nchrp_rrd_372

Upload: hoahairau

Post on 07-Aug-2018

222 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/20/2019 nchrp_rrd_372

    1/13

    NATIONAL COOPERATIVE HIGHWAY RESEARCH PROGRAM

    Responsible Senior Program Officer: Amir N. Ha

    March 20

    SENSITIVITY EVALUATION OF MEPDG PERFORMANCE PREDICTION

    This digest summarizes the findings from NCHRP Project 1-47,“Sensitivity Evaluation of MEPDG Performance Prediction.” It was preparedby Amir N. Hanna, NCHRP Senior Program Officer, from the contractor’s final report authored by Charles W. Schwartz and Rui Li of the Universityof Maryland, College Park, Maryland, and Sung Hwan Kim, Halil Ceylan,and Kasthurirangan Gopalakrishnan of Iowa State University, Ames, Iowa.Charles W. Schwartz served as principal investigator.

    Research Results Digest 372

    INTRODUCTION

    The AASHTO interim edition of the Mechanistic-Empirical Pavement DesignGuide (MEPDG) Manual of Practice waspublished in 2008. This Guide and relatedsoftware provide a methodology for theanalysis and performance prediction offlexible and rigid pavements based onmechanistic-empirical principles. For givenpavement structure, material properties,

    environmental conditions, and traffic load-ing characteristics, the structural responsessuch as stresses, strains, and deflectionsare mechanistically calculated using multi-layer elastic theory or finite element meth-ods. Thermal and moisture distributionsare also mechanistically determined usingan Enhanced Integrated Climate Model.These responses are then used as inputs toempirical models for predicting pavementperformance in terms of distresses suchas cracking, rutting, faulting, and smooth-

    ness. These empirical models have beencalibrated using data from the Long-TermPavement Performance (LTPP) databasefor in-service pavements that are represen-tative of the conditions encountered in theUnited States.

    The performance predicted by thesemodels depends on the values of inputparameters that characterize the pavement

    materials, layers, design features, and condi-tion. However, because these input param-eter values are expected to differ from thosefor the constructed pavement, the predictedperformance will also vary to some degreedepending on the input parameter valuesEarlier studies conducted to relate predictedperformance to differences in input param-eter values have not addressed this rela-tionship in a systematic manner to identify

    the relative influence of input parametervalues on predicted performance. Alsothese studies have not considered thecombined effects of variations in two ormore input parameter values on predictedperformance in a comprehensive man-ner. Research was needed to determinethe degree of sensitivity of the perfor-mance predicted by the MEPDG to inputparameter values and identify, for specificclimatic region and traffic conditions, theinput parameters that appear to substan-

    tially influence predicted performanceIn this manner, users can focus efforts onthose input parameters that will greatlyinfluence the pavement design. NCHRPProject 1-47 was conducted to address thisneed; this digest summarizes the findingsof this research.

    The original research version of theMEPDG software (Version 0.7) was

  • 8/20/2019 nchrp_rrd_372

    2/13

    2

    taneously across the entire problem domain for eachpavement type; the results were used to evaluatethe mean and variability of the sensitivities as wellas any potential interaction effects among designinputs. Artificial neural network (ANN) responsesurface models (RSMs) were fit to the GSA results

    to permit evaluation of design input sensitivitiesacross the entire problem domain. The key findingsfrom the study are based on the GSA results.

    A normalized sensitivity index (NSI) was adoptedas the quantitative metric. The NSI is defined as theaverage percent change of predicted distress relativeto its design limit caused by a percent change in thedesign input. For example, considering sensitivityof total rutting to granular base resilient modulus, ifa 10% reduction in base resilient modulus increasesthe total rutting by 0.01875 in. with respect to its

    design limit of 0.75 in., then NSI = 0.01875 × 100/ (0.75 × 10) = 0.25. In other words, the 10% reduc-tion in base resilient modulus will increase ruttingby (–0.10)  (–0.25)  0.75 = 0.01875. At NSI = 1,the percent change in distress relative to its designlimit equals the percent change in the design input.The mean plus/minus two standard deviations valueof NSI (i.e., NSIµ±2σ) was adopted as the rankingmeasure because it incorporates both the mean sen-sitivity and the variability of sensitivity across theproblem domain. The following design input sensi-tivity categories were defined for the GSA results:

