Regional prediction method for periodic weighting of large-dip-angle working face

Abstract

The invention provides a regional prediction method for periodic weighting of a large-dip-angle working face, and belongs to the technical field of coal mining. According to the method, a twinborn model corresponding to a physical entity is created by using a digital twinborn technology, and collected data are analyzed and predicted in combination with multiple algorithms in machine learning. Regional division is performed on a working face through a Mean Shift clustering algorithm, and an algorithm model is constructed for hydraulic support resistance data of different regions, so that prediction precision and reliability are improved. And meanwhile, the prediction results of the twinborn model and the algorithm model are subjected to consistency test, so that the accuracy and consistency of the prediction results are ensured. According to the regional prediction method for the periodic weighting of the large-dip-angle working face, dynamic monitoring and high-precision prediction of the periodic weighting under complex geological conditions are achieved through cooperation of digital twinning and multiple algorithms, and a reliable basis is provided for coal mine intelligent management, disaster early warning and safety decision making.

Description

Technical Field

[0001] The invention relates to the technical field of coal mining, and in particular to a regional prediction method for periodic pressure on a large-angle working face. Background Art

[0002] In the process of coal mining, the mining direction gradually advances to deep and difficult-to-mine coal seams, accompanied by the emergence of ground pressure and various geological disasters. Especially in the large-angle working face, due to the complex geological conditions and abnormal structure, the frequency of pressure phenomena has increased significantly. At present, the coal mining industry faces the important challenge of how to accurately predict the periodic pressure phenomenon to achieve preventive control. In recent years, with the rapid development of digital twin technology and machine learning, these emerging technologies have provided new ideas and methods for the prediction of periodic pressure in coal mines. Digital twin technology mainly creates a virtual mirror of the physical system, so that the status, environment, state and other information of the mine can be monitored and updated in real time, forming a dynamic prediction and analysis platform. Machine learning uses a large amount of historical data to perform pattern recognition and prediction through algorithm models to realize the analysis and judgment of the potential risks of pressure phenomena. By combining digital twin technology with machine learning, complex working face pressure prediction can be achieved faster.

[0003] In the high-angle working face, due to its special geological structure and ore rock characteristics, the pressure phenomenon not only shows randomness, but also shows obvious regional characteristics. Due to the differences in geological parameters such as pressure and stress in different regions, there are significant differences in the frequency and intensity of periodic pressure. Therefore, traditional prediction methods often cannot effectively capture the complex characteristics of these regions, resulting in prediction results that are inconsistent with the actual situation. In view of this situation, it is necessary to propose a new pressure prediction method. Summary of the invention

[0004] The purpose of the present invention is to provide a regional prediction method for periodic pressure on a large-angle working face, aiming to solve a series of problems faced in mining large-angle working faces through digital twin technology and machine learning algorithm technology, including but not limited to the fact that traditional predictions mainly rely on empirical rules and simple mathematical models, lack effective modeling of complex geological environments and dynamic working conditions, and the inability of mines to fully utilize sensor data and advanced data analysis technology, which reduces the timeliness and reliability of prediction results. Through this invention technology, it is possible to accurately capture the law of pressure changes on complex working faces, significantly improve the accuracy of pressure prediction, identify potential risks in advance, improve work safety, and thus optimize resource allocation and improve work efficiency.

[0005] To achieve the above object, the present invention provides a regional prediction method for periodic pressure on a large-angle working face, comprising the following steps:

[0006] Step S1: The perception layer collects data through sensors arranged on the mine working face and hydraulic supports, and transmits the data to the information layer through the communication network via the interaction layer;

[0007] Step S2: After the information layer performs data cleaning and correlation analysis on the collected data, it obtains the processed data and transmits it to the model layer;

[0008] Step S3: The model layer is set with a digital twin weighting prediction model. The digital twin weighting prediction model includes a twin model and an algorithm model, and trains and optimizes the constructed twin model and algorithm model;

[0009] Step S4: Conduct a consistency test on the prediction results of the trained and optimized twin model and algorithm model.

[0010] Preferably, in step S1, the sensors include pressure sensors arranged on the hydraulic supports and underground cameras arranged in the roadway.

[0011] Preferably, in step S2, data cleaning includes deleting duplicate data, filling missing values, and handling outliers.

