Evaluation of the Siometer as a Device for Measurement of Pavement Profiles

Transcription

1 112 TRANSPORTATION RESEARCH RECORD 126 Evaluation of the Siometer as a Devie for Measurement of Pavement Profiles EMMANUEL G. FERNANDO, ROGER s. WALKER, AND ROBERT L. LYTTON Highway engineers have always been onerned with providing pavements of aeptable servieability. The servieability of a highway segment, whih is iargeiy a funtion of pavement roughness, is a widely u ed riterion for deiding when pavements are in need of rehabilitation. For this appliation, various statistis are urrently used as indiators of pavement servieability, the most ommon being the present servieability index. These statistis are largely determined from measurements of pavement roughness. Various devies and proedures have been developed for aomplishing these measurements. [ pratial n ity, devies for measuring pavement roughne s must be apable of providing repeatable measurements at normal highway speeds. In addition, devies that do not require diffiult alibration proedures, that possess the apability for field proessing of the data olleted, and that are relatively inexpensive to own, operate, and maintain ar most desirable. The Siomer, whih is urrently us d by the Texas State Department of Highway and Publi Trau portation (SDHPT) for v<i luation of pavement riding quality, hold pl'omise a an instrument for the routine olletion of profile data on a networkwide sale. The Texas SDHPT has reently begun investigating the profilemeasuring apability of the Siometer. A unique feature of this devie is the statistial modeling proedure for haraterizing the vehile on whih it is installed, whih lends portability to the Siometer. In it, the parameters of the statistial model are determined in a selfalibration proedure that is run before profile data are olleted. To evaluate the appliability of the Siometer as a devie for profile measurements, profile measurements with the Siometer were ompared with those from a profilometer. Pavement roughness is the prinipal determinant of riding quality as pereived by the road user. In order to provide roads that offer a smooth and omfortable ride, a transportation ageny requires measurement tehniques for quantifying pavement surfae roughness. An evaluation was made of a profilemeasuring devie known as the Siometer. This devie, developed by Dr. Roger Walker of the University of Texas at Arlington, is used by the Texas State Department of Highways and Publi Transportation (SDHPT) for evaluating the riding quality of pavement setions in the state. Pavement surfae profiles measured with the Siometer were ompared with those determined from the surfae dynamis profilometer (SDP). Over the years, the SOP has gained wide aeptane as a devie for evaluating pavement profiles. It is lassified as a Class 2 instrument by the World Bank (1) for the measurement of the international roughness index (IRI). The SOP was designed by General Motors and built by K. J. Law Engineers in Originally, it had as primary E. G. Fernando and R. L. Lytton, Texas Transportation Institute, Texas A&M University, College Station, Tex R. S. Walker, Department of Computer Siene Engineering, The University of Texas at i\rlington, i\.rlington, Tex sensors two aelerometers and two linear potentiometers, onneted to roadfollowing wheels. The aelerometers determine the amount and diretion of vertial aeleration experiened by the vehile, whereas the potentiometers and wheels measure the distane from the vehile body to the road surfae. A profile measurement is alulated by summing the double integral of the aelerometer signal and the displaement signal from the potentiometer (2). In the latest version of this devie, the potentiometers and roadfollowing wheels have been replaed by nonontat sensors. The SOP is apable of measuring profiles of onsiderable auray and onsisteny at normal highway speeds without the need for alibration. It has been used as a referene devie for measurement of present servieability index (PSI) within the Texas SDHPT. The prinipal statisti urrently used by the Department in omputing PSI from profile data is the rootmeansquare vertial aeleration (3). Although the SOP provides a fairly rapid and aurate method of determining pavement profiles from whih various roughness statistis an be omputed, it requires a large initial apital outlay and is relatively expensive to operate. Consequently, many state transportation agenies generally use responsi::lypt: road roughness measuring devies, suh as the Mays meter, for olleting roughness data on a networkwide basis. However, suh devies require periodi alibration, whih often entails signifiant effort. What is learly needed is a devie that an be used to ollet fairly aurate and onsistent profile data at normal highway speeds, that is