Research On Gas Concentration Abnormal Signals Identification Of Coal Mine Safety Monitoring System

The gas sensors are one of the most important sensors in coal mine safety monitoring system.Gas sensors can detect the coal mine gas emission status in real time,and the accuracy and health management level of the gas sensor is the key of modern coal mine safety production.However,the bad downhole environment such as high temperature,high humidity,dust,and strong jamming will led to the misstatement and omission accident of the gas sensors.Therefore,how real-time detecting a gas concentration abnormal signal,identifying its type,eliminating false alarm,and analyzing the reason of false alarm is very important.In order to solve the two problems with three kind of abnormal gas concentration signal,i.e.,the constant deviation type abnormal signal,collision type abnormal signal,and periodic pulse type abnormal signal detection and feature extraction classification,the coal mine safety monitoring system of gas concentration anomaly signal identification algorithm is proposed.The main Conclusions are as follows:(1)With the analysis of gas concentration abnormal signal identification characteristics,using data mining and signal analysis theory,one safety monitoring system of gas concentration abnormal signal identification method is put forward and the model is set up.This model can detect the gas concentration abnormal signal accurately,and feature extraction classification correctly.(2)The results indicate that the time domain parameter kurtosis has the high distinction degree for the three kinds of abnormal signal.In addition,the abnormal gas concentration abnormal signal can be divided into three patterns,such as collision type abnormal signal and periodic pulse type abnormal signal with the analysis of the abnormal gas concentration abnormal signal of coal mine safety monitoring system.(3)Based on genetic algorithm optimization support vector regression machine(GA-SVR)combined with adjustable operator analytical fuzzy decision index of abnormal signal detection method is proposed in the research of gas concentration abnormal signal detection means.Firstly,multiple related sensor signals are integrated to estimate time series of the diagnosed sensor.Secondly,the measured value and estimated value of the diagnosed sensor were compared to get the residual error.Thirdly,residual characteristics were decided through the analytical fuzzy fault decision index based on adjustable operators,and finally getting the diagnosis result.Through the offline experiment in coal mine monitoring and control system,compared with the artificial neural network,support vector regression machine algorithm can well solve the problem of gas sensor fault small sample training,and has the good approximation capability for nonlinear change of gas concentration.Genetic algorithm can be adaptive optimization of SVR's parameters selection,so as to improve the accuracy of support vector regression avoiding the blindness of selecting parameters manually.According to the simulation results,the proposed decision index could effectively overcome the shortage of fault identification based on single threshold and reduce false alarm rate and underreporting rate of the diagnosis system.(4)In the research on gas concentration abnormal signal Feature extraction and classification method,gas monitoring abnormal signal identification experiment platform was set up,and a lot of abnormal signal samples are obtained with a large number of experiments based on this platform.Wavelet de-noising DFT spectrum analysis method,wavelet packet energy feature vector,and Hilbert Huang transform method of feature extraction performance experiment were carried out;the classification of the binary tree support vector machine classifier performance is validated.The experimental results show that the wavelet packet energy feature with vector means are affected by the large sample point selection and point in time and degree of abnormal signal.Hilbert-Huang transform method for short-term mutant abnormal data has good recognition effect,and the timing of the signal abnormalities were instructed by the instantaneous frequency of signal,the constant deviation type abnormal signal cannot be decomposed by Hilbert Huang transform method.The feature of the constant deviation type abnormal signal cannot be extracted by this method.The abnormal signal features can be extracted by the method based on wavelet noise reduction combined with DFT transform spectrum as well,the distinguish of DFT transform amplitude spectrum distribution is obvious,there are a little influence for identification results on the change of the starting time of signal sampling.The abnormal signal features of coal mine safety monitoring system can be extracted by the method based on wavelet noise reduction combined with DFT transform spectrum,and the classification accuracy of GA-PCA-SVC is above 96%.According to the gas sensor misstatement problem of coal mine monitoring and control system,from the constant deviation type abnormal signal,collision type abnormal signal and periodic pulse type abnormal signal identification perspective,the research on the gas density abnormal signal detection,feature extraction and classification method with the signal processing and artificial intelligence algorithms,promote the rapid detection and real-time forecasting ability of the coal mine monitoring and control system gas sensor misstatement,help the field engineers to finish complex troubleshooting,prevent greater loss because of false positives,Improve the security maintenance level of the coal mine safety monitoring system.