Multi-Modal Learning-Based Equipment Fault Prediction in the Internet of Things

• DOI: 10.3390/s22186722
• Publication Date: 2022-09-08T08:18:32Z
• Authors: Xin Nan; Bo Zhang; Changyou Liu; Zhenwen Gui; Xiaoyan Yin
• Keywords: Fuse (electrical); Robustness; Sensor Fusion; Concatenation (mathematics)

The timely detection of equipment failure can effectively avoid industrial safety accidents. existing fault diagnosis methods based on single-mode signal not only have low accuracy, but also the inherent risk being misled by noise. In this paper, we reveal possibility using multi-modal monitoring data to improve accuracy prediction. main challenge fusion is how fuse We propose a learning framework for low-quality and high-quality data. essence, are used as compensation Firstly, optimized, then features extracted. At same time, dealt with complexity convolutional neural network. Moreover, robustness algorithm guaranteed adding noise Finally, different dimensional projected into common space obtain accurate sample classification. Experimental results performance analysis confirm superiority proposed algorithm. Compared traditional feature concatenation method, prediction be improved up 7.42%.