SENSITIVITY ANALYSIS OF PAVEMENT PERFORMANCE PREDICTED USING THE M-E PDG

  • Ms Word Format
  • 70 Pages
  • ₦3000
  • 1-5 Chapters

SENSITIVITY ANALYSIS OF PAVEMENT PERFORMANCE PREDICTED USING THE M-E PDGABSTRACT The Mechanistic-Empirical Pavement Design Guide (M-E PDG) was developed by the National Cooperative Highway Research Program to improve the design and performance of the national highway system. The design procedure used in the guide is a major improvement over the previous design methods in terms of using parameters related to construction and operation of a pavement to predict the performance over a specified design life. These parameters must be entered in the M-E PDG software as input, which requires the user to gather a large amount of data. Therefore, sensitivity study of the input parameters on a pavement’s predicted performance is necessary for standardizing data collection and documentation by a pavement design agency, as well as improving pavement design with efficient utilization of resources to maximize performance. Evaluation of currently followed design procedures by the New England state agencies, evolution of a data collection methodology for Level 2 and 3 design and study of sensitivity of the collected data conforming to specified tolerances are conducted as part of the research presented in this thesis.                                             Table of Contents List of Figures ……………………………………………………………………………………………………………………..vi List of Tables………………………………………………………………………………………………………………………vii Acknowledgement………………………………………………………………………………………………………………..ix

1. Introduction…………………………………………………………………………………………………………1

1.1 Objective…………………………………………………………………………………………………………………….5 1.2 Research Significance…………………………………………………………………………………………………..6 1.3 Research Tasks…………………………………………………………………………………………………………….7

2. Literature Review…………………………………………………………………………………………………8

2.1 Critical Input Parameters that Affect Flexible Pavement Design………………………………………..8 2.2 Findings from Completed Research Activities on M-E PDG Implementation…………………….13 2.2.1 M-E PDG Implementation – Indiana Study………………………………………………………………………….14

3. Research Methodology………………………………………………………………………………………..19

3.1 Data Collection, Analysis and Input Data Generation……………………………………………………..19 3.2 Tolerances from State Design Specifications Documents ………………………………………………..20 3.2.1 HMA Gradation ……………………………………………………………………………………………………………….20 3.2.2 HMA Mix Stiffness…………………………………………………………………………………………………………..21 3.2.3 Subgrade / Base Resilient Modulus……………………………………………………………………………………..21 3.2.4 HMA Thickness ……………………………………………………………………………………………………………….22 3.2.5 Binder Content …………………………………………………………………………………………………………………22 3.2.6 Air Voids Percentage ………………………………………………………………………………………………………..23 3.2.7 Binder Type……………………………………………………………………………………………………………………..23 3.2 Input Value Selection for M-E PDG Runs……………………………………………………………………..24 3.2.1 Traffic Inputs……………………………………………………………………………………………………………………25 3.2.2 Climate Inputs………………………………………………………………………………………………………………….26 3.2.3 Asphalt Material Inputs……………………………………………………………………………………………………..27 3.2.4 Base Course and Subgrade Inputs……………………………………………………………………………………….28

4. Data Analysis – Statistical Approach…………………………………………………………………..30

4.1 Sensitivity Analysis – Version 1.0 Results…………………………………………………………………….33 4.1.1 Statistical Analysis of Data – New Hampshire ……………………………………………………………………..34 4.1.2 Statistical Analysis of Data – Connecticut……………………………………………………………………………40 4.1.3 Statistical Analysis of Results – Maine…………………………………………………………………………………45 4.2 Sensitivity Analysis – Thermal Cracking ………………………………………………………………………50 4.2.1 Effect of Asphalt Concrete Layer Thickness…………………………………………………………………………51 4.2.2 Effect of Air Void Content…………………………………………………………………………………………………51 4.2.3 Effect of Coefficient of Thermal Contraction and Average Tensile Strength…………………………….52 4.2.4 Effect of Climate………………………………………………………………………………………………………………55 4.2.5 PG Binder Grade………………………………………………………………………………………………………………56 4.3 Response Surface Methodology – Interaction Effects Study…………………………………………….58