[0012] Preferably, in step S3, the construction steps of the twin model are as follows:

[0013] Use SolidWorks software to analyze the motion postures of the hydraulic supports in the steeply inclined working face, and construct a three-dimensional solid model including the physical and mechanical structures and the hydraulic system. Set the motion postures of the hydraulic supports in actual applications, define the corresponding motion parameters and constraint conditions, and at the same time verify and compare the accuracy and precision of the model, and then adjust the existing errors and inconsistencies to simulate the actual working process of the hydraulic supports;

[0014] Convert the constructed three-dimensional solid model into an STL interface file, import it into Maya software to adjust the material, texture, lighting and shadows of the model, and perform animation design to optimize the rendering effect;

[0015] Create a virtual scene of the steeply inclined working face in Unity3D, configure the background, lighting and sound effects, use C# scripts to program the actions of the hydraulic supports and control the parameters of the hydraulic system, and set the sensors and communication modules to achieve data exchange and information transfer, thereby completing the construction of the twin model.

[0016] Preferably, in step S3, the construction steps of the algorithm model are as follows:

[0017] Extract the original time-series data of the upper, middle and lower area hydraulic supports in a specific time period from the hydraulic support online monitoring system as a data set, and perform data preprocessing, including data cleaning, conversion, feature selection and extraction, and smoothing and noise reduction processing;

[0018] Use the MeanShift clustering algorithm to analyze the regional characteristics of the steeply inclined working face and divide the working face area;

[0019] Construct multiple machine algorithm models for the hydraulic support resistance data in different working face areas;

[0020] Compare the prediction accuracies of the algorithm model prediction results in each working face area. Based on the mean square error, root mean square error, mean absolute error, mean absolute percentage error, and coefficient of determination, select the algorithm model with the highest prediction accuracy in each working face area for the final prediction.

[0021] Preferably, in step S4, when the prediction results of the twin model and the algorithm model are consistent, the visualization result is displayed, and relevant control decisions are made according to the prediction results. When the prediction results of the twin model and the algorithm model are inconsistent, the digital twin abutment pressure prediction model is updated and the test is repeated.

[0022] Therefore, the present invention adopts the above-mentioned sub-region prediction method for the periodic abutment pressure of the steeply inclined working face, and the beneficial technical effects are as follows: it can accurately monitor the abutment pressure situation of the steeply inclined working face, which is beneficial to the further realization of intelligent management in coal mines. At the same time, the accurate abutment pressure prediction results help decision-makers make correct decisions before the environment changes, thereby improving the safety and intelligent level of mine management. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is the architecture diagram of the sub-region prediction system for the periodic abutment pressure of the steeply inclined working face;

[0024] Figure 2 It is the flow chart of the sub-region prediction method for the periodic abutment pressure of the steeply inclined working face;

[0025] Figure 3 It is the construction diagram of the twin model;

[0026] Figure 4 It is the analysis diagram of the mechanical model;

[0027] Figure 5 It is the simulation curve diagram of the twin model;

[0028] Figure 6 It is the flow chart of the sub-region abutment pressure prediction of the algorithm model;

[0029] Figure 7 It is the clustering division diagram of the steeply inclined working face area;

[0030] Figure 8 It is the comparison diagram of the load prediction in multiple areas of the working face; among them, Figure 8 (a) in it is the No. 3 hydraulic support; Figure 8 (b) in it is the No. 27 hydraulic support;Figure 8 In (c), it is hydraulic support No. 43; Figure 8 In (d), it is hydraulic support No. 66;

[0031] Figure 9 It is a comparison chart for predicting the abutment pressure in multiple areas of the working face; Figure 9 In (a), it is hydraulic support No. 3; Figure 9 In (b), it is hydraulic support No. 27; Figure 9 In (c), it is hydraulic support No. 43; Figure 9 In (d), it is hydraulic support No. 66. Specific implementation manners

[0032] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0033] Unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the ordinary meanings understood by those of ordinary skill in the field to which the present invention belongs.

[0034] Embodiment 1

[0035] I. Architecture of the sub - area prediction system for periodic abutment pressure in a large - dip working face.

[0036] In view of the problems such as the complex geological environment, lagging prediction methods and insufficient accuracy existing in the current large - dip working face, the present invention proposes a sub - area prediction method for periodic abutment pressure in a large - dip working face. This method combines various sensor data features, industrial Internet information interaction technology, a fusion module of digital twin and machine learning, etc., and constructs a perception layer, an interaction layer, an information layer and a model layer.