relatively inexpensive to own and operate, and that does not require diffiult alibration proedures. A devie that has the potential of offering all of these advantages is the Siometer. The appliability of this devie for measuring pavement profiles is evaluated in the following setions. THE SIOMETER The development of the Siometer was initiated by Walker during the early 197s. A unique feature of this devie is the statistial modeling proedure for haraterizing the vehile on whih it is installed. Through this proedure, the influene of the vehile on the measurement proess is identified and removed ( 4,5). The statistial model is parameterized with the Siometer's onboard miroomputer using vertial aelerations of the vehile measured at fixed distanes as the vehile is driven down the road. Vertial aelerations are obtained from an aelerometer that is housed in a small ase and installed in the trunk of the vehile. One the parameters of the vehile ar determined, the Siometer is alibrated and

2 Fernando et al. ready for profile measurements. The vehile is then driven over the roadway setions for whih profiles are to be determined and the resulting aelerations are measured. The differenes between the atual measurements and those predited from the statistial model are used to estimate the road profile by integrating the aeleration differenes with the time between suessive samples. The primary appliation of the Siometer within the Texas SDHPT is for evaluation of riding quality. Thus, the devie beame known as the Siometer beause its primary output is the servieability index (SI) for a partiular pavement setion, even though the SI is alulated using statistis derived from the predited road profile. The devie is portable and an be easily transferred from one vehile to another. Furthermore, beause it implements a selfalibration proedure, the devie, in theory, an be installed in any vehile beause the effet of the vehile is modeled in the same proess. The urrent version of the Siometer used by the Texas SDHPT is the R68 system manufatured by Mirosher Inorporated. The R68 system onsists of three omponents, namely (a) a sensor unit, (b) a main ontrol module, and () a laptop omputer for storing the results. The system omputes and displays SI and predits the pavement profile. The sensor unit and the main ontrol module urrently ost $2,. The laptop omputer an be purhased separately by the user from any other vendor. The sensor unit inludes the aelerometer and a distanemeasuring signal. The aelerometer is housed in a small ase that is weighed down with a sandbag and mounted vertially inside the trunk of the vehile, where it measures the vertial aeleration. The signal from the aelerometer is transmitted to the main ontrol module where it is digitized in aordane with the distane signal and proessed. The main ontrol module ontains two Motorola 68 miroproessors working in parallel. One performs inputoutput operations and the other performs numerial omputations. The data storage omponent is a portable laptop omputer. A ommuniations program provides the interfae between the ontrol module and the laptop omputer. This program and the personal omputer provide the means of obtaining ontinuous SI or profile measurements. The entire Siometer system is portable and an be easily installed in most standard vehiles. An enhaned version has also been developed that implements the South Dakota method of measuring longitudinal profiles. The South Dakota profiler, urrently onsidered by many to be a Class 2 instrument, is beoming a popular devie for measuring pavement profiles. This devie measures pavement profile elevations by the use of an aelerometer and aousti sensors, whih perform the same funtion as the laser probes in the SDP. The South Dakota profiler differs from the SDP in this respet and also in the proedure used for integrating the aelerometer signal. The urrent version of the devie measures longitudinal profile elevations at the inner wheelpath and also provides estimates of pavement rutting using data from the aousti sensors. At present, the roll of the vehile is not onsidered in the determination of pavement rutting although numerous tests indi.ate reasonable agreement between rut depth data obtained manually and rut depth estimates from the profiler. Beause the Siometer an easily implement the South Dakota profiler onept by the simple installation of aousti sensors in the test vehile, it has reently been upgraded for this purpose and is undergoing evaluation by the Texas SDHPT. In addition, to identify the influene of vehile roll, the possibility of using additional aelerometers in onjuntion with up to five aousti sensors is being onsidered. This improvement will require modifiations to the Siometer hardware, but onsideration of vehile roll will provide the Texas SDHPT with