| Gas monitoring system is an important defense line to ensure the safety production of coal mine.Gas sensor is the source device of gas monitoring system.Its own operation state directly affects the accuracy of real-time measurement of gas concentration,and then affects the reliability of coal mine gas monitoring system.Comprehensive utilization of mine sensor data,mining the potential relationship between the data,so that the gas sensor has the ability of real-time diagnosis,gas monitoring system with high reliability and high intelligence is one of the functional objectives of the construction of intelligent mine.Based on the national key research and development project,combined with a coal mine engineering project in Zhengzhou,this thesis extracts the historical gas concentration time series data from the coal mine monitoring system,explores the temporal and spatial correlation of gas concentration time series at different locations,carries out the research on the online monitoring method of gas sensor performance,and puts forward a kind of gas concentration abnormal reason based on temporal and spatial correlation and information fusion.This method can realize on-line performance monitoring of gas sensor,reduce the probability of missing report and false report caused by gas sensor fault,diagnose the cause of abnormal gas concentration in real time,and facilitate managers to take measures in time.It has certain theoretical significance and engineering application value for strengthening the reliability and intelligence of gas monitoring system and improving the safety mining level of coal mines in China.The main contents of this thesis are as follows:1.This thesis introduces the morphological characteristics of eight common gas concentration time series,namely:stable state,coal and gas outburst,gas after blasting,local air stop,sensor calibration,sensor deviation fault(constant deviation,constant gain),sensor significant mutation fault.The stationarity of eight typical gas concentration time series is verified by ADF/PP characteristic root test.2.When the gas concentration in coal mine roadway is abnormal,the space-time correlation of gas in the process of roadway migration is analyzed quantitatively and qualitatively,and the calculation formula of lag step in the process of gas migration is determined.The correlation coefficient and correlation strength are introduced to quantify the spatiotemporal correlation strength of gas concentration time series under eight different conditions,which can help to diagnose the causes of abnormal gas concentration.3.Based on the historical gas concentration time series database data of a coal mine in Zhengzhou Coal Industry(Group),an effective ARMA-RNN gas concentration time series hybrid prediction model is established.The model can effectively predict the time series of gas concentration under stationary state.The traditional threshold method is improved,and the residual energy and residual kurtosis are introduced as the fuzzy evaluation indexes to find the abnormal values of gas concentration time series.The ARMA-RNN prediction model combined with the improved judgment method constitutes an effective gas concentration anomaly detection system.4.A feature extraction method of gas concentration time series based on composite multi-scale entropy analysis and Fisher discriminant method is proposed.The spatio temp oral correlation strength of gas concentration time series is integrated into the feature vector,and the recognition method based on PSO-SVM is used to complete the recognition and diagnosis process.5.The first mock exam based on composite multiscale sample entropy and the diagnostic results of the recognition and diagnosis model based on the compound multi-scale permutation entropy are fused by using D-S evidence theory.The shortcomings of the single model are overcome,and the accuracy of diagnosis is improved.This thesis has 31 pictures,14 tables and 76 references. |
PDF Full Text Request