Evaluating structural conditions of existing, in-service pavements is a part of the routine maintenance and rehabilitation activities undertaken by the most departments of transportation (DOTs). In the field, the pavement deflection profiles (or basins) gathered from the nondestructive falling weight deflectometer (FWD) test data are typically used to evaluate pavement structural conditions. Over the past decade, interest has increased in a new class of computational intelligence system, known as artificial neural networks (ANNs), for use in geomechanical and pavement systems applications. This report describes the development and use of ANN models as pavement structural analysis tools for the rapid and accurate prediction of layer parameters of Iowa pavements subjected to typical highway loadings. ANN models trained with the results from the structural analysis program solutions have been found to be practical alternatives. The ILLI-PAVE, ISLAB2000, and DIPLOMAT programs were used as the structural response models for solving the deflection parameters of flexible, rigid, and composite pavements, respectively. The trained ANN models in this study were capable of predicting pavement layer moduli and critical pavement responses from FWD deflection basins with low errors.

The developed methodology was successfully verified using results from long-term pavement performance (LTPP) FWD tests, as well as Iowa DOT FWD field data. All successfully developed ANN models were incorporated into a Microsoft Excel spreadsheet-based backcalculation software toolbox with a user-friendly interface. The final outcome of this study was a field-validated, nondestructive pavement evaluation toolbox that will be used to assess pavement condition, estimate remaining pavement life, and eventually help assess pavement rehabilitation strategies by the Iowa DOT pavement management team.