Abstract: To solve the problems of few identifiable fault types and low diagnosis accuracy of single-source sensor fault diagnosis, by utilizing current and vibration sensor data, we proposed a multi-source feature selection and fusion method based on the sequential forward selection (SFS) and fuzzy C-means (FCM) clustering. This method evaluates the clustering performance by adjusting the Adjusted Rand Index (ARI), and selects and fuses the extracted multi-source sensor features to obtain the optimal feature set. Based on this, nine types of circuit breaker faults are simulated and divided into three classes. Support vector machine (SVM) is used to classify the single-source sensor features and the multi-source fusion features separately so as to verify the effectiveness of the proposed method. Moreover, three other common classifiers are used for comparison experiments. The results show that the multi-source fusion features have significantly higher recognition accuracy than the single-source features, reaching 95.0%, 92.5%, and 96.5%, respectively, in the three classes of faults, and they can achieve similar results under multiple classifiers, which is effective and universal. The proposed method provides a new approach for circuit breaker fault diagnosis in the context of multi-source sensors.