Accepted Manuscript

Influence of computation algorithm on the accuracy of rut depth measurement

Di Wang, Augusto Cannone Falchetto, Matthias Goeke, Weina Wang, Tiantian Li,
Michael P. Wistuba

PII: S2095-7564(17)30077-6
DOI: 10.1016/j.jtte.2017.03.001
Reference: JTTE 117

To appear in: Journal of Traffic and Transportation Engineering (English

Edition)

Please cite this article as: Wang, D., Cannone Falchetto, A., Goeke, M., Wang, W., Li, T., Wistuba, M.P.,
Influence of computation algorithm on the accuracy of rut depth measurement, Journal of Traffic and
Transportation Engineering (English Edition) (2017), doi: 10.1016/j.jtte.2017.03.001.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

1 Original Research Paper

3 Influence of computation algorithm on the

PT
4 accuracy of rut depth measurement

RI
5

SC
6 Di Wanga,b, Augusto Cannone Falchettoa,*, Matthias Goekea, Weina Wangc,

7 Tiantian Lib, Michael P. Wistubaa

8
U
AN
a
9 Department of Civil Engineering, Technische Universität Braunschweig, Braunschweig 38106, Germany
b
10 Key Laboratory for Special Area Highway Engineering of Ministry of Education, Chang’an University, Xi’an
M

11 710064, China
c
12
D

School of Civil Engineering, Chongqing Jiaotong University, Chongqing 400074, China

13
TE

14 Highlights

15 ·The multipoint laser detection technology for rut depth measurement was applied.
EP

16 ·The difference value between straight-edge method and wire line method was calculated.

17 ·The effect of rutting shape and rut depth magnitude on the accuracy of rut depth measurement was
C

18 analyzed.
AC

19
20
21 Abstract

22 Rutting is one of the dominant pavement distresses, hence, the accuracy of rut depth measurements

23 can have substantially impact on the maintenance and rehabilitation (M & R) strategies and funding

24 allocation. Different computation algorithms such as straight-edge method and wire line method, which

1
 ACCEPTED MANUSCRIPT

25 are based on the same raw data, may lead to rut depth estimation which are not always consistent.

26 Therefore, there is an urgent need to assess the impact of algorithm types on the accuracy of rut depth

27 computation. In this paper, a 13-point-based laser sensor detection technology, commonly accepted in

28 China for rut depth measurements, was used to obtain a database of 85,000 field transverse profiles

PT
29 having three representative rutting shapes with small, medium and high severity rut levels. Based on

30 the reconstruction of real transverse profiles, the consequences from two different algorithms were

RI
31 compared. Results showed that there is a combined effect of rut depth and profile shape on the rut

32 depth computation accuracy. As expected, the difference between the results obtained with the two

SC
33 computation methods increases with deeper rutting sections: when the distress is above 15 mm

34 (severe level) the average difference between the two computation methods is above 1.5 mm, normally,

35
U
the wire line method provides larger results. The computation suggests that the rutting shapes have a
AN
36 minimal influence on the results. An in-depth analysis showed that the upheaval outside of the wheel

37 path is a dominant shape factor which results in higher computation differences.

M

38
39 Keywords:
D

40 Pavement distress; Multipoint laser detection; Straight-edge rut depth; Wire line rut depth;
TE

41 Rutting shape; Rut depth magnitude.

42
EP

43
*Corresponding author. Tel.: +49 531 391 62064; fax: +49 531 391 62063.
E-mail addresses: di.wang@tu-braunschweig.de (D. Wang), a.cannone-falchetto@tu-bs.de (A.
C

Cannone Falchetto), m.goeke@tu-braunschweig.de (M. Goeke), wwn0816@yeah.net (W. Wang),

litiantian_2009_9@126.com (T. Li), m.wistuba@tu-bs.de (M. P. Wistuba).
AC

2
 ACCEPTED MANUSCRIPT

44 1 Introduction

45 Rutting is one of the most significant distresses of asphalt pavement. It consists of a permanent surface

46 deformation in the wheel path occurring when pavement materials are under high loading and shear

PT
47 (Haas and Norman, 2001; McGhee, 2004; Sousa et al., 1991). This phenomenon can significantly

48 impact roadway safety since rainwater may fill ruts, eventually leading to loss of traction and friction due

RI
49 to hydroplaning. Therefore, timely decisions and correct solutions for maintenance and rehabilitation (M

50 & R) need to be identified for minimizing the detrimental effects of this distress.

