Transport Research Laboratory
PUBLISHED PROJECT REPORT PPR538
Measuring skid resistance without contact
2009 - 2010 progress report
by A Dunford (TRL)
Prepared for: Project Record:
Client:
Contactless Microtexture Assessment
Transport Research Foundation
Copyright Transport Research Laboratory December 2010
This Published Report has been prepared for Transport Research Foundation. Published
Project Reports are written primarily for the Client rather than for a general audience
and are published with the Client’s approval.
The views expressed are those of the author and not necessarily those of Transport
Research Foundation.
Name
Date
Approved
Project
Manager
A Dunford
25/10/10
Technical
Referee
A Wright
25/10/10
Published Project Report
When purchased in hard copy, this publication is printed on paper that is FSC (Forestry
Stewardship Council) and TCF (Totally Chlorine Free) registered.
TRL
PPR538
Published Project Report
Contents
Executive summary
i
1
Introduction
1
2
Experiment methodology
2
2.1
Specimens
2
2.2
Surface photographs
3
2.3
Surface replication
4
2.4
Scanning electron microscopy
5
2.5
Surface texture measurements
6
3
Digital images
8
4
Scanning electron microscopy
12
5
Surface texture measurements
14
5.1
Excel analysis
14
5.2
Texture analysis using data visualisation software
15
5.3
Texture analysis using bespoke texture analysis software
18
6
Summary
21
7
Discussion
22
8
Conclusions and recommendations
24
Acknowledgements
24
References
24
Appendix A
TRL
Description of the Wehner-Schulze machine
26
PPR538
Published Project Report
Executive summary
This report describes the latest stage in a programme of research, undertaken by TRL for
the Transport Research Foundation (TRF) to investigate the possibility of carrying out or
aiding skid resistance measurement without contact with the road surface. Although
previous stages of this research sought solely to analyse images of aggregate surfaces
and detect visual changes possibly related to changes in skid resistance, the work
reported here has an emphasis on investigating surface texture characterisation using
both qualitative and quantitative measurement.
An aggregate specimen was polished in the laboratory to provide nine discrete levels of
polish, with associated measurements of friction. At each level of polishing, high
resolution images were taken and analysed to provide a record of the process and to
provide a link with previous stages of the research. Although the correlation between
parameters derived from the images and friction that has been demonstrated before was
shown to hold, the analysis also showed that there are likely to be more factors involved
in the changing appearance of the surface than just its texture.
At each polishing stage, replicas of the aggregate surface were taken. As well as
providing a permanent record of the surface texture, the replicas allowed inspection by
scanning electron microscopy (which is not possible on the aggregate specimen because
it is too big for the SEM chamber). Clear changes in the surface texture, apparently due
to the polishing process, were detectable at various scales.
Surface texture was measured using an optical microscope that combines the small
depth of focus of an optical system with vertical scanning to provide topographical
information derived from the variation of focus. Surface topography was analysed with
three different software packages, using a number of standard roughness parameters.
The effect of various wavelength filters was explored and this suggested that
determination of the scale of texture most relevant to skid resistance is important, even
within the relatively narrow range present on a single stone surface. It was shown that
that the distribution of peaks, rather than their heights, is also likely to be a relevant
factor.
Recommendations are made for further use of the specimens prepared in this work and
for development of the methodology that will allow more conclusive texture
measurement analysis in future.
TRL
i
PPR538
Published Project Report
1
Introduction
Routine monitoring of the skid resistance of road surfaces is an important component of
ensuring that the road surface is maintained in a safe and serviceable condition. Various
different types of equipment have been developed for measuring skid resistance, but all
share a common principle of measuring the forces generated when a rubber tyre or
slider is forced to slide across a wetted road surface. This presents a number of
limitations that affect the safety and efficiency of data collection and the quality of the
data obtained. These could potentially be overcome through the development of a
contactless method of characterising the skid resistance properties of road surfaces.
Such a method could be applied throughout the UK and in the many other countries that
conduct skid resistance measurements.
A research programme is being undertaken by TRL for the Transport Research
Foundation (TRF) to determine whether detailed imaging of the road surface has
potential to be applied to the measurement of skid resistance. The ultimate objective of
this work is to develop a method suitable for implementation on a traffic-speed survey
vehicle. This report describes progress made during 2009 – 2010. The work reported in
this document is part of the Transport Research Foundation’s ongoing research
programme.
In the previous stage of the project (Dunford A. , 2009), work concentrated on collecting
a large number of images of road surfaces from a moving vehicle in order to compare
parameters derived from the images with the skid resistance of the surface over
extended continuous lengths of surface. It was found that the image analysis algorithms
were very sensitive to image quality, particularly the quality of focus, achieved by the
image collection system. It was concluded that development of a traffic speed system
will rely on considerable investment in the development of image collection equipment.
However, it was also concluded that, before committing such a significant investment, it
is necessary to have a more complete understanding of the characteristics within the
surface texture affecting the skid resistance that the image collection system would aim
to quantify.
The research presented in this report, therefore, revisits the fundamental principles of
comparing microtexture with skid resistance in anticipation of future equipment
development.
TRL
1
PPR538
Published Project Report
2
Experiment methodology
In order to relate changes in surface texture to changes in skid resistance, surfaces with
a range of textures or skid resistances must be observed.
2.1
Specimens
A number of levels of skid resistance could be achieved by using a variety of materials –
for example, skid resistance on a smooth glass surface compared with skid resistance on
a sheet of glass paper, or skid resistance on a polished limestone surface compared with
skid resistance on a granite surface. However, in order to reduce the potential number
of variables arising from differing structures of materials, in this research a range of
levels of skid resistance have been achieved using one material, polished to different
extents.
