关键词:

矿井提升机 , 制动系统 , Transformer神经网络 , 故障诊断 , 自注意力机制

Abstract: As the ' throat ' connecting the upper and lower mines, the operation state of the mine hoist directly affects the production efficiency and safety of the mine, and the braking system is an important guarantee for the stable operation of the hoist. In order to reduce the dependence on expert experience and fully exploit the complex relationship between data, a fault diagnosis method of lifting braking system based on improved Transformer neural network is proposed. Firstly, the fault phenomena and causes of the braking system are analyzed, and the monitoring parameters are determined. Secondly, an improved Transformer fault diagnosis model is built, and a multi-layer self-attention mechanism is used to capture the correlation and fault relationship between the monitoring data of the mine hoist. The pooling layer is introduced into the Transformer model to reduce the parameters of the model and alleviate the risk of over-fitting. Finally, the experimental research is carried out based on the actual operation data of the hoist. The Adam optimizer is used to update the model parameters. The results show that the accuracy of the improved Transformer fault classification prediction can reach 97.5 %, which is 6.1 %, 10.0 % and 14.8 % higher than that of Transformer, CNN and LSTM neural network, respectively.