USF Libraries
USF Digital Collections

Development and evaluation of an inertial based pavement roughness measuring system

MISSING IMAGE

Material Information

Title:
Development and evaluation of an inertial based pavement roughness measuring system
Physical Description:
Book
Language:
English
Creator:
Hu, Fengxuan
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Repeatability
spectral analysis
Iri
Rn
Speed impact
Dissertations, Academic -- Civil Engineering -- Doctoral -- USF
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Abstract:
ABSTRACT: Roughness is an important indicator of pavement riding comfort and safety. It is a condition indicator that should be carefully considered when evaluating primary pavements. At the same time, the use of roughness measurements plays a critical role in the pavement management system.There are many devices used for roughness evaluation. The major tools used for road roughness quantify are the road profilers. In the thesis research, in order to obtain useful pavement surface condition data for pavement evaluation, an inertial based pavement roughness measuring system was developed with the combination of modern sensor technology and computer technology. The research will focus on the development of new method to get the profile in order to improve the repeatability of the inertial based pavement roughness system, the hardware design and the software development which is used for data sampling and data analysis. Finally maximum entropy spectral analysis method was used to evalu ate the road profile spectrum.In order to get evaluate the accuracy and correction of the laser profiler system, different roughness devices (including Dipstick, direct type profiler and the laser profiler developed) were operated in the test sites. The research focused on several performance measures, such as repeatability (before and after new method analysis), impact of operating speed and sample interval, correlativity and etc. IRI from these devices were analyzed to evaluate the correlativity between these devices. Some regression models were developed in this research. Test results show that the new method can improve the repeatability of the profiler system. The laser profiler system has good repeatability and the operating speed and sample interval do not have a significant impact on the inertial based roughness measuring system. With the reliable results, the system is ready to be used in the field application within the speed and sample interval range. Through the spectrum an alysis, it shows that the spectrum has a qualitative relation with pavement roughness conditions.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2006.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Fengxuan Hu.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 106 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001798404
oclc - 162019203
usfldc doi - E14-SFE0001641
usfldc handle - e14.1641
System ID:
SFS0025959:00001


This item is only available as the following downloads:


Full Text

PAGE 1

Development and Evaluation of an Iner tial Based Pavement Roughness Measuring System by Fengxuan Hu A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Civil & Environmental Engineering College of Engineering University of South Florida Major Professor: Jian John Lu, Ph.D. Ram Pendyala, Ph.D. Manjriker Gunaratne, Ph.D. Steve Polzin, Ph.D. Lihua Li, Ph.D. Date of Approval: May 31, 2006 Keywords: repeatability, spectral an alysis, iri, rn, speed impact Copyright 2006, Fengxuan Hu

PAGE 2

DEDICATION This dissertation is gratef ully dedicated to my wife, Hong Xu, who made all of this possible, for her unfailing love and efforts.

PAGE 3

ACKNOWLEDGMENTS I am here to show my sincere gratitude to my major professor, Dr. Jian John Lu for his guidance in my academic studies. I would like to take this opportunity to thank Dr. Manjriker Gunaratne, Dr. Ram Pendyala, Dr. St eve Polzin and Dr. Lihua Li for taking the time and effort to ensure my work was of quality. Their suggestions and wise counsel were invaluable. Special thanks also go to Mr. Pan Liu, Mr. Zhenyu Wang and Mr. Fatih Pirinccioglu for their help dur ing the field data collection.

PAGE 4

i TABLE OF CONTENTS LIST OF TABLES iii LIST OF FIGURES v ABSTRACT viii CHAPTER 1 INTRODUCTION 1 1.1 Research Background 1 1.2 Basic Concepts 3 1.2.1 Profiles 3 1.2.2 Profile Index 5 1.2.3 Roughness Definition 5 1.2.3.1 International Roughness Index (IRI) 6 1.2.3.2 Ride Number (RN) 7 1.2.4 Signal Processing, Filter, Power Spectrum Analysis 7 1.3 Research Objective 8 CHAPTER 2 LITERATURE REVIEW 9 2.1 Overview 9 2.2 Roughness Measurement Systems 12 2.2.1 Class I System 14 2.2.1.1 Rod and Level 14 2.2.1.2 Dipstick 15 2.2.2 Class II System 16 2.2.2.1 K. J. Law Profilometer 16 2.2.2.2 APL Profilometer 17 2.2.2.3 South Dakota Profiler 18 2.2.3 Class III System 19 2.2.3.1 BPR Roughometer 19 2.2.3.2 Laser Profiler 20 2.3 SDHPTs Results 22 2.3.1 Background 22 2.3.2 Correlation Analysis and Roughness Calibration Models 24 2.3.3 Conclusion 27 2.4 Computation Methods of Inertial Based Profiling System 28 CHAPTER 3 METHODOLOGY 30 3.1 Introduction 30 3.2 Methods to Get Road Profile 30

PAGE 5

ii 3.3 IRI and RN Models and Algorithms 32 3.3.1 Quarter Car Model 32 3.3.2 Calculation of IRI 35 3.3.3 Calculation of RN 38 3.4 Digital Filter 41 3.5 Maximum Entropy Spectrum Analysis 47 CHAPTER 4 SYSTEM DEVELOPMENT 53 4.1 System Requirements 53 4.2 Hardware Description 58 4.2.1 Laser Profiler Introduction 58 4.2.2 Hardware Architecture 61 4.3 Software Development 62 4.3.1 Overview 62 4.3.2 Calibration Module 64 4.3.3 Data Sample Module 67 4.3.4 Data Analysis Module 68 4.3.5 Configuration Module 72 CHAPTER 5 SYSTEM VERIFICATION 75 5.1 Principle of Sensors Measurement 75 5.2 Sensors Verification 77 5.3 Software Output Verification 79 CHAPTER 6 DATA COLLECTION AND DATA ANALYSIS 82 6.1 Consideration for Field Data Collection 82 6.1.1 Test Consideration for Direct Type Profiler 84 6.1.2 Test Consideration for Laser profiler 84 6.2 Data Analysis 85 6.2.1 Repeatability 85 6.2.2 Impact of Sample Speed 88 6.2.3 Impact of Sample Interval 90 6.2.4 Correlation Analysis 92 6.3 Maximum Entropy Spectral Analysis 95 CHAPTER 7 SUMMARY, CONCLUSION AND RECOMMENDATION 101 7.1 Summary 101 7.2 Conclusion 103 7.3 Recommendation 104 REFERENCES 105 ABOUT THE AUTHOR End Page

PAGE 6

iii LIST OF TABLES Table 5-1 Incremental Encoder Calibration Results 77 Table 5-2 Encoder Verification Results 78 Table 5-3 Distance Sensor Calibration Results 78 Table 5-4 Distance Sensor Verification Results 79 Table 5-5 IRI Results 80 Table 5-6 RN Results 81 Table 6-1 IRI Values Between Repeated Runs with Trend Remove Algorithm 85 Table 6-2 RN Values Between Repeated Runs with Trend Remove Algorithm 86 Table 6-3 Re Values of IRI and RN for Each Section with Trend Remove Algorithm 86 Table 6-4 IRI Values Between Repeated Runs without Trend Remove Algorithm 87 Table 6-5 Re Values with and without Trend Remove Algorithm 87 Table 6-6 IRI Values at Different Operating Speeds 88 Table 6-7 RN Values at Different Operating Speeds 88 Table 6-8 IRI Values of Laser High Speed Profiler at Different Sampling Interval 91 Table 6-9 RN Values of Laser High Speed Profiler at Different Sampling Interval 91

PAGE 7

iv Table 6-10 IRI Values Collected by Laser Profiler, Dipstick and Direct Type Profiler 93

PAGE 8

v LIST OF FIGURES Figure 1-1 Longitudinal Profile 4 Figure 2-1 Rod and Level 14 Figure 2-2 Dipstick 15 Figure 2-3 APL Profilometer 17 Figure 2-4 BPR Roughometer 19 Figure 2-5 ICC Laser Profiler 21 Figure 2-6 ARAN Laser Profiler 21 Figure 3-1 Quarter-car Model 33 Figure 3-2 IRI Roughness Scale 38 Figure 3-3 Subjective Rating Scales for Roads 39 Figure 3-4 Sensitive of RN to Wave Number 41 Figure 3-5 Pavement Profile 42 Figure 3-6 Measured Profile 42 Figure 3-7 Filtered Profile 43 Figure 3-8 Effect of Moving Average Filter 44 Figure 4-1 The Inertial Based Laser Profiler 59 Figure 4-2 Laser Profiler 60 Figure 4-3 Prototype of Laser Profiler 60 Figure 4-4 System Diagram of Laser Profiler 61

PAGE 9

vi Figure 4-5 LIPRES Main Interface Form 63 Figure 4-6 System Software Architecture 63 Figure 4-7 Detail Functions of System Calibration Module 64 Figure 4-8 Accelerometer Calibration Form 65 Figure 4-9 Laser Height Distance Sensor Calibration Form 65 Figure 4-10 Pulse Sensor Calibration Form 66 Figure 4-11 Laser Height Sensor Amplifier Calibration Form 66 Figure 4-12 Sample Setup Form 67 Figure 4-13 Data Sample Form 68 Figure 4-14 Data Analysis Module Details 69 Figure 4-15 Data Analysis Form 69 Figure 4-16 Profile Display Form 70 Figure 4-17 Whole Section Analysis Results 70 Figure 4-18 Subsection IRI 71 Figure 4-19 Subsection IRI Results Form 71 Figure 4-20 System Configuration Form 72 Figure 4-21 Detail of System Configuration Functions 72 Figure 4-22 System Parameters Setup Form 73 Figure 4-23 Filter Setup Form 73 Figure 4-24 Roughness Index Parameters Form 74 Figure 5-1 Rotary Incremental Encoder 76 Figure 5-2 RoadRuf Output 81 Figure 6-1 Operating Speed Impact on IRI 89

PAGE 10

vii Figure 6-2 Operating Speed Impact on RN 90 Figure 6-3 Sample Interval Impact on IRI 91 Figure 6-4 Sample Interval Impact on RN 92 Figure 6-5 IRI Correlation Between Laser Profiler and Dipstick 94 Figure 6-6 IRI Correlation Between Laser Profiler and Direct Type Profiler 95 Figure 6-7 Spectral Curve of Good Pavement Condition 96 Figure 6-8 Spectral Curve of Bad Pavement Condition 97 Figure 6-9 Pavement Profile Curve with IRI = 2.19 m/km 97 Figure 6-10 Pavement Profile Curve with IRI = 4.36 m/km 98 Figure 6-11 Spectral Curve of Pavement Profile with High RN 99 Figure 6-12 Spectral Curve of Pavement Profile with Low RN 99

PAGE 11

viii DEVELOPMENT AND EVALUATION OF AN INERTIAL BASED PAVEMENT ROUGHNESS MEASURING SYSTEM Fengxuan Hu ABSTRACT Roughness is an important indicator of pave ment riding comfort and safety. It is a condition indicator that should be carefully considered when evaluating primary pavements. At the same time, the use of roughness measurements plays a critical role in the pavement management system. There are many devices used for roughness evaluation. The major tools used for road roughness quantify are the ro ad profilers. In the thesis re search, in order to obtain useful pavement surface condition data fo r pavement evaluation, an inertial based pavement roughness measuring system was deve loped with the comb ination of modern sensor technology and computer technology. The research will focus on the development of new method to get the profile in order to improve the repeatability of the inertial based pavement roughness system, the hardware desi gn and the software development which is used for data sampling and data analysis Finally maximum entropy spectral analysis method was used to evaluate the road profile spectrum. In order to get evaluate the accuracy a nd correction of the laser profiler system, different roughness devices (including Dipstick, direct type profiler and the laser profiler

PAGE 12

ix developed) were operated in th e test sites. The research focused on several performance measures, such as repeatabili ty (before and after new me thod analysis), impact of operating speed and sample interval, correlativity and etc. IRI from these devices were analyzed to evaluate the correlativity betw een these devices. Some regression models were developed in this research. Test resu lts show that the new method can improve the repeatability of the profiler system. The laser profiler system has good repeatability and the operating speed and sample interval do not have a significant impact on the inertial based roughness measuring system. With the reli able results, the system is ready to be used in the field application within the speed and sample interval range. Through the spectrum analysis, it shows that the spectru m has a qualitative relation with pavement roughness conditions.

PAGE 13

1 CHAPTER 1 INTRODUCTON 1.1 Research Background Pavement roughness is one of the most important performance measures for pavement surface performance conditions. Pavement roughness is also an important indicator of pavement riding comfort and safety. Roughness condition has been used as the criteria for accepting new construction of pavement (including ov erlay) and also as the performance measure to quantify the surf ace performance of existing pavements in a pavement management system at both networ k level and project level. For example, roughness can be used for dividing the network into uniform sections, establishing value limits for acceptable pavement condition, and setting maintenance and rehabilitation (M&R) priorities. Roughness measurements are us ed to locate areas of critical roughness and to maintain construction quality control. The need to measure roughness has brought a wide of instruments on the market, covering range from rather simple devices to quite complicated systems. In the past decades, roughness measurement instruments had become the everyday tools for measuring road roughness. The majority of States now own pavement roughness measurement systems. A substantial body of knowledge exists for the field of system

PAGE 14

2 design and technology. There are also many proven methods for analyzing and interpreting data similar to the measurement results obtained from these systems. By far, the major tools app lied in the road roughness quant ify is the road profilers. A variety of devices are ava ilable today to measure a road profile. These devices range from the hand-held Dipstick profilers, highspeed, vehicle-based profilers and ResponseType Systems. The former devices are based on mathematical modeling of the measured pavement surface profiles so the result indices are repeatable. However, the latter systems that were also called as road meters are always a passenger car, a van, a light truck, or a special trailer. Engineer inst alled devices to record suspen sion stroke as a measure of roughness, normally it is a transducer that accumulates suspension motions and is known as response-type road roughne ss measuring system (RTRRMS). Response-type indices are vehicle dependent and are not repeatable, even when the same vehicle is used -due to change in the vehicle's characteristics over time and drivers driving behavior. At the same time, difficulties exist in the correlation and transferability of measures from various instruments and the cali bration to a common scal e, a situation that is exacerbated through a large number of factors that cause variations between readings of similar instruments, and even for the sa me instruments. The need of correlation and calibration led to the advent of the Intern ational Road Roughness Experiment (IRRE) in Brazil in 1982, which was also led to pub lish of International Roughness Index (IRI). The research leading to the developmen t of roughness measuring equipment dates back more than 60 years. Early profilers were time and labor consuming, required testing at very slow speeds. With the help of the development of sensors technology and computer technology, it is no longer the case now adays. In this research, in order to

