Predicting pavement performance under traffic loading using genetic algorithms and artificial neural networks to obtain resilient modulus values

In many developing countries, the lack of high performance equipment and/or lack of knowledge on how to operate these specialized equipment obligates pavement engineers to continue using pavement design methodologies based on empirical tests that do not represent the dynamic nature of repeated traffic loads nor simulate the actual conditions of a pavement structure under loading in the field. For the realistic prediction of pavement performance it is very important to accurately characterize the mechanical behavior of unbound material layers and subgrade soils. In pavement analysis using elastic layered theory, material properties in terms of dynamic elastic modulus and Poisson's ratio are the major input parameters. The dynamic elastic modulus of pavement materials or resilient modulus (MR) is measured by conducting repeated load triaxial tests typically not available to highway authorities in developing countries.;A material model for the reliable prediction of the dynamic elastic moduli of the various materials used in pavement design to describe the performance of soil layers subjected to repeated wheel loads in a multilayered pavement system is developed in this study. This material model provides MR design values in a practical and low-cost manner as predictions are made from commonly performed laboratory tests in developing countries. The model is capable of accounting for the effects on performance of different soil types, the effect of environmental conditions such as moisture content and degree of saturation, as well as different levels of confining stress and wheel loads. A new material model that accommodates new data sets for both granular (cohesionless) and cohesive materials by readily incorporating them into the predictive model is presented. An ability of the proposed model to use small data sets and reduce the bias towards predominant data sets is demonstrated. The proposed model is also able to account for incomplete sets of input parameters.;Pavement performance under traffic loading is predicted by implementing a pavement response model that uses genetic algorithms and artificial neural networks to determine the mechanical behavior of the different layers in a multilayered pavement system. The combination of the material model with pavement response model offers an efficient, reliable, and low-cost methodology for the design and analysis of pavements that is more easily accessible to practice. Given the importance of the proposed methodology for developing countries, its applicability beyond the geographical boundaries of the U.S. is accomplished by including soil samples from tropical countries in Central and South America. Mechanistic design information is provided to transportation agencies in developing countries with much different design cultures.