    Hypersensitive, NSIµ±2σ >  5; Very Sensitive, 1

  • 8/20/2019 nchrp_rrd_372

    3/13

    3

    rutting, and total rutting. International RoughnessIndex (IRI) was not considered explicitly in theinitial triage because it is a function of the otherprimary distresses. Reflection cracking and fatiguecracking of cement-treated bases (CTB) were notconsidered.

    For rigid pavements, the triage process exam-ined the influence of design inputs on faulting andtransverse cracking for JPCP and punchouts, maxi-mum crack width, and load transfer efficiency (LTE)for CRCP; IRI was not considered explicitly in theinitial triage.

    Some of the design inputs are correlated and/orhave other characteristics that warranted special treat-ment. For HMA pavements, these included unboundmaterial properties, HMA dynamic modulus andbinder properties, and HMA low temperature proper-ties. For concrete pavements, these included unboundmaterial properties, concrete stiffness and strength,water/cement ratio, dowel diameter and steel depth,and edge support condition.

    Ranges for inputs for the flexible and rigid pave-ment were identified for use in the sensitivity analy-ses. The baseline values for these inputs were usedto define the OAT sensitivity indices, and the mini-mum and maximum values provided the range forthe GSA simulations.

    Base Cases

    Sensitivity analyses were conducted for the fullranges of all model inputs and outputs that were con-sidered practical. Seventy-five base cases that coveredthe commonly encountered ranges of pavement types,climate conditions, and traffic levels (5 pavementtypes × 5 climates × 3 traffic levels) were selected toevaluate the entire solution domain for the MEPDG.For these base cases, sets of climate conditions andtraffic levels were used as global inputs in the sensi-tivity analyses for all pavement types. Five climatezones were considered—hot-dry, hot-wet, temperate,

    cold-dry and cold-wet—and three traffic levels thatwere classified based on AADTT values and truckvolume as low (15,000) were used in all GSAs. The resulting15 base cases for the sensitivity analyses for eachpavement type were abbreviated in the form of“XXY” in which “XX” designates the climate (CD= cold-dry; CW = cold-wet; T = temperate; HD =hot-dry; and HW = hot-wet) and “Y” designates thetraffic level (L = low, M =medium, H = high). For

    example, CDH designates the cold-dry climate atthe high traffic level.

    Sampling and Analysis Process

    For the OAT local sensitivity analyses, small

    perturbations above and below the baseline valueswere considered individually for each design inputand the corresponding distresses were predictedusing the MEPDG software. These values were usedto compute the sensitivity metrics.

    For the GSA simulation inputs, Latin HypercubeSampling (LHS), which is a widely used variant ofthe standard or random Monte Carlo method, wasadopted. The LHS approach reduces by a factor of5 to 20 the required number of simulations as com-pared with the conventional Monte Carlo method.

    The GSA simulations provided predictions of

    pavement performance at random discrete locationsin the problem domain; these were fitted with con-tinuous RSMs that provided a means for computingderivatives for sensitivity indices. The primary met-rics selected to quantify sensitivity of model outputsto model inputs were a point-normalized sensitivityindex for the OAT analyses and the regression coef-ficients from normalized multivariate linear regres-sion and a point-normalized sensitivity index for theGSAs.

    OAT SENSITIVITY ANALYSIS RESULTSIn an OAT analysis, each potentially sensitive

    design input was varied individually for a givenbase case or reference condition to quantify andassess the local sensitivity of the predicted output(distress). The OAT local sensitivity analyses wereperformed to confirm the assessments from the ini-tial triage of design inputs and to identify sensitivedesign inputs for use in the GSA simulations. Resultsof the OAT analyses and the results were used to cal-culate NSI values for each design input-pavementdistress combination for each of the 15 base cases foreach pavement type.

