Geometric Descriptors of Road Surface Texture in Relation to Tire/Road Noise

The paper deals with the determination of geometric parameters in order to study the relationship between the tire/road noise and the texture of road surfaces. The approach was found to be an alternative to the classical spectral analyses and the numerical simulations of the tire/road contact. Texture parameters were derived from previous works in LCPC related to the influence of the microtexture of road surfaces on the skid resistance. Use of these parameters was justified by the consideration of generation mechanisms of rolling noise. Texture, rolling noise and absorption measurements were performed on 12 road surfaces. The measuring devices and the test methods were presented. The texture profile analyses, including the spectral and geometric approaches, were presented. Definitions of the geometric parameters were given. Correlation between the noise and texture spectra showed similar results to those published in previous works. Fair tendencies were found between the global noise level at 90 km/h and the geometric parameters. Unexpected results obtained on the porous asphalt surfaces were partially explained by the attenuation effect, which was quantified by using existing models. Results issued from the correlation between the third octave-band noise levels and the geometric parameters corroborated those of the spectral analyses. INTRODUCTION The noise generated by tire/road contact is the major contribution to light vehicle noise emission at typical cruising speeds, and consequently produces a considerable nuisance in the environment. The physical phenomena generating this noise are both vibratory and aerodynamic. The vibratory mechanisms are responsible for low frequency emission (< 1 kHz). They come from the tire deformation during the contact with the pavement surface, and from the impact between the rubber treads and the surface. The aerodynamic mechanisms are responsible for high frequency emission (> 1 kHz). They correspond to the successive compression/relaxation cycles of the air trapped inside the contact patch, called “air pumping”. Both tire and road surface parameters influence the noise emitted. The main parameters of road pavements regarding noise are texture and porosity. Rough texture increases the tire vibration and consequently increases its sound radiation. Air pumping effect is maximum on a smooth pavement, but highly reduced by porosity. In the seventies, Sandberg and Descornet (1) established correlation relations between tire/road noise spectrum (sound pressure level as a function of temporal frequency) and road texture spectrum (height of the texture profile as a function of spatial wavelength), on an experimental basis. They identified two zones corresponding respectively to both generating mechanisms: one in low frequency range (below 1.5 kHz) of positive correlation related to radial vibration of the tire, the second in high frequency range of negative correlation related to air pumping effect. Some more recent researches on the subject, used sophisticated numerical codes (Finite Element Method, Boundary Element Method...) to develop comprehensive models for tire noise prediction where the input data were 2D (2) or 3D (3) texture profiles. By use of mechanical contact model, the enveloped texture profile is transformed into surface displacement or strain at the contact patch, and the resulting vibration and radiation of the tire were derived after significant calculation with structural dynamic models. In this paper, a simple and comprehensive characterization of texture profiles is proposed in relation to tire/road noise. The final aim is to define relevant geometric descriptors in order to assess the relationship between the road surface texture and the rolling noise. RESEARCH METHODOLOGY Recently, LCPC developed a profile analysis method to characterize the geometry of microtexture asperities by means of two angular parameters and relate it to the skid resistance (4)(5). The microtexture is defined as surface asperities which height ranges from 0.001 mm to 0.5 mm and width is less than 0.5 mm (6). These asperities are required mainly under wet conditions to breakdown the thin water film and to enhance the creation of effective areas of contact with the tire treads. The angular parameters are related to the shape and relief of the indenters, which were defined as asperities in contact with the tire. Results obtained on road surfaces showed fair correlation between the two angular parameters and the low speed friction (5). It was proposed to use the shape and relief parameters as descriptors of the macrotexture of road surfaces and relate it to the rolling noise. The macrotexture is defined as surface asperities that range from 0.1 mm to 20 mm in height and from 0.5 mm to 50 mm in width (6). The idea is supported by the similarities between the generation mechanisms of friction and rolling noise, both taking into account the deformation of the tire by the Anfosso-Lédée and Do 3 road surface irregularities. The objective was to validate the relevance of the descriptors by correlation with noise levels. Parallel with the geometric analyses, a spectral analysis was performed. The aim was to check if former results of spectral correlation in (1) were still valid for current road surfaces and tires. EXPERIMENTAL PROGRAM Test Sites Tests were performed on 12 road surfaces located on the test tracks of LCPC and INRETS. The surfaces were composed of (the maximum aggregate size in millimeter is indicated in brackets): 2 dense asphalt concrete (10 mm); 4 surface dressing (1.5mm, 4 mm, 10 mm and 14 mm); 2 porous asphalt concrete (10 mm); 1 porous cement concrete (10 mm); 1 very thin asphalt concrete (10 mm); 1 cement concrete; 1 smooth epoxy surface. Tire/Road Noise Measurements A controlled pass-by method (CPB) was chosen for noise evaluation, according to the standard procedure (7). In this method, the pass-by noise of a vehicle driven at a constrained speed is measured at a 7.5 m distance from the vehicle path and 1.2 m above