SC
51 The accuracy of rut depth measurement can substantially impact the reliability of performance

52 evaluation, on the selection of M & R strategies, as well as, on the allocation of funding. In some

53
U
countries such as China, the pavement industry is still in the construction climax and the time for
AN
54 massive scale pavement M & R has not come yet. Hence, in the next ten years, the transportation

55 department will face the challenge of implementing a consistent M & R program with an efficient use of
M

56 resources.

57 The original definition of the rut depth is based on the manual straight-edge measurement, however,
D

58 the straight-edge length varies depending on region and countries. Hence, the rut depth estimation is
TE

59 not always consistent. In the US, American Association of State Highway and Transportation Officials

60 (AASHTO) and the Highway Design and Maintenance Standards Model (HDM-III) relies on a 1.2 m
EP

61 straight-edge bar to measure rut depth (Lous, 1995). The American Society for Testing and Materials

62 (ASTM, 2010) suggested a flexible straight-edge between 1.83 m (6 feet) and 3.66 m (12 feet); the
C

63 straight-edge length should stretch across the highest points between adjacent upheavals. However,
AC

64 common straight-edge length was established for a 1.83 m (6 feet) lane in the long-term pavement

65 performance (LTPP) (Miller and Bellinger, 2014). Among the countries using international units (UI)

66 Denmark uses 1.8 m straight-edge, other institutions in countries such as the United Kingdom

67 (Transport Research Laboratory, TRL), Australia, New Zealand, South Africa and some of the British

68 Commonwealth of Nations define the rut depth under a 2.0 m straight-edge (Lous, 1995),. A different

69 approach is used by Strategic Highway Research Program (SHRP, US), Sweden (Stroup et al., 2004)

70 and China (RIOH, 2007), for which the minimum length of a straight-edge is as wide as the driving lane

3
 ACCEPTED MANUSCRIPT

71 of interest. Japan uses the pulling line method, which is based on a flexible wire rather than on a

72 straight-edge (Joseph, 2001).

73 In order to obtain rut depth in a more time-efficient manner, the automatic laser detection technology

74 was introduced. The very first measurement devices consisted only three or five laser sensors, so that

PT
75 the transverse section could only be approximated as a discrete profile. The corresponding computation

76 method was named pseudo rutting (NCHRP, 2004) or AASHTO method (Cole and Shippen, 2005). The

RI
77 assumption of the two methods are similar: the instrumented vehicle runs along the centerline of the

78 driving lane. The wheels are located within the wheel paths, and the laser sensors are just right above

SC
79 the maximum rut section. The rut depth is defined as the relative height difference between the central

80 sensor and the sensors on both sides. These two methods were mainly used in the US (Vedula et al.,

81
U
2002); however, due to the limitations of the assumption and the measurements inaccuracy, it was later
AN
82 superseded.

83 As the technology developed to more refined laser systems, more and more sensors have been
M

84 installed in the detection devices. Simpson (2001a) suggested that a measured transverse section can

85 be considered as a continuous profile when the number of laser sensors is more than nine and such a
D

86 configuration is commonly available in most of the current detection systems. Based on these types of
TE

87 measurement device, two new computation methods were implemented: straight-edge rut depth and

88 wire line rut depth (AASHTO, 2001). Cenek et al. (1994) and Lous (1995) evaluated the difference
EP

89 between the two computation methods showing that the wire line method leads to a larger result when

90 the wheel paths are relative wide. Simpson (2001b) and Li (2006) performed a qualitative comparative
C

91 analysis demonstrating that the shape of the transverse section affects the rut depth results.

92
AC

A different study from the Federal Highway Administration (Joseph, 2001) found that the rut depth

93 obtained from a wire line method is larger than that measured with a 1.8 m or 1.2 m straight-edge.

94 Bennett and Wang (2003) and Wu (2007) explored the possibility of differentiating the transverse rut

95 profile into W-shape and U-shape sections. While no significant differences were observed for W-shape

96 profile, for U-shape sections the wire line method was associated to larger rut depths compared to the

97 straight-edge method. Although the straight-edge and wire line methods were the most commonly used

98 algorithms in rut depth computation, the existing studies provided only qualitative evidence that the

4
 ACCEPTED MANUSCRIPT

99 rutting shape affected the rut depth accuracy. In addition, the separation of rut profile into two classes,

100 W-shape and U-shape, appeared to be too simplistic for thoroughly addressing the impact of the

101 computation algorithm on the actual accuracy of the rut depth.

