Analytic Network Process for Local Scour Formulas Ranking with Parametric Sensitivity Analysis and Soil Class Clustering

Abstract Predicting scour depth around bridge piers during flood events was identified as one of the most important scientific interests of researchers dealing with local scour, wherein, for this purpose, a variety of empirical formulas were proposed. The redundancy and dissimilarity of these formulas turned the selection of the most appropriate one among them into a complicated task. This study focuses on developing a novel framework for comparing several formulas already available for local scour assessment. To achieve this purpose, 18 distinct formulas were selected and then tested using a large set of field data under three major particle size distributions that a streambed could comprise. The approval process is performed using the analytical network process, in which the conventional weights were revised with the help of several analyses, including statistical, sensitivity, and correlation. The findings of this study show that the effectiveness of a given formula varies depending on the class of soil studied, and the best performance is obtained for scouring formulas containing mainly geometrical parameters.