the apability of measuring transverse pavement profiles in addition to longitudinal profiles. EVALUATION OF PAVEMENT PROFILES MEASURED FROM THE SIOMETER The Texas SDHPT has reently begun investigating the profilemeasuring apability of the Si m ter. The findings presented herein are based on re. ults obtained thus far. In order to evaluate the appliability of the Siometer for profile measurements, the SDP was used as a referene. In this evaluation, nine bituminous test setions were seleted on whih profile measurements using the SDP and the Siometer were made. Three of the test setions were smooth, three were rough, and the other three were intermediate. All setions were.2 mi in length. The servieability indies alulated from the SDP profiles on the nine seleted setions are shown in Table 1. All setions, with the exeption of TC7 in Tarrant County, are loated within the general viinity of Austin, Texas. The pavement profiles of the nine setions were measured using the SDP and Siometer of the Texas SDHPT. The SDHPT's SDP is similar in design to that originally built by K. J. Law exept that the potentiometer and roadfollowing wheel ombination has been replaed with two nonontat Selom laser probes. This feature has redued maintenane problems assoiated with the mehanial roadfollowing wheels and has allowed profile measurements to be onduted at faster highway speeds. In addition, data aquisition and proessing apability was upgraded to take advantage of improvements in hardware tehnology and thus allow data redution to be onduted in the field. Consequently, roughness statistis and profile data an now be obtained as soon as a run is ompleted on a partiular highway segment. For eah test setion seleted, two profile measurements were obtained from eah devie. Profile elevations were taken at.5ft intervals along eah.2mi setion. Beause the Siometer was portable, the devie ould be installed inside the SDP van. This allowed profile measurements to be made simultaneously on both devies for any given run, thus eliminating errors assoiated with runtorun variations, suh as differenes in wheelpaths traked between runs, differenes in vehile trak widths, and differenes in starting times between profile measurements. All measurements were taken at 2 mph in an attempt to traverse the same wheelpaths eah time a run was made on a partiular setion. On two of the rough setions (Setions 1 and 4), yellow dots painted at regular intervals on the wheelpaths were used to guide the diretion of travel between runs. In order to establish a benhmark for evaluating Siometer profiles, a omparison of the profiles from repeat runs of the SDP was initially made. Figure 1 shows a omparison of 113

3 114 TRANSPORT AT/ON RESEARCH RECORD 126 TARLF. 1 TF.ST SECTIONS WHERE PROFILE MEASUREMENTS WERE COLLECTED Setion TC7 Loation Deker Lake Road West, approximately.2 miles west of FM 973 Deker Lake Road East, approximately.3 miles west of FM 973 U.S. 183 South, 1.5 miles north of Burleson Road U.S. 183 North, 1.1 miles north of Burleson Road at oneway sign at rossover north of reek Peare Lane West, approximately.9 miles east of FM 973 FM 685 North, approximately.2 miles north of Phillips 66 gas station FM 973 South,.56 miles south of Shmidt Lane FM 3177 South, at Texas Heritage Center sign U.S. 183 frontage road, west bound, near interset ion with U.S. 157, in Tarrant ounty, north of Arlington Present Servieability Index (PSI)* * average PSI from 2 SDP runs on setion measured left wheelpath profile elevations from repeat runs of the SDP on Setion 1. The orrelation oeffiient r between the measured profile elevations was determined to be.985 (as shown in Figure 1) with a standard error of the estimate of approximately 91 mils. Similarly, standard errors of estimate and orrelation oeffiients between measured profile elevations from repeat runs of the SDP on the other test setions were alulated. The results are presented in Table 2. The orrelation oeffiients and standard errors of estimate shown in Table 2 were ompared with the orresponding statistis alulated using Siometer and SDP profile elevations measured during a given run (Table 3). In general, the orrelation oeffiients between SDP and Siometer profiles taken during the same run are omparable with the orrelation oeffiients between orrespondi11g SDP 1epliaie runs. in addition, for six of the nine test setions (i.e., Setions 1, 7, 12, 4, 42, and TC7), the standard errors of estimate alulated using SDP and Siometer profiles are somewhat better than those alulated using SDP repliate profiles. Figures 2, 3, and 4 show the generally favorable agreement obtained between SDP and Siometer profiles for data measured from the left wheelpaths of Setions 1, 7, and 4, respetively. An overall measure of the agreement between Siometer and SDP profile elevations was obtained by alulating the overall orrelation oeffiient between measured profile elevations from the two devies. Figure 5 shows a omparison of all measured profile elevations from the Siometer with the orresponding profile elevations from the SDP. The overall orrelation oeffiient between measured profiles taken during the same run from the two devies was determined to be.971, as shown in Figure 5. This is siightiy greater than the overall orrelation oeffiient of.96 between profile

