An Efficient and Accurate Genetic Algorithm for Backcalculation of Flexible Pavement Layer Moduli (SJN: 134481)
Researcher: Ernie Pan – University of Akron
Mechanical properties of pavement materials have always played a primary role in pavement engineering. They are of importance in designing new pavement, estimating pavement remaining life, and helping decision-makers select the optimal reconstruction and/or rehabilitation strategy. Moreover, in the newly released Mechanistic Empirical Pavement Design Guide (MEPDG), material properties, i.e. modulus and Poisson’s ratio, form the basic inputs in the forward calculation of pavement responses, i.e. displacement, strain and stress. Therefore, it is obvious that the accuracy of modulus input will directly influence the pavement performance, and thus the safety of travelers on the highway. Backcalculation has been a challenging and yet urgent topic and has gained much attention from engineers and researchers. Conventionally, various backcalculation programs were proposed. However, as is well known, the solutions from most programs are not unique due to certain intrinsic drawbacks in their programs where either the search algorithms are not universal or the fitness function is not suitably defined. Consequently, there are many sets of pavement moduli satisfying the same level of fitness; however, most of them are local optimization except for one genuine solution. These local optimal solutions often mislead pavement engineers in their daily design activities. The genetic algorithm (GA) has the unique advantage of global searching and has been gaining more and more attention in backcalculation. However, the backcalculation program based on GA needs tens of thousands of computation trials before achieving satisfactory results, which usually implies unbearable running time. In backcalculation, every computation trial is a forward calculation, thus the speed of forward calculation plays a decisive role in the speed of a backcalculation program. Because of the large amount of computation tasks, the backcalculation programs based on GA are currently still limited to theoretical investigation. Furthermore, most available programs are only able to analyze pavement with very few homogeneous layers, e.g. four layers in MODULUS and five layers in BACKGA-ANN. However, it is well-known that asphalt material is very sensitive to time and temperature, and the aging- and temperature-related functional graded moduli (FGM) need to be included in asphalt concrete. Such stiffness gradient will consequently affect pavement responses significantly, thus functionally graded stiffness should be considered in backcalulation.
The objective of this research is to investigate how the stiffness gradient (FGM) affects backcalculation of flexible pavement. Stiffness gradient (FGM) profiles of pavements have been extensively reported in pavement engineering; however, no work has been conducted to investigate the influence of the FGM on pavement backcalculation.