Research On Fault Diagnosis Method Of Mine Gas Sensor Based On Multi-sensor Data Fusion

The gas sensor is one of the most important sensors in the coal mine safety monitoring and control system.The monitoring system completes the real-time monitoring of the underground gas concentration through it.The accuracy of gas sensor output signal plays a vital role in the performance of the whole monitoring system and the safety of coal mine production.However,the gas sensor is in a harsh environment of high temperature,high dust,high humidity and strong interference for a long time.It is prone to impact,periodic interference,jamming,deviation,drift and other faults,resulting in misstatement and omission accident of the gas sensor.Therefore,it is of great significance to study how to detect the abnormal signals of gas sensors in time,and distinguish the fault types of the abnormal signals,so as to analyze the fault cause of the gas sensor.Firstly,the time-domain characteristic parameters such as mean,root mean square,peak,variance,crest factor,kurtosis coefficient and skewness coefficient are calculated for the abnormal signals generated by the five common fault states of gas sensors.Through comparison and analysis,the kurtosis coefficient is used as the main time domain characteristic index for abnormal signal detection.Secondly,for the abnormal information of the detection signal from the gas sensor output signal,the multi-sensor data fusion technology is used to fuse the output data of the gas sensor of the typical coal mining face with the output data of the wind speed sensor,the temperature sensor and two gas sensors in different positions.The prediction model of support vector regression machine(PSO-SVR)based on particle swarm optimization is constructed.The predicted value of the gas sensor is compared with the measured value to obtain the residual,and the energy and kurtness of the residual are used as the judgment.Therefore,the lack of single threshold determination abnormality is avoided,and the accuracy of abnormal signal detection is improved.Then,For the abnormal signal of the gas sensor,the wavelet packet fractal algorithm is used to extract the fault feature of the abnormal signal.The gas sensor abnormal signal is decomposed,denoised and reconstructed by wavelet packet transform to obtain the reconstructed signals of different frequency bands,and then the fractal dimension of the reconstructed signals in each frequency band is calculated by fractal theory as the feature vector of fault.The gas sensor fault feature vector extracted by wavelet packet fractal has good reliability and stability.Finally,the binary tree support vector machine optimized by particle swarm optimization is used to classify and recognize the fault feature vectors extracted by wavelet packet fractal,and the fault classification of the gas sensor is completed.By comparing the fault identification accuracy of BP neural network classification algorithm and traditional SVM classification algorithm with the fault recognition accuracy of this method,the results show that the proposed fault diagnosis algorithm for the gas sensor has better fault classification and recognition effect,and achieves the purpose of improving the fault diagnosis ability of gas sensor.