International Journal of Pavement Engineering · 2005 · 95 citations · 15 references
EngineeringCivil EngineeringComputational EngineeringFinite Element AnalysisMachine LearningPavement EngineeringAsphalt MaterialsHighway PavementsNonlinear Structural AnalysisPavement Structural AnalysisStress-strain BehaviorStructural Response AnalysisBackcalculation Of Layer ModuliElastic Layered TheoryElastic ModuliFlexible Pavement AnalysisFlexible PavementsPavement Response ModelingSofteningStress Distribution In Pavement Layers
The study uses artificial neural networks to rapidly and accurately predict critical responses and deflection profiles of full‑depth flexible pavements under typical highway loadings. The authors employed the ILLI‑PAVE finite‑element model, incorporating nonlinear stress‑dependent subgrade soil behavior, and trained artificial neural networks on its solutions to predict pavement responses. The trained ANN models accurately predicted asphalt tensile strains, subgrade stresses, and pavement layer moduli from falling‑weight deflectometer data, achieving low average absolute errors compared to ILLI‑PAVE analyses. Keywords: artificial neural networks, falling‑weight deflectometer, full‑depth flexible pavements, nonlinear stress‑dependent subgrade response.
Abstract This paper describes the use of artificial neural networks (ANNs) as pavement structural analysis tools for the rapid and accurate prediction of critical responses and deflection profiles of full-depth flexible pavements subjected to typical highway loadings. The ILLI-PAVE finite element program, extensively tested and validated for over three decades, was used as an advanced structural model for solving the critical responses of full-depth flexible pavements. Unlike the linear elastic layered theory commonly used in pavement layer backcalculation, nonlinear stress-dependent subgrade soil response models were used in the ILLI-PAVE program to account for the softening nature of fine-grained subgrade soils under increasing stress states. ANN models then trained with the results from the ILLI-PAVE solutions have been found to be viable alternatives. The trained forward calculation ANN models were capable of predicting asphalt tensile strains and subgrade compressive stresses/strains with low average absolute errors of those obtained directly from the ILLI-PAVE analyses. ANN backcalculation models developed in this study were also capable of successfully predicting the pavement layer moduli from the falling weight deflectometer (FWD) deflection basins and they may be used in the field for rapidly assessing the condition of pavement sections during the FWD testing. Keywords: Artificial neural networksFalling weight deflectometerFull-depth flexible pavementsNonlinear, stress-dependent subgrade response Notes ¶Email: alper@iastate.edu §Email: tutumlue@uiuc.edu ∥Email: mrthomps@uiuc.edu Additional informationNotes on contributorsAlper Guclu ¶ ¶Email: alper@iastate.edu Erol Tutumluer § §Email: tutumlue@uiuc.edu Marshall R. Thompson ∥ ∥Email: mrthomps@uiuc.edu
15