Heat transfer performance prediction of Taylor–Couette flow with longitudinal slits using artificial neural networks

Abstract

A numerical analysis is performed to examine the heat transfer performance of turbulence inside coaxial cylinders with complex slit surfaces. The comparison is conducted for various number, width, and angle of slits. The synergistic relationship between the fluid temperature and velocity fields is explored following the field synergy principle. The back-propagation neural network (BPNN) coupled with a genetic algorithm (GA) is developed to predict the thermal performance of Taylor-Couette flow. Lastly, the slit structures are optimized by the particle swarm optimization (PSO) algorithm. Results indicate that the modification of slit structure leads to a remarkable difference in heat transfer performance of Taylor-Couette flow. The Nusselt number increases and then decreases with increasing the slit width, while a smaller slit angle strengthens the heat transfer properties. This principle remains applicable when the slit number changes as well. The smaller synergy angle is observed in the annular gap of the high heat transfer performance model which is owed to the excellent synergistic relationship between the temperature and flow fields. The proposed GA-BPNN model makes a remarkably accurate prediction of the Taylor-Couette flow heat transfer performance. Compared with linear regression, the correlation coefficient (R2) has increased by 24.55 %. According to the PSO algorithm and GA-BPNN model, optimal heat transfer performance is achieved with a slit structure of N = 12, w = 11.33 mm, and β = 60°. The maximum improvement in heat transfer capacity for a given range of Reynolds numbers is 16.35 %.