5. Results and Discussion………………………………………………………………………………………..66

5.1 General Pavement Design Inputs………………………………………………………………………………….66 5.2 Traffic Inputs …………………………………………………………………………………………………………….67 5.3 Climate Inputs……………………………………………………………………………………………………………70 5.4 Pavement Layer Inputs ……………………………………………………………………………………………….71 5.5 Material Inputs – Asphalt Material Inputs……………………………………………………………………..74 5.6 Material Properties – Unbound Layer Inputs………………………………………………………………….77 5.7 Normalization of Distresses – Comparison between States………………………………………………77

6. Conclusions………………………………………………………………………………………………………..86

6.1 State – Specific Recommendations for Tolerances on Material Properties…………………………88 6.2 Recommendations for Future Work………………………………………………………………………………89 References………………………………………………………………………………………………………………………91 Appendix A Subgrade Resilient Modulus……………………………………………………………………………….95 Appendix B M-E PDG Runs – Implementation Plan………………………………………………………………..96 Appendix C Input Values and Factor Levels for New Hampshire – Level 3………………………………..99 Appendix D Results of M-E PDG Runs………………………………………………………………………………..100 Appendix E Thermal Cracking Results – Level 3…………………………………………………………………..113 Appendix F General Linear Model Layout for New Hampshire Level 3 Data……………………………115 Appendix G Response Surface Methodology – Experiment Design, Results and Contour Plots…..116  

1. Introduction

In 1996, the National Cooperative Highway Research Program (NCHRP) launched Project 1-37A to develop a new design guide for pavement structures. The design guide recommended by the project team in 2004 is based on mechanistic-empirical (M-E) principles. Pavement performance is determined by the prevailing traffic and environmental conditions, pavement structure and material properties. With increasing traffic volume on highways, development of new material specifications and need for more reliable performance predictions, a correspondingly efficient pavement design methodology is required.   The Mechanistic Empirical Pavement Design Guide (abbreviated M-E PDG) is therefore considered by the Federal Highway Administration (1) an important factor for improving the national highway system. The M-E PDG uses performance prediction models to predict pavement performance over a specified design life. The design guide can therefore be used to design pavements such that their performance is maximized over the design life. The FHWA has developed a Design Guide Implementation Team (DGIT) whose mission is “To raise awareness, assist, and support State Highway Agencies and their industry partners in the development and implementation of the new mechanistic-empirical Design Guide”. A Lead States Group has also been established and includes: Arizona, California, Florida, Kentucky, Maryland, Minnesota, Mississippi, Missouri, Montana, New Jersey, New Mexico, Pennsylvania, Texas, Utah, Virginia, Washington, and Wisconsin. These states are actively pursuing implementation of the M-E PDG, have obtained upper management support of the process and have documented research findings from their studies on the M-E PDG (2).   The M-E PDG is a significant improvement over the previous AASHTO design guides for pavement design. The M-E design guide is built into the form of software that is capable of using various parameters involved in pavement design as input data and predicting the performance of the pavement over a specified design life period. Briefly stated, the M-E performance prediction model consists of four sub-models: the environmental effects model, pavement response model, material characterization model, and performance prediction model (Figure 1). The model is termed mechanistic due to the mechanistic calculation of stresses, strains, and deflections of a pavement structure, which are the fundamental pavement responses under repeated traffic loading and environmental conditions. The empirical component of the M-E PDG relates the pavement responses to field distresses and performance using existing empirical relationships, widely known as transfer functions (3). The design process is an iterative procedure that starts with a trial design and ends when predicted distresses meet the acceptable limits based on desired level of statistical reliability.   Figure 1 Mechanistic Empirical Design Procedure incorporated in the M-E PDG The predictive equations used by the M-E PDG for predicting the performance of flexible pavements are enlisted in Table 1 (4). Table 1 M-E PDG Response Prediction Equations and Transfer Functions

Distress Type Equations Terms in equation
Fatigue Cracking Asphalt Institute Model Fatigue cracking predictive equation 3.9492                                                        1.281  = 0.00432*k *C1     1 f                                               1       εt   E  f =Number of repetitions to fatigue cracking εt =tensile strain at critical location E=stiffness of material k1=correction for asphalt layer thickness C=laboratory to field adjustment factor
Bottom-Up Cracking Bottom-up cracking transfer function (measured in % of total lane area) FCBottom = C1 *C1′ +6000C2 *C2′ *log10 (D*100)) * 601 1+e   FCBottom =Bottom-up cracking (%) D=bottom-up fatigue damage C1, C1’, C2, C2’=calibration factors

