validating the fatigue endurance limit for hot mix asphalt pptx

The concept ofan endurance limits widely recognized in many areas of materialscience, «specially thatof rove metals, The endurance it isusully auld from the relationship bf stain to load

NATIONAL COOPERATIVE HIGHWAY

RESEARCH PROGRAM

REPORT 646

Validating the Fatigue

Endurance Limit for

Hot Mix Asphalt

“ARCH BOARD

TRANSPORTATION RESEARCH BOARD 2010 EXECUTIVE COMMITTEE

ot Miche 8 Mets, Dr of Taition Nok Coal es Cou Gvenmets ington

Ves Cinn Nell Pere, tlm yn sceighay Adin, Ble

Tang Men Ses nce Dern Mii De akon ind C0, Nol

hi of calor os ng nbn eny Kr Pf Depron i Eh ont in, Cn fr Train tac, ney

‘ard Ned Hele Pee, Cnr Cla ily, Nain DC andl ben Dir Caos DOT Sasa

Peek fabn Dias sour DOR io i

‘Sanda Roeloun, rer ing Uy of rina, Taco “rnp Rowe Pe, Carp Pic WalMart, ne, Me LA

Seven Tcl hs Opi fry Marne Rese rupee

Nene: (Gee) Scat esha tel fea Ci IS Le MO

Bevel A Sats Ga Monje Chi sein fr Maplin hea apd Tron Ahr, Aan GA vid Setae, Pini Moran ah LLC Pin, PA

DanlelSpering rer hl Eiri incr, Bary Eon Cov, ir of Cara Dar on Eres

(lade! Wan: Bn Cot Cn Chain Ean Uni of Tes, Ain

Ex oFFicio memaeRs

Peter Appel ie

‘Sort Bull Pscon Eris onl Ue Pfc Pons tao Xo

no eskn of gern Wosingon DC

‘Anne er, Admin Foal ater Car Sty Anat UDOT

Elnard ambryrs Pre onf CED Aor aneroon Ronson DE Iohn Hovey, ce Decor Ameri Lc ofS Hay al Tramporaon fal, Washing, DC

Dav: Maun Dpaty Aish Min eatin USOT

Le AMntrienUADOT

Pee opal nme Fla nat David Sud nso Neon igay Tai Se Aang UDOT kimono SPOT

kecphC Smle,Adeiztmuer Fel aad hms USO

Pals Totenberg, ust are for Tremor Poy UDOT abit, Van Aner Gens sung) ho grand Comma Geel US Amy Corps Đgiser Won

i US Cont an US Depron Hom

NATIONAL COOPERATIVE HIGHWAY RESEARCH PROGRAM

NCHRP REPORT 646

Validating the Fatigue Endurance Li

for Hot Mix Asphalt

Bris D, Powell Ray Brown

Semel H Carpenter Sudip Bhatacharee™

Saced Maghsoodion

oth prose yb AdreiAezkiendl Sun Hugl srl Tngotdanilds in tpn the el gy ia

TRANSPORTATION RESEARCH BOARD

NATIONAL COOPERATIVE HIGHWAY

RESEARCH PROGRAM

Syemats well eined rec provides te mow itv

‘clentng gh of ighray tampon evap cen

roti are bet đulil dưnngh v coonlenal mm of

m1

n of ae Highway an Tamistvio Đi

"hen MB em chư Hôm! ghey rv poem

replying madecnuiotftrảnkgee TWaprugem =ugrodelom

snaunted by the Auction Wo sinter the rv prog

‘pecs in hihaytasportaion mate 1 bring the Sings oF

The props doped on he i of eet onde rid

EraiddsdeiitrtescfMehgdvep nl amporiton erent

Revcah Cumcanl he ord the Ameren Asa of Ste

The nee for highway rsuch ate many andthe National

Ccoopeive Highwny Reh Progen snient

‘oot to the ition af igh rsporaton poles of

THE NATIONAL ACADEMIES

‘Advisers othe Nation on Science, Engineering, and Medine

he Naina Academy of Scences is priate, nonprofit, self pepeunting society of distinguished scholars engage i scetii and enpinerng reach dedicate to the farterace of cence and tehology 2nd t therefor the general well Ôn tác

"suthrity ofthe charter granted ot hy the Congress 186, he Academy hat» mandate that reser to adie the eer {goverment once and technical mater De Ralph I Cieone is resident ofthe National Academy of Sciences,

The National Academy of Eagineringvas eas in 1964, under the carte ofthe National Academy of Secs, pre erpuiation of outstanding einer tc autonomous int ainistration and inthe selection ofits member cing wth he

“National Academy l Sdemex the eponsbity lo aising the federal gonenment The National Academy of Engineering loi sponsors engineering programs aimed at meeting pation need encoutgesedeation and research, and recoaizes the superior achievements o engineers De Chaves M Vesti president ofthe Nation Academy of Eaginrine

The nate of Medicine was established in 1970 by the National Academy of Scene oscar the services of eminent members

of appoprne professions in the examination of plcy mates pertaining othe heath ofthe public The ludiete acts under the responsibilty gen to the National Academy of Sciences by its congresional care tobe an aise to the federal government and, omits ve inate, o identity sues of mesial cae, sere, and education, i Harvey V ineberg i president of the Ina of Medicine

“The National Reseach Counc was organize by the National Academy of Scns in 1916 to aso the boa sommanity of

‘eins an ecology withthe Academy’ porposes of fartering hnowiedge and advising the federal government, Fantonig in acowdance with genta pices determined bythe Academy the Counc has become the rincpal operating agency of ba he Naval Academy af Since’ and the Navowal Academy af Eagacring ia providing serie othe overeat the publ, ahd the scientific and engineering communi The Counc is administered oly by both the Acdenier and the Unite of esc Dr Ralph iserane and Dr Chale M Ves ae chr and vie chi, respectively ofthe Nationa Reseateh Cun

‘The Transportation Research Bards one fs mor divisions of he National Research Coun The mision ofthe Transport tion Research Board isto provid ederahip in transportation inavtion and progres through recach and afrmition exchange,

‘condocted within a setting hati objective iterdsiplinry, and muliodal The Boud’s vied actives anual engage about 2,00 eaginees scents, ae other transportation rsarchets and peactones Kom the public and pate sectors sod academia, lof whom contribute her expertise inthe public interest The program is supported by tat tanporation departments, federal ences including the component administrations ofthe US Department of Transportation, spd other organizations and india SIsntresed ia the development of tasporatin wie TRB org,

wawwenational-academies.org

COOPERATIVE RESEARCH PROGRAMS

(CRP STAFF FOR NCHRP REPORT 646

FOREWORD

By Edward T Harrigan

Stat Ofer

‘Transportation Research Board

"his report presents the findings of research performed to investigate the exitence of a faxgue endurance limit for hot mix asphalt (HMA) mixtures the elect of HMA mixture sharacteritics on the endurance init and the potential forthe limit’ incorporation in Steuctral design methods for exile pavements, The report descibes the research pet= formed and includes proposed standaed practices wsing various experimental and analy tal procedures for determining the endurance limit of HMA mixtures Thus, the eport will bbeofimmedit interest to material nd structural design engincers in stat highway agen-

‘es and engineerin the HMA construction industry

‘Many wellconstrcted flexible pavements witha thick HMA structure have been in service for 10 or more years without any evidence of bottom-up fatigue cracking This ld xpviencesogess that an endurance i, that a evel of strain below which atigue damage does not occur for any number of oad repetitions, sa valid concep for HMA mixtures The concept ofan endurance limits widely recognized in many areas of materialscience,

«specially thatof rove metals, The endurance it isusully auld from the relationship

bf stain to load repetitions to lure, Defining an endurance limit for HMA anstes wl result in more ecient structural design of flexible pavements built with mistuees of Yantn§ propenies For instance, small increases in the binder content of HMA mites provide longer fatigue lives (presumably because of a higher strain level forthe endurance list ofthese mincutes) er factors likely to determine he value ofthe fatigue endurance lit fora given HMA mixture are the incorporation ofa modifier inthe asphalt binder, the aggreyate Iype and gradation, the asphalt binder grade, and the misture'syolumetsc properties Previous research sugested thatthe fatigue behavior of flexible pavements consistent withthe existence of an endurance limit with an approsimate value of 70 microstsins, However, few laboratory studies corroborate this value Moreover, pavement design approaches, including the Mechanistic-Empirial Pavement Design Guide (MEPDG) {developed in NCHRP Projects 1-374 and 1-10, donot ly incosporat the endrance limit

‘concept, Ths is because, to date, rserch in the fatigue of HMA mixtures hasbeen limite to strain eels well above the hypothesized value of 70 microstans

"The objectives ofthis research were (1) ts the hypothesis that there isan endurance lint inthe ftgue behavior of HMA mixtures (2) measurethe value ofthe endurance limit fora representative range of HMA misturesand (3) recommend a procedure to incorporate the effects ofthe endurance limit into mechanistic pavement design methods The research eas performed by the National Center for Asphalt Technology, Auburn University

‘Auburn AL, with the asstance ofthe following oefanizations: Applied Research Associates,

Inc, Round Rock, TX; the Asphalt Institute, Lexington, KY; the Univesity of Hino,

‘Urhana-Champaign, iL and the University of New Hampshire, Durham, NH

“The report fully documents the design and conduct ofa extensive laboratory program

‘of beam fatigue and uniasal tension testing that experimentally confirmed the existence of

fn FIMA fatigue endurance imit and quantified how the value ofthe iit infuenced by HMA mixture and binder properties, Based on these results, «practical definition of the endurance limit was developed, along with a methodology to estimate the endurance limit

in he laboratory Analysis in-service pavements by the eseaech team also demonsteted the existence ofthe endrance limit ai indicated that polymer madfeation of asphalk binders improves the fatigue performance of HMA mistures and flexible pavement, Finally, sensitivity analyses were conducted that indicated that the value ofthe endurance limit can afet the recommended thickness of perpetal pavements designed with the MEPDG and Perkond methodologies signal

‘his report includes six appendices as fllows:

Appendix A: A Proposed Standard Practice for Predicting the Endurance Lins of Hot Mix Asphalt (HMA) for Long Life Pavement Desi

Appendix: A Proposed Standard Practice for Predicting the Endurance Limit of Hot Mix Asphalt (HMA) by Pseudo Stain Approach:

Appendix C: A Proposed Standard Pace fr Extrapolting LoncLife Beam Fatigue Tests Using the Ratio af Disipted Energy Change (RDEC)

Append D:A summary of resus ofthe beam gue esting accomplished durngthe pois Appendix En analytical method for constuction ofa damage characteristic curve through caleltion of pseudo stains and

‘Appendix FA proposed design for a nelaborstory udyt develop a precision statement for AASHTO 121, Standard Method of Fest for Determining he Fatigue Life of Compated TIM Ashnl (HMA) Subjected to Repeated exer Bending

“The proposed standard practices are under considerstion for posible adoption by the AASHTO Highway Sulbcommitce on Materials andthe AASHTO Joint Techaeal Commit: tee on Pavements