4 Fernando et al. 115 N = 2198 OBS ::J r = ui.r::. (.) ts > w t<l Profile Elevation (inhes, r.un 1) FIGURE 1 Comparison of left wheelpath profile elevations from repeat runs of the SDP on Setion 1. TABLE 3 CORRELATION COEFFICIENTS AND TABLE 2 CORRELATION COEFFICIENTS AND STANDARD ERRORS OF ESTIMATE BETWEEN STANDARD ERRORS OF ESTIMATE BETWEEN REPEAT PROFILOMETER AND SIOMETER MEASUREMENTS PROFILOMETER MEASUREMENTS TAKEN DURING THE SAME RUN Standard Error Standard Error Correlation of Estimate Run Correlation of Estimate Setion Wheelpath Coeffiient (mils) Setion Number Wheelpath Coeffiient (mils) 1 left left right right left left right right left left right right left left right right left left right right left left right right left.% left right right left left right right TC7 left left TC7 right right left right left right elevations from repeat runs of the SDP. In addition, the over 31 2 left right all standard error of the estimate between orresponding pro 4 1 left.99{) 64 file elevations from the Siometer and the SDP was alulated 4 1 right to be approximately 9 mils. The same statisti alulated 4 2 left using orresponding profile elevations from repeat SDP runs 4 2 right was determined to be approximately 17 mils left right The slightly lower orrelation oeffiient between profile 42 2 left elevations from repeat SD P runs and the higher standard error 42 2 right of the estimate obtained are largely attributed to variations TC7 1 left in wheelpath traked between runs of the instrument. It is TC7 1 right TC7 2 left also likely that differenes in starting times between repeat TC7 2 right runs would have ontributed to the slightly higher variation

5 116 TRANSPORTATION R ESEARCH RECORD 126 en.. () > w ;;:: E ;; 1 N = 2197 OBS r = "...r SDP Profile Elevation Cinhes) FIGURE 2 Comparison of profile elevations measured with the SDP and Siometer for the left wheelpath of Setion 1 (Run 1)..6 ;;...4 ().2 :;:; ell > w.2.4 en N = 2196 OBS r = J<....r SDP Profile Elevation Cinhes) FIGURE 3 Comparison of profile elevations measured with the SDP and Siomeler for the left wheelpath of Setion 7 (Run 1). between orresponding profile elevations from the SD P. However, to ompensate for the effet of this fator, the SD P profiles from repeat runs were initially lined up before alulation of the statistis presented. This was done by means of rossorrelation analysis wherein profiles from repeat SDP runs were shifted relative to eah other until a maximum ross orrelation was obtained. The lose agreement between Siometer and SDP profiles taken under idential operating onditions lends redibility to the Siometer's approah for estimating pavement profiles. The essential element of this tehnique is the selfalibration sheme for parameterizing the statistial model of the vehile on whih the devie is installed. The alibrated statistial model provides a way of separating the vehile's ontribution to the measured vertial aelerations from the input attributable to the road profile. In essene, the road profile is estimated from integration of the differenes between measured aelerations and those predited from the statistial model. For this study, the right and left sides of the SDP van were modeled differently so that the statistial models for the right and left wheelpaths were different. In estimating pavement profiles with the Siometer, measured aelerations from the aelerometers mounted inside the SDP van were used in the omputations. Thus, the

6 1.6 Iii 1.2..r: ().8 :;::::;.4 ro > w.!!!.4 Qi.8 E en 1.2 N 2197 OBS r = SDP Profile Elevation (inhes) FIGURE 4 Comparison of profile elevations measured with the SDP and Siometer for the left wheelpath of Setion 4 (Run 1). 2 N = 79,56 OBS r =.971 en. > w?. lii Qi E i:i) 1 2'1;..rr 2 SOP Profile Elevation!inhes) FIGURE 5 Comparison of Siometer profile elevations with SDP profile elevations. 2