PAGE 15

3 obtain useful pavement surface condition data for pavement evaluation in the State of Florida, an inertial-based road roughness m easurement system was developed with the combination of the modern sensors and co mputer. The pavement roughness measuring system uses the vertical acceleration, laser profile and longitudinal distance sensor to measure the profile, filter the profile to include only those waves of interest, and mathematically compute all major types of roughness index. 1.2 Basic Concepts 1.2.1 Profiles The evaluation of the entire pavement su rface is required to define roughness completely. However, for most purposes, r oughness can be divided into three profile components of distortion: transverse, longitudinal, and horizontal. Of particular interest are variations in profile that impart acceleration to the vehicle or occupant and thus influence comfort and safet y. Here, the research will fo cus on longitudinal profiles. Figure 1-1 shows the exampl e of longitudinal profile. Distortions of the pavement surface can generate both vertical and lateral acceleration in the vehicle. Vertical acceler ation is the major contributing factor to occupant comfort and derive s from longitudinal distortion of the pavement profile. Lateral accelerations are the re sult of vehicle roll and yaw. Roll results from rotation about the longitudinal axis of the vehicle whil e yaw is the rotation a bout the vertical axis. The curvature of the roadway, which contribut es to yaw, is normally handled through

PAGE 16

good geometric design. Roll results from diffe rential transverse pavement elevations. Under severe conditions, it can impart an undesirable level of vertical acceleration. Figure 1-1 Longitudinal Profile It is possible to take many profiles for a road, each along a different line. However since approximately 70 percent of vehi cles travel in a well-defined wheel path with the right wheel located 2.5 to 3.5 feet from the paveme nt edge, The wheel tracks of automobiles and trucks are approximately 6 an d 7 feet apart, respectively. Therefore, line measurement of the longitudinal profile on the wheel path provides the best sample of road surface roughness. Furt hermore, comparison between the two wheel paths can provide some measure of the transverse variations that affect roll. Based on the pavement roughness definition, it is concluded that road roughness evaluation requires measurement of the longi tudinal profile of the pavement in the vehicle wheel path. The profile of a road, pavement, or ground can be measured along any continuous imaginary line on the surface an d in order to obtain repeatable measures. It helps to make the line phys ically by using paint. For e ngineering interpretation, the 4

PAGE 17

5 measurements are usually handled with mathematical model that generates a summary statistics, ranged from power spectru m to some type of roughness index. 1.2.2 Profile Index A profile index is a summary number calculated from the data that make up a profile. The profile index is por table, reproducible and stable with time. Almost all road profiling system include two summary roughne ss statistic like, In ternational Roughness Index (IRI) and the estimate of Mean Panel Rating value Ride Number (RN). Although there are also some others roughness indices are used, but IRI and RN are the most used roughness indices because they portable and repr oducible and they are stable with time. So the other roughness indices are not widely av ailable in the form of software and they correlate so highly with IRI, we w ill focus on the former two indices. 1.2.3 Roughness Definition From an auto drivers point of vi ew, pavement roughness is a phenomenon experienced by the passenger and operator of a vehicle. According to the definition (E867) of the American Society of Testing and Materials (A STM), roughness is the deviations of a pavement surface from a true planar surface with characteristic dimensions that affect vehicle dynamics, ri de quality, dynamic loads, and drainage, for example, longitudinal profile, transverse profil e, and cross slope. This definition covers the factors that contribute to road roughness and it is also ve ry broad. However, it does not provide a quantita tive definition or standard scale for roughness, so it still requires a measurement and analysis method for quantif ying distortions of the pavement surface.

PAGE 18

6 Once the measurement and analysis method is selected, individual agencies can establish interpretation scale to determ ine the severity of the roughness level. At the same time, pavement roughness consists of random multifrequency waves of many wavelength and amplitudes. Longitudinal roughness has been defi ned as "the longitudi nal deviations of a pavement surface from a true planar surface w ith characteristic dimensions that affect vehicle dynamics, ride quality and dynamic pave ment load". Pavement profiles, detailed recordings of surface characteristics, are frequently used to characterize roughness. There are several causes of pavement roughness: traffic loading, environmental effects, construction materials and built-in c onstruction irregularities. All pavements have irregularities built into the surface during construction, so even a new pavement that has not been opened to traffic can exhibit roughness. The rough ness of a pavement normally increases with exposure to traffic loading and the envi ronment. Short-wavelength roughness is normally caused by localized paveme nt distress, that is, depression and cracking, at the same time the long-wavele ngth roughness is normally caused by environmental processes in combinati on with pavement layer properties. 1.2.3.1 International Roughness Index (IRI) The International Roughness Index (IR I) was established in 1986 by the World Bank and based on earlier work performed by NCHRP. It was first introduced in the International Road Roughness Experiment (IR RE) that was held in Brazil. IRI is calculated from a measured longitudinal road profile by accumulating the output from a quarter-car model or directly derived from a class 1 or class 2 instruments and divided by the profile length to yield a summary roughness index with units of slope. The IRI has

PAGE 19

7 been reported to be relevant as an indicator of pavement serviceability, independent of the particular equipment used to measure it, it is internationally and geographically transferable and time stable. IRI is often us ed as an accepted standard against which roughness measuring systems are calibrated. 1.2.3.2 Ride Number (RN) Ride Number is a profile index intended to indicate ride ability on a scale similar to PSI. The longitudinal profile measurements taken with a profiler are processed using a computer program to obtain the RN, which matches the mean panel rating of a rating panel. Rider Number is an esti mate of Mean Panel Rating and us es the 0 to 5 scale. It is a nonlinear transform of PI. It is ideally calcul ated from the profiles in the left and right wheel paths of automobiles. The method was to be provided as porta ble software similar to that available for the IRI. Details of Ride Number are handled in computer software. 1.2.4 Signal Processing, Filter, Power Spectrum Analysis Modern profilers produce sequence of numb ers called as a signal. The outputs of the transducers in the prof ile are converted to number s and processed by computer. Signal processing is the mathematical an alysis and transformation of signals. There are mainly two reasons for the signal pro cessing: the first is to improve the quality of a measurement by eliminating unwanted noi se from the data, and the second is to extract information of interest from the signal.

PAGE 20

8 A profile can be considered consists of different wavelengths, varying from a few inches to hundreds of feet. To analyze a profile for roughness, it is important that the profile be filtered to include onl y those waves of interest. A di gital filter is a calculation procedure that transforms a series of number s (a signal) into a ne w series of numbers. 1.3 Research Objective The primary objective of this research is to develop the inertial based roughness system. The goal is to identify the factor s that affect roughness measurement, quantify their effect on repeatability and accuracy. The spectrum analysis was used to evaluate and compare the road roughness condition. More specifically, the objective will consist of the following: 1). to develop the hardware and softwa re of the inertial based laser roughness measuring system; 2). to use a new method to improve and test the repeatability of the developed laser profiler; 3). to test the impacts of operating speed on the repeatability of the laser profiler; 4). to check the sample interval impacts on the road roughness measuring system; 5). to develop the regression models of correlation of the laser roughness measuring system and the direct type profiler, Dipstick; 6). to conduct maximum entropy spectral an alysis of the measured road profile.

PAGE 21

9 CHAPTER TWO LITERATURE REVIEW In order to understand the pavement roughness and pavement roughness measurement problems, current roughness meas uring system situation and an overview of the past studies are presen ted in this chapter. It al so includes the importance of pavement profile, the research about the de velopment of roughness measurement system, the significance of relate d topics and the potentia l study topics expected. 2.1 Overview A lot of studies and research have been done on the subject of pavement roughness since 1960s. In the late 1960s, after Spangler and Kelley developed GMR profilometer at the General Motors Resear ch Laboratory, the routine analysis of pavement profiles began. NCHRP sponsored a study of response-type road roughness measuring system such as the BPR roughometer and vehicles equippe d with Mays rider meters in the late 1970s. An objective of the study was to develop calibration methods for the response type systems. The best correla tion was obtained by using the Golden Car. In the late 1970s, when many state and federal ag encies in charge of monitoring pavement conditions began using profilers to judge the se rviceability of roads, profiling technology found broad application beyond research in the United States. A major advantage of

PAGE 22

10 profilers is that they are capab le of providing a stable and transportable way of measuring roughness. In other words, roughness valu es produced by a vali d profiler can be compared to values from prior years and values measured by other valid profilers. Unfortunately, insufficiencies in profile r design, data processing techniques, and operational practices have compromised the accuracy of profile measurement. Routine analysis of road profiles be gan shortly after General Motor (GM) profilometer was developed in the late 1960s by Spangler and Kelley [Sayers et al, 1986]. Like high-speed profiler today, it could measure true profil e over a range of wavelengths affecting vehicle vibrations. One of the first research applicati ons for this type of system combined measured road profiles with a quart er-car computer model that replicated the Bureau of Public Roads (BPR) Roughometer, a one wheeled trailer with a road meter. GM licensed K. J. Law, Inc. to market the device commercially and continue its development. A commercial version was soon available that included a quarter-car analysis to summarize roughness of the measur ed profiles. Users of early K. J .Law profilometers could choose of two quarter-car data sets. In the late 1970s, NCHRP sponsored a study of response-type road roughness measuring systems such as the BPR Roughomete r and vehicles equipped with Mays ride meters [Gillespie et al, 1980] The results were published in NCHRP 228. An objective of the study was to develop calibration methods for the response-type systems. The researchers, Gillespie and Sayers, concluded that the only valid method was calibration by correlation against a defined roughness index. Considerable research was performed using simulations and experiments to compare alternative referen ce roughness indexes. The candidate analyses included vehicle simulation with 10 alternative sets of

PAGE 23

11 parameters. The best correlati on was obtained by using a vehi cle simulation with a set of parameter values that is often called the Golden Car. In 1982 the World Bank initiated a correlation experiment in Brazil called the international road roughness experiment (IRRE ) to establish correlation and a calibration standard for roughness measurement [Sayers, 1991]. In processing the data, it became clear that nearly all roughne ss-measuring instruments in use through the world were capable of producing measures on the same scale, if that scale had been selected suitable. Accordingly, an objective was added to the re search program: to develop IRI. The main criteria in designing the IRI were that it be re levant, transportable, and stable with time. To ensure transportability, it had to be measured with a wide range of equipment, including response-type systems. To be stab le with time, it had to be defined as a mathematical transform of a measured prof ile. Many roughness definitions were applied to the large amount of test data obtained in the IRRE. The Golden Car simulation from the NCHRP project was one of the candidate references considered, under the condition that a standard simulation speed would be needed to use it for IRI. After processing the IRRE data, the best correlations between a profile index and the response-type systems were found with two vehicle simulations base d on the Golden Car parameters: a quartercar and a half-car. Both gave essentially th e same level of correlation. The quarter-car was selected for the IRI because it could be used with all profiling methods that were in use at that time. The consensus of the researchers and participants is that the standard speed should be 80 km/hr (49.7 mph) because at that simulated speed, the IRI is sensitive to the same profile wavelengths that cause vehicle vibrations in normal highway use. The research findings were highly encouraging and led the World Bank to publish guidelines

PAGE 24

12 for conducting and calibrating roughness meas urements. The researchers (Sayers, Gillespie, Queiroz and Paterson) prepared instructions for using various types of equipment to measure IRI. The guidelines also include computer code for calculating IRI from profile. A companion report desc ribed the IRRE, using many analytical comparisons of algorithms and some sensit ivity analyses. In 1990 FHWA required the IRI as the standard reference for reporting pavement roughness. 2.2 Roughness Measurement Systems Pavement profile may be measured in th e field and evaluated or summarized by computer, or it can be processed through a m echanical response type device. The need to evaluate roughness of pavements was rec ognized in the 1920s. The concept of the functional performance of pavements was deve loped at the AASHO Road Test in the late 1950s. The most straightforward techniques for measuring the profile of a pavement are with precision rod and level survey. However, it is time consuming, costly and limited to the evaluation of short length of pavements. So there are many kinds of pavement roughness measuring system in the United St ates. Generally, the roughness measuring system can be divided into three classes: Class I. Manually operated instruments accurately measure short wavelength profiles of the pavements. The measurement interval is less than equal to 1 foot, and the maximum error is 1.5 percent bias, or 19 inches/mile. In a Class I survey, the longitudinal profile of the wheel path is measured manually using a rod and level, Transportation Road Research La boratory (TRRL) Beam, Face Dipstick, or

PAGE 25

13 similar high-precision device. The measur ed profile is used as a basis for calculating the IRI. Class II: Dynamic direct profiling instruments that employ a variety of methods to produce elevation data. The measurement interv al is less than or equal to 2 feet, and the maximum error is 5 percent bias, or 44 inches/mile. The profile of one or both wheel paths is measured using eith er contact or non-contact profilometers that have been calibrated on sections with profiles determined from a class device. Examples of these instruments include th e APL trailer, the GM profilometer, the K.J. Law profilometer, and the South Dakota profiler. Class III: Response-Type Road Ro ughness Measurement System (RTRRMS), which accumulates suspension deflections (axle to body or acceleration values) from the roadway surfaces. The maximum e rror associated with the operation of these instruments is 10 percent, or 32 to 63 inches per mile, and the measurement interval is the test section length. The measures from class III device must be correlated with IRI using equations deve loped experimentally for each device. The class III profiler must be calibrated to sections whose profiles have been determined form a class I or class II pr ofiler. Examples of these instruments include the Mays Ride Meter, the Walk er Roughness Device (Siometer), the BPR Roughometer, ICC Laser High-Speed profile r and the Automatic Road Analyzer (ARAN) unit. Class I and class II include instruments used in the measurement of the shorter wavelengths contained in the pavement surf ace profiles. The instruments within these classifications possess the highest resolution and the smallest acceptable maximum error.

