A health state assessment method for IGBTS under non-monitoring conditions

In the absence of historical monitoring data, traditional time-series-based methods for evaluating the health state of Insulated Gate Bipolar Transistors (IGBTs) are rendered inapplicable. To overcome this limitation, this paper proposes an innovative assessment framework specifically designed for scenarios lacking monitoring data. First, a polynomial-enhanced Wiener process is introduced to construct a nonlinear degradation model capable of accurately capturing the complex ageing behavior of IGBTs. Second, a consistency-based similarity metric is utilized to identify the most representative ageing trajectory from a set of reference patterns, ensuring alignment with the degradation characteristics of the target device. Finally, health state estimates derived from multiple parameters across different reference devices are clustered and filtered using the K-Means algorithm to enhance assessment robustness. Experimental results show that the proposed method achieves accuracy levels exceeding 95 percent under various ageing stress conditions, thereby confirming its effectiveness and resilience in data-scarce environments.