Description and Evaluation of Alaska's Pavement Rating Procedure

Transcription

1 Transportation Researh Reord Desription and Evaluation of Alaska's Pavement Rating Proedure ROBERT L MHATTIE AND BILLY G CONNOR Pavement ondition rating methods used on Alaska's roadways sine 1978 are desribed and examined The methods are intended to provide the speifi performane data neessary to optimie onstrution and maintenane planning and the alloation of available funds Rating elements inh1de simplified measurements of ride roughness, fatigue (alligator) raking, pathing, and rut depth These features are reported individually and are also ombined with traffi data to indiate more general levels of roadway servieability Field evidene shows that a high degree of variability exists in the measurement of raking, pathing, and rutting Coeffiients of variation above 2 perent were estimated for eah type of rating element from experimentally repeated measurements On a given road setion estimates of fatigue raking made by 15 rews differed by up to twie the alulated average Rut depth measurements were typified by alulated standard deviations of about half the mean value Report findings suggest that great are be exerised on future pavement performane inventories Standardiation tehniques are sug gested that should improve manual rating methods Mehanied or eletroni data aquisition tehniques must be developed to eliminate human error The Alaska Department of Transportation and Publi Failities (ADOTPF) initiated use of newly developed pavement rating proedures during its 1978 highway inventory The purpose of this study is to evaluate the statistial validity of individual measurements that omprise it The urrent Alaskan rating attempts to quantify surfae fatigue raking, pathing, and wheelpath rutting as an aid to planning design, onstrution, and maintenane The amount of error assoiated with measurements of pavement distress is examined, and improvements are suggested that an be inorporated into future inventory work rhe researh data base used onsisted of data and experiene aumulated from two omplete inventories of the Alaskan paved highway system onduted during The study also examines results of repetitive sampling onduted speifially for this projet on five typial pavement setions loated near Fairbanks, Alaska DEVELOPING ALASKAN PAVEMENT RATING PHILOSOPHY During the winter of the planning division of ADOTPF deided to revise its existing highway inventory proedure to fill the need for aurate, quantitative data for programming highway maintenane and onstrution funds The department's researh setion was ommissioned to produe a pratial inventory that would stand the srutiny of statistial evaluation As a first step, the literature was researhed to see how other states and foreign transportation agenies had negotiated the same ground A method for rating pavements was first developed for use in the AASHO Road Test of the late 195s to early 196s Pavements are lassified numerially based on the subjetive observations of engineering speialists and normal highway users (!,) The rating sale was arbitrarily set between and 5, where is extremely poor and 5 is perfet The key distress manifestations seleted are surfae deterioration, ride roughness, rutting, raking, and maintenane pathing This rating tehnique produes a number termed present servieability rating (PSR) for lassifying a given setion of road Figure 1 (2) indiates the number of individual raters neessary I Statistially I to estimate the true value Of PSR by using the ompletely subjetive AASHO method This figure indiates that for one or two raters the error assoiated with estimation of PSR is greater than 1 The error an range ±1 from the true value; therefore, the full range of possible estimation is two, whih represents one-third of the total to 5 sale The AASHO researhers then took the next logial step of onverting the rating from a subjetive to an objetive method by deriving a regression equation that losely mathes PSR panel sores Independent variables for the regression equation onsisted of standardied measurements of fatigue raking area, maintenane path area, wheelpath rut depth, and longitudinal surfae variation (roughness) The road surfae ondition values alulated by the regression equation are termed present servieability index (PSI} PSI= log (I + SV) - l38rd 2 - ODl (C + P) (!) where sv mean slope variane in the two wheelpaths as measured absolutely by a longitudinal profilometer (in/mile x 1 6 ), RD mean rut depth ( i n), C + P raking + pathing (ft'/1, ft 2 total surfae) Most pavement rating methods developed subsequent to the AASHO study, inluding Alaska's, are related in some degree to the original AASHO form and were intended to provide key performane feedbak to the overall pavement management proess Generation of Alaska's rating sheme was expedited by a summary and ritique of highway ageny pavement management praties A federally sponsored workshop was held in Tumwater, Washington, in November 1977 to examine the existing state of the art in the field of pavement management systems (PMS) United States and Figure 1 Estimating PSR _12-3"'-----'4,-5 6_, _78-- 9, lo Permissible Error in PSR