PAGE 26

The disadvantage of class I and class II devices is the low operating speed and the need to close the facility while the measurements ar e performed. Class II is a category based on the assumption that pavement surface ride quality can be direct related to the passengers perception of the vehicles vi bration at a certain frequenc y band, rather than to the absolute surface profile. That is, the passengers are more sensitive to the vertical acceleration of vehicle body than to the actual elevation chan ges of the pavement surface. 2.2.1 Class I System 2.2.1.1 Rod and Level Rod and level (shown in Figure 2-1) is cal led static because the instruments are not moving when the elevation measures are taken. It is conventional surveying equipment consisting of a precision rod, a level for establishing the horizontal datum, and a tape to mark the longitudinal di stance for elevation measurement. Figure 2-1 Rod and Level 14

PAGE 27

2.2.1.2 Dipstick The Dipstick (shown in Figure 2-2) is a device developed, patented, and sold by the Face Company. It is the simplest devices fo r measuring the profile of the pavement. It consists of an inclinometer mounted on a frame; a handle and a microcomputer are mounted on the Dipstick. The Dipstick is walked along the line being profiled. The distance between the two support feet are 305 mm apart. To get the profile along the ground, the surveyor leans the device so all of its weight is on th e leading foot, then raising the rear foot slightly off the ground. Then you pivot the device 180 degree about the leading foot, locating the other foot (for merly behind) in front, along the line bein g profiled. The computer monitors the sensor continuously. When it senses the instrument has stabilized, it automatically records the change in elevat ion and beeps, signaling that the next step can be taken. Figure 2-2 Dipstick 15

PAGE 28

16 The reference elevation is the value calcu lated for the previous point. The height relative to reference is deduced by the angle of the device relative to gravity, together with the spacing between its supports. The longitudinal distance is determined by multiplying the number of measures made with the known spacing. Data analysis for IRI computations is computerized and a continuou s scaled plot of surface profile can be printed. However, the Dipstick does not have the capability to generate RN measurements. 2.2.2 Class II System 2.2.2.1 K. J. Law Profilometer This profiler is a refined version of the original GM-type iner tial profiler. The original GM profiler was deve loped in the 1960s using iner tial reference concept. The original model consisted of two spring-loaded, road-fo llowing wheels mounted on arms beneath the vehicle. These arms were held in contact with th e road by 300-lb spring force. A linear potentiometer measured th e relative displacement between the road surface and a computed inertial reference. Ve hicle frame motion is measured by doubling integration of the signal from accelerometer s, which are mounted on the frame over each of the rear wheels. These accelerometers sens e the vertical motions of the vehicle body relative to an inertial reference. Frame mo tion is added to the relative displacement motion. Improvements to the original profiler are designed by K. J. Law Engineers Inc., including "the conversion to a digital instrumentation system, a non-contacting road

PAGE 29

sensor, and a digital, spatial-based pro cessing method for computing the measured profile. The processing method produces profil e measurements that are independent of measuring speed and changes in speed during measurement." Profiles are measured in real time by a non-contacting op tical displacement measuring system and precision and accelerometers in the right and left wheel paths. 2.2.2.2 APL Profilometer The Longitudinal Profile Analyzer (shown in Figure 2-3) was developed by the French Road Research Laboratory. It consists of a towed trailer with a combination of instrumentation and build-in mechanical proper ties that allow longitudinal profile to be measured. The profile reference is provide d by an inertial pendulum instead of an accelerometer. This pendulum is centered by a coil spring and amped magnetically. A low voltage displacement transducer is loca ted between the pendulum and the arm of the road wheel. Figure 2-3 APL Profilometer 17

PAGE 30

18 As the trailer wheel moves up and down in response to the road roughness, the angle between pendulum and wheel frame is measured and converted to a vertical distance measurement, which is recorded at specified distance intervals. Due to the mechanical nature of the device, measurements must be performed at constant speed; the response is quite sensitive to the speed. Meas urement of the profile distortions that are significant for highway pavements requires op erating the APL at approximately 13 mph. 2.2.2.3 South Dakota Profiler The South Dakota Profiler was develope d by the South Dakota Department of Transportation in 1981. It is typically mounted in a small to mid-sized van and measures pavement profile and rut depth. Mounted on th e front of the initial vehicles are an accelerometer and ultrasonic sensor for profile measurement in on wheel path and three ultrasonic sensors for the meas urement of the rut depth. Prof ile elevation measurements are reported at 1 feet interval and rut depth elevations are m easured and reported at 2 feet intervals. Testing speed can range up to 65 mph. Roughness output has been reported by S outh Dakota profiling system by a PSI value computed form the measured profile data. Profile data are processed nearly instantaneously by the system software usin g correlations between measured profile values and rating panel values from surveys conducted in South Dakot a. It also has the capability to generate IRI fr om measured profile data.

PAGE 31

2.2.3 Class III System There are two basic designs of response-type road roughness measuring systems or devices: these measuring the displacem ent between the vehicle body and axle, and those that use accelerometer to measure the response of the vehicle axle or body. In reality, these devices measures the response of the vehicle to the roughness of the road; hence, the term RTRRMS to describe this class of measuring equipment. 2.2.3.1 BPR Roughometer The BPR roughometer (shown in Figure 2-4) was first introduced in 1925, and was recognized as being the best high-speed roughness-measuring device available at that time. It consisted of a single wheeled trailer that is towed by a car or a light truck at a speed of 20 mph. The wheel is mounted on leaf springs supported by the trailer frame. Pavement surface contours cause the sensing whee l to oscillate vertica lly with respect to the frame. The vertical movement is accumula ted using a numerical integrator, yielding a roughness statistic in terms of in/mile. Figure 2-4 BPR Roughometer 19

PAGE 32

20 After some period of use, it was learned that the equipment was highly susceptible to changes in temperature and to the condition of its bearings and other mechanical components. In addition, it has a resonant frequency problem that, it excited, produced erroneous results. Vibrations were commonly noted at high roughness levels. As a result, its use has gradually declined. 2.2.3.2 Laser Profiler The Laser Profiler uses an infrared laser and precision accelerometer to obtain an accurate, precise profile measurement at speed s up to 65 MPH. It uses the measurement to calculate a profile index (PI), international roughness index (IRI), and ride number (RN), which is used to rate the surface smoothness. The system also generates a profilograph-type plot with defect locations and must grin d lines, which tells the user where the roughness exists and what corrective action to ta ke. There are many companies that produce laser profilers. Figure 2-5 and Figure 2-6 are the pictures of ICC laser profiler and Automatic Road Analyzer (AR AN) of Roadware Group Inc. respectively. The laser profiler consists of industrial PC with printe r, precision accelerometer, laser height sensor, data acquisition sub-system and distance measuring instrument. The axlemounted accelerometer is not as sensitive to the vehicle parameters as the displacement type devices. Movement of the axle in response to road roughness depends on the amount of tire distortion and the upwar d vertical force generated wh en the tire hits a bump and the downward vertical force of the vehicle suspension. If th e force of the suspension on the axle is greater than the upward force generated by the bump, then the tire maintains contact with the pavement so the axle provides a reasonable tracking of the pavement

PAGE 33

surface. The output of the accelerometer can be integrated twice to obtain an estimate of the vertical axle movement. However, this in tegration process can magnify the effect of undesired noise in the signal. Generally th e axle mounted RTRRMS s use a measure of the root-mean-square acceleration of the axle to quantify pavement roughness. The data collected is not affected by vehicle variation such as sp eed, weight and suspension. Figure 2-5 ICC Laser Profiler Figure 2-6 ARAN Laser Profiler 21

PAGE 34

22 The Profiler offers many benefits over th e conventional method of measurement. It doesnt require any set up or break dow n and operates at speeds up to 65 mph. This eliminates the need for lane cl osures or traffic control to te st existing pavements. When the Profiler is used on an all terrain vehicl e it is so lightweight it can test pavements before they have completely set up. The Prof iler can be provided on any vehicle required by the user. The system collects data in realtime as it traverses the pavements surface. The raw data is processed and the results are out put in standard or metric units on the flat panel display or graphics printer and are saved on a hard drive. The software of Laser profiler incl udes digital band-pa ss filters passing wavelengths of 1 foot to 300 feet, digita l high-pass filters passing filters passing wavelengths of 2 feet or less, and statisti cal models generating the reported roughness statistics root mean square vertical acceleration (RMSVA), mean absolute slope. The laser profiler provides surface profile, IRI, Serviceability Index (SI) and Ride Number output. 2.3 SDHPTs Results 2.3.1 Background In 1982, the Texas State of Department of Highways and Public Transportation (SDHPT) initiated a network-level pavement evaluation system (PES) [Lu et al, 1991]. Around the same time, the Federal Highway Ad ministration (FHWA) re quested that each state change its regulation on pavement desi gn policy and procedure; in effect, each state was required to establish a pavement ma nagement system (PMS), two important functions of which are pavement evaluation and field data collection.

PAGE 35

23 In attempting to satisfy the requirement of the FHWA, the SDHPT purchased an Automatic Road Analyzer (ARAN) in 1987. Capable of operating at 30 to 50 mph under normal traffic conditions, the ARAN unit is equipped with several important pavement surface condition surveying subsystems. Three ARAN unit subsystems in partic ular were evaluated: the roughness measuring subsystem, the rut depth measuring subsystem, and the orientation measuring subsystem. For the roughness measuring subsystem evaluation, the main activities included an analysis of repeatability, corre lation, index report interval effect, testing speed effect, dynamic and static response, accuracy, and error. For the subsystem modeling development, the main work was conducted using methodologies to develop (1) calibration models for the problems found during the subsystem evaluation and (2) other models c onsidered useful in implementing the ARAN unit. The ARAN units roughness outpu ts are Root Mean Square Vertical Acceleration (RMSVA), Mean absolute slope (MAS), and other parameters. In the roughness measuring subsystem, the two-part subsystem is divided according to its hardware or its software. The hardware consists of axle and body accelerometers, analog signal amplifiers, analog low-pass filt ers, and a 12-bit analog-to-digital (A /D) converter. The software consists of digital band-pass filters passing wavelengths of 1 foot to 300 feet, digital high-pass filters passing wavelengths of 2 feet or less, and statistical models generating the reported roughne ss statistics (RMSVA, MAS a nd so on). These roughness statistics are described below:

PAGE 36

Root Mean Square Verti cal Acceleration (RMSVA) RMSVA (Root Mean Square Vertical Acceleration) is defined as: N iia N RMSVA1 2)]([ 1 where a(i) is the i th discrete value of filtered vertical acceleration (which must be spatially filtered to remove any DC bias); and N is the number of sample points. MAS is the cumulative value of the abso lute vertical axle or body displacement divided by the vehicles travel ed distance. Mathematically: N iiZX L T N MAS1 2)()( 2 1 where: T =e lapsed time in a test section (station), second; L =station length, miles; X =sample interval of raw acceleration values; N=L/ X; and Z(i) =height calculated by double integrating with this equation; thus Z(i)=Z(i-1)+a(i)+a(i-1) 2.3.2 Correlation Analysis and Roughness Calibration Models Field Tests The modified K. J. Law profilometer was used as the standard roughness reference instrument with IRI output for th e correlation analysis. It was necessary to conduct field experiments to verify whether adequate correlations with the Texas reference, the modified K. J. Law profilo meter, were being ac hieved. It was also 24

PAGE 37

25 appropriate to determine the Texas S DHPT performance boundaries of the ARAN roughness subsystem in terms of testing speed s, pavement types, roughness levels, and report interval. In order to obtain the correlation and calibration models for the roughness measuring subsystem of the ARAN unit, 29 test sections were selected in the Texas SDHPT study. These sections consisted of bot h rigid and flexible pavements and were evaluated with both the modified K. J. Law profilometer and the ARAN unit. The 26 flexible test sites were located in the Austin area. The rigid pavement test sections were newly constructed CRCP and had not yet been opened to the public at the time field tests were conducted. The models developed for this research were based on the combined data collected from both the flexible and the rigid pavement test sections. The test sites were selected because th ey could provide the broadest range of roughness levels and could be safely run at the 50 mph testing speed. The smooth sites were needed to ensure that the subsystem had the resolution necessary to measure smooth pavements correctly, while the rough sites ensured that th e subsystem could handle the large amplitudes generated when traveling down rough pavement. The medium sections allowed data points to be locat ed between the two extremes. Test Consideration of the Modified K. J. Law Profilometer (1) Testing speed. The most frequently used operational speeds of the profilometer are 20 and 50 mph. Therefore, each test section was run at the testing speeds of 20 and 50 mph.

PAGE 38

26 (2) Number of repeat runs. Three repeat runs were made for each test section and testing speed. The mean values of the reported r oughness statistics were calculated and used as the summarized statistic. This wa s done to cancel the operational bias. (3) Raw data reporting interval. The raw data reporting interval of the profilometer is set at 6 inches. The summary statistics are reported for the entire length of a test run. Test Considerati on of the ARAN Unit (1) Testing speed. The ARAN unit is designed for operation in the normal traffic speed range. The field tests were conducted at speeds of 30, 40 and 50 mph for each test section. (2) Number of runs. Three repeat runs were made at each test speed on each test section. The mean values of the repeat runs were calculated and taken as the summary statistic. (3) Report interval. If a test pavement sectio n is divided into L s ubsections, and in each subsection M data are sampled by the roughness measuring subsystem, then the statistical roughness outputs RMSVA, MAS are reported from these M acceleration data. In the ARAN unit, the re ported statistics of each test section are the mean values of the statistical roughness outputs calculate d in each subsection (L). The factor of report interval does not have a statisti cally significant effect on the roughness statistics. Therefore, it was not critical to choose a speci fic report interval for the correlation analysis and calib ration. The report interval of 0.005 mile was chosen for every test because as much data as possible per test run was desired. (4) Raw data sampling interval. The raw data sampling interval is not adjustable in the ARAN unit. The data summary interval of the roughness measuring subsystem is 6 inches.

PAGE 39

2.3.3 Conclusion (1) It shows the roughness statistics co llected from the field tests using the ARAN unit and the modified K. J. Law profilometer, respectively. These test sections are divided into three roughness-le vel groups, as previously me ntioned. This wide roughness distribution makes the correlation analysis results suitable across the wide roughness levels that are normally found in the Texas highway network. The linear model proposed for the research evaluation effort is: Roughness (Prof) = A + B x Roughness (ARAN) where A and B are constant. Two statistical indices showing the correla tivity of two instruments are used. One is the R square value and the other is the ro ot mean square error (RMSE), defined by )( 11 ii N iyx N RMSE where N is the number of test section (N=29), is the estimation of the roughness statistic of the profilometer at i ix th test section, and is the roughness statistic measured by ARAN unit at i iy th test section. (2) Different testing speeds of the ARAN unit result in different correlation models in terms of parameters A and B. This indicates that the testing speed has a direct impact on the roughness statistic meas ured and reported by the ARAN unit. (3) The correlation models developed from these field tests are speed-dependent. If no effect-canceling model is implemented, the correlation models should be used only for a given operation speed. (4) Report interval does not have a si gnificant impact on th e roughness outputs. 27

PAGE 40

2.4 Computation Methods of Inertial Based Profiling System Several computation methods are available to get the profile. The methods are Spanglers method used in K. J. Law pr ofilometer, the Swedish Road and Traffic Research Institutes (VTI) profiling method, and the Pennsylvania Transportation Institutes (PTI) profiling method. The Spanglers profile computation me thod is based on the following equation: dsds V Y YWP2)( Here, P is the computed profile; (W-Y) is the height measurement; V is the vehicles speed measurement; ds is the integration distance interval. Y is the vertical acceleration. In VTI profiling method, the acceleration signal is integrated first and high-passed by a second order filter. Then the height measurement is differentiated and high-pass filtered. The profile slope was obtained and mapped into spatial domain from time domain. Finally, the slope profile is integrated and the profile is obtained. The PTI profiling method was develope d by Pong. The profile computation algorithm includes a double integration routine for processing acceleration signals and a high pass filter routine to remove the unwante d low frequency profile s. Three steps for signal processing were performed. First, accel eration signals were double integrated over the time period for the vehicl e to pass the distance sampling interval. Second, both the integrated signals and the height signals pa ss digital filters to remove profiles with wavelength longer than 300ft. Fina lly, the profile is the sum of both the filtered integrated acceleration and height signals. 28

PAGE 41

29 From the calculation procedures of these profiling computation methods, the phase shift was introduced. At the same time the errors measured in the acceleration would be enlarged due to double integrati on. The new computation method needs to be developed to remove the phase shift intr oduced by these three computation methods.

