A hydraulic support roof beam load distribution inversion method

Abstract

The application provides a hydraulic support roof beam load distribution inversion method, relates to the field of intelligent monitoring and digital twinning of a fully mechanized coal mining face support system, and the method comprises the following steps: acquiring real-time column pressure data of a hydraulic support during operation; inputting the real-time column pressure data into a pre-constructed mechanical mapping agent model, and outputting a real-time load distribution result of a roof beam of the hydraulic support.

Description

Technical Field

[0001] This invention belongs to the field of intelligent monitoring and digital twin technology of support system for fully mechanized coal mining faces, and particularly relates to a method for inverting the load distribution of hydraulic support top beam. Background Technology

[0002] As coal mining extends to deeper levels, the surrounding rock conditions at the working face become increasingly complex, significantly increasing the risk of dynamic disasters such as rock bursts. As a core support device ensuring the safety of the mining area, the load-bearing state of the hydraulic support's top beam directly affects the stability of the entire support system.

[0003] Currently, the monitoring of loads on hydraulic supports has the following limitations: 1. Reliance on physical sensors, making maintenance difficult: Traditional methods often involve installing column-type pressure sensor arrays on the top beam to directly monitor force distribution. However, the downhole environment is characterized by high humidity, high dust levels, and strong impact vibrations, making sensors mounted on the top beam highly susceptible to damage or signal drift, resulting in high maintenance costs and insufficient data reliability.

[0004] 2. Severe model simplification and insufficient accuracy: Some existing technologies attempt to invert loads using cylinder pressure, but these are often based on two-dimensional vector closed-loop equations and the assumption of ideal hinges, treating the top beam as a rigid body. This rigid model ignores the local stress concentration in the top beam caused by asymmetric loads in three-dimensional space, and also fails to consider the rigid-flexible coupling effect of hydraulic support component deformation, resulting in inversion results that cannot accurately reflect the real load transfer mechanism.

[0005] 3. Lack of distribution inversion and uncertainty quantification: Existing monitoring methods mostly focus on the overall working resistance (i.e., the simple sum of column pressure), and cannot accurately invert the specific load distribution in different areas of the top beam (such as the front end, rear end, left and right sides). In addition, existing inversion methods are mostly deterministic outputs, lacking the ability to assess the reliability (uncertainty) of the inversion results.

[0006] Therefore, there is an urgent need for a method that can reduce reliance on sensors, take into account structural flexibility and deformation, and accurately invert the load distribution and uncertainties of the top beam. Summary of the Invention

[0007] To address the aforementioned technical challenges, a method for inverting the load distribution of a hydraulic support top beam based on digital twins and physical Gaussian process regression is provided. This method integrates a rigid-flexible coupled dynamic model with a Gaussian process regression algorithm to construct a data-driven surrogate model, enabling high-precision inversion and working condition identification of the array-like load distribution of the top beam using only column pressure.

[0008] In a first aspect, the embodiments of this specification provide a method for inverting the load distribution of a hydraulic support top beam, comprising: acquiring real-time column pressure data of the hydraulic support during operation; inputting the real-time column pressure data into a pre-constructed mechanical mapping proxy model, and outputting the real-time load distribution result of the hydraulic support top beam; wherein, the mechanical mapping proxy model is based on a Gaussian process regression algorithm; the nonlinear mapping relationship of the mechanical mapping proxy model is determined by a mechanical mapping dataset; and the data characteristics of the mechanical mapping dataset are determined by simulation data of the rigid-flexible coupling dynamic model of the hydraulic support under multiple working conditions.

[0009] This solution overcomes the traditional reliance on physical sensors for the top beam in monitoring. By simply reading easily accessible and relatively stable column pressure data, the load distribution of the entire top beam can be indirectly and accurately inverted through a proxy model that integrates data and mechanisms, achieving a technological leap from "point" monitoring to "area" perception. Furthermore, by utilizing a rigid-flexible coupling model as the data source, it compensates for the shortcomings of traditional rigid body models that ignore component deformation, significantly improving inversion accuracy.

