Research On Early Warning Method Of Coal And Rock Combination Failure Based On GA-BP Neural Network

Coal and rock dynamic disasters are sudden damage or instability of coal and rock caused by disturbed actions of coal and rock.Common dynamic disasters related to coal and rock include coal and gas outbursts,rock bursts,and rock bursts.The deformation and fracture of coal and rock is not only affected by its own physical and mechanical properties,but more importantly,it is affected by the combined structure of coal and rock.Therefore,considering the roof and floor rocks and coal seams as a system to form a “rock-coal-rock”(RCR)combination,the study of its mechanical properties and characteristics of damage and instability is more in line with the actual situation,and it is also useful for preventing coal and rock.Dynamic disasters are of great significance.This paper conducts uniaxial compression experiments on typical ternary coal-rock assemblies with different height ratios,and analyzes the mechanical characteristics,instability failure characteristics and acoustic emission response laws of coal-rock assemblies under uniaxial load through acoustic emission monitoring technology.In-depth analysis of acoustic emission characteristic parameters and optimization of precursor signals are carried out.The BP neural network and genetic algorithm are used to establish a coal-rock combination failure precursor identification model to realize the precursor warning of coal-rock combination failure and instability.The main research results are as follows:(1)Uniaxial compression mechanical properties and acoustic emission response law of coal-rock assembly.The mechanical properties of the combination are between the rock sample and the coal sample.As the thickness of the coal seam increases,the uniaxial compressive strength and elastic modulus of the composite body increase within a certain range;the final failure mode of the composite body presents coal and rock co-shear failure,coal rock "H" tensile failure,and coal rock " X” type co-shear failure;the time of the increase inflection point of the combined acoustic emission ringing count and absolute energy accumulation curve is advanced.Before the destabilization of the combination occurs,the number of signals in each frequency band of the peak frequency increases,and the 195?205k Hz frequency band forms a dense distribution band,and the proportion of low-frequency signals rises faster than high-frequency signals.The law of the temporal evolution of acoustic emission signals can be monitored in real time and the incubation and development of coal and rock dynamic disasters,so as to realize early warning of coal and rock dynamic disasters.(2)Optimization of precursor signals for failure of coal-rock assembly under uniaxial compression.The characteristic parameters and the two-step clustering method are used to classify the acoustic emission signals generated during the whole process of uniaxial compression of the coal-rock assembly,and the signal types with the evolutionary abnormality of the destruction precursor are selected as the precursor signals of the coal-rock assembly destruction.The time intensity ft is used to quantify the damage precursor signal,and the threshold is defined according to the evolution characteristics of the signal time series.(3)A model for identifying precursors of coal-rock combination failure based on BP neural network is established.According to the principle and structure of BP neural network,determine the input layer and output layer parameters,and set appropriate network training parameters.The prediction results show that the AE precursor information identification model of uniaxial compression failure of coal-rock combination based on BP neural network can make good predictions for unknown types of acoustic emission signals,but the prediction results are volatile and unstable.(4)An optimization model for the identification of failure precursors of coal-rock composite bodies based on GA-BP neural network is established.Aiming at the problems of the BP neural network prediction model,genetic algorithm is selected to optimize the BP neural network.The result proves that the prediction error of the discriminant model based on GA-BP neural network is smaller than that of BP neural network,and it has higher signal recognition accuracy and good optimization effect.This paper combines BP neural network,genetic algorithm and coal-rock combination uniaxial compression failure precursor early warning problem,and proposes a coal-rock combination failure precursor identification model based on GA-BP neural network,which can be combined with the time intensiveness threshold.Realize effective and intelligent early warning of coal-rock combination damage,and provide an effective idea for preventing coal-rock dynamic disasters in actual coal mine production.