IMPLEMENTATION OF NEURAL NETWORK MODELS TO PREDICT RESISTANCE FLOW THROUGH POROUS MEDIA WITH CONVERGENT BOUNDARIES

Abstract

Experimental and Neural network modeling investigation were completed to study the resistance to non uniform flow through porous media with convergent boundaries. The Experimental observations were made with Convergent Flow Permeameter using crushed rock as the media and water as fluid to anlyse the resistance of flowing fluid with different radial lines and with different ratio of radii. The present investigation aims to develop NNCB models to predict the optimal solution by its ability to capture non linear interacts among various parameters of the system. Feed Forward Back Propagation Neural Network models have been used in the present study for prediction of resistance flow at different radial lines with different ratios of radii and also to predict the relationship between friction factor (fk) and Reynolds number (Rk) for flow in porous media with converging boundaries, using intrinsic permeability as the characteristic length. The results shows that the ANNs can be very efficient tools for predicting resistance flow and also it is possible to obtain good ANN model performance even with extremely simplified architectures involving a very few input variables.