In principle, any material could be used, but clearly, since the ultimate goal is to
measure skid resistance on road surfaces, the most relevant materials are aggregates
that are typically used in pavement surface courses. According to a survey of UK
quarries (Thompson, Burrows, Flavin, & Walsh, 2004), the most quarried rock types
used as road construction aggregate are greywacke, dolerite and basalt.
TRL stores a stockpile of dolerite aggregate for use as control stone in the polished stone
value (PSV) test, which measures the polishing resistance of aggregates used in road
surfaces. Skid resistance and polish resistance of this material is well established, and it
is known to be of an homogenous nature (i.e. one stone is as similar to the next as is
possible within a naturally occurring material).
Polishing and measurement of skid resistance was achieved using the Highways Agency’s
Wehner-Schulze machine, which is described in more detail in Appendix A. A regime of
polishing was used to achieve a discrete number of polishing levels on the same
specimen.
The aggregate was polished in several stages, following a logarithmic
pattern: 0; 30; 90; 300; 900; 3,000; 9,000; 30,000 and 90,000 passes of the machine’s
polishing rollers. This resulted in 9 discrete polishing levels, which are referred to by
number (0 – 8) for the remainder of this report. At each polishing level, a set of
photographs of the specimen surface (Section 2.2) and a set of replicas of the specimen
surface (Section 2.3) were taken before friction was measured using the Wehner-Schulze
machine.
Figure 2.1 shows the evolution of friction on the sample surface as it was gradually
polished. The coefficient of friction measured reduces very quickly at first, but then
requires proportionally more passes to achieve further reduction. For example, a
reduction in friction of 0.1 was achieved in the in the first 9,000 passes, and a reduction
of a similar amount was achieved in the following 81,000 passes.
TRL
2
PPR538
Published Project Report
1.0
0.9
Coefficient of friction, µPWS
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0
10000
20000
30000
40000
50000
60000
70000
80000
90000
100000
Number of polishing passes
Figure 2.1 Evolution of friction throughout progressive polishing
Figure 2.2 shows the same friction measurements at each stage against nominal
polishing levels 0 to 8.
1.0
0.9
Coefficient of friction, µPWS
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0
1
2
3
4
5
6
7
8
Polishing level
Figure 2.2 Evolution of friction measurements by polishing stage
2.2
Surface photographs
In order to take photographs of identical areas of the specimen surface at sufficient
resolution to track any changes occurring to the surface texture as the specimen was
polished, the following equipment was used:
Nikon D200 10 megapixel digital SLR
Sigma 50mm F2.8 DG macro lens
Copy stand
Specimen mounting turntable
Flash lamp set to evenly illuminate the area of interest on the specimen.
TRL
3
PPR538
Published Project Report
The specimen was mounted into the turntable and carefully marked so that its position
could be accurately replicated with respect to the camera position in subsequent
photographs.
Photographs were taken of six areas of the surface around the
circumference of the aggregate specimen within the area swept by the polishing rollers
of the Wehner-Schulze machine. Each photographed area encompassed at least 5
aggregate particles so that a total of 30 individual aggregates surfaces could be
analysed. The resolution achieved in the photographs was approximately 10µm per pixel
– within the ‘microtexture’ range of surface texture.
This photographic method of analysis has been used in previous stages of work carried
out for the Transport Research Foundation (Dunford & Viner, 2006), and it was not
intended to be the main focus of the present experiment. However, the photographs
offer an accessible record of changes occurring on the aggregate surface, and the
analysis carried out provides a link to earlier stages of this work.
2.3
Surface replication
In order to examine the surface texture within the same area of aggregate before and
after polishing, and to retain a record of the polishing state, replicas of the surface were
made at each level of polishing. A commercial replication compound, designed for
engineering inspection purposes, (Microset Products Ltd, Nuneaton) was used. This two
part polymer system is mixed as it is dispensed onto the surface from a cartridge to form
a semi-viscous liquid which cures quickly to form a flexible solid compound that can be
peeled from the surface to give an inverse relief (mirror image) replica. Normally used
to perform metallography of machine components such as inspection of microstructure,
micro-cracking and pitting, the replicating compound has better than 0.1µm resolution,
according to the manufacturer.
At each stage of polishing, three of the six photographed areas were replicated. This
was achieved by surrounding the area of interest with a plastic ring of diameter 50 mm
and flooding the surface within the ring with replicating compound. The working life of
the compound is 30 seconds which was sufficient time for it to coat the surface within
the ring and also self-level to give a flat base to the finished (inverted) plaque. Figure
2.3 shows one of the prepared plaques; the replicated surface texture of each aggregate
particle can be easily seen.
Figure 2.3 Replicated aggregate specimen
When each replica plaque was removed from the surface, the area with which it had
been in contact was darker than the surrounding aggregate. This was queried with the
manufacturer who thought it could either be that a monomolecular layer of silicon
compound had been left on the surface, or that the replica plaque had removed very fine
detritus from the surface leaving behind an intricately cleaned area. The replicated
areas can be clearly seen on the aggregate specimen shown in Figure 2.4.
TRL
4
PPR538
Published Project Report
Figure 2.4 Aggregate specimen used in experiment
2.4
Scanning electron microscopy
Scanning electron microscopy was used to check the fidelity of surface replication. It
was also used to examine the surfaces before and after polishing to gain a qualitative
assessment of the polishing process.
A small disc (50mm diameter) was cut from an additional aggregate specimen, and the
replicating compound was applied to its surface using a plastic ring to form a boundary
so that the replication liquid could pool over the whole area. Scanning electron
microscopy (SEM) was then used to compare the replica to the aggregate surface.