[0010] In one embodiment, the rigid-flexible coupling dynamic model of the hydraulic support is constructed through the following steps: establishing a rigid multibody dynamic model of the hydraulic support; performing flexible preprocessing on the key components of the hydraulic support, calculating modal parameters and generating a flexible body modal neutral file; generating a flexible body component using the flexible body modal neutral file and replacing the corresponding rigid component in the rigid multibody dynamic model; establishing the mechanical connection relationship between the flexible body component and adjacent components through node mapping, thereby forming the rigid-flexible coupling dynamic model of the hydraulic support.

[0011] This approach integrates mechanical motion, hydraulic transmission, and structural flexibility to construct a high-fidelity electromechanical-hydraulic system simulation model. This model can realistically simulate the dynamic response of hydraulic supports under complex working conditions, providing a physically accurate "data foundation" for proxy models.

[0012] In one embodiment, the mechanical mapping dataset is determined through the following steps: constructing a virtual working condition instruction set including uniformly distributed load, eccentric load, and impact load; driving the rigid-flexible coupling dynamic model of the hydraulic support to execute the virtual working condition instruction set; synchronously recording the column pressure data during the simulation process as input features, and recording the load distribution data of the preset array nodes on the top beam as output features; normalizing and quality-verifying the recorded data to form a mechanical mapping dataset.

[0013] This solution constructs a comprehensive "column pressure - top beam distributed load" mapping dataset covering typical working conditions of hydraulic supports, ensuring that the trained surrogate model can not only handle conventional working conditions, but also adapt to extreme working conditions such as off-center loading and impact, thus improving the model's generalization ability.

[0014] In one implementation, the mechanical mapping proxy model is determined through the following steps: using a mechanical mapping dataset as training samples; selecting a radial basis function as the kernel function for Gaussian process regression and initializing hyperparameters; calculating the covariance matrix using the training samples and optimizing the hyperparameters by maximizing the marginal likelihood function to establish a nonlinear mapping relationship from column pressure to top beam load distribution, thereby obtaining the mechanical mapping proxy model.

[0015] This scheme utilizes the powerful nonlinear mapping capability of Gaussian process regression (GPR) to efficiently establish the complex relationship between column pressure and multi-region load distribution in the top beam. Compared with traditional physical equation solving, this method has a fast calculation speed and can meet the real-time monitoring requirements at the millisecond level.

[0016] In one embodiment, the real-time load distribution results of the hydraulic support top beam include: the predicted mean of the load distribution in each region of the top beam; and the predicted variance calculated based on the covariance matrix of Gaussian process regression, wherein the predicted variance is used to characterize the uncertainty range of the inversion results.

[0017] This approach not only outputs the predicted load values ​​but also quantifies the uncertainty (i.e., reliability) of the predictions. This provides a crucial reference for downhole safety decision-making, allowing operators to assess the reliability of the current inversion results and avoid misjudgments caused by model prediction biases.

[0018] In one embodiment, after acquiring the real-time column pressure data of the hydraulic support during operation, the method further includes a step of performing working condition identification, which includes: extracting the left column pressure value and the right column pressure value from the real-time column pressure data; calculating the pressure non-uniformity coefficient based on the left column pressure value and the right column pressure value; and determining the numerical range to which the pressure non-uniformity coefficient belongs, so as to identify the current working condition type of the hydraulic support.

[0019] This solution can automatically identify whether the stent is currently in a normal support, off-center load, or impact state, thus upgrading the support status from passive monitoring to active intelligent perception and classification.

[0020] In one embodiment, the step of calculating the pressure unevenness coefficient based on the pressure values ​​of the left and right columns, and determining the numerical range to which the pressure unevenness coefficient belongs, includes: using a formula ; Calculate the pressure unevenness coefficient α, where x1 and x2 are the pressure values ​​of the left and right columns, respectively; execute the following judgment logic: when α < 0.1 and the pressures on the left and right are balanced, the current working condition is determined to be a normal support working condition; when 0.1 ≤ α ≤ 0.3, the current working condition is determined to be a slight off-center load working condition; when α > 0.3, the current working condition is determined to be a significant off-center load working condition.

[0021] This solution establishes a quantitative identification criterion based on the pressure non-uniformity coefficient, which can accurately identify the severity of load non-uniformity and provide data support for targeted on-site frame adjustment operations.

