Computational Modeling and Simulations of Condition Deterioration to Enhance Asphalt Highway Pavement Design and Asset Management
Date of Award
Ph.D. in Engineering Science
A nation’s economy and prosperity depend on an efficient and safe transportation network for public mobility and freight transportation. A country’s road network is recognized as one of the largest public infrastructure assets. About 93 percent of 2.6 million miles of paved roads and highways in the United States (U.S.) are surfaced with asphalt. Longitudinal roughness, pavement cracking, potholes, and rutting are the major reasons for rehabilitation of asphalt roads. Billions of dollars are required annually for the maintenance and rehabilitation of road networks. If timely maintenance and rehabilitation are not performed, the pavement damages inflicted by heavy traffic repetitions and environmental impacts may lead to life threatening condition for road users. This research is focused on asphalt pavement condition deterioration progression modeling and computational simulations of uncracked and cracked asphalt pavement-subgrade models. The research objectives are to (1) evaluate and enhance asphalt pavement condition deterioration prediction models, (2) evaluate modulus backcalculation approaches for characterizing asphalt pavement layers of selected test sections, (3) develop three dimensional-finite element (3D-FE) asphalt pavement models and study impacts of cracking on pavement structural responses, and (4) implement pavement condition deterioration models for improved structural design and asset management of asphalt highway pavements. The historical asphalt pavement database records of the Long-Term Pavement Performance (LTPP) research program were used to develop asphalt pavement condition deterioration progression models, considering LTPP regions and maintenance and rehabilitation history. The enhanced condition deterioration prediction equations of the International Roughness Index (IRI), rutting, and cracking distresses were developed and evaluated in this research for LTPP datasets of 2,588 for IRI, 214 for rutting, and 2,240 for cracking. The LTPP regions and major maintenance intervention criteria were comfactors considered in all multiple regression equations. The IRI prediction equation also considered the IRI measurement location factor. Additionally, the rutting prediction equation includes additional factors of in situ modulus of pavement layers and base layer type. In comparison, the U.S. national mechanistic empirical pavement design guide (MEPDG) performance prediction models do not include maintenance and rehabilitation and climatic factors which present major limitations of the MEPDG method of pavement thickness design. Both regression analysis and Artificial Neural Network (ANN) analysis methods were used and the results were compared. The IRI multiple regression equation shows R of 0.633, which is slightly lower compared to the ANN IRI model’s R of 0.717. The IRI predictions using the enhanced multiple regression equation are comparable with the ANN results for verification data sets. The prediction equations from multiple regression modeling and ANN modeling of rutting distress show high R values above 0.93 and 0.94, respectively, and reasonably accurate result of model database and verification section. These model equations have got higher R value compared to the MEPDG’s R value. A new cracking model namely Unified Cracking Index (UCI) was developed in this research by combining all crack types which is not available in the MEPDG. The overall UCI combines the densities (% crack area per total area) of the alligator, block, longitudinal, and transverse cracking types. This approach is practical and easy to implement with intervention criteria of maintenance and rehabilitation for life-cycle asset management of asphalt highway pavements. The UCI equations using multiple regression for log transformation and using sigmoidal transformation for the model database shows the correlation, R, of 0.551 and 0.511 respectively, with 19.5 and 4.1 percent errors in predictions compared to the measured LTPP data. In comparison, the ANN model for UCI shosignificant improvements in R value (0.707) with 14.6% error. It also shohigh R value (0.861) and low error for the verification data sets. The MEPDG method includes separate models of alligator crack, longitudinal crack (defined as fatigue induced crack in the MEPDG), and transverse crack. In comparison, this research developed prediction equations not only for alligator, longitudinal, and transverse cracks but for block crack too. Individual ANN model of cracking (alligator, block, longitudinal, transverse) also shoreasonably accurate results. In situ modulus values of existing pavements are other important material inputs for pavement structural response analysis of overlay thickness design. Several modulus backcalculation software, based on the layer elastic static analysis theory, were evaluated in this research for selected LTPP highway sections. The comparisons indicated that the backcalculated modulus values in the LTPP database were generally unreasonable using the EVERCALC 5.0 software. Overall, the backcalculated modulus values using BAKFAA 2.0 and PEDD/UMPED were generally reasonable for all pavement layers. It was also shown that the thickness design of longer lasting pavement performance depends on seasonal layer modulus values considering extreme weather and climate attribute. In order to create a structural response database for pavement-subgrade subjected to design truck axle load, the 3D-FE models of uncracked and cracked asphalt pavement layer were developed using the LS-DYNA finite element software. The structural responses such as surface deflections, stresses and strains at different depths in the pavement-subgrade model were analyzed for critical locations. A full factorial experiment for six independent variables at two levels was designed, and the simulations for 64 treatment combinations were executed for the uncracked model. The results of the 3D-FE models shocomparable results with previous studies using the LS-DYNA software and the outputs of the GAMES linear elastic program. An extended analysis was conducted on the cracked model to study the effect of full depth cracked on effective viasphalt modulus values. Based on the full-depth cracked 3D-FE model results, low-level modulus of weak pavements shoa higher reduction of 81.0 % in the asphalt modulus compared to the compared to the asphalt modulus of the uncracked 3D-FE model, while the high-level modulus and thick pavement shoa low reduction of 13.5 % in the asphalt modulus of the uncracked pavement model. The development of the enhanced pavement condition prediction equations provide significant improvements over the MEPDG method, such as consideration of maintenance and rehabilitation history and climatic regions, using larger number of LTPP datasets, compared to model data sets used in the MEPDG. Therefore, the developed equations are more appropriate for the pavement structural design and asset management of asphalt highways. This implementation will contribute towards longer-lasting asphalt highway pavement assets to serve the public, improve safety, support efficient supply chain and economic growth.
Mohamed Jaafar, Zul Fahmi Bin, "Computational Modeling and Simulations of Condition Deterioration to Enhance Asphalt Highway Pavement Design and Asset Management" (2019). Electronic Theses and Dissertations. 1643.