PT
102 2 Objective and research approach

103 In this paper, the effect of computation algorithms on the accuracy of rut depth measurements is

RI
104 evaluated with the aim of identifying the effective influence of rut depth and rutting shapes to provide

105 estimation on the reliability of the decision for timely M & R actions. For this purpose, two typical rutting

SC
106 shapes obtained from field observations and one virtual rutting shape were used to analyze the impact

107 of rut measurements accuracy for different rut depth magnitude. In the present research, the

108
U
straight-edge rut depth method refers to a 2.0 m straight-edge.
AN
109 This paper is organized as follows. First the 13-point laser bar device and the experimental

110 measurements are introduced. Then, straight-edge rut depth and wire line rut depth algorithms are
M

111 described. Rutting severity level and the seven profile shapes which are conventionally observed in

112 China are presented. Within this set of profiles, the three shapes which showed different rut results
D

113 between the two computation methods (straight-edge and wire line) are further analyzed to understand
TE

114 the influence of the rut depth accuracy. Finally, the profile characteristics affecting the measurement

115 accuracy are identified and the one-dimension rutting shape indexes are proposed for further research.
EP

116 3 Rut depth measurement technology and computation algorithms

C

117 In this section, the 13-point-based laser bar and an extensive dataset of road profiles from the Jiangsu
AC

118 province in China are introduced together with the two most common rut depth computation methods

119 (RIOH, 2007).

120 3.1 13-point-based laser bar and experimental measurements

121 In the recent past, the Research Institute of Highway (RIOH), which is part of the Chinese Ministry of

122 Transport (MOT) has developed an automatic road detection system, named “multifunctional

123 high-speed highway condition monitor system” (CiCS, 2010), equipped with a 13-point laser sensor

5
 ACCEPTED MANUSCRIPT

124 (Fig. 1). This device is based on the relative height measurement at discrete points, where the laser bar

125 is 2300 mm wide and is installed on the vehicle at 300 mm above the pavement. Nine vertical laser

126 sensors are unevenly positioned along the laser bar, with more sensors in the wheel path area and less

127 in the non-wheel path zone. Two additional, oblique laser sensors are installed on both left and right end

PT
128 of the laser bar so that an overall detection width of 3600 mm can be achieved. The detailed layout and

129 laser spacing are shown in Fig. 2.

RI
U SC
AN
130

131 Fig. 1 Instrumented vehicle with 13-point laser bar in the road measuring.
M

132
D
TE
C EP

133
134 Fig. 2 Layout of laser sensors.
AC

135 Table 1 shows the original data of a representative rut transverse profile collected by the

136 13-point-based laser bar. Each profile data consists of three lines. In the first row, "G40" indicates the

137 identification number of the roadway; "A" represents the upper line, while "104" identifies the stake mark

138 which means the profile is located in K1+120.82 m. In the second row, each point shows the vertical

139 height of the pavement measured by the laser sensors with an accuracy of 0.1 mm. The measurement

140 system calculates the relative height of the laser sensors; therefore, the reference point of the profile is

6
 ACCEPTED MANUSCRIPT

141 the lowest sensor which is marked by zero. The third row is the horizontal position of the detecting laser

142 points in units of 1 mm. The sensor D7 in the middle (Fig. 2) is represented by 2500 mm so that the width

143 of the measurement is the difference between the first and the last coordinate.

144 Table 1 Original data of 13-point-based laser bar

PT
Row Raw data Unit
1 G40A, 112, 082 cm

RI
2 258, 267, 157, 118, 102, 145, 224, 195, 114, 47, 0, 26, 122 0.1 mm
3 699, 1073, 1343, 1625, 1875, 2125, 2500, 2875, 3125, 3375, 3659, 3944, 4322 mm
145

SC
146 3.2 Rut depth calculation algorithms

147 The 13 raw discrete points are used to reconstruct an approximate continuous rut’s cross section relying

148
U
on MATLAB (MathWorks, 2015) as follows: each discrete elevation point is connected by a straight line
AN
149 one-by-one to the cross-sectional shape and both endpoints are connected and extended to determine

150 the raw baseline. Then, the corresponding rut depth can be computed with the straight-edge method
M

151 (Hadley and Mayers, 1991) or wire line method (RIOH, 2007, 2008).

