Intelligent tunneling control method and system for coal mine

Abstract

The invention provides a coal mine intelligent tunneling control method and system, and belongs to the technical field of coal mine intelligent tunneling control. In order to solve the technical problems that the current coal mine tunneling operation depends on manual operation, the labor intensity is high, the tunneling efficiency is low and the safety risk is high, the adopted technical scheme is that a control system of the tunneling machine is started, all control modules and parts are initialized, and data of a tunneling face and the surrounding environment are collected in real time through a sensor module; the control data processing module is used for fusing and analyzing the collected data and constructing an environment model; controlling the intelligent decision-making module to respectively carry out coal rock identification, tunneling path planning and self-adaptive adjustment of tunneling parameters according to the constructed environment model, and then outputting a decision-making judgment result; the execution control module automatically controls the heading machine to act according to the output decision judgment result, autonomous heading is carried out, and an operator controls the man-machine interaction module to carry out monitoring or intervention at the same time; the device is applied to coal mine tunneling operation.

Description

TECHNICAL FIELD

[0001] The application provides a coal mine intelligent tunneling control method and system, and belongs to the technical field of coal mine intelligent tunneling control. BACKGROUND

[0002] The existing coal mine tunneling equipment relies on manual operation, and has problems of high labor intensity, poor working environment, low tunneling efficiency and high safety risk. With the increasing demand for intelligentization of coal mines, how to realize intelligent control of tunneling equipment and improve tunneling efficiency and safety has become a problem to be solved in the industry. SUMMARY

[0003] In order to solve the technical problems in the background art, the technical scheme adopted by the application is to provide a coal mine intelligent tunneling control method, which comprises the following control steps:

[0004] Step one: start the control system of the tunneling machine, initialize each control module and component, and collect tunneling face and surrounding environment data in real time through the sensor module;

[0005] Step two: the control data processing module analyzes the data collected in step one, uses extended Kalman filtering to fuse laser radar, inertial navigation and coal rock identification data, and constructs an environment model;

[0006] Step three: the control intelligent decision module outputs a decision-making result after processing coal rock identification, tunneling path planning and adjustment and optimization of tunneling machine operation parameters according to the constructed environment model;

[0007] Step four: the execution control module automatically controls the action of the tunneling machine according to the output decision-making result, performs autonomous tunneling, and the operator controls the man-machine interaction module to monitor or intervene at the same time.

[0008] The specific method for coal rock identification processing in step three is:

[0009] The formation of rock fragments and the load data of the tunneling machine under various working conditions are obtained through experiments, including:

[0010] The normal force F is calculated N And the rolling force F R The relationship between the penetration, rock compressive strength UCS and cutting force is obtained by a polynomial fitting method, and the linear analysis relationship between the cutter load and the cutter thrust and torque is combined to obtain the physical relationship between the thrust or torque and the compressive strength or penetration, which are respectively denoted as Th and Tor, and the calculation formula is:

[0011] ;

[0012] Wherein, p is the penetration, n is the total number of cutters, and ri is the distance from the i-th cutter to the center of the tool head;

[0013] When the cutter spacing is s, as the penetration p increases, the coal rock crack continues to expand downward and on both sides, when the penetration p increases to a critical value, the cracks generated by adjacent cutters of the roadheader on the coal rock connect with each other, forming rock fragments, it is defined that under the condition that the cutter spacing s is the same, there is a critical penetration s / p value that produces rock fragments, when the penetration p is greater than the value, the rock between the cutters can be successfully broken, and the calculation formula of the critical penetration s / p is:

[0014] ;

[0015] Based on the above formula, the rock breaking critical penetration rule is obtained, which is used as a rock breaking physical constraint condition for identifying coal rock.