Figure 2.5 shows low magnification (composite) images of a stone (top) and its replica
(bottom). Because the replication process provides a ‘mirror image’ of the surface, the
SEM images of the replica have been transposed left-right so that the orientation
matches.
Figure 2.5 Low magnification images of stone (top) and replica (bottom)
TRL
5
PPR538
Published Project Report
Using the shape of the stone and its edges for navigation, several corresponding
positions were found on both the stone and on the replica. Figure 2.6 and Figure 2.7
show two features of the surface which are clearly visible and easily identifiable on both
the stone (left) and the replica (right). The scale information in the black bar on the
bottom of each image indicates a length of 20µm.
Figure 2.6 Feature 1 – stone (left) and replica (right)
Figure 2.7 Feature 2 – stone (left) and replica (right)
This preliminary experiment allows some confidence in the replication technique for
aggregate surfaces, and shows that small features between 5 and 10µm are definitely
replicated. However, there is some loss of definition on the smallest features suggesting
that the resolution of the replication compound may not be as high as 0.1µm when
measuring this type of surface.
In order to directly compare the aggregate surfaces before and after polishing, a similar
technique was used to identify specific features on the before-polishing replica and then
find the same features on the after-polishing replica. Section 4 describes some of the
changes that were observed.
2.5
Surface texture measurements
Texture measurements were made using an Alicona Infinite Focus microscope at the
National Physical Laboratory (NPL) (Figure 2.8) on the surfaces of aggregate particles
replicated at each of the nine polishing levels.
TRL
6
PPR538
Published Project Report
Figure 2.8 Alicona Infinite Focus microscope
The microscope combines the small depth of focus of an optical system with vertical
scanning to provide topographical information derived from the variation of focus. Using
a beam splitting mirror, light emerging from a white light source is inserted into the
optical path of the system and focused onto the specimen via the objective lens.
Depending on the topography of the specimen, the light is reflected in several directions
as soon as it hits the specimen. Light rays emerging from the specimen and hitting the
objective lens are gathered by a light sensitive sensor behind the beam splitting mirror.
Due to the small depth of field of the optics only small regions of the specimen are
sharply imaged when the lens is at any given height. The lens is moved vertically along
the optical axis while continuously capturing data from the surface until each region of
the specimen has been imaged in sharp focus. Algorithms are used to analyse the
variation of focus along the vertical axis and convert the acquired sensor data into 3D
information.
Three regions on each of five aggregate particle surfaces were measured on each of the
three replica plaques taken at each polishing level (45 scanned areas for each polishing
level). Data from the microscope consisted of 3D coordinates for points within a
rectangular region with dimensions approximately 2.8 mm by 2.2 mm. The spacing
between measurement points was approximately 3.5 µm in all three dimensions. The
manufacturer claimed resolution depends on the objective lenses used, and 3.5 µm is
well within the device’s operational range.
The following three chapters describe results from digital imaging, scanning electron
microscopy and surface texture measurements.
TRL
7
PPR538
Published Project Report
3
Digital images
Figure 3.1 shows two images extracted from larger photographs of the specimen surface
before polishing (left) and after all polishing (right), for one of the three areas that were
photographed but not replicated. Figure 3.2 shows a similar pair of images for one of
the three areas that were replicated. In the first pair of images, from a non-replicated
area, the aggregate surfaces appear darker after polishing (i.e. the image on the right is
darker). In the second pair, the difference between the two images is less marked,
although both appear darker than the image of the unpolished, non-replicated area (top
left). It has already been noted that the replicating compound leaves behind a darkened
area (Figure 2.4).
Figure 3.1 Images from non-replicated areas before (left) and after all
polishing
Figure 3.2 Images from replicated areas before (left) and after all polishing
Figure 3.3 shows pixel intensity histograms for photographs taken at each of the
polishing levels 0 to 8 within the non-replicated areas. Figure 3.4 shows equivalent pixel
intensity histograms for photographs of the replicated areas. Each photograph has just
over 10 million pixels, and each pixel can have a discrete intensity value from 0 to 255.
The graphs show a smoothed representation of the number of pixels with each level of
intensity, averaged over the three photographs taken at each polishing level. In the
non-replicated areas, the distribution skews towards lower pixel intensities (darker
images), but in the replicated areas, the distribution of pixel intensities remains broadly
the same throughout the polishing process. Note also that the non-replicated areas are
lighter before polishing, and contain a broader range of pixel intensities (the replicated
areas are darker and more homogenous). These measurements are consistent with
observations based on the images above.
TRL
8
PPR538
Published Project Report
Frequency
(number of pixels having each intensity)
150000
100000
0
1
2
3
4
5
6
50000
7
8
0
0
50
100
150
200
250
300
Pixel intensity
Figure 3.3 Intensity histograms for non-replicated areas.
Frequency
(number of pixels having each intensity)
150000
100000
0
1
2
3
4
5
6
50000
7
8
0
0
50
100
150
200
250
300
Intensity
Figure 3.4 Intensity histograms for replicated areas
Image processing analysis algorithms were developed in previous stages of this work to
quantify texture using variation in pixel intensity, either compared to surrounding pixels,
or compared to a nominal threshold (Dunford & Viner, 2006). The algorithms were
applied separately to the replicated and non-replicated areas, and it was found that the
output parameters did not vary with polishing level for the replicated areas. The
following results have therefore been calculated using only the photographs of nonreplicated areas.
The first algorithm developed in previous research inspects each image and counts the
number of contiguous pixels that fall below a predefined intensity level. In effect, this
measures the homogeneity of the image. Figure 3.5 shows the average number of
pixels contiguously falling below an intensity value of 100 for each polishing level.
