Assessing heuristic models through k-fold testing approach for estimating scour characteristics in environmental friendly structures

Abstract

Scouring downstream of grade control structure is a common problem. W-weir structures are recommended for controlling, reducing soil erosion, and development aquatic organisms habitat. W-weir structures can provide energy dissipation, increased aquatic habitat, grade-control, and bank and bed stabilization through the diversion of conveyance with the river-channel center. The current study aimed at analyzing the scour geometry downstream of W-weirs in the sinuous channels using heuristic gene expression programming (GEP) and support vector machine (SVM) models. A comparison was also made between the GEP and SVM with traditional multi variable linear regression (MLR) technique. The cross-validation technique (i.e. k-fold testing) was used to assess the developed models. Observations revealed that the scour typology affects the maximum scour depth value. The obtained results showed that both the GEP and SVM models give promising results in simulating the scour depth and scour length magnitudes. Assessing the models confirmed that using the k-fold testing cross validation is very necessary as all the available patterns can participate in training and testing stages, which reduces the risk of over fitting.

Keywords:

• Cross validation
• gene expression programming
• scour
• support vector machine