Coal mine fully mechanized working face digital twin monitoring system based on virtual-real fusion

Abstract

The present application belongs to the technical field of safety monitoring, and discloses a coal mine fully mechanized working face digital twin monitoring system based on virtual-real fusion, which comprises: obtaining the running state data of the fully mechanized working face hydraulic support group, extracting the impact response characteristics of the hinged point based on the same, and combining the shield beam posture data to construct the wear gap distribution atlas; calculating the force transmission attenuation coefficient and the end displacement deviation value according to the wear gap distribution atlas, and constructing the pressure coordination relationship benchmark model; continuously collecting the correction pressure signal through the sliding window, extracting the overrun event in combination with the benchmark model, and carrying out secondary verification based on the gap state to screen the leakage candidate support; the digital twin monitoring platform drives the three-dimensional model to realize visual leakage reproduction, and generates a diagnosis report; the present application realizes the accurate real-time monitoring of the hydraulic support leakage, and provides scientific decision support for the predictive maintenance of the coal mine fully mechanized working face.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring technology, and more specifically, to a digital twin monitoring system for fully mechanized coal mining faces based on virtual-real fusion. Background Technology

[0002] In existing coal mine fully mechanized mining face monitoring technologies, mainstream methods typically rely on single-sensor data acquisition and simple threshold alarm mechanisms. They employ static monitoring point layouts and passive fault detection strategies, lacking fusion analysis of multi-source heterogeneous data from hydraulic support groups and deep perception of equipment wear status. This results in significant lag in monitoring systems when dealing with progressive fault evolution, complex operating condition changes, and the coupled effects of multiple factors. With the deepening of intelligent coal mine construction and the rapid development of unmanned mining technology, the traditional "post-fault maintenance" monitoring model is increasingly unable to meet the urgent needs of efficient and safe production, accurate fault prediction, and intelligent decision support. It cannot accurately identify trends in equipment health status changes and lacks forward-looking maintenance guidance mechanisms. This technological bottleneck not only restricts the operational efficiency and safety assurance level of intelligent fully mechanized mining equipment but also seriously hinders the transformation and upgrading of coal mine monitoring technology towards higher precision, stronger predictability, and better human-machine collaboration.

[0003] In view of this, the present invention proposes a digital twin monitoring system for fully mechanized coal mining faces based on virtual-real fusion to solve the above problems. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: A digital twin monitoring system for fully mechanized coal mining faces based on virtual-real fusion includes: The data acquisition module is used to acquire real-time operating status data of the hydraulic support group in the fully mechanized mining face during the coal mining cycle. The real-time operating status data includes the pressure timing signal of each support column, the triaxial vibration acceleration signal of the hinge point of the four-bar linkage, the pitch angle data of the shield beam, and the sequence of coordinated action commands between adjacent supports. The parameter generation module extracts the impact response characteristics of each hinge point based on the triaxial vibration acceleration signal, and constructs a wear gap distribution map using the pitch angle data of the shield beam as a geometric constraint; and calculates the displacement deviation value at the end of the shield beam and the force transmission attenuation coefficient of the support column based on it. The benchmark modeling module corrects the gap error of each pressure timing signal based on the force transmission attenuation coefficient of the support column to obtain the corrected pressure timing signal. It also aligns and corrects the action time in the coordinated action command according to the displacement deviation value of the end of the shield beam to obtain the corrected coordinated action timing. Finally, it constructs a benchmark model of pressure coordination relationship based on the corrected pressure timing signal and the corrected coordinated timing. The dual verification module is used to continuously collect the corrected pressure time sequence signal of each support in a sliding time window of preset duration, and extract pressure coordination deviation exceeding the limit event by combining the pressure coordination relationship benchmark model. Based on the wear gap distribution map, the pressure coordination deviation exceeding the limit event is verified a second time to obtain the effective pressure anomaly time sequence. The diagnostic generation module performs anomaly screening on each stent based on valid abnormal time series to obtain leakage candidate stents. It synchronously reports the column number of the leakage candidate stent and the stent attitude data at the corresponding time to the digital twin monitoring platform. Based on the column number of the leakage candidate stent and the stent attitude data at the corresponding time, it performs visual leakage reproduction and generates a diagnostic report.

