New stochastic modeling strategy on the prediction enhancement of pier scour depth in cohesive bed materials
Abstract Scouring around the piers, especially in cohesive bed materials, is a fully stochastic phenomenon and a reliable prediction of scour depth is still a challenging concern for bridge designers. This study introduces a new stochastic model based on the integration of Group Method of Data Handling (GMDH) and Generalized Likelihood Uncertainty Estimation (GLUE) to predict scour depth around piers in cohesive soils. The GLUE approach is developed to estimate the related parameters whereas the GMDH model is used for the prediction target. To assess the adequacy of the GMDH-GLUE model, the conventional GMDH and genetic programming (GP) models are also developed for evaluation. Several statistical performance indicators are computed over both the training and testing phases for the prediction accuracy validation. Based on the attained numerical indicators, the proposed GMDH-GLUE model revealed better predictability performance of pier scour depth against the benchmark models as well as several gathered literature studies. To provide an informative comparison among the proposed techniques (i.e. GMDH-GLUE, GMDH, and GP models), an improvement index () is employed. Results indicated that the GMDH-GLUE model achieved = 6% and = 3%, demonstrating satisfying performance improvement in comparison with the previously proposed GMDH model.