    The OAT analyses provided the final rankingand sensitivity categories of the design inputs basedon the maximum absolute NSI calculated for thedesign input for any pavement type, base case, ordistress. The overall sensitivity values for the designinputs scenarios were categorized as

    • Hypersensitive (HS, NSI > 5);• Very sensitive (VS, 1

  • 8/20/2019 nchrp_rrd_372

    4/13

    4

    • Sensitive (S, 0.1

  • 8/20/2019 nchrp_rrd_372

    5/13

    5

    sensitive (S, 0.1

  • 8/20/2019 nchrp_rrd_372

    6/13

    Table 1 Sensitivity of design inputs by distress type for new HMA pavements.

    Distress Input Category Hypersensitive (NSI > 5) Very Sensitive (1

  • 8/20/2019 nchrp_rrd_372

    7/13

    Distress Input Category Hypersensitive (NSI > 5) Very Sensitive (1

  • 8/20/2019 nchrp_rrd_372

    8/13

    Table 2 Sensitivity of design inputs by distress type for new JPCP.

    Distress Input Category Hypersensitive ( NSI> 5) Very Sensitive (1

  • 8/20/2019 nchrp_rrd_372

    9/13

    Distress Input Category Hypersensitive (NSI > 5) Very Sensitive (1

  • 8/20/2019 nchrp_rrd_372

    10/13

    10

    estimates of concrete stiffness and strengthgains with time can cause large errors in pre-dicted pavement distresses.

    • The sensitivity values for faulting, transversecracking, and IRI were similar in magnitude, but

    the range for faulting was significantly largerthan that for transverse cracking and IRI.

    • Transverse cracking is highly sensitive to thewidened slab edge support condition but notother edge support conditions (i.e., no edgesupport and tied shoulder edge support with80% LTE).

    Findings for CRCP

    The sensitivity of predicted performance to vari-ations in design inputs of CRCP varied by distresstype (e.g., faulting vs. crack width vs. crack LTE vs.IRI) but the results suggest some broad conclusions.Table 3 summarizes the level of sensitivity by mate-rial category (e.g., concrete/steel, base, and subgrade)for each distress. The findings were as follows:

    • Concrete and steel layer properties (concretethickness, strength and stiffness properties,reinforcing steel properties, concrete unitweight, coefficient of thermal expansion, sur-face shortwave absorptivity) were most con-sistently the highest sensitivity design inputs

    followed by the base and subgrade properties;traffic volume was also an important designinput.

    • The largest sensitivity values for CRCP weresubstantially larger than those for JPCP.

    • The magnitudes (mean and standard devia-tion) of the highest sensitivity values forpunchouts and crack width were substan-tially greater than the values for crack LTEand IRI.

    • The 20-year to 28-day ratio for modulus ofrupture was the third-ranked sensitive designinput for punchouts (and thus for IRI). Thisis of concern because the 20-year modulus ofrupture can only be estimated.

    General Remarks

    Consideration of the high sensitivity or criti-cal design inputs will depend on the specific design

    input. While some high sensitivity inputs can bespecified with a small tolerance (e.g., HMA thick-ness, concrete thickness, lane width, and steelproperties), other properties need to be measuredor estimated. For example, mix-specific labora-tory measurement of HMA dynamic modulus and

    Poisson’s ratio and concrete modulus of ruptureand modulus of elasticity may be appropriate forhigh-value projects. However, simple methods fordefining surface shortwave absorptivity for asphaltand concrete surfaces and thermal conductivityand heat capacity of stabilized bases are not read-ily available and realistic values for specific pavingmaterials have not been established. For these rea-sons, project-specific sensitivity studies to evaluatethe consequences of uncertain input values may beappropriate for some situations.

    FINAL REPORT

    The contract agency’s final report, “SensitivityEvaluation of MEPDG Performance Prediction,”gives a detailed account of the project, findings, andconclusions and includes further information on thesensitivity analysis methodology and the results ofthe different analyses. The report is available onlineat trb.org by searching NCHRP01-47_FR.pdf-.

