Only Divide Data: Mechanical Fault Severity Diagnosis Method of High Voltage Circuit Breaker

摘要

We present ODD, a new approach to circuit breaker operating mechanism fault diagnosis with vibration signal. Prior work adopts complex pre-processing and extraction method. Besides, it puts features into classifiers to perform fault identification. Instead, we take fault diagnosis as a regression problem to figure out fault severity. A simple time-domain method extracts features directly from original signal. Then features are put into weighted support vector regression(WSVR). Since the whole diagnosis process is simple, the number of parameters needed to be tuned is small. Our ODD model has a good performance on operating mechanism fault severity diagnosis. The no-load test results show that the maximum diagnostic error for coil power supply voltage fault is less than 8% of rated voltage. As for the closing spring's compression/stretching severity, ODD's error is less than 5mm. In addition, ODD runs extremely fast, it outperforms other diagnosis method in speed, including EMD and its improved method.