PAGE 42

30 CHAPTER 3 METHODOLOGY 3.1 Introduction In the thesis research, an inertial-based auto matic pavement roughnessmeasuring system was develope d. The methods used in th e system are described in detail here. This chapter consis ts of three sections. The fi rst section will explain the methodology used to get th e pavement profile from th e vertical acceleration, longitudinal distance and height distance. Then the quarter car model, the algorithms of calculating International Roughne ss Index (IRI) and Ride Number (RN) are described in details. Finally the digital filter and power spectrum analysis will be explained and the discussion of their application in the pavement measurement system. 3.2 Method to Get Road Profile Inertial profilers compute profile from a combination of the output of three sensors: a height sensor, an accelerometer and a longitudinal dist ance sensor. Vertical acceleration measured at a point fixed on the ve hicle body is integrated twice to construct floating reference height. The height sens or, mounted in the same position as the

PAGE 43

accelerometer, measures the distance from th e floating reference height. The height sensor signal is subtracted from the height of the floating reference to compute the profile elevation. The longitudinal distance measurem ent is needed to associate a position with each profile elevation. This me thod of measuring profile was invented by Elson Spangler and William Kelly. It is described mathematically by the following: x tdsdsVsAxHxZ2/) ( Where x is longitudinal distance, xZ is the computed profile, xH is the height sensor measurement, the term with integral is the floating refe rence derived from vertical acceleration and forward speed V. The acceleration is divided by forward speed squared to convert it into spatial acceleratio n in units of 1/length. The height sensor measurement is the distance from the ve hicle to the ground and should always be negative. ) ( sAt All inertial profilers use a discrete adap tation to compute profile. The procedures we used to compute the profile are: 1). Calculate the bias in the accelerometer signal and remove it. This step helps minimize error in the integration that follows. 2). Convert temporal acceleration to spatial acceleration. tA sA 3). Integrate the spatial acceleration once to obtain slope. This is done with a recursive finite difference equation: )()1()( iAiSCiSs a a 31

PAGE 44

Where is the sample interval and Sa is the component of the slope profile measured by the accelerometer. The first term includes a drift removal coefficient: L is usually set the three times the l ongest wavelength of interest. LC / 4).Differentiate the height sensor signal (H) once to obtain slope: )()1( )( iHiHC iSh Where is the component of the slop prof ile measured by the height sensor. hS 5). Combine the slope from the height sens or and accelerometer signals to get the slope of the road profile (S): ) ()( iSiSSh ah If the final goal of the profile measurement is to get IRI or RN, this resu lt can go directly into the calculation. 6). Integrate the slope profile to obtain elevation. The integration is performed backward in this step to can cel the phase lag introduced in the computation of the slope profile. This method of profile computation cancels the phase shift associated with integration by moving forward through the profile in steps 1 to 5, then backward in step 6. 7). Use the trend remove algorithm to get rid of the profile trend. 3.3 IRI and RN Models and Algorithms 3.3.1 Quarter Car Model The concept of quarter-car simulation as a method for analyzing pavement profile data was originally an attempt to simu late the output of the BPR roughometer. Subsequently, vehicle simulation studies at the University of Michigan demonstrated that 32

PAGE 45

full-car and half-car simulation models do not provide an advantag e over the quarter-car simulation and are computat ionally much complicated. The parameters of the quarter-car that are shown in Figure 3-1: Figure 3-1 Quarter-car Model these parameters include the major dynamic effects that determine how roughness causes vibration in a road vehicle. The masses, springs, and dampers are defined by the following parameters: the sprung mass of the vehicle body; the suspension spring and damper (shock absorber) constants; the Unsp rung mass of the suspension, tire, and wheel; and the spring constant of the tire. Theore tical correctness woul d require a damper constant for the tire. However, practical application generally ignores this term. Mathematically, the behavior of a quarte r-car can be described with two-second order equations: 0)(.. .. ZZKZMZMUt U U S S and 0)()(.. ..USS US S S SZZKZZCZM 33

PAGE 46

34 Where Z = road profile elevation, Zu = elevation of unsprung mass (axle), Zs = elevation of sprung mass (body), Kt = tire spring constant, Ks = suspension spring constant, Cs = shock absorber constant, Mu = unsprung mass (axle), and Ms = sprung mass. The double dot notation above the elevati on terms represents acceleration while the single dot repr esents velocity. To simplify the equations, the paramete rs are normalized by the sprung mass, Ms. The following values for the normalized parameters define the Golden Car data set: K1 = Kt / Ms = 653, K2 = Ks / Ms = 63.3 C = Cs / Ms = 6.0 M = Mu / Ms = 0.15 Since RTRRMS devices generally measure the movement between the vehicle axle and body, simulation require s calculation of the differen ce in elevation between the body and axle in response to the road profile and forward motion of the vehicle. This is accomplished by integrating the difference in the velocities between the sprung and unsprung mass; producing the quarter-car statistic, QCS:

PAGE 47

dtZZ C QCST O U S..1 (1) The term C represents either the total time required to traverse the section of road or the length of the section, L. If the time factor is used to normalize the quarter-car statistic, the calculation results in an aver age rectified velocity, while a distance base yields the average rectified slope. There are several acceptable numerical techniques for the solution of Equation 1. However, the linear nature of the equations permits an exact solution with the state transition matrix method. Historically, two sets of vehicle parameters have been used for computing quarter-car statistics for calibration of RTRRMS devices. A set representing the original BPR Roughometer trailer was used for several years, until research at the Highway Safety Research Institute (HRSI) produced an updated set of vehicle parameters. The World Bank recommends the HSRI vehicle parameters and has termed the quarter-car statistic computed as the in ternational roughness index, IRI. Although the mathematical base for quarte r-car simulation is somewhat complex, computer programs are readily avai lable for performing the calculation. 3.3.2 Calculation of IRI The calculation of the internationa l roughness index (IRI) is accomplished by computing four variables as functions of th e measured profile. (T hese four variables simulate the dynamic response of a reference vehicle, shown in Figure 3-1, traveling over the measured profile.) The equations for the four variables are solved for each measured 35

PAGE 48

elevation point, except for the first point. The average slope over the first 11m (0.5 sec at 80 km/h) is used for initializing the va riables by assigning th e following values: 1/11 0 11/)(' 4 2 1 3 1 dxa ZZ YYZZa (2) where Ya is the a-th profile elevation point that is a dist ance of 11m from the start of the profile, Y 1 is the first point, and dx is the sample interval. (Thus, for a sample interval of dx=0.25m, Equation 2 would use the difference between the 45 th elevation point and the first elevation point to establish an initial slop e for the IRI computation.) The following four-recursive equations are then solved for each elevation point, from 2 to n (n =number of elevation measurement): 1 4 3 2 1 114 13 12 11 YPZsZsZsZsZ (3) 2 4 3 2 1 224 23 22 21 YPZsZsZsZsZ (4) 3 4 3 2 1 334 33 32 31 YPZsZsZsZsZ (5) 4 4 3 2 1 444 43 42 41 YPZsZsZsZsZ (6) Where: slopedxYYYii/)(1 41' jposition previous fromZZjj (7) Sij and Pj are coefficients that are fixe d for a given sample interval, dx, thus, the equations 3 to 6 above are solved for each position along the wheel track. After they are solved for one position, Equation 7 is used to reset the values of Z1', Z2', Z3' and Z4', for 36

PAGE 49

the next position. Also for each position, the rec tify slope (RS) of the filtered profile is computed as: 13ZZRSi The IRI statistic is the average of the RS variables over the length of the site. Thus, after the above equations have been solved for all points, the IRI is calculated as: n zi iRS nIRI )1/(1 The above procedure is valid for any sample interval between dx=0.25 m and dx=0.62 m (2.0ft). For shorter sa mple intervals, the additi onal step of smoothing the profile with a 0.25m moving average is r ecommended to better represent the way in which the tire of a vehicle envelops the ground. Then the IRI is calculated by solving the equations for each average point using coefficients in the equations appropriate for the smaller interval. The computed IRI will have units cons istent with those used for elevation measures and for the sample interval. For example, if elevation is measured as millimeters and dx is expressed in meter, then the IRI will have the preferred units: mm/m=m/km=slope*10. The coefficients used in Equations.17 through 20 are calculated from the equations of motion that define a quart er-car model. In the general case, they are specific to the vehicle model parameter values simulation speed, and the sample interval. The IRI summarizes the roughness qualities th at impact vehicle response, and is most appropriate when a roughness measure is desired that relates to: overall vehicle 37

PAGE 50

operating cost, overall ride quality, dynamic wheel loads, and overall surface condition. Figure 3-2 shows IRI ranges represented by different of road. For the specific case of IRI, define d by the NCHRP 228 parameters [Wambold, 1980] and a standard 80km/hr si mulation speed, they depend only on the sample interval. Complete instructions for measuring IRI are available in Sayers, gillespie, and Queiroz [Janoff et al, 1990]. The instructions include listings of comp uter programs that solve the equations of motion and also computer pr ograms that calculate the coefficients. Figure 3-2 IRI Roughness Scale 3.3.3 Calculation of RN RN is the result of two NCHRP research es performed in the 1980s by Janoff to investigate the effect of road surface roughne ss on ride comfort. Th e objective of these researches was to determine how features in road profil es were linked to subjective opinion about the road from members of the public. During two studies, spaced at about a 38

PAGE 51

5-year interval, mean panel ratings (MPR) we re determined experimentally on a 0-to-5 scale for test sites in several states. The 0 to 5 scale as shown in Figure 3-3 was used for a large-scale road test conduc ted by AASHO in the 1950s, in which roads were subjected to mixed traffic and researchers trac ked the condition of the pavement. Figure 3-3 Subjective Rating Scales for Roads Longitudinal profiles were obtained from leftand righ t-wheel tracks of the lanes that were rated. RN is an estimate of MPR. The mathematical procedure developed to calculate RN is described in NCHRP Repor t 275, but not in complete detail. In 1995, some of the data from the two NCHRP projects performed and a panel study conducted in Minnesota were analyzed again in a pooled-fund study initiated by the Federal Highway Administration to develop and test a practical mathematical process for obtaining RN based on objective measurement, not subjective rating. The method was to be provided as portable software similar to th at available for the IR I, but for predicting MPR rather than IRI. 39

PAGE 52

RN is a nonlinear transform of a statistic called profile index (P I). PI is calculated from one or two profiles. The profile is filtered with a moving average with a 250-mm (9.85-in) base length. The movi ng average is a low-pass filter that smoothes the profile. The computer program does not ap ply the filter unless the profile interval is shorter than 167mm (6.6 in). The profile is further filtered with band-pass filter. The filter uses the same equations as the quarter-car model in the IRI. However different coefficients are used to obtain the sensitivity to wave numbe r shown in the last figure. The quarter-car parameters for the PI calculation are: K 1 = K t / M s = 5120, K 2 = K s / M s = 390, C = C s / M s = 17 M = M u / M s = 0.036 The filtered profiles are reduced to yield PI, which should have units of dimensionless slope (ft/ft, m/m, etc). Then, PI is transformed to RN. RN is defined as an exponential transform of PI according to the equation: )(1605PIeRN If a single profile is being processed, PI is calculated directly. If two profiles for both the leftand right-wheel tracks are processed, PI values from the two wheel tracks are averaged with the following equati on, then previous equation is applied. 222 R LPIPI PI 40

PAGE 53

Figure 3-4 shows the sensitivity of RN to wave number. The maximum sensitivity of RN is for a wave number of 0.164 cycle/m (0.05 cycles/ft), which is a wavelength of about 6 meters (20 ft). The IRI has a great se nsitivity to a wavelength of 16 meters (wave number of 0.065 cycle/m). The figure shows th at RN has a low sensitivity to that wavelength and even lower sens itivity for longer wavelengths. Figure 3-4 Sensitive of RN to Wave Number 3.4 Digital Filter The true profile is continuous. It is a slice of the pavement or ground surface. Instruments that produce continuous measures are called analog, because the measure is analogous to the variable of interest. Alte rnatively, a continuous variable is often represented with a sequence of numbers calle d as digital signal. Most of the modern profilers produce sequences of numb ers stand for the pavement profile. The reason why we try to apply the digita l filter technology is because a profile can be thought as consisting of different ki nds of signals. For example, as shown in 41

PAGE 54

Figure 3-5, it is a part of tr ue pavement profile. Even t hough there is a general slope cover from the start point to the end point, you can still find a lot of micro-texture exists in this section of the pavement. Figure 3-5 Pavement Profile After a measurement is made, all we know ab out the road profile are the numbers that make up our measurement and we can plot the curve of the profile form the result of measurement as shown in the Figure 3-6. Figure 3-6 Measured Profile If we want to get the picture about the micro-texture condition of that pavement, we can apply the digital filter to process above data and get the profile like Figure 3-7: 42

PAGE 55

Figure 3-7 Filtered Profile Signal processing is the mathematical an alysis and transformation of signals. A digital filter is a calculation pro cedure that transforms a series of numbers (a signal) into a new series of numbers. In or der to make practical use of a profile measurement, it is necessary to filter the sequence of numbers that makes up the profile. Profiling information is used to evaluate the condition of pavements and to manage road networks. A profile consists of different wavelengths varying from a few centimeters to hundreds of meters. It is also necessary to filter pr ofile data to view diffe rent types of profile features, i.e., the profile are filtered to include only those waves of interest. When analyzing pavement profile, it is desirable to remove the long wavelengths when the road trend is desired, to remove the short wavelengt hs if expected to ge t the pavement details. In summary, digital filter is the mathematical analysis and transformation of signals. Signals are filtered mainly for two reasons: 1) to improve the quality of measurem ent by eliminating unwanted noise 2) to extract information of interest from the signal. To make practical use of a profile measurement, it is almost mandatory to filter the sequence of numbers that make up the profile. A much simpler filter, commonly used in profile analysis, is called the moving aver age. A moving average filter simply replaces 43

PAGE 56

each profile point with the av erage of several adjacent point s within some base length. For a profile p that has been sampled at interval, a moving average smoothing filter is defined by the summation: X B i X B ij ave ifljp N p2 2 )()( 1 where, fp is the smoothed profile B is the base length of the moving average aveN is the number of samples included in the summation. The ratio B/(2 X ) is often not an integer. The round-off method influences the effective base length of the moving average. The effect of a moving average filter can be demonstrated in Figure 3-8. The effect is to smooth the profile by averag ing out the point-by-point fluctuations. Figure 3-8 Effect of Moving Average Filter There are four basic filter types; Low-pa ss, high-pass, band-pass and band-stop. There are also two types of filters: Finite impulse response (FIR) and infinite impulse 44