[0005] The technical effects and advantages of this invention, based on a virtual-real fusion digital twin monitoring system for fully mechanized coal mining faces: This invention utilizes wavelet packet decomposition technology of vibration acceleration signals to simultaneously extract the impact response spectrum, energy distribution, attenuation characteristics, and phase features of the hinge point, providing a richer and more reliable mechanical condition data foundation for accurate assessment of wear conditions compared to traditional single pressure monitoring. It innovatively introduces a gap inversion algorithm and a force transmission attenuation compensation mechanism, quantitatively correcting the previously neglected mechanical wear effects and significantly improving the accuracy of pressure measurement and the timing of coordinated actions. A three-dimensional index benchmark model is constructed, including cut position partitions, coordinated action stages, and upper and lower bounds of allowable deviations, accurately describing the spatiotemporal variation of the pressure coordination relationship between adjacent supports, solving the technical problem that traditional fixed threshold methods are difficult to adapt to complex working conditions. Furthermore, a staged pressure coordination analysis mechanism is established by combining the mechanical mechanisms of hydraulic systems and four-bar linkages, deeply revealing the load-bearing capacity. The dynamic evolution of pressure transmission patterns under key operating conditions such as static, rising and falling column movements, and pushing processes was investigated. Sliding window monitoring and secondary verification technologies were employed to achieve multi-timescale fault diagnosis capabilities, ranging from second-level anomaly detection to hourly leakage trend prediction. A remaining life estimation and priority decision-making mechanism based on pressure decay curves was established, effectively supporting the shift from passive maintenance to predictive maintenance. Through active and passive leakage differentiation and pressure drop slope verification technologies, combined with dynamic verification methods for gap states, leakage location accuracy was significantly improved and the false alarm rate was significantly reduced. Ultimately, a complete technical solution integrating multi-source information perception, wear compensation correction, collaborative relationship modeling, and digital twin visualization was formed, providing precise equipment health assessment tools and scientific maintenance decision support for intelligent safety monitoring of fully mechanized coal mining faces. Attached Figure Description

[0006] Figure 1 This is a schematic diagram of the digital twin monitoring system for fully mechanized coal mining faces based on virtual-real fusion, as described in this invention. Detailed Implementation

[0007] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0008] Example 1 Please see Figure 1 As shown in this embodiment, the digital twin monitoring system for fully mechanized coal mining faces based on virtual-real fusion includes: Real-time operational status data of the hydraulic support group in the fully mechanized mining face is acquired during the coal mining cycle. This real-time data is continuously collected throughout the entire coal mining cycle via a distributed sensor network. The coal mining cycle refers to the complete operational period from coal cutting by the mining machine and support relocation to the completion of roof support. The pressure timing signals of each support column reflect the dynamic changes in the roof pressure borne by the column. These signals are collected in real-time by pressure sensors installed on the piston rods of the columns, and they reflect the roof load transmission characteristics and the support's load-bearing status. The triaxial vibration acceleration signal of the hinge point of the four-bar linkage is acquired by a MEMS triaxial accelerometer. The sensor is fixed at the hinge point of the front link, rear link, shield beam, and top beam. The vibration signal contains information on the wear state and impact characteristics of the hinge point. The pitch angle data of the shield beam is measured by an angle sensor to measure the pitch angle of the shield beam relative to the horizontal plane with an accuracy of ±0.1°. The angle data reflects the current attitude of the support and the geometric configuration of the four-bar linkage. The coordinated action command sequence between adjacent supports is recorded by the centralized control system. It includes the action types of each support, such as raising the column, lowering the column, and moving the support, as well as their corresponding execution time. The sequence reflects the coordinated operation mode and timing relationship of the support group.

[0009] Impact response features of each hinge point were extracted based on triaxial vibration acceleration signals, and a wear gap distribution map was constructed using the pitch angle data of the shield beam as a geometric constraint. The displacement deviation at the end of the shield beam and the force transmission attenuation coefficient of the support column were calculated based on the wear gap distribution map. Impact response feature extraction is a key technology for making the wear information contained in the vibration signal explicit. By identifying the impact vibration mode generated at the hinge point during action switching, the degree of gap expansion is quantified. The pitch angle of the shield beam, as a geometric constraint, transforms the local gap of each hinge point into a displacement contribution in the global coordinate system, ensuring the spatial consistency of gap quantification. The wear gap distribution map is organized in the form of a two-dimensional index matrix, with row indices representing support numbers and column indices representing hinge point numbers. Matrix elements store the gap expansion amount and confidence level of the corresponding hinge point. The map comprehensively describes the wear state distribution of the entire working face support group. The displacement deviation at the end of the shield beam reflects the impact of the cumulative gap effect on the positioning accuracy of the support, and is calculated using a kinematic error transmission model. The force transmission attenuation coefficient of the support column quantifies the impact of the increased frictional dissipation caused by the gap on the accuracy of pressure monitoring, and a mapping relationship between the gap and force loss is established through dynamic analysis. These two parameters provide a quantitative basis for subsequent data correction.