2 62 Transportation Researh Reord 938 Canadian representatives were invited providing they were atively implementing and, therefore, experiened in a PMS program At the time ADOTPF was attempting to devise a rating method for asphalt onre'te pavements, the Tumwater onferene r ert was by far ' the most omprehensive soure of information onerning rating shemes available ( 3) The Tumwater report not only disussed vario field methods but also ompared them ritially Rating system elements were suggested that provided the best input to the overall PMS Assuming that PMS would be the ultimately intended use of pavement inventory data, the following onsensus emerged from the Tumwater onferene: 1 Ride roughness should be rtd objetively? Strutural apaity shot1ld he rated, but whether to rate strutural apaity :on the basis of defletion tests or: surfae distress measurements was not learly deided 3 Pavement distress should be rated This inand eludes measurement of rut depth, raking, pathing 4 Rut depth measurements were onsidered along with skid testing to provide an indiation of road safety Rut mea!jp,re!llents should, therefore, be inluded i n any, h g hw ay, rating sheme 5 pe u!-'s, of,a \ g l e lassifiation nt1mber sph as PSI was said to' provide a valid measure of pavement ondition 6 Little standardiation of terminology and measurement tehnique exists among the available systems of pavement rating when these systems are examined in detail Eah of the preeding points was onsidered seriously before devlopment of the Alaskan atitig system Table 1 (3) indiates the salient features of the road rating- methods used by the us states and Canadian provines represented at the Tumwater: onferene The objetives and basi rating elements listed below were hosen by ADOTPF from bakground researh and a definition of departmental needs They guided the development of Alaska's inventory rating by providing use targets Only the most ommonly reognied pavement ondition indiators were seleted for onsideration as elements in Alaska's rating proedure ADOTPF deided that a pavement ondition (!;a\ng) must 1 Provide information for planning and ordering the priority of rehabilitative design and maintenane of existing pavements, 2 Provide information on he relative ondition of total highway mileage within,various jurisditions for budgetary apportion;;,-t purposes, and 3 Provide design feedbak information Table 1 Pavement monitoring features and evaluation Strutural Ageny Surfae Condition Roughness of Ride Skid Resistane Capaity Rating System Primary Deision Criteria Ariona California Florida Kentuky New York Pennsylvania Texas Utah Washington Ontario Saskathewan Crak survey Pavement ondition survey based on alphanumeri rating Strutural defets of raking, rutting, and pathing Mays ridemeter on annual basis Ride sore not part of pavement distress Mays ridemeter orrelated with CHLOE profilometer Used as feedbak for Roughness index ordesign defiienies related to PSI Use ride-quality meter or GM profilometer Vehile response profiler is heart of system Distress survey based on vehile-mounted amera-visual distress rating Pavement distress based on I I observed parameters Pavement ondition survey every 2 years overing entire network Pavement ondition rating, I to 2 year yle Annual surfae ondition rating Mays ridemeter used to develop servieability Mays ridemeter orrelated with Surfae Dynamis profilometer PCA roadmeter on I-mile inrements PCA roadmeter on all setions Subjetive riding omfort index PCA roadmeter at intervals of l month to l year Mu meter-5 ft at eah mile post Measured periodially ASTM skid trailer Skid trailer Mu meter, 5-mile setions tested every 2 miles ASTM skid trailer for high aident loations-onsidered separately Dynaflet-3 loations/mile Road rater for speifi design evaluation Road rater Dynaflet for ritial loations Dynaflet for prediting remaining life Limited use of Benkelman beam Dynaflet, random sample loations in need of rehabilitation Benkelman beam data used for overlay design Pavement management information system Alphanumeri rating ombines severity and extent of defets Combined ride rating and defet rating Correlation of several fators for design input Pavement servieability system, based on orrelation with known servieability levels Relative design, ratio of allowable I SK axle loads to those predited for next 2 years Present servieability index Combined strutural rating and ride sore Subjetive, pavement ondition rating Condition rating system used to order priority of projets for overlay or saling Compares major maintenane alternatives Defets ompared to repair strategies and osts Adjusted pavement rating evaluated for priority programming Input used to develop overlay design Aimed at identifying budget needs, failed pavements, effetiveness of expenditures All highways must arry their traffi safely and omfortably Overall priority ranking for preventative rehabilitation Tabulate rehabilitation strategies and osts based on pavement ondition Required overlay predition based on expeted performane Preventative maintenane is primary goal 3 Photo logging of entire system

3 Transportation Researh Reord Literature review plus ommon sense pointed to the need for a rating method that would haraterie the road ondition adequately and allow a high degree of reproduibility at a minimum ost The data must provide true reproduible harateriation of pavement ondition that hanges from year to year in a rational manner (ie, pavements should not appear to heal anomalously with time unless maintenane has atually been done) The rating tehnique, therefore, had to be as simple as possible and inlude the largest pratial sampling of eah road setion The following were hosen as rating parameters by Alaskan researhers: 1 Fatigue raking (alligator raking), 2 Major pathing (at least full lane width), 3 Wheelpath rut depth, and 4 Ride roughness as measured by the Mays ridemeter Fatigue raking was seleted as a rating parameter beause it is an exellent indiator of strutural ondition and load-life potential Design- 1 ife vehile load apaity is said to be reahed when signifiant alligator raking beomes apparent Fatigue raking is also often assoiated with unaeptable rutting, rough rides for the vehile, and desintegration of the pavement surfae Major pathing needed to repair a host of problems, inluding fatigue raking, embankment settlement, and rutting, gives a general piture of the maintenane effort required on a given road setion Pathing is also a prinipal soure of surfae / roughness and usually beomes raked and potholed with time Wheelpath rutting is generally onsidered important in terms of driver safety and travel osts A onsensus of available literature indiated that rutting deeper than approximately 5 in is a safety haard that an ause hydroplaning on wet road surfaes at high vehile speeds Rutting also has an effet on vehile steering and redues the mehanial life of hassis omponents Deep rutting usually aompanies advaned alligator raking and signifies