[0016] The specific method for identifying coal rock based on the rock breaking physical constraint condition is:

[0017] The relationship between the equation constraint of the rock breaking physical constraint condition, the roadheader thrust or torque, the rock compressive strength and the penetration, and the inequality constraint of the critical penetration are introduced into the DNN deep neural network respectively, wherein:

[0018] The inequality constraint is used to evaluate the rationality of the sample, the sample meeting the inequality constraint condition is regarded as a normal sample, and the sample violating the inequality constraint condition is defined as an abnormal sample, and the weight calculation formula of the abnormal sample is:

[0019] ;

[0020] Wherein, Loss i and Loss i ′ represent the contribution of the i-th normal or abnormal sample to the loss function, n is the total number of samples, y′ and y i are the network calculation result and the actual result of the i-th sample respectively, μ1 is the weight of the sample control constraint, and its value is between 0 and 1;

[0021] The equality constraint represents the quantitative relationship between parameters, and the quantitative relationship is a positive reference for data mining, the variable calculated by the equality constraint is given a weight μ2 and added to the calculated value for modification;

[0022] The expression for introducing the equality constraint into the DNN deep neural network is:

[0023] ;

[0024] Wherein, Loss i is the part of the loss function caused by the i-th sample, n is the total number of samples, yi and y ic represent DNN deep neural network calculation value and equality constraint condition, y' is actual value.

[0025] The specific method of introducing the rock breaking physical constraint condition into the DNN deep neural network is as follows:

[0026] Three constraint conditions are respectively established in the DNN deep neural network: CP critical permeability model, TH measured thrust model and TOR torque data model, wherein the CP model is an inequality constraint, and the TH and TOR models are equality constraints;

[0027] The calculation formula of the loss function based on the above constraint conditions is as follows:

[0028] ;

[0029] Wherein, p i is the measured penetration of the ith sample, p lim is the critical penetration of the rock block, E Th and E Tor represent the calculation errors of the cutter thrust and torque respectively, Th and Tor represent the measured thrust and torque data of the sample respectively, Th' and Tor' represent the mapping calculation results based on the physical rules and data mining respectively, Th p and Tor p represent the thrust and torque calculated by the TH and TOR rules respectively, µ1, µ2, µ3 represent the weights of the three constraint conditions respectively;

[0030] Finally, a rock-machine mapping model based on physical rules and data mining is constructed under the constraint condition of physical rules.

[0031] The specific method of adjusting and optimizing the running parameters of the heading machine in step three is as follows:

[0032] According to the physical rules and the rock-machine mapping model, and the rated values of the thrust, torque, belt conveying capacity and critical penetration of the heading machine, the constraint conditions are determined to obtain the rated value range of the running parameters of the heading machine as the constraint condition of the optimized operation;

[0033] The total heading cost of the heading machine is calculated as follows:

[0034] The total heading cost is specifically the sum of the time-related cost Cs and the tool-related cost Cc, wherein the time-related cost Cs is obtained by multiplying the time by the daily heading cost, and the tool-related cost Cc is obtained by multiplying the number of tools by the unit tool cost, and the calculation formula is as follows:

[0035] ;

[0036] Wherein, C1 is the daily cost of the heading machine construction, including heading machine equipment rental cost, employee wages, accommodation and other expenses; C2 is the single cutter cost, L, A, PR and H f are the driving distance, cross-sectional area, penetration rate and cutter life, respectively;

[0037] μ is the utilization rate of the heading machine, i.e. the ratio of daily working time to total time; t is the total time per day;

[0038] The penetration rate PR and the heading machine life are functions of the working parameter, wherein the penetration rate PR is the product of the working parameter P and the rotational speed, and the calculation formula is:

[0039] PR = p rpm;

[0040] The expression of the rock-mechanical mapping model is:

[0041] ;

[0042] Wherein, Mu and Vi refer to indicators related to the heading machine mud and vibration;

[0043] Based on the obtained rated value range of the heading machine operating parameters, the established objective function model is globally optimized and searched to obtain the optimal operating parameters.

