Application of Meta-Heuristic Models for Local Scour Evaluation

Abstract

One of the most important factors in bridge destruction is scour around piers of bridges. The scour depth evaluation around the bridge pier is very critical in bridge design. Although there are several empirical formulas to predict depth scour, these formulas usually have not had accurate results. Therefore, there has been an increase in use of soft computing tools such as neural networks (ANNs), fuzzy inference system (FIS) and adaptive network fuzzy inference system (ANFIS) for this purpose. Model trees are one of the prevailing data mining tools. In the present study, for prediction of scour depth, model trees (M5′ algorithm) and regression trees (CART algorithm) with using experimental data of scour measurement in clear water condition were employed. Moreover, these models were used for two cases; the first one with original (dimensional) data and the second one with non‐dimensional data and the results were compared with the results of six empirical formulas. The results showed that model and regression trees are more efficient than the empirical formulas to predict scour depth.