TRL
9
PPR538
Published Project Report
Average number of contigous pixels with
intensity less than 100
0.6
0.5
0.4
0.3
0.2
0.1
0
0
1
2
3
4
5
6
7
8
Polishing level
Figure 3.5 Average number of contiguous dark pixels for each polishing level
The second algorithm developed in previous research is based on the conventional
roughness parameter Ra, which measures the average deviation of a surface away from
a notional mean level. Figure 3.6 shows the results obtained when the parameter is
produced by calculating the average deviation of each pixel’s intensity from the mean
pixel intensity of the surrounding 80 pixels.
0.6
"Ra" using pixel intensity and
81 pixel moving average
0.5
0.4
0.3
0.2
0.1
0
0
1
2
3
4
5
6
7
8
Polishing level
Figure 3.6 Average “Ra” at each polishing level
In both cases, the algorithms have been applied to the whole area photographed but
with the region between aggregate particles masked and excluded.
It can be seen that the parameters change with increasing polishing. Figure 3.7 shows
both parameters plotted against friction with trend lines demonstrating a strong
correlation between the parameters and the measured friction. In the case of the R a
based parameter, the correlation is stronger when the final polishing stage is not
included in the trend line (the broken red line).
TRL
10
PPR538
Published Project Report
0.8
Parameter derived from images
y = -1.5534x + 1.2597
R² = 0.9601
0.6
'Ra'
0.4
Dark pixels
y = 0.5502x + 0.1166
R² = 0.6916
0.2
y = 0.9577x - 0.1449
R² = 0.9114
0
0
0.2
0.4
0.6
0.8
1
Coefficient of friction, µ
Figure 3.7 Correlation between friction and parameters derived from images
TRL
11
PPR538
Published Project Report
4
Scanning electron microscopy
Scanning electron microscope images of the surface from replica plaques of the beforeand after-polishing states can be difficult to interpret because the replica surface is in
negative relief.
The changes expected due to polishing will primarily affect the
uppermost surfaces of the stone and consequently these will be represented by the
lowermost surfaces of the replica plaques. However, when comparing SEM images
collected before and after polishing, there are some very clear changes in the texture of
the surface. Figure 4.1 shows a section of replica surface, at low magnification, on the
plaques before (left) and after polishing. Subjectively, the image after polishing appears
smoother in general. It is possible to identify specific areas of the images that appear to
show change on different scales (a length of 500µm is indicated in the black information
bar at the bottom of each image).
Figure 4.1 Texture changes between before-polishing (left) and after-polishing
(right) at 100x magnification
Higher magnification images are shown in Figure 4.2 for a second area of the surface.
Again, the general smoothness of the after-polishing surface compared to the beforepolishing surface is apparent. There are several distinct planes in the image, and the
effect of the polishing process is noticeable across their full width – examples highlighted
with yellow ovals are clearly smoother in the image after polishing.
Figure 4.2 Texture changes between before-polishing (left) and after-polishing
(right) at 200x magnification
The images in Figure 4.3 show a particular feature that has clearly changed on more
than one scale. The edges of the triangular plane have become more rounded while the
surfaces are distinctly smoother, and an additional fracture line has appeared
horizontally across its centre.
TRL
12
PPR538
Published Project Report
Figure 4.3 Texture changes between before-polishing (left) and after-polishing
(right) at 400x magnification
It is likely that the changes at the smallest scale will be affected by the petrography of
the particular rock – being largely due to the size of grains and the mechanisms for
polishing and fracture they exhibit. It is suggested that further qualitative investigation
could use SEM images to study how surface texture on different aggregates changes as
they are polished to help explain why skid resistance reduces more quickly on some
aggregates than on others.
TRL
13
PPR538
Published Project Report
5
Surface texture measurements
Analysis of surface texture measurements made using the focus variation scanning
microscope was undertaken using three different software packages: Microsoft Excel,
data visualisation software IDL by ITT Visual Information Systems and bespoke texture
analysis software MountainsMap by Digitalsurf.
5.1
Excel analysis
Excel was first used to calculate a basic version of the roughness parameter, Sa, which is
the arithmetic mean of absolute departures of a surface from a mean plane (British
Standards, 2008). In the first instance, the assessment lengths lx and ly were set to be
the extent of the area measured: 2.8 mm and 2.2 mm respectively. In practice, this
was achieved by calculating the average of all 500192 height measurements within each
area and then summing the absolute differences between each measurement and this
average before finally dividing by the number of points. Equation 5.1 gives the standard
method of calculation for Sa.
5.1
In an attempt to account for some element of the form of the surface, that is
wavelengths larger than the expected microtexture, the same calculations were made for
octants of the areas scanned. This effectively applies a basic filter by setting the
assessment lengths lx and ly to 0.7 mm and 1.1 mm respectively. The graph in Figure
5.1 shows the average calculated Sa for each polishing level for the full areas (blue
columns in the background using the axis on the left) and for the area octants (red
columns in the foreground using the axis on the right).
100
50
Average Sa - whole area
Average Sa - area octants
90
49
80
48
70
47
Sa / µm
60
50
46
40
45
30
44
20
43
10
0
42
0
1
2
3
4
5
6
7
8
Polishing level
Figure 5.1 Basic calculations of Sa for each polishing level
There is no clear relationship between the calculated values of Sa and the level of
polishing, regardless of the use of octants. Values of Sa calculated using whole areas
were approximately twice the values of S a calculated using octants of the same areas.
The reason for this difference may be that long wavelength features contribute to the
calculated roughness parameter Sa. It is therefore important to apply filters to the data
in order to isolate the scale of texture responsible for changes in skid resistance from the
underlying shape. This has been further investigated in the following sections.