PAGE 57

response (IIR). In general FIR filters can be designed to have exact linear phase and there is also great flexibility in shaping their ma gnitude response. In addition, FIR filters are inherently more stable and th e effects of quantization errors are less severe than IIR filters. Conversely, IIR filters require fewer co efficients than FIR filters for a sharp cutoff frequency response, and analogue filters can only be modeled using IIR filters. The method of digital filter design is bu ilt upon a more fundamental approach that is call Fourier series method. This method is based on the fact that the frequency response of a digital filter is periodic and is therefor e represented as a Fourier series. A desired target frequency response is selected and e xpanded as Fourier seri es. This expansion is truncated to a finite number of terms that are us ed as the filter coefficients or filter orders. The resulting filter has a frequency response that approximates the original desired target response. Digital filters can be implemented in two ways, by convolution (called finite impulse response or FIR) and by recursion ( called infinite impulse response or IIR). The general form of the digita l filter difference equation is: N i N i i iinybinxany01)()()( where y(n) is the current filter output, the y(n-i)s are previous filter outputs, the x(n-i)s are current or previous filter inputs, the a i s are the filters feed forward coefficients corresponding to the zero s of the filter, the b i s are the filters feedback coefficients corresponding to the poles of the filter, and N is the filters order. II R filters have one or more nonzero feedback coefficients. That is, as a result of the feedback term, if the filter has one or more poles, once the filter has been excited with an impulse there is always an 45

PAGE 58

output. FIR filters have no non-zero feedback coe fficient. That is, the filter has only zeros, and once it has been excited with an impulse, the output is present for only a finite (N) number of computational cycles. The recursive filter is described by a difference equation given by: ...]2[]1[...]2[]1[][][2 1 2 1 0 nybnybnxanxanxany By using the z-transform, we can system s transfer function can be given by: ... 1 ... ][3 3 2 2 1 1 3 3 2 2 1 10 zbzbzb zazazaa zH So the major task of filter design here is to calculate the value of all coefficients, and the major steps to get all the coefficients are described as follows: Step 1. Specify a desired frequency response ) ( dH including the magnitude and cutoff frequencies, for example, the cutoff frequencies for band pass filter are L and H Step 2. Specify the desired number of filter orders N. Step 3. Compute the filter coefficients h(n) for n=0, 1, 2,..., N-1 using 2)]sin())[cos(( 2 1 )( dmjm Hnhd The coefficients h(n) for an ideal band pass filter are calculated as: 0 )]sin()[sin( 1 0 )( m m m m m nhL H LH where: 46

PAGE 59

evenn Nn oddn Nn m 2/ 2/)1( During the research period, the Chebyshev filter was developed. The Chebyshev filter is a mathematical strategy for achievi ng a faster roll-off by allowing ripple in the frequency response. Digital filters that use this approach are called Chebyshev filters. 3.5 Maximum Entropy Spectrum Analysis Road profile usually does not contain id entifiable sinusoids. It means that an arbitrary profile shape can be constructed ar tificially by adding together a series of sinusoids with different wavelengths, amplitude s and phases. If a prof ile is defined with N equally spaced elevation points, then it can be duplicated math ematically with N/2 sinusoids. Because there are so many sinusoi ds being added, their individual amplitudes are not large. A mathematical transform ca lled a Fourier transform can be used to compute the amplitudes of the sinusoids that could be added togeth er to construct the profile. The Fourier transform can be scaled such that it shows how the variance of the profile is distributed over wave numbers asso ciated with sinusoids. When scaled in this manner, the analysis is power spectral analysis. Power comes from its application in electronics, where it is applie d to voltages. The variance of a voltage is proportional to power in a resister, so the power spectral an alysis illustrates the distribution of power over frequency. The mathematical calculati ons can be applied to road profiles. Characteristics of profile are importa nt when evaluating pavement roughness. Traditional research is based on analysis in the time or space domain, and its scope is sometimes limited because of lack of methodology or lim itations of those domains. 47

PAGE 60

Maximum entropy spectral analysis was first proposed by J.P. Burg in 1967, which has developed in the past several decades. It is also called modern spectrum compared with classical FFT method. In the research the maxi mum entropy spectral analysis will be used to evaluate and compare the road roughness condition related to profile under different road conditions. Below are some basi c concepts of spectral analysis. Suppose pavement profile sampled at time interval T can be abstracted as a discrete sequence, called the discrete profile sequence or aaaain12,,.... where a= i i th profile data sampled at time interval T (sec). Consider the inverse discrete Fourier tran sformation of the discrete data sequence defined by Oppenheim: a N Heik jikN k N 12 0 1 / ( i=1,2.N) where N = length of the sequence (number of data points in the sequence), ai = i th profile data, Hk = weights (k =0, 1, 2,., N-1), and j= ()/ 112 a can be considered a weighted summation of sine function e. If a new variable is defined by i jikN 2 / wk N (k= 0,1,2,.N-1) wkk 2 / then 48

PAGE 61

1 01N k jiw k ikeH N a ( i=1,2.N) and iwjiw ek k iwjkcos sin)( Usually, the variable is called frequency and is within the range [0, From this Equation, it is known that the larger the the more sine components with frequency wthe discrete acceleration sequence { } contains. Mathematically, it can be proved that wk 21 ()/ N ] N Hk k ai HHwaekki jiw k i () [ wNNNk N 02421 ,/,/,...,()/ ] In other words, is the discrete Fourier transformation of { } and the function of frequencyw. The equation implies that the discrete sequence {a} in the space domain can be transferred into the frequency domain sequence {H }, and characteristics of sequence { } can be analyzed in the fr equency domainthat is, knowing, one can analyze the characteristics of {a}. Since Hw is an imaginary sequence, a real function is defined by Hwk( ) ) ) ) ai k i ( wk ai Hwk( i k( SwHwkk()() 2 where is called the spectral de nsity function of sequence{ }. To calculate the summation should be from Swk( ) ) ai Hwk( to + In practical engineering cases, sequences of N are finite because one cant collect infinite se quences of data. The 49

PAGE 62

50) spectral density function Sw thus should be estimated from { } by some estimation model. k( ai In the area of spectral function estima tion, several mathematical methods are available, such as fast Fourier transf ormation (FFT), maximum likelihood spectral estimation, and maximum entropy spectral estim ation (MESE). The MESE method is one of the best. The MESE method was introduced by Bu rg in 1968. Like maximum likelihood spectral estimation, MESE is a kind of esti mator of parameter estimation. Consider a discrete sequence {a}with sequence length N and sample in terval T. If the sequence is a stationary, zero mean, approximately normally distributed, and band-limited stochastic process, then entropy of the sequence is defined as i B BdwwSBB H )](ln[ 4 1 )2ln( 2 1 where B is band width of the sequence and S( w) is the spectral de nsity function of the sequence, or SwTRmejmTw m()() R(m) is defined as the autoco rrelation function of sequence{a} i R(m)=E{a i aim } So entropy is obtained by B B jmTwdwemRTBB H ))(ln( 4 1 )2ln( 2 1 Suppose the values of autocorrelation R(m) are given for m = 0, 1, 2.M. then the corresponding extension of autocorrelati on function is defined by convolution sum

PAGE 63

RmRmkbk k M()() 1 (m>M) or, equivalently, Rmkbk k M() 00 ( b01 m> M) The method that Burg introduced maxi mizes entropy H with respect to R(m) ( mM ) with restrained condit ion, so that parameter bb can be obtained. Mathematically, this can be expressed as bM 12,,... H Rm () 0 mM Rmkbk k M() 00 It can be proved that with the conditions, sequence{ } can be related by the following auto-regression mode l, called AR(M) model: ai abababaiiiMiM ei 1122... where M is the order of the AR(M) model and { } is an approximately normally distributed disturbance with zero mean value. The estimator of the parameter ( can be obtained by the Yule-Welker equation: ei bbbM 12,,...,) RB=P where R is the autocorrelation matrix of sequence{ }. R is called the Toeplitz matrix: ai where 51

PAGE 64

)0( )1( )1_()( )1()0( )2()1( )1()2( )0( )1( )()1( )1()0( R R R mRMR R R MR MR MRMR R R MRMR R R and M Mb b b B1 1. 1 P PM 0 0 0 . where PE{(eMi )}2 Finally, with all parameters estimated by MESE algorithm, the maximum entropy spectral density function can be expressed by Sw PT beM m jmTw m M() 11 2 52

PAGE 65

53 CHAPTER 4 SYSTEM DEVELOPMENT There are primarily two types of equipment measure Road roughness in the United States: direct type profiler, and re sponse type road roughness measuring system. Ideally, the road profiling method gives accura te measurement of the pavement profile along a reference path and the response-type method can be operated at high speed and needs to be calibrated using Dipstick or dire ct type profiler. The results from inertialbased system are acceptable if calibrated accurately. This chapter describes the high speed lase r profiler design. It demonstrates the decisions made in the design, th eir rationale and the particular details of the hardware and software design. It is felt that the overall system description is necessary and then the profiler software, also called the Laser In ertial-based Pavement Roughness Evaluation System (LIPRES) is described. 4.1 System Requirements To obtain information from a measured pr ofile, there are two basic requirements: 1) The system must be capable of sampli ng the relevant information present in the true profile.

PAGE 66

54 2) The computer software must exist to pr ocess the measured values to extract the desired information (such as a summary index). Because pavement roughness is defined over a pavement profile and the length of it is normally five hundred to six hundred feet the size of data needed to get the profile will consume a lot of memory due to the small sample interval. It is complicated if the user want to process the profile to get what he wants. With the development of electronic and computer technology, we can use the el ectronic device and computer system to provide the data logging and data analysis process easily. However, there are still several factors need to be taken into account: How the computer connect to the electri c part of devices, that is, how the computer get the signal of the electric part of the sensors. Other factors to be considered are th e easy use of sampling under the testing environment and software compatibility. Based on the above pavement roughness measurement requirements, we followed following approaches: The profiler system consists of a vehicle; a laptop computer is connected with several sensors through A/D card with PCI interface and combined with specific software. The power supply for the sensor is from voltage inverter connected from the vehicle electric outlet. A profiler is an instrument used to produ ce a series of numbers related in a welldefined way to a true profile. It measures th e components of true profile that are needed for a specific purpose.

PAGE 67

55 A profiler works by combining three ingredients: a reference elevation, a height relative to the reference, and Longitudinal distance. The factors that may affect to get a true profile include the p hysical configuration of the systems, the capabilities of the sensors, and the manner in which the sensor signals are processed. There are minimum performa nce specifications for components of the profiling system, standards for equipmen t design, procedures for ensuring that components are functioning properly and some improvements for error detection and diagnosis. Sample interval is important factors when measuring profile. It is relate to well established sampling theory, but are intimatel y tied with the selection of the minimum sample interval necessary for measuring profile. The profile is computed from a combina tion of longitudinal distance, height and acceleration measurements. The height and acceleration measurements require special conditioning, because they are random signals. The accelerometers in a profile are analog sensors. They output voltage that is cont inuous and proportional to the acceleration. Height sensors can make a measurement a finite number of times in a second. Usually, the height sensor measures over 1k sample per second. It is impractical to record and use all of the data that is measured by the sensor s in a profiler. Thus, data are digitized and recorded into computer at discrete interv als. The longitudinal distance between points that are digitized and fed into the profile comp utation algorithm is the sample interval of the profile.

PAGE 68

56 A crucial step that must be performed on the height sensor and accelerometer signals before the data are reco rded is anti-aliasing. Aliasing is a problem inherent in all digital sampling. As a consequen ce of digitizing at a given sample interval, some high frequency components of quantity being measured will contaminate the lower frequency components. This problem is solv ed by filtering the sensor signals. There are mainly 3 types of sensors used in a laser profiler: accelerometer, height sensor and longitudinal distance measurement. The requirement for each sensor and sensor specification ar e specified as follows. Accelerometers Accelerometer is used in a high speed prof iler to establish an inertial reference from which relative height measurements are made. The vertical acceleration of the host vehicle body is integrated twice to establish vertical position. This is used as a floating reference height, and the height sensor measur ement is subtracted from it to get a road elevation. Accelerometer should be oriented vertica lly. The Accelerometer is mounted just above the height sensor. Thus, the accelerometer is not always perfectly vertical when the vehicle body undergoes pitch and roll as it trav els over uneven roads. An error occurs if the vehicle pitches and accelerates longitudinally at the same time. But this error is small if the lateral and longitudinal accelerations are held under 0.1G. The accelerometer we used in the profiler measures the true vertical acceleration accurately even if the vehicle body is tilted. The accelerometer in the profiler used has a total range of -10g to 10g. This is enough for normal road use.

PAGE 69

57 To properly capture the wavelength range of interest for measurement of road roughness, the accelerometer has a natural fr equency of 100 Hz, which can measure a wavelength of 0.3m at a travel speed of 100 km per hour, which is the limit of the profiler. Height Sensor The height sensor in the profiler measures the vertical distance from the vehicle to the road. This value is subtracted from a floating reference height, measured by the accelerometer, to get road elevation. It is la ser distance sensor by means of triangulation. A spot of invisible light is projected onto th e road surface. It is reflected through a lens mounted at an angle onto a light sensitive displacement sensor. The height sensor has a fast respond speed of over 1000 samples per se cond so that the profiler can get a sample interval less than 10 mm at a speed of 60 km per hour. Sayers study shows that the resolution required of the profile for accurate measurement of IRI less than 3.0m/km, a resolution of 1mm and a sample interval of 500mm or less was required. On roads rougher than 5 m/km a resolution of 2.5 mm was permissible. The resolution required of the he ight sensor is about the same magnitude. The resolution of the height sensor used in the profiler is 0.4mm, which meets the requirement for measuring road profile and can be used at an inertial based roughness measurement system. Longitudinal Distance Sensor The distance-measuring instrument is on e the three major types of transducers that make up a profiler. Distance must be m easured properly to obt ain accurate roughness statistics, but it must also be correct from an operational standpoint. In network

PAGE 70

58 monitoring, applications roughness is often measured over very long distance, such that a small bias in longitudinal distance measurem ent may build up to a large net error. The error throws off distance accounting and the l ongitudinal positioning of each segment. In project level applications, m easurement of new constructi on corrective action is often recommended at specific locations. Thus, accu rate measurement of longitudinal distance relative to fixed landmarks is very important. In the laser profiler, distance traveled is measured by a pulse sensor on one of the vehicle wheels. The configuration is to inst all an exciter right with equally spaced notches on the back side of the disc brake rotor of one of the wheels, and then the distance sensor is mounted to the board. Ro tation of the wheel is measured by detection of pulses as the wheel rotates. The pulse sensor can measure up to several thousands pulses for each round. During normal operation, each pulse is associated directly with a fixed travel distance through the rolling radius of the tire. The rolling radius is the effective radius of the tire when the vehicle is moving. It is generally smaller than the unloaded radius of a tire, but larger than the radius of a loaded but stationary tire. Because the rolling radius can not be measured statically, the distance pulse sensor must be calibrated. This is done by traveling a know distance and counting the pulses. 4.2 Hardware Description 4.2.1 Laser Profiler Introduction The objective in the development of the lase r profiler is to develop a high speed, inexpensive, easily to operate device. The profiler can measure the roughness of a long

