On Neural Modeling of Heat Exchange in Heat Exchangers (Recuperators) with the Systems of Plane-Parallel Impingement Jets for Machine Building and Metallurgical Productions

摘要

The paper presents an artificial neural simulation network used to forecast the coefficient of heat exchange between the transfer surface and the systems of plane-parallel impingement jets in the heat exchangers for air heating in heat-treatment and thermal furnaces in machine building and at the metallurgical factories. The Levenberg-Marquardt algorithm is used for the network training in this study. A number of network structures are treated with four nodes in the input layer (in the function of which the constructive and operative parameters were considered), one node in the output layer and with various amount of neurons in the hidden layers. At the four input variables and a single hidden layer, the network structures had the following form: (4-2-1), (4-4-1), (4-8-1), (4-9-1), and with the two hidden layers were as follows: (4-2-2-1), (4-4-4-1), (4-8-8-1), (4-9-9-1). It has been stated as the research result that the following network structure enables obtaining the best predicted results out of the studied ones: (4-9-9-1). The regression analysis of the predicted results and target outputs has been performed, and the regression equations and the correlation coefficients have been obtained. The comparison is drawn between the results obtained when using the neural simulation and the experimental data