7 Chapter 2 State of Practice

7 Arguments forthe Existence ofthe Eodurance List

9 ators lectin Fatigue Life

9 Strteiesto Produce Long-Life Pavernents

12 Laboratory Fatigue Tests and Analysis Methods

13 Laboratory Stdes ro Quantify the Endurance iit

13 Moding Fatigue and Relationship to Fie Performance

15 Chapter 3 Research Plan

15 Introduction

HP Test Methods

20 Chapter 4 Beam Fatigue Test Results and Analyses,

20 _Extraplation Methods to Predict Fatigue Life

39 ristence ofthe Endurance Limit

m imate of Pein of Beam Fatigue Teste

42 Indiret Tele Sirength a Sureogte for Endurance

Limit Determination

44 Chapter 5 Uniaxial Tension Results and Analyses

4 Test Specimens

44 Dynamic Modus an Phase Angle Maser Crves

8 Damage Charsteritic Carve

53 Baluntion of Endrance Lint

59 Chapter 6 Examination of LTPP Database for Indications

of an Endurance Limit

59 Including the Endurance Lit Design Premise ito

‘Mechanistc-Empircal-Based Pavement Design Procedures

62 Defining the Endurance Limit—A Survivability Analyse

8% Updated Serves Anahys Using LTPP Data

ster 7 Sensitivity of Pavement Thickness

to the Endurance Limit Summary of Predicted Endurance Limits

Enmate of Shift Factors between Laboratory Tests ad

Fed Perfomance Sensivty of Mechanistc- Empirical Pavement Design Methods

tothe Endrance Limit Considerations for facorpraing the Endurance Limit

Into M-E Design Procedures Chapter 8 Conclusions and Recommendations

Conclsions

Recommendations

References

APPENDIX A Proposed Standard Practice for Predicting the

Endurance Limit of Hot Mix Asphalt (HMA) for Long-Life Pavement Design

APPENDIX B Proposed Standard Practice for Predicting the

Endurance Limit of Hot Mix Asphalt (HMA) by Pseudo Strain Approach

APPENDIX € Proposed Standard Practice for Extrapolating

Long-Life Beam Fatigue Tests Using the Ratio

of Dissipated Energy Change (RDEC) APPENDIX D NCHRP 9-38 Beam Fatigue Data

APPENDIX E Construction of Characteristic Curve

APPENDIX F NCHRP 9-38 Beam Fatigue Round Robin

AUTHOR ACKNOWLEDGMENTS

‘The seach deste erin as pcre under NCHRP Proj 98 he Naina Center for Thám Prowl and Michiel Andersn served s Co-Prncipal Investigators Brown ad Prowel ere

“onlay by NCAT when ts rena ea

“The ceerchteen ncadd the phan inst the University of Nee ampsbice:Appied Resch Await nthe Univers Min ad, te, Advanced Mater LLC The phat tute ated in bev fag ting a Phan Land Ho he projet fo Dail ove nial tenon teingandanlen tthe Univer of New Hampi Harold Von Quits with Apled Research MEPDG Soc Carpenter vera nn tg testing Phe of he ty wth Shi She tthe Univers oilinoand promded experi o ples hing the rato diated energy: Ban Prove

‘wsepimaiy respons forth tara content ofthe rernng estas and forming h a

The project tea appreciate the wpport ad cai anitae of Br-We Tat ofthe Unvesy of alone ich My f SEM Atri a ill Maupin, raf he Virginia Tanportion

hd đendopabeter understanding of he varity fem fig They ab uted in he

‘alsin ofthe dat protools fo determining th endurance

‘Other individ provided ialuble tance othe projec Boe en Tal provided sigpiicant techni suingtc withthe snl and three stage Well method Chris Wagner (FHWA) suite withthe MEPDG, Buzz Powell roids techn at fom the NCAT Tes rk Davi To a chr Wiis proved advice an uence for eelating il factors bn om fd sin mse

SUMMARY

Validating the Fatigue Endurance Limit

for Hot Mix Asphalt

Hot-mix asphalt (HMA) pavements have been designed primarily to resist rotting of

‘hesubgrad and bottom-up fatigue cracking nclasial pavement design, as design load

applications increase, pavement thicknes also must increase, There isa proving belief that

‘bottom-up fatigue cracking doesnot occur for thick pavemtents The concept of the HMA

Fatigue endorance imit—a level of strain belo which there isn cumulative damage over

mingefinite number oflod cycles—isproposedto explain this occurrence Therefore, a

tional pavement thickness, grestrthan that quired o keep stra levels at the bottom of

the HMA layer below the endurance lini, would not provide addtional if This concept

has significant design and economie implications

Research nthe 19708 observed that the log lo relationship between stain and bending

‘icles ora number of IMA mixes converged below 7 microstain (ms) at high number

‘ofloading cycles Field studies in the United Kingdom recommended an HMA thickness

Fange between 79 in and 15.4 i, for a long-ife pavement depending on Suc factors as

binder tfines Thesis of thick pavements was observed to increase wih ime, most

Iie dueto binder aging Other stadieshave documented theabsence ofbottom up Fatigue

‘racking in thick pavements andthe common occurrence of top-down cracking, Ta theory, samples tested at a strain level below the endurance limit should last for an

indefinite or infinite number of loading cycles It isimposibe to test samples oan infinite

numberof cyles Therefore a practical definition of the endurance limit, o a laboratory

life epresentative ofthe endurance imi was needed A capacity analysis indicated that at

‘minimum sue spacing, alanecareying 100% trucks 4 hoursada,?daysa week for 0 years

‘ould canrya maximum of 329,376,000 tucks that would produce 1,317,5D1,000 ane ep-

‘tions (neglecting the steer axle However a more likly traffic stream would produce

‘maximum of approximately 500 million ale oad repetitions

‘Studieshave indicated thata shift factor, ranging fom 4 to 100, must be applied to relate

laboratory and ld fatigue performance There are many reasons that probably lead to the

nee for shift factor with two primary fitors being res periods and healing Ashi factor

‘of 10 was recommended inthe Stratege Highvay Reseatch Program (SHRP) by Leahy eta

{4} Considering thi shi factor, laboratory testing 50 milion cycles woul equate

approximately 500 milion loading cycles in the eld o approximately the maximum prob

bleloading in a 0-year period, This wis considered practical target fr evaluating param

‘ers indicating an endrance lit

‘An experimental pla was developed using a single aggregate gradation, wo levels of

binder content optimum and optimum plus0.7%), and to binder grades (PG 67-22 and

676-22) The aggregates, radatons, optimum binder content, and binder gradessources

‘matched those used in the base layers ofthe 2003 National Cente for Asphalt Technology

(NCAT) Test Track steuctral etions, Using the same mixtures asthe NCAT Test Track

allowed calibration ofthe fatigue shift actor Samples at optimum asphalt content were

‘compacted to 70.5% air void while samples at optimum pls asphalt content were com: pacted 08.3: 059% air voids eam fatigue and uniaxial tension testing were performed on the four mixtures In Phas beam gue tests were performed on two aldiional mixes

al optimum asphalt content using PG 38-28 and PG 6-2,

Beam fatigue testing was conducted aeconling to AASHTO T321, Atleast two replicates

‘wer tested at 80,400, 200,07 100 msuntithefatguelivesoftwo repicatesata givenstrain level exceeded 50 milion cles The time to testa singe sample o 30 million load eylesis approximately 50 days this pont a log-log regression was performed between stain and fatigue life usingal ofthe data for which samples led in les than 50 milion cycles The strain level that corresponded toa fatigue i o 50 million cycles ws predicts Tso a tional beams were tested at this strain level

‘numberof extrapolation techniques were considered to prec the numberof ees to fllure(N) for Beams thatarenot tested ole inclding the exponent model described

in AASHTO TS21, 3 logarithmic regression, ratio of dispated energy change, and single and three-stage Weibull models The extrapolations were evaluated using samples that had Tong ftguelves—in exces of 12 milion cycle —but failed efore 50 milion cycles, The logarithmic regression and ratio of dispated energy change consistently overestimated Ny len by several orders of magnitude The exponential model consistently underestimated [tthe entice lading history to fate was not wed in the calculation This suggested

‘theexponential model was nota good technique for extrapolation fom a number of load ing cles les than N, Extrapoations using the single-stage Weibull model were gener ally distibuted around the line of quality and provided the best prediction In one case, the three-stage Weibull model provided s more accurate prediction of Nand it always provided a etter ft to the stifnes vers loalin cycle data

or each misture,loglog regresion plots were created using the dats for samples that lei ss than 50 milion cycles Prediction limits were determined forlower sta evel

“The extrapolated N for samples that didnot fl within 50 milion eyes indicated fatigue lies that werelongerthan thos indicated bythe pedton limits fom the egession of san ples tested at “normal” strain levels Ths deviation indicates the existence of an endurance limit for HMA,

1 standard practice was developed to predict the endurance limit based on tess con: ducted at normal strain levels (above the endurance limit) based on statistical prediction limits, The estimate ofthe endurance limit isthe one-sided, 95% confidence lower preic- tien mi fr the strain level that produces fatigue life of 50 milion cycles Confismation teste at the predicted strain level ae recommended The predicted endirance limits deter

‘ined using this methodology ranged from 75 to 20 ms Siffer binders tended o produce higher endurance limits, Optimum pls binder content combined with lower sample al voids produced slight inereasein the endurance limi

‘Amini round-robin was conducted to assess the variability ofbeam ftigue tet at normal strain levels and extrapolations tlw stain evel N varie signicanly depending onthe

‘train evel at which the samples are tested The logbsse 10 transformation of N, wis used in

‘he round robin analyses to prodacea relatively constant variability across the range of stain levels evaluated, Ona log basis for normal stein fatigne tests the repeatability (within) standard deviation vas determined tobe 0.248 and the reproducibility (between-lab) tan

‘lard deviation was determined tobe 0318 This resls in within- and between-ab coe

‘lent of vatation of 585 and 6%, respectively The single-stage Weball extrapolations Ina the lowes variability within and beeen abs

‘Uiail tension tests were performed on clinica samples cored and sawed from speci sens compacted inthe Superpave gyratory compactor, Testing inched complex modulus,

‘monotonic, and fatigue tess in uniaxial tension Analysis ofthese data was done using vis

<codasticand continuum damage mechanies principles o ident the fatigue endorace lit

‘ofthe PG 67-22 optimum, PG67-22 optimum plas, PG 76-22 optim and PG76-22 opt

‘mim pls mixtures Frequency sweeps at varios temperatures were run to measure the

‘complex dynamic modulus ofeach mix The dynamic modulus and phase angle master

‘curves were constructed fom these dats and the relaxation master curve obtained through Tinar viscoelastic conversion, The damage characteristic curve of each mix was obtained by running uniaxial monotonic ess to failure or by running constant amplitude fatigue tests tofllre The characteristic damage curve was used to predict the numberof exces ofl tre at iferent stain amplitudes to determine the fatigue endurance limit of the mixture

“Theestimated endurance limits from his method varied depending on whether the genes lize power law or exponential model was wed oft the data in the characteristic curve, The estimated endurance limits forthe generalized power aw ranged from 164 to 261 ms,

‘while those forthe exponential model anged rom 47 to 96 ms The relative rankings ofthe binders evaluated in Phase I appeared tobe revered from the beam fatigue tess,

‘Asecond methodology wis employed whereatigue tests wth increasing train amplitude were un in uniaxial tension to dirty Menthe fatigue endurance init ofthe mints The tess versus stain lading history forma hysteresis loop, the area of which isthe dis sipated energy per le The use of pseudo strain instead of engineering strain in consti tive analysis emoves the hysteretic fet of viscoelasticity Ifthe induced stain levels are Tow enough not to induce damage (eg, below the endurance lint) then the hysteresis loop

<allapses since the dspated energy is nly due othe viscoelastic response ofthe materi Thecstimated endurance limit using this methodology ranged from 11540230 ms However, ferences were observed between thestrin applied through the crosshead andthe induced

‘strain measured on the samples These difrences made t dificult to presely increment the induced strain in order to acurately determine the endrance lin

“Analyses were performed of LTPP data for indications ofthe endurance init A 1995 survivability analysis of data om the general pavement studies (GPS) experiments known

48 GPS-1 and GPS-2 indicated an endurance limit of approximately 65 ms, However,

‘when the nalysis was updated, and LTPP special pavement studies (SPS) data were added From SPS-1, no endurance limit was indicated During the period between the orginal and updated analyses, LTPP changed the definition of longitudinal crackingin the whee path

“This may have resulted in cracks previously identified as top-down cracking tobe ec fed as bottom-up fatigue cracks, Forensic investigations on the thicker pavements in the study, exhibiting cracking, ae required to idetily the type of cracking Stl, separation

‘vas indicated between the survivability of pavements with tensile tains atthe bottom of the HMA layer les than 150 ms and those equal to or greater than 150 ms This may ind

<atethat neglecting top-down cracking, the endurance limit may be lssthan 150 ms Addi- tional cracking dat were presented which demonstrates the iyproved aig performance

‘ofmixes containing polymer modified binders, sini to that indicated ia the beam fatigue testing program

“Analyses were conducted on data fom the structural sections of the 2003 NCAT ‘Test

“Teac o estimate the shit factor between laboratory and field performance Thee of the ight structural sections exhibited fatigue cracking in exces of 40% ofthe wheel-path tea

‘orapproximately 20% ofthe otal lane area during the test track loading eee (application