[0010] Based on the force transmission attenuation coefficient of the support column, gap error correction is performed on each pressure timing signal to obtain a corrected pressure timing signal. Simultaneously, the action timing in the coordinated action command is aligned and corrected according to the displacement deviation value at the end of the shield beam to obtain a corrected coordinated action timing. A pressure coordination relationship benchmark model is then constructed based on the corrected pressure timing signal and the corrected coordinated action timing. Gap error correction is a core step in eliminating wear effects and improving the accuracy of monitoring data. The correction of the pressure timing signal is achieved by multiplying the measured pressure by a compensation gain, which is the reciprocal of the force transmission attenuation coefficient. The corrected signal reflects the actual pressure borne by the column from the top plate. The alignment correction of the coordinated action timing compensates for the action delay caused by the gap. Due to the existence of the gap, the support needs to eliminate the gap before generating effective displacement, causing the actual action completion time to lag behind the command time. The alignment correction calculates the delay time corresponding to the gap, shifting the theoretical time to the actual time to ensure the accuracy of the timing benchmark. The pressure coordination relationship benchmark model describes the correlation pattern of pressure between adjacent supports under normal operating conditions. The model uses the cut-off location partition and the coordinated action stage as indexes to store the statistical characteristics and allowable deviation range of pressure difference under the corresponding operating conditions, providing a dynamic benchmark for anomaly detection.

[0011] The corrected pressure timing signal of each support is continuously collected within a preset sliding time window. Combined with a pressure coordination relationship benchmark model, pressure coordination deviation exceeding the limit event is extracted. A secondary verification of the pressure coordination deviation exceeding the limit event is performed based on the wear gap distribution map, resulting in a valid pressure anomaly event sequence. Pressure coordination deviation exceeding the limit event extraction is the initial anomaly detection; by comparing the real-time pressure difference with the allowable deviation range of the benchmark model, moments significantly deviating from the normal pattern are identified. Secondary verification is a crucial step in reducing the false alarm rate. The equivalent error band of the pressure difference is calculated using the real-time updated gap status. If the deviation amplitude falls within the error band, it is determined to be a false anomaly caused by gap fluctuation and is discarded; if it exceeds the error band, it is confirmed as a real anomaly and retained. The valid pressure anomaly event sequence is a high-reliability anomaly record that has undergone double screening. Each record contains key information such as the event start time, duration, involved support number, and maximum deviation amplitude, providing accurate input for fault diagnosis.

[0012] Based on the effective pressure anomaly event sequence, anomaly screening is performed on each support within a single coal mining cycle to identify leak candidate supports. The column numbers of these candidate supports, along with their corresponding support attitude data, are synchronously reported to the digital twin monitoring platform. Anomaly screening focuses on diagnosis from the event layer to the equipment layer. By statistically analyzing the frequency, duration, and pressure change characteristics of each support participating in anomaly events within the coal mining cycle, active fault sources are identified. Leak candidate supports are those that simultaneously meet the three criteria of high-frequency participation, long duration, and continuous pressure decrease, reflecting a typical leakage pattern in the column hydraulic system. The column number precisely locates the faulty component; a support may contain 4-6 columns, and the number indicates the specific leaking column. The support attitude data at the corresponding moment includes geometric parameters such as the shield beam pitch angle, column extension / retraction, and the angles of each member of the four-bar linkage, recording the support spatial configuration at the time of leakage. Data synchronous reporting is transmitted in real-time to the digital twin monitoring platform at the ground monitoring center via industrial Ethernet. The OPCUA transmission protocol ensures data integrity and real-time performance.

[0013] The digital twin monitoring platform visualizes and recreates leaks based on the column numbers of candidate leaking supports and their corresponding posture data at specific times, generating a diagnostic report. Visualized leak reproduction is a core application of virtual-real fusion technology. By driving a 3D digital twin model, it recreates the support state at the moment of the leak, providing an intuitive presentation of the fault. The digital twin model is built based on the support's CAD model, containing complete geometric parameters, kinematic constraints, and physical properties, and can respond to real-time data from the field. During the reproduction process, the reported posture data is input into the model's kinematic solver to solve for the spatial position and posture of each member, driving the model to the target configuration. Simultaneously, the gap status of each hinge point is read from the wear gap distribution map, and the gap size is rendered at the hinge point in a visual annotation format. The leaking column is highlighted in red and a flashing warning symbol is applied to visually emphasize the fault focus. The diagnostic report is a structured output of fault information, including spatiotemporal fault location, severity assessment, development trend prediction, and maintenance recommendations, providing decision support for on-site handling and planned maintenance.

