Artificial Neural Networks Approach for Cross-Flow Heat Exchanger Fouling Modeling

The traditional estimation methods such as fundamental equations, conventional correlations or developing unique designs from experimental data through trial and error have limits in thermal engineering due to the complexity of problems addressed. Thereby, the purpose of the present work is to explain the effective utilization of the Artificial Neural Networks (ANN) model in heat transfer applications for thermal problems, like fouling in a heat exchanger. The application of the ANN tool with different techniques and structures shows that it is an effective and powerful tool due to its small errors in comparison with experimental data. The feed-forward network with backpropagation technique was implemented in Mechanical Engineering Technologies and Applicatithis study. Based on sensitivity analysis, the performance of the network trained was tested, validated and compared to the experimental data. The results achieved by sensitivity analysis show that ANN can be used reliably to predict fouling in a heat exchanger.