CN-121808354-B - Working face ore pressure multi-parameter intelligent monitoring and early warning system

CN121808354BCN 121808354 BCN121808354 BCN 121808354BCN-121808354-B

Abstract

The invention relates to the technical field of working face ore pressure multi-parameter intelligent monitoring and early warning systems, and particularly discloses a working face ore pressure multi-parameter intelligent monitoring and early warning system. The system comprises a multi-source sensing data acquisition device, a pulse sequence coding unit, a bionic pulse neural network early warning model, a dynamic attention regulation and control module and an early warning information output terminal, wherein the multi-source heterogeneous data such as hydraulic support resistance, microseism energy and acoustic emission frequency are converted into a pulse event sequence similar to biological nerve signals, a simulated neuron discharge mechanism and a bionic model of synaptic plasticity are utilized to identify rock stratum instability precursors, and a dynamic attention mechanism is combined to adaptively strengthen key signals and inhibit noise. By adopting the technical scheme, the multi-parameter coupling characteristic can be captured with high sensitivity, the early warning accuracy and the early warning amount are improved, the false alarm rate is reduced, and the reliable guarantee is provided for the safety of deep coal mines.

Inventors

YUE TAO
YANG CHAO
LI XIANLONG
Du dongxing
DONG RUIKAI
WU YUJIE
WU LEI
HUANG YINGJUN
HAN WEI
HAO XIAOYU
YAN RENQIANG
Tang Gonglong

Assignees

内蒙古双欣矿业有限公司

Dates

Publication Date: 20260512
Application Date: 20260312

Claims (8)