[0022] In one embodiment, the step of performing working condition identification further includes: monitoring the total pressure value and its time-series change rate of real-time column pressure data; when an abnormal increase in the total pressure value is detected and there is an instantaneous peak value exceeding a preset threshold, the current working condition type is determined to be an impact load working condition.

[0023] This solution can respond to instantaneous changes in pressure, quickly identify dynamic phenomena such as pressure on the working face roof, and provide early warning of rockburst hazards.

[0024] In one embodiment, after outputting the real-time load distribution results of the hydraulic support top beam, the method further includes a visualization step based on digital twins, which includes: constructing a three-dimensional visualization model corresponding to the physical hydraulic support; mapping the real-time load distribution results to the corresponding grid area of ​​the top beam in the three-dimensional visualization model; generating a real-time load distribution heat map on the three-dimensional visualization model and updating the support posture synchronously.

[0025] This solution establishes a closed-loop feedback chain integrating "sensing-inversion-visualization," allowing operators to view the support posture and the thermal diagram of the top beam stress in real time through an intuitive 3D visualization interface, greatly improving the intuitiveness and efficiency of monitoring.

[0026] In one embodiment, the method further includes a closed-loop verification step of the mechanical mapping proxy model, which includes: controlling the physical hydraulic support test bench to perform a preset loading condition, collecting measured column pressure data and measured top beam load data; inputting the measured column pressure data into the mechanical mapping proxy model to obtain a predicted load distribution; calculating the error between the predicted load distribution and the measured top beam load data, and if the error exceeds a preset threshold, updating the mechanical mapping proxy model using the measured column pressure data and the measured top beam load data.

[0027] This solution establishes a closed-loop verification and iterative system encompassing "high-fidelity simulation - data-driven proxy - physical experimentation." The physical test bench provides a repeatable and quantifiable physical verification environment, which not only verifies the accuracy of the proxy model but also utilizes measured data to inversely optimize the model, ensuring the reliability of the inversion system in practical engineering applications.

[0028] In a second aspect, embodiments of this specification provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method as described in any of the first aspects.

[0029] Thirdly, embodiments of this specification provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in any of the first aspects. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0031] Figure 1 A flowchart illustrating the method for inverting the load distribution of the hydraulic support top beam is provided for the implementation of this specification. Figure 2 A detailed flowchart illustrating the method for inverting the load distribution of the hydraulic support top beam is provided for the implementation of this specification. Figure 3 This is a flowchart illustrating the hierarchical construction of a rigid-flexible coupling dynamic model of a hydraulic support according to an embodiment of this application. Figure 4 This is a flowchart of the top beam load inversion and working condition identification based on physical Gaussian process regression according to an embodiment of this application; Figure 5 This is a flowchart illustrating the construction and closed-loop verification process of a digital twin of a hydraulic support according to an embodiment of this application. Figure 6 This is a schematic diagram of the structure of an electronic device provided for the implementation of this specification. Detailed Implementation

[0032] Unless otherwise defined, the technical or scientific terms used in the embodiments of this specification shall have the ordinary meaning understood by one skilled in the art to which this specification pertains. The terms "first," "second," and similar terms used in the embodiments of this specification do not indicate any order, quantity, or importance, but are merely used to avoid confusion of constituent elements.

[0033] Unless the context otherwise requires, throughout this specification, "a plurality of" means "at least two," and "including" is interpreted as open-ended or encompassing, that is, "including, but not limited to." In the description of this specification, terms such as "one embodiment," "some embodiments," "exemplary embodiment," "example," "specific example," or "some examples" are intended to indicate that a particular feature, structure, material, or characteristic associated with that embodiment or example is included in at least one embodiment or example of this specification. The illustrative representations of the above terms do not necessarily refer to the same embodiment or example.

[0034] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this specification.

[0035] As described in the background section, with the deepening of coal mining, the surrounding rock conditions at the working face become increasingly complex, significantly increasing the risk of dynamic disasters such as rock bursts. Hydraulic supports, as the core support equipment for ensuring the safety of the mining area, have their top beam load-bearing capacity directly affecting the stability of the entire support system.