that pavement strutural soil layers (base or subbase) have been loaded beyond apaity This ondition is aggravated through use of materials subjet to extensi ve moisture-related softening (thaw weakening) Ride roughness is measured beause it is the harater is ti of pavement that is of primary onern to the driving publi The ombination of differential settlement and leveling pathes is ommon to all parts of Alaska and is the major ause of roughness felt by the driving publi Ride roughness is measured objetively on a ontinuous basis by using available tehnology suh as the Mays ridemeter Some reognied surfae distress features were disregarded in order to simplify the rating proess These inlude raveling, longitudinal raks, thermal raking, shoving and bleeding, potholes, and defletion Statewide skid measurements in 1975 indiated that the materials used in Alaskan roadbuilding provided onsistently high skid numbers Reasons for this inlude a high degree of aggregate hardness and limited potential for asphalt bleeding beause of Alaska's relatively ool air temperatures Testing of defletion statewide will ultimately beome part of the normal inventory proess This proess began in 1982 and will require approximately five years per s tatewide yle The falling weight defletometer is urrently being used to ollet inventory data RATING AND SCORING PROCEDURES USED SINCE 1978 A disussion of pavement rating methods used by ADOTPF sine 1978 and an explanation of how field data are manipulated for purposes of soring and reporting follow Figure 2 illustrates the manner in whih raw field data are transformed into a useful pavement inventory report Develo pment o f Field Me t hods The rating proess is done as two separate operations, eah requiring the use of a two-person rew Phase 1 onsists of measurement of ride quality The Mays ridemeter trailer is urrently being used beause of the relatively low ost of gathering data and its reasonably good repeatability The trailermounted meter provides a standardied vehile, suspension, and tire type In phase 2 the surfae distress features are measured These inlude alligator raking, full-lane pathing, and rut depths The Mays ridemeter an provide a ontinuous sampling of highway roughness at 5 mph automatially Studies of the repeatibility of this test have been made by others and are beyond the sope of this report The objetive nature of r idemeter measurements suggests that they be onsidered a relatively reliable element of the urrent pavement inventory Methods for measuring alligator raking and major pathing were evaluated initially on seven setions of roadway near Fairbanks Eah setion was divided into 1-mile subsetions that were rated independently Full-width pathing was harateried on the basis of total length (density) ; fatigue raking was typified by both density and severity A type-1 or type-2 lassifiation was adopted for raking of lesser or greater severity Alligator raking was defined as raking that is visible while driving 7 to 1 mph It is measured as the total perentage of the road setion length that exhibits raking, regardless of wheelpath loation Histograms were onstruted from field data (Figure 3) to show the frequeny distribution of fatigue raking for the subsetions within eah mile The distribution of raking is strongly polymodal (showing no single mean value) and bounded to both the O and 1 perent ourrene level Distributions of fatigue raking are obviously non Gaussion in harater Based on these data Table 2 gives the probability of a random seletion of a 1-mile sample that will predit the true mean ondition of eah setion of roadway The probability is obviously small in all ases In view of these data, a 1-, 2-, 3-mile or more length of paved road ould not be rated aurately for fatigue raking based on measurements in a randomly seleted subsetion several hundred feet long The normal assumption of a 1 to 2 perent sampling density is of no value in this ase Fatigue raking, therefore, muo;t be measured by ontinuous observation through eah mile of roadway All data olleted subsequent to the initial trial have supported this deision Full-width pathing was observed to have a distribution of ourrene similar to that of fatigue raking and it was also deided that this feature ould be properly harateried only by ontinuous observation The frequeny of rut depth measurement was also examined briefly before development of the rating method through multiple readings taken on eah of eight 1-mile-long pavement setions near Fairbanks Rut depth averages ranged from 185 to 244 in The standard deviations of the sample ranged between 16 and 35 perent of the sample means and the plotted frequeny distributions of rut depth measurements appeared reasonably indiative of normal

4 64 Transportation Researh Reord 938 (Gaussian) behavior Based on these trials the assumption was made that rut depth measurement ould be evaluated by normal statistial tehniques Sampling f requeny was addressed through the statistial method used for estimating a true mean value from a small sampling An estimation of true population average is given by (2) where x s N t true population average (ie, true average rut depth), average rut depth as determined f rom sampl e, standard deviation of sample, number of measurements onstituting the sample, and Student's t-value for a given onfidene level and N Figure 2 Elements of Alaskan pavement inventory FIELD DATA ACQUISITION DATA DATA PROCESSING PRESENTATION--->- I Normalied (Ranked) Raw Data Normalied (Ranked) Raw Data Normalied (Ranliled) Raw Data Calulate Surfae Condition Valu e (CV) Calulate Surfae Normelled Condifion Index I--'-= ==:-' (SCI) (Ranked) SCI ; Tratt1 Aident CJata e Traffi Vo lumrtjttotfl Capaity Data Normalied (Ranked) Raw Data 6 Figure 3 Alligator rak frequeny distribution (/)!! :i 4 e a ;:: 2 No t 3 No 2 (/) 5 ::i; 3 No 3 No 4 o ollj1ii-1+1+-i _ g u G (/) 5 :i % Craking No 5 u (/) ::i; e a ;:: % Craking No6 u (/) "O % Craking % Craking EXAMPLES: No7 - -1%: 4 Subsetions % Craking % Craking Note: Figure shows variation in amount of n1king in subsetion measurements within seven setions Exept for setion 1, all were 1 mile long and inluded 1-1 /1-mile subsetions, 1 was 17 miles long and inluded 17-1 /1-mile subsetions,