1. The mining pressure multi-parameter intelligent monitoring and early warning system for the working face is characterized by comprising a multi-source sensing data acquisition device, a sensor matrix and a sensor matrix, wherein the multi-source sensing data acquisition device is configured to acquire an original physical signal of a fully-mechanized working face through the heterogeneous sensor matrix and convert the original physical signal into a standardized digital signal stream; the pulse sequence coding unit is connected with the multi-source sensing data acquisition device and is configured to map the digital signal stream into a pulse event sequence with a time stamp according to the change rate and the amplitude characteristic of each physical quantity to form a space-time coding format representing the dynamic process of rock stratum fracture; The bionic pulse neural network early warning model is connected with the pulse sequence coding unit and is configured to receive a multichannel pulse event sequence, and carries out nonlinear integration processing on the pulse event sequence based on a calculation rule of a simulated biological neuron discharge mechanism so as to identify a multiparameter coupling mode corresponding to a rock stratum instability precursor; the dynamic attention regulation module is integrated in a feedback control loop of the bionic pulse neural network early warning model and is configured to adaptively adjust weight distribution of each input channel according to the space-time distribution characteristic of the current input pulse stream, so as to strengthen attention to key precursor signals; The early warning information output terminal is connected with the bionic impulse neural network early warning model and is configured to receive an early warning judgment result and generate a visual warning prompt or a linkage control instruction according to the risk level; The bionic pulse neural network early warning model has synaptic plasticity learning capability and is connected with a historical disaster database storing typical accident evolution samples, the model is configured to automatically optimize connection weights among neurons in a training stage by simulating learning rules of long-term enhancement and long-term inhibition, when pulse combination frequently occurs in an accident precursor sequence, the model is configured to automatically adjust and increase synaptic weights among corresponding neurons so as to improve the recognition speed and the confidence of the mine pressure instability precursor, and the model also comprises a reference correlation learning module, is configured to acquire production parameters including the current of a coal cutter motor, the vibration frequency of a cutting part and the load of a scraper conveyor, convert the production parameters into a pulse sequence, learn a time domain response function between a production pulse sequence and the mine pressure monitoring pulse sequence through a background cancellation layer, and cancel pulse components caused by normal production operation by using inhibitory neuron synapses; The dynamic attention regulation module comprises a space-time correlation evaluation subunit and a weight self-adaptive correction subunit, wherein the space-time correlation evaluation subunit is configured to carry out sliding window statistics on pulse activities of all input channels, calculate pulse emission density of each channel in a current time window and calculate correlation scores of pulse activities among different sensor channels, the weight self-adaptive correction subunit is configured to reduce synaptic gains of channels judged to contain equipment noise or environmental interference or to increase discharge threshold values of corresponding input neurons of the channels according to evaluation results, and the dynamic attention regulation module is configured to judge that the current stage is in a high risk evolution stage when stent pressure, microseism and acoustic emission are in a same time period to generate pulse density cooperation rise, so that trigger thresholds of the decision layer are automatically reduced and higher weight is given to the associated channels.
2. The mining pressure multi-parameter intelligent monitoring and early warning system for the working face according to claim 1, wherein the multi-source sensing data acquisition device comprises a bracket pressure monitoring sub-module, a microseism monitoring sub-module and an acoustic emission monitoring sub-module; The support pressure monitoring submodule comprises a pressure transmitter arranged on a hydraulic support upright column cylinder body and is configured to acquire circulating resistance data of a working face support; The microseism monitoring submodule comprises a vibration pickup sensor array arranged on the periphery of a working face gallery or a goaf and is configured to capture vibration waveform signals generated by rock stratum fracture; the acoustic emission monitoring submodule comprises a high-frequency acoustic sensor and is configured to capture elastic waves in the development process of rock micro-cracks; The multi-source sensing data acquisition device is internally integrated with a signal conditioning circuit, an analog-to-digital conversion circuit and an anti-electromagnetic interference filter circuit adopting a three-stage inductance-capacitance filter structure; The multi-source sensing data acquisition device is also integrated with a zero automatic calibration circuit, and is configured to perform reference calibration on the pressure sensor in a support pressure relief stage so as to eliminate temperature drift and zero drift of the sensor; the multi-source sensing data acquisition device adopts a redundant array design, at least two mutually-prepared sensors are deployed at each monitoring point, and an internal health degree evaluation unit monitors the electrical characteristics of the sensors.
3. The mining pressure multi-parameter intelligent monitoring and early warning system for the working face is characterized in that the pulse sequence coding unit is configured to adopt a double coding strategy of a release rate and a time sequence, the pulse sequence coding unit adopts release rate coding logic aiming at a hydraulic support resistance signal, when the absolute value of support resistance exceeds a first preset threshold or the change rate exceeds a second preset threshold, pulse release frequency of a corresponding channel is increased, the pulse sequence coding unit adopts time sequence coding logic aiming at a microseismic energy and an acoustic emission frequency signal, the energy intensity of the signal is coded by utilizing a time interval of accurate arrival of the pulse, the earlier the initial pulse release is when the energy is larger, a global synchronous clock is arranged in the pulse sequence coding unit and is configured to ensure that sensing pulses from different physical positions are marked with time stamps under a unified time reference system, the error range of the time stamp is less than or equal to 1 millisecond, and the pulse sequence coding unit is further configured to insert a health status flag bit into a generated pulse data stream when a sensor is switched in a fault.