7 118 operating onditions under whih the SDP and Siometer profiles were taken were as lose to being ompletely idential as an be arranged. In this way, the omparisons between the SDP and Siometer profiles learly demonstrate the degree of apability of the Siometer's method of measuring pavement profiles. Judging from the results obtained, the Siometer's approah, based on measured vertial aelerations oupled with a statistial model of the vehile, leads to profiles that are omparable to those obtained from the SDP, whih is based on measured vertial aelerations and the use of nonontat probes (lasers) for determining the distane between the vehile and the ground at any given time. EVALUATIOt,J OF PROFiLE POWER SPECTRA The omparison of measured profiles between the SDP and the Siometer forms a basis for evaluating the appliability of the Siometer as a devie for measuring pavement profiles. However, the evaluation should not stop here beause differenes in the frequeny ontent of two pavement profiles may exist that are not readily apparent from a visual examination of the measured profiles. One an piture pavement profiles as onsisting of the sum of a variety of waveforms of different frequenies and amplitudes. Waveforms of low frequenies or long wavelengths may be identifiable from a visual examination of a partiular pavement profile. However, the highfrequeny omponents will in all likelihood be masked beause of the sales involved. Consequently, to obtain omplete information on the frequeny ontent of a partiular pavement profile, its power spetrum must be evaluated by means of spetral analysis. A power spetrum is a graph of the frequeny (as the absissa) versus the power, whih is the square of the amplitude of eah frequeny. In this way, the dominant frequenies or wavelengths within the profile an TRANSPORTATION RESEA RCH RECORD 126 be identified. In addition, by omparing the harateristis of two profiles in the frequeny domain, the similarity in the waveform omposition of the two profiles an be evaluated. A spetral analysis was onduted to determine the power spetra of the measured SDP and Siometer profile elevations. Figures 6 and 7 show the power spetra for the left wheelpath profiles of Setions 1 and 7, respetively. The higher the power at a given fre4ueny, the more dominant are the waveforms of that partiular frequeny within a given pavement profile. The results shown in Figures 6 and 7, whih are typial of those that were obtained for all of the other profiles, illustrate the reasonable agreement between the power spetral densities of orresponding SDP and Siometer profile elevations. In these figures, the power spetral density (PSD) is expressed in db units, defined herein as 1 1og 1 (amplitude squared per yle per foot). In order to evaluate the agreement between SDP and Siometer power spetral densities, the overall orrelation oeffiient between the PSDs was determined. Figure 8 shows the PSDs of Siometer profile elevations and the orresponding PSDs of SDP profile elevations. Power spetral densities determined from SDP and Siometer profiles taken during the same run were ompared. The overall orrelation oeffiient between SDP and Siometer power spetral densities was determined to be.99. This value ompares favorably with the overall orrelation oeffiient of. 993 between the PSDs of profile elevations from repeat SDP runs. In addition, a rootmeansquare statisti that provides an overall measure of the math between the amplitudes of SDP and Siometer power spetra was alulated from the following expression: II L (Y; Y/)2 RMSD (1) n 7 : 'O '(ii > 5 iii \ \ '\ "" a. (/) ;: 3 't'... osop + Siometer Frequeny (yles/foot).6.8 FIGURE 6 Power spetra of pavement profiles measured with the SOP and Siometer for the left wheelpath of Setion 1 (Run 1).

8 Fernando et al : "O > "iii 5 Cl iii.... Cl) \ 1% l... 3 y,,. ;:: ".oi;j ;:\ "' DSDP + Siometer Pr """ Frequeny (yles/foot> FIGURE 7 Power spetra of pavement profiles measured with the SDP and Siometer for the left wheelpath of Setion 7 (Run 1). 1 : "O 8 N = 234 OBS Cl r =.99 Cl) :;:::; ro 6 > jjj 4 a: Qi E U5 2 1";" SOP Profile Elevation PSD (dbl FIGURE 8 Comparison of power spetral densities of Siometer profile elevations with power spetral densities of SDP profile elevations. 8 where RMSD = rootmeansquare deviation, mils; Y; = SDP amplitude, mils; Y/ = Siometer amplitude, mils; and n = number of observations. Using Equation 1, the RMSD assoiated with the Siometer power spetra was determined to be 2.46 mils with 2,34 observations. A similar statisti alulated from the power spetra between repeat SDP runs was found to equal 3.87 mils with 1,17 observations. On the average, therefore, the amplitudes of the waveforms assoiated with Siometer profile elevations deviated from the amplitudes of the orresponding SDP waveforms by approximately 2.5 mils. Similarly, the amplitudes of the waveforms from repeat runs of the SDP differed, on the average, by about 4 mils. The higher RMSD obtained between amplitudes of power spetra from repeat SDP runs is again indiative of the effets of variations in wheelpaths traked between runs of the instrument. Judging from the statistis presented, it is evident that the Siometer power spetra ompare favorably with the orresponding SDP power spetra.