5 Transportation Researh Reord This equation is an expression of the entral limit theorem, whih desribes the distribution of sample means about a true mean In modified form the equation an be expressed as follows: (µo - X)/S = T y'n (3) The error in estimating true rut depth average (ie, u - X) is small in relation to the sample standard deviation (at a given level of onfidene) when the term T/IN is minimied Figure 4 is a plot of N versus T/IN used to selet sampling frequeny for the initial inventory runs in 1978 Flattening of the urve beginning between N = 4 and N = 7 suggested that a sampling of at least four loations would be neessary to ensure that the error in estimating true mean rut depth would be less than 2 standard deviations of the sample Figure 4 indiates that the error of estimating true mean rut depth is about 16 x S for N = 4 Beause S of the trial road setions averaged approximately 5 in, errors in estimating rut depth during inventory work would be expeted to be no larger than ±16 x 5 (ie, 8 in) This auray was onsidered good enough for beginning the pavement inventory proess Fewer than 4 readings per mile were required in the 1978 rating method if rutting was generally observed to be less than 25 in Summary of Required Measurement Frequenies On the basis of limited field trials it was deided that pavement distress, exept for rut depth measurement, should be harateried by ontinuous ob- servation of the entire road Three field seasons of field data olletion have reinfored the idea of using 1 perent sampling The frequeny of measurements neessary to determine average rut depth adequately was alulated from a preliminary statistial assessment Measurement of ruts was known to be a disproportionately time-onsuming job when ompared with other distress observations The hope was that the experiene in aumulation would show that no more than four sets of readings would be required per mile of road Road Condition Soring Using Alaska's Pavement Rating Alaska uses its pavement inventory data to onstrut a mile-by-mile summary report listing individual ondition sores (perentage of raking, rut depth, perentage pathing, and ride roughness) and also a ombined ondition value (CV) sore The CV is analogous to the AASHO PSI and provides a single numerial desriptor of a given road setion The CV is alulated from the inventory data in the following way: CV= [Mays ridemeter sore (ranked)+ Surfae ondition index franked)] + 2 (4) Ranked data indiate that the data have been transformed mathematially into a perent-worse-than sore before alulation of CV This provides a normaliing of raw sores on a to 1 (worst to best) sale Perentage worse than = [(I /2E + L)/N] x 1 (5) where Table 2 Proability of sampling true mean performane Probability of Seleting I -Mile Perentage of Range Aeptable with Car- Total Area ret Perentage Craked From To of Crak I E L N number of statewide setions rated the same, number of &tatewide setions rated worse, and total number of statewide setions rated The Mays ridemeter sore shown in Equation 4 is derived diretly through the perentile ranking equation from raw Mays ridemeter data Surfae ondition index (SCI) is alulated by means of Equation 6 and then transformed to a perentage-worsethan ranking through the ranking formula SCI = 138R 2 + OOl(A+P) (6) Figure 4 Plot of relative error fator It/Vil) versus number in sample 1 8 : :;::< : ' ::::W:<;" ' '" : : - : ::: '< : ASSUMING; t/"vn@ the 95 ile level of onfidene The 4bovl! assumption is derived from the Central Llnit Theor-em as eipressed by J1 " X ± ts/\1""11"' ' t s Atual average rut depth tudent's "t" t-ible value depending on sample sie and onfidene level Standarrl deviation of sample Numl>er of measurements per umpl e 4 2 Curv-Qgins to flatten within this interval of 4 to 6 measurements per setion Therefore, the range of error in estimatin9 the atual mean rut depth be!jins to f'lininde --'"-& L Number of Measurements per Sample (n)

6 66 Transportation Researh Reord 938 where R A p average rut depth {in), perentage of road setion that is alligator raked, and perentaqe of road setion that is overed by full-width pathing Analysis of Field Ota An indiation of measurement variabilities between rews is given through the oeffiient of variation (Cv) assoiated with eah distress type: Cy =(SD/Mean value) x 1 (8) In addition to reporting a summary of the previously disussed information, the pavement inventory report also inludes, for multiple mile setions, the ranked sores of a volume/apaity ratio and the setion's aident rating value Finally, the ondition value plus apaity and aident sores are ombined in the form of a geometri mean to produe a omposite value alulated as Composite value sore= [Condition value x Capaity (ranked sore) x Aidents (ranked sore)] 11 3 (7) This omposite sore, like CV, is used mostly for generalied administrative planning and programming purposes By ombining the three parameters in this manner the lowest value may affet the alulation signifiantly Thus, if any of the three values has a very low sore, it aused that mile to be flagged Figure 5 is a sample page from the 1979 inventory summary REVIEW OF ALASKA'S PAVEMENT RATING METHODS BASED ON RECENT FIELD STUDIES After the Alaskan pavement rating method had been in use for 2 years a more detailed evaluation of its onstituent measurements was thought neessary A field study was begun in 198 to investigate the repeatability of raking and pathing measurements made by different rating rews Frequeny of measurements neessary to estimate true mean rut depth was also reviewed Method of Study ;ind D;it;i Aquisition Five roadway setions were seleted near Fairbanks that reflet the average range of road surfae onditions ommonly enountered Eah of the setions was rated by 15 different two-member rews using the urrent standard Alaskan proedure Members were drawn mostly from the middle-level professional and tehnial ranks of road design, maintenane, rightof-way, and