4. The mining pressure multi-parameter intelligent monitoring and early warning system according to claim 3, wherein the input coding layer is configured to receive the pulse event sequence and distribute the pulse event sequence to corresponding neuron receptive fields, the characteristic mapping layer comprises multiple groups of neuron clusters, each group of clusters is configured to generate sensitive responses to a mining pressure appearance mode, leakage integration triggering model logic is introduced into the space-time integration layer, each virtual neuron node has a membrane potential attribute, the model is configured to increase the membrane potential according to weight when input pulses arrive and enable the membrane potential to exponentially decrease with time according to a preset attenuation constant during no-pulse input, and the model is configured to control the neurons to issue output pulses to the next layer, filter scattered interference pulses and generate resonance responses to signals with space-time correlation only when the membrane potential is accumulated to exceed a triggering threshold value within a preset time window.
5. The mining pressure multi-parameter intelligent monitoring and early warning system of claim 4, wherein the early warning information output terminal comprises a multi-stage risk decider and a response execution matrix, the multi-stage risk decider is configured to map output pulse density of the bionic pulse neural network early warning model into four risk levels, namely a green safety state, a blue attention state, a yellow early warning state and a red emergency state, respectively, and the response execution matrix is configured to execute differentiation actions according to the risk levels, and comprises: Sending early warning information to the explosion-proof mobile terminal in a blue state; Driving the roadway audible and visual alarm in a yellow state; in red state, issuing emergency instruction to the fully mechanized mining automatic centralized control system, and forcedly executing shutdown or emergency pressure relief operation; The early warning information output terminal is also integrated with a three-dimensional geomechanical simulation feedback system, and is configured to extract peripheral pulse space-time characteristics and convert the peripheral pulse space-time characteristics into mechanical parameters when risk points are identified, and input numerical simulation software to predict risk evolution trend within 15 minutes in future.
6. The mining pressure multi-parameter intelligent monitoring and early warning system of claim 5, wherein the system adopts a distributed edge computing architecture and comprises an edge side processing node cluster deployed at a fully-mechanized mining face gateway and a transfer point and a central cloud decision server positioned at a ground monitoring center, wherein the edge side processing node cluster comprises an embedded system-level chip and is used for completing conversion of a physical signal to a pulse event at a physical perception front end, so that the data volume of a pulse data packet transmitted to the central cloud decision server is smaller than 1/100 of the original sampling data volume, the central cloud decision server is loaded with the bionic pulse neural network early warning model, the model adopts a deep circulation pulse neural network architecture and comprises a recursive neuron layer with feedback connection and is used for carrying out time sequence memory on the mining pressure evolution trend, and the central cloud decision server also comprises a self-evolution module and is used for periodically evaluating the accuracy rate and automatically correcting the synaptic connection threshold in a neural network through an enhanced learning algorithm by using a marked false report or missing report pulse sequence.
7. The mining pressure multi-parameter intelligent monitoring and early warning system for the working face is characterized by further comprising a distributed optical fiber sensing network, wherein the distributed optical fiber sensing network is paved in a roof of the working face and a roadway side part, the distributed optical fiber sensing network is configured by an optical fiber sensing host to convert an optical echo signal into an equivalent pressure pulse stream and is connected with a pulse sequence coding unit, the dynamic attention regulation module is configured to perform space convolution operation on pulse activities of adjacent optical fiber sensing points according to a space topological relation so as to position a stress concentration area and an evolution track, the early warning information output terminal supports an augmented reality display function, and is configured to superimpose a mining pressure distribution thermodynamic diagram and an early warning level on a digital twin model of the working face in real time by utilizing a three-dimensional modeling technology, and display a stress concentration area and a fracture crack propagation path in surrounding rock through explosion-proof augmented reality glasses.
8. The mining pressure multi-parameter intelligent monitoring and early warning system of claim 7, wherein the multi-source sensing data acquisition device is packaged in an explosion-proof shell with explosion-proof capacity, an internal circuit board is coated with three-proofing paint for preventing moisture and salt mist, and a connecting cable adopts a carbon fiber reinforced tensile armoured structure and a rotary self-locking joint; the communication links between the devices adopt a redundant ring network structure and are configured to enable data to automatically bypass through a redundant path when a main optical cable is subjected to single-point fracture, and the switching time is not more than 50 milliseconds; The bionic pulse neural network early warning model is deployed in an industrial server provided with an artificial intelligent acceleration board card, adopts an asynchronous logic circuit design at the bottom layer of hardware, and is configured to generate calculation power consumption only when pulse signals arrive; the system is provided with an intrinsically safe standby power supply configured to maintain monitoring data for no less than 72 hours in a power outage state; the system also comprises a wireless ad hoc network emergency search and rescue auxiliary module which is configured to form a low-power-consumption positioning network by utilizing residual electric quantity when the main communication network is paralyzed, and provide a post-disaster roadway collapse distribution prediction graph according to a surrounding rock rupture sequence recorded before disaster.