[0036] Currently, the monitoring of loads on hydraulic supports has the following limitations: 1. Reliance on physical sensors, making maintenance difficult: Traditional methods often involve installing column-type pressure sensor arrays on the top beam to directly monitor force distribution. However, the downhole environment is characterized by high humidity, high dust levels, and strong impact vibrations, making sensors mounted on the top beam highly susceptible to damage or signal drift, resulting in high maintenance costs and insufficient data reliability.

[0037] 2. Severe model simplification and insufficient accuracy: Some existing technologies attempt to invert loads using cylinder pressure, but these are often based on two-dimensional vector closed-loop equations and the assumption of ideal hinges, treating the top beam as a rigid body. This rigid model ignores the local stress concentration in the top beam caused by asymmetric loads in three-dimensional space, and also fails to consider the rigid-flexible coupling effect of hydraulic support component deformation, resulting in inversion results that cannot accurately reflect the real load transfer mechanism.

[0038] 3. Lack of distribution inversion and uncertainty quantification: Existing monitoring methods mostly focus on the overall working resistance (i.e., the simple sum of column pressure), and cannot accurately invert the specific load distribution in different areas of the top beam (such as the front end, rear end, left and right sides). In addition, existing inversion methods are mostly deterministic outputs, lacking the ability to assess the reliability (uncertainty) of the inversion results.

[0039] Therefore, there is an urgent need for a method that can reduce reliance on sensors, take into account structural flexibility and deformation, and accurately invert the load distribution and uncertainties of the top beam.

[0040] Based on the above inventive concept, the following is an exemplary description of the hydraulic support top beam load distribution inversion method provided in the embodiments of this specification.

[0041] This embodiment provides a method for inverting the load distribution of a hydraulic support top beam. This method integrates digital twin, rigid-flexible coupling dynamic simulation, and machine learning techniques. Figure 1 As shown, the method specifically includes the following steps: Step S101: Obtain real-time column pressure data of the hydraulic support during operation.

[0042] In this step, real-time column pressure data refers to the real-time pressure values ​​of the hydraulic cylinders inside the support columns (usually including the left and right columns) during actual downhole operation or test bench testing of the hydraulic support. This data can typically be read directly from the existing hydraulic control system of the hydraulic support or from additional pressure sensors. These data will serve as the input feature variables x1 and x2 for the subsequent inversion model.

[0043] Step S102: Input the real-time column pressure data into the pre-built mechanical mapping proxy model and output the real-time load distribution result of the hydraulic support top beam.

[0044] In this step, the mechanical mapping proxy model is a data-driven model built based on machine learning algorithms, specifically employing the Gaussian Process Regression (GPR) algorithm. The core function of this model is to establish a nonlinear mapping relationship between the "input (column pressure)" and the "output (top beam load distribution)." It is important to note that the training data for this mechanical mapping proxy model (i.e., the mechanical mapping dataset) does not entirely rely on expensive and difficult-to-obtain full-sample downhole measurement data, but is primarily determined by simulation data from the rigid-flexible coupling dynamic model of the hydraulic support under multiple operating conditions. This rigid-flexible coupling dynamic model differs from traditional pure rigid body models by introducing the flexible characteristics of key components, thus generating high-fidelity mechanical data containing component deformation information. The model output specifically includes the load prediction mean and its uncertainty variance for nine regions arranged in a three-row, three-column array on the top beam.

[0045] This embodiment is a further refinement and improvement of Embodiment 1, detailing the specific processes of constructing the rigid-flexible coupling model, generating the dataset, training the surrogate model, and identifying operating conditions and verifying closed loops. For example... Figures 2 to 4 As shown, the method in this embodiment includes the following detailed steps: Step S201: Construct a rigid-flexible coupling dynamic model of the hydraulic support.

[0046] This step aims to build a high-fidelity "digital prototype" as the source for generating training data. The specific process is as follows: Figure 3 As shown, the model construction spans both the front-end user environment and the MWORKS+ANSYS runtime support environment, and is divided into the following four levels: Front-end / User Layer (Geometric Processing and Definition): First, the geometric model is processed in 3D modeling software (such as UG). This layer primarily clarifies user requirements and objectives, prepares the 3D geometric model, defines material properties, and sets working conditions and boundary conditions. The 3D model of the hydraulic support undergoes geometric repair and feature simplification. By identifying and suppressing non-critical features such as small-scale fillets, chamfers, and bolt holes that do not affect overall stiffness, the geometric topology is optimized, laying the foundation for subsequent high-quality mesh generation. Subsequently, material mechanical parameters, including basic properties such as elastic modulus, Poisson's ratio, and material density, are accurately defined for the simplified geometric model to ensure that the material model can realistically reflect the mechanical behavior of the component under actual loads. Finally, the load conditions, boundary conditions, and required output variables are defined according to the simulation objectives.