152 However, there exists a road camber in the highway, in China. Normally, the slope angle, α , is
D

153 approximately 3‰ in the non-super elevation section, hence, the original baseline is not horizontal. In
TE

154 China, seven typical rutting shapes were identified in "field test methods of subgrade and pavement for

155 highway engineering" (RIOH, 2008), but their base lines are a horizontal. As shown in Fig. 3, there are
EP

156 two definitions for rut depth: perpendicular to the datum of the elevation measurements, which is the one

157 related to the horizontal line, and the other one is perpendicular to the measurement bar (straight-edge
C

158 or wire) associated to the raw baseline. It is obviously that the different definitions lead the differences of

159
AC

related rut depths.

160

7
 ACCEPTED MANUSCRIPT

161 Fig. 3 Implication of the baseline.

162 Bennett and Wang (2003) suggested that there is a cosine relation between the two rut depths, since

163 α is a very small value, then the cos( α ) 1. Further calculation showed that the difference value

164 between the two results does not exceed 0.01 mm while the rut depth is less than 50 mm. Therefore, the

PT
165 different baselines have a minimal influence on the results. In this paper, we correct the original baseline

166 to horizontal as the national standard suggested, therefore, in this paper the baseline is a horizontal line.

RI
167 3.2.1 Straight-edge method

168 The straight-edge rut depth algorithm is based on the Strategic Highway Research Program (SHRP)

SC
169 algorithm in Hadley and Myers (1991). The analysis starts at sensor one (D1) (Fig. 2) which is the

170 closest to the pavement kerb. It progresses until the rutting in one wheelpath is established. It is then

171
U
repeated for the second wheelpath starting at the right most sensor and moving downwards. Once a
AN
172 viable placement point had been established, the vertical distances of all intermediate placement points

173 were established.

M

174 In this paper, according to the SHRP project (Simpson, 2003), the straight-edge is defined as an

175 imaginary straight ruler which stretches across the road profile, however, the length of the straight-edge
D

176 is limited. In this study, the straight-edge is defined as 2.0 m long. As shown in Fig. 4, the straight-edge
TE

177 touches the highest points/peaks of the cross section, while the wheel path is beyond the ruler’s length,

178 and the straight-edge only goes across the highest points between adjacent upheavals.
EP

179 (a)
C
AC

180

181 (b)

182
183 Fig. 4 Illustration of straight-edge method in different rutting shapes. (a) First rutting shape. (b) Second rutting shape.

8
 ACCEPTED MANUSCRIPT

184 3.2.2 Wire line method

185 According to the Highway Performance Assessment Standards (RIOH, 2007), the wire line is defined as

186 an imaginary line which stretches across the entire road profile, while both ends of the line overlap with

187 the endpoints of the cross section. According to Fig. 5, the wire line touches the highest points/peaks of

PT
188 the cross section, while the rut depth is given by the maximum vertical distance between the road profile

189 and the wire line.

RI
190 (a)

U SC
191
AN
192 (b)
M

193
D

194 Fig. 5 Illustration of wire line method in different rutting shapes. (a) First rutting shape. (b) Second rutting shape.
TE

195 4 Rut depth magnitude and typical rutting shapes in China

EP

196 In this section, the standard category of rutting magnitude used in China and the typical rutting shapes

197 identified by the Chinese national standard are presented. Then, the four shapes which led to different

198
C

results in two computation methods are analyzed and illustrated.

AC

199 4.1 Rutting severity level definition

200 The rutting severity magnitude is commonly defined in terms of different ranges in rut depth (Fwa and

201 Ong, 2008; Li, 2012). According to the Chinese national standards (RIOH, 2007), a pavement is

202 affected by rutting phenomena when it presents a permanent deformation of 10 mm or larger. And when

203 the depths are lower than 10 mm, the pavement won’t be affected by rutting phenomena. Rutting

204 between 6 and 10 mm can be associated to the small level, and rutting between 10 and 15 mm is

9
 ACCEPTED MANUSCRIPT

205 defined as medium level rutting. For these two cases, M & R judgement is allocated based on the

206 Pavement Condition Index (PCI) and special maintenance treatment for rutting is not required.

207 Deformations over 15 mm are considered in the category of high severity, then maintenance actions

208 have to be planned immediately.

PT
209 In this study, an expressway road, 10.7 km long, affected by rutting distress, was selected and

210 investigated. The road was designed and built in 2008, in Jiangsu province of China. The pavement is

RI
211 21 m wide and no maintenance or re-vamping was performed previously. The layers’ structural package

212 is as follows: 5 cm of SMA-16 wearing course, 7 cm AC-25 binder layer and 10 cm AC-30 base layer.