‘6f 10 milion equivalent sngle-axe loads [ESALs}) Forensic investigations indicated that the cicking in one of these sections resulted rom debonding of he HMA layers fourth section had limited crackingand cumulative damage was estimated at 0.7 a the end ofthe Toading cles

‘Allo the sractral sections were instrumented to measure strain tthe bottom ofthe HMA layer The measured strain data, pavement temperatures, and ase repetitions were used wth the fatigue transfer functions developed as part ofthe beam fatigue testing to

‘alculate incremental damage on an hourly bass Shift actors between 42 and 7538 were

‘aeulated between oratory and ld performance The shift actor of 2 was determined

fora polymer modified section witha total aphal thickness of only 4.8, Similar acu tions sere performed with PerRoad an the Mechanistic- Empirical Pavement Dsign Guide (MEPDG) The shift factors determined using Perfoad ranged from 6.7 to 450 Damage in the MEPDG was calculated using the nationally ealibate fatigue model, ot the rans Function developed as part of the beam fatigue esting Shift ators for to sections ana Iyzed with the MEPDG ranged fom 0.7 to 1,0 The predicted craddngBasellon 9019 ch ability was much higher, Overall, ts conchaded that the assumption ofa shit ficor of between laboratory and field performance was reasonable

Analysts were conducted using the MEPDG and PerRoad to determine their sensitivity {othe measured fatigue endurance limit, Analyses were conducted with both the NCAT Test Track raft (very limited range of axle weights) and the MEPD default tuck rac es sification No I for principal arterials The perpetual pavement thickness determined with both programs was sensitive tothe measured endurance, «SO ms change in he endurance limit resulted in approsimately 47-0 8-n, change in pavement thickness forthe MEPDG

‘oratin change in pavement thickness for PerRoad

Usngthe endurance mits predicted from beam fatigue txts conducted apart afthestady and the default afc classification, the perpetual pavement thicknes determined with Pee Road was approximately the same the 20- and 40-year conventional (no endurance limit) MEPDG or 1993 AASHTO Pavement Design Guide pavement thicknesses The MEPDG perpetual thickness was approximately 50 thicker The overall conclusions atthe end of

‘or 40-year period were significantly diferent, With the comventonal desis, the pave ments would have filed in bottom-up fatigue with cracking over 20% ofthe lane area at

‘0% reliability while no cracking would be expected ifthe endurance limit was considered, The implementation ofthe endurance limit a singe vale for 3 given mis appears es: sonable fOr ME design programs that wse equvaen temperatore concepts such ak PerRoad Forthe MEPDG, however, temperature asa function of depth saluted on an hours basis

‘throughout the design ife or analysis perio This can result in bigher peak temperatures, with conespondinghighersttainsat the bation ofthe asphalt yer, tan equialent temper tre methods

Research conducted outside this study indicates thatthe endurance limit based on lb: oratory testing varies aa function of temperature One interpretation of thismay be that 3mics bllt lo hai and therefore its fatigue capacity” are reater at higher temperature

‘Afi study of data fom the NCAT Test Track indicates pavements can withstand a dist bution of strains without incurring ctmlatve dammage, Both ofthese concepts have merit and both are recommended for future research aimed atbtterimplementingthe endurance limit in the MEPDG,

CHAPTER 1

Introduction and Research Approach

Introduction

Fatigue racing hasbeen an ese inthe design and pr

formance of ht mix asphalt (HMA) pavements ever since

hot mix apa pavements began tobe wet hslong been

undetstood that pavement structures that are to thi alia

fig under repeated loads This type of tructral flare

rests rom fatigue ofthe HMA mintures and results inthe

Fouation of alligator cracking

There area numberof properties in the mist that en

bbeadjsted to improve the resistance o fatigue ut the most

important property by fri the thickness ofthe HMA and is

resulingeffectonthe tense strinat the bottom ofthe HMA

Teisthisstanatthe bottom eftheasphalhatsthe primary

‘tute ofthe botom-up fatigue cracking Uimately amet

isneededto determine he expected stan atthe bttomefthe

asphalt yer more accurately anda determine the ect of

this expected strain onthe fatigue resistance of he mitre

Determination ofatigue lie based on laboratory tess gener

ally doesnot provide an accurate predicorofwhatisobserved

inthe field and shift ctor mst be applied There are many

reasonsthat roby lead tothe esd or asi ctor, nl

ing sing, est periods, healing, densifeation under tai,

temperature tictuatons concep of constant stress versus

constant sin, simulation of eld compaction versus labors

tory compaction and property changes deo other enon

rental conditions such as aging ofthe binder and moisture

<Samage The shi ator can besgaicant 151020 ies

‘ormore ofthe liboratory estimate of the fatigue Hie)

Mach research wis psformed inthe 1960s nd 187050

study fatigue of HMA Until rece yearsi¢ was believed tht

fatigue cacking aways started atthe Botton ofthe HMA

yến sn propagated up through the HMA ayers centoaly

reaching the sae Howser recently there sevens that

Some projects have exhibited friguealure due to top-down

‘racking Generally, itisbelleved hat any modification tathe

six properties that wilimprove the resistance ta baton {racking wl also iniprove the resistance top-down crack ing Whether its op-down cracking or botom-up cracking, thelikaly couse of fiiguedsessis tens steansin the HMA

‘Mest pontin he pavement structure,

“Thee isnow amc wider of polymer modified spas than in past years, Therehave een very fe ftigaesteson mixturescontaning medi spas anditisnot clear how uch improvement, f any canbe obtained from modified

“Typically tgue data are anslyedby plotingthe nember ofcylesta file verse the masimam tensile train or srese inthe HMA specimens, thar been shown tit plating these daa on algo lg of ees aur versus og of pied Stes or stain) pot illpeovide approximately stag ie Ieisvery ime consumingto conduct atgue estat very high rnumbesof cess, general researchers have applied ily high tres or asin values thatthe numberof plest fal tre wll ot be so high o provide excessive ts tne Ice

‘generally believed that approximately 1 spevinens mast be tested for each mittre being evaluated to provide suitable relationship between applied sves or svain and number of celeste

‘Since mest ofthe testing has been done at higher tess corsa eves, there snot been much work ookst the txpected performance a lower stes a mịn len even though thes lower es levels ar tp in the eld Geer" ally bes lineis etre forthe data thes lately high strain or sres levels andthe performance of mistres

at lower tres or sain eels extrapolated fom the dat ence there is ver litle understanding ofthe actual per~ formance that would be expected at hese lowe eves Some work asindcatedthatheremightbe level ofstress orstesin blow eich no damage occurs othe test pein

“Thestesostain vel below which no tigue damage oi inating rom thebotomofhe pavement stricture o<curshas

teen termed tig ena fa pavement designed

and contrasted so that under repeated ra loads m0 dam

age occurs then thatpavement shoud last indefinite without

‘stactr fire, Thispavemen ilstillequreovelyson

sme regular basi to maintain the srfce in good operating

‘condition, bu the pavement structure should provide avery

longi

Research Problem Statement

‘Workisneededto determine there isan endurancelimit

for MA andi's0, how hisinformation should be wedin

pavement design, One would not want to desig all pave

‘ments, segardles of afi level otha the endurance lit

isnot exceeded since ths would result in the sume design

‘hicknes regards of afi evel For ery igh tall

‘leit might he desirable to design the pavement 0 thatthe

fendurance limits not exceeded, but fr lv tac evel

{shel that this approach woud be 100 expensive How-

er, similar concept could be employed for pavements

Subject to very low traileelsta ene along Assay,

the pavement should be designed t provide he lowest ie

‘ele cots,

Hence, there isa need frst ofall to determine i here is

an endarance limit for HMA Onee the endarance limits

deteemined itsimportatto determine how his endurance

limit would into the new mechanistic pavement design

procedures

Objectives

“The abjestives of thisstady ae to

‘Test he hypoess that there sam endurance initia the fatigue behavior f FIMA mintres and messes ale fora representative ange of HMA mistres

Recommend a procedare to incorporate the eects of the endurance limit into mechanistic pavement design methods

Thị steady inched a erature review, laboratory testing, and analysis field at A mix design usedto construct svc Turalsectons tthe 2003 National Cater fo Asphalt Techcl-

‘oy (NCAT) Test Track, rept Far sesons othe mdsiga were vaunted encompassing he eect ofinceased 2sphal binder content snd polymer modifeation, The mi tures with higher agphat binder content were prepared at higher densest saulate the innprovement in compaction

“expected in the ld, Beam fatigue and nial tension ts: ing were conducted to determine fatigue lil, Beam Fig resting wasconductedatproresvly wer strain eves ut fatigue ie in exces F530 millon les was achieved Data from the Long-Term Pavement Performance (LTPP) study were aad to deernine i they supported the endorance limit concert

CHAPTER 2

State of Practice

Arevew ofthe exiting literature was conducted to answer

thefoloxingspoiic questions:

1 Whatis an endurance init?

2 Wht ld or laboratory stots support the existence of

anendaranc limit?

3 What HMA materia factors fect fatigue Me and conse

quently might affect the endurance iit for diferent

—

4, Nhat xe the best methods oF measuring fatigue ie inthe

bbamtong'

5 What analsi methods should be used o analyze fatigue

dain onder o identify the endurance limit fora given

LAtrenendousxolumeofierturevasideniledrdhielto

thefaalorsafecủngfatguelie,mehedeofmeisudngfaBue

lie and analyzing laigue dau, Thưc veygoodsuntietol'

thisliesture were denied, onety Eppsand Monit (2)

produced inthe erly 19708 and two ational sommares

Produced spat of the Suaegic Highway Research Program

(SHRP) Q, 3) Therefore an atempr was ot made to sum-

nivel ofthe reference elated factors aetna

ead Stig measurement

Arguments for the Existence of

the Endurance Limit

Pavements have ben designed primarily ores reting

ofthe subgrade and bottom-up fatigue cracking Incas

«al pavement desig, as design lond applications increase,

pavement thickness mst also increase Thee isa growing

bai hat for thick pavements, bottom-up fatigue cracking

<esnot orc The concept fan endrance limits been

eveloped,represemtingastrain evel esting roma com:

bination of HMA stifiness and thickness, below which

‘bottom-up cracking will ac initiate Therefore, adtionl

pavement thickness, greater than that required to ep

“Srsinsbelow the endurance it, would not provides ional fe, This concept has significant design and eco- omicimpliations

The concept ofthe endurance Limit was originally devel- ped for metas 5), Bret etal (3) describe the endorance limi fr metals being stress belo which, fr uncracked materia the plot of stress versus cycles to failure becomes eeatilly horizontal and fatigue doesnot occur igure 2.1 itsrates the theoreti concep ofthe endrance iit asit woul be applied 10 HMA

“The concept ofthe endurance limit was fist implemented for paving materials by the Portland Cement Assocation

‘An examination of fatigue tests conducted by various researchers on Portland cement concrete beams cussed inBalad and Snyder (6) and Hang (indicated tht ifthe stress ratio fe kept below D45, the concrete will have an essentially infinite aigue life The sess ato was defined

‘ste cai ofstressinduced in the concrete pavement tothe onerte's modulus of ruptore A mamma of 10 million

«cles toile was ured forthe majority ofthis testing one simple ws tested to approximately 20 milion cycle) (Fg- ue32)

‘Monismith nd MeLean (8) fest proposed an endurance limit of 70 mirostrin (ms forasphat pavements 35 served tat the log-log elationship between stain and bending cyles converged below 70 ms at approximately

5 million cycles Maupin and Freeman (9) noted a similar

‘onvergence Using lowsrain design principles, Monisitk and MeLean (8) designed a pavement structure thatineressed the fatigue lif ofthe pavement from 1210 approximately 19-plus years

Tn the ie, Nan (10) he United Kingdoms (UK) and Nghe ai [in Japan proposed concep for long hile pavements for which csi! botom-up fatigue cracking would not occur Nunn (10) dens long-life pavements as

‘hose that sta east 40 yens without stucra seogthen-

smentl roads that had carved pt 20 milion standard aes,

‘When thisstudy was conducted these ltonships were bing

ewtapobtedte move than 20 milion standard aes Nunn

(10) evaluated the most heal aveled pavements the UK,

ost of which had carrie in excess of 10 milion standard

le lạ enlsle the then current design system, Non 10)

