Fault Diagnosis of Main Pump in Converter Station Based on Deep Neural Network

As the core component of valve cooling system in a converter station, main pump plays major role ensuring stable operation valve. Thus, accurate and efficient fault diagnosis according to vibration signals is positive significance for detection failure equipment reducing maintenance cost. This paper proposed new neural network based on classify four faults one normal state pump, which consisted convolutional (CNN) long short-term memory (LSTM). Multi-scale features were extracted by two CNNs with different kernel sizes, temporal LSTM. Moreover, random sampling was used data processing imbalanced data, meaningful symmetry. Experimental results indicated that accuracy 0.987 obtained from test set, average values F1-score, recall, precision 0.987, 0.988, respectively. It found performed well multi-label superior other methods.