Frictional Pressure Drop Prediction Using ANN for Gas-Non-Newtonian Liquid Flow Through 45° Bend

Applicability of Artificial Neural Networks (ANN) methodology was investigated using experimental data obtained from our earlier experimental studies on the frictional pressure drop per unit length for gas-non-Newtonian liquid flow through 45° bend. This approach proved its worth when rigorous fluid mechanics treatment based on the solution of first principle equations is not tractable. The proposed approach towards the prediction is done using a Multilayer Perceptron (MLP), which is trained with backpropagation algorithm with the help of four different transfer functions in a hidden layer. Statistical analysis confirms that the transfer function 1 with 20 processing elements in the hidden layer gives the best prediction.