Description

Working face ore pressure multi-parameter intelligent monitoring and early warning system Technical Field The invention belongs to the technical field of mine safety monitoring and intelligent early warning, and particularly relates to a working face mine pressure multi-parameter intelligent monitoring and early warning system. Background Along with continuous acceleration of intelligent mine construction, safety monitoring and disaster early warning of fully mechanized mining face become core technical support for guaranteeing efficient production of coal mine. Accurate monitoring of mine pressure display can effectively prevent dynamic disasters such as roof accidents, rock burst and the like, and is a safe foundation stone for realizing unmanned or less-unmanned exploitation. Under deep complex geological conditions, the intensity and unpredictability of the rock stratum movement are obviously increased, and problems are raised on the data processing precision and early warning sensitivity of a monitoring system. The mine pressure multi-parameter monitoring generally covers physical information of various dimensions such as hydraulic support resistance, microseismic energy, acoustic emission frequency and the like. By collecting and comprehensively evaluating the multi-source heterogeneous sensing data in real time, a dynamic sensing system for the surrounding rock state of the working face can be constructed, and auxiliary decision support is provided for mine disaster prevention and control. The traditional monitoring system often depends on single physical quantity threshold judgment logic, and when the underground severe electromagnetic interference and equipment loss are faced, a large number of false alarms are easy to generate, so that the public confidence of early warning information is influenced. The traditional statistical analysis method is difficult to extract deep features in mass data, so that the system has insufficient perception of weak abnormal signals before rock stratum instability, and cannot capture the nonlinear coupling rule of multiple parameters in space-time dimension. The existing system lacks the capability of carrying out bionic level fusion processing on multi-source information, and the real stress evolution track is difficult to reflect by simple numerical superposition, so that the early warning advance is limited and the reliability is low. Therefore, a multi-parameter intelligent monitoring and early warning system for the mine pressure of the working face is needed. Disclosure of Invention The invention aims to provide a multi-parameter intelligent monitoring and early warning system for the mine pressure of a working face, which can solve the problems of difficult feature extraction and high false alarm rate in the background technology. In order to achieve the above purpose, the technical scheme adopted by the invention is as follows: The utility model provides a working face ore deposit pressure multiparameter intelligent monitoring and early warning system, includes multisource sensing data acquisition device, pulse sequence coding unit, bionical pulse neural network early warning model, dynamic attention regulation and control module and early warning information output terminal, wherein: The multi-source sensing data acquisition device is configured to acquire original physical signals of hydraulic support resistance, microseism energy and acoustic emission frequency of the fully-mechanized coal face in real time, convert the original physical signals into digital signals and then transmit the digital signals to the pulse sequence coding unit; the pulse sequence coding unit is configured to receive the digital signals from the multi-source sensing data acquisition device, and map different types of heterogeneous data into pulse event sequences with time stamps according to the change rate and the amplitude characteristics of each physical quantity, so as to form a space-time coding format similar to biological nerve signals; The bionic pulse neural network early warning model is configured to receive a multichannel pulse event sequence generated by the pulse sequence coding unit, and perform nonlinear integration processing on the pulse event sequence based on a calculation rule simulating a biological neuron discharge mechanism so as to identify a multiparameter coupling mode corresponding to a rock stratum instability precursor; The dynamic attention regulation module is configured to adaptively adjust weight distribution of each input channel according to the space-time distribution characteristic of the current input pulse stream in the running process of the bionic pulse neural network early warning model, strengthen attention to key precursor signals and inhibit the influence of noise interference components; The early warning information output terminal is configured to receive an early warning judgment result output by the bionic impulse neural network