SC
213 This road was tested with the 13-point-based laser bar and the cross section profile was measured and

214 recorded every 20 cm.

U
AN
215 4.2 Typical profile shapes in China

216 In order to analyze the sensitivity of computation methods, the rutting shapes which results in different
M

217 rut depth computations need to be first identified. Due to the computation approach, the differences are

218 observed only for lane without upheaval in center. In this case, the wire line method provides larger rut
D

219 values than the straight-edge method. Seven typical rutting shapes in China are identified in the current
TE

220 national standard (RIOH, 2008). According to the computation approaches, four profile shapes show

221 differences in rut depth: Types 4, 5, 6 and 7 (Fig. 6).

C EP
AC

222
223 Fig. 6 Rutting profile shapes in China.

224 The rutting shapes depend on a variety of factors which are linked to a series of dominant

225 phenomena. According to previous studies (N.D. Lea International Ltd, 1995; Sha, 2001), the dominant

226 phenomena can be divided into four types: structure deformation (SD), plastic deformation (PD),

10
 ACCEPTED MANUSCRIPT

227 surface abrasion (SA) and densifications deformation (DD). Since a pavement structure consisting of a

228 thin asphalt surface with thick semi-rigid base is widely used for highway pavements in China, and due

229 to the combined interaction of climate and construction with traffic load, the major rutting phenomena in

230 China can be restricted to PD and DD. In this study, only three typical shapes were collected from the

PT
231 database, which could be considered as PD (Types 3, 5) and DD (Type 6) from Fig. 6. As shown in Fig.

232 7, the rutting shapes associated to this specific configuration and conditions are illustrated. And in the

RI
233 figure, the value of deformation (D) refers to the rut depth.

234 (a)

U SC
AN
M

235
236 (b)
237
D
TE
EP

238
239
C

Fig. 7 Rutting shape collected from database. (a) Plastic deformation rutting shape-Type 5. (b) Densifications

240 deformation rutting shape-Type 6.

AC

241 In order to analyze the rutting shapes effect, one extra rutting shape needs be simulated in this

242 research. Zhu (2007) studied the common rutting shape in heavy traffic load in China and the author

243 pointed out that Type 7 rutting shape is very uncommon, since it could be observed only for rural road

244 when a traffic overload occured. Hence, in this study, only Type 4 rutting shape is necessary to be

245 virtually reconstructed. The simulation process is illustrated in Fig. 8.

11
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
246
247 Fig. 8 Simulation process of Type 4 rutting shape.

248 The simulation process is as follows.

M

249 • Type 4 rutting transverse sections with different rut depths were drawn according to the

250
D

definition of Chinese national standard (RIOH, 2008).

251 • The 13-point-based measurement process was simulated with the actual sensors spacing
TE

252 distribution of the 13-point-based laser bar configuration.

253 • The rutting section was reconstructed by connecting the discrete elevation points and the
EP

254 maximum rut depth was then calculated using both straight-edge and wire line methods.
C

255 5 Assessment of rut depth magnitude and rutting shape effect

AC

256 In order to study the effect of the rut depth magnitude and rutting shape on the accuracy of rutting

257 measurements, Types 5 and 6 rutting shapes were used. Type 4 rutting shape was also included in the

258 analysis based on reconstructed profile with a simulated rut depth between 6 mm and 19 mm with 1 mm

259 interval. The two computational methods, straight-edge and wire line, were implemented into a MATLAB

260 (Mathworks, 2015) code and applied to different profiles to compare the potential differences in the

261 estimations of rut depth.

12
 ACCEPTED MANUSCRIPT

262 5.1 Rut depth magnitude effect

263 Fig. 9 presents the differences of rut depth for Types 4, 5 and 6 rutting shapes. In the horizontal axis,

264 each number represents the rut depths magnitude, for example, “8” represents the rutting sections

PT
265 which show rutting larger than 8 mm but smaller than 9 mm.

RI
U SC
AN
266
267 Fig. 9 Differences in rut depth between straight-edge and wire line methods.
M

268 The progressive increasing differences for all the three rutting shapes can be observed. For small and

269 medium rut depth levels, the average difference is below 1.5 mm, hence, both computation methods are
D

270 acceptable. However, much higher differences (up to 2.7 mm) can be found for larger rut depth (19 mm).
TE

271 Such a severe discrepancy is expected to significantly affect the accuracy of the measurements and,

272 therefore, the specific computation method needs to be carefully selected as a satisfactory balance
EP

273 between road users’ safety and maintenance costs. According to previous researches (Fwa and Ong,

274 2008; Guo et al., 2013), the rutting can easily induce the occurrence of hydroplaning risk, and the safety
C

275 critical rut depth will decrease compared with the dry condition, hence, the authors suggest engineers to
AC

276 use the wire line method in M & R decision.