TRL
14
PPR538
Published Project Report
5.2
Texture analysis using data visualisation software
In order to apply more complex filtering to the texture data, IDL data visualisation
software was used to plot 3D representations of the measured surfaces and apply simple
filters before calculating roughness parameters such as Sa, Sq and Sp (British Standards,
2008). Sq, defined by Equation 5.2, is an RMS equivalent to Sa, and for normally
distributed texture they are linearly related. S p reports the maximum peak height above
the average surface plane.
5.2
Figure 5.2 shows an example of one of the 45 areas scanned at polishing level zero
(before polishing). It was found that corrupt data points were present, probably due to
highly reflective spots or other features causing the microscope’s measurement software
to fail. The software has displayed the corrupt data points with surface heights of -9999
in Figure 5.2. Figure 5.3 shows the same surface with the corrupt data points replaced
by interpolated values. All filters and calculations reported herein were applied to these
‘cleaned’ versions of each surface area.
Figure 5.2 Surface profile before polishing
Figure 5.3 Cleaned version of surface profile before polishing
TRL
15
PPR538
Published Project Report
A calculation of Sa on the whole of each surface area before any filtering was applied,
shows similar results to those obtained using Microsoft Excel, as would be expected
(Figure 5.4).
100
90
80
70
Sa / µm
60
50
40
30
20
10
0
0
1
2
3
4
5
6
7
8
Polishing level
Figure 5.4 Average Sa calculated using IDL for all areas before filtering
The additional functionality of IDL, compared with Excel, was used to apply several filters
to the surfaces, with wavelengths based on the standard sampling lengths recommended
in ISO4288: 0.08 mm, 0.25 mm and 0.8 mm (British Standards, 1998) with an
additional intermediate wavelength of 0.5 mm. Figure 5.5 shows, for example, the same
surface from Figure 5.3, with a 0.08 mm moving average filter applied. This is a low
pass filter in the wavelength domain (i.e. wavelengths less than 0.08 mm pass). Figure
5.6 shows the same surface again with a 0.08 mm to 0.25 mm moving average
bandpass filter applied. Note that the size of the area scanned limits the size of the filter
that can be applied and the larger the wavelength, the smaller the remaining useable
dataset.
Figure 5.5 Surface profile with 0.08 mm moving average low pass filter applied
Figure 5.6 Surface profile with 0.08-0.25 mm moving average bandpass filter
applied
Roughness parameters were calculated for every surface area at each polishing level for
each filter applied. Table 5.1 shows the average calculated parameter values.
TRL
16
PPR538
Published Project Report
Table 5.1 Calculated parameters for all filtered surface areas
Moving average
Lowpass
Sa
Sq
Sp
Polishing
level
0.08 mm
0
1
2
3
Moving average
Bandpass
0.25 mm
0.080.25 mm
0.250.5 mm
0.50.8 mm
29.77
69.98
42.79
162.33
262.56
41.76
103.65
62.10
171.86
195.51
29.73
71.14
43.29
158.82
482.42
42.95
106.57
64.12
218.26
341.12
4
33.71
77.80
48.94
194.41
534.65
5
20.92
45.63
29.94
201.70
401.14
6
29.64
66.40
43.19
188.27
369.70
7
38.08
91.45
55.25
231.61
378.45
8
35.59
88.45
52.16
226.97
520.07
0
32.47
71.10
46.41
163.05
263.30
1
44.81
104.41
65.28
172.33
196.48
2
32.64
72.39
46.84
159.75
482.74
3
45.33
107.32
67.00
218.66
341.63
4
36.28
79.31
52.35
195.32
534.93
5
23.59
47.65
33.84
202.15
401.54
6
32.80
67.89
46.72
188.77
370.11
7
40.39
92.37
58.76
231.94
378.89
8
38.41
89.33
55.25
227.54
520.28
0
75.28
114.43
100.99
197.42
293.18
1
100.61
165.80
124.16
208.25
224.70
2
40.82
38.46
40.79
176.40
512.53
3
103.08
176.15
137.38
263.16
376.73
4
47.63
63.36
61.86
210.28
566.94
5
71.32
108.64
86.79
238.90
432.47
6
80.43
124.90
101.66
226.07
402.63
7
92.59
162.13
129.53
271.71
411.70
8
71.04
105.10
85.31
263.74
551.88
The values for Sa, normalised to maximum values, for each of the filters are plotted in
Figure 5.7. There are no discernable trends for changing roughness for increased
polishing, and the same is true for Sq and Sp which exhibit very similar patterns.
TRL
17
PPR538
Published Project Report
1.20
Sa / normalised to maximum
1.00
0.80
0.60
0.40
0.20
Moving average Lowpass 0.08 mm
Moving average Lowpass 0.25 mm
Moving average Bandpass 0.08-0.25 mm
Moving average Bandpass 0.25-0.5 mm
Moving average Bandpass 0.5-0.8 mm
0.00
0
1
2
3
4
5
6
7
8
9
Polishing level
Figure 5.7 Average normalised Sa for surfaces after application of filters
5.3
Texture analysis using bespoke texture analysis software
A commonly used, industry standard, texture analysis software package, MountainsMap
by Digitalsurf was used to calculate a wider range of roughness, spacing and surface
volume parameters using a number of similar surface filters. Some of the parameters
calculated are shown in Table 5.2. Two types of filter were used: a surface levelling filter
using subtraction from a fitted least square plane; and a Gaussian low pass filter with
upper wavelengths 0.025 mm, 0.08 mm, 0.25 mm and 0.8 mm.
