Prediction of scour at abutments using piecewise regression

Scour is one of the main causes of bridge failure and scour depth prediction is thus a very important issue. Although many studies have been carried out to develop formulas for the prediction of scour depth at abutments, the existing formulas do not yield accurate results, mainly due to the complexity of the scour phenomena around abutments. This paper reports on a study of abutment scour using piecewise multiple linear regression analysis. First, scour depth data were collected from the literature and the existing formulas were evaluated and compared. To develop new formulas, the conventional governing parameters were selected as the independent variables. The piecewise multiple linear regression technique was applied to investigate the relationship between dimensionless scour depth and the governing dimensionless parameters in several subdomains. Errors of the proposed formula and previous models were also compared. Compared with existing formulas, the error measures in the proposed method show a significant improvement in prediction accuracy.