Optimization of Thermohydraulic Characteristics of Converging-Diverging Microchannel Heat Sinks Based on BP Neural Network
2026-99-2516
11/20/2026
- Content
- As the heat flux of high-power electronic devices keeps increasing continuously, the structural design optimization of microchannel heat sinks has turned into the core factor for enhancing heat dissipation performance. The conventional design scheme of straight channels with uniform cross-section has an inherent conflict between thermal resistance and pumping power consumption, and the optimization of a single channel structure is hard to satisfy the performance demands under the coupling effect of multiple parameters. This research puts forward a performance prediction approach for microchannel heat sinks that integrates Computational Fluid Dynamics (CFD) and Back Propagation (BP) neural network. This investigation includes three representative channel structures: uniform height, converging type and diverging type, and further analyzes the nonlinear impact mechanism of channel width and wall thickness on thermohydraulic performance. A neural network model with high prediction accuracy was trained by constructing a high-fidelity database containing 3750 sets of samples. In terms of the data that falls within the effective operating scope of the model, the coefficient of determination (R2) corresponding to the model’s predicted data is entirely higher than 0.995, the relevant correlation coefficient R exceeds 0.99, and the root mean square error (RMSE) values for thermal resistance and pressure drop are lower than 0.00288 K/W and 6285 Pa individually. It indicates that the model achieves a high degree of correlation in both target parameters and has excellent robustness. Verified by tests, the model demonstrates the ability of auxiliary optimization and can replace the traditional CFD method. The time consumed for data prediction is shortened from hundreds of hours to seconds, with an efficiency improvement of 100,000 times, providing an efficient theoretical tool for the refined geometric design of microchannel heat sinks.
- Citation
- Wang, Q., "Optimization of Thermohydraulic Characteristics of Converging-Diverging Microchannel Heat Sinks Based on BP Neural Network," 2026 5th International Conference on Aerospace, Aerodynamics and Mechatronics Engineering (AAME2026), Nanchang, China, April 17, 2026, https://doi.org/10.4271/2026-99-2516.