    ACKNOWLEDGMENTS

    The work presented herein was performed underNCHRP Project 1-47 and was guided by NCHRPProject Panel C1-47, chaired by Dr. Kevin D. Hall,University of Arkansas-Fayetteville, with mem-bers Dr. Michael E. Ayers (formerly with AmericanConcrete Pavement Association); Dr. Imad Basheer,California DOT; Dr. Mohamed K. Elfino, VirginiaDOT; Ms. Laura Fenley, Wisconsin DOT; Mr. Geof-frey Hall, Maryland State Highway Administration;Mr. Kent Hansen, National Asphalt Pavement Asso-ciation; Dr. Tommy E. Nantung, Indiana DOT; and

    Mr. Richard Zamora, Colorado DOT. Mr. Tom Yu andMr. Stephen Maher provided liaison with the FHWAand TRB, respectively. Dr. Amir N. Hanna served asthe responsible staff officer. The final report was pre-pared by Dr. Charles W. Schwartz and Dr. Rui Li ofthe University of Maryland and Dr. Sung Hwan Kim,Dr. Kasthurirangan Gopalakrishnan, and Dr. HalilCeylan of Iowa State University.

    (text continued from page 5)

  • 8/20/2019 nchrp_rrd_372

    11/13

    Table 3 Sensitivity of design inputs by distress type for new CRCP.

    Distress Input Category Hypersensitive (NSI > 5) Very Sensitive (1

  • 8/20/2019 nchrp_rrd_372

    12/13

    Crack LTE PCC/Steel Properties 28-Day Modulus ofRupture (+2.4)Thickness (+1.6)28-Day Indirect TensileStrength (–1.3)Percent Steel (+1.0)

    Crack LTE Base Properties

    Subgrade Properties

    Other Properties Bar Diameter (–1.5)

    IRI PCC/Steel Properties Thickness (–9.0)28-Day Modulus of Rupture (–7.4)

    20-year to 28-day Modulusof Rupture (–3.5)Unit Weight (–3.2)Percent Steel (–3.0)28-Day Indirect TensileStrength (+2.4)28-Day Elastic Modulus (+2.1)Water-to-Cement Ratio (+1.9)Cement Content (+1.4)Coefficient of ThermalExpansion (+1.2)

    Base Properties Resilient Modulus (–1.2)

    Subgrade Properties Resilient Modulus (–1.2)

    Other Properties Bar Diameter (+1.9)Traffic Volume (+1.6)Steel Depth (+1.5)

    NSIµ±2σ values are given in parentheses.

    Table 3 (Continued)

    Distress Input Category Hypersensitive (NSI > 5) Very Sensitive (1

  • 8/20/2019 nchrp_rrd_372

    13/13

    Transportation Research Board

    500 Fifth Street, NW

    Washington, DC 20001

    These digests are issued in order to increase awareness of research results emanating from projects in the Cooperative Research Programs (CRP). Person

    wanting to pursue the project subject matter in greater depth should contact the CRP Staff, Transportation Research Board of the National Academies, 50

    Fifth Street, NW, Washington, DC 20001.

    COPYRIGHT INFORMATION

    Authors herein are responsible for the authenticity of their materials and for obtaining written permissions from publishers or persons who own the copyrig

    to any previously published or copyrighted material used herein.

    Cooperative Research Programs (CRP) grants permission to reproduce material in this publication for classroom and not-for-profit purposes. Permission

    given with the understanding that none of the material will be used to imply TRB, AASHTO, FAA, FHWA, FMCSA, FTA, or Transit Development Corporatio

    endorsement of a particular product, method, or practice. It is expected that those reproducing the material in this document for educational and not-for-pro

    uses will give appropriate acknowledgment of the source of any reprinted or reproduced material. For other uses of the material, request permission from CR

    Subscriber Categories: Materials • Pavements

    ISBN 978-0-309-25893-

    9 7 8 0 3 0 9 2 5 8 93 7

    9 0 0