Coal-pillar-free self-forming roadway intelligent design method and system based on machine learning

Abstract

The invention discloses a coal-pillar-free self-formed roadway intelligent design method and system based on machine learning, and relates to the technical field of coal mining, and the method comprises the following steps: S1, building a 110 construction method intelligent design platform; s2, building a surrounding rock geomechanical model, and monitoring surrounding rock deformation and pressure change in the model; s3, the surrounding rock stability and rock goaf breaking and swelling sufficient filling are judged, and decisions such as support parameters and equipment type selection are completed; s4, establishing a working face digital model; s5, anchor-cutting-support-protection-control integrated self-adaptive collaborative intelligent control is formed through a robot cluster, and analysis is carried out through a working face digital platform; and S6, performing optimization to form a decision optimal model of intelligent coal mining of the 110 construction method. According to the intelligent design method and system for the coal-pillar-free self-forming roadway based on machine learning, the labor cost, the equipment investment and the material consumption are reduced, the total construction cost is reduced, and innovation and development of the underground engineering technology in China are promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of coal mining, and in particular to an intelligent design method and system for a self-forming lane without coal pillars based on machine learning. Background Art

[0002] At present, my country's intelligent mining technology is in its early stages, mainly focusing on the application of the 121 method. Although the robot system can complete more construction tasks, improve the production efficiency of mines, replace traditional manual operations, and reduce dependence on technical workers, the intelligent 121 method mining environment still faces problems such as coal pillars in the mining area and large lanes, low coal recovery rate in the mine, waste of resources, low lane excavation rate, and tight mining and excavation continuity.

[0003] The coal pillar-free self-forming tunneling technology, namely the 110 method of cutting the top to form the tunnel, is one of the effective technologies to maintain the sustainable development of my country's coal resources. It can reduce or avoid the above-mentioned problems to a large extent. The 110 method innovatively utilizes the mine pressure, the crushing and expansion characteristics of the roof rock mass and the tunnel space to realize automatic tunneling and coal pillar-free mining. This method not only eliminates tunnel excavation and reduces safety risks, but also uses the top cutting and pressure relief technology to cut off the pressure transmission between the tunnel and the goaf, thereby improving the stability of the tunnel. Compared with the traditional 121 method, the 110 method eliminates the stress concentration caused by coal pillars and weakens the periodic pressure on the working face, which is beneficial to prevent and control disasters such as impact ground pressure and rock bursts, and ensure safe production in mines.

[0004] In order to realize the comprehensive intelligence of mining, it is necessary to conduct in-depth research on the above issues. Only by integrating the concept of intelligent mining with the 110 method can we truly solve the various problems in the mining process. At present, there are problems in how to perceive the parameters of the designed mine in the design; how to judge whether the design method proposed by the system is executable; how to make analysis and decisions based on the actual conditions of the working face; how to optimize the tunneling design according to real-time monitoring data when executing control. Summary of the invention

[0005] The purpose of the present invention is to provide an intelligent design method and system for a self-forming lane without coal pillars based on machine learning, so as to solve the problems of waste of resources and low lane excavation rate.

[0006] To achieve the above object, the present invention provides an intelligent design method for self-forming lanes without coal pillars based on machine learning, comprising the following steps:

[0007] S1. Program the design of all projects that have implemented the 110 construction method, conduct machine learning on the data in the database on the platform, and establish an intelligent design platform for the 110 construction method;

[0008] S2. By designing the surrounding rock geomechanical parameters, mine information, and stope conditions of the working face coal mine, a surrounding rock geomechanical model is constructed. At the same time, the deformation and pressure changes of the surrounding rock in the model are monitored, and the monitoring results are compared with the actual on-site monitoring data;

[0009] S3. Based on the model data and machine learning content, make judgments on the stability of the surrounding rock and the sufficient filling of the rock goaf, and independently complete the support parameters and equipment selection decisions based on the judgment results;

[0010] S4. Establish a digital model of the working face and a two-lane support model, as well as a digital twin model of the top cutting, anchor drilling, support, rock blocking, and top control of the completed lane, to achieve real-time process mapping of the completed lane, and integrate a digital intervention model that integrates geological, panel, disaster, and succession guidance factors;

[0011] S5: According to the decision content of S3, it is transmitted to the robot cluster to complete intelligent top cutting, drilling and anchoring, rock blocking and temporary support at the working face, and the application effect is transmitted to the working face digital platform in real time, and its effect is evaluated and fed back to the perception information section for analysis;

[0012] S6. Based on the analysis results of the perception information section, the production process is deduced, and combined with the production conditions fed back by the equipment, the planning execution effect and production process are evaluated and the model is optimized through autonomous learning, so as to achieve the optimal decision-making model for intelligent coal mining of the 110 method, optimize the working face and design safety assurance.

