| The coal mine flood is one of the main security disasters in mine construction and production process, as well as methane gas and roof. Some precursor characteristics, such as rock mass stress variation, water pressure rise, will generate during water inrush accompanied by the release of acoustic emission (AE) signals. A wealth of information of rock internal state is contained in AE signals. The collected AE signals are processed and extracted precursor feature information as a basis for monitoring and forecasting of water inrush. This will give some guidance to curb China’s water inrush incidents.In this paper, through theoretical analysis and experimental research, from signal processing’s angle, mine interference signal recognition, AE signal pre-processing and precursor feature acquisition methods for water inrush are studied separately based on AE monitoring. The main work and innovation are as follows:(1) The characteristics of mine AE and interference signals, as well as the suitable noise reduction methods are studied. Considering the characteristics of mine AE signal, a variety of noise reduction methods are analyzed, and the noise processing to AE signal by wavelet is proposed. Through theoretical analysis and experimental simulation, Evaluating noise reduction effect by using SNR and RMSE, it can be seen that sym8 and coif5 wavelets are suitable, and the de-noising results are better than the other cases by selecting 6-layer wavelet decomposition, rigrsure threshold, sln mode and hard threshold processing.(2) Aiming at the limiting saturation problem in AE signal, a shape repair method is proposed. Experimental verification and effect appraisal to emulation limiting saturation signal are carried on. It can be drawn the conclusion that the shape repair is a better recovery method to limiting saturated. This method plays an important role in acquiring complete AE waveform and further feature extraction.(3) Wavelet characteristic power spectrum coefficient and feature vector are proposed for extracting the signal feature. Furthermore, neural network recognition method based on wavelet feature vector is also proposed. Through practice analysis, wavelet feature vector can express the signal’s characteristic by using fewer parameters, can be used to identify different types of interference signals and response signals. Wavelet feature vector is also very beneficial for restraining the instability of the response signal and suppressing interference signals in the same frequency band. Neural network based on wavelet feature vector has merits of simple structure, short running time, and is also advantageous in realizing the real-time pattern recognition.(4) The extracting method of waveform characteristics of AE signals is proposed by using wavelet packet in experiments of watery and drying coal rock under different stress rates. From the experimental results, the change characteristics of the AE event number, average amplitude and maximum amplitude are quite obvious under different watery situations and different stress rates.(5) Wavelet feature coding and condition time series characteristics of AE signal are proposed. The basis of wavelet feature coding is elaborated. The method of feature coding is put forward. The feasibility of wavelet feature coding has confirmed from code scheme’s availability and consistency. The condition time series is obtained by chronological arranging feature coding, which acquired from each signal analysis. The continuous feature coding cannot only differentiate mutually, and contains continuous distribution characteristics of energy, makes the waveform characteristics more orderly. This will lay an important foundation for the time sequence analysis of AE event’s evolution in mine water inrush. |
PDF Full Text Request