‘once the following

+ For pavements in exes of 71 (180) hick, ring

tended to occur inthe HMA ayers

+ Surface inated cracking was common in high-rafic pavements, but there was litle evidence of Botom-up Thlgue Sunleelnilitel tac tended ostop ata depthof

“in (00 mm) + Te vas observed thatthe stifins of thick pavements was incrssing with time, mos likly vet binder aging, This

‘wouldnt endo occurifthe pavement was weakening due

to secured damage

‘+ Aminimum thiknes for alongle pavement was recom rendedas79in, withamaximum thickest of 134i, This rangeisbasedon avaretyoffacorsoch abner sites

in seh Figure 2.2 Mustration of endurance imi for concrete pavements (7)

Nighizawa etl (11) reported an endurance iit of 20 ms

‘bse onthe anal ofn-service pavements in Jpn Sm

lr tan evel rthebowom otthe mỹhalt byerofbeneen

‘and 18 ms were caculitedbssed on back-clelated st

ness data from healing weight deflecometer Bralengcifc

pavement in Kansas (12) Others (13 14 eeprt sii ind

Ings, parieohrhpth stsenec ofbofomanp Iigue cracking

inthik pvemenlsandthecommmoneetrrenccoltop-dow

cracking

Factors Affecting Fatigue Life

Asigniicant amount of aie research was conducted in

the 1960s and 1970s Epps and Monismith (1) provide a

summary ofthe effes resulting from binder stifaess,

"alt content agaregate ype, aggregate gradation and ác

‘oid content, Table 2.1 indatsthe elaine affects ofthese

components The authors conclude that binder stifnessand

Sievoid content havea larger influence on fatigue ethan

‘sugegate type and gradation The SHRP A-44 research (3)

noted that angulir aggregates tended 0 produce both

ster mises an longer fatigue lives Harvey and Tei (I5)

‘vlusted the effect of air voids and asphalt content on

Fatigue life In most previous evaluations, constant com

aston effort ay used to produce sample This ests in

Sroids being highly corelated with the ash content of

{ven mixin which ase there ite relationship Between

asphalt content and fatigue fe In this instance, speci a

vod lees were targeted, reeling in elationships between Tsdgoe li and both air void content and aps conten Voids ile wih binder (VFB) have been a ypcl parame- ter usedin tigue life prediction equations Hareeyand Ts (15 caution agains the use of VFB since various combina- tion fair voids and asphalt content wil produce he same V8, Maupin and Freeman (8) note that there was litle increase in ftgue lie resulting fom an increase of 0.3% binder, bt significant increases weve sen with an increase

of 1.0% binder, Long-Life Pavements

A numberof strategies have been put orth to potted Ieinood of eonstracting = longelfe pavement, including: polymer modification, rch Bottom layers and high-modls Asphalt bases

Polymer Modification Fatigue testing and analyses of aapalt mixtures made with modified asphalt binders hasbeen performed in ex fralstudies In 1988, Goodeich prsentedaneaty study on the Fatigue performance of polymer madifel me I7) this study three unmolfedaspak binders with deren em perature susceptibilities and two modified spa binders— Produced singonebasespultand.wolevelsof modification Table 21, Factors affecting the stiffness and fatigue behavior of hot mix

10

(vot identified) were evaluated asng mintare ague ests

The intent wast correlate binder properties wth mitre

fatigue pecformance Laboratory testing was conducted using

Alevural bean atgue at 25°C and 1.67 He loding frequency

in contrlled stress mode Test were conducted t a initia

strain evel of 400, and ndings indicated tht the fatigue

lives ofthe two modified asphalt binders were an order of

‘magnitude greater than the fatigue life of one ofthe unmed

‘ed asphalt bindees (produced from the sae source asthe

tse asphalt usd to crete the mie apt binders) The

ase asa properties per tole important inthe perors

ance of the modfed apa binders, Mates made wth one

amadBed aphatbinder vậh lo temperature ssc

"ạphad sppresinatdl tuoto tưeeneethefadguelf sfthe

polymer modiied mitures

Daring the Strategie Highway Research Program (SHRP),

the A003 contactor elite thew of he Renal beam

fatigue testa a mixture performance test for fatigue The

‘modified asphalt mistres experiment (MAME), described

fn SHRP Report A-401, was performed to determine if the

fatigue chaise of modified mixtres could be ra

sted using the eral beam fatigue es (8) Asphalt mistres

‘were made sing one aggregate source, thre aspal binders

(AAF-1,AAG-1,and AAK-1)and tweemoaifers(identied

33 M-05,N-415, and M416) Test resell indicated that

‘the ation of modifier M-405 to cach ofthe thre asphalt

binders decreased the fatigue life compared to the unmed

‘ied asphalt mixes The addition of modifiers M-41S and

[Mls hada nitive fet onthe fatigue life of aies nade

vith AAG=I, but substantially increased the aig ie (by

pproximely thre to five times) of mistores made with

[AAK-I compared to the unmodified mixtures, A validation

study was peeormed using sib whee track estson mistures

‘nade with AAG-I and the thee moaifes, Although the

fess were similar forthe M-405 modi (decrease in

fatigue if) the Mel15 and MeA16 mdr resulted in an

Increase in atiguelife fom the sab wheel tack es Thiswas

contrary tothe ndings ofthe flexural beam ftigue es,

‘Shortly afer the implementation of the Superpave

performancesgraded asphalt binder ests and pecfction,

‘ses recognized that the properties of modified asphalt

binders may nt be characterized properly using the Super

ve binder ss This ed to the fing of NCHRP Project

5-10, "Saperpave Protocol for Modified Asphalt Binders”

{in 1996, The esearch 19) was condactedby the Unversity

of Wisconsin-Madison, Asphalt Insitute ané NCAT, undet

the direction of Hussain Bab

‘The NCHRP9-10 esearch evausted the effectiveness ofthe

lnnermediste temperate sitfiness requirement (C'sin 8S

5000 KP) by performing exural bea fag tes on four

‘sggregate sractres (gravel and limestone, coarse an fine

fadtion) using nine diferent aiodied tsphalt binders

Testing was conducted on mistores using 10 Hz insoidal Toading eth a peak-to-peak strain of 0 ms The tes tem perature wae selected foreach mistore asthe temperate

‘where the intermediate temperature requirement Was met {sin 8= $000 KP} So, unk other studies in which the tempezatare was xed (usualy 1 20°C), he viscous compar net ofthe shear moduls wat fixe,

"Arie examination ofthe dati the eportindicts that the fatigue lives ofthe modified miatures could be signif any diferent Mistuses made with « PG 82-22 asphal binder produced using radial styrene-butaden

(588) moder had tig lives st 24°C that were 0 0 Se times thefigue lives tested at 32°C) of minturesmae with

sn unmodified (aided) PG 82-22 apa binder Since the Grain B value wat the sume fr etch ofthese mixtures thật respective test temperatures, he researches considered the intermediate temperature criterion nthe PG binder speci=

‘ation to be inadequate fr asessng the Ftigue characteris Tics of asphalt binders

mm SHRP (18) ndatedastrangrelationship between temper tre and fatigue life a gven strain level with fatigue lie

‘ecreasing asthe temperature decreased Thus, it could be hypothesize thatthe diference in ftgue life between the Shai modifed PG 42:22 mistores and the oxidized PG82-22 mixtures woul be reterifthe test emperateres had been gual

Monismith et al (20) reported on the development of the esgn and specications for the California 1-710 eeabilts ion he stady, AR-800 (roughly equitalenti0a PG 61-10, (FPG 64-6) and PBA-6a (PG 61-40) asphalt binders were sed to prepare metres for testing, When tested ot 20°C

ng the procedure described in AASHTO TS21, the mes sured aigue ile ofthe PBA-Sa stares wa approximately

Sn order of magnitude (10 times) greater than the Fatigue leo the AR-8000 mistures Thiseltonshipseemed tobe affected bythe applied stain resulking in an increased i= ference between the two sets of fatigue ves at higher stn deel

Lee etal (21 reported on laboratory craluation ofthe sffets of aggregate gradation and binder type on mixture fatigue if Using uniaxial tension fatigue est ress (on ducted a 25°C) anda viscodasti fatigue model, he ators

‘lt that mintores made withan S8S-modfed PC 76-22

‘spal binder ad 10 times greater fatiguelife than mintres made withan unmeaifed asphalt binder, regardless fare

¬¬ Research performed in 2002 bythe Asphalt usdufefr the Asphalt Pavement alliance evaluated the posit of fatigue endurance limit by testing mixtures made with two asphalt binders (PG 64-22 and PG 76-22) and two asphalt binder contents at various stain lvl The results, shown in

Figure 2 indicat thatthe matures made with the PG 7622

_sphalebiader have aprosinately an onder ofmagnitade Le,

Time) greater tg ethan the mistres made with the

PG 6H-2 asphalt binder

‘Von Quintus{23) conducted astudyo quantify thefts

‘ofpolymer modified asphals Based on alterature review, Von

‘Quintsreported that PMA mistores generally lst about 25%

Tonge than conventional mixtares Some prematre floes

“hud concen, However, masta the flares were” found

to occur pir tothe adoption ofthe Performance Graded (PC)

binder specication and canbe traced back o inferior com

straction (fr example, high ar voids) inferior materials,

andlor inadequate desig thickness” The stud also notes

“One ofthe more important findings fom the recent field

‘experiments i that many ofthe PMA pavements ace not

‘exhibiting fatigue cracking or have less load-elated crack

ing than the conta sections (snmodifed minure)”

Tn summary there appersto be significant historical data

ingiatng that the laboratory fatigue performance of modi

fe asphalt mixtures is eater than mintues made with

smoslifed asphalt binders In some reported cases mo

fed aaphaltmintres have exhibited an order of magnitude

areatr fatigue life compared to unmosited asphalt mix tures The fatigue characteristic appa to be dependent on

the base asphalt binder sed for moction,

Rich Bottom Layers

‘The concept ofa rich bottom lye originated from the

-Ausalian experience (24) and was explored duelog SHRD

‘experiments Two potential benefits are creted through

the use of rich baton lye: nceased asphalt binder on

tent and decreased ir voids inthe baton yer (asa result of

‘easier compaction ceated by th addtional asphalt inde)

Several known Iitersture soures confirm the rationale

for these concepts Inthe mix design fatigue experiment

reported in SHRP Report A-04, one asphalt-aggregate iuet(RB apgrepateand AAG-1 aspal binder) was used

to prepare specimens a thre different level fair aide and two diferent levels of asphalt binder content (4.3 and

‘08 The effect of asphalt binder conteat on fatigue ie was

a primary focus ofthis study since the 8 x2 expanded test, eopram experiment did nt evaluate asphalt binder content bea varibl, A separate 2% 2 plo est program wed to asphalt binder content defined as optimum and high—with the bigh asphalt binder conten established a 0.6% higher than the optimus asphalt bindee content For the 2 2 pit test program, statisti analysis of revaltsFor

‘ral beam igae(contolestres and contrled sein and oer tess indicated tht asphalt content did not signifi-

‘antl afer fatigue ie (18)

The eau ofthe mix design Fatigue experiment indicate

‘hatanalt binder content signieamly feted flexor ti nes (decreasing by 8% as asp binder content increased)

‘nd fatigue if—inereaing the fatigue life by 67% a5 the sphalbinde content inceasediom 48% 10 6% Asin ther txperiments, increasing the ar vod content rele i 2

‘ecresse in ero sttinessof 3% and a deste in aie Ife by 459 a aie wos increased from 4% to 86 (18), Harvey and Tas (25) conducted study onthe effects of asphalt content and ai wold content on misture fatigue and siliness, Samples were produced at Give asphalt contents 4.0%, 4.59%, 5.0%,538,and 6.0% by weight of oggreate and three anges of atid 15 03%, $85 106, and 764099,

In this experiment, a constant compaction effort was not used, oa voids were independent of asphalt content Con- Stan strain ets were performed at wo tan level: 30 3nd

150 ms Analysis the dts nated that higher gph on tentand ower ai vodscesltedin longer atgue ves Lower Asphalt content and ower sr voids esulted in bighe inl sities, nstead of esing stlfiness in miatre ati lie pre- Aton model, the asthrs recommended evslvating the