[0014] In an embodiment of the present invention, the process of constructing a wear gap distribution map includes: Vector synthesis is performed on the triaxial vibration acceleration signals at each hinge point. Peak detection is then performed on the vector synthesis results to extract the impact peak point and its corresponding timestamp at the moment of action switching. Signal segments of preset durations before and after the impact peak point are extracted as the time-domain waveform of a single impact response. The synthesis process first performs a square summation and square root operation on the triaxial acceleration components at each sampling moment to obtain the synthesized acceleration time series. Then, peak points are detected and extracted using a set peak detection threshold. The peak points correspond to the hinge point collision events at the moment of support action switching, such as the contact impact between the shield beam and the top beam hinge point when the lifting column action is started, or the sudden change in motion between the base and the shield beam hinge point during the pushing action. The timestamp and amplitude of each peak point are recorded as time positioning markers for the action events. Signal segments are then extracted forward and backward from each peak point (the preset duration is usually 50 milliseconds to ensure that the main attenuation process of the impact response is included) to form the time-domain waveform of a single impact response. This time-domain windowing method ensures that subsequent analysis focuses on the impact event itself and eliminates interference from irrelevant signals.

[0015] Based on wavelet packet decomposition, the frequency range of the time-domain waveform of the impact response is divided into several sub-bands, and the center frequency, energy proportion and decay time constant of the sub-band with an energy proportion greater than a preset energy threshold are encapsulated as the hinge point impact response feature vector. The decomposition process employs a Daubechies wavelet basis to perform 3-4 levels of wavelet packet decomposition on the time-domain waveform of the impulse response, dividing the frequency range into 8-16 sub-bands. Each sub-band corresponds to a specific frequency range. Then, the energy percentage of each sub-band is calculated, where energy is defined as the sum of the squares of the wavelet coefficients of that sub-band, and the energy percentage is the energy divided by the total energy of all sub-bands. An energy threshold is set, meaning only sub-bands with energy percentages greater than the threshold are retained for subsequent analysis, filtering out weak noise frequencies. For the retained sub-bands, three key parameters are extracted: center frequency, energy percentage, and decay time constant (obtained by exponential fitting of the wavelet coefficient envelope curve of the corresponding sub-band, reflecting the decay rate of that frequency component). Finally, the center frequency, energy percentage, and decay time constant of all retained sub-bands are arranged sequentially to form the hinge point impulse response feature vector, comprehensively characterizing the frequency structure and time decay characteristics of the impulse response, providing rich information for gap inversion.

[0016] The impact response feature vector is compared parameter-by-parameter with the baseline feature vectors of corresponding hinge points and action types in the pre-constructed impact response baseline feature library to obtain the impact response energy offset. The impact response baseline feature library is a reference standard established by collecting impact response data of new supports (without wear or with slight wear) under standard working conditions, and includes typical impact response characteristics of different hinge point positions and different action types (lifting, lowering, pushing, etc.). The comparison process first retrieves the corresponding baseline feature vector from the impact response baseline feature library according to the current hinge point number and action type; then, a parameter-by-parameter comparison is performed to calculate the center frequency offset, energy proportion offset, and decay time constant offset respectively; the center frequency offset is defined as the difference between the measured center frequency and the baseline center frequency; the energy proportion offset is defined as the ratio of the measured energy proportion to the baseline energy proportion; the decay time constant offset is defined as the difference between the measured decay time constant and the baseline decay time constant; then, the offsets of each sub-frequency band are weighted and synthesized to finally calculate the impact response energy offset, the calculation formula of which is: ; In the formula, Indicates the energy offset of the impact response; Indicates the number of reserved sub-bands; Indicates the first The weight of a sub-band (equal to the normalized value of the energy proportion of that sub-band in the baseline characteristics). and Representing the measured and baseline values ​​respectively Sub-band center frequency; and These represent the energy percentages of the i-th sub-band in the measured and baseline measurements, respectively. and Represent the measured and baseline attenuation time constants of the i-th sub-band, respectively; , , These represent the weighting coefficients for frequency offset, energy offset, and decay time offset, respectively, in this application. This reflects the dominant role of energy offset in gap determination; the energy offset integrates deviation information from multiple frequency bands and parameters, providing quantitative input for gap inversion.