Table 5.2 Roughness and spatial parameters calculated using MountainsMap
Parameter
Description
Sq
Root mean square height of the surface
Ssk
Skewness of height distribution
Sku
Kurtosis of height distribution
Sp
Maximum height of peaks
Sv
Maximum height of valleys
Sz
Maximum height of the surface
Sa
Arithmetical mean height of the surface
Spd
Density of peaks – number of local maxima per
areas
The graph in Figure 5.8 shows average values of Sa measured at each polishing level for
all the filters applied, normalised to the maximum value of Sa. As observed in the
previous sections, there are no clear trends for changes in Sa with increased polishing,
regardless of the filter applied. The graph in Figure 5.9 shows normalised values for all
calculated parameters measured on the software levelled surfaces. The pattern of
changing roughness with polishing is similar for all but the skewness measure (S sk), and
TRL
18
PPR538
Published Project Report
to a lesser extent, the kurtosis measure (Sku). The difference is highlighted by, but not
entirely attributable to, a slightly outlying average value at polishing level 7. The
distribution of peak heights is therefore an important measurement for future
consideration.
1.2
Sa / normalised to maximum
1
0.8
0.6
0.4
Original surface
Levelled Surface
Gaussian 0.25mm
Gaussian 0.025mm
Gaussian 0.8mm
Gaussian 0.08mm
0.2
0
0
1
2
3
4
5
6
7
8
9
Polishing levels
Figure 5.8 Average normalised Sa for each polishing level after application of
various filters
Roughness / normalised to maximum
1.2
1
0.8
0.6
0.4
0.2
Sa Levelled Surface
Sq Levelled Surface
Sp Levelled Surface
Sv Levelled Surface
Sz Levelled Surface
Ssk Levelled Surface
Sku Levelled Surface
0
0
1
2
3
4
5
6
7
8
9
Polishing level
Figure 5.9 Average normalised roughness parameters for each polishing level
on software levelled surfaces
Shaw (2007), in work carried out at the Health and Safety Laboratory comparing slip
resistance with roughness in an attempt to improve correlation with polishing level,
combined some of the above parameters. For example, the graph in Figure 5.10 shows
the average values of maximum peak height divided by the peak density (S p/Spd) for all
surface areas at each of the polishing levels after a 0.8 mm Gaussian filter had been
applied (the trend was apparent for the same parameter after application of the other
filters, but it was clearest for 0.8 mm).
TRL
19
PPR538
Published Project Report
1.2
Roughness/Peak density
1
0.8
0.6
0.4
0.2
0
0
1
2
3
4
5
6
7
8
9
Polishing level
Figure 5.10 Average Sp divided by Spd for all surfaces at each polishing level
after application of 0.8 mm Gaussian filter
Although the Sp/Spd parameter can be justified physically – larger or more isolated peaks
generate rougher surfaces than smaller or more densely packed peaks – the trend is not
apparent for either Sp or Spd individually. However, this initial investigation shows that
combined relationships should be considered further as a way of charactering surface
texture.
TRL
20
PPR538
Published Project Report
6
Summary
A methodology developed in earlier stages of this research (Dunford & Viner, 2006),
using laboratory equipment to simulate trafficking, was used to provide a range of levels
of skid resistance on one surface made up of aggregate particles. At each polishing
level, high resolution digital photographs and surface replicas were taken before a
friction measurement was made.
Bespoke image analysis software was used to analyse the photographs to investigate
changes in the appearance of the surface texture with increasing polishing. Algorithms
measuring pixel intensity and its localised variation showed that, on areas that were not
replicated, the darkness and homogeneity of pixels could be correlated to the level of
polishing and to the friction measured.
The surface replica plaques were used to make off-site observations of surface texture
with a scanning electron microscope (the original specimen would not fit in the SEM
chamber) and make off-site measurements of surface texture. The replicas also allow
retention of a permanent record of aggregate surface texture throughout the polishing
process.
Scanning electron microscopy was used to verify the resolution of replication. This was
achieved by using prominent features of the surface (such as stone edges or distinctive
ridges) for navigation so the exact location on a real stone sample and a replica could be
compared. Although the resolution of replication was not as high as the manufacturer
claim, features as small as 5 µm were easily identifiable.
Scanning electron microscopy and the same navigation technique was then used to
compare identical areas from the replica of the before-polishing surface with the replica
of the after-polishing surface to observe changes in the surface texture as it is polished.
The images showed distinctive changes had occurred at more than one scale of texture.
Quantitative measurements of surface texture were made on the replica plaques by
scanning a large number of discrete areas with a focus variation measurement
microscope. Standard roughness parameters applied without filtering to the areas
measured did not demonstrate a change in surface roughness due to polishing. Various
filters were applied to the surface in an attempt to account for changes in surface form.
Some weak trends for changes in roughness parameters were found for skewness and
for a compound of maximum peak height and peak density. These weak relationships
were only found for individual filters – a levelled surface using a least-square fit and a
0.8 mm Gaussian filter respectively.
TRL
21
PPR538
Published Project Report
7
Discussion
It is expected that the effect of polishing is to change (most likely reduce) the
microtexture on the surface. The image analysis algorithms use the supposition that
pixel intensity, and intensity variation, can be used as a surrogate for surface texture.
