Efficient response models for rigid airfield pavement systems design

Rezaei Tarahomi, Adel
Major Professor
Halil Ceylan
Committee Member
Journal Title
Journal ISSN
Volume Title
Research Projects
Organizational Units
Journal Issue
Civil, Construction, and Environmental Engineering

The Federal Aviation Administration (FAA) has recognized that its current rigid pavement design model, reflecting a single slab loaded at one edge by a single aircraft landing gear, does not adequately account for top-down cracking, meaning that one of the major observed failure modes for rigid pavements is poorly accounted for in the rigid pavement design procedure FAA Rigid and Flexible Iterative Elastic Layer Design (FAARFIELD). To expand the FAARFIELD design model beyond the currently-used reduced one-slab model, since practical alternatives to running the 3D-FEM stress computation as client software are needed, this study seeks to fill this research gap by developing a surrogate computational response model or procedure (suitable for implementation in FAARFIELD 1.4-TDC) that returns a close estimate of the top-down bending stress computed by the 3D-FE model for combined vehicle and temperature loading of rigid airport pavements.

A synthetic database has been generated by conducting batch runs of FAA finite-element analysis software (FEAFAA 2.0), and this database contains data from thousands of multiple-slab rigid pavement cases with associated critical tensile stresses at the slab top induced by either mechanical-only or combined mechanical and temperature loading, with critical responses that include tensile stresses in both x and y directions along with principal tensile stresses. Artificial neural networks (ANNs) have been employed to develop a surrogate top-down slab bending-stress prediction model using non-linear input-output mapping of the database. Surrogate response models were trained for each of the Airbus and Boeing aircraft provided in the FEAFAA 2.0 library. This has been accomplished by developing software for automating entry of the database obtained by conducting FEAFAA batch runs, to train the ANN models using different architectures and algorithms, to control the ANN input parameters, and to collect training results. Both accuracy and robustness of the models were validated through independent testing and sensitivity testing. A new ANN tool for rapid analysis of nine-slab rigid airfield pavements that replicates the top-down critical stresses obtained from direct finite element solutions was developed. In addition, a new airfield pavement design approach was proposed which employs Bayesian optimization along with the trained ANN models.