[0044] To realize the intelligent heading control method for coal mines, a control system is adopted, comprising:

[0045] A sensor module for real-time acquisition of geological, equipment posture and environmental parameters of the heading face, the sensor module comprising a laser radar, an inertial navigation unit, a coal and rock identification sensor and an environmental monitoring sensor;

[0046] A data processing module for fusing, filtering and feature extraction of sensor data, generating a heading environment model;

[0047] An intelligent decision module for analyzing the environment model based on deep learning or expert system algorithm, generating heading path planning and operation parameters, the intelligent decision module adopting a convolutional neural network to classify and identify the front coal and rock image, and automatically adjusting the heading parameters according to the identification result;

[0048] An execution control module for automatically controlling the propulsion, cutting and supporting actions of the heading machine according to the decision result;

[0049] A human-computer interaction module for realizing system state display, parameter setting and remote monitoring.

[0050] The advantages of this invention compared to existing technologies are as follows: This invention provides an intelligent tunneling control scheme for coal mines. During intelligent tunneling control, constraints for the control parameter optimization model are established through physical rules. Rock-breaking physical constraints are introduced to establish a dual-drive rock-machine mapping model. By adjusting and optimizing the tunneling machine's operating parameters, the optimal decision judgment result is output to control the tunneling machine's operation. This invention can automatically adjust tunneling parameters according to changes in tunneling geology, adapt to complex coal and rock environments, realize autonomous operation of tunneling equipment, reduce manual intervention, and improve tunneling efficiency and operational safety. Attached Figure Description

[0051] The present invention will be further described below with reference to the accompanying drawings:

[0052] Figure 1 This is a flowchart illustrating the steps of the intelligent tunneling control method for coal mines according to the present invention. Detailed Implementation

[0053] like Figure 1 As shown, the intelligent tunneling control method for coal mines provided by this invention mainly includes the following control steps:

[0054] Step 1: Start the control system of the tunneling machine, initialize each control module and component, and collect data on the tunnel face and surrounding environment in real time through the sensor module;

[0055] Step 2: The control data processing module analyzes the data collected in Step 1, and uses extended Kalman filtering to fuse the lidar, inertial navigation, and coal and rock identification data to construct an environmental model;

[0056] Step 3: The intelligent decision-making module, based on the constructed environmental model, performs coal and rock identification, tunneling path planning, and adjustment and optimization of tunneling machine operating parameters, and then outputs the decision judgment results.

[0057] Step 4: The execution control module automatically controls the tunneling machine's actions based on the output decision judgment results, and performs autonomous tunneling. The operator controls the human-machine interaction module to monitor or intervene at the same time.

[0058] Furthermore, the intelligent decision-making module internally provides coal and rock identification and parameter adaptive optimization algorithms, as well as multi-sensor data fusion algorithms based on deep neural networks (DNN), to calculate the optimal operating parameters of the tunneling machine, specifically including:

[0059] (1) The physical rules of rock breaking were calculated;

[0060] Linear cutting tests and numerical simulations are two commonly used methods for studying rock breaking mechanisms and physical laws. Numerical simulations are more likely to construct various rock mass conditions, while linear cutting tests more closely resemble the actual rock breaking process. Each has its advantages and they can complement each other well.

[0061] The present application adopts numerical simulation method and linear cutting test to simulate and test the rock breaking process under different cutting parameters (penetration, cutter spacing, etc.) and rock compressive strength. Specifically, in order to explore the breaking process and the evolution of the roadheader cutter load under different working conditions, a plurality of different strength rock masses are used in the linear cutting test and numerical simulation, and a plurality of different penetration and cutter spacing combinations are set;

[0062] Through the test, the formation of rock broken pieces and the roadheader load data under various working conditions can be obtained, including the calculated normal force FN and rolling force FR, and the relationship between penetration, UCS and cutting force can be obtained by polynomial fitting method. Combined with the linear analysis relationship between cutter load and cutter thrust and torque, the physical relationship between thrust / torque and compressive strength / penetration can be obtained, which are respectively denoted as Th and Tor correlation, and the expression is:

[0063] ;

[0064] Wherein, p is the penetration, n is the total number of cutters, r i is the distance from the ith cutter to the center cutter;

[0065] When the cutter spacing is s, with the increase of the penetration p, the cracks continue to expand downward and on both sides, and when the penetration p increases to a certain value, the cracks generated by adjacent cutters are connected with each other to form rock broken pieces.