[0047] Multibody dynamics modeling layer (rigid model construction): In a multibody dynamics simulation platform (such as MWORKS.Sysplorer), a rigid model containing components such as the top beam, columns, and base is established. This layer mainly completes the modeling of rigid body components, the definition of constraints and excitations, the definition of kinematic pairs, and the application of drives and loads. Based on the geometric model processed in the front-end layer, the kinematic pair constraints between the main components such as the top beam, columns, and base are precisely defined, and the complete system topology is constructed. The platform's built-in model library is used to establish the initial model of the hydraulic system (such as hydraulic cylinders and control valve pipelines), and external excitations such as top plate loads are applied. This stage requires model assembly verification and preliminary kinematic simulation.

[0048] Flexible body preprocessing and neutral document generation layer: For critical components that significantly impact mechanical transmission (such as the piston, outer cylinder, middle cylinder, and top beam), flexibility processing is performed using finite element analysis software (such as ANSYS). Through mesh generation and remote point definition, geometric preprocessing, assignment of material properties, and modal analysis, a flexible neutral file containing the piston, outer cylinder, middle cylinder, and top beam is generated. First, detailed connection point definitions are performed on the imported component geometry. Remote points with "deformable" behavior are created at the assembly interfaces between components, and their degrees of freedom are precisely constrained according to the actual connection type. Mesh generation is then performed, with local refinement applied to complex areas. Flexible configuration settings are configured using the APDL command flow, defining the modal extraction method. Modal analysis is then performed to calculate the component's multiple natural frequencies and mode shapes in the free state, and to analyze the modal participation factor and effective mass distribution. After the solution is completed, a standard modal neutral file (.mnf) is generated through a dedicated interface. This binary file encapsulates complete flexibility information for the component, including its mass matrix, stiffness matrix, and mode shapes.

[0049] Rigid-flexible coupling integration layer: Using the aforementioned neutral file, flexible body components are generated in a multibody dynamics platform to replace the original rigid parts. Mechanical connections between the flexible body and surrounding components are established through node mapping technology, ultimately forming a rigid-flexible coupled dynamic model. Specifically, the .mnf file is imported using the flexible body model library function to create high-precision flexible body components, which are then used to replace the rigid piston, outer cylinder, middle cylinder, and top beam in the system one by one. This model is solved by the platform's built-in high-performance solver, realistically simulating the dynamic response of the hydraulic support under complex working conditions. Through flexible body model replacement, model verification and settings, and rigid-flexible body connections, the coupled model assembly and integration are completed. Finally, a co-simulation is run, outputting the column-top beam mechanical mapping data for various working conditions.

[0050] Step S202: Generate a mechanical mapping dataset.

[0051] Build an instruction set: Design a set of virtual load case instructions covering "uniformly distributed load", "eccentric load case" (left eccentric / right eccentric / diagonal eccentric) and "impact load".

[0052] Simulation drive: Use the above instruction set to drive the rigid-flexible coupling model to run.

[0053] Data Recording and Processing: Simulation data is systematically extracted in the MWORKS.Sysplorer post-processing environment. The column pressure (x1, x2) during the simulation is recorded synchronously as input, and the load data (y1 to y9) of the preset array nodes on the top beam (e.g., nine coordinate point sensors arranged in three rows and three columns) is recorded as output. The data is normalized and quality checked to form a structured sample set. Complete records of column pressure data and synchronously acquired top beam load distribution data are made for each working condition, forming a comprehensive multi-condition column pressure-top beam load distribution mapping dataset covering typical working states of the hydraulic support.

[0054] Step S203: Construct and train the mechanics mapping proxy model.

