done previously. Florence February 23 rd 2001 Paper pag. 39

40 CONCLUSION The expected testing conditions effects, particularly on wet asphalts (the biggest sub-group of available data), have been fully retrieved. In some cases, the ranking among tyres with same structure and different tread compound, provided with trailer tests, was coherent with that obtained with a group of braking tests on a vehicle. The influence of testing conditions, such as velocity, speed, water and tread depths, ground surface, is definitely correct in the case of wet asphalt tests, and gives much information about the variation of available friction. The influence of the different devices used can be recognized, but it is not judged high enough to spoil the validity of tests; in any case, it is automatically taken in account in the statistical evaluations. Different surfaces also dramatically influences the performance, especially in the case of snows and ices, where the effect of the other variables (tread depth, velocity) cannot be investigated (it was out of the purposes of the test activity). The quantitative characterisazion of surface would need a more exhaustive activity. Dry VERT activity has been mostly focused on critical conditions regarding available friction, such as wet, snowed and iced surfaces; thus, only a few number of tests has been performed on dry asphalt, with different tread compounds but also very different structures (from 175/65R14 up to 225/45R17). However, the most relevant effect seems to be a difference between the sensitivity of µ MAX and slide values in relation to the tyre properties, generating a µ MAX /Slide ratio generally proportional to µ MAX : Wet A big amount of tests has been carried out on wet surfaces. Those of them which have passed the interpolation and check phases can be are 724 tests with different tyre sizes and structures, tread depths, water depths, speeds, surfaces. This group of data is statistically very relevant and leads to an excellent coerence of results if referred to the testing conditions. Particularly: - The main effect of reduction of peak force value with the increase of water depth and speed and with the reduction of the tire tread depth, is fully noticeable and is in the correct directions. The effect of the tread thickness appears to be stronger than the water depth one. - Always for µmax value, Tyre different properties effect also is evident; the worse behaviour of winter sizes on wet is acceptable, while more surprising is the collocation of 225/45 R17 Summer tyres among more normal sizes. The same ranking one can notice in the leverage plot can be retrieved in the table of parameter estimates. - The most relevant effects on braking behaviour slope at small slippages are the changes in surface and tyre type. Speed and water depth effects are much smaller, but in the correct directions. Florence February 23 rd 2001 Paper pag. 40

41 - With increasing speed, the peak/slide ratio tends to increase, emphasising the gain due to the use of an hypothetical ABS system. - This consideration is comfirmed by the vehicle tests performed with and without use of ABS system, on dry and wet asphalt. Below, the found general effects of wet testing conditions on performance parameters are summarised: Snow For what concerns trailer tests on ice for different tyre sizes, the main found evidences are: - Generally, the effect of the speed on the achieved performances seems to be nearly null. - Very (obviously) different behaviour between summer and winter tyres retrieved. - Strong influence of the tyre type also withing the group summer and the group winter - Definitely strong influence of the surface (type of snow) on behaviour The tyres with same structure and different compounds have highlighted a strong influence of the surface behaviour. Ice - The specific device does not seem to influence the test in a dramatic way; - A strong dependence on surface type has been highlighted. Vehicle The braking tests on XANTIA, executed by CETE, on wet and dry surfaces, activating and deactivating the ABS system, were generally in good agreement with the compounds ranking defined with the trailer tests. On the other side, the attempt of establishing a direct relation between trailer tests results (µmax and slide values) was much less productive. Florence February 23 rd 2001 Paper pag. 41

42 ACKNOWLEDGMENTS Acknowledgments are due to all the VERT partners, without whose co-operation it would have been impossible to collect a so relevant and useful amount of information. CONTACT - Diego Donadio Pirelli Tyres Tel diego.donadio@pirelli.com - Diego Speziari Pirelli Tyres Tel diego.speziari@pirelli.com REFERENCES - JMP3.1 Statistics and Graphics Guide, SAS Institute Inc. - Techniques for determining the parameters of a two-dimensional tire model for the study of the ride comfort, G. Matrascia, Tire science and technology, 1997, Vol. 25, Number 3 Florence February 23 rd 2001 Paper pag. 42