Feature Extraction And Recognition Of Coal Mine Microseismic And Blast Signals

Coal mine microseismic signals contain rich information of coal and rock burst, which have been widely used in the research of microseism locating and rock burst monitoring. However, the underground environment of coal mine is complicated and changeable where the interference factors are numerous. Microseismic signals induced by blasting(blasting signals) are majority and easily confused with the microseismic signals induced by coal and rock burst(coal mine microseismic signals). This paper focuses on the automatic identification of mine microseismic and blasting signals. The differences were analyzed that of mine microseismic and blasting signals in frequency distribution and nonlinear characteristics. Then, the significant different characteristics of two kinds of signals were obtained. Finally, the automatic identification model of coal mine microseismic and blasting signals was established and verified by field tests. The main research results are as follows:(1) Based on fast Fourier transform and multi fractal, the differences were analyzed that of coal mine microseismic and blasting signals in frequency distribution and nonlinear characteristics. The results show that the spectrum distribution of coal mine microseismic signals is more concentrated, and the low frequency component is dominant. The spectrum distribution of blasting signals is relatively dispersed, and the high frequency component is dominant. The singularity exponent distribution range and the spectral width of coal mine microseismic signals are less than the blasting signals. It shows that the distribution of probability measure and the local index fluctuation of burst signals are more intense than mine microseismic signals.(2) Based on Hilbert-huang transform(HHT), the three-dimensional energy spectra of two kinds of signals was obtained, and the corresponding relationship between sampling points(time), frequency and instantaneous energy was clearly displayed. The research shows that the instantaneous energy of each IMF component of coal mine microseismic signals is mainly distributed in the low frequency region of below 100 Hz. The instantaneous energy of blasting signals is mainly distributed in the high frequency region of 100~150Hz. The duration of instantaneous energy of coal mine microseismic signals is generally longer than blasting signals. Coal mine microseismic signals decay more slowly than blasting signals. In the decay process of instantaneous energy of coal mine microseismic signals, there are a number of jagged fluctuations. It shows that the attenuation process of coal mine microseismic signals is not smooth, and the tail wave is more developmental.(3) Attenuation curve of following peak could be obtained by using the power function to fit the envelope curve of the peak(the maximum amplitude) point to the noise level(end moment) point of mine microseismic signals and blasting signals. Then, the attenuation coefficient and fitting accuracy of two kinds of signals attenuation curves could be expressed by the power exponent and the correction coefficient of fitting results. In this way, the attenuation process of two kinds of signals was descripted quantitativly. The results show that the proportion of signals whose fitting precision is more than 0.8 is 90%. The distribution range of attenuation coefficient of two kinds of signals is 2~25. In most cases, the attenuation coefficient and fitting accuracy of mine microseismic signals are less than blasting signals.(4) The differences were analyzed and figured out that of two kinds of signals in dominant frequency, duration, attenuation coefficient, fitting precision, the first peak slope and the maximum peak slope. The research reveals that the dominant frequency, duration and attenuation coefficient are significant features to distinguish two kinds of microseismic signals. The fitting precision and the first peak slope can be used as auxiliary features to distinguish two types of microseismic signals. Based on dominant frequency, duration and attenuation coefficient characteristics, the automatic identification model of the coal mine microseismic and blasting signals was established by using Fisher linear discriminant method. The recognition accuracy of model was verified by field tests. The results show that the recognition accuracy rate of the automatic identification model is more than 85%, which can basically meet the requirements of mine microseismic monitoring with data processing.The research results have important theoretical and practical significance for improving the automatic recognition efficiency of the mine microseismic and blasting signals, reducing the workload of manual identification and extracting precursory information of rockburst.