2

oratory fatigue ie nd then predicting pavement pron

ance by incorporating the lect of mite stiffness onthe

predicted pavement strains resting from layered laste

analysis Using this methodology the authors show the poten

‘albenefis of rich bottom ayer-Iisls important tore

ognize that lower in-place a Woids increase sins while

‘esting in longer fate ves, apposite conventions wie

«dm that suggests that ser mines Pave shorter fatigue ive,

Monismith etal (20) reported on the development of

the design and specifications forthe California 1-710 reba

biltation tn their sty AK-8000 (roughly equivalent toa

PG 64-10 or PG 64-16) and PBAM6a (PG O40) agphalt

binders were seo prepare mintres for testing ato com

binations of asphalt binder content and ae void content

‘Mines ere prepared with higher phat binder content

and 3% lower sir voids, When tested t 20°C sing the proce

‘ore described in AASHTOT321,the messoredatige lie of

‘he mites with 0.85 higher asphalt binder content and lower

ale void contents was apposite two anesthe atgue ie

‘of mites prepared lower sapal binder content

Anderson and Benen (26 reportedon at caating

the influence of wos in mineral aggregate {VMAI on ml

ture performance Since VMA i elated o aphal binder

contets natures wit high VM had asphalt inde cơn

tentsthat were approsimatly 1.85 higher than the low WMA

situres, Flexural beam fatigue testing conducted at 20°C

fd 500 ms insted tha the mies ith he igher pale

binder contents had two tines renter lboratry fatigue lite

‘han the lw asphalt binder content mies

Harvey etal (24 repotedonCaliforia'sexperences with

the design and construction of long-life asphale pavements,

‘The authors reported tht most ful depth aephal lon 1E

sigs willinciudea i faigue-resistant bottom ye This

layerstermeda i oom layer dsgned to havea very ow

aie wid content approximately 0 to $9) ules of the

Inyeriaconsideratonsineit intended to reduce the over

allshicnessof the HMA layers Thelow ar oid content also

duces permeability and improves moisture resistance The

benefit of ich bolton layer te maniized with thickness

ange of 89 mm to 75 mm nos has alo adopted thin

‘concept (4)

(Generally the stifness of « mix can be increased with

Increased compaction Farther, inceeased compaction gen

erally increases fatigue ile a the same strain level Ths

‘ncreased compaction specications or ower ifs esl in

both ower strains de to nce stitnes and alo increased

fatigue life fora given strain level, producing « more eco

‘omnia pavement The unbound lyre must be slfiently

site low high degree of compton in the bottoms HMA

layers

‘study conducted by Maupin (27) examined the impact

ofagphalt content on durability of Virginia surface mistre,

Testing was conducted on 9.-mm and [25-mm nominal ravimum aggregate size (MAS) mistres to examine the fier of increasing asphalt binder content, Flexural beam fatigue test conducted at 0 ms indicated slight increase

in fatigue life wit the 0% higher asphale content A sub

an inceeae in aig if was noted with 1.0% higher ssphat content

High Modulus Base Europeans have used si binders to produce high modu- Tas ase ayer (10, 1.28.29) Corte (28) reported the we

‘of high modulus asphalt mistres in France The ist use of highsmodus asphalt concrete (HMAC) occured around

1980 nly these mistures were sed for sengihening or rehabilitation where pavement thicknes was constrained or instance by bridge clearance) The use increased in 1985.Tt

‘was found that locally alae weak aggregates could be

‘sed with tif binder, These atures were designed with relatively high asphalt binder content an low voids (Less than 66) Constant stein fatigue ets indicate that HMAC mises are more fatigue resistant than conventional base mistuces,Thisibeleved tobe duct th higher asphalt con tent and lower vide found inthe EMAC mistuts Cortes Findings match the Findings i simile stadies where lower sirvoids increase stifiness, but also appear to inereas fatigue Ife (20,21) The stipes ofthese ayers reduce the strain at the bottom ofthe asphalt layer using les thickness than

<onventonalaspls Cracking canbe a problem with these ries, Corte (2) discusses inder ests 10 sine elke Thood of racking

Laboratory Fatigue Tests and Analysis Methods

‘The SHRP AODS-A projet 2,3) erated seven methods

‘of measuring laboratory fatigue lie Repeated fad flexure tnd direct tension tests received the highest rankings A tmethodology was developed to evilate fatigue ie using fexural beam fatigo tess conducted constant sain mode tI0 He Thin pavements ar generally subjected toa mode

‘of loading best represented by constant stain Thick pve tents re generally eepeseted by a mode of loading most

‘osly represented by constant sees, However, the SHIP [ADDS-A researchers recommended constant stan est for allpavement loading conditions Thisrecommendation was

‘based upon the act tht fatigue evaluations are made inthe contest ofthe pavement stature (eg by caleuaingexpeted

‘traingat the bottom of 3 given pavement structore) then

“Constant ses andl constant stain tests give similar ean ings (325,10), AASHTO T321 isthe current standard for

‘beam aig ets

The et tension test was eliminated ery in the SHRP

03-4 research ductus aligning and gripping the

specimens (2), However, research under te dretion of Kim

Sppeat to have improved this technique 31-33) Previous

rescarchers noted that only a prtion of asampe’s dissipated

energy i most key causing damage (84.35) Dale and

kim (32) developed a method using a characteristic curve

From which fatigue ie canbe predicted apa from monet

‘ic unas tension tests characteristic curves enested

by modsing vicoeistic material Behavior using Schapery’s

correspondence principle, continuum: damage mechani,

nd work potent theory Thi method may be sed to more

rapidly determine the endurance iit ofa mistre-Uniowal

teason testing would at require the production of an HMA

‘beam and instead would use asamp more closely related to

‘hose ein contented for simple performance tetsrelated

mà

avpin nd Freeman (9) evaluated five simpletetstope

it aigue ie niet tenets resus were found to be

‘coveted othe eoefliients wed in standard equations to

Pred tiguelifeinboth constant stress and constant strin

modes of testing Von Quintshas indicated that longlife

pavement may be designed where the srsin atthe bottom

‘ofthe asphalt ayer i ss than 1.0% ofthe indiect tensile

strength lore sean Thus indirect tensile strength may

iso provide rap sreening tool to evaluate the endrance

Timi of given mitre

Laboratory Studies to Quantify

the Endurance Limit

As nerestin longi pavement res lborstorystules

‘begun tot to validate the existence ofthe endurance iit

and develop methods of determining it fora given mixture

‘Ghrlanand Carpenter (34 proposed the we of he dspated

“nergy ratio (DER) to deine the essence of an endance

Tmt The disputed energy fo agen Fatigue ce is cle

Ite asthe arexofthe tess staalysteresis loop (3 38}-DER

is spl the rato of dispate energy trom one cyl tothe

rex: Daring the course of fag tes, three regions ofthe

DER curve versus loading eyeles may be identified: a ini

val dawaward tend, plateau witha nearly constantenergy

inpetsand allure region where the dissipate energy apy

increases which oscors at approsimatly 40% of the inital

sine (34, 35) Carpemer ti: (06) conducted addtional

beam fatigue esti he range of 700 100 ms wth samples

‘being tested o between 38 and 46 mon ces They con

led hat low stan texting inthe ange of70 ms esuted in

eraerlmal long atigue lif” Researchers at the Univer

sity of lino nde the detion of Carpenter, have con

Gucted a: number of low sin beam fntigue test tht

indicated a breakin fatigue ie behavior for samples with

„ Faigue ives in exces of 1 lion cles (37) Tests wero sondoeedthat ndeated that period vera (loads ce singstrin eeeinexcessf the endurance ii) would not Substantially reduce the fatigue lie where the majority of the load cycles were les than the endurance limit (371 The Asphalt estate conducted study to identiy the endurance lito the Asphalt Pavement lance (3,38).Beam samples were ested to maximum of milion cycles Extrapelatons Suppor the presence ofan endurance iit between 70 and

TH me

Shen and Carpenter (40) developed + new method for

<keterminng/predictng the endurance lint Their esearch indicated a linear relationship between the lg ofthe plateau Yalueof the ratio of dspated ney curve (previous) caled DER) an he lg of eyes o $Me nil stifines for both oral and Tow Tbelow the endurance iit) stain levels Fanher,they bdlete this methodology can be used toe trapolate he failure point for tests conducted in site as 540,00 cycles Atenttive plateau value o 8 57E-9 was iden-

‘ed asinicatng the endurance iit, This appear vo bea Promising technique fr analyring the endrance imi,

Modeling Fatigue and Relationship

to Field Performance

‘The NCHRP 1-574 Design Goide has instteted an

‘enhanced version ofthe Asphalt Institute Model for fatigue life (11) The eabancements were developed ta better predict the performance of thin pavementsinacoastant stain mode loading, With hit mods, pavement hikes wl contin= ullinerease wih increasing design traf loads The ites ture indicates number of acorsthat must be accounted for ifanaltenate appeoach for atigue ie determination, which incorporates the endarace lin developed,

Previous studies indicate a coelcient of variation of approximately 40% for beam fatigue teste (3,30) Tisai ably must be accounted fr when making fatigue life re- Aitions and extrapolations Several methods or accounting for this variability have been proposed (30, 2-11) For Fatiguele predictions below the endarance ni, viec bility of design trafic estimates should also be considered (G0, 43) However ifateuly infinite fatigue eis estimated bythe endurance lint, design trai liability ay be of leser mnportance, Harvey etl (43) developed 3 method that incorporates the variability frm laboratory fatige tests; materials production, and construction Material pro- duction varibilty includes asphalt content and in-place ir olds: Construction variation i represented by variation in pavement thicknes ad subgrade support Mont Caro sm Ultion is used to entimate the combined affect of meri

‘nd construction variation, Savard etal (4) report on 3 French methodology to determine an acceptable stan init

“

tse on variability in the ftgue test and subgrade sup

port, This methodology could be used shift he messured

fedrance limit fora mitre rom 508% reliability of the

Inboratry test dats to an acceptable level of reliability for

the constructed pavement

Fatigue tess ae typically conducted at 20°C However,

pavement damage oriet wth temperatire and theres

{ng changes in stifines ofthe HMA layer Methodologies

have been developed to sb fatigue fe results a 20°C to

an equivalent annual or monthly temperature (30, 48-15)

Overloads, or strain levels exceeding the endurance intact

ost ikely to occur either inthe warmest samme months

(37) ec when the spring thaw occur, depending on environ

mental conditions,

Finally, in-service pavements tend to ave ong Fatigue lives than those nicated by Laboratory et Ths, ft factor is applied to nboratory fatigue le to predict fed fatigue ite, 43-44, 46-47) The sbi factor ie belived

to aecouat for such tings asthe affect of tst periods and healing Shit factors may range om 10t0 10043) Leaby

st al (6) recommend shift factor of 10 for upto 10% fatigue cracking in the wheel path, Harvey tal, 43) cl

‘ated hf factors for California conditions The shi fa tor calclated with thei equation increases with decreasing

‘Men levels, A sht factor of 16.6 would be elelated for

70 ms Pierce and Mahoney (17) note that Washington

‘State utes shift factors between 4 and 10 Smaller shift fa tore ae used for ticker pavements

CHAPTER 3

Research Plan

Introduction

Based upon the review ofthe literature, controll Ibo

ratory expernnentl plan was developed The experimental

plan was developed withthe primary objective of esting the

hypothesis that there isan endaranceliit for HMA mistres

‘Asasecondary objective sone of the HMA mesa proper

tes that fet the endsrance it were inestigated

‘A working definition of the endurance limit was deve:

‘opedasafameork fr testing within he experinentl plan

‘Although the endarance limi i deined aan eset i

frit fatigue if for tal los testing for an infinite ies

impractical The endurance init must be defined in practi

<a sable terms iit so ave meaning For example the

literature ae defined 401050 years resonable istimeto

‘beconsideredasalonghstng or perpetal, pavement Hence,

<eteinng stain eel that resin #005 years (or

ren more of pavement eis avery practical wayto deny

the endurance it

‘The Highway Capacity Mana tates tht the maximo

numberof passenger sr per hour pr ane fora rey at 3

fre ow speed of 5 mph 2,880 (48) a rlig teen,

single ruck or bus would replace 2.5 passenger cas (1)