9 12 TRANSPORTATION RESEARCH RECORD ' ' '" F f J!j T.,,, ')J ::, " "' 'Y' P,11.. r.i....._,i,....8 :2.6 (.) :;:::; a.4 a Qi... (.) V. " " o Between SOP runs + SOP vs. Siometer Frequeny (yles/foot) FIGURE 9 Correlation oeffiients between roughness power spetral densities aross the frequeny domain. 1.4 I CJ) ::?: a: Cl _J.4.2 \ o Between SOP runs +SOP vs. Siometer \ K «I \ 1 \ \ I laoo'\_\, h:i,_,yt Y\. "' 19"' ::. _ "' _, J IQ!. ' "' l!i 'I'...,. I "' Frequeny (yles/foot) FIGURE 1 Rootmeansquare deviations between amplitudes of profile spetra aross the frequeny domain. However, while this may be true, the statistis presented only provide an overall measure of the agreement between SDP and Siometer profiles. It is also important to evaluate the agreement between profiles frequeny by frequeny. Consequently, the orrelation oeffiients and RMSD values were also ompared frequeny by frequeny. Figure 9 shows the orrelation oeffiients aross the frequeny domain, between PSD values from repeat SDP runs, and between PSD values from orresponding Siometer and SDP runs. Figure 1 shows the RMSD values. It is generally observed that the Simeter po\ver spetra ompare favorably with the SDP power spetra. However, at a frequeny of.125 yles/ft (about 3.7 Hz at 2 mph), the agreement is not as good ompared with the other frequenies. At.125 yles/ ft, the orrelation oeffiient between Siometer and profilometer PSD values drops to about.65 as shown by Figure 9. This result suggests that a fundamental response frequeny of the vehile has not been removed and that a need exists for finetuning the proedure to parameterize the statistial model of the vehile so that better agreement between the power spetra of Siometer and SDP profile elevations may be ahieved vvithin the entire frequeny range.

10 Fernando et al. 121 EVALUATION OF LOAD PROFILES PREDICTED FROM SIOMETER ROAD PROFILES Pavement surfae roughness affets the vehile dynami loadings that are imparted to the pavement. Consequently, it is also appropriate to ompare the load profiles assoiated with SDP and Siometer pavement profiles. After all, the dynami loadings produed will affet pavement servie life, and it is of value to know how the predited dynami load profiles differ from eah other. This would provide another basis for judging the aeptability of the Siometer as a devie for profile measurements. A vehile simulation program developed at Texas A&M University was used to predit the dynami loadings produed by a given vehile running over the measured SDP and Siometer profiles. The vehile modeled was a tratorsemitrailer (3S2) ombination with a 12,lb steering axle load and a 34,lb tandem axle on eah of the drive and trailer axles. The measured profiles for Setions 1 and 7 were used in the analysis. Two different vehile speeds, 45 and 27 mph, were used in the simulation. Figure 11 shows axle loads predited using profiles from repeat SDP runs on Setions 1 and 7. In the simulation, dynami axle loads were evaluated at.5ft intervals along a given setion for all five axles of the tratorsemitrailer ombination. The overall orrelation oeffiient between axle loads assoiated with profiles from repeat SDP measurements was.99. Similarly, dynami axle loads predited using Siometer profiles were ompared with those predited using orresponding SDP profiles. Figure 12 shows the axle loads evaluated using profiles from the two devies. The overall orrelation oef 6 5 C\i :J 4 CL (f) (/). :S2 3 'O rn _J u.e 2 rn > 1 1 N = 41,8 OBS r = :. ' Dynami Load (kips, SDP run 1) FIGURE 11 Comparison of dynami axle loads predited using repliate SDP proflle measurements Gi Qi 5 4 i:i5 ui. 32 "CJ 3 rn.3 u.e rn N = 82, 16 OBS r = Dynami Load (kips, SOP). FIGURE 12 Comparison of dynami axle loads predited using Siometer proflles with dynami axle loads predited using SDP profiles. fiient between dynami axle loads was.952. This value ompared favorably with the overall orrelation oeffiient of.99 between axle loads assoiated with profiles from repeat SDP measurements. In addition, the rootmeansquare deviations between dynami axle loads predited from SDP and Siometer profile elevations was 986 lb for 82,16 observations. This statisti was determined using Equation 1 with Y; being the dynami axle load predited using SDP profile elevations and Y;' the dynami axle load assoiated with Siometer profiles. A similar statisti alulated between dynami axle loads