Abstract
A novel temperature distribution visualization evaluation method for direct bonded copper (DBC) substrates in medium-high voltage power modules was developed based on square-wave (SW) excitation electroluminescence (EL). Firstly, an EL device with a planar electrode structure was fabricated, and the temperature dependence of the EL integral spectrum was analyzed using the Gaussian multi-peak fitting method. Subsequently, gradient temperature experiments statistically validated the equivalence between the RGB ratio of luminescence images and the integral spectral peak R12, establishing a foundation for RGB-based temperature characterization. Then, systematic investigations of temperature-sensitive parameters were conducted to identify key factors influencing the EL thermosensitive characteristics. Experimental validations and quantitative analysis of temperature measurement sensitivity were performed under applied electric field intensity, prolonged high-temperature testing (2 hours at 300.15K, 372.65K and 407.65K), and thermal-cycling conditions (from 298 K to 450 K for 9 times). A temperature inversion model based on the Gaussian process regression (GPR) algorithm was established using the image G/B channel intensity ratio as the parameter. Finally, the proposed model’s accuracy in temperature field reconstruction was verified on an EL device with spatial temperature gradient distribution and two DBC substrate samples with distinct heating configurations.