Performance optimization of a heat exchanger with coiled-wire turbulator insert by using various machine learning methods

In present study, heat transfer augmentation and pressure loss in a double pipe concentric type heat exchanger with a coiled-wire turbulator inserted in it, are discussed in terms of regression analysis by using various Machine Learning (ML) methods. The non-dimensional design parameters in question are; the Reynolds number (Re), the thickness (e/d), the pitch (p/d) and the length (l/d) of the coiled-wire. Nusselt number (Nu), friction factor (f) and thermal performance factor (eta) are the extracted results of the experiments, namely the outputs of the system. In the regression analysis four well-known methods are applied; Supported Vector Regression (SVR), Gaussian Process Regression (GPR), Random Forest (RF) and Multilayer Perceptron Network (MLP) (a kind of Artificial Neural Network (ANN)). The Multi Linear Regression (MLR) method is also applied to the results for comparison. The regression coefficient (R-2), mean square error (MSE), mean absolute error (MAE) and root mean square error (RMSE) are the performances obtained by each method. The findings demonstrated that, among the suggested techniques, the MLP and GPR models in particular can be effective tools for calculating Nusselt number, friction factor and thermal performance factor for the selected heat exchanger. The R-2 values for Nusselt number, friction factor and thermal performance factor in the MLP model are respectively found to be 1, 0.95 and 0.98. In terms of all performance criteria, the order from best to worst-performing method is as follows: MLP, GPR, SVR and RF. In addition, it was observed that all four methods applied give better results than MLR does.