predited using repliate SDP profiles was 446 lb for 41,8 observations. On the average, therefore, the dynami axle loads assoiated with Siometer profiles differed from the orresponding axle loads assoiated with SDP profiles by 986 lb. This value is 8.2 perent of the nominal stati axle load of 12, lb on the steering axle of the vehile used in the simulation, and approximately 5.8 perent of the nominal stati axle load of 17, lb on eah axle of the drive and trailer tandems. The results therefore indiate reasonable agreement between SDPbased and Siometerbased dynami axle loads. The power spetra of the predited dynami loads were also evaluated to hek the degree of similarity in the frequeny ontent of the SDP and Siometer load profiles. Figures 13 and 14 show the load power spetral densities assoiated with the two devies for profile measurements made on Setion 1. The load power spetral densities were determined using the predited dynami axle loads for the lead axles of the drive and trailer tandems, at a simulation speed of 45 mph. As seen from the figures, there is good agreement between the load PSD values assoiated with SDP and Siometer profiles. 5 6

11 122 TRANSPORTATION RESEARCH RECORD LI "O >. 9 +' "iii 111!'1 le6j,, Iii... I u 7 a. en \... ;:: 5.,. 't' "O a o SOP + Siometer ""'Ff ijl!jl _J ,,.._, Frequeny (yles/foot) FIGURE 13 Power spetra of dynami axle loads assoiated with profiles measured with the SDP and Siometer on Setion l (Run 1), for the leading axle of the trator drive tandem assembly. "\1 LI "O I Z' 9 "iii \ u 7 a. (/) Qj '" "'t'l:j j k DSDP + Siometer ;:: * 5 "O., a _J Frequeny (yles/foot) FIGURE 14 Power spetra of dynami axle loads assoiated with profiles measured with the SDP and Siometer on Setion 1 (Run l), for the leading axle of the trailer tandem assembly. In order to evaluate the agreement between SDP and Siometer load power spetral densities, the overall orrelation oeffiient between PSD values was determined. Figure 15 shows the load power spetral densities assoiated with SDP and Siometer profiles. An overall orrelation oeffiient of.982 was determined, as indiated in the figure. This value ompares favorably with the overall orrelation oeffiient of.997 between power spetral densities assoiated with repeat SDP profile measurements. In addition, the rootmeansquare deviation between the amplitudes of SDP and Siometer load spetra was determined to be lb for 2,6 observations. A similar statisti between the amplitudes of load spetra assoiated with repeat SDP profile measurements was found to equal lb with 1,3 observations. Consequently, the amplitudes of the waveforms assoiated with the Siometer and SDP load power spetra differ on the average by about 3 lb. Similarly, the amplitudes of the waveforms assoiated with load puwer spedra frum

12 Fernando et al : 8 "O N 26 OBS r.982 Cl) "O 6 tu...j ii> 4 E U Profilometer Load PSD (dbl FIGURE 15 Comparison of load power spetral densities assoiated with Siometer and SDP profiles. to.8 \t.li /\. r'\,i.... v \ I 'I" :2.6 () tu Qi....4 () Frequeny (yles/foot) o BETWEEN SDP RUNS t SDP VS. SIOMETER FIGURE 16 Correlation oeffiients between load power spetral densities aross the frequeny domain. repliate SDP profiles differ on the average by about 13 lb. These values suggest that the Siometerbased load power spetra mathes fairly with the orresponding SDPbased load spetra. The similarity in the load spetra assoiated with SDP and Siometer profile measurements was also evaluated frequeny by frequeny. Figure 16 shows the orrelation oeffiients between repliate load PSD values assoiated with repeat SDP runs and the orrelation oeffiients between load PSD values assoiated with Siometer and SDP roughness measurements. The trends observed are similar to those shown in Figure 9 of the orrelation oeffiients between power spetral densities of SDP and Siometer profile elevations aross the frequeny domain. The orrelation oeffiients aross the frequeny domain between SDP and Siometer load power spetral densities are generally aeptable. However, at a frequeny of.125 yles/ft, the orrelation oeffiient dereases to slightly less than.5. This derease oinides with the derease at this same frequeny in the orrelation oeffiient between PSD values of SDP and Siometer profile elevations (see Figure 9). This result again points to a need for refining the vehile modeing proedure on whih the Siometer is based.

