Enhancement of convective heat transfer in Taylor-Couette flow with elliptical slits using machine learning and particle swarm optimization

Taylor-Couette (T-C) flow commonly occurs in the annular gaps of rotating machinery, and improving its heat transfer performance is essential for effective thermal management. However, existing empirical correlations often have limitations in both efficiency and accuracy. To address this, this study integrates machine learning with optimization algorithms to refine T-C flow configurations that incorporate elliptical slits, aiming to develop a more efficient and precise optimization approach. Four machine learning methods are compared against a predictive correlation to assess their prediction accuracy for T-C flow. The particle swarm optimization (PSO) algorithm is subsequently applied to determine the optimal slit parameters. The results indicate that the Genetic Algorithm-Back Propagation Neural Network (GA-BPNN) model is the most suitable model, showing the highest agreement between predicted and simulated values. By incorporating the PSO algorithm, the optimal slit width of 11.33 mm, slit depth of 12.48 mm, and slit number of 12 are obtained. The predicted results agree well with experimental data, exhibiting a relative error of only 2.99%. Compared to the rectangular slit model, the optimized elliptical slit enhances the Nusselt number by 17%. The methodology and findings presented in this study provide a methodological and technical reference for optimizing and enhancing T-C flow systems.