Development of a Model to predict the efficiency of industrial heat exchanger By Artificial Neural Network (ANN)

This paper presents a predict method of industrial heat exchanger efficiency using artificial neural network (ANN). The ANN-developed model can expect the maintenance time prior to reaching the critical time of fouling. In fact, Fouling influence heat exchanger performance and cause sudden mechanical failure. So, must be studied fouling behavior, which is very complicated for heat exchangers because of the difficulty to monitor growth fouling. Three approaches are used for estimating the efficiency and evaluating the performance of the industrial heat exchangers. The first approach, C-factor (experimental method) depends on the pressure drop and volumetric flow rate. The C-factor approach gives results that are relatively accurate but it needs too much reading for a long period. The second approach, thermal analysis (traditional method) is a complicated mathematical model because needs many assumptions and design aspects that give approximated results. The third approach, ANN (modern method) is a very sensitive technic to evaluate the performance of industrial heat exchangers. This work, Uses the Feed-Forward neural network (FFNN) Configuration with the Bayesian regularization (BR) algorithm. Using 285 readings and measurements practical during operating the heat exchanger for training and testing processing to build up model architecture neural network. The maximum deviation between results ANN-based correlation, Thermal analysis, and comparing by experimental results of C-factor is 9.8 % and 33.6 % respectively. Based on ,the good results this assisted model reference strategy of ANN can be used to predict the efficiency of the heat exchangers. The examined network architecture by using 62 readings for another heat exchanger within acceptable certainties. Consequently, the ANN is flexible and capable to update in terms of new sets of weights and biases when the validity range changes in the same network.