[0017] The material parameters of the contact surface at the hinge point are obtained, and the gap expansion is obtained by combining the impact response energy offset. The inversion process first obtains the material parameters of the contact surface at the hinge point from the support design file and on-site measurements, including the elastic modulus E and Poisson's ratio. Surface roughness coefficient of friction These parameters determine the mechanical behavior and energy transfer characteristics of contact collisions. Then, Hertz contact theory and impact dynamics models are applied to establish the mapping relationship between gap size and impact response energy. For example, increasing the gap at the hinge point leads to an increase in the relative velocity before collision (longer free travel), thereby increasing the collision kinetic energy and impact response energy. Simultaneously, increasing the gap changes the effective contact area and contact stiffness of the contact surface, thus altering the frequency characteristics and attenuation characteristics of the impact response. Based on these physical mechanisms, the gap expansion is constructed. With energy offset The inversion formula between them: In the formula, Indicates the amount of gap expansion; The inversion coefficient is determined by a combination of contact surface material parameters and geometric parameters, and is obtained through offline calibration experiments. The equivalent mass at the hinge point is calculated using the mass of the support members and the kinematic relationship. The contact stiffness is calculated based on the material's elastic modulus and the geometry of the contact surface according to Hertz theory. The rationality of the inversion results is checked. If the calculated gap expansion is negative or exceeds the physical maximum wear, it is judged as an inversion anomaly, and the extrapolated value of historical gap data or the average value of nearby hinge points is used instead. The final gap expansion is a quantitative value of the current wear state of each hinge point, providing basic data for subsequent error propagation analysis.

[0018] The collected pitch angle data of the shield beam is converted into the geometric coordinates of each hinge point of the four-bar linkage in the corresponding attitude. Using these coordinates as constraints, the gap expansion at each hinge point is projected onto the displacement space at the end of the shield beam, and the comprehensive displacement contribution of each hinge point gap to the end of the shield beam in the current attitude is calculated. The conversion process first reads the pitch angle of the shield beam. , defined as the angle between the shield beam and the horizontal plane; then, based on the link length parameters of the four-bar linkage (top beam length L1, front link length L2, rear link length L3, shield beam length L4) and the relative positions of the hinge points, the absolute coordinates of each hinge point are solved using closed-loop vector equations; a plane coordinate system is established with the center of the support base as the origin, the horizontal direction to the right as the positive x-axis, and the vertical direction upward as the positive y-axis; the (x, y) coordinates of each hinge point are calculated by solving the closed-loop vector equations; next, based on the geometric position coordinates of each hinge point, the gap expansion is projected onto the end of the shield beam using differential kinematics; the gap expansion can be regarded as a small displacement disturbance at the hinge point, which is transmitted to the end of the mechanism through the Jacobian matrix; specifically, for the j-th hinge point, its gap expansion is... A small displacement is generated in a random direction, which is transmitted to the end of the shield beam through the linkage kinematics, resulting in an end-effector displacement contribution. Due to the randomness of the gap direction, a conservative estimate is adopted, assuming that the gap exerts its influence in the most unfavorable direction (i.e., the direction with the largest Jacobian transfer gain). The vector superposition of the end displacement contributions of all hinge points is performed, and the coupling effect of the gaps at each hinge point is considered to obtain the comprehensive displacement contribution. , where M is the total number of hinge points; the comprehensive displacement contribution reflects the combined impact of the gaps between all hinge points on the positioning accuracy of the end of the shield beam under the current support posture.

[0019] The expansion amount of the gap at each hinge point within each coal mining cycle is stored using a two-dimensional index based on the support number and the hinge point number. At the same time, the corresponding shield beam attitude interval identifier is recorded to form a wear gap distribution map. The stored procedure employs a two-dimensional array or hash table structure. The first dimension index is the support number, and the second dimension index is the hinge point number (such as the main hinge point, like the top beam-shield beam hinge point). Each storage unit contains the gap expansion value of that hinge point, the measurement timestamp, and the corresponding shield beam attitude interval identifier. Because the support undergoes actions such as raising, moving, and lowering the support during the coal mining cycle, the shield beam attitude changes dynamically, and the gap expansion may also change with the attitude (e.g., the hinge point experiences greater force and more severe wear under certain attitudes). Therefore, recording the attitude interval identifier ensures the accuracy of the query. The map data is updated according to the coal mining cycle. After each cycle, the new gap measurement result overwrites the old value of the corresponding storage unit, while historical data is retained for wear trend analysis. The final wear gap distribution map is a dynamically updated, globally covered wear state knowledge base that supports quick queries for gap expansion values ​​under any support, any hinge point, and any attitude, providing a data foundation for real-time data correction and predictive maintenance.