 

Longitudinal (Top-Down Fatigue) Cracking Top-down cracking transfer function (measured in feet per lane mile) FCTop =              (7.0−10003.5*log10 (D*100)) *10.56 1+ e                       FCTop =Top down cracking (feet/mile) D=top-down fatigue damage
Permanent Deformation in Unbound Materials Tseng-Lytton Model Permanent deformation of the layer/sublayer in inches 1εε0r e− ρβευh δa (    ) = β  δa (     )=Permanent deformation of layer (in) N=number of traffic repetitions ε0 ,β,ρ=material properties ευ=Average resilient strain from primary response model εr =Resilient strain imposed in laboratory test to obtain above properties h=thickness of layer/sublayer β1 =Calibration factor Subgrade: β1 =1.35 Granular Base: β1 =1.673
Permanent Deformation in Asphalt Layer NCHRP 9-19 Superpave Model Permanent deformation of the asphalt layer (measured in inches) εp = k1 *10−3..4488T 1.5606 0.479244 εr εp =Accumulated plastic strain at N load repetitions εr =Resilient strain of asphalt as a function of mix properties, temperature and loading time N=number of load repetitions T=temperature (0F)
Thermal Cracking NCHRP 9-19 Superpave Models – TCMODEL Thermal crack length transfer function (measured in feet per mile lane) C f = β1 *    logC hac   σ  C f =Observed amount of thermal cracking β1 =Regression coefficient determined through field determination N(z)=standard normal distribution evaluated at z σ=standard deviation of log of crack depth in pavement C=crack depth hac =Thickness of asphalt layer

  The models used by the M-E PDG for predicting roughness of the pavement (IRI) are a function of the base type, and are developed using key pavement distresses mentioned above. The development and calibration of the IRI prediction models is discussed in detail in the M-E PDG (5). The inputs required by the M-E PDG software for pavement performance prediction are classified into four categories:

  • Traffic inputs
  • Climatic inputs
  • Pavement structure inputs
  • Material property inputs

  Inputs are classified according to a hierarchy system where the designer can select the level of advanced material properties and data accuracy based on the economic impact of the project. The selection is also a function of the state-of-knowledge and availability of the data. A summary of the hierarchal design input levels is given below (4):

  • Level 3 represents the lowest level of the hierarchy system and provides the lowest level of reliability; the inputs consist of default or user-selected values obtained from national and regional experiences such as LTTP sites.
  • Level 2 represents a higher level in the hierarchy system and provides more reliability than Level 3. Design inputs are based on laboratory test data and/or default predictive equations. This level is expected to be used on pavement design projects of higher significance.
  • Level 1 represents the highest level in the hierarchy system and provides the highest degree of reliability. Design inputs are generally site specific and are determined from material testing and/or in-situ measurement.

  The current distress prediction models incorporated in the M-E Pavement Design Guide were calibrated and validated using field performance of selected pavement sections throughout the United States, among which are numerous Long Term Pavement Performance (LTPP) test sites. Thus, the coefficients incorporated in the prediction models can be regarded as national averages derived from the performance measured from the sites selected for the calibration. The M-E PDG also provides state agencies the flexibility to adjust the model calibration coefficients derived from local calibration conducted on performance data of pavements in the state to improve the accuracy of prediction.   Detailed explanation of the M-E PDG software and its components is available in reports and published works on research conducted for developing implementation plans for various states in the US (6, 7, 8 and 9). An explanation of the design procedures in the AASHTO design guides as well as the M-E PDG is available in published work (7, 8 and 9).   State highway agencies (SHA) require a well-defined implementation strategy to make a transition from their current design practices to the M-E PDG. This is necessitated by the requirement of data pertaining to a large number of parameters that are needed for pavement design using the M-E PDG. Different SHA have different data collection and storage procedures based on the availability of resources such as instrumentation, level of expertise and data required for their current pavement design method. An integral part of developing an implementation strategy is the knowledge of input parameters required by the M-E PDG and the significance of their effect on prediction of a pavement distress type. The data for the research work presented in this thesis is derived from the project NETC 06- 1 – “New England Verification of NCHRP 1-37A Mechanistic – Empirical Pavement Design Guide with Level 2 and 3 Inputs”. The main objective of the project is to provide the New England states, namely New Hampshire, Connecticut, Maine and Rhode Island with an implementation strategy for smooth transition to the M-E PDG in terms of adequacy of data collection techniques and existing design specifications for the design of flexible pavements and AC overlays.