[0013] Preferably, S1 comprises the following steps:

[0014] S11. Process the data from the previous 110 method application cases. Data processing includes data cleaning, feature engineering and selection, oversampling, data analysis and data standardization;

[0015] S12. Use the BP neural network model for training and prediction. During the training process, the error between the predicted value and the true value is calculated, and the error is passed back through the network layer by layer. The weights and biases are adjusted according to the error of each layer, and the network parameters are continuously updated iteratively until the predetermined training goal is achieved.

[0016] Preferably, in S11, synthetic minority oversampling technology SMOTE is used to optimize sample distribution characteristics, oversample the original data set, and the method for standardizing the oversampled data is as follows:

[0017]

[0018] in, represents the sample mean, s represents the sample standard deviation, and n represents the sample size.

[0019] Preferably, in S12, a particle swarm algorithm is used to optimize the initial weights and thresholds of the neural network. There are N particles and a Z-dimensional solution space. Initially, a population consisting of N particles is randomly initialized. After calculating the individual extreme value and the global extreme value, the speed and position of the particles are updated according to the following formula:

[0020]

[0021] u t+1 =u t +u t+1 ;

[0022] Among them, ω t 、c 1 t 、c 2 t They represent the extent to which the particle motion is affected by the local optimal solution and the global optimal solution, respectively. t is the velocity vector, u t is the position vector, p j is the individual extreme value vector, p k is the global extreme value vector.

[0023] Preferably, in S12, a particle swarm algorithm is used to optimize the parameters of the initial weights and thresholds of the neural network, and a 6-fold cross validation method is used to train and predict the model.

[0024] Preferably, the model evaluation indicators are the mean absolute error EMA, the root mean square error ERMS and the determination coefficient R 2 To comprehensively evaluate the prediction effect of the model, the calculation method is as follows:

[0025]

[0026] Preferably, in S2, the designed working face is modeled using the three-dimensional modeling software FLAC3D, and the deformation of the surrounding rock and the change of the confining pressure are monitored during the simulated excavation process to generate a visual chart.

[0027] Preferably, the robot cluster in S5 is a 110 method intelligent lane formation four-machine equipment system consisting of a complete set of core equipment including a 110 method intelligent slit drilling rig, a 110 method intelligent constant resistance anchor cable drilling rig, a 110 method multifunctional drilling rig support and a 110 top cutting and side guard support.

[0028] The pillar-free and self-forming lane intelligent system based on machine learning includes a perception and recognition layer, an analysis and decision-making layer, and an execution and control layer.

[0029] Therefore, the present invention adopts the above-mentioned coal pillar-free self-forming lane intelligent design method and system based on machine learning, which has the following beneficial effects:

[0030] 1. The method adopted in the present invention can fundamentally realize the rational development and utilization of mine resources, which can not only reduce manpower and improve efficiency, but also avoid problems such as waste of resources and low tunnel excavation rate, and comprehensively form a new model of intelligent mine construction with Chinese characteristics.

[0031] 2. The 110 construction method safe and efficient lane forming robot system adopted by the present invention can realize real-time monitoring and early warning of the construction site during the construction process, effectively preventing accidents. In addition, the robot has an autonomous risk avoidance function and can take timely measures to ensure construction safety when encountering danger; the system adopts automation and intelligent technology to achieve efficient operation of lane construction. The robot can accurately complete various construction tasks, reduce manual intervention, and greatly improve construction speed and quality.

[0032] 3. The method adopted by the present invention can not only reduce labor costs, equipment investment and material consumption, and reduce the overall construction cost, but also help promote the innovation and development of my country's underground engineering technology and improve my country's competitiveness in the international underground engineering field.