[0066] In addition, under the condition that the cutter spacing s is the same, there is a critical penetration for generating rock broken pieces, and when the penetration p is greater than the value, the rock between the cutters can be successfully broken. It has been proved through tests that under the same rock mass condition, the value of critical s / p is almost a constant, usually between 10-20, and the critical s / p value can also be determined by regression method, and the expression is:

[0067] ;

[0068] Based on the above formula, the CP (critical penetration) rule is obtained, which is used to establish the constraint condition of the control parameter optimization model.

[0069] (2) The above constraint condition is incorporated into the DNN deep neural network;

[0070] The above rock breaking physical constraints can be divided into two categories, namely equality and inequality. Specifically, the relationship between the roadheader thrust / torque, rock compressive strength and penetration is an equality constraint, and the calculated critical penetration (i.e. the minimum penetration required for rock broken pieces to form) is an inequality constraint.

[0071] The methods of introducing the above two types of constraints into DNNs are different. First, inequality constraints are used to evaluate the rationality of samples. Samples that meet the inequality constraint conditions are considered normal samples and have larger weights during training; while samples that violate the inequality constraint conditions are considered abnormal samples and have lower weights during DNN training:

[0072] wherein the weight calculation formula of the abnormal sample is:

[0073] ;

[0074] wherein Loss i and Loss i ′ represent the contribution of the i-th normal or abnormal sample to the loss function, n is the total number of samples, y′ and y i are the network calculation result and the actual result of the i-th sample respectively, μ1 is the weight of the sample control constraint, and its value is between 0 and 1.

[0075] Reducing the weight of abnormal samples will correspondingly increase the influence of normal samples on the rock machine mapping based on physical rules and data mining.

[0076] Unlike inequality constraints, equality constraints represent quantitative relationships between parameters, which are usually positive references for data mining. The variables calculated by the equality constraint are assigned a weight μ2 and added to the calculated value to modify it.

[0077] Through continuous training, the difference between the mapping based on physical rules and data mining and the physical is continuously reduced, which can ensure accuracy and reduce the risk of overfitting; the expression for introducing equality constraints into DNNs is:

[0078] ;

[0079] wherein Loss i is the part of the loss function caused by the i-th sample, n is the total number of samples, y i and y ic represent the DNN calculation value and the equality constraint condition, and y′ is the actual value.

[0080] As mentioned earlier, it is necessary to introduce rock breaking physical constraints into DNNs, and three constraint conditions are established: CP critical permeability model, TH measured thrust model, and TOR torque data model, wherein the CP model is an inequality constraint, and the TH and TOR models are equality constraints.

[0081] The calculation formula of the loss function based on the above constraint conditions is:

[0082] ;

[0083] where p i is the measured penetration of the ith sample, p lim is the critical penetration of the rock fragments, E Th and E Tor represent the calculation errors of the cutter thrust and torque, respectively, Th and Tor represent the measured thrust and torque data of the sample, respectively, Th' and Tor' represent the mapping calculation results based on the physical rules and data mining, Th p and Tor p represent the thrust and torque calculated by the TH and TOR rules, respectively, and µ1, µ2, µ3 represent the weights of the three constraint conditions.

[0084] (3) Establish a double-drive rock-mechanical mapping model;

[0085] After introducing the rock breaking physical constraints into the DNN, the network structure and hyperparameters need to be determined according to the characteristics of the data set. Specifically, the ReLU activation function is used between the hidden layers of the DNN, and the number of neurons in the input layer and the output layer is set to the number of input parameters and output targets, respectively. To build a DNN, the number of neurons in the hidden layer, the learning rate, the structure loss weight, and the weight of the physical rule constraints need to be determined. Through repeated experiments, multiple DNNs with different hyperparameters and structures are constructed, and then the training accuracy of each network model is compared to select the network with the highest accuracy. Finally, a rock-mechanical mapping model based on physical rules and data mining is constructed with the physical rules as the constraint conditions.