[0037] As shown in Figure 1 , it is the architecture diagram of the sub - area prediction system for periodic abutment pressure in an inclined working face, including:

[0038] The perception layer is mainly used to detect the state of hydraulic supports, the mine working environment and the surrounding rock stress state. Sensors are arranged in the actual coal - mining working face to collect experimental data. Pressure sensors are arranged at key parts of the hydraulic support, such as hydraulic cylinders, columns and cross beams, to monitor the pressure changes of the roof abutment pressure in real time during the mining process. At the same time, underground cameras are arranged in the roadway to monitor the operation state of the equipment and, combined with the monitoring system, realize remote control. The layout of these sensors comprehensively considers different working environments of the mine and the characteristics of each production stage to ensure that the actual abutment pressure situation can be timely feedback under various loads.

[0039] As the key connection between the perception layer and the information layer, the interaction layer plays an important role in information transmission, data processing and control execution. The interaction layer mainly involves two aspects of communication connection and edge control to ensure the efficient and intelligent operation of the system.

[0040] In a communication connection, communication devices are the basis for realizing data transmission, including various sensors, data collectors, and communication modules. These devices are responsible for collecting key parameters from a steeply inclined working face, such as hydraulic support pressure and its corresponding time, etc., and through appropriate signal processing, convert the data into a transmissible format. To achieve efficient data transmission, the interaction layer adopts a variety of communication networks, including both wired networks (such as Ethernet) and supports wireless networks (such as Wi-Fi, NB-IoT, etc.). This flexible network architecture enables rapid and effective information transfer in various complex working environments. The effectiveness of the communication layer also depends on the reasonable application of data transmission protocols. Commonly used protocols such as MQTT, CoAP, and HTTP, etc., can achieve standardized communication between different devices. Protocol parsing ensures the correctness and accuracy of the transmitted data, enabling the received data to be quickly understood and processed.

[0041] Edge control runs on edge computing nodes close to the data source, processing and analyzing data from the perception layer. They can perform on-site immediate analysis, use machine learning algorithms for status evaluation and trend prediction, saving bandwidth and shortening response time. The edge gateway connects edge devices to the cloud, responsible for data forwarding, caching, and protocol conversion, managing real-time and historical data, and ensuring the efficient operation of the system. It also adapts the data protocols between different devices. Edge control refers to the control strategy executed in real-time by edge computing nodes. Based on the processing and analysis of real-time data, edge control can make decisions quickly and execute corresponding operations. For example, when the sensor in the perception layer detects abnormal pressure, edge control will immediately adjust the working face or activate an alarm to reduce risks. With advanced communication devices and networks, the interaction layer can ensure real-time and efficient information transfer; through edge services, edge gateways, and edge control, the system can achieve rapid response and intelligent decision-making, thereby improving the safety and stability of the steeply inclined working face.

[0042] The information layer serves as a bridge for real-time mapping and interaction between physical entities and virtual twins, mainly responsible for data collection, cleaning, correlation analysis, and optimization of algorithm models. The construction of the information layer is crucial for ensuring data quality and the accuracy of analysis results. Data cleaning is the core link of the information layer, including identifying and deleting duplicate data, filling in missing values, and handling outliers, aiming to improve effectiveness by removing redundant, incorrect, and irrelevant data. These processes ensure that the data used for subsequent analysis is accurate and reliable, making the model training and prediction results more credible. The cleaned data is further processed through operations such as classification, clustering, and regression to extract effective information, exclude outliers, provide a high-quality data foundation for subsequent analysis and applications, and facilitate the application and processing of related twin models and algorithm models in the later stage.

[0043] The model layer is set with a digital twin for roof weighting prediction model, which consists of a twin model and an algorithm model. The two interact closely through twin data.

[0044] The twin model integrates two specific models: a load model and a simulation model. Among them, the load model is a mathematical representation established based on the coupling dynamics and load change characteristics between the surrounding rock and the support system, and can also provide relevant data support for the subsequent simulation model. The simulation model mainly constructs a model through software such as SolidWorks in combination with actual motion parameters and is used to represent the actual working process.

[0045] The algorithm model is divided into data algorithm, model algorithm and application algorithm, and the algorithm model is mainly trained through historical load data or simulation data. The specific process is as follows: the data algorithm preprocesses and smooths and denoises the hydraulic support data, and then is used for the training and testing of the model algorithm; the trained algorithm is optimized for hyperparameters to achieve accurate prediction of the roof weighting situation; the application algorithm analyzes the prediction results and formulates an intelligent decision-making plan.

[0046] II. Steps for periodic roof weighting prediction.

