Theoretical And Experimental Study On Coal-rock Interface Identification Based On Multi Information Fusion

Intelligent and unmanned mining are the inevitable development trend of the coal mine in the future while coal-rock interface identification is one of the key technologies,which realizing the intelligent control of the shearer,improving the efficiency of coal mining and prolonging the service lifetime of the shearer.Due to the complicated geological conditions of the working face,the signal is seriously disturbed to be identified.The realization of feature extraction and accurate recognition of the cutting signals are the key and urgent theoretical and technical problems.Thus,the effective and accurate identification of coal-rock interface is achieved by developing cutting experiment platform,extracting and identifying cutting characteristic signals and building identification model based on multi information fusion decision.Vibration signals,acoustic emission information and infrared thermal image features are generated accompanied by cutting coal and rock.Meanwhile,the cutting current also changes with the cutting load.The accurate extraction and recognition of the feature signals for multi signals are the significant basis and premise to build quality feature databases and identify the coal-rock interface accuracy.During actual mining process,considering the structure distribution and cutting characteristics of coal and rock,seven kinds of coal and rock specimens with different cutting coal ratios are poured.Furthermore,the physical and mechanical properties of specimens are determined.Besides,mechanical system,control system,data acquisition and analysis system of coal-rock cutting experiment platform are developed.The proposed studies provide the conditions and basis for the subsequent development of coal-rock cutting and signal extraction.Cutting experiment of coal-rock specimens is carried out to test multi feature signals.According to the change of current signal,the root mean square(RMS)value of three phase currents is used to reflect the real-time load.The time domain,frequency domain and wavelet packet analysis are used to analyze the vibration and acoustic emission signals and extract features.Furthermore,wavelet reconstruction of vibration signals and acoustic emission signals in different frequency bands are achieved and then the feature of acoustic emission energy and the RMS amplitude of vibration are extracted.To extract the peak values of flash temperature and obtain the temperature-frequency curves of temperature fields,the temperature field and flash temperature characteristics of cutting surface are analyzed.The change regulation of multi feature signals with different coal cutting ratios is determined by extracting,identifying and analyzing the vibration,acoustic emission,currents and the feature of infrared signals.Additionally,the sample databases of multi cutting signals are established to provide significant numerical basis for establishing the optimization model of membership function and optimizing the membership function based on minimum fuzzy entropy.By analyzing the fuzzy characteristics of the multi cutting signals,the optimization model of membership function based on the minimum fuzzy entropy is established.Combined with obtained sample databases and particle swarm optimization,the optimization of membership function thresholds is realized.Multi information fusion decision identification model for coal-rock interface is built based on optimized membership functions.Meanwhile,the basic probability assignment function,fusion rule and fusion decision rule of the model are studied.In accordance with the membership characteristics of cutting signals,an adaptive weight coefficient distribution model for coal-rock identification based on fuzzy membership is established,which can optimize and modify the weight coefficient of each evidence body under the conditions of no conflict,the relevance of single or multi evidence body and no association of evidence.The fusion results show that the model can effectively improve the accuracy of fusion results and reduce the uncertainty.Through laboratory random coal-rock interface cutting,the fusion decision identification model of coal-rock interface is used to fuse the test samples.Meanwhile,quantitative analysis method is uesd to compare the recognition accuracy of single signal and multi information fusion method.According to the characteristic of recognition results,the fusion result is optimized.The laboratory results show that the identification accuracy of coal-rock interface is significantly improved by using the multi information fusion decision model and the total recognition error reduces to 1.89%.In particular,the results of coal-rock interface identification are highly similar to the actual trajectory.The results validate that the multi information fusion decision model of coal rock interface has high recognition accuracy,as well as reducing overall recognition error.As a consequence,the proposed model could identify the coal-rock interface accurately and quickly.Meanwhile,it provides the theoretical basis and technical premise to realize the automation and intelligent mining of fully mechanized coal face.