Application of artificial intelligence approaches for predicting Nusselt Index in plate heat exchangers

In various industrial heat transfer processes, the Nusselt Index is a critical parameter for the design, analysis, and optimization of heat exchangers. Consequently, the precise calculation of the Nusselt Index is vital for a comprehensive understanding of the system's thermodynamic behavior. This research creates a predictive model utilizing artificial intelligence (AI), featuring contemporary algorithms namely Randomizable Filtered Classifier (RFC), Bagging, M5 Pruned (M5P), Multilayer Perceptron (MP), Reduced error pruning tree (REPT), and Additive Regression (AR), to assess the Nusselt Index over a broad spectrum of Reynolds number in plate heat exchangers. The surface enhancement factor, Reynolds number, corrugated profile aspect ratio, and chevron angle are included as input variables in different AI models, with the Nusselt Index as the output variable. The results showed that the RFC technique, with a correlation coefficient (R) of 0.983, a Nash-Sutcliffe efficiency (NSE) of 0.983, a Kling-Gupta efficiency (KGE) of 0.978, and a mean absolute error (MAE) of 2.47, performed best in the test dataset. The results reveal that the MP method is more effective than the Bagging technique across the R, NSE, and KGE criteria. The Bagging method performs better when evaluated based on the prediction error criterion. Consequently, the model developed in this study has the potential to enhance the optimal design of plate heat exchangers significantly. By doing so, it not only aims to improve operational efficiency but also seeks to minimize experimental costs associated with the design and testing processes.