The progressive integration of power electronics in modern industrial applications, for example in the automotive industry, the field of renewable energies or aviation, has resulted in an increased demand for the reliability and safety of systems, given the inherent risks associated with these technologies. The failure of an electronic component can have a significant impact on the safety of an entire system. This is particularly relevant in safety-critical or on-board applications, such as those found in aircraft and automobiles, where uninterrupted monitoring without the need for external intervention is essential. It is therefore crucial to be able to identify any potential damage to these components at an early stage. The current state of the art encompasses a multitude of methodologies for the detection of damage, including thermal measurements, vibration analyses and the monitoring of electrical parameters and variables. Despite their efficacy in numerous applications, these methods are nonetheless associated with significant drawbacks. In general, the utilisation of additional sensors or hardware components is a common requirement of these methods, and can result in increased costs, a larger space requirement. Additionally, this can contribute to increased system complexity. Moreover, external factors, including temperature, load fluctuations, humidity, and electromagnetic compatibility (EMC), can negatively impact the reliability of the detection results. It is therefore imperative to develop integrated methods that do not necessitate additional hardware, yet still facilitate precise early detection of damage.This paper presents an innovative method for the early detection of damage in power MOSFETs, which is based on the analysis of thermal RC networks. The proposed method models an ideal RC network configuration, which is adapted to the specific thermal characteristics of the device under test (DUT). This is achieved by utilising the system’s thermal properties. The periodic evaluation and comparison of system impedance curves with the characteristic properties of the network enable the reliable identification and classification of potential damage or changes within the system. The effectiveness of the method was evaluated and verified using a large number of test boards with various integrated defect patterns. To enhance the precision of the outcomes, the experimental data was reinforced through the incorporation of exhaustive finite element simulations. Simultaneously, the experimental data was employed to assess the simulation models.The results indicate that the methodology employed is statistically sensitive with respect to a range of potential defects, including die-attach failures, solder joints and printed circuit boards (PCBs). The method for the early detection of damage in power electronics “on-board” using thermal RC networks thus represents a promising alternative to existing technologies. Moreover, the method can be implemented without the necessity for additional hardware and can be readily integrated into existing systems. Significant alterations in the time and value range of the thermal RC network provide a statistically robust foundation for the early detection of even minor damage. The early warning function is particularly advantageous in industrial applications, as it facilitates the timely planning of maintenance work and the avoidance of unplanned downtime. This results in increased system reliability and optimised cost efficiency. Nevertheless, further research is necessary to confirm industrial viability and optimize the method for a wider range of applications.Keywords: automated damage detection, thermal RC networks, functional safety, MOSFET, early detection.
Reim et al. (Wed,) studied this question.