“Thus one would expect amaximum of 940 tucks per howe,

2,540 tucks per day, ora maximum of 329,376,000 racks

in 40-yeaperiod Soc a case might representa dedicated

trucklane running at capacity 2 hous day, 7 daysa week,

5365 days year, am unlikely occurrence By comparison, the

‘very heavily traveled seston of Intestate 710 in California

<atied a maximum of 3,680 trucks per day i the design

Tan (20, By calelatng the appropriate eanyvehicle adj

rent itor and determining ts impact on afc Hows (8),

ried tac steams with 25% and 50% trucks would pro

<docea maxima of 148,219,200 (10,152 racks per day and

28511800 tuck in a 4-year period respectively

Consider, fr example, an FHWA Chass vehicle ov fve-ale

single walr, which typialy consists of two tandem aus

Cs wehil woul provide our load rpetionsto the pve

‘ment fora maximum total of 1.170800 ale load repet- tions in a 40-year pevod This epesensa theoretical max= run loading where every tracksfuly loaded and the design Taneisat maximum capacity for 24hoursaday.7 dysaweek fora40-yearperio Ths oaing condition would beexpected tobe even more severe than a dedicated ruck ne, A siniae methodology was used by Maboney o calcul the naxi-

‘mum smber ESAs expected in 0-year peded(uaptbc lished data,

In atu mixed teat streams the highest percentage of teuck tend tobe aout 0, which woud reduc thế n3:

‘um aumber of load repetitions a 940,473,600 or a mati- rum mumber of load eepetitions of 582,576,800 for trucks Even the most heavily traveled highways do not maintain taf steams at capacity 24 hoursa day and not

A iruchs are loaded, The fact that all trucks ae nt fully loaded ilstratedby a Washington DOT study of 10 weigh- inemotion sites over a one-year period, which inated that the typical number of ESALs fora Class 9 vehicle was 1.2160) 11.2 "desig load” ales were applied per tuck for the maximum number of techs per lane ina 40-year period (629,476,000), a otof 395,251 200 oad repetitions would

be applied Als, in winter months in many pats ofthe Country, the pavement stiffness i very high, and this results

in sigaiicantly lower strains Therefor, it sa reasonable assumption that the maximum posible number of load repetitions expested in 3 40-year period is approximately

500 milion, This ould be considered a practi target for evaluating parameters (tain or energy) indicating an endurance lin

6

Research conducted during the SHRP recommended ashi

factor of ID betwee aborstory beam fate els nd field

erformanceyequtingt0 10% erckngin the whed-path 6)

Considering this sit str, hboratary testing 30 milion

eles would equate ro approximately 800 milion lading

‘eles inthe feldorapprosimatey the masimucn posible

loading in a 40-year perio Based on these analyses, a mix

thatprovided 30 milion cylesor more fatigue fein the

Inboratory ws considered 10 be indicative of longclie

pavement pavements were dvgned to haves strain level the en

rane it hen pavements, reels of tral, wold

be designed with approximately the same thickness of HMA

(assuming the same underlying suppor) This approach is

contrary othe way pavements haven designed inthe past

Sd isunlily oe cot fective for fate designs Hence,

‘tis more reasonable, expecially or highways with ow to

dium afi to design s pavement forthe expected rac

during an extended pevod of years (4010 $0) than to

simply design at the endurance Ui

Practical speaking highways with wer rac eves cn

be designed with ess pavement structure and til have long

lives Hence, the amount of trafic hast be srt clement

Jn the design process Inthe past sere hasbeen very ite

fatigue testingatlw strains (very high eeest alle) and

‘his study was designed to identify the relationship beeen

stain and ylesto failure a these ery low stains (high ples

toile) Inthe proces of evaluating law sans twa fl

‘hat the “endurance int” would be better denied,

Two est procedures were uilialto evaluate the existence

ofthe endurance knit: beam fcgu tests and uniaxial ten

fon tess, Beam fatigue test have een the mot widely used

‘method fr testing fargu inthe United Sates Uniaxial ten

Sion tests have provided an alternative that allowed a more

“ndamenbl amlyjs by modtin the vixevladie maeral

Lchavior uring Schapens correspondence prinspl, contin-

‘dum damage mechanics, and work potential teary

Materials The erature indicate that he primary material proper tes afetng fatigue face binder content, binder tins,

nd ir oid content The lieature indicated that aggregate {ridition, type, shape, and angularity have more lied

‘iets, Diferent nominal rain aggregate ses with their corresponding ifering minima VMA requirements willtend to produce dering volumes of asphalt binder Phase

A falfactoral experiment was conducted 1 evaluate the

‘existence fan endurance it and o enti factor fest ing the endurance int Two main factors were included in theexperiment Two aditional ators were fed These fc tor along with thei evel ae flows:

‘ofthe mixes forthe experiment, The mies were 19.0 mm NMAS granite mixture at optimum asphalt content with

‘oth neat PG 67-22 and SBS modiid PG 7-22 binder The verge eld gradations, shown in Table 1, were used the target gradation forthe aborstory stad Previous research

‘on fatigue as indicate hat fatigue esl are relatively nen sie to gradation, Therefor, it wa felt hat the ute of the

‘verge gradation would be appropiate

‘The base mix was placed four ifr sections NS and Nt

ad wo isin the other sections The asphalt content othe Base layer fr section NS was intentionally sta 0.5% above

‘optimum Treo, tse not included inthe veage

‘Table 22 High and intermediate temperature test

data for PG 68-22 binder

The PG 67-22 wea the 2003 NCAT Tes Tacks nom

standard grade usin the southeastern United tts The igh

temperate andintermeiate temperature binder tes data or

‘hen bier used inthe 2003 NCAT Tet Track ar shown

in Tale 32 The data aes thatthe PG 67-22 sed tthe

[NCAT Test Trackalso meesthe properties of PG 6-22

Following the procedures developed using SHRP and de

aibedin AASHTO RS allestutes underwent shoce-ern

‘ging for 4h at 135°C before compaction This short-term

ig procedure lows for absorption ofthe asphalt binder

into the agregte and simulates the aging that ogc ring

Production atan HMA fait Sample preparation affects the measured fatigue life To

reduce arbi all ofthe samples tested in the study were

‘ixedand compactedby NCAT- Individual beams wert cn

eted using a linear kneading computor for bet iu

testing Samples were then wet see o speed dimensions

Cylindrical samples were compacted using the Superpave

_prtory compactor fr uni tension testing, These er wore ltr cored and sawed to sie once they reached the

Univers of New Hampshite'slaberatory amples werecae

fully packed for shipping to oer iboats

Theale wid contents othe optimum apa conten sm

Pes weretargetedst7 £0.59, An experiment wis conducted

oases the eect eduction nai wos sing the same

‘constant tes compaction efor that wold esl fom the

‘optimum plu asphalt content 3.7% reduction in ai wide

‘vas observed, resulting in target ar voids content of 33-©

(0.56 forthe pimhum plus asphalt content spl

Phase i

Additonal testing was completed atthe end of Phase to

‘examine the variably of beat fatigue testing and seul

tio of the endurance limit and the affect of inde radeon

the endurance Two additonal binder grades, PG 38-28

nd PG 61-22, were wilized inthe previously described mix

tore optimum apa conten

“To die prcison statement has not ben developed for

beam fatigue testing fll round-robin according to ASTM

‘Ce isbeyondthescope ofthis project Asmaler cle round

robin was conducted provide an estimate ofthe vray

——

„

‘Test Methods Flexural Beam Fatigue Testing Four-poat beam fatigue testing was conducted according

tw AASHTO T321, “Determining the Fatigue Life of Com paced Fotis Asphalt (HMA) Subjected to Repeated Fes tal Bending.” nts procedure, beam specimens (380-mm length, 62-mm wth, 30-ram height areladed under stain- onuclled conditions using sinusoidal loading at 10 Hz The Tteratce indeed that beam fatigue teste were the most omen sed form of fatigue est in the United States The Titeratare also ndistd that beam aie tess were sensitive

to material properties,

“Testing ts condctd in constant stein mode, Akhowgh theltersture indicated that constant sires teste maybe mote appropriate or thick pavements, it als indated that pave- rentsneve prfonninatruecoastansvessmanne, whereas the peformunce of thick pavements can be approximated Ibyconstantstrain ets, Farther he sift mix performs the best in constant tess testing but hiss usualy not the ease inthe eld Iti fe hat mitre sities ascounte fora the analysis when caleulatng the stain at he bottom ofthe HMA layer,

achfthecellsin the experimental pan (Table3.1) wast

be tested tsi sean levels begining on the igh nd of range 38 fllows

+ Am, + sao ms,

—¬

+ yo msand + same Atleast selicteweretetfor each cell Once thefiigue lve ofboth eplictes ata given strain level exceeded 0 nil- lan cys the neat lowers eve vasaot tested AASHTO

121 indiates typi stain evel between 250 and 750 m5

“Theliteatresoagess that the endorance it in the bo ratory ison the order of70 ms (6 36} and pos up0 200 ms inthe field (11) The 80 ms strain lve was aed so that at least one strain level would be investigated that wa believed tobe below the endurance li,

Two replat tests were performed a eck stain level

“This provided a mania total of 12 datapoints to it the relationship betwen train and cycles flr, deal the research eam would have tested hve replicates ateach stain level, However, there was concern over headin Hie this wold ake at ow stain vel the research fam were assured thatthe ogg eltonshipsfor strain or energy con- c4 remained straight ines ow stain evel thre epli-

‘ates would have been preferable (51) However, since this

16

study was tryingto dent breakor cure in those elton

Ships it wa felt that eter points at mare level provided

ore information

Testing was conducted to flr (reduction in stiiness

‘of S0%) ora minimunn of $0 milion cycle Since the goal

ofthisstudywasto determine the eistence ofan endurance

limit the stain eves were Bing altered to better define the

‘endurancelinit For instance, the PG 61-2 mix tepimam

sephlt content ested st 100 ms ad fatigue lives in exces

(of 50 milion ces, but when tested at 200 ms fled poe

{0 50 milion eyes average 20445922 les) Tero,

sas devided that it was more informative to perorm tests at

"sm nermediae stain eve between 10D and 200m instead

of conducting tessa stein levels ess than 100 ms a this

‘example, 17D me was selected as the pot where the leg-oệ

‘elatonshiphetween stain and cls to flr, developed at

higher strain eves (80010200 ms), predicted aigue it of

50 milion cycles

‘Thee betn fatigue devices were usd to conduct the text

ing, The tay begin with NCAT using single IPC Global

beam fatigue device and the Asphalt Insite wing a Cox &

Sons beam fie fatre in an Intraken hyravlic load

feame NCAT ltr added a second [PC Glbal bes fatigue

device The Asphalt nstite hal some diicaie testing t

low strain level testing to geater than 1 milion cele

toile ith thei Interlaken yal lod frame, Conse

«quent the Asphalt nsiute aso obtained an TPC Global

beam fitigu device RugersUnvestyalso tested a PG 67-22

opium pls asphalt content beam a 200 me usng an

IC Global besm gue device,

In Phase Ito of th abs sed Cox 8 Sons tren x

servo-hyrauic frame and the remaining four abs sed LPC

{Globa's pneumatic system to conduct the bean atigue ts

Testing in Phase I was conducted a 0D and 40 an and the

strain evel zereenting the average othe predicted endrance

Tit oral ofthe abs testing given mi

Uniaxial Testing

Amethodologyby which the material response under:

fous vniasal tense testing conitons (ype of loading and

temperature) can be predicted from the material response

btaned fron single testing condiion hasbeen proposed

toy Danie sn Kin 32) The basisof this methodology ina

sharsctertc core that describes the redaction in materi

gry as damage increase The charactrisiccuveisgen

trated by modeling the vicoslssti material behavior using

Schapery’s correspondence principe, continuum damage

‘estan and work potential theory The chasceriticcarve

at any combination of temperature and loading conditions

{optic versusmonotenc, amplitudelat, quency) where

‘scolastic behavior dominates the material response can

be found by uring the time-temperature superposition Principle and the concept of educed time

(Chehab etal (33) demonscated that the viscoelastic time temperature si facorsareapplicbleto mixtures wth gow ing damage Therefore, the shit factors determined trom

