An improved Multi-channels Information Fusion Model with multi-scale signals for fault diagnosis

One of the challenges with Industrial artificial intelligence is that extracting features from signals collected from multiple sources of different components to train a more robust model. However, the acquisition of raw signals from complex mechanical systems shows a multi-rate sampling trend. To address the issue mentioned above, this paper presents a Multi-channels Information Fusion Model (MCIFM), which combines multisource learning into the basic CNN model. The results of Experiment 10 and the comprehensive comparative analysis of traditional CNN and traditional multi-source feature extractors show that this method can extract effective features from multi-rate sensors.