The encouraging results from analysis of non-replicated areas (Figure 3.5) supports this
supposition, and the long term aim of using high resolution images to assist
measurement of skid resistance. However, the difficulty in achieving the same results
on areas that were replicated emphasises the need to further investigate the physical
changes that occur in the surface texture that actually lead to visual changes. If the
effect of replication is simply to clean the aggregate surface then the implication is that
image analysis only works because texture is highlighted by contrasting detritus trapped
within the aggregate texture. Although in-service roads are not likely to be clean, this
phenomenon could present issues for calibration and consistency when using images of
the road in a practical system. Furthermore, since it is not possible to detect visual
changes in replicated areas it is more difficult to link visual changes with surface texture
changes unless both analyses can be carried out on the aggregate surface rather than on
aggregate and on replica surfaces respectively. An additional advantage of analysing
aggregate surfaces rather than replicas would be the potential use of other
measurement techniques such as energy-dispersive X-ray spectroscopy (EDX) to
determine the chemical composition of the surface and identify any individual
compounds that are more susceptible to deformation than others.
The qualitative observations of changes to surface texture before and after polishing
made using scanning electron microscopy are the most compelling outcome from this
work (Section 4). The intention here was to illustrate the sorts of texture changes that
could be detected through measurement of surface roughness. However, it has not been
possible to demonstrate conclusively that measurements of roughness can be made to
correlate with polishing level or friction (discussed below). Therefore, the methodology
for SEM inspection should be developed further. In the first instance, the surface
changes observed between the first and last polishing levels could be tracked throughout
the polishing process. This could be repeated for other aggregates of the same rock
type, to investigate the reasons for some aggregates’ higher or lower resistance to
polishing. The rock type used in the work reported here, dolerite, has one of the widest
ranges of polishing resistance of the rock types quarried in the UK (Thompson, Burrows,
Flavin, & Walsh, 2004).
In comparison with the photographic and scanning electron microscope analyses, the
surface texture measurements (Section 5) were more ambiguous. The methodology
assumed that a sufficiently large number of measurements would allow characterisation
of the surface as a whole, and therefore comparison with Wehner-Schulze friction
measurements. However, although an approximate record of the location of each
measured area was kept, no attempt was made to undertake surface texture
measurements on the specific areas that had been viewed by SEM, so it is not possible
to check that measured areas displayed physical signs of change. It is therefore
impossible to verify that the areas measured were replicated from polished parts of the
aggregate surface or if they consisted too much of sunken (and therefore not affected by
polishing) features on the aggregate.
An additional impediment to developing a quantitative roughness measurement is the
unknown effect of the different surface wavelengths.
It is certainly likely that
measurements of roughness made without any filtering will be heavily influenced by long
wavelength features such as the slope of the surface. The clearest trend was found after
applying a relatively long low pass wavelength filter that removed only the largest
features which is at odds with the accepted theory that skid resistance influencing
features are much smaller. It is noted that the use of roughness parameters that
consider distribution and density of peaks in addition to measurements of peak height
was more successful than the use of parameters that consider peak height alone. In
particular, the use of skewness aligns well with the finding from photographic analysis
TRL
22
PPR538
Published Project Report
that also shows increasing skew in pixel intensity values for increasing polishing level
(Figure 3.3), but this must be viewed in the context of the above observations on the
effect of replication. These are significant findings and future work should develop the
use of this type of parameter further.
TRL
23
PPR538
Published Project Report
8
Conclusions and recommendations
The long term goal of this research is to use high resolution images to assist with the
measurement of skid resistance, through assessment of road surface microtexture. The
image analysis work carried out during this stage of research showed, by demonstrating
correlation between a parameter derived from images of aggregate surfaces and friction
measured, that this may still be achievable (Section 3). However, the analysis also
showed that there are probably more factors involved in the changing appearance of the
surface than the texture alone and these would need to be more fully understood before
developing a system based on this approach.
The main aim for the work carried out under this project was to show a link between
changing texture and changing friction on the surface of aggregate particles. Qualitative
evidence for this had been provided by scanning electron microscopy (Section 4), but
quantitative measurements have been inconclusive (Section 5).
It is recommended that future research could include the following elements:
Further scanning electron microscopy using the current set of replicas to identify
features that change during polishing, and track those changes throughout the
process
Measurement of surface texture on the exact surface areas that have been
observed using SEM to link qualitative and quantitative measurements
Development of methodologies
o
to isolate areas that have been polished from areas that do not come into
contact with polishing rollers
o
to enable photographic image analysis, SEM and texture measurement on
the same surface regions
o
for analysis of in-service road surfaces.
Acknowledgements
The work described in this report was carried out in the Infrastructure Division of the
Transport Research Laboratory. The author is grateful to A Wright who carried out the
technical review and auditing of this report, to R Leach, E Bennett and C Giusca at NPL,
T Parry, P Shipway and N Neate at the University of Nottingham and H Viner and G
Helliwell at TRL for advice and assistance offered during the project.
References
British Standards. (1998). BS EN ISO 4288. Geometric Product Specification (GPS) Surface texture - Profile method: Rules and procedures for the assessment of surface
texture. London: BSi.
British Standards. (2008). BS ISO 25178-2. Geometric Product Specification (GPS) Surface texture: Areal. Part 2: Terms, definitions and surface texture parameters.
London: BSi.
Danzl, R., Helmli, F., & Scherer, S. (2009). Focus Variation - A new technology for high
resolution optical 3D surface metrology. Proceedings of the 10th international conference
of the Slovenian society for non-destructive testing. Ljubljana.
Dunford, A. (2009). PPR393. Measuring skid resistance without contact; 2008-2009
progress report. Crowthorne: TRL.
TRL
24
PPR538
Published Project Report
Dunford, A., & Viner, H. (2006). Measuring skid resistance without contact. Retrieved
January 6, 2009, from Transport Research Foundation publications:
http://www.transportresearchfoundation.co.uk/publications.htm
Shaw, R. (2007). An examination of novel roughness parameters to be used in
conjunction with the HSA slip assessment tool (SAT). London: Health and Safety
Executive (HSE).
