Determination of Aerated Steps Number over Broad-Crest Stepped Spillways under Jet Flow Regime by using Artificial Neural Network

Recently, particular attention has been paid to the stepped spillways due to the increasing effect of energy dissipation and the reduction of cavitations risks with the development of Roller Compacted Concrete (RCC) technique. Flow regimes on the spillways divide into three groups, namely skimming, jet and transition flow. Compared to the numerical methods, the majority of performed studies in this field have been done using experimental methods, which requires a lot of time and expenditure. This paper has attempted to train an Artificial Neural Network (ANN) which includes two layers for determination of the number of aerated steps over broad-crest stepped spillways under the jet flow regime condition using MATLAB. The best network was trained entailing one neuron on each layer. Results show that the number of aerated steps can be obtained by having the critical depth, steps height, chute length, chute angle, and total number of steps. The obtained nonlinear relationship was validated using testing data. The predicted results have been compared to experimental data. Proper conformity of outputs and experimental data has been investigated using SPSS. The P-Value of Man-Whitney Test of the training and testing data compared to experimental data are 0.89 and 0.77, respectively. Good agreement was shown between experimental data and predicted results using ANN.