[0033] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a schematic diagram of the intelligent design platform of the construction method 110 of the present invention;

[0035] Figure 2 It is a data processing flow chart of the present invention;

[0036] Figure 3 It is the topological structure diagram of the BP neural network of the present invention;

[0037] Figure 4 Establish a flow chart for the BP neural network model of the present invention

[0038] Figure 5 The cross-validation principle diagram of the present invention

[0039] Figure 6 This is the process diagram of the four-machine linkage lane forming method 110 of the present invention. DETAILED DESCRIPTION

[0040] The technical solution of the present invention is further described below through the accompanying drawings and embodiments.

[0041] Unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0042] Example

[0043] See also Figure 1-6 The present invention provides a method for intelligent design of a self-forming lane without coal pillars based on machine learning, which includes the following steps:

[0044] S1. Based on the previous data on the promotion and application of the 110 construction method, develop an intelligent design platform for the 110 construction method, and build a massive data resource management system with all-round acquisition, full network aggregation, and full-dimensional integration. Program the design of all projects that have implemented the 110 construction method, and conduct machine learning on the data in the database on the platform, so as to establish a 110 construction method intelligent design application support service platform to achieve unified data, unified services, and unified applications for each project.

[0045] The data resource management system needs to use the Sqoop open source system to migrate the data stored in the traditional database to the Hbase distributed database, such as the mining site conditions, mine information, surrounding rock geological conditions, and the supporting equipment data in the 110 method application cases in the past. Figure 2 As shown in the figure, the data from the previous 110 construction method application cases were processed, and the data processing included data cleaning, feature engineering and selection, oversampling, data analysis and data standardization. Data cleaning mainly uses the Spark distributed computing system to remove missing values ​​and data error values ​​caused by human factors; feature engineering refers to data that is highly correlated with the target data, such as mining depth, mining height, coal seam inclination, and working face inclined length; oversampling optimizes the sample distribution characteristics to unify the data range within a certain range; data standardization is to remove the influence of different sample dimensions, reduce data differences, and improve the stability of the data set.

[0046] The synthetic minority oversampling technique (SMOTE) is used to optimize the sample distribution characteristics and oversample the original data set. The method for normalizing the oversampled data is shown in the following formula:

[0047]

[0048] like Figure 3 As shown in the figure, the BP neural network model is used for training and prediction. During the training process, the error between the predicted value and the true value is calculated, and then the error is passed back through the network layer by layer. The weight and bias are adjusted according to the error of each layer, and the network parameters are continuously iterated and updated until the predetermined training goal is achieved.

[0049] like Figure 4 As shown in the figure, the main structure of the BP neural network is determined, and the particle swarm algorithm is further used to optimize the initial weights and thresholds of the neural network. Assuming that there are N particles and a Z-dimensional solution space, a population consisting of N particles is randomly initialized at the beginning. After calculating the individual extreme value and the global extreme value, the speed and position of the particles are updated according to the following formula:

[0050]

[0051] u t+1 =u t +u t+1 ;

[0052] In the formula, v t is the velocity vector, u t is the position vector, p j is the individual extreme value vector, p k is the global extreme value vector.

[0053] At the beginning of particle motion, ω t and c 2 t Set a larger value to ensure particle diversity and fully search the entire space; in the later stage of iteration, particles should converge to the global optimal solution, and to ensure the particle convergence effect, at this time ω t and c 1 t Should be smaller and c 2 t Larger. ωt and c 1 t 、c 2 t The update method is as follows:

[0054] ω t =[(ω 0 -ω 1 )cos(πt / T)+(ω 0 +ω 1 )] / 2;

[0055]

[0056] In order to further expand the search space, improve particle diversity, and avoid falling into the local optimal solution, after each update, according to a certain probability β t Initialize the particles and use random number a to determine the particle mutation direction, particle mutation method and β t The calculation method is as follows:

[0057]

[0058] βt=[(β 1 +β)-(β 1 -β 0 )cos(πt / T)] / 2;