277 It is, therefore, undeniable that rut depth plays an important role for the selection of the most

278 appropriate computation method. To further evaluate this effect, a sensitivity coefficient analysis based,

279 analysis of variance analysis (ANOVA) was used (Shi, 2012). Statistical significance level was set to

280 α = 0.05. This parameter is associated to the output of statistical analysis, p-value, which represents
281 the parameter discriminating the actual significance of the specific test. When p-value is smaller than

282 the significance level ( α = 0.05), then it may be concluded that there is a statistical significant difference

13
 ACCEPTED MANUSCRIPT

283 among the groups; otherwise, the groups compared are statistically equivalent.

284 Table 2 ANOVA on rut depth.

Analysis output Sum of squares df. Mean square F Sig.

Between groups 7.913 12 0.659 6.819 0

PT
Within groups 3.771 39 0.097
Total 11.684 51
285

RI
286 As shown in Table 2, F-test statistics is 6.819, and the corresponding probability p-value is 0 (bold

287 section in Table 2). This suggests that the rut depth has a statistical significance impact on the

SC
288 computation algorithm.

289 5.2 Rutting shape effect

U
AN
290 As previously mentioned, the peculiar rutting shape represented a notable factor (Bennett and Wang,

291 2003; Wu, 2007) affecting the rut depth measurement accuracy. Hence, ANOVA was used to analysis

292
M

its significance.

293 Table 3 ANOVA on rutting shape.

D

Analysis output Sum of squares df. Mean square F Sig.

Between groups 1.362 2 0.454 2.112 0.111
TE

Within groups 10.322 36 0.215

Total 11.684 38
294
EP

295 As shown in Table 3, F-test statistics value is 2.112 and the corresponding probability p-value is 0.111

296 (bold section in Table 3), which means that rutting shape is not affecting the measured accuracy.
C

297 Therefore, it is necessary to study the difference between each pair of rutting shapes. A multiple

298
AC

comparison statistical test based on the least significant difference (LSD) method was used (Shi, 2012).

299 Table 4 shows the comparison results between each type of three different rutting shapes.

300
301
302
303
304

14
 ACCEPTED MANUSCRIPT

305 Table 4 LSD-based multiple comparisons of three rutting shape types.

95% confidence interval

Rutting shape type Mean difference Std. error Sig.
Lower bound Upper bound
4 and 5 0.02308 0.18189 0.900 -0.3426 0.3888
4 and 6 0.38077 0.18189 0.042 0.0151 0.7465

PT
5 and 6 0.40385 0.18189 0.031 0.0381 0.7696
306
307 As shown in Table 4, p-value is 0.900 for group of Types 4 and 5, which means that there is no

RI
308 statistical significance within this group. However, p-value are 0.042 and 0.031 (bold section in Table 4)

SC
309 for groups of Types 4 and 6, 5 and 6, respectively. Hence, there is a significant influence of the rutting

310 shape for these two groups confirming a substantial similarity between Types 4 and 5.

311 As mentioned before, one of the dominant rutting causes is plastic deformation, which leads to

312
U
upheaval on the right side along the driving direction. As illustrated in Fig. 6, the upheaval outside of the
AN
313 right wheel path is present in both Types 4 and 5, while there is no upheaval in Type 6. Hence, the

314 results are most likely associated to the presence of the upheaval outside of the right wheel path, which
M

315 will lead to smaller differences between the two computation algorithms.
D

316 6 Conclusions
TE

317 In this paper, the effect of computation algorithms on the accuracy of rut depth is evaluated with the aim

318 of identifying the influence of rutting shapes, and rut depth on the reliability of the decision for timely
EP

319 maintenance and rehabilitation decision and activities. For this purpose, two field rutting shapes and

320 one simulated rutting profile typically observed in China were selected to analyze the impact on rut
C

321 measurements accuracy. Based on the analysis performed, the following conclusions can be drawn:
AC

322 (1) According to the current Chinese national standard (RIOH, 2008), four profile shapes with

323 upheaval show different rut depth estimation between the two computation algorithms.

324 (2) The rut depth estimation obtained with the wire line method results are, in most cases, larger

325 than those derived from the straight-edge method.