[0055] like Figure 4 The "Gaussian Process Regression Modeling Module" on the left and the "Model Performance Evaluation Module" in the middle show that the Gaussian Process Regression (GPR) algorithm is used. First, the dataset is divided into training and test sets in an 8:2 ratio. The radial basis function (RBF) is selected as the kernel function, which can be expressed as: ; The covariance matrix is ​​calculated using the dataset obtained in step S202, and the hyperparameters (such as length scale l, signal variance) are optimized by maximizing the marginal likelihood function. Noise variance Complete model training. During model training, the covariance matrix is ​​calculated using the training dataset. K : ; New data points Covariance vector between training data and training data : ; The self-covariance of the new data point : ; The key assumption in Gaussian process modeling is that the data can be viewed as samples drawn from a multivariate Gaussian distribution: ; Data y and the point to be predicted The joint distribution of is a multivariate Gaussian distribution, where the covariance matrix is ​​determined by the kernel function. , , constitute.

[0056] Given data, a certain predicted value What is the probability of it, that is, the conditional probability? This probability also follows a Gaussian distribution: ; The hyperparameters are optimized by maximizing the marginal likelihood function, which has the following logarithmic form: ; This establishes a nonlinear mapping relationship between column pressure and top beam load distribution.

[0057] Next, model performance is evaluated, specifically a systematic assessment of the trained model. If the prediction accuracy metrics (MAE, RMSE) and uncertainty quantification metrics (PICP, PIMWP) do not reach preset thresholds, the kernel function is reselected and redefined; if the thresholds are reached, the output model is used for subsequent calculations. Specifically, the mean absolute error (MAE) and root mean square error (RMSE) are used to evaluate prediction accuracy. ; ; Meanwhile, the probability of predicted interval coverage (PICP) and the average predicted interval width (PIMWP) are used to assess the uncertainty quantification capability: ; ; Set evaluation thresholds: MAE < 5%, RMSE < 8%, PICP > 95%. If the thresholds are not met, adjust the kernel structure and retrain; if the thresholds are met, the model passes validation.

[0058] Step S204: Online monitoring and load inversion.

[0059] In actual operation, real-time column pressure data is collected and input into a trained surrogate model. The mean of the posterior distribution is used as the predicted value to predict the load conditions in nine regions of the top beam. The best estimate is the mean of the distribution: ; in For new data points The covariance vector between the training data and the training data.

[0060] At the same time, the prediction variance is output using the characteristics of GPR. : ; in, This is the covariance of the new data points. This prediction variance directly represents the range of uncertainty in the current inversion results.

[0061] Step S205: Intelligent identification of working conditions.

[0062] The system automatically determines the operating condition type based on real-time pressure data, with the following logic: Figure 4 The "Working Condition Recognition and Result Output Module" on the right is shown below: Calculate the pressure non-uniformity coefficient: .

[0063] Normal support: when α < 0.1 and the left and right pressure values ​​are within the normal range.

[0064] Slightly uneven load: when 0.1≤α≤0.3.

[0065] Significant off-center loading: When α>0.3, the system should issue an alarm, indicating that there may be local breakage of the roof or incorrect posture of the support.

[0066] Impact load: When the total pressure value is detected to rise sharply in a very short period of time and the peak value exceeds the preset safety threshold, it is determined to be a risk of rockburst.

[0067] Step S206: Visualization and closed-loop verification.

[0068] This system, through the feedback of working condition commands and mechanical parameters, realizes a closed-loop system consisting of a virtual twin, a high-fidelity rigid-flexible coupling dynamic model, a Gaussian process proxy model, and a physical hydraulic support test bench. Its overall architecture and closed-loop verification logic are as follows: Figure 5 As shown, the details are as follows: Digital twin visualization: such as Figure 5 As shown in the module above, a digital twin model of the hydraulic support is constructed. First, the 3D models of the hydraulic support and scraper conveyor are built in UG software, converted to FBX format, and imported into Unity3D. In Unity3D, a hierarchical structure is used to clarify the parent-child relationships between the various components of the equipment, and a Mesh collider is constructed for the roof coal seam. A kinematic model of the support based on a finite state machine is developed using C# scripts to implement the logic control of column lowering, support shifting, column raising, and conveyor pushing. Data interfaces are configured: physical device information is obtained through the System.IO.Ports.SerialPort serial communication module. A network communication link is established between Unity and PyCharm (running the proxy model) based on the TCP protocol. The Unity client encapsulates real-time operating data in JSON format and sends it to the PyCharm client; the PyCharm client performs load inversion and sends the results back. Finally, a heat map is rendered in real-time on the top beam of the virtual support to visually display the stress concentration area.