materials setions, but only four raters had previous pavement rating experiene Raters with previous experiene were drawn from the department's researh and development setion Eah rew of raters was given the same introdution to pavement rating and direted from one pavement setion to another by the instrutor Ratings by eah rew required a full day and the sequene of pavement setions remained onstant throughout the duration of the study Considered important was that the sequene of setions not hange beause this assured that the sun angle relative to the viewer remained onsistent for eah rew for eah setion Sun illumination was known through aumulated field experiene to greatly affet pavement rak visibility To max1m1e the observational abilities of eah rating rew all ratings were performed from a light truk or van A nearly vertial windshield ombined with a relatively high seating position allowed the most advantageous view of pavement surfae of any standard type of vehile Eah setion was inspeted at under 1 mph in order to identify and measure raking Rut depths were measured in eah of the four wheelpaths every 2 mile Distanes were measured with an eletroni odometer apable of 1-ft resolution where SD is the standard deviation In general, a small Cv of approximately 5 to 1 perent indiates that a good estimate of a true mean value is possible from relatively few individual measurements Cv values assoiated with measurement of all pavement distress indiators were onsidered very high This tends to ontradit the initial hypothesis that, beause of the simpliity of the rating method, reproduibility of ratings among rews ould be taken for granted The following estimates of Cv were alulated from projet data Item Type l alligatoring Rut depth, alulated average Rut depth, alulated standard deviation Avg Cv (%) The signifiane of the foregoing listing should not be understated as the uniformity of Cv from setion to setion indiated Type 2 (severe) alligator raking and full-width pathing are not listed beause their infrequent ourrene within the test setions did not provide an adequate sampling to allow a good evaluation of differenes among rating rews Based on these limited observations, however, the variability in measuring pathing length is somewhat lower than for alligator raking that has Cv of perhaps 1 to 2 perent A lear distintion between type l and type 2 illig;itor riking was not easily made by the rating rews The tendeny, exept in the most obviously severe ases, was to plae all raking into the type l ategory Most rews apparently seleted a lower severity lassifiation whenever the question of degree of damage arose This problem an probably be remedied to some extent during the instrution proess by speifially advising that pavements be rated ritially The large amount of variability observed in the olleted data is given in Table 3 In view of the similarity in training and bakground among these experimental raters and previous inventory rews, these variabilities ould be expeted on pavement setions throughout the state Table 3 summaries the variation in data for the setions tested The variation in all the pavement distress measurements is large when onsidering the range i n raked length When onsidering the variation as a perentage of setion length, the variation is less The maximum variation between the mean and maximum values is 7 perent It an be argued that the alligator raking expressed as a perentage of setion length need only be determined to be within 1 perent of the true perentage for inventory purposes If it is assumed that the mean is the true value, then all five setions meet this riterion More detailed measurements may be neessary for design proesses The overall effet of variations in rew measurements on determinations of rut depth is magnified beause ADOTPF usually reports maximum rut depth in terms of average plus two standard deviations For example, the mean, mean + l standard deviation, and mean + 2 standard deviations are given for the fol

7 Figure 5 Sample pavement inventory LOCATION CONDITION ELEMENTS TEnMINI SECTION os RIDE CRACKING PATCHING AOT LENGTH MILE (in/mi) ( t/mi) (/mi) FAP 35 Parks Highway 312 (State Poute 17) JCT Old Nenana Hwy Ester JrT SECTION AVER1GES I I RUTTING CONDITION (in/1) VALUE SJ{) PERFORMANCi VALUES S"ERVICE ACCIDENT I COMPOS!TE VALUE VALUE VALUE , >i m '8, DI g : m ell, DI () ::>" : (), a ID w CD JCT FAS 649 GEIST ROAD g ! SECTION AVERAGES !28 I 2 I 6 II JCT FAU AIRPORT SPUR I I FAP 35 PARKS HIGHWAY (STATE ROUTE 17) SECTION AVERAGES l JCT STEESE AttD RICHARDSON HWYS l SECTION AVERAGES l "'

8 68 Transportation Researh Reord 938 Table 3 Observed variation in pavement distress measurements Pavement No Range Type 1, alligator raking ft ,434 ft ft ft ft Type 2, alligator raking 1 None deteted ft 3-13 ft 4 None deteted 5 None deteted Full-width pathing ft ,42 ft ft 4 None deteted None deteted Rut depth, average inner wheelpath in in in in in Rut depth, SD inner wheelpath in in in in in Rut depth, average outer wheelpath in in in in in Rut depth, SD outer wheelpath in in in in in Range as Perentage Avg as Perentage Average of Length of Length 14 ft ft ft ft ft ft ft ft ft ft 36 ft ft ft 7-4 ft ft 4 in 21 in 18 in 17 in 21 in 2 in 16 in 8 in 8 in 11 in 9 in 24 in 15 in 18 in 27 in 55 in 24 in 9 in 8 in, 16 in lowing setions by using Table 3 diltil outer wheel path Se- Mean Mean + SD Mean+ 2 SD tion Ji!!:l (in) (in) The foregoing examples demonstrate a wide range of unertainty as to the measured depth of rutting even though alulated mean values are low Disussion of Measurements of Alligator Craking In several of the following figures the variations in measurements have been normalied This normaliation step is used so that various road setions an be ompared diretly even though eah has a different mean rut depth or length of alligator raking Normaliation of soring, (eg, perent of alligator raking and average rut depth) is aomplished as follows: Normalied perentage of alligator raking= (A-B)/C (9) where A = perentage of alligator raking as measured by an individual rew on a