1.1 Objective

The objective of the research is to obtain a statistically well-defined relationship that explains the sensitivity of obtained performance prediction values to variation in input parameter values. The research focuses on mechanistic-empirical design of flexible pavements and asphalt concrete (AC) overlays. M-E PDG design can be classified into three hierarchal levels based on the accuracy of the input data and reliability of predicted pavement distresses. The study aims at studying sensitivity of input variables whose values are entered at Level 2 and 3 of design into the design guide.   The design guide requires data for a large number of variables as input to determine pavement responses and predict pavement distresses such as longitudinal and thermal cracking, rutting in addition to roughness (IRI) using performance prediction models. A value needs to be entered for each input variable. Input values can vary within a permissible range according to input standards built into the guide. Sensitivity study for the entire allowable range of values is a redundant process as most values are not implemented in design by agencies. Therefore, an initial identification of the critical variables which affect pavement performance and the ranges within which they can be varied are selected. A valid statistical representation of the sensitivity of predicted distresses to variability in inputs is then necessary to determine the level at which data collection should be implemented by state agencies for collecting data on a particular input variable.

1.2 Research Significance

A study of the sensitivity of the predicted distresses to the critical input variables is necessary to assess the relative significance of each variable. Sensitivity of the predicted distresses to change in value of the input parameters must be analyzed such that the design inputs can be adjusted to maximize pavement performance. M-E PDG input variables can be either quantitative variables, whose values are entered as numbers or qualitative or categorical variables whose values are non-numeric. Furthermore, the quantitative variables can be continuous or discrete. Therefore, the application of statistical models to the data is a challenging task which requires careful selection of values at which each input parameter must be entered for studying its sensitivity. The sensitivity study is conducted for flexible pavements and asphalt concrete overlays, but the methodology described can be applied to rigid pavement design and pavement rehabilitation as well.   Sensitivity studies conducted so far on M-E PDG prediction data have not established a valid statistical model so that the significance of the effect of an input variable can be ‘definitely’ quantified. Previous research involved in evaluating the effect of variation of a single parameter (6, 7, 8, 9, 10 and 11). However, pavement performance, which is a function of many input variables, depends on the combined influence of the values of the entire set of variables. The research conducted herein is an attempt to statistically analyze M-E PDG prediction data, as well as study the interaction effects of different variables. Different alternatives to study interaction effect of input variables are analyzed and a response surface methodology is selected to design the experiment for the interaction effect study.

1.3 Research Tasks

The tasks performed to conduct the research are briefly stated below:

  1. Review literature on similar attempts to study sensitivity of input variables on predicted pavement performance
  2. Identify input variables that are critical to each type of pavement distress that is predicted by the M-E PDG, and their corresponding input variables in the software
  3. Develop an initial set of input variables for a selected pavement section for each state – a Long Term Pavement Performance (LTPP) section is identified and all input variables that represent existing condition and characteristics of the highway are obtained
  4. Obtain various parameter tolerances from the state highway agencies’ design specification manuals and determine the range within which each input variable can be varied using the tolerance limits
  5. Conduct runs on the M-E PDG with each set of input variables and compile the predicted distress data
  6. Conduct a graphical and statistical analysis on the compiled data to determine the level of sensitivity of each input variable on pavement distress
  7. Evaluate adequacy of data collection and parameter tolerances, and suggest sources of information and data extraction procedures for use as input data for M-E PDG

SENSITIVITY ANALYSIS OF PAVEMENT PERFORMANCE PREDICTED USING THE M-E PDG    

0 Shares:
Leave a Reply

Your email address will not be published. Required fields are marked *

You May Also Like