Machine learning-based surrogate models for finned heatsink thermal modeling

With the continuous increase in power density in modern power converter, there is a growing focus on thermal system design, as its performance is a key factor influencing power density and determining the reliability of power converter. As the main heat dissipation component in the power conversion field, the heatsink plays a significant role in improving the reliability of power converters. However, it is difficult to forecast the accurate thermal performance of the device in field use. Thus, the purpose of this work is to present and propose a methodology for heatsink modeling that are based on high-performance computing and machine learning. The developed ML-based surrogate models can predict the thermal performances of the heatsink without a complicated analytical model.