‘complex modulus master cuve construction canbe wstdta

‘hilt the characterntie aves twaioos temperatures toa Single refeence temperature Complex moguls (eequensy

“cep) testing was conducted at five temperatures, “10°C, (°C, 10°C, 20°C, and 30°C to develop the maser curve, Uniaxial frequency sweep testing was conducted with 3 mean sires of roto prevent the accumulation of perma nent deformation It is interesting to note that Daniel and Kim (22) recommend the following for testing

‘Once the character curve and waco hit Fctors rekon, the behavior ofthe miter emperatres nd Toading ratelampliudes canbe predicted The number of lest fale odie amplitudes and temperatures were then predicted using the character curve nd the sift factors were determined from complex modulus testing Selected continuous ees to file tests were performed tower the predicted wales, Te continuous cycles fire test consists of constant crosshead sin amplitude haer= sine loading applied ontinvouly tothe spesien inthe ten sie deecion uni failure occurs Frequencies of 1 Iz and

10 Hzare used forthe aigue testing The amplitude ischosen toachieve fills ofthe specimen ata desied number oes

‘sed onthe fact hatte higher the amplitude the ase the specimen wl il or thi stad, test at 10 Ha were ase oe theverfction fgets alow comin with thebsam fatigue rests, Beciose of machine compliance, even when onsant san testsare conducted the sample rechesamited

‘mode of loading comparable tel pavements

Due olimitatonsia computer tenor andthe need for

‘a eeasonably ast data acguston eater capture the neces Sary information, only snapshots of data canbe acquieed uring the damage est Inthe continvous cyl Fagus,

‘ne-Scond sapsots of data at ate of 100 pas per yee {1.00 poins per second forthe 10 Hz loading eequency) were collected on logarithmic scale up toa tne increment

‘0f 2 10 minutes, depending upon the projected aur tine

ofthe spedaen I specimensare expected fallin a shorter

fount of time, the tne between soecessive snapehol was

reduced in an attempt to acquie data close othe atl lk

tre pont and to adequately describe the changing material

Behavior a damage grows inthe specimen

Sapte 150 min al by 75 mi in diameter were coved

fiom gyrtory simples fr testing Prior a texting, steel end

ates wet lued tothe specimen using Devcon Plastic Stel

epoxy A gluing fig was used 1 minnie any eects

‘due o unparallet specimen ends, Four loot core ype nest

‘arabe iffeenil transformers (LVDTs) were mounted 0

the specimen surface t 30° rad ntervals using a 100mm

sige length Additionally, two spring-loaded LVDTs were

Tndunted 0" apart to measure the plate-to-plate defor

mations The ram and LVDT deformations and load cll

measurements were clleced using a National Instruments

ata acquisition board and Labview software,

Testing wa pectormes using closed-loop servo-hydsutic

testing seem A89 RN (2,008) or 9 KN (200046) load

call wis used depending upon the anticipated testing oad,

The temperate was controled with an environmental cha

berthat esau nitrogen for coolingand feedbacksyetem

that maintains the tmperature daring testing An example ot

= led unas ension fatigue spl isso in are 3,1

Indirect Tensile Testing

‘The terture indicates ha parameters from the ines

tensile strength est AASHTO 7322, may be correlated with

parameters ented tthe endure limit This est a on

sidered asa posible surrogate est which could be conducted

more expedient than the long-duration beam fatigue tes

Tndvest tense tests were conducted onthe Phat mie

Figure 3.1 Uniaxial tension fatigue sample

”

2»

CHAPTER 4

Beam Fatigue Test Results and Analyses

This chapter describes he beam aie testing conducted

to confirm the existence of the endranc iit Oe ofthe

most important aspects of his research ia practical defn

tion fr the endurance iit Asphalt tures spy cannot

be tested foram infinite fatigue ein the laboratory Testing

a1 10 He, approximately one milion lsd! repetitions can be

applied to beam fatigue sample ina given dy The primary

joi ofthis testing was to conim the existence ofa fatigue

fndurance lit In order to accomplish this goat was ne

sary to develop method forestimating the endrance iit

‘hrovgh seserstd testing na reasonable period of ime A

secondary goal wis an estimate ofthe waist associated

with beam frgue testing and its potential impact on pave

‘ent thickass design

Theft portion af Cpe discusses methods or extra

latinghe fatigue falre point fr strain levels that didnot

‘esul in fale in les than $0 milion loading cycles The

smethods were applied to samples that had fatigue lives in

‘acess of 10min, but es han 39 lion loading cycle,

The second portion ofthe chapter presents the data coleted

fn Phae [8 well 8 atonal binder grades tested with

‘he same istrein Phas I Evidence ofthe existence of a

fatigue endurance ini resented foreach ofthese mitre,

The third portion of the chaper describes a limited round

robin conducted toassessthe aii oft testing

‘he rection ofthe endurance iit Finally niet eile

texte were investigated ea surrogate fr boa ftgu et

‘As scusedia Caper a single gradation nd aggregate

type was used frallothe esting, A al-ictoril experimen

shown in Table was ont in Phase to esate the

tistence of an endurance limit and oientiy factors als

Ing the endurance mit Twe main factors were included in

the expriment—bindesgeade and asphalt content Binder

trade was varied at wo levels: PG 6722 and PG 76-22, As

noted previously, the PG 67-23 also met the requirements

fof PG 6422, Asal content wae varied at two levels: opt

‘num and optimum plus 0.7% Optimum asphale content

‘was determined at Nang = 0 gyrations fw adlitin, the

‘samples sr void content were reduced from 7.02055 to

333 0398 for the optimum pus samples The vos for the

‘opti pls samples were reduced simul the expected Improved densification in the fed Advantages of optimus, plas rch boon lyers are bleed to inclade beter com Pactbiliy renter resistance tigue damage andimprowed oltre suscptiby,

Extrapolation Methods

to Predict Fatigue Life Asdlncosed previo twas desided prior tothe start of testing that beam gue est would be terminate 3 mil

on ofl ashi ctr of 1D was applied the et results from a sample ested to 50 milion cele, it would then be

‘estimated that the pavement could withstand 500 milion Toaing cycles the corresponding strain level Based on

‘apactyanalsiof lane, this then represents areaonable maximum number of lading cycles that might occur in a 40-year period For the practical defiition othe endurance lta 40yerifle wis considered tobe indiativ of long life pavement takes approximately 50 days to test single sample to 50 milion eyes, Additional anaes wil be đc

‘ust ltr to evaluate the existence of theoretical or uly infin le endurance lini

Forsamples that led in es than 0 mien xlet 50% Finite, the number of cycles to fire was deter mined rom the data aquison software conlling the es However, ifthe test was ermnated prior to reaching 50%

of nia stitins, either de to n equipment probes oF toeaching sO mill les an extrapolation procedure ws bse to etimate the numberof lading sks Np correspon cđngte 50% of inal sifness Leal, a method of exteap

‘lation would be identified that could be use to shorten the beam fatigue testing procedore wed to determine the

«endorance limit Then, samples could be tested 04 milion

Table 41 Experimental design

«eles as done in the Asphalt Institue study (38 28) 0 pos

Shy 10 milion cycles and the fatigue ie ator dose othe

‘rdarance limi predict,

Tn Phase I, testing wes conducted st progressively lower

stein level tl wo sumples a given strain level reached

5 milion crdes without reaching 50% oftheir iii it

ess (failure) Instead of testing at lower sa evel below

thot proving igo if of es 0 milion ces, sa

sere tesed at thestrain vl predicedto provides aigue

eof 30 milion yee The goal ofthis ational point was

to elp define the wansition fom “normal” suai test

tests blow the apparent fatigue endurance mi or la”

For the PG 67-22 mix at optimam asphalt content, the

ats from 800 through 200 me were used to estimate the

‘ein level that would result in fatigue life of 50 mlion

‘les linear repression was performed between the Log

‘of meand the Log, ofloading cycles to $0% inal tif,

‘The '=98.6 for Equation |, Using Equation I twas deter

mined that strain level of 166 ms should produce a aigue

lite of $0 million de, The va rounded to 170 ms for

testing purposes Testing was conducted at this stein Level

to better define the endurance lint

constant stain utd in beim Rlguetet n9) br of eyes to 50% oni sities and

When esting the PG 67-22 mina optimum asphalt content

3170s the rst eplicte aed in 947 mon ses The

‘second epi wa 35% of itn tifines 30 milion

eles Therefore testing wasextendedtoseeifhe failure point

‘could bedetermined However a1 60 millon cys, Sample 23

ful retained 53% ofits intl tines Therefor, esting ws

saspended at 60 milion les,

For the PG 67-22 mixat optimum asphalt content samples

tested a 200 spd 170 me were used oinesigateexrapas

tion techniques These sta levels and si steain levels

forthe othermines sed in Phat | provided long stig ives

{in excesof 10 milion les) while shaving defined i:

ure point tht could be ued investigate the accuracy ofthe

‘extrapolation Fv ecaiques were investigated for extapo

lation: exponential mol logarthnic model, Weibull func- tio, theestage Weibull function, and rato of dissipated energy change (RDEC), Fach ofthese dese below, and

‘examples are provided ofthe predicted fatigue lives Fatigue Life Extrapolation Using AASHTO T321 Exponential Model

|AASHTOTI21 specfisan exponential model Equation 2) forthe calelatonofeylesto 50% intial silfines 3 follows

a where

‘= samplesttines (MPa), constant and

n= number of oad cycles

‘solving for the constants in Baton 2 (52), Sample of the PG 67-22 mixat optimum asphalt conten, which wastesedat 170m, wasselecedasan example Iwas desirable oseletasumple that hodaslonga atgueifeas pos sbleandhod ached 50%} finial sins within 50 ion

«eles as ele that an extrapolation method that worked well at high stain levels may aot prov to be as acurat at

‘strain levels closer to the anticipated endurance limit Fi 0e 4.1 shows ist the loading ye versus samplestfnss fit determined sing Equation 2, The coeficentsfor Equa tion 2 were i usng the data up 104 million cyeles, 10 il liom eels and fur (247 million cycles) scan be seen From Fgare 4.1, when al ofthe data upto flare was used

to fit the mode, the model provides a reasonable estimate

of tiga ie

Logarithm of Loading Cycles versus Stiffness

logarithmic model (uation 3) using the natural log

sithn of loading cyles versus stiffs was evaluated as one

aeentiveto the exponential mode

where,

'S=the sample sifines at loading ple mand

‘vand Bare eression constants

When all ofthe fatigue data ate use to ft a logarithmic model, theslope of he fitedine athighernumbesefloading

‘elesmay befaterthan the actual dat, This esd nove

‘station ofthe fatigue life Thissiloseatedin Figure $.2 for

‘Sample Sof the PG 67-22 mix at optimum agphalk conten,

“Thefts tothe logavthnic model usingjut he re 10 milion

‘les and using al ofthe data ae inctinguishable onthe Hot Note that in Figure 4.2 the ogrthnic model provides

‘ery high R values, but the fited mods des not match the experimental data ata high numberof eycles

However, by eliminating a portion ofthe early loading

<jles,a good match tothe data can generally be obtained,

a

Figure 4.3 Logarithmic model fits for PG 67-22 at optimum, Sample 6

patculny at ow sein levels, In Figure 4.3, logsithie

‘odel were ito the data rm Sample ofthe PG 67-22 mix

St opimm asphalt content, which was ested t 200 ms In

Figite 4 logarithmic modelare shown ncudngal of the

inal loading eyes an excluding the fst milion loading

‘les, This provides better ito dat, bu il would tend

to overestimate the fatigue ie Further the numberof aly

Tending cc hat are not ncaed most be determined by

taland err Not tata the logarithmic modelssbown

in Figures 4.2 and 43 provide high Ras, even when the

Fito the data atahigh number of eles nt very good This

suggests that R values alone are not adequate to evalste

‘rapoiton models isbekieved that the poor at igh

umber of jles result from the fat that the data are cl

lected using logvthnic peogesion Tat the sampling

rateshighs every 10 eyes, when the est sinited bt ay

beevery lion cjlesat ahh umber feces Thus there

sare more dts pois oft in the early portion ofthe curv,

Rowe and Bouldin (53), when examining Ss fom the expo

neal model, concluded that aigue dita should be taken

‘vith every 586 retion in tines

Fatigue Life Extrapolation Using

‘the Power Model

the ratio of dsipsted energy, developed by Shen and

Carpentee (0) abo requires thatthe umber fest 50%

finial sincere lle in onder tater the platen

‘als Shes 50 recommendsa power model (Equation 4) for

mm

where, (= Draco the power ofthe intercept rom regression flog (5) verses og) ad

B= the slope om regression of log) vers log (0)