[0069] Closed-loop verification on the test bench: Verification and iteration are carried out using a hydraulic support test bench. For example... Figure 5The peripheral "closed-loop verification and iteration" calibration loop shows that the surrogate model is an "efficient substitute" for the high-fidelity model, while the high-fidelity dynamic model is the "data foundation" of the surrogate model. The physical hydraulic support test bench receives initial pose and column pressure inputs, executes commands such as uniformly distributed load, off-center load, and impact load, and feeds back the measured load data of the top beam and columns to the system. Initial parameter input: The test bench receives the "hydraulic support initial pose input" command sent by the digital twin and adjusts the mechanical structure to the corresponding pose. Multi-condition execution: The test bench executes load application according to the "condition command set." In the "off-center load condition," it independently controls the hydraulic circuits of the left and right columns; in the "impact load," it uses a high-frequency response servo valve to simulate the pressure on the top plate. Data acquisition and feedback: The test bench's built-in high-frequency sensor network synchronously acquires the measured "column load data" and "top beam load data." All data undergoes filtering, noise reduction, and time-series alignment processing through a high-speed data acquisition system.

[0070] Closed-loop iteration: The measured column pressure is input into the surrogate model, and the error between the predicted load and the measured load is calculated. If the error (e.g., MAE) exceeds the allowable range (e.g., 5%), the measured data is added to the training set, and the surrogate model is updated online (retraining), thus realizing a closed-loop verification loop of "simulation-surrogate model-experiment".

[0071] In one exemplary embodiment of this specification, an electronic device is also provided, such as Figure 6 As shown, the electronic device may include: a processor 310, a communication interface 320, a memory 330, and a communication bus 340, wherein the processor 310, the communication interface 320, and the memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute the hydraulic support top beam load distribution inversion method, which includes: Obtain real-time column pressure data of the hydraulic support during operation; The real-time column pressure data is input into a pre-built mechanical mapping proxy model, and the real-time load distribution results of the hydraulic support top beam are output. The mechanical mapping proxy model is based on the Gaussian process regression algorithm; the nonlinear mapping relationship of the mechanical mapping proxy model is determined by the mechanical mapping dataset; and the data characteristics of the mechanical mapping dataset are determined by the simulation data of the rigid-flexible coupling dynamic model of the hydraulic support under multiple working conditions.

[0072] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0073] In addition to the methods, apparatus, and devices described above, the hydraulic support top beam load distribution inversion method provided in the embodiments of this specification can also be a computer program product, which includes computer program instructions that, when executed by a processor, cause the processor to perform the steps in the hydraulic support top beam load distribution inversion method according to various embodiments of this specification as described in the "Exemplary Methods" section above.

[0074] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this specification. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages.

[0075] Furthermore, embodiments of this specification also provide a computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor of the steps in the hydraulic support top beam load distribution inversion method according to various embodiments of this specification as described in the "Exemplary Methods" section above.

[0076] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this specification can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0077] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0078] The embodiments described above are merely illustrative of several implementation methods outlined in this specification. While the descriptions are specific and detailed, they should not be construed as limiting the scope of the solutions provided in this specification. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this specification, and these all fall within the scope of protection of this specification. Therefore, the scope of protection for this patent should be determined by the appended claims.

Claims

1. A method for inverting the load distribution of a hydraulic support top beam, characterized in that, include: Obtain real-time column pressure data of the hydraulic support during operation; The real-time column pressure data is input into a pre-built mechanical mapping proxy model, and the real-time load distribution results of the hydraulic support top beam are output. The mechanical mapping proxy model is based on the Gaussian process regression algorithm; the nonlinear mapping relationship of the mechanical mapping proxy model is determined by the mechanical mapping dataset; and the data characteristics of the mechanical mapping dataset are determined by the simulation data of the rigid-flexible coupling dynamic model of the hydraulic support under multiple working conditions.

