Prediction of head loss on cascade weir using ANN and SVM

Abstract

In this study, the support vector machine (SVM) technique was applied to predict the head loss of flow on the cascade weirs. To this end, related data-set was collected in the literature. To compare the performance of SVM with other type of soft computing techniques, the multilayer perceptron neural network as common type of artificial neural network models was developed. To derive the most effective parameters on mechanism of head loss, sensitivity analysis was carried out on the both applied models. Results indicated that performance of SVM with coefficient of determination (R² = 0.98) and root mean square error (RMSE = 2.61) is suitable for predicting the head loss of energy and in comparison with the MLP performance, the accuracy of SVM is a bit more accurate. During the preparation of SVM model, it was found that the radial basis function as kernel function had satisfactory performance. The sensitivity analysis declared that the drop number, number of steps, and ratio of the critical flow depth to the height of steps are the most effective parameters on predicting the head loss of flow energy.

Keywords:

• Cavitation
• neural network
• critical flow
• drop number
• sensitivity analysis