[0086] (4) Optimize the tunneling operation parameters;

[0087] The tunneling efficiency is represented by the time-dependent cost affected by the penetration rate, and the tunneling consumption is represented by the tunneling machine-dependent cost affected by the service life of the tunneling machine. By reasonably setting the penetration and the number of revolutions per minute of the full-face hard rock tunneling machine, the tunneling speed and the service life of the tunneling machine can be effectively improved, and the operation cost of the tunneling machine can be reduced.

[0088] Specifically, the penetration and the number of revolutions per minute of the tunneling machine are taken as the optimization variables, and the time-dependent cost and the tunneling machine-dependent cost are taken as the optimization objectives of the total excavation cost of the tunneling machine.

[0089] In addition, to ensure the safety of the tunneling machine, the penetration parameters of the tunneling machine, such as the thrust, the torque, the belt conveying capacity, and the critical penetration for generating rock fragments, must be controlled within the rated value range, and they are taken as the constraint conditions of the above optimization.

[0090] The tunneling machine operation parameter decision optimization method provided by the present application based on multi-constraint and multi-objective optimization specifically includes:

[0091] 1) The constraints are determined according to the physical rules and the rock-machine mapping based on data mining and the rated values of the machine operating parameters (including thrust, torque, belt conveying capacity and critical penetration). The four constraints ensure that the above-mentioned machine indicators are within the rated range, so that the machine can efficiently and safely break rocks.

[0092] 2) The total tunneling cost Ct can be calculated as the sum of the time-related cost Cs and the tool-related cost Cc, where the time-related cost Cs is derived from the time multiplied by the daily tunneling cost, and the tool-related cost Cc can be derived from the number of tools multiplied by the unit tool cost, as follows:

[0093] ;

[0094] where C1 is the daily cost of the machine construction, mainly including the machine equipment rental cost, employee wages, accommodation and other expenses; C2 is the single tool cost, L, A, PR and H f are the driving distance, cross-sectional area, penetration rate and tool life, respectively; μ is the utilization rate of the machine, i.e. the ratio of daily working time to total time; t is the total time per day (24 hours / day). Among these parameters, L and A are constants. The penetration rate and the machine life are functions of the working parameters, where PR is the product of the working parameter P and the rotational speed, i.e. PR = p rpm.

[0095] The expression of the double-drive rock-mechanical mapping model is:

[0096] ;

[0097] where Mu and Vi refer to the machine mud and vibration-related indicators.

[0098] Using the feasible range of operating parameters obtained in step 1 above, a global optimization search is performed on the objective function model established in step 2, which can not only meet the safety requirements of the machine tunneling, but also minimize the tunneling cost.

[0099] Further, the present application aims to optimize the operation parameters of the heading machine. First, the reasonable range of these operation parameters under different rock conditions needs to be determined to ensure normal operation without causing mechanical damage. If the operation parameters (penetration and revolutions per minute) are globally optimized without considering the feasible parameter range, two situations can occur. First, the operation parameters can exceed the rated values, resulting in abnormal damage to the cutting machine, slipping or clogging of the belt conveyor. The other situation is that inappropriate operation parameters can cause it difficult to form uniform rock fragments during rock cutting, thus causing serious damage to the cutting machine and shortening its service life. Therefore, considering the constraints and calculating the reasonable range of operation parameters is crucial to achieve multi-scheme and improve the reliability and stability of optimization. In addition, parameters such as penetration should meet the requirements of the heading machine for breaking rock to ensure that rock fragments are produced rather than powder. According to the above analysis, the penetration of the heading machine should be greater than the critical penetration corresponding to the current rock mass condition, otherwise the rock mass cannot be broken into fragments, resulting in low efficiency.

