Exploring Turbulence and micro-scale mixing mechanisms for enhancing jet impingement heat transfer using micro-roughness elements: A data-driven and numerical analysis

Abstract

This study uses machine learning to quantify micro-scale mixing mechanisms that enhance jet impingement heat transfer. Experiments and computational fluid dynamics (CFD) simulations tested flat, discrete protrusion, and continuous V-groove surfaces under Reynolds numbers(ReN) from 10,000 to 27,500 and nozzle-to-plate distances(NPD) between 2 and 5 times the diameter of the nozzle. An integrated approach combining experimental data, CFD, and four neural network models was used for comprehensive turbulence analysis. The neural networks, trained on the combined sixty datasets, showed high predictive accuracy with R-squared over 0.999 and Mean Absolute Error over 1e-7. The results highlight that micro-scale turbulence, characterized by Reynolds stress, dominates over operating parameters such as ReN and NPD in enhancing heat transfer. Discrete protrusions actively disrupt the thermal boundary layer, promoting vigorous mixing, while weaker turbulence and insulation effects in V-grooves contribute less. Percentage change analysis shows protrusions are more effective at extracting energy from jet and generating turbulence at smaller NPDs, but V-groove performance increases more strongly with rising distance. This data-driven analysis provides insight into surface roughness-induced mixing mechanisms and compares key turbulence parameters to assess thermal performance. The advanced understanding will aid in developing optimized designs for improved heat transfer in practical applications.