2. The method according to claim 1, characterized in that, The rigid-flexible coupling dynamic model of the hydraulic support is constructed through the following steps: Establish a rigid multibody dynamics model for the hydraulic support; The key components of the hydraulic support undergo flexible preprocessing, modal parameters are calculated, and flexible body modal neutral files are generated. The flexible body modal neutral file is used to generate a flexible body component, which then replaces the corresponding rigid component in the rigid multibody dynamics model. The mechanical connection relationship between the flexible body component and adjacent parts is established by node mapping, thereby forming the rigid-flexible coupling dynamic model of the hydraulic support.

3. The method according to claim 1, characterized in that, The mechanical mapping dataset is determined through the following steps: Construct a virtual load case instruction set that includes uniformly distributed load, eccentric load, and impact load; The rigid-flexible coupling dynamic model of the hydraulic support is driven to execute the virtual working condition instruction set. The column pressure data during the simulation process is recorded as input features, and the load distribution data of the preset array nodes on the top beam is recorded as output features. The recorded data is normalized and its quality is verified to form the mechanical mapping dataset.

4. The method according to claim 1, characterized in that, The mechanical mapping proxy model is determined through the following steps: The aforementioned mechanical mapping dataset was used as the training sample. Radial basis functions are selected as the kernel function for Gaussian process regression, and hyperparameters are initialized. The covariance matrix is ​​calculated using the training samples, and the hyperparameters are optimized by maximizing the marginal likelihood function to establish a nonlinear mapping relationship from column pressure to top beam load distribution, thereby obtaining the mechanical mapping proxy model.

5. The method according to claim 1, characterized in that, The real-time load distribution results of the hydraulic support top beam include: Predicted mean values ​​of load distribution in each region of the top beam; And the prediction variance calculated based on the covariance matrix of Gaussian process regression, which is used to characterize the uncertainty range of the inversion results.

6. The method according to claim 1, characterized in that, After acquiring the real-time column pressure data of the hydraulic support during operation, the method further includes a step of performing working condition identification, which includes: Extract the pressure values ​​of the left and right columns from the real-time column pressure data; Calculate the pressure non-uniformity coefficient based on the pressure values ​​of the left and right columns; Determine the numerical range to which the pressure non-uniformity coefficient belongs in order to identify the current operating condition type of the hydraulic support.

7. The method according to claim 6, characterized in that, The step of calculating the pressure unevenness coefficient based on the pressure values ​​of the left and right columns, and determining the numerical range to which the pressure unevenness coefficient belongs, includes: Using formula ； Calculate the pressure non-uniformity coefficient α, where x1 and x2 are the pressure values ​​of the left column and the right column, respectively; Execute the following decision logic: When α < 0.1 and the pressure on the left and right sides is balanced, the current working condition is determined to be a normal support working condition. When 0.1≤α≤0.3, the current operating condition is determined to be a light off-center load condition; When α > 0.3, the current operating condition is determined to be a significant off-center load condition.

8. The method according to claim 6, characterized in that, The step of identifying the operating condition also includes: Monitor the total pressure value and its time-series rate of change of the real-time column pressure data; When an abnormal increase in the total pressure value is detected and there is an instantaneous peak value exceeding a preset threshold, the current working condition is determined to be an impact load condition.

9. The method according to claim 1, characterized in that, After outputting the real-time load distribution results of the hydraulic support top beam, the method further includes a visualization step based on digital twins, which includes: Construct a 3D visualization model corresponding to the physical hydraulic support; The real-time load distribution results are mapped to the corresponding grid area of ​​the top beam in the three-dimensional visualization model; A real-time load distribution heat map is generated on the three-dimensional visualization model, and the support posture is updated synchronously.

10. The method according to claim 1, characterized in that, The method further includes a step of performing closed-loop verification on the mechanical mapping proxy model, which includes: The physical hydraulic support test bench was controlled to perform preset loading conditions, and measured column pressure data and measured top beam load data were collected. The measured column pressure data is input into the mechanical mapping proxy model to obtain the predicted load distribution; Calculate the error between the predicted load distribution and the measured top beam load data. If the error exceeds a preset threshold, update the mechanical mapping proxy model using the measured column pressure data and the measured top beam load data.

Images