speifi road setion, B average perentage of alligator raking alulated from the measurements of all rews on the above setion, and C standard deviaton value alulated from the maourmnto of illl rews on the ilbove setion Figure 6 shows how normalied sores of individual rews rank in relation to alulated average values on all five pavement setions This plot indiates the ability of ertain rews (eg, 7 and 8) to see more damage than others Conversely, rew 14 saw muh less raking in all five pavement setions than the alulated average Figure 6 inludes the instrutor's subjetive assessment of eah rew in terms of (a) ommuniation between rew members [rated low (L), moderate {M), and high (H)] and (b) initial impression of rating ability (rated fair, good, and expert) Note that rews 2 and 1, rated expert by the instrutor, had at least a full season's rating experiene and were inluded for purposes of omparison with the other rews Although Figure 6 indiates that some rews ould apparently see more pavement damage than others, this differene was not aounted for in obvious attitudes or abilities Note, however, that rew 8, whih saw muh more pavement damage than rew 14, also rated higher in the instrutor's opinion Best results are obtained when onversation onerning the rating proess is enouraged between rew members, espeially during the first few days of inventory The data in Table 4 attempt to delineate reasons for differenes among rew ratings The samples have been broken down into a stratified format and ross indexed in terms of rew ommuniations and weather and pavement surfae ondition at the time of rating The numbers given in Table 4 as X (harateristi sample average) and SD (harateristi sample standard deviation) have been normalied, as

9 Transportation Researh Reord Figure 6 Range of variation in eah rew's measurement of type 1 alligator raking w er Level ot Dleuaelon Conerning 2 M M Rating Methoda Between Crew () H H en I Member a w ::::> M H J M H <( +1 L > M M <( Good w Fair er u en I 1 Good Fair, Good i== -1 <( Good > Fair Fair w Fair / L I Good lnatrutor1 Opklk>n of Ctow a: <( -2 Good Fair HOTE: Aotual Ratlfll I lndmdut aat&on1 lndlo t d llr -( ) <( I- en CREW NUMBER Fair Table 4 Analysis of type 1 alligator raking Cloudy, Rain, Very Disussion of Slightly Wet Wet Road Weigl1ted Avg Rating Methods Statisti Sunny Cloudy Road Surfae Surfae of Rows Ative x SD N Moderate x 3-3 SD 8 II 6 N Little x 1-7 SD 6 8 No samples N 4 1 Weighted avg of olumns x 4 SD I No samples previously desribed, thus allowing all five pavement setions to be onsidered in the same analysis The ombinati on of a slightly wet (SW) pavement surfae and a highly ommuniative rew resulted in more visible raking and a harateristi average of +8 SD above the overall sample average Also, in examining the weighted (for sample number) averages of both rows and olumns, good rew ommuniation and a slightly wet road surfae are individually assoiated with inreased damage observation Surfae Wetness The effet of a slightly wet surfae in optimiing the visibility of alligator raking is fairly obvious to even the asual observer and an often ause hairline alligator raking to stand out in vivid detail On the other hand a very wet road surfae, suh as obtained during or shortly after a rainstorm, amouflages all but severe raking Observations of raking should be disontinued during rainstorms or other periods when the pavement surfe is overed by free water Table 4 generally assoiates the least observed raking with a very wet (VW) surfae ondition The ideal, slightly wet surfae ondition is reated when the road surfae is dry exept in and around individual raks In this ase, water stored in the raks during rain- fall will keep the adjaent pavement wet longer than in areas of no raking The previous disussion leads to the onlusion that pavement ratings ould be done best shortly after a rainstorm; however, a dry road ondition represents the more normally enountered situation Beause of the need for a standard rating proedure rak measurements should be made only on dry pavement Sun Angle Illumination effets due to variations in vertial and horiontal sun angle are known to affet rak visibility strongly Experiene indiates that optimal lighting onditions are provided by a more or less head-on sun inidene Frontal light tends to shade and, therefore, darken the visible side of rak segments that are perpendiular to the observer and most easily viewed This has the net effet of maximiing apparent tone and texture differenes between raked and unraked pavement The travel diretion hosen for the experimental ratings produed over-the-shoulder lighting on four of the five pavement setions, whih is usually onsidered a worst-ase viewing ondition Eah test setion was examined at approximately the same time of day by eah rew to ensure a onsistent sun angle

10 7 Transportation Researh Reord 938 Figure 7 Range of variation in measurement of full-width pathing en : - " (/) > Q) : Cl <1l Q) "C ::? <1l E "C,g : II (/) : (ij ;: en <1l : > :::J Crew Number B Disussion of Full-Width Pathing Measurementa The ourrene of full-width pathing within the test setions was somewhat limited Data from setion 3 indiate that differenes in pathing measurements among rews may be about half those expeted from observations of raking, The distribution of normalied sores indiated in Figure 7 represents only the three test setions that atually ontained pathing The variation among rews is markedly less pronouned than for alligator raking Pathing appears to be more easily measured than alligator raking even though both are evaluated in a similar way In most ases pathing, at least new pathing, is atually seen quite easily Observation onditions that provide the best view of alligator raking also tend to make pathed areas stand out Again, very wet surfaed roads resulted in the most variable measurements among rews, and raks are most easily seen on a slightly wet pavement Regardless of the better viewing ondition afforded a slightly wet surfae, the dry road ondition is most ommonly enountered in field work and is, therefore, suggested as