“The power modethas similar shapet the logarithmic model Shen (4) alo repre the nedto eliminate a number ofa tak eylesto aban 2 good io the slope thigh numbers of

‘cles, Faire to eliminate some ofthe inital les results inan overestimation ofthe fatigue lif Additonal discussion

‘nthe use of the power model and its applistion tthe ratio

‘of disputed energy willbe provided later in ths eport Fatigue Life Extrapolation Using

‘the Weibull Survivor Function fea, ute dats canbe modeled using a Weibull dsri- bution The Weibull functions cnmnonly used in elbilty nginecringf estimate sorial fe Ts et ($5) applied the Weibull survivor fonction to FMA beam fatigue dat,

“The generalized equation forthe Weibull uncion sgven by Equation 5

where, R0) =the reliability at time t wher might be time ot another life parameter suchas leading eels, theslope,

the minimum tie, and the character fe

2%

Tea et al (55) appli specie case of the Welbul fac

tiên where the minim fe, 8, wt aur oe this

‘se the characteristic ife= Uh andthe Weibull fonction

‘impliesto Equation 6 Since the bea igus loading cycles

‘te applied at constant requeny of 10H, oading cytes,

ashe substated Fortine,

‘Theses rato canbe used to characterize fatigue dam

age The fines ratio is the stiffness measured at ce a,

Alive bythe iia stiffness, determined atthe Sth cycle,

ai (56 sates that ata given cyl, he beaming tested

has probability of survival past cycle meq tothe SR

times 1005 Thus SRin canbe substued for St) Tsi (56)

presents the deriation of Equation 7, which allows the cle

nd shape parameters for laboratory beam tiga dita tobe

‘etermined by near regression

Figure 44 Weibull survivor function for PG

Figure 44 shows an example ofthe data from the ono lod me samples from the PG 67-22 optimum mistore in the fem of Equation 7, Tal etal (55) observed thatthe

‘oneave dom shape, exhibited by Sample 13, "imped tht the atigu damage ate isslowed down and attens outwith incresse eptitons and thas causes frther damage ser

4 certain numberof repetitions.” This behavior is believed tobe indiative of the endurance limit

Fatigue Life Extrapolation Using Three-Stage Weibull Function Inthe previous section, the Weibal survivor fanction was presented asa method fr modeling the fatigue ie of beam Fjgue ets tn ater sections, it willbe demonstrated thatthe single-stage Weibal function generally provides good est rat ofasample’ gue life andisreprodcible when cales- Ine by diferent sbortorie There are, however cases for Which he siagle-stoge Webul function apparently undee- predisfigoe ife, Sample 13 in Figure 4.48 one such exam

Pe (but Sample 13 snot specially labeled on the pot It

‘should be noted that with the exception of two examples analyed by Tsai the tvee-stage Welbul analyses were not ondiacted vat after the completion ofall of the Phase 1

nd testing

"To improve upon the accuracy ofthe single tge Weibull Function, Tsai etal (57 developed a methodology fr iting three-stage Weibull curve Tsai eta (57) theoried tha a

lot of loading cycles versus stint ratio coud be divided

into thre stages inal heatingandtemperstre equiv,

rack inttion, and crack propagation a the ease of low

sirin tts blow the endorance limit) the third tage does

not appear to represent crack propagation, but rather con

‘ave don sae with 3 redcedrate of damage

"AWeibull function sft each ofthese stages asshown in

Using series of mathematical manipulation (97) is

sand’ can be calelated sequential follows:

8) 08)

an itedln-I(SR)) foreach le isused the ines ane tion, A "good gene” is defined sean optimus st of param- testo minimize the fines function a set of nputeangesis Fist determined for 1, and, The input canges aredermined by visual inspection athe dat, inple linear regressions are performed foreach ag o determine slopes {ndoffet An camples shown in Fire 4.3, Tolerance te

‘pple tothe parasers determined by inspection toset in {aleanges for each coefficient The input ranges and test ta are entered into s FORTRAN program, Nistagexe, devel

‘ped Tai The program randomly generates tof param

‘ters ogee within the input range, Te ness parameter

is calealated for each et of ones and theses of genes are ranked, Good gens are mated and bd genes ae dicarded and replaced with neve genes The ye of producing gene, ranking genes By the residual som of squares, mating 3nd

“

placing genes, contnucs uni the peited numberof en

‘tations complete (38) The stage progr yl takes

5010 60 smnutes to complet 750 generations, depending on

727622 optimum Pu Sample 9200

(toning oye) Figure 46 Three-stage Weibull ft to transformed data,

the sie of the dats et, The complete calculation procedure tothe Nisage progam

Ssdeserbed in Appendix A, Proposed Standard Prictice for

Predicting the Endurance Limit of Hot Mix Asphalt (HMA)

(7622 optimum Ps Sample # 20m

tending eee Figure 4.7 Three-stage Weibull fit to stiffness data

for Long-ite Pavement Design The NCHRP 9-38 research team developed a Mictosoft Excel spreadsheet to Slve the hve tage Webul parameters which produces sin results Figures 40 ad 4.7 show examples of the three-stage Weibul ft This methodology provides a good ft to both

normal and lw strain aig data, In some cases, only one

‘orto stgesare fief three stages arentally denied

Ratio of Dissipated Energy Change (RDEC)

Dispated energy is a measure ofthe energy that ost

‘he material or altered through mechanical work, heat gener

ation, oF damage othe sample Other eseachers have used

“cumulative dissipated energy to define damage withina spec

ine, assuming that alo the dsipated energy is espons

ble forthe damage The approach sugested by Ghusin and

{Carpenter (3 considers that only portion ofthe dsipted

nergy isesponsbefor actual damage

“Typically thee regions te observed in he RDEC aly

sis as shown in Figure 48, Region | vepretents the inital

“setling” ofthe spl where teat of change of disiated

nergy decreases n Region Itherateof change of disipated

‘nergy teaches a platen representing a period where the

amountof damage ocuring tothe sample isconstant Finally,

in Region I, sample instal Begin ak the rate of change

of disipated energy rapidly increases A lower dissipated

‘nergy ratio (DER) plateau value implies tht less damage

isoccurting pr cycle Therefore, sample with low DER

Pateau value would be expected to have longer fatigue fe

than a ample with a high DER plateau value Shen and

Carpenter (40) refined this technique and segs that the

DEC plateau value (PV) should be caleultedat the number

afgelentatpreducsdữeltheintilsungledifneafN)

‘APVot837E-9 waspeopasedby Shen and Carpenter sind

tre ofa ong ie pavement (1)

The RDEC aalyss procedure is dsrbed in Append C,

Proposed Standard Practice for Esrapolting Lang Life Beam

Fate Test Using the Ratio of Dispated Energy Change

[RDEC).RDECs eatin of disipated energy change between

two dita points divided by the numberof les between the

n two dt points that the average to of disputed energy change per lading cele This i writen follows

DE-DE

ele a, comparing io next cycle bi

sh = load cyclea and respectively (the yee count between cee aandb for RDEC clcultion wll

‘ary depending on the data aquisition sot wae) and

DE, DE, = thecsipated energy producedin ood cyte,

and respectively

‘The dsiated energy for ach loading yes determined lyymessuring the ares within the tes-srain byteresisoop foreach captured load pulse This methodology is used by the IPC Global beam ftigue device used inthe study by NCAT, Asphalastte andthe Universo lino Alter ively the dissipated energy can be calulated according

to Equation 1

hers,

cổ =maximim tensile stress in eye

"= maim tensile strsn in see B= 300s,

= loading equency Hr, and

1 = time aginsecomds been peak loadand peakdeec- tion in seconds

inka,

‘Figure 48 Typical REC versus loading cycles plot and the indication of PY

”

Khi

{eng ets Figure 49 DE versus LC chart for one IDO mix

with fitted curve

and PV.Acurvefiting procedure isrecommended to obain

the best it equation forthe dissipated energy-Loading cycle

dat Its esumed tat the regeston equation of disipated

snety-losding le relationship follows power law rel

tionship, Ax (sindiested in Figure 43) The key forthe

‘curve iting process iso obtain lope (inthe power aw

‘elation plot) of he curve, which an Best represent the

Crgial Carve Ia general there ae two rales for ealting

‘the goosnss ofthe ited curv: (1) high Kesquae value,

and 2) correct tend of the DE-LC curve This ssi to

the procedures described previously or iting logarithmic

‘or power modelo thestlfnes-loading epee curvesand i

llstrate in igure 4.10,

‘Theaverage DEC foran arbitrary 100 yclesateydeacan

be simpy calculated using thefllowing equation

“ay, Figure 4.10 Dissipated energy versus various power models

Inthe RDECapproach PV is defined asthe RDEC wale at the 50% sities reduction failure point (Na) Thezfare, the PV vale canbe obtained wing Eaton 16

09) Here,the PV value dependsonly onthe factor ofthe regressed power awe DE-LC curve and the defined flare pol, Ni For longi tess, where Ni ae not known the sifines loading ec cure fst needed tobe extrapolated to deter= mine Ns estingin the cleltion being bed on double satapehiioe

thủng this approach, hen and Carpenter (40) demon= strated unique PV-Nfcueve forall HMA mixes a normal strain/damage level esting, pads of the testing cond tion loading odes, and mixture types, The tse sed fore tshlishing this lationship at ll norm testing were cared tor beyond the fire point (ie the Ni vals are known) Als, aking the ress fom long-term fatigue esting, Shen dnd Carpenter 40) demonszated tha the unig relton-

Sang Cro

% loading yee for raw data and

shipbetween Pea valid fr low strain ting nother

‘words the PV-Nfrcltionshipis unique forthe whole loading

‘ange including both normal and ow sainidamage eve

“An evo analysis wat performed fortwo dispaed energy

ore iting routines sing 1 fliois DOT mistre tested

Tow stain levels nd the results are presented in Figure 4.1,

All the samples were tested to extended load repetitions

(ie, $10 30 milo load eeptitions) Hence, the NF values

‘obtained from the tities reduction curve are reasonable,

nd the comparison canbe focused on the eror that could

beinvoled duet iting the DE-LC curve Two PV-Nines

ste shown in Figure 4.1, One is with the Ps obtained from

highest fining ofthe dispar enegyloading cycle curves

‘heather ith the PVs from vial tingtha represent

ssspated energy loading cycle curve's bes trend (especially theestension end) They ar close at elativelyhigher PV

(shorter fatigue ite} With the decrease of PV increase a

fatigue life), he PV calculated from the highest Hing

ives grenter value compared to the wales abtined fom

‘sual fitting For love stain testing, the sent that give

higher R: does noc necessarily represent the eal tend othe

‘curve, which cou induce ero, Therefore, fr low stain

Tong fatigue testing, the "highest R? iting” rue i not best

suited: Formos cases, the initial segment ofthe DE-LC curve

has tobe eliminated and ony the ater segment that gives

4 good extension vend ofthe curve should be used, since

itis more eepresentative forthe actual logeter fatigue

mm

Comparison of Fatigue Life Methods This extended discussion on fatigue life extapoation techniques isprovded asanintroductionfature method- logis for identifying the endurance limit la this study, tess were conducted ¥ a matrtmum of 50min eels

‘order confirm the existence ofthe endrance lit As noted previous takes approximately we monthstocom- Pleteasingletest at ths high numberof jes This extended test me isnot practical fr routine determination ofthe tmdurance Hit One alternative to determine the stain level that corespandstotheendranc mit for given min ture would be to conduct beam fatigue tessa low train level toa more limited number of les (perhaps less han

10 milion or approximately 10 days) and extrapolate the Ais, Thus «model would be Gt othe sifies vera oad ingcyle data andthe number of lerequiredvreash 5% othe itil tines would be extrapolted A significant vation fom aog-og pot of san versus cycles to fllure

‘woul inicatethe sal eveleoresponding othe endurance

Ts his wil be shown ater in the section on Exitenee of