Experimental analysis of a mixed-plate gasketed plate heat exchanger and artificial neural net estimations of the performance as an alternative to classical correlations

In this study, experiments are performed to test the thermal and hydraulic performance of gasketed plate heat exchangers (GPHE). A heat exchanger composed of two different plate types is used for the experiments, for a Reynolds number range of 500-5000. The results are compared to the experimental results obtained for plate heat exchangers which are composed of plates that have the same geometry instead of mixing two different plates. Two methods are used to investigate the thermal and hydraulic characteristics based on the obtained experimental data. One of them is the classical correlation development for Nusselt number and friction factors. Artificial neural networks (ANNs) are also used to estimate the performance as an alternative to correlations. Different networks with various numbers of hidden neurons and layers are used to find the best configuration for predictions. The results show that, artificial neural networks can be an alternative to experimental correlations for predicting thermal and hydraulic characteristics of plate heat exchangers. They give better performance when compared to correlations which are very common in heat transfer applications. Especially for mixed plate configurations studied in this research, where different plate types are used as a combination in the complete heat exchanger, it is difficult to obtain a single correlation that represents all the plates in the heat exchanger. However, when ANN's are used, it is easier to predict the performance of mixed plate HEX and the predictions are more reliable when compared to correlations. (C) 2016 Elsevier Masson SAS. All rights reserved.