PAGE 71

distance section with speed from 20 to 60 km per hour. It is supposed to be a relatively precise machine suitable for measuring su rface roughness within a class III profiler confidence applied on the new pavement c onstructions, reconstruction or overlay. The profiler is an instrument designed to facilitate the efficient collection and presentation of continuous paved surface information, including distance, profile and International Roughness Index (IRI), Ride Nu mber (RN) values. It works by combining three ingredients: (1) a reference elevation; (2 ) a height relative to the reference; and (3) longitudinal distance. Figure 4-1 shows a simplif ied view of the three ingredients in an inertial based laser profiler. Figure 4-1 The Inertial Based Laser Profiler The profiler enables accurate recording of measurements for road profile, level for surfaces such as paved roads, footpaths, runways, building slabs and sporting surface. Figure 4-2 is a photo of the laser profiler. Figure 4-3 shows the prototype of the laser profile which you can see the how the sensors mounted onto the vehicle. These 59

PAGE 72

compact and easy-to-use devices are installed onto the vehicle and travel over the road surface to be surveyed. The computer inside the vehicle collects and records the sample signal from each sensor mounted on the ve hicle. After sampling, it calculates and displays graphics and tables results. The system comes with a complete software package that provides all the data acquisition, data processing work. It can also output all major roughness indices. Figure 4-2 Laser Profiler Figure 4-3 Prototype of Laser Profiler 60

PAGE 73

4.2.2 Hardware Architecture Figure 4-4 shows a schematic diagram of the hardware architecture of the laser profiler. When the operator collects the pave ment roughness data, he can sit inside the card and works on the software to begin sampling. The samp led data come from three sensors: the longitudinal distance is from the pulse sensor, vertical height is from the laser height senor, and acceleration is from the accelerometer. All the data will be transmitted to the desktop computer. The power is from the high capacity battery, and by which also powers all sensors. The computer will process the data and output the process result. There are also several voltage convert ers that convert the ba ttery voltage to the voltage each sensors needed. The provided desktop computer is with the necessary communication cable and power cable. Data Sampling Data Processing: IRI and RN System Configuration Accelerometer Height Distance Senso r A/D Converter Transducer PCI Interface Desktop Computer Longitudinal Distance Senso r Digital Counter Figure 4-4 System Diagram of Laser Profiler The inertial based profiler needs three parts to be installed: 1). Laser height sensor and accelerometer are usually installe d on the front bumper of the vehicle, 2). 61

PAGE 74

62 Pulse sensor needs to be inst alled on one of the vehicle front wheels, 3). The computer is installed inside the vehicle which cont ains the AD converter and the software. The laser profiler uses the AD converter that provides the PCI interface to the computer. The AD converter has voltage analog input and counter digi tal input. It gets the voltage from height distance and acceleration and collects the counts that vehicle traveled from pulse sensor. The analog input can make 100k samples per second and it can count the pulses to 2.5M Hz. 4.3 Software Development 4.3.1 Overview The software transforms the computer and data acquisition hardware into a complete data acquisition, analysis, and calib ration system. Its name is LIPRES (Laser Inertial-based Pavement Roughness Evaluation Sy stem). LIPRES is developed in Visual Basic and runs on a personal computer under Window 9x, window 2000 and Windows XP operating systems. LIPRES has four main functions: (1) Data Sampling, (2) Data Processing, (3) System calibration and (4) configuration. Th e user-friendly GUI takes the user through the software set-up features in different screens effortlessly with the context help providing useful instructions. Figure 4-5 is the main menu of the system. The system calibration and configuration are included in the param setup menu.

PAGE 75

Figure 4-5 LIPRES Main Interface Form The software architecture and function of the developed software can be summarized in Figure 4-6. Figure 4-6 System Software Architecture 63

PAGE 76

4.3.2 Calibration Module After the hardware system is installed in a vehicle, the calibration process need to be done before using it. Due to the initial voltage of accelerometer and laser distance sensor is unknown, field position calibration is required. Ther e are 4 types of calibration of the inertial based laser profiler system: accelerometer, laser height distance sensor (including 2 calibrations) and pulse sensor. User can finished the calibration procedure according to the instruction in the calibration forms. Figure 4-6 is the details of the system calibration module. System Calibration Accelerometer Encoder Laser Distance Sensor Laser Distance Amplifier Figure 4-7 Detail Functions of System Calibration Module Figure 4-8 is accelerometer calibration form First, put the whole system in a reference surface, then press Start button to start record accelerometer voltage at current position. After 20 seconds, system will record the accelerometer voltage in each text column. If the operator can compare the voltage readings to the original voltage provided by the accelerometer manufacture, then he can press the save button to save the results. If not, the operator can cl ick cancel button and try another time. 64

PAGE 77

Figure 4-8 Accelerometer Calibration Form Figure 4-9 is laser height distance sensor calibration form. The calibration process is similar to the accelerometer calibration st eps. By doing these calibration, the system get the voltage of accelerometers and height se nsors at the reference surface. The voltage readings can be considered as the zero position voltage readings. Figure 4-9 Laser Height Dist ance Sensor Calibration Form 65

PAGE 78

The pulse sensor calibration is the calibration of length of one pulse. The profiler can travel at a defined distan ce, then the system can record how many pulses that the vehicle traveled. By dividing the pulses from th e distance, then the length of each pulse is known. It is required to do the calibrated pr ocess twice for accuracy purpose. Figure 4-10 is the pulse sensor calibration form. Figure 4-10 Pulse Sensor Calibration Form In order to get the accurate of the amplifier of the laser height sensor, it is required to perform the amplif ier calibration (Figure 4-11). Figure 4-11 Laser Height Sensor Amplifier Calibration Form 66

PAGE 79

4.3.3 Data Sample Module The Data Sample Module executes in a real -time environment. It functions as the resource of the data acquisition and data st orage. After operator i nput the configuration parameters such as: Observer name, Data collection site name, Data collection date, one side or two side sample, maximum data coll ection length and normal, forward mileage or backward mileage sampling. All th ese selections are in data sample setup form, as shown in the Figure 4-12. Figure 4-12 Sample Setup Form Data sample form is the main user interface for data acquisition. When the Sample module receives the digital and analog data from the PCI interface, then the software can write the data into the hard drive. The operator can select the sample interval he wants from parameter setting modu le, the distance sensor will trigger the AD converter and the software to process the data collection. The software will gather 4 types 67

PAGE 80

of data: accelerometer, height, longitudinal distance and speed. The speed is calculated by dividing the time elapsed of the distance tr aveled. To minimize the errors caused by the pulse sensor, operator can input actual distance at a specified location. After sampling, the speed and distance will be modi fied according to each distance that the operator input during sampling. It is shown in Figure 4-13. Figure 4-13 Data Sample Form 4.3.4 Data Analysis Module The data analysis module is the most im portant part of the LIPRES software. It can calculate the profile from speed, distan ce, acceleration and height data, and then output the roughness results. The pavement prof ile can be filtered in order to get the wavelength of interest, reviewed for all sections or any interested section part, 68

PAGE 81

summarized to get the roughness indices and di vided to get the subsections analysis results in this module. The details of data processing module are shown in Figure 4-14. Data Analysis Whole Section Data Analysis Subsection Data Analysis IRI Results (IRI Plot & Results) RSD Results (RSD Plot & Results) RN Results (RN Plot & Results) ETD Results (ETD Plot & Results) Whole Length / Part Length Roughness Results Figure 4-14 Data Analysis Module Details The following (Figure 4-15) is the main data analysis interface. Figure 4-15 Data Analysis Form Figure 4-16 is the calculated profile of one sample data using whole section analysis. 69

PAGE 82

Figure 4-16 Profile Display Form Figure 4-17 is the roughness index resu lts for the whole sampled section. Figure 4-17 Whole Section Analysis Results Figure 4-18 and Figure 4-18 are the IR I curves and roughness output for each subsection of a sub-length of 50m with a total length of 300 meters from one sampled data. 70

PAGE 83

Figure 4-18 Subsection IRI Figure 4-19 Subsection IRI Results Form 71

PAGE 84

4.3.5 Configuration Module Figure 4-20 is the main configuration form. Figure 4-20 System Configuration Form The system configuration module function is shown in Figure 4-21. System Parameters Setup Sensors Sample Setting Sample Interval Calculation Params Figure 4-21 Detail of System Configuration Functions Before the data analysis, operator was required to input the configuration parameters that include the longe st wavelength to be kept in the data for further analysis, the value of the subsection length and the criteria for the el evation roughness analysis. 72

PAGE 85

The user can view the calibration results and change some other parameters. The parameters include how to display the speed sample interval, each pulses length if system is installed in a new vehicle. Figure 4-22 shows the system parameters setup form. Figure 4-22 System Parameters Setup Form User can change the parameters during data processing to get the expected results. The parameters are shown in Figure 4-23. Afte r done the correlation analysis, user also can change the roughness index parameters to get the correct roughness index. The parameters are shown in Figure 4-24. Figure 4-23 Filter Setup Form 73

PAGE 86

Figure 4-24 Roughness Index Parameters Form 74

PAGE 87

75 CHAPTER 5 SYSTEM VERIFICATION In the inertial based laser profiler system there are encoder distance unit and laser distance sensor. The encoder distance unit is used to collect the longitudinal distance of the pavement and speed of the vehicle. Lase r distance sensor is used to measure the distance between the mounting board and th e road surface. According to field and laboratory test results, it is proven that the accuracy of the encoder distance unit and laser distance sensor is enough for the system. In order to calibrate the sensors, we collected several sets of data in lab and fields. Analysis was done based on the data collected. This chapter will focus on two parts: sensors calibration and the system verification. 5.1 Principle of Sensors Measurement There are mainly two kinds of commercially available speed/l ongitudinal distance measurement sensors. One is Doppler radar se nsor such as Doppler Radar Speed Sensor from GMH engineering and Model 38000 Speed Sensor by AG Express Electronics Inc. Radar speed Sensor uses a microwave (radar ) signal and the princi ple of Doppler shift (measuring the changes in the frequency of reflected signal) to measure ground speed. The sensor output is frequenc y. The other one is optical rotary incremental encoders. A code disk is the heart of the incremental encode rs. As the code disk rotates in front of the

PAGE 88

stationary mask, it shutters light from the LED. The light that passe s through the mask is received by the photo-detector, which produces pulses in the form of a quasi-sine wave. The encoder electronics convert the sine wave into a squa re signal and is ready for transmission to a counter. The price of D oppler sensor is usually much higher than incremental encoders. Based on the cost/per formance comparison, the rotary incremental encoder was selected as the longitudinal di stance measurement unit (as shown in Figure 5-1). Figure 5-1 Rotary Incremental Encoder The laser distance sensor uses the lase r triangulation to measure the distance accurately. The laser puts a small bright spot on the object. The receiver of the sensor detects the position of this spot. The angle in creases with the distance of the object. The sensor basically measures this angle and th en calculates the distance. The laser distance sensor used in the profiler provides an analog output voltage proportional to the distance between the sensor and the object. The m easurement range is from 100mm to 500 mm. The supply voltage for the sensor is from 12 volts to 28 volts. The relationship between 76

PAGE 89

77 the voltage output and distance measured is pr oportional. As we tested in the lab, it was found that the voltage output of the se nsor changes with the supply voltage. 5.2 Sensors Verification It was necessary to conduct lab and fiel d experiment to test the incremental encoder accuracy and calibrate the laser dist ance sensors amplifier of the output to measured distance. Incremental encoder is applied to measur e the longitudinal dist ance. The output of the sensor is measured using the counter input of the AD converter. It is mandatory to get the distance per pulse. There are two steps for this test. First, several predefined test distance in the field was setup to get the dist ances per pulse after the encoder is installed on a vehicle. The vehicle moves along a line drawn on the road, then the test program will get the pulses traveled for the distance. Second, after get the distances per pulse, the vehicle will travel at several different distances to test the encoder accuracy. The calibration results and ve rification results are show n in Table 5-1 and Table 5-2. The values are the average value of 5 repeated runs. From the results, we can conclude that incremental encoder has good accuracy to be used in the profiler. Table 5-1 Incremental Encoder Calibration Results Distance Counter Distance per Count 50 25343 0.00197293 100 50787 0.00196902 150 75884 0.001976701 Average 0.001972883

PAGE 90

78 Table 5-2 Encoder Verification Results Measured Distance (m) Actual Distance (m) Measured Difference (m) 80 80.52 0.52 140 139.83 -0.17 200 200.237 0.237 The amplifier of the laser distance sensor is calibrated in the following way: First, set the laser distance sensor to a specified distance to a reference surface and get the output voltage at this position. Then add a small block with standard thickness to the reference surface and get the voltage output. By doing this for several times, then use a linear regression to get the re lationship between the voltage difference and the thickness of the block. The slope of the regression result is the sensor amplifier of the voltage to the actual distance. Second, move the laser di stance sensor the any position and get the voltage output, then move the sensor to any position inside the measure rage, get the voltage. Then multiply the amplifier with the vo ltage difference to get the distance. If the calculated distance is the same as the measured distance, then the calibrated amplifier can be used in the profiler. The results are show in Table 53 and Table 5-4. The values in the table are average value of 5 repeated runs. Table 5-3 Distance Sensor Calibration Results Measured Height (cm) Voltage Differe nce (v) Calculated Amplifier (cm/v) 5.478 2.24312 2.44214 11.078 4.53613 2.44216 16.546 6.77521 2.44214 Average 2.44215

PAGE 91

79 Table 5-4 Distance Sensor Verification Results Measured Distance (cm) Calculated Distance (cm) Measured Difference (m) 2.048 2.047 0.001 7.502 7.503 0.001 10.218 10.215 0.003 15.176 15.179 0.003 From the calibration and verification results, the laser distance sensor is accurate and good for profiler purpose. 5.3 Software Output Verification One of the most important concerns dur ing the system development period is the accuracy of the system result output. It de termines the reliability of the system. The major reason of the LIPRES softwa re output verification is because the output patterns of the system include graph output of which shows the profile of the pavement section, the table outputs of which include all important indices like IRI, RN number and the two types of output are all need to be verified with the standard tools. RoadRuf software is an integrated se t of computer tools for interpreting longitudinal road roughness profile data. It is intended to provide well-tested profile analyses such as the International Roughness Index (IRI) and Ride Number (RN) in a user-friendly package suited for immediate use by profile users. It is also intended to provide a benchmark for developers of profiler analysis software.