[0047] Figure 2 It is a flowchart of the sub-region prediction method for periodic roof weighting in a large dip angle working face, which can more detailedly explain the specific prediction process.

[0048] The present invention proposes to combine digital twin technology with machine learning theory for predicting periodic roof weighting in a mine working face to achieve precise, dynamic and efficient support measures.

[0049] The specific steps are as follows:

[0050] Step 1: The relevant devices in the working face perception layer collect the data of the physical entity hydraulic support in real time, and transfer the data to the information layer through communication connection via the interaction layer.

[0051] Step 2: After preprocessing the collected data, the information layer transmits the data to the twin model and the algorithm model respectively for simulation and training.

[0052] Step 3: Conduct a simulation experiment test on the twin model through the hydraulic support data to achieve visual display.

[0053] Step 4: Divide the preprocessed hydraulic support data into a training set and a test set, so as to conduct training evaluation and testing on the algorithm model, and then generate an algorithm prediction result.

[0054] Step 5: The large dip angle working face weighting prediction digital twin system calls the algorithm model through the application program interface (API) for load prediction, and transmits the results to the Unity3D interface. The digital twin system is integrated with the actual hydraulic support through the API method, realizing real-time data interaction and feedback of prediction results.

[0055] Step 6: Conduct a consistency test on the results of the twin model and the algorithm model. If the results are consistent, the prediction results are displayed in real time, and relevant control decisions are made based on the prediction results. If the results are inconsistent, the prediction model is updated again to improve the prediction accuracy, and the system is improved, and then the consistency test is repeated until the test results are consistent.

[0056] Step 7: The reasonable prediction data displayed in real time by the twin model will be re-imported into the database to optimize the database.

[0057] Step 8: The prediction results of the twin model directly interact with the physical hydraulic support entity and realize automatic control, decision optimization and feedback.

[0058] III. Construction and mechanical analysis of the twin model.

[0059] Figure 3 It is a construction diagram of the hydraulic support twin model, which specifically describes the construction process of the twin model in digital twin.

[0060] During the construction process of the hydraulic support twin model, software such as SolidWorks and 3DsMax were used to establish the twin model, and by importing these models into the Unity3D software, the docking and real-time interaction between the physical entity and the virtual model were realized. The specific steps are as follows: First, use SolidWorks software to analyze the motion postures of the hydraulic supports in the large dip angle working face, and at the same time import the relevant structural parameters of the hydraulic supports collected in the information layer, such as the column angle, column length, roof beam length, base length, balance bar length, etc., and construct a three-dimensional solid model including the physical mechanics structure and the hydraulic system. During this process, the motion postures of the hydraulic supports during the actual working face mining process, such as the spatial pose state, the motion state of the hydraulic cylinders, and the relative motion of the key components, were set, and the motion parameters such as geometric motion parameters, dynamic response parameters, and mechanical parameters, as well as the constraint conditions such as stroke limit, angle limit, bearing capacity limit, stability constraint, and kinematic constraint, were defined. At the same time, the accuracy and precision of the model were verified and compared, and then the existing errors and inconsistencies were adjusted to accurately simulate the actual working state of the hydraulic supports. After the adjustment is completed, the model is converted into an STL interface file.

[0061] Next, the obtained STL interface file is imported into Maya software, where the material, texture, lighting, and shadows of the model are adjusted to realistically display the surface features of the hydraulic support. At the same time, relevant animation designs are carried out to achieve the smoothness of the movement process and state changes, further optimize the rendering effect, and enhance the realism and visual beauty of the model. Finally, a virtual scene of a steeply inclined working face is created in Unity3D to simulate the actual working environment, and the background, lighting, and sound effects are configured. Meanwhile, C# scripts are used to program the actions of the hydraulic support and control the parameters of the hydraulic system, and sensors and communication modules are set up to achieve data exchange and information transmission, thus ensuring real-time interaction between the model and the physical hydraulic support.

[0062] The generated digital twin system for roof weighting prediction in the steeply inclined working face calls relevant algorithm models for prediction through the application programming interface (API) and transmits the results to the Unity3 system. Then, the digital twin system is integrated with the actual hydraulic support system through the API method to achieve real-time data interaction of the hydraulic support group and feedback of the prediction results. Specifically, the data part in the digital twin system receives real-time data, makes predictions through the prediction model in the digital twin system, and then feeds the prediction results back to the physical part of the digital twin system to update the model parameters and states, thereby realizing real-time prediction and management of the roof weighting prediction of the hydraulic support.

