2 Literature review
2.5 Factors affecting skid resistance
2.5.1 Surface texture
One of the most popular and enduring models relating macrotexture to skid resistance was developed at Pennsylvania State University (Leu & Henry, 1978). The Penn State model is designed to describe the variation of skid resistance with vehicle speed and it incorporates constants dependent on both macrotexture and microtexture. It was to be used theoretically to predict skid resistance at any speed using measurements of macrotexture alone given an initial measurement of skid resistance on the surface in question. The model was empirically derived using results from skid resistance testing at various speeds and the theory that the gradient of a line fitted to these results is dependent on the macrotexture of the surface.
The experimental, ‘Skid Number’ (SN), data was produced using a locked wheel friction tester, and features of the data such as the Skid Number Gradient (SNG) with respect to the speed of testing, V, and Percentage
Skid Number Gradient (PSNG, which is simply SNG divided by SN, and multiplied by -100 because the gradient is always negative) are used to derive the model as follows:
2.25 2.26
Integrating this from zero to any speed, and assuming PSNG is independent of speed:
2.27 And this yields the Pennsylvania State University model for skid resistance-speed behaviour:
2.28 where c0 is SN0, the skid resistance at zero (or, practically speaking, low) speed, which is thought to be related to microtexture, and c1 is , the percentage skid number gradient, which is independent of speed and is related to surface macrotexture. Curves using this model, which have been fitted to experimental data, are similar to those shown in Figure 2.24.
The behaviour is in part because the surface’s ability to remove water from the contact area becomes crucial as the test speed increases and the time available to remove water is reduced.
Figure 2.24 Friction speed curves using the Penn State model
0 20 40 60 80
0 20 40 60 80 100 120
Friction Number
Speed / km/h
High microtexture, low macrotexture Low microtexture, high macrotexture
to be correlated to one another and a reasonable non-linear relationship with c1 was found:
2.29 For microtexture, values of root mean square height, calculated from surface profiles, were compared with measurements made with the portable skid resistance tester. The root mean square heights were calculated as a function of the longest wavelength and a range of different wavelength cut-offs were examined. The best correlation was obtained for a cut-off wavelength of 0.5 mm.
Forster’s experiments with microtexture measurement using white light line projection (1989) have been mentioned already. He measured surface profiles on 87 samples of asphalt and Portland cement concrete taken from in-service roads and demonstrated a correlation (with correlation coefficient of 0.73) between shape factor (Equation 2.17) and skid resistance, measured using the portable skid resistance tester. It is noted that 100 profiles, each approximately 2 mm long, are sufficient to characterise the microtexture of each sample. The correlation was improved by additional consideration of the percentage contact area between the tester’s rubber slider and the sample surfaces, he concluded that a combination of microtexture and macrotexture characterisation was required to properly estimate skid resistance.
In a joint programme of research investigating the relation between skid resistance and microtexture of pavement surfaces (Himeno, Nakamura, Kawamura, & Saito, 2000) used 17 specimens, measuring 300 mm × 300 mm × 30 mm (height x width x depth), which were prepared from asphalt mixes of varying aggregate gradings. Friction on each specimen surface was measured using a Dynamic Friction Tester (DFT) and the microtexture of the surface of each specimen was measured over four 30 mm lengths, along the path of the rubber tips of the DFT, using a laser profiler (10 m horizontal and 1 µm vertical resolution). The profiles were characterised using Mean Profile Depth (MPD) and a newly proposed parameter, Profile Depth Index (PDI). PDI is defined as a “weighted average of frequency of the wave” and further description of the parameter is very poor, although it is apparently very similar to MPD because its relationship with friction is nearly identical. The results of the dynamic friction tests found that the friction coefficients of the specimens ranged between 0.47 and 0.81 and there was an inverse relation between the
measured dynamic friction coefficient and both the MPD and PDI. In both cases a correlation coefficient, r, of 0.72 was reported.
