Bayesian calibration of ILLI-THERM for temperature prediction within airfield concrete pavement

Mechanistic models (e.g., finite-element analysis, discrete-element analysis) are commonly used techniques to understand the behavior of materials. Although these models have been successfully utilized to simulate engineering problems, their outputs may not match the field data because of inherent uncertainties. One method to improve the accuracy of mechanistic models is calibration. In this paper, Bayesian calibration framework is applied to the mechanistic model ILLI-THERM, which uses the Enhanced Integrated Climatic Model to predict temperature profile within a pavement. The temperature data used for the calibration were collected by the Federal Aviation Administration (FAA) for a concrete taxiway at John F. Kennedy airport in New York, US. The primary objectives of this project were to study curling in concrete slabs (wet-freeze climatic region) and measure total strain and load-induced strain in slabs under multi-gear aircraft such as A-380, B-777, B-747. This paper focuses on material-related uncertainties. Hence, ILLI-THERM was calibrated for material characterization parameters only. Adaptive Monte Carlo Markov Chain was used to generate samples for estimating posterior distributions. The calibration resulted in improving the prediction performance of ILLI-THERM; root mean square error (RMSE) is reduced by 19%. Furthermore, the posterior distribution resultant from the Bayesian calibration indicated the significance of each material properties on temperature distribution within the pavement.