Research on Fatigue State Detection Technology of High-Voltage Circuit Breaker Spring Mechanism Based on Vibration

Abstract

Aiming at the difficulty in effectively evaluating the health condition of spring operating mechanisms of high-voltage circuit breakers after tens of thousands of operations, this research is carried out along the main line of “signal perception – evolution law – mechanism modeling”. At the signal perception stage, a vibration signal-based condition detection technology is developed to realize online monitoring of the spring operating mechanism. At the evolution law stage, the fatigue evolution law of the spring is revealed through tens of thousands of cycles of operation tests, and the long-term variation trend of mechanical characteristics is identified. At the mechanism research stage, the influence mechanism of key geometric parameters on the mechanical properties of helical springs is clarified. The research results can provide comprehensive theoretical and technical support for the optimal design, condition assessment and fault diagnosis of high-voltage circuit breaker spring operating mechanisms, and possess remarkable engineering application value. Furthermore, this paper proposes a short-time energy-to-entropy ratio (STEER) double-threshold event extraction algorithm as the core diagnostic method. The algorithm computes the STEER for each VMD-decomposed intrinsic mode function (IMF) by dividing signals into overlapping frames (64 samples, 1.28 ms) and calculating the ratio of short-time energy to short-time spectral entropy within each frame. A composite STEER is obtained by summing across all effective IMFs. Double thresholds (T_high $={0. 9} \cdot$ R_max, T_low = 0.8·R_max) are then applied: the high threshold identifies impact event centers, while the low threshold extends boundaries in both directions to delineate complete event intervals. Post-processing merges overlapping intervals and discards spurious detections shorter than 5 frames (0.064 ms). This algorithm effectively amplifies the contrast between impact events and background noise, enabling accurate extraction of switching time parameters and segmental energy features for subsequent FCM clustering and SVM-based fault identification, achieving reliable discrimination among normal operation, spring fatigue, damper failure, and base looseness conditions.