• P-ISSN 0974-6846 E-ISSN 0974-5645

Indian Journal of Science and Technology


Indian Journal of Science and Technology

Year: 2015, Volume: 8, Issue: Supplementary 9, Pages: 1-6

Original Article

Artificial Neural Network based Prediction of Pressure Drop in Heat Exchangers


Objectives: The aim of the present work is to predict the value of pressure drop for different inlet-outlet configurations of an air cooled cross flow heat exchanger using artificial neural network. Methods: Configuration and mass flow rate are given as inputs. The pressure drop is calculated based on numerical simulations for various mass flow rates and different configurations. The numerical results are validated with experiments. This numerical data obtained for 4 different flow rates for each of the 24 configurations is used in training a back propagation based neural network to predict the pressure drop. Results: In the present work Levenberg-Marquardt algorithm is used to train the network. Two inputs are specified in the present study. Mass flow rate and type of configuration are assigned in the input layers and the output is the pressure drop across the heat exchanger. The number of hidden layers is fixed as 10 after a series oftrial and error. The data are taken from the simulations and are fed to the network. 80% of data is taken for training and 20% are taken for validation and testing. Using the feed forward perception network the input is propagated and the error in the output is back propagated to modify the weight. Sigmoid transfer function is used as the activation function for the hidden layer and is represented in Equation 3. It can be concluded that the neural network is able to predict the pressure drop for varying input parameters. The neural network gave a mean regression coefficient to be 0.97. The regression plot for the training, validation, testing data has been shown in Figure 6. Conclusion: As the regression value approaches 1, it may be concluded that ANN is capable of predicting the results for different set of input parameters within the training range.
Keywords: Artificial Neural Network, Flow Maldistribution, Heat Exchangers, Pressure Drop, Radiator


Subscribe now for latest articles and news.