Artificial neural network (ANN) modeling of COD reduction from landfill leachate by the ultrasonic process

In the study, the use of an artificial neural network (ANN) has been applied for the prediction of COD removal from landfill leachate by the ultrasonic process. The configuration of the backpropagation neural network giving the lowest mean square error (MSE) was a three-layer ANN with a tangent sigmoid transfer function (tansig) at a hidden layer with 14 neurons, linear transfer function (purelin) at the output layer and the Levenberg–Marquardt backpropagation training algorithm (LMA). The ANN predicted results are very close to the experimental data with the correlation coefficient (R²) of 0.992 and the MSE of 0.000331. The sensitivity analysis showed that all studied variables (contact time, pH, ultrasound frequency and power) have strong effect on COD removal. In addition, ultrasound power is the most influential parameter with relative importance of 25.8%. The results showed that modeling neural network could effectively predict COD removal from landfill leachate by ultrasonic process.