[0063] To ensure the accuracy of the digital representation of the load model in the twin model, a planar mechanical analysis is carried out to comprehensively understand the stress conditions of the hydraulic support in the mine roadway under the load conditions of the steeply inclined working face, which helps to more accurately simulate the actual working state of the hydraulic support under periodic roof weighting. As Figure 4 shown, it is the analysis diagram of the mechanical model of the hydraulic support.

[0064] During the mining process of the steeply inclined working face, affected by the movement and periodic roof weighting of the overlying roof of the working face, the hydraulic support will have a tendency of relative movement due to the roof weighting and the subsidence of the floor. As the contact mode and load characteristics between the working face roof and the hydraulic support change, the behavior of the hydraulic support also changes, resulting in subsidence, sliding, and rotation. Through in-depth analysis of the load and instability phenomena of the mechanical model of the hydraulic support during periodic roof weighting, it is more helpful for the accurate construction and effective application of the twin model in aspects such as structural parameter optimization, dynamic characteristic analysis, data interaction optimization, intelligent decision support, and visualization display.

[0065] As Figure 4 shown, during the movement of the roof, as the overlying roof collapses and exerts pressure, the position state of the hydraulic support will change. First, the subsidence amount of point O of the hydraulic support along the z-axis direction is denoted as z o , and at the same time, the rotation angle around point O is denoted as According to the elastic foundation theory, the distributed load q at the upper and lower edge positions of the hydraulic support base along the dip direction can be obtained. A and q B The calculation formula is as follows:

[0066]

[0067] Among them, k o represents the floor foundation coefficient (kN·m -3 ); c represents the length of the hydraulic support base; α represents the coal seam dip angle.

[0068] Through the above formulas (1) and (2), the resultant force F of the normal load of the floor on the hydraulic support and the acting position x2 at the current position of the hydraulic support can be obtained as follows: N and the acting position x2 are:

[0069] F N = ack o z o (3);

[0070]

[0071] Among them, a represents the width of the hydraulic support.

[0072] According to the mechanical model of the hydraulic support as shown in Figure 4 , the equilibrium condition of the hydraulic support under the condition of a large dip working face can be obtained as follows:

[0073]

[0074] Among them, F R represents the friction force between the hydraulic support and the roof rock stratum; F F represents the friction force between the hydraulic support and the floor rock stratum; G represents the weight of the hydraulic support; P represents the working resistance of the hydraulic support; x1 represents the acting position between the hydraulic support and the roof load; b represents the height of the hydraulic support; h o represents the height of the center of gravity of the hydraulic support.

[0075] Since the working resistance of the hydraulic support is much greater than the gravity of the hydraulic support itself, for the convenience of theoretical solution, the influence of the rotation angle of the hydraulic support on its own gravity is ignored in the subsequent analysis. According to formulas (5) to (7), the settlement amount z O , the rotation angle and the friction force F F between the hydraulic support and the floor can be calculated as follows:

[0076]

[0077] F F= Gsinα - F R (10);

[0078] The analysis of the above mechanical model helps to identify the key parameters and their effects on the performance of the hydraulic support. At the same time, through the optimization and adjustment of these parameters in the twin model, the model is closer to the actual working conditions, improving the prediction accuracy and adaptability of the model. It is also convenient to master the overall dynamic characteristics of the hydraulic support, providing an accurate basis for the dynamic simulation of the twin model, enabling the twin model to truly reproduce the actual motion state of the hydraulic support.

[0079] IV. Verification of the twin model.

[0080] Since the accuracy of the hydraulic support's prediction of the roof weighting has a relatively large impact on the entire twin model, in order to verify the rationality of the twin system and ensure the correct progress of modeling, it is necessary to conduct a verification experiment on the model. After setting the parameters of the twin system, a simulation experiment is carried out to judge whether the time-load curve and the roof weighting manifestation are consistent. The simulation curve of the twin system is as Figure 5 shown.

