Module Parasitics-Based Current and Temperature Sensing Using Explainable Neural Networks.

Area of Science:

• Electrical Engineering
• Materials Science
• Computational Intelligence

Background:

• Traditional current and temperature sensing in power semiconductor modules rely on dedicated sensors.
• Parasitic components within modules offer alternative signal sources for monitoring.
• Artificial Neural Networks (ANNs) show potential for complex signal analysis.

Purpose of the Study:

• To investigate the application of simple ANNs for current measurement and junction temperature determination in power semiconductor modules.
• To utilize parasitic components (e.g., parasitic inductances, on-state voltage, turn-on delay time) as signal sources.
• To develop a more interpretable ANN approach for power module monitoring.

Main Methods:

• Extraction of signals from parasitic components within power semiconductor modules.
• Application of simple artificial neural networks (ANNs) for signal processing and information extraction.
• Development of a method to enhance the interpretability of the ANN models.

Main Results:

• Demonstrated the feasibility of using parasitic component signals for current and temperature monitoring.
• Successfully applied ANNs to extract desired information from complex, parameter-affected signals.
• Introduced a method to improve the transparency of ANN models in this application.

Conclusions:

• Simple ANNs can effectively monitor power semiconductor modules using parasitic components, reducing reliance on conventional sensors.
• The proposed method enhances ANN interpretability, addressing a key challenge in their practical application.
• This research paves the way for more integrated and intelligent power module monitoring systems.