Genetic functions-based modelling for pier scour depth prediction in coarse bed streams

Bridge pier scour in coarse bed streams is a significant problem that needs to be addressed in bridge design. To accurately predict the scouring process by means of inductive modelling, extensive studies have been conducted using both laboratory and on-site field data. Several techniques have been used previously, including conventional regression models as well as more complex models such as those based on artificial intelligence techniques. In this research, a relatively new technique based on genetic algorithms, named genetic functions (GF), was investigated to predict pier scour depth in coarse bed streams. The primary motivation was to obtain a relatively simple and compact explicit functional approximation for pier scour depth prediction in coarse bed streams using on-site measurements. The data used in the model development contained a total of 125 on-site measurements from coarse bed streams. The performance of the GF-based technique was compared with other empirical models based on regression artificial neural networks and gene expression programming. The performance of the GF-based model was found to be highly encouraging in predicting bridge pier scour depth. GF have the added advantage of providing a relatively simple, easy to use and explicit functional expression for pier scour depth.