Data-Driven Design for Targeted Regulation of Heat Transfer in Carbon/Carbon Composite Structure

Targeted regulation of heat transfer in carbon/carbon composite structure is built for cooling electronic device. A three-dimensional data-driven design model coupling genetic algorithm (GA) with self-adaption deep learning for targeted regulation of heat transfer in built structure is proposed. The self-adaption deep learning model predicts the temperature of built structure closer to optimal value in GA model. The distributions of pore and carbon fiber bundles in built structure are optimized by the proposed model. The surface temperature of electronic device in the optimized structures is 19.1%–27.5% lower than that in the initial configurations when the porosity of built structure varies from 3% to 11%. The surface temperature of electronic device increases with an increase in porosity. The built structure with carbon fiber bundles near the surface of electronic device and pore distribution in the middle of structure has a higher heat dissipation capacity compared with that in the initial configuration. Besides, the computation time of the proposed model is less than one tenth compared with that of the traditional genetic algorithm.