Characteristics and Classification of Microseismic Signals in Heading Face of Coal Mine: Implication for Coal and Gas Outburst Warning

Microseismic monitoring technique has been proved to be an effective method to predict catastrophic disasters in underground engineering. However, it is difficult to identify the microseismic signals of coal fracture due to the complex environment and operations in coal mine, which hinders the application of microseismic monitoring on the early warning of coal and gas outbursts (outbursts). In this work, microseismic signals in the heading face of Xinyuan coal mine were recorded under different operations, including drilling of holes, roadway support, moving of belt conveyor, driving activities, blasting and no work. The records show that the microseismic events occur frequently under driving activities, while a few microseismic events happen during the no work period. The characteristics of time domain, frequency domain and time-frequency for different microseismic signals were then analyzed based on the fast Fourier transform and short-time Fourier transform methods. The results indicate that the low-frequency waveform is well developed under the drilling of holes, roadway support, moving of belt conveyor and no work period. A short and discontinuous high-frequency waveform exists in the work of roadway support, moving of belt conveyor and no work period, which belongs to the coal fracture microseismic signals. Driving activities have massive low-frequency and high-frequency waveforms, but the high-frequency waveform corresponds to the coal fracture microseismic signals. The high-frequency waveform existing in the blasting work presents an obvious periodic feature. Combined with the long short-term memory (LSTM) network, different MS signals were further classified after extracting the maximum amplitude, dominant frequency, rising time, duration time and period from separation waveforms. Compared with the k-nearest neighbor and support vector machine models, the classification result of LSTM has the lowest loss value and highest accuracy rate of about 0.18 and 85%. The microseismic events of coal fracture identified by the LSTM have a better sequential response to gas concentration, which can be used to warn the outbursts in advance. This study provides a theoretical support for the early warning of outbursts and is meaningful for the safety and efficient production of coal mines.

Automated summary

Microseismic monitoring technique was used to analyze microseismic signals in coal mines. Different microseismic signals were classified using the LSTM network. The results showed that the LSTM classification had the lowest loss value and highest accuracy rate. The identified coal fracture microseismic events had a good sequential response to gas concentration, which could be used to warn the outbursts in advance.