[0081] In Figure 5 , the twin model can clearly show that as the fully mechanized coal mining face advances continuously, the main roof undergoes periodic fracture, causing the load on the hydraulic support to increase significantly and even reach the peak. At the same time, the "masonry beam" structure formed by the broken blocks of the main roof undergoes periodic movement changes, thus forming the periodic roof weighting phenomenon. The load curve of the twin model can also accurately show the various stages of the periodic roof weighting process, as follows:

[0082] Initial stage: The hydraulic support and the roof are not fully in contact or only have the initial support force, and the working resistance is relatively low. The load pressure of the hydraulic support is generally about 15 Mpa. As the working face gradually advances, the geological conditions and stress states around the hydraulic support will change. The exposed roof fractures and exerts an additional load on the hydraulic support, and the concentrated pressure of the roof on the hydraulic support during local strong roof weighting causes the working resistance of the hydraulic support to gradually increase to a relatively high peak. At this time, the peak load is the end-of-cycle resistance. After the peak load, due to large-scale fracture or caving of the roof, or passive / active unloading of the hydraulic support, the working resistance of the hydraulic support drops rapidly, which also indicates the end of a cycle of roof weighting. When the working face continues to advance to the next position, the roof will fracture and sink again, and the working resistance will gradually rise again, starting a new cycle of roof weighting process.

[0083] From the simulation curve results of the twin model, it can be seen that the change curve of the twin model is basically consistent with the actual dynamic change trend, and the correlation error is also within the acceptable range. The dynamic characteristics in each stage meet the experimental requirements. This proves the reliability and effectiveness of the twin model in predicting the weighting phenomenon in the simulated inclined stope environment.

[0084] V. Construction of the algorithm model.

[0085] Figure 6 It is the flowchart of the sectional weighting prediction of the algorithm model, which mainly explains the construction process of the algorithm model and the relevant analysis results. By using the MeanShift clustering algorithm to analyze the resistance of hydraulic supports and using the algorithm model to predict the resistance of hydraulic supports, the relevant weighting results are finally obtained, and then applied to the prediction and analysis of the weighting situation in the large dip angle working face. The specific steps are as follows:

[0086] (1) Define the sample data. Extract the original time series data of hydraulic supports numbered 1 to 67 in the upper, middle and lower regions from a certain place from January 2, 2024 to February 6, 2024 from the on-line monitoring system of hydraulic supports as the data set, and the sampling interval is 10 minutes.

[0087] (2) Data preprocessing. During the coal mining process, emergencies such as periodic weighting and coal wall spalling will cause sharp changes in the load of hydraulic supports. Through wavelet decomposition, the local characteristics in the signal can be revealed, which is more convenient for subsequent signal analysis. On this basis, the extracted hydraulic support data is cleaned, transformed, feature selected and extracted, and smoothed and denoised to optimize the training and prediction performance of the model.

[0088] (3) The regional characteristics analysis of the large dip angle working face shows that the periodic weighting and fracture of the roof of the large dip angle working face have significant zoning and time series characteristics. In order to effectively cope with these complex laws, various algorithm models can be adopted to predict according to the performance of hydraulic supports during the weighting process.

[0089] (4) The data category division adopts the clustering algorithm based on MeanShift. First, the bandwidth parameter is reasonably set to determine the neighborhood search range. Through kernel density estimation, the weighted average value of each data point within its neighborhood is accurately calculated. Then, the mean shift vector of each data point is accurately calculated, and this vector accurately points to the direction of density increase. Subsequently, according to the mean shift vector, the positions of the data points are continuously iteratively updated until the position changes of all data points are lower than the preset threshold and the convergence terminates. After convergence, the adjacent data points are accurately merged into the clustering centers, thus initially determining the clustering results. After that, the rationality of the clustering results is evaluated, and the abnormal or unreasonable classification results are deeply optimized until the expected requirements are met. Finally, the clustering results are visually visualized and comprehensively evaluated to fully support subsequent in-depth analysis and scientific decision-making.

[0090] Based on the monitoring situation of the time load sequence data of the hydraulic supports in the working face, the MeanShift clustering algorithm is used to autonomously divide the density of the sample set. It is mainly based on the load differences borne by the hydraulic supports and their different positions, so as to accurately distinguish the working face area. When selecting the monitoring points of the hydraulic support column loads, the principle of equidistant distribution is followed. In the working face with a total of 67 hydraulic supports, the time load sequences of the 5th, 11th, 17th, 23rd, 29th, 35th, 41st, 47th, 53rd, 59th, and 65th hydraulic supports are monitored respectively. At the same time, to ensure the continuity of the working face area division, the above-numbered hydraulic supports are used as the central supports, and the central support and its first 3 and last 2 hydraulic supports are divided into the same area. The entire working face area is divided according to this principle. The division results are as Figure 7 shown. Using the two variables of the hydraulic support load parameter and the hydraulic support number as the input parameters of the clustering algorithm, the hydraulic support groups in the steeply inclined working face can be divided into four areas, and the boundaries of each area in the working face are relatively obvious. The division results of the steeply inclined working face area are the 1st - 20th hydraulic supports, the 21st - 37th hydraulic supports, the 38th - 50th hydraulic supports, and the 51st - 67th hydraulic supports. The data of the 3rd, 27th, 43rd, and 66th hydraulic supports are respectively selected as the research objects to train and predict the algorithm model.

