Feature Extraction and Fusion of Multi-sensor Data Method for Electromagnetic Pulse Valve Diagnosis

The health status of the Electromagnetic pulse valve is of great significance for the operation of the bag dust removal system. The object of this paper is to investigate the intelligent monitoring and fault diagnosis method based on multi-sensor data fusion of Electromagnetic pulse valve. It is difficult to extract high quality features from single sensor data under complex operating conditions. However, multi-sensor data fusion is an effective method to improve the accuracy and robustness of fault diagnosis under complex operating conditions. Therefore, this study considers both pressure and electrical switching signals and proposes a multi-sensor data fusion method for Electromagnetic pulse valve fault diagnosis, called attention-based fusion network model (AFCNN). This model uses multiple one-dimensional convolutional neural networks (CNN) to extract depth features from multi-sensor data, and assigns different weights to the features based on learning global information using an attention mechanism. This fusion strategy can obtain comprehensive and representative fault information from multi-sensor signals. Finally, the experimental results show that the testing accuracy of the proposed method reach to 92.5%, which is obviously higher than that of single-sensor results and other fusion methods. It proves that the proposed method has superior diagnosis performance.