Thompson, A., Burrows, A., Flavin, D., & Walsh, I. (2004). The Sustainable Use of High
Specification Aggregates for Skid Resistant Road Surfacing in England. East Grinstead:
Capita Symonds Ltd.
Woodbridge, M. E., Dunford, A., & Roe, P. G. (2006). PPR144, Wehner-Schulze machine:
First UK experiences with a new test for polishing resistance in aggregates. Crowthorne:
TRL.
TRL
25
PPR538
Published Project Report
Appendix A Description of the Wehner-Schulze
machine
This Appendix provides a brief description of the Wehner-Schulze (W-S) test equipment
and test procedure. More details can be found in TRL report PPR144 (Woodbridge,
Dunford, & Roe, 2006), the report on its initial evaluation in the UK.
The W-S test equipment, shown in Figure A.1, was developed during the 1960s in
Germany, at the Technical University of Berlin (TUB), as an alternative laboratory test
procedure for assessing the polishing of aggregates in road surfacings. At that time in
Germany, it was considered that the Polished Stone Value (PSV) test was not
satisfactory because it gave relatively small numerical differences between different
aggregates used in Germany and had poor reproducibility. Since then, however, the
reproducibility of the PSV test has been much improved and, especially in the UK,
considerable experience of the relationship between PSV and skid resistance has been
developed.
Figure A.1 Wehner-Schulze test equipment
As with the PSV test, the W-S procedure is designed to simulate accelerated polishing on
road surfacing materials and test the friction provided by the specimen before and after
that polishing. An important difference between the PSV test and the W-S procedure,
however, is that the latter uses large, flat specimens (usually 225 mm diameter) that
can be obtained from actual road surfaces, asphalt test specimens manufactured in the
laboratory or laboratory-manufactured test plates using aggregate alone. The test is
carried out using a purpose-designed machine that is now available commercially.
There are essentially three processes involved in the complete W-S procedure: friction
testing, polishing and grit-blasting. The specimen (a 225 mm core or a 320 mm by 260
mm rectangular slab) is held in an aluminium mould and attached firmly to the mounting
table in the machine so that the table and specimen surfaces are accurately parallel.
The mounting table can slide between the friction testing station and the polishing
station.
The friction measuring head comprises a metal support onto which three sole plates with
attached rubber sliders are fitted at a regular spacing, each slider being 30 mm long and
TRL
26
PPR538
Published Project Report
14.5 mm wide. In the standard test, the measuring head is accelerated until it is
rotating at 3000 rpm, which is equivalent to a tangential speed for the rubber sliders of
100 km/h. Just before the head has reached the target speed, water at 10°C is sprayed
on to the test surface to attain a theoretical water film thickness of about 0.5 mm and
the assembly is dropped onto the surface of the test specimen from a height of about 10
mm, imparting a pressure of 0.2 Nmm-2, equivalent to 2 bar (29 psi) in tyre pressure.
The test head decelerates to a stop while a proximity sensor system records the rotation
of the head and torque transducers in the mounting table continuously measure the
reaction force. The data is sent directly to a dedicated computer that automatically
calculates the coefficient of friction (using an assumed static load) and speed at any
instant and generates a smoothed friction/speed curve for the test. The standard
reported value is the coefficient of friction at 60 km/h. Before each friction test on a
sample, a friction test on a ‘calibration plate’ of rippled toughened glass is carried out
and this should generate readings within closely defined boundaries.
At the polishing station, three rubber-covered conical form rollers are lowered into
contact with the test surface.
During the polishing operation, each roller is
independently forced onto the test surface at a contact pressure of approximately 0.4
Nmm-2, equivalent to 4 bar (58psi), typical of the tyre pressures of a commercial vehicle.
The mounting bearings are engineered to provide some friction so that, although the
rollers are free to rotate, there is some drag, giving a slight slip of 0.5 to 1.0%. Grooves
about 2 mm wide, 2 mm deep and about 20 mm apart are cut in the roller rubber,
running from the apex to base of the rollers, to simulate tyre treads. In the standard
test, the roller head is rotated at 500 rpm for 1 hour, giving a total of 30,000 revolutions
of the head and 90,000 roller passes over the sample surface. A suspension consisting of
about 5% quartz powder in 95% tap water is mixed at a controlled temperature of 20°C
in a separate tank and is pumped onto the specimen during this process. This replicates
the detritus on a road surface and assists in the polishing process.
The grit blasting stage is usually used to ‘roughen’ the specimen surfaces in order to
simulate the action of winter weather. For this purpose, a custom-designed grit blasting
cabinet is used. The cabinet has several automatic settings which control the duration
and evenness of the blasting over the specimen surface. The grit blasting process can
also be used to clean excess bitumen from new asphalt specimens.
A full cycle of the test (as developed by TUB) for specimens taken from the road has the
following stages:
•
•
•
•
•
•
•
•
Friction test core.
Polish for 1 hour.
Friction test.
Grit blast.
Friction test.
Polish for 1 hour.
Friction test.
Friction test ‘to the limit’.
Determine in situ friction.
Simulate summer polishing.
Simulate winter weathering.
Determine lowest friction.
The initial test on cores taken from the road is carried out because, in Germany (unlike
the UK), there are contractual requirements for skid resistance on newly-laid surfaces.
Experience with the equipment in Germany has suggested that the final level of friction,
after both polishing stages have been carried out, simulates the state of skid resistance
that occurs after four to six years of traffic on aggregates in situ.
Further
experimentation in Germany has mirrored findings using the PSV test in the UK and
observations on the road: the level of skid resistance drops very quickly to an
equilibrium level that reduces only slowly over the remainder of the test, or road service
life.
TRL
27
PPR538