[0020] In embodiments of the present invention, the calculation process for the displacement deviation value at the end of the shield beam and the force transmission attenuation coefficient of the support column includes: The hinge points within the four-bar linkage are arranged into an ordered node chain according to the kinematic transmission path. Using the gap expansion of the first node in the ordered node chain as the initial input, the gap expansion is sequentially converted into the contribution of the transmission error to the end displacement of subsequent nodes using a Jacobian matrix. The error is then accumulated node by node, outputting the displacement deviation value at the end of the shield beam. During the arrangement process, the typical transmission chain of the hydraulic support four-bar linkage is first identified based on the kinematic transmission path, forming an ordered node chain. Then, the gap expansion of the first node N1 in the ordered node chain... As the initial input, the displacement disturbance generated by this gap in a random direction is transmitted to node N2 through the rear link; the transmission process uses the Jacobian matrix method, where the Jacobian matrix J1 describes the linear mapping relationship between the small displacement at node N1 and the displacement at node N2; the elements of the Jacobian matrix are calculated using the partial derivative method from geometric parameters such as link length and current attitude angle; then, the gap expansion amount of node N2 itself is... The generated local displacement disturbance and the transmitted displacement Vector superposition is performed to obtain the cumulative displacement error at node N2; then, this cumulative error is propagated to node N3 through the Jacobian matrix J2, and so on, node by node, until it is propagated to the end of the shield beam; the final output displacement deviation value at the end of the shield beam is: In the formula, This indicates the displacement deviation value at the end of the protective beam; The Jacobian matrix represents the displacement of the j-th hinge point with respect to the end point. The vector represents the Euclidean norm. Since the Jacobian matrix is ​​related to the support posture, the end displacement deviation values ​​are different under different shield beam pitch angles. Therefore, it is necessary to calculate and store them separately in each posture range. The end displacement deviation value quantifies the impact of the gap on the support positioning accuracy and provides a quantitative basis for displacement compensation of the support movement.

[0021] The hydraulic pressure transmission path of the column at each hinge point in the four-bar linkage is analyzed. The correlation between the gap expansion at each hinge point and the increase in frictional dissipation during hydraulic pressure transmission is established. The total attenuation of the overall force transmission is calculated by summing the frictional dissipation increments along the column force transmission path at each hinge point. Using the current rated bearing capacity of the column as a normalization benchmark, the force transmission attenuation coefficient of the support column is output. The analysis process first establishes the static equilibrium equation of the support, and the load of the top plate acting on the shield beam... The force is transmitted through the shield beam to the hinge points of the front and rear links, which in turn transmit the force to the base and the column; the column provides the supporting reaction force. To balance the load on the top plate; under ideal frictionless conditions, ,in The lever amplification factor is determined by the geometry of the four-bar linkage; however, gaps and friction exist at each hinge point, resulting in an increase in frictional dissipation. With gap expansion The correlation between them is established through a tribological model, and the specific correlation formula is as follows: In the formula; Indicates the coefficient of friction; This represents the normal contact force at the hinge point; This represents the diameter of the hinge pin; then, the frictional dissipation increments at each hinge point are accumulated and summed along the force transmission path to obtain the total force transmission attenuation. ; Actual measured column pressure Forces transferred from the actual roof load to the columns The relationship between them is: Therefore, the force transmission attenuation coefficient Using the rated bearing capacity of the column as the normalization benchmark, the final force transmission attenuation coefficient is obtained, which provides a quantitative compensation basis for the gap error correction of the pressure signal.

[0022] In an embodiment of the present invention, the process of correcting gap error includes: The effectiveness of the current force transmission attenuation coefficient of the support column is verified to determine whether it is within a preset reasonable range. If not, it is saturated and limited to the boundary of the preset reasonable range. The verification process first sets the preset reasonable range, which is obtained from historical data statistical analysis; among them, the reasonable range of the force transmission attenuation coefficient is set as follows: Then, based on its calculation of the final... Perform range validation, if If the value is too low, it is considered abnormally low, possibly due to an excessively large gap expansion, an overestimated friction coefficient, or a calculation error. In this case, saturation processing is performed to force the setting. ;like If the value is too high, it is considered abnormally high, possibly due to an underestimation or negative value of the gap expansion amount. In this case, saturation processing is performed, and the value is forcibly set. ;like If the value is within a reasonable range, it is considered a valid value and remains unchanged. Saturation processing ensures the numerical stability of subsequent correction calculations and avoids over-correction or under-correction caused by abnormal coefficients.