[0091] (5) Build an algorithm model to predict the periodic weighting situation of the working face. First, classify different types of data, and regard the cyclic end resistance of the measured curve as a periodic change feature. To effectively train the model, 80% of the data is selected as the training set, and the remaining 20% is used as the test set. During the model construction process, considering the regional characteristics in the actual production process of the steeply inclined working face, and comparing the algorithm prediction accuracies of different regions.

[0092] Due to the particularity of geological structure and the influence of steeply inclined working face, the load data of hydraulic supports show obvious non-linear regionalization characteristics and time-series dependence characteristics. In this case, Random Forest (RF) and Support Vector Regression (SVR) perform well in modeling non-linear relationships, while Long Short-Term Memory (LSTM) is good at capturing long-term dependence relationships. Therefore, it is necessary to compare the performance advantages and disadvantages of the three algorithms and select the most appropriate model for prediction analysis. Finally, the prediction results of the algorithm model are visualized and consistency tested through a twin model to verify the rationality of the model, which is conducive to the continuous learning and optimization of the model itself.

[0093] To more comprehensively evaluate the accuracy of the three prediction models, the evaluation indicators adopted include: Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Coefficient of Determination (R 2 ). The comparison results of the prediction accuracies of multiple algorithms are shown in Table 1. After comparing and analyzing the prediction accuracies of the three models, algorithm models with higher accuracies are respectively selected for prediction. Among them, the RF model is used to predict the X1-X20 hydraulic supports, the RF model is used to predict the X21-X35 hydraulic supports, the SVR model is used to predict the X36-X50 hydraulic supports, and the SVR model is used to predict the X51-X67 hydraulic supports. The comparison of the prediction accuracies of specific algorithm models can be seen in Figure 8 .

[0094] (6) Subsequently, in order to improve the prediction performance of the model, ISSA is used to optimize the hyperparameters of the above-mentioned related algorithms. Through steps such as initial hyperparameter setting, iterative update, performance evaluation, and optimal solution selection, the efficiency and accuracy of the algorithm model are improved, the search strategy and local information utilization rate are enhanced, and the optimal hyperparameter combination is found. At the same time, the optimized model is evaluated. If the predicted value meets the requirements, the model is exported and its performance is tested and evaluated. If it does not meet the requirements, the parameters are adjusted to optimize the model and the model is optimized again. Finally, in order to reflect the advantages of the optimized model, its parameters are compared with the initial model, as shown in Table 2, indicating that the overall prediction accuracy of the model has been further improved after being optimized by ISSA. ISSA refers to "Improved Sparrow Search Algorithm Integrating Cauchy Mutation and Opposition-Based Learning".

[0095] Table 1 Comparison Results of Prediction Accuracies of Multiple Algorithms

[0096]

[0097] Table 2 Comparison Results of Prediction Accuracies of Optimized Multiple Algorithms

[0098]

[0099]

[0100] (7) Output the prediction results and analyze. Output the resistance of the hydraulic support predicted by the algorithm model, and analyze the weighting situation of each hydraulic support during the advancement of the working face. Use the time-weighted average working resistance as the weighting criterion. The sum of the average value of the weighted working resistance of each cycle of the hydraulic support and its one-fold variance is used as the criterion for judging the roof weighting. The calculation formula is as follows:

[0101]

[0102] Among them, p′t represents the weighting criterion; represents the average value of the initial support force; represents the mean square deviation.

[0103] VI. Analysis of simulation results.

[0104] Visualize the load conditions of the hydraulic supports during the advancement of the hydraulic supports in different regions. The comparison results of the load predictions in multiple regions of the working face are as Figure 8 shown. Through the comparison results of the actual and predicted weighting, it is found that the present invention can accurately predict the working resistance of the hydraulic support. The prediction accuracy of the regions with obvious periodic weighting during the advancement of the working face can reach more than 95%, and accurate prediction and early warning can be achieved.