PAGE 92

80 RoadRuf was developed at The University of Michigan Transportation Research Institute (UMTRI) with funding from the Federal Highway Administration (FHWA) under a research project called Interpretation of Road Roughness Profile Data. RoadRuf includes advanced analysis capabili ties to support research activities. In addition to IRI and RN analysis, the software includes an interactive X-Y plotter, and a set of custom filters. Based on the above reasons, we choose the RoadRuf software as our result verification tools. During the verification period of the study, we collected 5 sites of the data. Table 5-5 and Table 5-6 show the IRI and RN resu lts collected from the field tests using the LIPRES software. L1, L2, L3 are the test resu lts of left wheel path. R1, R2, R3 are the test results of right wheel path. The distance between the left wheel path and right wheel path is 1.6 m. All the data from the five co llected sites are inputte d into the RoadRuf and results are compared with the LIPRES so ftware output. Figure 5-2 shows the output screen of the RoadRuf software after running the site1 data. Table 5-5 IRI Results IRI Results Site No# L1 L2 L3 R1 R2 R3 1 4.47 4.26 4.17 3.17 3.82 4.72 2 4.91 5.92 4.92 4.63 5.29 5.88 3 4.11 3.35 3.65 3.97 3.63 3.7 4 4.85 4.67 4.56 4.74 4.56 4.6 5 4.76 4.85 4.95 4.6 5.02 4.91

PAGE 93

Table 5-6 RN Results RN Results Site No# L1 L2 L3 R1 R2 R3 1 2.04 2.12 2.21 2.58 2.33 1.92 2 1.99 1.64 2.02 1.81 1.57 1.32 3 1.98 2.55 2.39 2.3 2.48 2.52 4 2.35 2.48 2.19 2.19 1.97 1.74 5 2.08 2.06 2.14 2.05 1.89 1.98 We can find the results matched very well, totally the same. It shows the LIPRES softwares functions are reliable correspond to both the smooth road condition and rough road condition. Figure 5-2 RoadRuf Output 81

PAGE 94

82 CHAPTER 6 DATA COLLECTION AND DATA ANALYSIS In order to verify the system function of two direct type roughness-measuring systems, we collected several sets of da ta under different conditions using several different profilers (laser prof iler, direct type profiler) to evaluate the repeatability, accuracy, and correlation of the systems. Sa tisfactory results were show after the repeatability, impact of speed effect and sa mple interval analysis. Correlations of the roughness indices are conducted be tween laser profiler and dire ct type profiler. Finally, Spectral analysis was performed over different pavement roughness conditions. 6.1 Consideration for Field Data Collection It was necessary to conduct field experime nt to verify whether the laser profiler functions well enough to match the design requir ements and also to test the correlation results between the laser profiler and direct type profiler system. In order to verify function of laser profiler, field data collection was performed in Tampa Bay area. The true profile of each se ction was measured using the direct type walking profiler. The elevations from the profile measurement were converted to the IRI and RN standard statistic, representing th e sections true roughne ss. The laser profiler

PAGE 95

83 then was operated along the same sections at different speed and IRI and RN was also obtained for each section. The test sites consist of flexible pavements only. The models developed for this research are based on the combined data collected from the flexible pavements. The test sites were selected be cause they can provide the broad range of roughness index levels and could be safely run at the 50km per hour testing speed. The smooth sites were need to ensure that the system has the resolution necessary to measure smooth pavement road (like new pavement road, etc) correctly, while the rough sites ensured that the system could handle the la rge amplitudes generated when traveling on rough or damaged pavement. The medium secti ons allowed data points to be located between the two extreme pavement conditions. A calibration section should be straight with no curves or turns, and should have as flat a grade as possible. Based on the suggestions from FDOT Each section has a length of 150 meters, which have been commo nly used in the United States and many other parts of the world. Measurement obt ained from a specific vehicle depends on primarily on the vehicle condition. If a new vehicle was used for test purpose, a calibration process needs to be done before test In this study, the same vehicle was used through the data collection. Each test sites should be located on a lightly traffic road and has a sufficiently wide roadway. For the direct t ype profiler, an entire lane of the section was used. If a section has higher traffic volume, it should have at least two lanes in each direction. Extra care should be taken to protect the surveyor in term of tr affic rerouting signs, flagging and traffic cones. Most of test sites were blocked when the direct-type system survey was conducted.

PAGE 96

84 6.1.1 Test Consideration for Direct Type Profiler a) Testing speed: the direct type profiler is designed for operations in the walking speed range. b) Number of repeat runs. Three repeat r uns were made for each wheel path at each test section. The mean values of the re ported roughness statistics were calculated and used as the summarized statistic. c) Sample Interval: The sample interval fo r direct type profile r was fixed at 30 cm. The filter is set to none filter status. The summary statistics were reported for the entire length of a test run. d) Testing distance: A total testing di stance for each section was 150 meters. 6.1.2 Test consideration for Laser Profiler a) Testing speed: The laser profiler is desi gned for operation in the normal traffic speed range. The field tests were conducted at speeds of 30, 50 and 70kmph for each test section. b) Number of Runs: Five repeat runs were made at each test speed on each test section. The mean values of the repeated r uns were calculated and taken as the summary roughness statistic for correlations with the rough ness statistic from di rect type profiler. c) Data sampling interval: The data sampling interval is adjustable in the system. The data sample intervals of the roughness measuring system are 2cm, 4cm, 8cm and 10cm. d) Test distance: The total distance for each test section was 150m. It is the same distance as the one tested using direct type profiler.

PAGE 97

85 6.2 Data Analysis 6.2.1 Repeatability Repeatability refers to the capability of a measuring device to obtain statistically similar results form repeated runs with m easuring conditions unchanged. Repeatability is one of the most important quality measures used to evaluate the performance of a measuring device. For the laser high speed pr ofiler, since different operating speeds and sampling rates were used, thus, for each combination of operating speed and sampling rate, five repeated runs were used to obta in IRI and RN value on each calibration section. Table 6-1 shows the IRI and RN values obtained by the laser high speed profiler from repeated runs on each calibration sec tion using the trend remove algorithm. The difference of IRI and RN values between repeated runs was used to quantify the repeatability of laser high speed profiler. From Table 6-1 and Tabl e 6-2, it can be seen that the over-all average difference of IRI between repeated runs was 0.055 and the overall-all average difference of RN between repeated runs was 0.058 m/km. Table 6-1 IRI Values Between Repeated Runs with Trend Remove Algorithm 1 2 3 4 5 Max. Diff. Section 1 2.07 2.03 2.02 2.03 2.01 0.06 Section 2 2.47 2.51 2.47 2.48 2.47 0.04 Section 3 3.36 3.32 3.34 3.35 3.38 0.06 Section 4 4.61 4.61 4.65 4.70 4.62 0.08 Over-All Average Difference = 0.055

PAGE 98

Table 6-2 RN Values Between Repeated Runs with Trend Remove Algorithm 1 2 3 4 5 Max. Diff. Section 1 3.26 3.20 3.27 3.29 3.26 0.09 Section 2 3.02 3.00 3.02 3.01 3.04 0.04 Section 3 2.69 2.65 2.66 2.65 2.67 0.04 Section 4 1.52 1.53 1.50 1.51 1.47 0.06 Over-All Average Difference = 0.058 In order to test the repeat ability of the roughness output of the laser profiler, the repeatability index used for the evaluation was: x xx NN i 1 2)( 1 Re Where N is the number of the repeat runs, is the output of the i ix th run, and x is the mean value of thes. The quantity Re represents the relative difference between the test runs. If Re is small, for example 5 pe rcent, it can be said the systematic and operational repeatability of the lase r profiler on a test section is good. ix Table 6-3 shows the resulting calculated Re values for each test section. All the Re values are less than 2 percent, which indicates that the repeatability of the laser profiler is relatively reliable. Table 6-3 Re Values of IRI and RN for Each Section with Trend Remove Algorithm Section 1 Section 2 Section 3 Section 4 Mean Values Re (IRI) 1.001% 0.63% 0.60% 0.74% 0.743% Re (RN) 0.93% 0.44% 0.56% 1.37% 0.825% 86

PAGE 99

87 Table 6-4 shows the IRI values obtained by the laser high speed profiler from repeated runs on each calibration section without the trend re move algorithm. Table 6-4 IRI Values Between Repeated Runs without Trend Remove Algorithm 1 2 3 4 5 Max. Diff. Section 1 2.10 2.34 2.13 2.29 2.15 0.24 Section 2 2.49 2.57 2.58 2.55 2.60 0.11 Section 3 3.45 3.75 3.6 3.37 3.69 0.38 Section 4 6.46 6.37 6.65 6.67 6.63 0.30 Over-All Average Difference = 0.258 The repeatability for each section w ith/without trend re move algorithm are calculated and shown in Table 6-5. Table 6-5 Re Values with and without Trend Remove Algorithm Section 1 Section 2 Section 3 Section 4 Mean Values Re (IRI) (without trend remove algorithm) 6.40% 2.35% 7.56% 4.66% 5.24% Re (IRI) (with trend remove algorithm) 1.001% 0.63% 0.60% 0.74% 0.743% From the table above, we can see that using the trend remove algorithm, the average repeatability was lowered from 5.24% to 0.74%. This indicates that the used trend remove algorithm improved the repeatability of the inertial base pavement roughness measuring system. The inertial based laser pavement roughness measuring systems have average repeatability of 5% from the past researches. The repeatability is decreased to 0.7% by using the trend remove algorithm. So a better repeatability can be obtained using trend remove algorithm.

PAGE 100

88 6.2.2 Impact of Sampling Speed Since the laser high speed profiler uses an accelerometer to measure vehicle vertical acceleration of ARAN profiler, the operating spee d of the laser profiler may not have certain impact on the roughness measuremen ts of the laser profiler. To evaluate the speed impact of the laser high speed profiler, different oper ating speeds were used in field data collection. Actually, for a given operating speed and sampling interval, three repeated runs were used on a calibration section. Thus, average value from the repeated runs was used. Table 6-6 and Table 6-7 show th e IRI and RN values of laser profiler at different operating speeds. The relative errors are also shown in each table. Table 6-6 IRI Values at Di fferent Operating Speeds Relative Error (Take 50kph as Reference) 30 kph 50 kph 70 kph 30 kph 70 kph Section 1 2.44 2.42 2.48 0.83% 2.48% Section 2 2.43 2.34 2.38 3.85% 1.71% Section 3 4.65 4.72 4.75 1.48% 0.64% Section 4 2.16 2.10 2.14 2.86% 1.90% Table 6-7 RN Values at Di fferent Operating Speeds Relative Error (Take 50kph as Reference) 30 kph 50 kph 70 kph 30 kph 70 kph Section 1 3.03 3.04 3.08 0.83% 2.48% Section 2 2.92 2.98 3.01 3.85% 1.71% Section 3 1.50 1.52 1.47 1.48% 0.64% Section 4 2.91 2.99 3.05 2.86% 1.90%

PAGE 101

Figure 6-1 and Figure 6-2 grap hically present the same data. The values in the table were the average values of RN from repeated runs and at different sampling intervals. From the table and the figure, it can be concluded that the operating speed did not have significant impact on IRI and RN value because there were no significant differences among the IRI and RN values for a given calibration section. In fact, theoretically, the roughness measuring devices with the use of vertical accelerometer and laser distance sensor should not be sensitive to operating speed. 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5Section 1Section 2Section 3Section 4IRI Value (m/km) 30 kph 50 kph 70 kph Figure 6-1 Operating Speed Impact on IRI 89

PAGE 102

0 0.5 1 1.5 2 2.5 3 3.5 Section 1Section 2Section 3Section 4RN Value 30 kph 50 kph 70 kph Figure 6-2 Operating Speed Impact on RN 6.2.3 Impact of Sample Interval Sample interval may have certain imp act on the roughness out put, IRI and RN, of the laser high speed profiler. This research tried different sample interval to test whether the sampling rate had impact on IRI and RN Since the operating speed did not have impact on IRI and RN and th e laser high speed profiler showed good repeatability, IRI and RN values from different runs and di fferent operating speeds on a given calibration section were averaged so that the only factor, sample interval, could be evaluated. Table 6-8 and Table 6-9 present the IRI and RN valu es at different sample interval on several calibration sections. From the tables, it is clearly seen that IRI an RN value does not change as the sample interval changed. 90

PAGE 103

Table 6-8 IRI Values of Laser High Speed Profiler at Different Sampling Interval Sample Interval 8cm Sample Interval 12 cm Sample Interval 16cm Section 1 2.96 2.99 2.97 Section 2 2.69 2.71 2.72 Section 3 1.92 1.96 1.94 Section 4 3.36 3.34 3.32 Table 6-9 RN Values of Laser High Speed Profiler at Different Sampling Interval Sample Interval 8cm Sample Interval 12cm Sample Interval 16cm Section 1 2.66 2.66 2.69 Section 2 3.01 3.06 3.05 Section 3 3.25 3.30 3.29 Section 4 2.69 2.65 2.66 0 0.5 1 1.5 2 2.5 3 3.5 4 Section 1Section 2Section 3Section 4IRI Value (m/km) Sample Interval 8cm Sample Interval 12 cm Sample Interval 16cm Figure 6-3 Sample Interval Impact on IRI 91

PAGE 104

0 0.5 1 1.5 2 2.5 3 3.5 Section 1 Section 2 Section 3 Section 4RN Value Sample Interval 8cm Sample Interval 12 cm Sample Interval 16cm Figure 6-4 Sample Interval Impact on RN Figure 6-3 and Figure 6-4 gr aphically presents the difference of IRI and RN values at different sampling rates. It is cl early shown in the figures that IRI and RN do not change within certain sample inte rval as the sample interval changed. 6.2.4 Correlation Analysis A reliable measuring device should ha ve good correlation with standard reference. If a measuring device has a good co rrelation with standa rd reference and good repeatability, this device is said to be reliable with good measuring performance. For inertial based laser profiler, the IRI correlation with Dipstick and direct type profiler need to be analyzed. 92

PAGE 105

93 Dipstick has been considered standard device for field calibration because it has best accuracy performance as compared with other automated roughness measuring devices. If a roughness-measuring device has good correlativity with Di pstick, this device is considered having good correlation with sta ndard reference. Direct type profiler has been proved that it has good correlation with Dipstick on both IRI and RN. So it can be used to calibration the iner tial based laser profiler. The laser profiler and direct type walking profiler were operated in several test sections Repeated runs were performed and the average values from the repeated runs were used for correlation anal ysis. Table 6-10 presents the average IRI values from Laser profiler, direct type pr ofiler and the Dipstick. Table 6-10 IRI Values Collected by Laser Prof iler, Dipstick and Di rect Type Profiler Laser Profiler Dipstick Direct Type Profiler 3.4 3.23 3.213 3.68 3.36 3.420 1.51 1.8 1.689 1.9 2.41 2.367 7.03 6.49 6.399 6.39 5.97 5.734 3.18 3.26 3.642 4.35 4.06 4.182 6.43 5.46 5.900 5.04 5.3 4.792 Figure 6-5 shows the IRI correlation be tween Dipstick and laser high speed profiler. From this figure, it can be seen that the laser high speed profiler has good

