NEURAL-NETWORK METAMODELLING FOR THE PREDICTION OF THE PRESSURE DROP OF A FLUID PASSING THROUGH METALLIC POROUS MEDIUM

Abstract

The pressure drop across metallic porous medium is a critical element in cooling aerospace engineering application. This paper presents a metamodel based on artificial neural networks (ANN) for estimating the pressure drop through metallic porous media. The ANN is developed using experimental data obtained from an experimental bench, developed at PRISME laboratory, which ensure the monitoring of temperature, pressure and mass flow rate in stationary and transient conditions. For each case the gas pressure which crosses the metallic porous material is measured as a function of inlet gas pressure, gas mass flow rate and temperature. The optimal feedforward ANN architecture with error backpropagation (BPNN) was determined by the cross validation method. The ANN architecture having 35 hidden neurons gives the best choice. Comparing the modelled values by ANN with the experimental data indicates that neural network model provide accurate results. The performance of the ANN model is compared with a metamodelling method using multilinear regression approximation. 2/20 NOMENCLATURE m number of hidden nodes T dimensional temperature (K) n number of input nodes t target value O i simulation output measures, corresponding to X i W ij connection weights between the input layer and the hidden layer P gas pressure (Pa) W jk connection weights between the hidden layer and the output layer p number of output nodes X i particular input parameter setting, i q gas mass flow rate (kg s-1) Subscripts S transfer function in inlet s number of patterns out outlet