Study On Time Series Data Mining And Group Recognition System Of Electromagnetic Precursor Information Of Rock Burst

Electromagnetic radiation technology has been used to forecast rock burst in some coalmines in our country. It is the development trend of forecasting rock burst by EMR technology in the future that how to recognize the precursor information of EMR monitoring data for rock burst effectively, study the EMR precursor characteristics of rock burst, and forecast rock burst accurately.In the paper, we have studied the EMR characteristics of coal or rock samples in the experiments, with field tests analyzed the EMR responding laws of different tectonics and the stress change on the working face, mined and analyzed the EMR monitoring data of rock burst by time series data mining (TSDA) technology, recognized the EMR precursor information of rock burst quantitatively, and established the group recognition system of the EMR precursor information of rock burst.EMR characteristics of coal or rock samples during the pre-peak phase and the post-peak phase of complete stress-strain experiments and in the friction experiments were studied. EMR signals which are produced by the deformation of coal or rock samples increase with the increase of stress during the pre-peak phase of the complete strain-stress process, and firstly increase and then decrease with the reduction of stress during the post-peak phase. In the friction experiments, the intensity of EMR signals are made to be enhanced with the increase of shear stress between the friction surfaces and the increase of loading rate .EMR responding laws of different tectonics and stress change on the working face are tested in the field and analyzed. EMR testing results can reflect the laws of the stress distribution and change on the working face. EMR signals produced by coal or rock around the tectonics such as faults and folds are usually abnormal.EMR precursor data of rock burst is analyzed by the traditional time series analysis method, and the similarity of EMR precursor sequence is measured by the data mining method of variable cluster. Research results show that EMR data of rock burst is a non-white noise and stationary sequence in which there is the precursor information to be extracted. ARMA model can be used to fit and forecast this sequence. According to the similarity measurement, the monitoring region of rock burst is divided into several areas with different risk, and the area with more risk is confirmed as the important area for predicting and treating rock burst. We have established the discrimination and warning criterion of EMR intensity abnormity including two statistical parameters of mean and variance, put forward the discrimination indexes of EMR individual and group abnormity, and recognized the EMR precursor information of rock burst quantitatively. The results show that this discrimination and warning criterion can estimate the dangerous degree of EMR abnormity, and these discrimination indexes can recognize the EMR precursor effectively.Based on the quantitative recognition of the EMR precursor information of rock burst, the group recognition system of EMR precursor is established and demonstrated. The demonstration results show that the three recognition levels of this system, including individual abnormity recognition, sub-group abnormity recognition and group abnormity, all can extract and recognize the EMR precursor information of rock burst effectively.The innovations of the paper are the time series data mining of EMR precursor of rock burst, and the group recognition of the precursor information.In the paper, the recognition of EMR precursor information of rock burst has been studied systematically, having an important theoretical and practical significance of forecasting rock burst.