The performance of a power supply is highly susceptible to environmental variations, which directly impacts the proper functioning of nonlinear loads such as lasers and LEDs. This paper proposes an evaluation index-based multiphysics coupling modeling (EIMCM) technique that enables comprehensive characterization of various physical properties for power supplies. By correlating heat, stress, electricity and magnetism, the technique explores the intrinsic connections between different physical fields and establishes a reliability evaluation model. For the large number of elements and indicators in the evaluation, a back-propagation neural network (BPNN) is employed to train the multi-objective parameters, thereby improving the performance of the module. To demonstrate the capabilities of the proposed methodology, a 600 W power supply for driving laser loads is designed and optimized for heat dissipation, strain, output power, and regulated voltage accuracy. The simulation and experimental results exhibit strong consistency as a systematic error less than 4%, validating the effectiveness of the proposed EIMCM methodology.
.jpg "Xuliang Wang (he/him/his) photo")