Data filtering method for failure detection of mine catalytic combustible gas sensor based on BP neural network

Artificial intelligence technology has just started in the field of coal mine safety monitoring. The application of neural network technology in artificial intelligence to data identification and analysis in underground combustible gas detection can effectively improve the detection accuracy of sensors, filter invalid data and improve the level of safety monitoring. Due to the influence of chemical elements and environmental factors, catalytic sensors will have a certain characteristic response curve after detecting a certain concentration of combustible gas. Combined with BP neural network technology, this paper effectively classifies the sample data of characteristic curve and abnormal failure data, so as to identify the wrong data and filter it effectively, So as to improve the safety detection accuracy of the sensor. In view of the defects of BP neural network gradient descent algorithm, a sliding filter gradient descent method is proposed to make up for the problems of slow convergence and easy to fall into the minimum in the gradient descent algorithm. The experiments show that BP neural network technology has important practical significance in improving the detection accuracy of catalytic sensors.