the standard for inventory purposes Disussion of Rut Depth Measurement The approah taken initially to determine a sample number (as indiated in Figure 4) was a rough attempt to limit the possibility of gross errors Suffiient field data have sine been olleted to allow a more valid estimation of rut depth The problem of rut depth measurement an be addressed by normal statistial methods The prinipal questions asked are 1 How frequently must rut depth measurements be taken? and 2 Must rut depth measurements be taken in both inner and outer wheelpaths? Sampling Frequeny The frequeny of sampling must be high enough to ensure (to some speified onfidene level) that a alulated mean rut depth is reasonably lose to the atual mean rut depth Atual or population average in this ase is that value that would be measured from an infinitely large sampling Sampling tables available in referenes suh as the Chemial Rubber Company statistial handbook (4) indiate minimum sample numbers neessary to attain speifi levels of onfidene against either a type-1 or type-2 error being ommitted A type-1 error ours if statistial alulations indiate that the sample mean is not representative of the population mean, when in fat it is Conversely, a type-2 error ours when statistis indiate that the sample mean is representative of a population mean when it is not It was assumed that, for prediting the atual rut depth average from sample data, an error of no more than ±5 in would be allowable In most sampling situations, little onern is expressed over type-2 errors This philosophy leads to 5 perent level of type-2 error ontrol (ie, no ontrol and a signifiantly redued sample sie) Determination of sample sie is dependent on expeted standard deviation: therefore, it is important to onsider the magnitude of values that might ommonly be enountered Rut measurements made on the five Fairbanks test setions indiated a range of standard deviations from about 2 to more than 35 in assoiated with average rut depths between 2 and 4 in Irifiations of variability in rut measurement derived from the Fairbanks test setion data suggest that minimum sampling be based on a standard deviation perhaps as high as 3 to 35 in This magnitude of deviation plus 9 to 95 perent onfidene level against error results in a minimum sample sie in exess of 1 Rut depth measurement, therefore, begins to appear impossible exept through an automati rut-measuring devie apable of high-density sampling Alternatives Several soures of rut measurement data were used to onstrut funtional relationships among average rut depth, alulated standard deviation, and required number of sampling points This report substantiates previous ontentions <2l that true mean rut depth an be aurately harateried only through a

11 Transportation Researh Reord Figure 8 Comparison of 1978 with 198/81 pavement inventory data 1 CRACKING 1 PATCHING o 5 o "' - 5 / I ' :, ' : I : - :%/< ' ' ' RIDE SCORE CONDITION VALUE o 6 o "' Note: All numbers have been normalied to a -1 (worst-bes ) soring system large sampling Problem rut depths on the order of o4 to 5 in or larger would require an assumed standard deviation of at least 3 in Reasonable error onfidene levels indiate a sampling obviously greater than 1/setion Furthermore, the inability to predit whether inner or outer wheelpath represents the worst-ase ondition would require doubling of the sampling effort In dealing with this question the hoies are 1 Assume rutting to not be a problem and ease measurement, 2 Perform a few random measurements per mile at loations that appear from general observation to represent worst-ase onditions, or 3 Purhase or build an automati devie for rut measurement as desribed by Jurik () Deeply rutted setions are usually assoiated with severe alligator raking on most Alaskan road setions: therefore the measurement of both is unneessary Rutting within the state is rarely as deep as 5 in, whih is onsidered ritial in most literature soures Alternative l appears to be a reasonable ourse of ation at present Alternative 2 provides numbers and the numbers an, of ourse, be inluded in subsequent disussions of pavement ondition The numbers generated from alternative 2 have no basi statistial validity, however, and might be thought of as inventory garbage Alternative 3 is preferred if departmental poliy requires an aurate determination of rut depth A 1981 ost estimate for the purhase of an automati rut-measuring devie was $15, to $2, COMPARATIVE LOOK AT PREVIOUS INVENTORY DATA This setion looks at atual pavement inventory data in view of the findings of this paper Beause of the rather gross variability evident in the experimental measurement of raking and pathing, a diret mile-by-mile omparison between two previous inventories is made Figure 8 shows the apparent variation in pavement distress between 1978 and 198/81 As shown, these data have been normalied to provide a total soring range of to 1 (worst-best) Note that data at oordinates ( perent, perent) and (1 perent, 1 perent) are often lustered in the graphs for raking and pathing, whih aounts for the appearane of fewer than expeted individual points on these plots The same number of data points was reorded for all of the graphs A line of x = y has been inluded in eah plot and differentiates pavement setions that apparently or atually improved with time (points above the line) from those that beame worse (points below the line) Examination of plotted data indiates 1 A very high degree of overall satter and 2 An unusually large number of data points above the line of x y (ie, performane improvement with time) Taken together, these findings demonstrate a marked degree of randomness inherent in the rating proess The impliation of point number two is espeially signifiant in view of the onunon-sense assumption that pavement ondition deteriorates with time This assumed generality ould, of ourse, be altered by reonstrution, overlay, or areful pathing, and no attempt was made to remove speifi points that represent reonditioned pavement setions from the plots This should, however, aount for only a small perentage of total rated mileage A signifiant degree of randomness is suggested beause even setions that sored better than average in 1978 showed a very high rate of apparent improvement with time The likelihood that initially good

