Research and Application of Lifetime Feature Extraction for PCBs Based on Time-Frequency Transforms

With the increasing complexity and integration of printed circuit boards (PCB), coupled with the emergence of condition-based maintenance requirements, prognostics and health management (PHM) technology of PCBs has been developed, making the extraction of distinguishable lifetime features from circuits a research focus. Given the challenges associated with traditional measurements of physical parameters for sensitive devices and the impossibility of on-board evaluations, this paper proposes a method for extracting key signal lifetime features of PCBs based on time-frequency transforms. This method begins with wavelet packet decomposition and empirical mode decomposition (EMD) of power circuit signals in PCBs for time-frequency transformation. Signal features are then extracted using indicators such as information entropy, kurtosis, and signal energy percentage from the transformed signals. The extracted signal features are subsequently trend-filtered through Spearman correlation analysis to yield lifetime features. This method was applied to the circuit module of a switching power supply, and the extracted features were validated through accelerated life testing. The results indicate that the proposed method can effectively capture lifetime features that represent the current circuit state and exhibit a degradation trend.