Abstract:  Aiming at the problems such as the difficulty of fully fitting the characteristic parameters of hoist in different working environments and the difficulty of fault warning, a fault warning model of mine hoist based on the integration of long-short term memory neural network (long-short term memory, LSTM) and adaptive moment estimation algorithm (adaptive moment estimation, Adam) was proposed. Firstly, by analyzing the working principle and common fault manifestations of mine hoists, a prediction model of hoist characteristic parameters was established based on LSTM neural network, and the prediction model was trained and optimized by combining Adam optimization algorithm. Then, the actual operation data of a mine hoist was used to verify the performance of the built predictive model. The sliding weighted mean method was used to analyze the forecast residual, and the reasonable warning thresholds of several key feature parameters were obtained, and the elevator fault warning model was established. Finally, taking the brake system failure as an example, the effectiveness of the hoist fault warning model was verified by fault simulation experiments. The research results show that when the learning rate of the model is 0.015, the training effect is the best, the loss rate of the predicted model can reach 0.12%, and the parameter change trend can be better fitted. The mine hoist early warning model based on LSTM-Adam can accurately predict the trend of parameter change; combining with the prediction residual analysis results for fault early warning, it can achieve accurate early warning of hoist failure.
Key words: hoisting machinery; mine hoist; fault early warning; longshort term memory(LSTM); adaptive moment estimation(Adam); deep learning