[0100] In addition, key parameters such as thrust and torque of the cutter head should also be considered, which will affect the key structures of the heading machine (such as the cutter head, cutter head and main bearing). When the thrust and torque exceed the rated values, the heading machine will be overloaded, thus damaging the heading machine and the milling head; another problem is that when the rock mass breaks or the penetration is too large, the amount of sludge will increase significantly, causing the belt conveyor to slip and clog.

[0101] In summary, the four constraints for selecting the operation parameters of the heading machine include: rated thrust, rated torque, rated belt conveying capacity, and critical penetration of rock block preparation. Generally speaking, the critical penetration constraint is used to achieve a good rock breaking mode and high excavation efficiency; thrust, torque, and belt conveying capacity constraints are used to maintain the integrity of the cutter head, milling cutter, and belt conveyor, which control the operation parameters such as penetration and cutter head revolutions per minute. Based on the above four constraints, a double-drive rock-mechanical mapping model can be constructed.

[0102] The above gives the "contour lines" of the penetration and the number of revolutions per minute relative to the constraint condition extreme value. In addition, considering that the rock-machine mapping obtained by the learning-based method is usually implicitly expressed, therefore, at each constraint condition, with reference to the control parameter range and step of the actual boring machine boring, the feasible range of boring machine operating parameters is calculated; assuming that the rock mass parameters, soil geometry and vibration signals are known, the operating parameter combination (i.e. the penetration per minute and the boring machine revolutions per minute) can be evaluated, and whether these parameters meet the constraint requirements is determined according to the known rated value and rock-machine mapping; taking the thrust constraint as an example, according to the known rock-machine mapping, the corresponding thrust value of each working parameter combination is calculated, if the calculated value exceeds the rated thrust, it is considered that the corresponding operating parameters do not meet the thrust constraint condition, on the contrary, if the thrust value obtained by the operating parameter combination is lower than the rated value, it is considered that the constraint condition is met.

[0103] Through the above data processing process, the operating parameter combination that meets the thrust constraint condition can be obtained, and the operating parameter combination that meets all four constraint conditions is reasonable, so these parameters constitute the feasible region of the operating parameters, in this region, the optimal boring parameters under the premise of safe boring can be obtained through optimization search.

[0104] The application also provides a coal mine intelligent boring control system, which realizes autonomous operation of boring equipment, improves boring efficiency and operation safety through multi-sensor information fusion, intelligent decision and automatic control, and the control system specifically comprises:

[0105] Sensor module: including laser radar, inertial navigation, coal rock identification sensor, environment monitoring sensor, etc., for real-time acquisition of boring face geology, equipment attitude and environment parameters, wherein the sensor interface includes:

[0106] Laser radar: Ethernet / RS422 interface, data format is point cloud data packet;

[0107] IMU (inertial measurement unit): CAN bus or RS232 interface, data format is acceleration and angular velocity;

[0108] Camera: USB3.0 or GigE interface, data format is YUV / RGB image stream;

[0109] Coal rock identification sensor: custom interface, supporting digital / analog signal input.

[0110] Data processing module: fusing, filtering and feature extracting the sensor data to generate a boring environment model; the controller interface on which the data processing module is installed includes:

[0111] Industrial PC or embedded controller, with multi-channel Ethernet, CAN, USB interface;

[0112] PLC interface: for communication with the execution mechanism of the tunneling machine, hydraulic, motor, etc., supporting Modbus, Profibus, etc. Industrial protocols;

[0113] Human-computer interaction terminal: touch screen or host computer, Ethernet / RS232 interface.

[0114] Intelligent decision module: based on deep learning or expert system algorithm, analyzes the environment model, generates tunneling path planning and operation parameters.

[0115] Execution control module: according to the decision result, automatically control the propulsion, cutting, support and other actions of the tunneling machine; the execution mechanism interface includes:

[0116] Hydraulic pump / motor control: through PLC output digital / analog signal;

[0117] Support system control: relay or IO module interface;

[0118] Safety emergency stop and alarm: hardware IO direct connection to ensure safety.