[0023] The reciprocal of the force transmission attenuation coefficient of the support column is used as the pressure compensation gain. The data from each sampling point in the pressure time-series signal corresponding to each support column is multiplied by the corresponding pressure compensation gain to obtain the corrected pressure time-series signal. The compensation process first reads the original pressure time-series signal, which is a sequence of pressure data continuously collected by the column pressure sensor during the coal mining cycle. Then, compensation calculations are performed on each sampling point to correct the pressure time-series signal. Since the pressure compensation gain is greater than 1, the corrected pressure value is not less than the measured value, reflecting that the measurement underestimation caused by force transmission attenuation has been compensated. For the attitude changes of the support during the coal mining cycle, if different attitude intervals correspond to different force transmission attenuation coefficients, it is necessary to query the corresponding force transmission attenuation coefficient from the wear gap distribution map based on the real-time attitude and dynamically update the pressure compensation gain to achieve segmented compensation. The final corrected pressure time-series signal more accurately reflects the true changes in the roof load, eliminates the systematic error of gap wear on pressure measurement, and provides high-quality input data for subsequent pressure collaborative analysis.

[0024] In an embodiment of the present invention, the process of performing alignment correction includes: Obtain the displacement deviation vector at the end of the shield beam under the current gap state. Based on the target displacement of the support movement and the magnitude of the displacement deviation vector at the end of the shield beam, calculate the displacement compensation coefficient for the movement. The acquisition process first reads the gap expansion amount at each hinge point corresponding to the current support and current attitude range from the wear gap distribution map, and then calculates the displacement deviation vector at the end of the shield beam through Jacobian matrix transfer. Then, read the target displacement of the movement. This parameter is determined by the coal mining process and is usually the step distance of the support; the displacement compensation coefficient is defined as the proportion of additional displacement required to achieve the target displacement, and the calculation formula is: In the formula, The displacement compensation coefficient represents the frame-shifting motion; This represents the magnitude of the displacement deviation vector at the end of the shield beam; this coefficient is greater than 1, reflecting the amount of displacement that needs to be increased due to the presence of the gap.

[0025] The theoretical action times of adjacent supports are read from the coordinated action instruction sequence. Based on the support movement speed parameters and displacement compensation coefficient, the action delay time caused by the gap is calculated and added to the theoretical action time to obtain the corrected actual action completion time. The coordinated action instruction sequence contains the action arrangement of adjacent supports (such as the nth support and the (n+1th support), and each action includes the action type and theoretical action time. Parameters such as the duration of the action are also included. The reading process first extracts the theoretical action time of adjacent stents from the instruction sequence, such as the theoretical completion time of the pushing action of stent n. Then read the support frame movement speed parameters. This parameter represents the extension and retraction speed of the hydraulic push jack; based on the displacement compensation coefficient, the additional displacement caused by the clearance is calculated. Dividing the additional displacement by the frame movement speed yields the motion delay time. In the formula, The delay time is represented; then the delay time is added to the theoretical action time to obtain the corrected actual action completion time; the same correction process is applied to all supports and all actions in the coordinated action sequence.

[0026] The corrected actual action completion times are arranged by stent number and action type to obtain the corrected coordinated action sequence. The arrangement process first sorts the stents in ascending order by number (from stent 1 to stent 150). For each stent, the action events are arranged in chronological order of occurrence. Each action event record includes: stent number, action type (such as "column raising", "column lowering", and "push", etc.), corrected actual action start time, corrected actual action completion time, and action duration. The resulting corrected coordinated action sequence is a time-expanded event sequence, providing a time alignment benchmark for constructing the pressure coordination relationship baseline model, ensuring the accurate correspondence between pressure analysis and action stages.

[0027] In an embodiment of the present invention, the process of constructing a benchmark model for the pressure coordination relationship between adjacent stents includes: Based on the direction of the coal mining machine's movement along the working face and the length of the working face, the fully mechanized mining face is divided into several cutting position zones, and a corresponding zone index is set for each cutting position zone. The zoning process first calculates the average length of each zone based on the length of the working face and the preset number of zones; then, the space is divided into several cutting position zones along the length of the working face, such as zone 1 corresponding to 0-30 meters, zone 2 corresponding to 30-60 meters, and so on; and a unique zone index is set for each zone; the current cutting position of the coal mining machine is obtained in real time through its positioning system (such as UWB positioning), and mapped to the corresponding zone index according to the position coordinates; the purpose of zoning is that the geological conditions (such as coal seam thickness, roof hardness) and support stress characteristics may differ at different locations of the working face, and zoning modeling can capture these spatial variation characteristics, improving the adaptability and accuracy of the benchmark model.