326 (3) The rut depth leads to significant differences between the two computation methods. The

327 different values show a progressively increasing trend for all the three rutting shapes, it is up to

328 2.7 mm when rut depth magnitude is 19 mm.

15
 ACCEPTED MANUSCRIPT

329 (4) The selection of the computation methods should be carefully selected as a satisfactory balance

330 between road users’ safety and maintenance costs. Due to the safety consideration, the findings

331 of this study suggest engineers to use the wire line method in M & R decisions.

332 (5) The overall rutting shape does not significantly affect the rut depth measurement accuracy.

PT
333 However, the upheaval on the right wheel path represents a dominant factor which impact on the

334 results, and may lead to smaller differences between the two computation algorithms.

RI
335 According the analysis above, the difference estimation between two algorithms is a combined effect

336 of rutting shape and rut depth. For rutting shape, the findings of this study suggest the upheaval outside

SC
337 of the right wheel path is the dominant factor, however, it is not an accurate parameter and only

338 qualitative study is conducted in this paper. Therefore, the geometry of rutting shape characteristic

339
U
should be established and quantized analysis in further research. For rutting depth, the difference value
AN
340 between two algorithms is gradually changed in this paper, serious cases should be studied to find out

341 the critical magnitude. And additional transverse profiles derived from 13-point laser bars with more rut
M

342 types should be further analyzed, using the proposed method to quantify the potential errors and further

343 understand the impact of rut type on rut depth measurement error.
D

344 Acknowledgments
TE

345 The authors would like to thank Prof. Yichang Tsai from Georgia Institute of Technology for his technical
EP

346 support. The research was sponsored by China Postdoctoral Science Foundation (2014M562287), and

347 National Natural Science Foundation of China (51508034, 51408083, 51508064).

C

348 References
AC

349 AASHTO, 2001. Standard Practice for Determining Maximum Rut Depth in Asphalt Pavements.

350 AASHTOPP38. American Association of State Highway and Transportation Officials, Washington

351 DC.

16
 ACCEPTED MANUSCRIPT

352 ASTM, 2010. Standard Test Method for Measuring Rut-depth of Pavement Surfaces Using a

353 Straightedge. E1703/E1703M-10. American Society for Testing and Materials, West

354 Conshohocken.

355 Bennett, C.R., Wang, H., 2003. Harmonising Automated Rut Depth Measurements. California Transit

PT
356 Association, Sacramento.

357 Cenek, P.D., Patrick, J.E., McGuire, J.F., et al., 1994. New Zealand experience in comparing manual

RI
358 and automatic pavement condition rating systems. In: The 3rd International Conference on

359 Managing Pavements, San Antonio, 1994.

SC
360 CiCS, 2010. Equipment Introduction. Available at: http://www.xssj.cc/news/2010429/n785955.html

361 (Accessed 26 February 2016).

362
U
Cole, M., Shippen, N., 2005. Automated Data Collection Equipment for Monitoring Highway Condition.
AN
363 FHWA-OR-RD-05-10. Oregon Department of Transportation, Salem.

364 Fwa, T.F., Ong, G.P., 2008. Wet-pavement hydroplaning risk and skid resistance: analysis. Journal of
M

365 Transportation Engineering 134(5), 182–190.

366 Guo, X., Zhang, C., Cui, B., et al., 2013. Analysis of impact of transverse slope on hydroplaning risk
D

367 level. Procedia-Social and Behavioral Sciences 96, 2310-2319.

TE

368 Haas, R., Norman, W., 2001. Reinventing the pavement management wheel. In: 5th International

369 Conference on Managing Pavements, Seattle, 2001.

EP

370 Hadley, W.O., Myers, M.G., 1991. Rut Depth Estimates Developed from Cross Profile Data. AU-179.

371 Texas Research and Development Foundation, Austin.

C

372 Joseph, B., 2001. A Synopsis on the Current Equipment Used for Measuring Pavement Smoothness.

373
AC

Federal Highway Administration, Washington DC.

374 Li, F., 2012. A Methodology for Characterizing Pavement Rutting Condition Using Emerging 3D Line

375 Laser Imaging Technology (PhD thesis). Georgia Institute of Technology, Atlanta.

376 Li, Q., 2006. Highway Pavement Rutting Detection, Evaluation and Prediction Technology (PhD thesis).

377 Southeast University, Nanjing.

17
 ACCEPTED MANUSCRIPT

378 Lous, K., 1995. Modeling Rutting in Flexible Pavements in HDM-4. Available at:

379 http://www.lpcb.org/index.php/documents/pavement-deterioration1459-1995-rut-depth-modelling-r

380 ut-depths-in-flexible-pavement/file (Accessed 1 August 2016).

