Pages: 5-11

Abstract

In this study, using self-organizing systems of Group method of data handling (GMDH), the highest depth of scour around bridge piers in bed of sticky soils is estimated. GMDH network is developed by back propagation algorithm. To train GMDH algorithm with the above-mentioned method, quadratic polynomial is used in the structure of GMDH. Parameters affecting scour depth include the initial moisture content of cohesive sediments bed, clay percent of the sediments, flow Froude number, and shear strength of the sediments of the bed. These parameters are also specified as algorithm input variables. Assessment of GMDH gives accurate approximation of scouring in the bed of cohesive soils compared with empirical relationships based on regression model evaluation.