scour depth prediction around bridge abutment protected by spur dike using soft computing tools and regression methods

Scour depth around bridge abutment is a crucial parameter to design the protective spur dike. costly and time-consuming experiments make it difficult to evaluate the scour depth in the problems involving scour phenomena. however, soft computing and regression methods may be applied based on the experimental results. in this paper, a set of experiments is performed and a database including 127 records is collected to evaluate the relation between scour depth and five independent variables including abutment length, flow discharge, flow depth, spur dike length and spur dike distance from abutment to upstream. this paper presents a new application of the multi-layer perceptron neural network (mlp), group method of data handling (gmdh), non-linear regression (nlr) and multiple linear regression (mlr) to predict the scour depth. a sensitivity analysis is also performed to evaluate the influence of each variable on the scour depth. results indicate that the first three methods are efficient and accurate enough to be applied in practical applications with determination coefficient (r2) above 90%, while, the mlr has shown a poor performance in this paper. it is observed that mlp and gmdh outperform other methods based on the test data. however, explicit equation derived by nlr has a major advantage to be applied in the field applications without skilled operators and computer packages.