the ground. One small and new passenger car was used, fitted successively with two different tire sets, one rather noisy (large and stiff rubber) and one rather quiet (narrow and soft rubber). Each measurement consisted in eight runs per tire configuration, homogeneously distributed over the total speed range (from 70 km/h to 110 km/h). Each run was characterized by the maximum pass-by noise level and the vehicle speed measured with a radar. Linear regression on the eight noise level – speed data points was performed and the tire/road noise level at the reference speed 90 km/h was derived. The same analysis was done for each third octave frequency band ranging from 100 Hz to 5 kHz, to derive the noise third octave band spectrum at 90 km/h. The noise level and spectrum representative of the road pavement were taken as the average level between the two tire configurations. The global noise levels were used for geometric analysis, third octave band spectra were used for spectral analysis. This procedure is close to the one used by (1) when establishing the texture/noise relationship. Texture Measurements The texture measurements were done by means of a device developed at LCPC using a triangulation laser sensor. The spot diameter of the laser beam is 0.5 mm. Sensor measuring range is 20 mm with vertical resolution less than 5 μm. The profile positions are shown in the figure 1. On each test surface, 20 profiles of 1-meter length (L) were sampled every 0.05 mm along the wheel tracks. They were divided into two series of 10 profiles for each wheel track. The space between the individual profiles was 2 meters. On each profile, invalid points (dropouts) were first suppressed and replaced using the linear interpolation technique. The profile slope was then removed by means of the least squares method. ANALYSIS OF TEXTURE PROFILES Spectral Analysis Each texture profile was signal processed in order to get the texture power spectrum. It means that by a Fourier transformation, the texture profile is expressed as a function of spatial wavelength. The profile level Lx is expressed in dB according to the following : Anfosso-Lédée and Do 4         = ref x x a a L 10 log 20 (1) where ax : profile root mean square (m) aref (= 10 -6 m) : reference root mean square value of the texture profile amplitude x : subscript indicating a value obtained with a certain filter In the present case, third octave wavelength bands were considered. Texture spectral components were expressed in 21 third octave bands, ranging from center wavelength 250 mm to 2.5 mm. This limits are related respectively to the length (L) of the profile and to the spatial sampling of the measurement. The final texture spectrum representative of the road surface is obtained by averaging the spectra of the 20 profiles. A fuller description of the signal processing procedures can be found in (8). The next step was to correlate the noise spectrum and the texture spectrum of the 12 pavements. Thus for each of the 18 noise third octave bands (fi) and each of the 21 texture bands (λj), the noise level of the 12 pavements were plotted against the texture level. The regression line and the correlation coefficient was derived for each of the 18x21 couples (fi, λj). An illustration is presented in figure 2 corresponding to the noise frequency 630 Hz and the texture wavelength 160 mm. Each dot corresponds to a pavement. Profile Analyses by Means of Geometric Parameters The concept of the shape and relief description was fully developed in previous papers (4)(5), only some brief definitions are given in this paper. The analysis method developed in LCPC focused on the significant features of the profiles, that is, the asperities called “indenters”, which are in contact with the tire. Profile indenters were defined as being composed of a profile peak and its two neighboring left-right valleys (Fig. 3). The indenter shape was defined locally as the cotangent of its summit semi-angle (α). In order to take into account the relative positions of the indenters, the indenter relief was defined locally as the angle (θ) between the segment connecting the summits of two consecutive indenters and the horizontal. Beside the angular parameters, it was also possible to define the density as being the number of indenters per unit length. Peaks and valleys were defined as points respectively higher and lower than their neighboring left and right points. Formulae for calculating the shape and relief are the following: p x 1 p x p z 1 p z 1 tan − + − + − = θ (2) where zp, xp: height and abscissa of the p th peak.         − − + − − × = α + + − − − − e 1 e e 1 e 1 1 e e 1 e e 1 z z x x tan z z x x tan 2 1 (3) where ze, xe: height and abscissa of the e th extremum. . The profile analyses were performed by means of a MATLAB program developed at LCPC. The profiles were re-sampled at a sampling interval of 0.5 mm, meaning that the smallest indenter width that could be detected is 1 mm. Option is given in the program to analyze only a part of the profiles. The analysis threshold is defined as a fraction of the difference between the maximum and the minimum amplitudes (0: the whole profile is analyzed; 1: no part of the profile is analyzed). In this study, it was decided to use a threshold of 0.5, meaning that only the upper half of the profiles were analyzed. This choice was supported by the fact that only the parts of the profile in contact with the tire are of interest and the troughs of the profiles might be neglected. Anfosso-Lédée and Do 5 Example of calculation result on actual profiles is shown in the figure 3. For the purpose of clarity, only part of the measured profile (0.1 m length) was shown. On each profile, the indenters were detected and the related shape and relief values were calculated. Values of cotangent (α) and (θ) from the 20 profiles measured on each test surface were then regrouped from which the respective mean values were calculated to characterize the macrotexture of the test surfaces. The density was also calculated by dividing the total number of indenters by the total profile length (20 meters).