[0119] Human-computer interaction module: realize system state display, parameter setting and remote monitoring.

[0120] The above modules work:

[0121] Sensor module real-time acquisition of tunneling face geological, equipment posture and environmental parameters;

[0122] Data processing module for sensor data fusion, filtering and feature extraction, generate tunneling environment model;

[0123] Intelligent decision module based on deep learning or expert system algorithm, analyzes the environment model, generates tunneling path planning and operation parameters of intelligent decision module, uses convolutional neural network to classify and identify the front coal and rock image, and automatically adjusts the tunneling parameters according to the identification result;

[0124] Execution control module according to the decision result, automatically control the propulsion, cutting, support and other actions of the tunneling machine, realize autonomous tunneling;

[0125] Human-computer interaction module is used for display system state, parameter setting and remote monitoring, for operator monitoring and intervention.

[0126] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions recorded in the above embodiments can be modified, or some or all of the technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A coal mine intelligent tunneling control method, characterized in that: The control steps include the following: Step 1: Start the control system of the tunneling machine, initialize each control module and component, and collect tunneling face and surrounding environment data in real time through the sensor module; Step 2: The control data processing module analyzes the data collected in step 1, fuses the laser radar, inertial navigation, and coal rock identification data using extended Kalman filtering, and constructs an environment model; Step 3: The control intelligent decision-making module processes coal rock identification, tunneling path planning, and adjustment and optimization of tunneling machine operating parameters based on the constructed environment model, and outputs decision-making results; Step 4: The execution control module automatically controls the actions of the tunneling machine based on the output decision-making results to perform autonomous tunneling, and the operator controls the human-machine interaction module to monitor or intervene at the same time.

2. The intelligent tunneling control method for coal mines according to claim 1, characterized in that: The specific method for coal rock identification processing in step 3 is: Through experiments, the formation of rock fragments and the load data of the tunneling machine under various working conditions are obtained, including: The calculated normal force F N And the rolling force F R The relationship between the penetration, the rock compressive strength UCS and the cutting force is obtained by a polynomial fitting method, and the linear analysis relationship between the cutter load and the cutter thrust and torque is combined to obtain the physical relationship between the thrust or torque and the compressive strength or penetration, respectively denoted as Th and Tor, and the calculation formula is: ； where p is the penetration, n is the total number of tools, r i is the distance of the ith tool to the center tool head; When the cutter spacing is s, as the penetration p increases, the coal rock cracks continue to expand downward and to the sides, and when the penetration p increases to a critical value, the cracks generated by adjacent cutters of the tunneling machine connect with each other to form rock fragments. It is defined that under the condition that the cutter spacing s is the same, there is a critical penetration s / p value that produces rock fragments, and when the penetration p is greater than this value, the rock between the cutters can be successfully broken. The calculation formula of the critical penetration s / p is: ； Based on the above formula, the rock breaking critical penetration rule is obtained as a rock breaking physical constraint condition for identifying coal rock.

3. The intelligent tunneling control method for coal mines of claim 2, characterized in that: The specific method for identifying coal rock based on the rock breaking physical constraint condition is: The relationship between the equal constraint of the rock breaking physical constraint condition, the tunneling machine thrust or torque, the rock compressive strength, and the penetration, and the inequality constraint of the critical penetration are introduced into the DNN deep neural network, respectively, wherein: The inequality constraint is used to evaluate the rationality of the sample. The sample that meets the inequality constraint condition is regarded as a normal sample, and the sample that violates the inequality constraint condition is defined as an abnormal sample. The weight calculation formula of the abnormal sample is: ； wherein Loss i and Loss i ′ denotes the contribution of the i-th normal or abnormal sample to the loss function, n is the total number of samples, y' and y i are the network calculated result and the actual result of the i-th sample, respectively, and μ1 is the weight of the sample control constraint, the value of which is between 0 and 1. The equal constraint represents the quantitative relationship between parameters. The quantitative relationship is a positive reference for data mining. The variable calculated by the equal constraint is assigned a weight μ2 and added to the calculated value for modification. The expression for introducing the equal constraint into the DNN deep neural network is: ； wherein Loss i is the loss function part caused by the i-th sample, n is the total number of samples, y i and y ic represent the DNN deep neural network calculation value and the equal constraint condition, and y′ is the actual value.