Work at LCPC, using the indenter model (Figure 2.17) found that and , calculated from sets of 13 surface profiles, 80 mm in length measured on each of approximately 20 asphalt samples, using Equations 2.18 and 2.19, could be correlated with measurements of µ with correlation coefficients (R2) of 0.64 and 0.51 respectively. The work was extended to make use of a model defining the area of contact between a rubber tyre and a rough surface (Do, Marsac, & Delanne, 2004). Friction was calculated and compared against British Pendulum Number and peak friction from vehicle braking tests. It was concluded that, generally, the comparison was fair but that the scatter in the results showed that efforts are still needed before the prediction of friction from surface microtexture is solved. Some analysis of where improvements can be made is also presented: the resolution of the profile, at 10 µm, is sufficient but the shape and relief descriptors of the microtexture ( and ) could be improved. The comparison between texture and friction was improved when only the upper part of the profile is considered (roughly 0.5 mm to 1 mm) and the mathematical process of filtering the profile to achieve this should be improved and some physical proof or justification should be sought.
More recent work by the same research group (Do, Tang, Kane, & de Larrard, 2009) also included use of the indenter model. They measured 15 surface profiles, 76 mm in length, at only three stages during a polishing process using a Wehner-Schulze machine (Section 2.6) on two different aggregates: rhyolite and limestone. In addition to the indenter model, the standard roughness parameter Rq was used to characterise the surface texture in relation to the friction measured in the Wehner-Schulze machine.
Comparisons showed that a reduction in friction on the aggregate surfaces was accompanied by a reduction in roughness as characterised by both methods. The polishing action affected each of the two aggregates differently and two different polishing mechanisms were proposed: for the rhyolite aggregate, polishing affected asperity height, whereas for the limestone aggregate, polishing affected both asperity height and asperity
different minerals and a ‘differential polishing’ behaviour may occur whereby soft minerals are abraded more than hard minerals.
The Health and Safety Executive (HSE) carried out an investigation into slips and trips, the most common cause of major injuries in the work place, and the interaction of a pedestrian’s heel with the floor surface (Shaw, 2007). At present, the HSE use two primary measurements to help assess the slip resistance of floor surfaces - coefficient of friction, determined by portable skid resistance tester (PSRT, aka the pendulum test), and Rz surface microroughness. The Rz component is supposed to provide an indicator of how a surface is likely to perform when it is contaminated; it is typically measured by means of a portable roughness device using a stylus.
The development of affordable and portable roughness instruments has made it possible to measure a wider range of parameters on site than was traditionally available. So, the purpose of the investigation was to study the relationships between different roughness parameters and wet PSRT values to determine whether any of the other parameters could be used in the determination of surface slip resistance.
The HSE investigation recorded 11 roughness parameters on a number of floor samples in laboratory conditions. Parameter values were plotted against measured pendulum test values (PTV) in turn with varying degrees of agreement. Six roughness parameters showed some correlation with was 0.8901 – the highest correlation ever seen by HSE between wet PTV and another surface test. The graph is reproduced below in Figure 2.25.
Other spacing parameters were also used in place of RS but the correlation was not equalled. When the surface is artificially roughened, the highest peaks (and deepest troughs) may not necessarily be removed, and indeed may remain the highest peaks within the sampling length so parameters like Rz or Rp would not change. However, the roughening process is likely to introduce more peaks (and troughs) within each sampling length so RS would decrease. Figure 2.26 shows two example surfaces that have the
same Rz and Rp but different RS, the bottom one, in red, having lower RS perhaps representing the effect of acid etching.
Figure 2.25 Relationship between portable skid resistance tester values and composite roughness parameters
R² = 0.8901
0 10 20 30 40 50 60 70
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
PTV
Rp/ RS
y
yp1 l yv1
y
yp1 l yv1
characterising roughness for all applications. For example, a parameter that characterises two surfaces as equally rough may be sufficient for a structural chemist looking at bonding calculations because the heights of asperities are important, but insufficient for a mechanical engineer looking at wear because the shape and density of the asperities is important.
However, it was possible to relate the texture of surfaces to the generated skid resistance, and a mechano-lattice analogy was developed to analyse the stresses and strains on a surface and predict hysteretic friction. The experimental work was carried out primarily on dry, polished, roadstones and in this case it was noted that the friction between roadstones and sliding rubber is attributable to the energy losses in the deforming tyre.