[0059] In order to further improve the data utilization efficiency and the generalization performance of the model, the 6-fold cross-validation method is used to train and predict the model. Figure 5 As shown. First, the target data set is divided into 6 unrelated subsets. Five subsets are taken as training sets each time, and the remaining subset is used as the validation set. Six training and predictions are performed in sequence to obtain 6 prediction models. The average prediction effect of the 6 prediction models is selected to evaluate the model and parameters. The model evaluation index uses the mean absolute error (EMA), root mean square error (ERMS) and determination coefficient (R2) to comprehensively evaluate the prediction effect of the model. The calculation method is as follows:

[0060]

[0061] Step 2: Enter the surrounding rock geomechanical parameters, mine information, and stope conditions of the designed working face coal mine in the perception and recognition layer. Based on the previous design usage and the selected matching 110 working face design, further increase the design model attributes. Perform efficient digital modeling for machine learning of model digital design, and integrate the three-dimensional design platform, process simulation platform, and cost management platform. Use FLAC3D three-dimensional modeling software to model the designed working face, simulate the excavation process, monitor the surrounding rock deformation and confining pressure changes, and call the digital model from the database or server through asynchronous data interaction methods and FLAC3D three-dimensional modeling methods to perform three-dimensional visualization conversion to form a visualization model. Communicate, adjust, and modify with experts to form a design plan or solution, which will be returned to the Hbase sub-database for storage through data interaction.

[0062] In step three, the system judges the stability of the surrounding rock and the sufficient filling of the rock goaf according to the design scheme generated in step two and the content of machine learning. Then, based on the judgment results, it independently completes decisions on support parameters, equipment selection, scheme design, design optimization, etc.

[0063] Step 4: Collect important parameters of each equipment of the fully mechanized mining face through the underground industrial ring network, coal mining technology, geological structure, coal seam roof and floor, hydrology and other factors affecting mining, unify all data, apply digital twin technology, build digital twin models of top cutting, drilling and anchoring, support, gangue blocking, and top control of the lane, as well as two lane support models, realize real-time process mapping of the lane, and integrate digital intervention models of guiding factors such as geology, panel area, disaster, and continuation. In order to realize intelligent linkage control of the working face mining, transportation, support, fluid supply, video, personnel positioning and other systems, and realize intelligent collaborative control of fully mechanized mining equipment in the working face chute monitoring center.

[0064] Step 5: According to the decision content of step 3, it is transmitted to the robot cluster and completed on the working surface. Figure 6 The four-machine linkage tunneling process shown in the figure includes intelligent drilling and anchoring, top cutting, rock blocking and temporary support. The work results are transmitted to the working face digital platform in real time, and the results are evaluated and fed back to the perception information section for analysis.

[0065] Step six: Based on the analysis results of the perception information section, the production process is deduced, and combined with the production conditions fed back by the equipment, the planning execution effect and production process are evaluated, and the model is optimized through autonomous learning. By applying AI large model technology, deep learning and manual correction are carried out, mining is accurately driven, the tunneling process is effectively implemented, and the production process is optimized in real time to achieve the optimal decision-making model of the 110 method intelligent design system, guide actual production, and ensure production safety.

[0066] Therefore, the present invention adopts the above-mentioned coal pillar-free self-forming lane intelligent design method and system based on machine learning, which can fundamentally realize the rational development and utilization of mine resources, can not only reduce manpower and improve efficiency, but also avoid problems such as waste of resources and low tunnel excavation rate, and comprehensively form a new model of intelligent mine construction with Chinese characteristics. The system adopts automation and intelligent technology to achieve efficient operation of tunnel construction. The robot can accurately complete various construction tasks, reduce human intervention, and greatly improve construction speed and quality. The present invention can not only reduce labor costs, equipment investment and material consumption, and reduce the overall construction cost, but also help promote the innovation and development of my country's underground engineering technology and improve my country's competitiveness in the field of international underground engineering.

