| Large-scale disastrous ground pressure in complex goaf groups is a typical major source of danger in mines,and ground pressure monitoring is an important safety guarantee for mine safety production.Relying on the Guangxi key research and development plan "Intelligent monitoring,early warning and technology research on disastrous ground pressure in large and complex goaf groups in underground mines",this paper takes the Panlong lead-zinc mine in Guangxi as the engineering background,designs a ground pressure monitoring scheme for the Panlong mine,and provides a comprehensive analysis of the ground pressure monitoring scheme for the Panlong mine.The raw data collected by the ground pressure monitoring system is used as the research object to conduct research on acoustic emission signal noise reduction and signal classification and identification.The specific research contents and results are as follows:(1)The situation of goaf and ground pressure activity in Panlong lead-zinc mine is investigated,the necessity of ground pressure monitoring in Panlong lead-zinc mine is introduced,and a ground pressure monitoring scheme is designed for Panlong mine.The daily monitoring data of Panlong Mine was manually identified,and the waveforms generated by rock drilling operations,scraper operations,blasting operations,electric locomotives and other rail equipment operations,and stray current operations were identified.And established artificial datasets.(2)An improved CEEMDAN noise reduction method is proposed.This method is an improvement on the basis of the CEEMDAN method.Several IMF components are obtained by decomposing the CEEMDAN method.Adaptive culling.Through simulation experiments,it is found that the improved CEEMDAN noise reduction method can effectively eliminate noise components and irrelevant components,and can clearly decompose the effective signal.Compared with the original signal,more effective information is retained and the decomposition effect is better.In the application of Panlong mine,it is concluded that the improved CEEMDAN method can effectively eliminate noise signals and irrelevant components,and provide high-quality data for the next step of classification and identification of acoustic emission signals.(3)In order to realize the multi-classification of underground acoustic emission signals,an improved CEEMDAN-FCM acoustic emission waveform identification and classification method is proposed.The multi-permutation entropy of the waveform is extracted as a characteristic parameter,a characteristic matrix is established,and the FCM is used for cluster analysis.The analysis of the test results shows that compared with the CEEMDAN-FCM method,the improved CEEMDAN-FCM method has higher accuracy and better clustering effect.Cleaner data.(4)In order to realize more intelligent waveform recognition and classification,combined with the Markov transfer field in mathematical statistics and the deep learning method convolutional neural network,a MTF-DCNN recognition and classification model of AE signal of underground surrounding rock mass and noise signal of mining operation was proposed.MTF comprehensively considers the time-position relationship.CNN performs more comprehensive analysis in automatic feature extraction and recognition and classification,and the accuracy rate is higher than 97%.To further verify the generalization ability of the model,the test and training sets are trained under different noise environments.The results show that the original model has certain antinoise ability and robustness,and the anti-noise ability of the model is significantly improved after intensive training.The comparative analysis with other recognition and classification methods shows that the various indicators of the MTF-CNN method are significantly better than the traditional methods CNN,SVM,ANN for the original image classification. |
