Running State Prediction of Belt Conveyor Based on Audio Feature Fusion Res-CNN-LSTM Network

Abstract: In order to solve the problems of inconvenient installation, poor stability, and lack of prediction of the operating status of the belt conveyor using contact sensors and machine vision to monitor the operating status of the mining belt conveyor, Res-CNN-LSTM based on audio feature fusion is proposed. Network-based belt conveyor operation state prediction method. Firstly, the audio signal is filtered and denoised, and then the one-dimensional Mel cepstrum coefficient MFCC0 of the signal is extracted by the Mel cepstrum method (MFCC) as the input of the network model; considering that the deepening of the network model will lead to over-fitting and performance degradation,The residual block is introduced to optimize the network model. Experiments are carried out on the laboratory belt conveyor platform. During the experiment, the network model parameters are continuously optimized. The results show that the audio signal can obtain more information about the operation status of the belt conveyor; compared with other models, the designed network model predicts the accuracy is the highest, and the training time is short; at the same time, it is verified that the model has high robustness under different working conditions.