[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. An intelligent design method for self-forming lanes without coal pillars based on machine learning, characterized in that: The following steps are involved: S1. Program the design of all projects that have implemented the 110 construction method, conduct machine learning on the data in the database on the platform, and establish an intelligent design platform for the 110 construction method; S2. By designing the surrounding rock geomechanical parameters, mine information, and stope conditions of the working face coal mine, a surrounding rock geomechanical model is constructed. At the same time, the deformation and pressure changes of the surrounding rock in the model are monitored, and the monitoring results are compared with the actual on-site monitoring data; S3. Based on the model data and machine learning content, make judgments on the stability of the surrounding rock and the sufficient filling of the rock goaf, and independently complete the support parameters and equipment selection decisions based on the judgment results; S4. Establish a digital model of the working face and a two-lane support model, as well as a digital twin model of the top cutting, anchor drilling, support, rock blocking, and top control of the completed lane, to achieve real-time process mapping of the completed lane, and integrate a digital intervention model that integrates geological, panel, disaster, and succession guidance factors; S5: According to the decision content of S3, it is transmitted to the robot cluster to complete intelligent top cutting, drilling and anchoring, rock blocking and temporary support at the working face, and the application effect is transmitted to the working face digital platform in real time, and its effect is evaluated and fed back to the perception information section for analysis; S6. Based on the analysis results of the perception information section, the production process is deduced, and combined with the production conditions fed back by the equipment, the planning execution effect and production process are evaluated and the model is optimized through autonomous learning, so as to achieve the optimal decision-making model for intelligent coal mining of the 110 method, optimize the working face and design safety assurance.

2. The method for intelligent design of self-forming lanes without coal pillars based on machine learning according to claim 1 is characterized in that: S1 includes the following steps: S11. Process the data from the previous 110 method application cases. Data processing includes data cleaning, feature engineering and selection, oversampling, data analysis and data standardization; S12. Use the BP neural network model for training and prediction. During the training process, the error between the predicted value and the true value is calculated, and the error is passed back through the network layer by layer. The weights and biases are adjusted according to the error of each layer, and the network parameters are continuously updated iteratively until the predetermined training goal is achieved.

3. The method for intelligent design of self-forming lanes without coal pillars based on machine learning according to claim 2 is characterized by: In S11, synthetic minority oversampling technology SMOTE is used to optimize the sample distribution characteristics, oversample the original data set, and the method for standardizing the oversampled data is as follows: in, represents the sample mean, s represents the sample standard deviation, and n represents the sample size.

4. The method for intelligent design of self-forming lanes without coal pillars based on machine learning according to claim 3 is characterized by: In S12, the particle swarm algorithm is used to optimize the initial weights and thresholds of the neural network. There are N particles and a Z-dimensional solution space. Initially, a population of N particles is randomly initialized. After calculating the individual extreme value and the global extreme value, the speed and position of the particles are updated according to the following formula: in t+1 =in t +in t+1 ; Among them, ω t 、c1 t 、c2 t They represent the extent to which the particle motion is affected by the local optimal solution and the global optimal solution, respectively. t is the velocity vector, u t is the position vector, p j is the individual extreme value vector, p k is the global extreme value vector.

5. The method for intelligent design of self-forming lanes without coal pillars based on machine learning according to claim 4 is characterized in that: In S12, the particle swarm algorithm is used to optimize the initial weights and thresholds of the neural network, and the 6-fold cross-validation method is used to train and predict the model.

6. The method for intelligent design of self-forming lanes without coal pillars based on machine learning according to claim 5 is characterized by: The model evaluation indicators include mean absolute error EMA, root mean square error ERMS and determination coefficient R 2 To comprehensively evaluate the prediction effect of the model, the calculation method is as follows:

7. The method for intelligent design of self-forming lanes without coal pillars based on machine learning according to claim 6 is characterized by: In S2, the three-dimensional modeling software FLAC3D is used to model the designed working face, simulate the excavation process, monitor the surrounding rock deformation and confining pressure changes, and generate visual charts.

8. The method for intelligent design of self-forming lanes without coal pillars based on machine learning according to claim 7 is characterized in that: The robot cluster in S5 is a 110 method intelligent lane formation four-machine equipment system consisting of a set of core equipment including the 110 method intelligent slotting drill rig, the 110 method intelligent constant resistance anchor cable drill rig, the 110 method multifunctional drilling rig support and the 110 top cutting and side guard support.

9. A system using the method for intelligent design of self-forming lanes without coal pillars based on machine learning as described in any one of claims 1 to 8, characterized in that: It includes perception and recognition layer, analysis and decision-making layer, and execution and control layer.

Images