An Innovative Method for Thermal Characterization of Automotive Electronic Devices Based on CNNs

In this paper, a new approach for the thermal characterization of power electronic devices is introduced. The new method is based on cellular nonlinear networks (CNNs). By using the proposed technique, the thermal impedance of the device can be derived by processing a series of thermal maps, which can be acquired from a single experimental trial. In contrast, the most common technique (the delta drain-source voltage method) used for thermal characterization requires that the device is turned off after each single measure. The approach is validated on a power electronic device used in automotive applications, proving that the CNN-based method has an accuracy comparable with that of the delta drain-source voltage method but is more efficient in terms of time and human resources needed for device characterization.