4. The intelligent tunneling control method for coal mines of claim 3, characterized in that: The specific method for introducing the rock breaking physical constraint condition into the DNN deep neural network is: Three constraint conditions are established in the DNN deep neural network: CP critical penetration model, TH measured thrust model, and TOR torque data model. The CP model is an inequality constraint, and the TH and TOR models are equal constraints. The calculation formula of the loss function based on the above constraint conditions is: ； where p i is the measured penetration of the ith sample, p lim is the critical penetration of the rock fragment, E Th and E Tor represent the calculation errors of the cutter thrust and torque, respectively, Th and Tor represent the measured thrust and torque data of the sample, respectively, Th' and Tor' represent the mapping calculation results based on physical rules and data mining, respectively, Th p and Tor p represent the thrust and torque calculated by the TH and TOR rules, respectively, and µ1, µ2, µ3 represent the weights of the three constraint conditions, respectively. Finally, a rock-machine mapping model based on physical rules and data mining is constructed with physical rules as constraint conditions.

5. The intelligent tunneling control method for coal mines according to claim 4, characterized in that: The specific method for adjusting and optimizing the operating parameters of the tunneling machine in step 3 is: According to physical rules and rock-mechanical mapping model, and the rated values of the thrust, torque, belt conveying capacity, and critical penetration of the boring machine, constraint conditions are determined, and a rated value range of the operating parameters of the boring machine is obtained as a constraint condition for optimization operation; The total boring cost of the boring machine is calculated as: The total boring cost is specifically a sum of a time-related cost Cs and a tool-related cost Cc, wherein the time-related cost Cs is obtained by multiplying time by a daily boring cost, the tool-related cost Cc is obtained by multiplying the number of tools by a unit tool cost, and a calculation formula is: ； Wherein, C1 is the daily cost of the tunneling machine construction, including the tunneling machine equipment rental cost, employee salary, accommodation and other expenses; C2 is the single cutter cost, L, A, PR and H f are the driving distance, cross-sectional area, penetration rate and cutter life, respectively; μ is the utilization rate of the boring machine, i.e., the ratio of daily working time to total time; t is the total daily time; The penetration rate PR and the boring machine life are functions of the working parameters, wherein the penetration rate PR is the product of the working parameter P and the rotational speed, and a calculation formula is: PR = p rpm; The expression of the rock-mechanical mapping model is: ； Wherein, Mu and Vi refer to indexes related to the boring machine mud and vibration; Based on the obtained rated value range of the operating parameters of the boring machine, a global optimization search is performed on the established objective function model, and optimal operating parameters are obtained.

6. A control system for implementing the coal mine intelligent tunneling control method of claim 1, characterized in that: It comprises: A sensor module for real-time acquisition of geological, equipment posture and environmental parameters of the boring face, the sensor module comprising a laser radar, an inertial navigation unit, a coal and rock identification sensor and an environmental monitoring sensor; A data processing module for fusing, filtering and feature extraction of sensor data, and generating a boring environment model; An intelligent decision module for analyzing the environment model based on deep learning or expert system algorithm, generating boring path planning and operation parameters, the intelligent decision module using a convolutional neural network to classify and identify the front coal and rock image, and automatically adjusting the boring parameters according to the identification result; An execution control module for automatically controlling the propulsion, cutting and supporting actions of the boring machine according to the decision result; A human-computer interaction module for realizing system state display, parameter setting and remote monitoring.

Images