13 124 CONCLUSIONS From the results of the evaluation onduted, the following findings are noted: 1. From an examination of the pavement profiles obtained from the same run, there is lose agreement between SDP and Siometer profiles. This finding suggests that for pratial purposes, the Siometer an show just as well as the SDP an, where the rough spots are on a partiular streth of highway. 2. From a omparison of predited load profiles, Siometer profiles an reasonably be used in onjuntion with a vehile simulation program for identifying those portions of a given highway segment that are likely to be subjeted to severe dynami loadings. 3. From the spetral analysis of SDP and Siometer profile elevations, the Siometer power spetra ompared favorably with the SDP power spetra. However, at a frequeny of.125 yles/ft, the orrelation oeffiient between power spetral densities of SDP and Siometer profile elevations dereased to approximtely.65, indiating a need for fine tuning the vehile modeling proedure on whih the Siometer is based. 4. From the spetral analysis of dynami axle loads assoiated with SDP and Siometer profiles, reasonable agreement between omputed load PSD values was observed. The results also suggest that improving the orrelation between power spetral densities of SDP and Siometer profile elevations at.125 yles/ft will lead to better agreement between SDP and Siometer load PSD values within the entire frequeny spetrum. Overall, the results obtained are promising and show the potential of the Siometer as an eonomial, pratial, and useful devie for olleting profile data on a networkwide sale. Future measurements using the Siometer and the SDP are planned to get more data to further verify the aeptability TRANSPORTATION RESEARCH RECORD 126 of the Siometer as a devie for profile tneasurements. Plans inlude measurements on portland ement pavement setions and reevaluation of the parameterization proedure for modeling the vehile in the measurement proess. ACKNOWLEDGMENTS This paper is based on results of a projet funded by the Texas SDHPT. REFERENCES l. M. W. Sayers, T. D. Gillespie, and C. A. V. Queiroz. The International Road Ro11g/111e.1 s Experi111e11tEstablishi11g orre/11/ion and a Cafibmtio dord for Measurements. World Bonk Tehnial Paper 45, The World Bank, Washington, D.C., E. B. Spangler and W. J. Kelly. GMR Road ProfilometerA Method for Measuring Road Profiles. Researh Publiation GMR 452, Engineering Mehanis Department, General Motors Corporation, Detroit, Mih., De D. W. MKenzie, W. R. Hudson, and C. E. Lee. Tlte Use of Road Profile. taii tis for May. Meter alibra1io11. < prativ Rsenrh Program, Texa rate Dep..rtmnt of JJi ghwy and Publi Transportation, Researh Rpol't 2511, Austin. ex.. eb R. S. Walker. A SelfCalibrating Roughness Measuring Proess. Researh Report 2791, Texn. State Deportment of Highways and Publi Transportation, Aus1i11, Tex., Aug R. S. Wnl.ker and T. P. Luat. The Walkl!r ffot1ghness Devie for Rougf/11ess Mea 11rements. Researh Report 479lF, The Unive r sity of Texas at Arlington, July The ontents of this paper do not neessarily reflet the offiial views or poliies of the Texas SDHPT or FHWA. Publiation oj this paper sponsored by Committee on Swfae PropertiesVehile Interation.