Research on intelligent sensing method for tulip contact status of circuit breakers based on lightweight multi-class support vector machine

Abstract Addressing the issues of data quality and detection accuracy in the pressure measurement of tulip contacts in medium-voltage vacuum circuit breakers, this paper proposes an intelligent monitoring system with a lightweight multi-class support vector machine (SVM) framework. The system adopts a senso-three-channel transmitter-Bluetooth-mobile terminal architecture to enable real-time acquisition, transmission, and visualization of pressure signals. Based on pressure time-series data from 10 kV circuit breakers, discrete Kalman filtering with a forgetting factor and information-maximization-based feature selection are employed to construct an efficient multi-class SVM classifier. Experiments on a self-built dataset of 408 contact pressure samples data demonstrate superior classification performance compared with random forest and K-nearest neighbors. Crucially, the model demonstrates high accuracy and robustness across diverse valid ranges, validated extensively over 375 threshold and width combinations. The proposed system provides a low-latency and highly generalizable intelligent sensing solution for practical condition monitoring of medium-voltage vacuum circuit breakers.