[0105] The comparison of the multi-region weighting predictions of the large dip angle working face is shown in Figure 9 . It can be directly seen from the figure that the method proposed by the present invention can well realize the regional weighting prediction of the working face, and the prediction phenomenon also conforms to the actual situation. The prediction results of the experiment show that in the large dip angle working face, the broken and fallen gangue in the middle and upper regions will slip, thus supporting the roof in the lower region, making the working face inclined. The weighting manifestation in the middle and upper regions is significantly more intense than that in the lower region, and the weighting step distance in the middle is greater than that on both sides, which conforms to the basic law of mine pressure during the mining process of large dip angle coal seams. Therefore, the prediction method proposed by the present invention can realize the high-precision dynamic prediction of the periodic weighting in different regions of the large dip angle working face.

[0106] It should be noted that the content not elaborated in detail in the present invention is all prior art and is well known to those skilled in the art.

[0107] Therefore, the present invention adopts the above-mentioned method for regional prediction of the periodic weighting of the large dip angle working face, and through the coordination of digital twin and multiple algorithms, realizes the dynamic monitoring and high-precision prediction of the periodic weighting under complex geological conditions, providing a reliable basis for the intelligent management, disaster early warning and safety decision-making of coal mines.

[0108] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions of the present invention or make equivalent replacements, and these modifications or equivalent replacements do not enable the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for predicting periodic weighting in a large dip angle working face by dividing regions, characterized in that It includes the following steps: Step S1: The perception layer collects data through sensors arranged on the mine working face and hydraulic supports, and transmits the data to the information layer through the interaction layer via a communication network; Step S2: After the information layer performs data cleaning and correlation analysis on the collected data, it obtains the processed data and transmits it to the model layer; Step S3: The model layer is provided with a digital twin weighting prediction model, which includes a twin model and an algorithm model, and trains and optimizes the constructed twin model and algorithm model; Step S4: Conduct a consistency test on the prediction results of the trained and optimized twin model and algorithm model.

2. The sub-region prediction method for periodic weighting in a large dip angle working face according to claim 1, characterized in that, In step S1, the sensors include pressure sensors arranged on the hydraulic supports and underground cameras arranged in the roadway.

3. The sub-region prediction method for periodic weighting in a large dip angle working face according to claim 1, characterized in that In step S2, data cleaning includes deleting duplicate data, filling in missing values, and handling outliers.

4. A method for predicting the periodic weighting of a large dip working face by sub-regions, according to claim 1, characterized in that In step S3, the construction steps of the twin model are as follows: Use SolidWorks software to analyze the motion postures of the hydraulic supports in the steeply inclined working face, and construct a three-dimensional solid model including physical and mechanical structures and hydraulic systems. Set the motion postures of the hydraulic supports in actual applications, define corresponding motion parameters and constraint conditions, and at the same time verify and compare the accuracy and precision of the model. Then adjust for the existing errors and inconsistencies to simulate the actual working process of the hydraulic supports; Convert the constructed three-dimensional solid model into an STL interface file, import it into Maya software to adjust the material, texture, lighting and shadows of the model, and perform animation design to optimize the rendering effect; Create a virtual scene of the steeply inclined working face in Unity3D, configure the background, lighting and sound effects, use C# scripts to program the actions of the hydraulic supports and control the parameters of the hydraulic system, and set up sensors and communication modules to achieve data exchange and information transmission, thus completing the construction of the twin model.

5. The sub-region prediction method for periodic weighting in a large dip angle working face according to claim 1, characterized in that In step S3, the construction steps of the algorithm model are as follows: Extract the original time series data of the hydraulic supports in the upper, middle and lower regions from the hydraulic support online monitoring system as the data set, and perform data preprocessing, including data cleaning, conversion, feature selection and extraction, and smoothing and noise reduction processing; Use the MeanShift clustering algorithm to analyze the regional characteristics of the steeply inclined working face and divide the working face area; Construct multiple machine learning algorithm models for the resistance data of the hydraulic supports in different working face areas; Compare the prediction accuracies of the algorithm models in each working face area. Based on the mean squared error, root mean squared error, mean absolute error, mean absolute percentage error, and coefficient of determination, select the algorithm model with the highest prediction accuracy in each working face area for the final prediction.

6. The sub-region prediction method for periodic weighting in a large dip angle working face according to claim 1, characterized in that, In step S4, when the prediction results of the twin model and the algorithm model are consistent, the visualization results are displayed, and relevant control decisions are made based on the prediction results. When the prediction results of the twin model and the algorithm model are inconsistent, the digital twin weighting prediction model is updated and the test is repeated.

Images