Mechanical Engineering / Sustainable Energy Session:

Title:            Application of neural networks for the prediction of heat transfer parameters in a multi pass cross flow heat exchanger
Authors:       Shivakumar A, Krishna Murthy, Srinivasa Pai P., Shrinivasa Rao B. R.
Abstract:       In the present paper applicability of neural networks was tested in order to correlate the experimentally determined heat transfer parameters of a multi pass cross flow heat exchanger. The waste heat from the IC engine is used to heat the water in a cross flow heat exchanger. Experiments were conducted for varied mass flow rates of cold water as well as varied flow rates of exhaust gases. Heat transfer characteristics like outlet temperatures, effectiveness and correction factors for the heat exchanger were determined. ANN modeling was done by using MLP (Multilayer perceptron). Experimental results were used for the ANN modeling. Network has been trained by using trainrp, trailm and trainscg which are the different training algorithms under MLP. Mean relative error (MRE) is the performance parameter selected for measuring the performance of the network. Network results exhibited a close tolerance with the experimental results. Results trained by trainlm were better than trained by other algorithms. MRE for the test data trained by trainlm for all the heat transfer parameters were within 5% showing its better prediction capability over other algorithms. These satisfactory results suggest MLP network can be used to predict the thermal performance characteristics of multi pass cross flow heat exchanger using limited number of experimental data.

Keywords:   Artificial Neural Network, Multilayer Perceptron, Cross Flow Heat Exchanger, Effectiveness
Pages:         350-354

Download Full Paper