PAGE 106

correlativity with Dipstick (R 2 = 0.9694). Thus, it is reas onable that the correlation between Dipstick and laser high speed profiler is good. y = 0.8169x + 0.6289 R2 = 0.9694 0 1 2 3 4 5 6 7 01234567Laser Profiler IRI (m/km)Dipstick IRI (m/km 8 ) Figure 6-5 IRI Correlation Between Laser Profiler and Dipstick Since direct type profiler showed good correlation with Dipstick, direct type profiler could be used as a standard refere nce to calibrate laser high speed profiler. During field data collection, the laser high speed profiler produ ced roughness data at different wavelengths (bandwidth), including 100 meter wavelength and full wavelength (unfiltered bandwidth). Usually, IRI should be processed from pavement surface longitudinal profile with wavelength bandwidth in the range of 60 meters to 50 meters. Thus, the correlation analysis was based on the filtered data with a 100 meters wavelength. Figure 6-6 shows the IRI correlati on of laser high speed profiler and direct type profiler. 94

PAGE 107

y = 0.8104x + 0.6562 R2 = 0.9847 0 1 2 3 4 5 6 7 01234567 IRI of Laser Profiler (m/km)IRI of Direct Type Profiler (m/km 8 ) Figure 6-6 IRI Correlation Between Laser Profiler and Direct Type Profiler Based on the correlation analysis results, it is found that the correlation between laser high speed profiler and direct type profiler is good. 6.3 Maximum Entropy Spectral Analysis As described before, the profile raw data were collected from each site. Then the LIPRES software gets the data and builds the profile. Spectral analysis was carried by the developed software. Conventional Fast Fourier Transform (F FT) method has the disadvantages that is has low resolution to the amplitude of some frequency; some amplitude of segment of certain wavelength ma y be weakened or even disappeared in the spectral curves. So maximum entropy spectral an alysis method was used in the software to analyze the spect ral response of the calculated profile. 95

PAGE 108

From the previous discussion, it is know that the laser profiler is sensitive to the wavelength from 0m to 25m. Thus, the maxi mum wavelength of 25m was chosen in the analysis. It would be expected that if pa vement condition was poor, the magnitude of the spectral density function in 0 to 25 meters would be large. If the pavement condition is good, it would results in relatively low magnitude of the spectral density function. Compare the shapes of Figure 6-7 and Figures 6-8, one road section is relatively smooth, the IRI value is 2.19. While IRI value for th e other one is 4.36, which is bad condition. We can see that the spectra l density function is affected by the pavement roughness situation significantly. In good pavement c ondition, the magnitude of the function was distributed in the 10 to 25 meters wavele ngth range. Major parts of the spectral are distributed in two ranges: one is in the range of 0 to 10 meters wavelength, and the other is in the range of 0 to 10 meters wavelength. Figure 6-7 Spectral Curve of Good Pavement Condition 96

PAGE 109

Figure 6-8 Spectral Curve of Bad Pavement Condition Figure 6-9 and 6-10 are the plot picture of the two corresponding profiles. Figure 6-9 Pavement Profile with IRI = 2.19 m/km 97

PAGE 110

Figure 6-10 Pavement Profile with IRI = 4.36 m/km From the comparison of the profiles and th eir spectrum, the bad pavement has the big amplitude in the sinusoids with wavelength of 10 25 meters. It is affected by the short wavelength, so the IRI is high. How ev er the good pavement is not much affected by the 0-10 meters wavelength and was most ly affected by the long wavelength, the amplitude of the sinusoids with the long wave length is low. That is the reason that the roughness index IRI is much lower. Two test sites have the almost same IRI value (3.06 and 3.13 m/km respectively), but their RN is different. The spectral curves for these two sites ar e shown in Figure 6-11 and Figure 6-12. From the spectru m analysis of the two paveme nt profiles, it is clear to find the reason why the road has different RN values. From the figures, the long wavelength parts contributed to the IRI, wh ile there are some minor differences on the short wavelength part. For the pavement in Figure 6-11, the amplitude short wavelength part is low. So the pavement has good ride quality and high RN value. While for the 98

PAGE 111

pavement in Figure 6-12, the amplitude of the short wavelength pa rt is a little high compared to the other one, so the RN va lue is lower than the one in Figure 6-11. Figure 6-11 Spectral Curve of Pa vement Profile with High RN Figure 6-12 Spectral Curve of Pa vement Profile with Low RN 99

PAGE 112

100 From the analysis, it shows that the ma ximum entropy spectral analysis shows a qualitative relation of different pavement roughness conditions. From the spectrums, the pavement roughness conditions can be easily id entified if the pavement conditions differ significantly. But if there conditions differenc es are slight, it is hard to tell which pavement section is better or worse. Then the roughness index IRI and RN are used to identify the pavement conditions quantitatively. From the spectrums, it is still easy to see which wavelength is the major part in a particular pavement section.

PAGE 113

101 CHAPTER 7 SUMMARY, CONCLUSION AND RECOMMENDATION 7.1 Summary An inertial-based pavement roughness ev aluation system was developed in this research and study. It is a vehicle mounted instrument and can be operated in speed between 25 kmph and 80 kmph. It is classified as a class III instrument for pavement roughness measurement.. It is designed to fac ilitate the efficient data collection and presentation of pavement surface informa tion, including pavement profile, grade and roughness index such as IRI and RN. It can be used for network level data collection and analysis of pavement roughness. The purpose of this research was to develop the inertial based pavement roughness evaluation system, the trend remove algorithm to improve repeatability, evaluate system repeatability, speed effect and sample interval effect, finally test whether the laser high speed profiler has good correlation to Dipstick and direct type profiler. To reach the purpose, standard roughness measur ing systems were used as references in fields to evaluate whether the laser high speed profiler had good correlation with these standard references. The reference measurements included IRI valuates collected by Dipstick, direct type profiler. However, si nce Dipstick do not have the function to

PAGE 114

102 produce RN values, only the direct type profile r was used to evaluate the laser high speed profilers performance in measuring RN values. Field tests were performed in Tampa, Florida. The corresp onding IRI values and corresponding RN values for each test sec tion were obtained. The laser high speed profiler was operated at different sampling interval and at different speed (30 kph, 50 kph, and 70 kph). All devices were operated for at least three repeated runs. After field data were obtained, data an alysis was performed to evaluate trend remove algorithm function and the measuring performance of the la ser high profiler in obtaining IRI and RN values. The performanc e was evaluated based on the impact of vehicle speed and sampling interval, repeatab ility, and correlativity with direct type profiler, etc. Linear regression analysis (correlation analysis) was performed to evaluate the IRI correlativity between laser high speed profiler and direct type profiler, Dipstick. The correlation analysis was also performed to ev aluate the RN correlativity between direct type profiler and laser high speed profiler. The maximum entropy spectral analysis show s a qualitative rela tion of different pavement roughness conditions. From the sp ectrums, the pavement roughness condition can be easily identified if the pavement condition differs significantly. Even if the pavement roughness conditions are slight, it is still can tell the small difference in RN value from the spectral analysis. From the sp ectrum of pavement profile, it is easy to see which wavelength is the major part in a particular pavement condition.

PAGE 115

103 7.2 Conclusion From data analysis, it was found that lase r high speed profiler showed satisfactory repeatability performances. Thus, for real da ta collection by lase r high speed pavement evaluation system, if the data collection proced ure is well controlled, there is no need to run these devices more than three repeated runs because the difference between different runs could be ignored at the network level. Fo r project level analysis it is better run the device more the three repeated runs and get the average va lue to get precise roughness index. Laser high speed profiler shows very good repeatability and also could be operated at different operating speeds (20 kph 100 kph) with very little difference in IRI and RN values for a given test section and sampling interval. However, any speeds beyond the speed range may not produce the sa me conclusion because the analysis was based on the speed range and no conclusion is supported if the speed is beyond the speed range. From the data analysis, it is proved that the computation method used in the roughness measuring system removed the phase shift introduced by other computation methods and the repeatability was increased with the Re value less than 2% by using trend remove algorithm. The laser high speed profiler and direct type profiler had good correlations at different sampling intervals and operating sp eed of the laser high speed profiler. Thus, the laser high speed profiler can be used in the field application.

PAGE 116

104 7.3 Recommendation Some recommendations are presented here based on the analysis and system development experience. They could improve th e performance of the system and make it more precise. From the pavement test, all the test sec tions do not have big cracks and bumps on the pavement. From the profile calculation pr ocedure, the phase shifts because of the bump. So the issue that how to eliminating th e bump phase shift still needs to be better understood. More researches need to taken in the future. Also, how cracks affect the pavement roughness also needs to be identified. From the spectral analysis of pavement pr ofile, it shows that spectral density has relationship with road roughness. Pavement s ections with the same IRI are likely to contain different roughness details. New di gital signal processing techniques may identify the different roughness details amon g such pavement sections and such researches may need to be done.

PAGE 117

105 REFERENCES 1. Bennett, C.R., Calibrating Road Roughness Meters in Developing Countries, Transportation Research Record 1536. 2. Gillespie, T.D., Everything You Always Wa nted to Know About the IRI, But Were Afraid to Ask, Presented at the Fourth Annual Road Profile Users Group Meeting, Lincoln, Nebraska. (1992) 13p. 3. Gillespie, T.D., Sayers, M.W. and Segel, L., Calibration of Response-Type Road Roughness Measuring Systems. NCHRP Report No. 228, Dec. 1980. 4. Hajek, J.J., Musgrove, G., A Switch to the Internationa l Roughness Index, 77 th Annual Meeting of Transportation Research Board, Washington D.C., 1998. 5. Janoff, M.S., Pavement Roughness and Rideability Field Evaluation, NSHRP Report 308, National Research Council, Washington, D.C., 1998. 6. Liu, C., Herman, R., Road Profiles, Ve hicle Dynamics, and Human Judgment of Serviceability of Roads: Spectral Fre quency Domain Analysis, Journal of Transportation Engineering, pp106-111, 1998. 7. Lu, J., Spectral Analysis of Vehicle Sp eed Characteristics. In Transportation Research Record 1375, National Resear ch Council, Washington D.C. 1992. 8. Lu, J., Carl, B.B., and Hudson, R., Evalu ation and implementation of the roughness Measuring Subsystem of the ARAN Unit., Un iversity of Texas at Austin, February, 1991. 9. Lu. J., et al., Speed Effect Analysis and Canceling Model of a Response-Type Road Roughness Measuring System, Transpor tation Research Record 1260, pp26-36, 1990. 10. Lu, J., Hudson, R., and Bertrand, C., "Eva luation of the Roughness System of the Automatic Road Analyzer", Transp ortation Research Record 1435, pp38-44, 1994. 11. Sayers, M.W., Steven, M. K., The Little Book of Profiling, September 1998.

PAGE 118

106 12. Shahin, M.Y., Pavement Management for Ai rport, Roads, and Parking Lots, 1994, 572p. 13. Oppenheim, A.V., Schafer, R.W., Buck, J. R., Discrete-Time Signal Processing, Second Edition, Prentice Hall Press, 1999, 870p. 14. Pong, M., Wambold, J., Evaluation of Co mputation Methods for AccelerometerEstablished Inertial Profiling Reference Systems, Transportation Research Record 1348, National Research Council, pp 8-17, 1992. 15. Press, W.H., Teukolsky, S.A., Vetterling, W.T. and Flannery, B.P., Numerical Recipes In C++: the Art of Scientific Computing, Second Edition, Cambridge University Press, 2002, 1032p. 16. Sayers, M.W., On the Calculation of International Roughness Index from Longitudinal Road Profile. Transportati on Research Record 1501, National Research Council, Washington, D.C., 1995. 17. Sayers, M.W., Gillespie, T.D. and Paterson, W.D., Guidelines for the Conduct and Calibration of Road Roughness Measurements. World Bank Technical Paper 46, The World Bank, Washington, DC., 1986. 18. Spapgler, E. and Kelly, W., GMR Road Profilometer-A Method for Measuring Road Profile, HRR121, Highway Research Board, 1966. 19. Karamihas, S.M. etc., Report 434 Guide lines for Longitudinal Pavement Profile Measurement, Transportation Resear ch Board, National Academy Press, Washington, D.C., 1999. 20. Standard Test Method for Measuring Longitudinal Profile of Traveled Surfaces with an Accelerometer Established Inertial Profiling Reference, Annual Book of ASTM Standards, Vol. 04.03, E950, pp702-706, 1996. 21. Sun L., Simulation of pavement roughness an d IRI based on power spectral density, Mathematics and Computers in Simulation, pp 77-88, 2003. 22. Yoder, E. J., Hampton, D., Pavemen t Profile and Roughne ss Measurements; A Review of Method, Purdue University, January, 1958.

PAGE 119

ABOUT THE AUTHOR Fengxuan Hu received a Bachelors Degr ee in Electrical Engineering from Wuhan Institue of Technology in Wuhan, China in 1993. He got his M.S. in Electrical Engineering from Tongji University in Shanghai, China in 1996. He enrolled in the Ph.D. program in Civil and Environmental Engineering Department at the University of South Florid a in 2001 as a graduate research assistant. Fengxuan Hus dissertation focuses on the de velopment and evaluation of road roughness measurement system. His research areas in clude transportation re levant instrument design, traffic engineering and the developmen t of Intelligent Transportation system.


xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001798404
003 fts
005 20070801101547.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 070801s2006 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001641
040
FHM
c FHM
035
(OCoLC)162019203
049
FHMM
090
TA145 (ONLINE)
1 100
Hu, Fengxuan.
0 245
Development and evaluation of an inertial based pavement roughness measuring system
h [electronic resource] /
by Fengxuan Hu.
260
[Tampa, Fla] :
b University of South Florida,
2006.
3 520
ABSTRACT: Roughness is an important indicator of pavement riding comfort and safety. It is a condition indicator that should be carefully considered when evaluating primary pavements. At the same time, the use of roughness measurements plays a critical role in the pavement management system.There are many devices used for roughness evaluation. The major tools used for road roughness quantify are the road profilers. In the thesis research, in order to obtain useful pavement surface condition data for pavement evaluation, an inertial based pavement roughness measuring system was developed with the combination of modern sensor technology and computer technology. The research will focus on the development of new method to get the profile in order to improve the repeatability of the inertial based pavement roughness system, the hardware design and the software development which is used for data sampling and data analysis. Finally maximum entropy spectral analysis method was used to evalu ate the road profile spectrum.In order to get evaluate the accuracy and correction of the laser profiler system, different roughness devices (including Dipstick, direct type profiler and the laser profiler developed) were operated in the test sites. The research focused on several performance measures, such as repeatability (before and after new method analysis), impact of operating speed and sample interval, correlativity and etc. IRI from these devices were analyzed to evaluate the correlativity between these devices. Some regression models were developed in this research. Test results show that the new method can improve the repeatability of the profiler system. The laser profiler system has good repeatability and the operating speed and sample interval do not have a significant impact on the inertial based roughness measuring system. With the reliable results, the system is ready to be used in the field application within the speed and sample interval range. Through the spectrum an alysis, it shows that the spectrum has a qualitative relation with pavement roughness conditions.
502
Dissertation (Ph.D.)--University of South Florida, 2006.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 106 pages.
Includes vita.
590
Adviser: Jian John Lu, Ph.D.
653
Repeatability.
spectral analysis.
Iri.
Rn.
Speed impact.
690
Dissertations, Academic
z USF
x Civil Engineering
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.1641