12 72 Transportation Researh Reord 938 pavements (soring 5 to 1) will be improved substantially within a period of 3 years through maintenane is slight SUMMARY AND CONCLUSIONS The development and evaluation of Alaska's inventory rating proedure for flexible pavements has been desribed Development of the system was based on the generally aepted priniples of pavement rating pratie as outlined in reent literature The Alaskan rating method attempts to measure basi elements of road quality from two important viewpoints: 1 The highway user--ride roughness and 2 The highway engineer--fatigue (alligator) raking, major (full lane width) pathing, and wheelpath rutting These rating features are reported on a mile- bymile summary both individually and in terms of a omposite servieability sore A onerted effort was made during the development of the rating method to keep all distress measurements as simple as possible but still provide adequate information for pavement management needs The rating method was evaluated through a speial field study and experiene aumulated during the 3 years sine its implementation Findings indiate a large variation in the abilities of different rating rews to haraterie the extent and severity of pathing and raking The range of variation in rak and path measurements obtained by 15 rews on 5 seleted pavement setions was found to be as muh as twie the mean measured value These differenes are apparently assoiated with the level ot interest in the task expressed by eah rew and weather fators that ontrol visibility of pavement surfae features Examination of previous inventory ratings onfirmed the data i;atter indiated by the experimental pavement setions The variation in rut depth measurements was large enough to require very high sampling frequenies A mehanied form of rut-measuring devie, apable of more than 1 measurements per setion in both inner and outer wheelpaths, is suggested Marked differenes between average depths of inner and outer wheelpaths require data from both loations in order to define the worst-ase ondition Conlusions The assumption that Alaska's pavement rating methods are simple enough to ensure a high degree of reproduibility is not demonstrated by the available data A great deal of variation is apparent in the field measurement of raking, pathing, and rutting This is indiated through examination of experimental data as well as from data olleted from previous inventory work The use of mahine measurements is suggested wherever possible in all phases of the rating proess The visual rating of pavements is a diffiult proess that requires areful and rigorous standardied tehnique Pavement rating instrutions must be formalied to inlude guidelines for training rating rews and ensuring aeptable pe-rformane Speifiations are neessary for standardiation of viewing height, aeptable 1ighting onditions, and vehile speed Reommendations The ability to quantify pavement performane is a requirement of almost any approah to pavement man- agement Alaska's pavement rating method should therefore be viewed as a tool to be improved rather than disarded Reommendations for improvement inlude 1 Phase out visual measurements of pavement distress as reliable mahine methods beome available: 2 Exept for very rough lassifiation purposes, disontinue rut measurements until sampling rates of more than 1/mile an be ahieved: and 3 Continue existing approah but with greatly inreased and improved rew training and a strit standardiation of observation tehnique An ideal form of instrution would inlude the use of standard road setions On these setions the rew would attempt to math the ratings assigned by experiened personnel A five-day tuning period is suggested for new rating rews Ratings performed during this first week would not be inluded in the inventory summary before verifiation by repeated observation Observation onditions for the inventory measurement of raking and pathing should be standardied: 1 Vehile speed of 6 mph or less, 2 Rating of only ompletely dry road surfaes, 3 Use of optimal sun inidene whenever possible for best illumination--a horiontal sun angle of ±7 degrees from head-on or a vertial sun angle of more than 1 and less than 6 degrees from the horiontal (this point should be emphasied even if it requires that the diretion of travel, ie, di- retion of the rater's view, be hanged), 4 Standardied viewing height of 5 5 ft ± 5 ft, and 5 use of utility van-type vehile that has a nearly vertial windshield ACKNOWLEDGMENT On behalf of the Alaska Department of Transportation, we wish to aknowledge tehnial assistane and funding support provided by FHWA ADOTPF Interior Region Division of Planning and Programming, Design and Constrution, and Maintenane and Operations are offered speial thanks for providing temporary use of their employees as rating rew members REFERENCES 1 The AASHO Road Test: Report 5, Pavement Researh HRB, Speial Rept 61E, 1962, 352 pp 2 EJ Yoder and MW Witak Priniples of Pavement Design, 2nd ed Wiley, New York, RL Terrel and RV Leler: Washington Department of Transportation Pavement Management Workshop: Tumwater, Washington FHWA, FHWA-TS , Sept The Chemial Rubber Company: Handbook of Tables for Probability and Statistis, 2nd ed (WH Beyer ed) CRC Press In, Vol 1, Cleveland, Ohio, R Jurik Automated Pavement Rut Depth Measuring System Alaska Department of Transportation and Publi Failities, Fairbanks, Rept AK-RD- 8-1, 1981 Publiation of this paper sponsored by Committee on Monitoring, Evaluation and Data Storage Notie: The ontents of this paper reflet the views of the authors, who are responsible for the fats and the auray of the data presented herein The ontents do not neessarily reflet the offiial views or poliies of the Alaska Department of Transportation and Publi Failities or FHWA This report does not onstitute a standard, speifiation, or regulation