Data-Driven Study on Thermal Shock Resistance Prediction of Copper Alloys

The use of copper in the modern energy sector is expected to grow five times as the sector transitions from the traditional power generation of nuclear and fuel-based systems to renewable energy. In industries, copper's high conductivity and excellent oxidation resistance are utilized in the extreme environments of heat exchangers and cooling systems. To handle the sharp thermal gradients and sudden temperature changes in those conditions, copper alloys need adequate thermal shock resistance. Hence, its optimization is crucial for the development of these alloys. In this study, a dataset of 2198 copper alloys was collected for thermal shock resistance prediction. Models based on six machine learning algorithms were prepared using composition and manufacturing processes as features. The model based on the multi-layer perceptron algorithm achieved an accuracy of 0.89 R2 scores. Further analysis was done to study the effects of different alloying elements and manufacturing processes with the help of correlation heatmaps and sequential feature selectors. Alloying additions of nickel, aluminum, and silicon and heat-treatment processes of hardening and precipitation hardening were found to be most effective in boosting the thermal shock resistance.