381 MathWorks, 2015. MATLAB R2015a. Available at: http://de.mathworks.com/ (Accessed 1 August

PT
382 2016).

383 McGhee, K.H., 2004. Automated Pavement Distress Collection Techniques: a Synthesis of Highway

RI
384 Practice. NCHRP Synthesis 334. TRB, Washington DC.

385 Miller, J.S., Bellinger, W.Y., 2014. Distress Identification Manual for the Long-term Pavement

SC
386 Performance Program. FHWA-HRT-13-092. U.S. Department of Transportation, Washington DC.

387 NCHRP, 2004. Guide for Mechanistic-Empirical Design of New and Rehabilitated Pavement Structures.

388
U
Project 1-37A. National Research Council, Washington DC.
AN
389 N.D. Lea International Ltd, 1995. Modeling Road Deterioration and Maintenance Effects in HDM-4.

390 Asian Development Bank, Vancouver.

M

391 Research Institute of Highway (RIOH), 2007. Highway Performance Assessment Standards. JTG

392 H20-2007. Ministry of Transport of the People’s Republic of China, Beijing.

D

393 Research Institute of Highway (RIOH), 2008. Field Test Methods of Subgrade and Pavement for
TE

394 Highway Engineering. JTG E60-2008. Ministry of Transport of the People’s Republic of China,

395 Beijing.
EP

396 Sha, Q.L., 2001. Premature Damage and Preservative Measure of Bituminous Pavement on

397 Expressway. China Communications Press, Beijing.

C

398 Shi, L.W., 2012. From Beginning to Professorial. Tsinghua University Press, Beijing.

399
AC

Simpson, A.L., 2001a. Characterization of Transverse Profiles. FHWA/RD 01-024. Federal Highway

400 Administration, Washington DC.

401 Simpson, A.L., 2001b. Measurement of Rutting in Asphalt Pavements (PhD thesis). The University of

402 Texas at Austin, Austin.

403 Simpson, A.L., 2003. Measurement of a rut. In: TRB 82nd Annual Meeting, Washington DC, 2003.

404 Sousa, J.B., Craus, J., Monismith, C.L., 1991. Summary Report on Permanent Deformation in Asphalt

405 Concrete. SHRP-A-318. U.S. National Research Council, Washington DC.

18
 ACCEPTED MANUSCRIPT

406 Stroup, G., Gudamettla J., Hays J., 2004. Profile, rut depth and cross slope measurements using the

407 ARRB walking profiler. In: TRB 83rd Annual Meeting, Washington DC, 2004.

408 Wu, H.Y., 2007. Research on the Method of Extracting and Calculating the Characteristics of Pavement

409 Rut (Master thesis). Harbin Institute of Technology, Harbin.

PT
410 Zhu, Y.S., 2007. Asphalt Pavement Rutting Prediction Model in Heavy Traffic (PhD thesis). Tongji

411 University, Shanghai.

RI
412 Vedula, K., Hossain, M., Reigle, J., et al., 2002. Comparison of 3-point and 5-point rut depth data

413 analysis. In: The Pavement Evaluation Conference, Roanoke, 2002.

SC
414
415

U
AN
M
D
TE
C EP
AC

19
 ACCEPTED MANUSCRIPT

PT
416
417 Di Wang has been a PhD student at Institut für Straßenwesen der TU Braunschweig (ISBS) of

RI
418 Technische Universität Braunschweig since May 2015. He was a research associate in the School of

419 Highway at Chang’an University, Xi’an, China. He received a Master’s degree and a Bachelor’s degree

SC
420 from Chang’an University. His research interests are characterization and modeling of asphalt materials

421 at low temperatures, asphalt materials recycling and diffusion process of rejuvenators and fresh binder

U
422 in the aged binder contained in reclaimed asphalt pavement materials.
AN
423
424
M
D
TE

425
426 Augusto Cannone Falchetto has been a research associate at ISBS since 2013. He had been a
EP

427 research associate in the Department of Civil Engineering at the University of Minnesota, USA from

428 2008 to 2013.He received his PhD degree in Civil Engineering (minor Statistics) in University of
C

429 Minnesota, USA. His research interests are characterization and modeling of asphalt materials at low
AC

430 temperature, asphalt materials recycling, and size effect and scaling of quasi-brittle material.

431

20