Investigation of the Accuracy and Applicability of Genetic Programming for Estimating Scour Depth Downstream of a Spillway

Abstract

Scouring is an important issue in the safety of hydraulic structures such as downstream spillways, dams, downstream sluice gates, bridge foundations, and pipelines. The scouring phenomenon downstream of composite spillways is one of the important issues in the design and safety assessment of hydraulic structures. Spillways and sluice gates are hydraulic structures that can control water levels and measure flow rates. For this reason, extensive studies have been conducted on these two structures, but these structures also have disadvantages. By combining these two structures, a combined spillway-sluice gate structure can be proposed to overcome these disadvantages. The scouring problem downstream of these structures in erodible beds is an important issue. If not controlled, it may cause instability of the structure and ultimately its destruction. For this purpose, many laboratories have studied the effect of various variables on the scour depth of these structures. Various models such as neural networks (ANN), group data hierarchy (GMDH) and regression equations have predicted the scour depth in hydraulically sensitive structures. In this study, the accuracy and application of genetic programming (GP) in the downstream of the dam spillway were investigated. From the studies conducted, it was concluded that the GP model when used to estimate the scour depth downstream of the spillway has a coefficient of determination (R^2 = 0.977) and also when this GMDH model is trained with genetic programming, the result has a higher coefficient of determination.