[0028] When the coal mining machine is in the corresponding cutting position zone, the corrected pressure time sequence signal within the corresponding time period is acquired. Synchronous time sequence segments of the pressure time sequence signals corresponding to the adjacent support columns are extracted according to the support number. Based on these, a cross-correlation coefficient sequence is obtained, and the pressure transmission delay between adjacent supports is determined by the peak position of the cross-correlation coefficient sequence. The acquisition process first records the corrected pressure time sequence signals of all supports within that time period when the coal mining machine is positioned in a certain zone. Then, for each pair of adjacent supports (such as support number n and support number n+1), pressure time sequence signal segments within the same time range are extracted. Next, the cross-correlation coefficients of the two signals are calculated, forming a cross-correlation coefficient sequence. Then, the peak value is searched in the cross-correlation coefficient sequence; the time delay corresponding to the peak position is the pressure transmission delay between adjacent supports. The pressure transmission delay is typically 0.5–3 seconds, with the specific value depending on factors such as the roof strata properties and support spacing. Determining the pressure transmission delay provides a time reference for subsequent time delay alignment and pressure difference calculation.

[0029] The pressure time-series signal is time-delay aligned using the pressure transmission delay, and the time-by-time pressure difference sequence is calculated. The mean and standard deviation of the pressure difference sequence after multiple mining cycles within the current cutting position zone are statistically analyzed. Based on these, upper and lower bounds for the allowable deviation of pressure coordination between adjacent supports under this cutting position zone are set. The alignment process first performs a time shift on the pressure time-series signal of support n+1 to obtain the time-delay aligned signal; then, the time-by-time pressure difference sequence is calculated. For the current cutting position zone, pressure difference data from multiple mining cycles (typically 5-10 cycles) are collected to form a large sample dataset. Statistical analysis is performed on the pressure difference at all sampling times, calculating the mean and standard deviation. Based on the mean and standard deviation, the 3σ criterion is used to set the upper and lower bounds for the allowable deviation. These upper and lower bounds define the normal fluctuation range of pressure coordination between adjacent supports under the current cutting position zone. Pressure differences falling within this range are considered normal, while those exceeding the range are considered abnormal.

[0030] Align and label the support action event times in the corrected coordinated action sequence with the pressure difference sequence, identify the phased change patterns of the pressure difference sequence in different coordinated action stages, and establish independent upper and lower bounds of allowable deviation for each type of coordinated action stage; the coordinated action stage refers to the typical action combination of the support group in the coal mining cycle, such as "coal cutting stage of the coal mining machine" (support bearing static load) and "column lowering stage of the front support" (pressure redistribution of adjacent supports), etc. The identification process first extracts the time markers of stent action events from the corrected coordinated action time sequence, and segments the pressure difference sequence according to the time of the action events, with each segment corresponding to a coordinated action stage. Then, the pressure difference sequence within each stage is independently statistically analyzed to calculate the mean and standard deviation specific to that stage. The analysis revealed that the distribution characteristics of pressure difference in different stages are significantly different: the pressure difference fluctuates little and the mean is stable during the static bearing stage; the pressure difference between adjacent stents increases instantaneously during the column lowering stage; the pressure difference changes rapidly during the column raising stage; and the pressure difference fluctuates greatly but has no obvious anomalies during the shifting stage. Based on the statistical characteristics of each stage, independent upper and lower bounds for allowable deviations are set for each stage. This staged modeling significantly improves the accuracy of anomaly detection and avoids misjudging normal staged pressure changes as anomalies.

[0031] A baseline record for the pressure coordination relationship benchmark model between adjacent supports is constructed by combining the partition index, the coordinated action stage, and the corresponding upper and lower bounds of the allowable deviation. The construction process employs a multi-dimensional index structure, with each baseline record containing the following fields: partition index (locating the spatial position of the working face); support pair number (identifying adjacent supports); coordinated action stage identifier (locating the coal mining cycle sequence); pressure transmission delay; mean pressure difference; standard deviation of pressure difference; upper bound of allowable deviation; lower bound of allowable deviation; data update timestamp; number of sample cycles M. The model is stored in a database or hash table, supporting quick lookup of the corresponding upper and lower bounds of allowable deviation by using the three-dimensional index of "partition index + support pair number + coordinated action stage". The baseline model is established through data accumulation over multiple cycles in the early stages of working face production and is periodically updated in subsequent production (e.g., every shift or daily) to adapt to the slow changes in geological conditions and equipment status, ensuring the timeliness and accuracy of the baseline model.

[0032] In an embodiment of the present invention, the process of obtaining the effective pressure anomaly time series includes: The sliding time window is advanced with a preset sliding step size. Within each time window, the corrected pressure timing signals corresponding to all adjacent supports are extracted. The corresponding cutting position partition index is determined based on the current cutting position of the coal mining machine, and the upper and lower bounds of the allowable deviation corresponding to the current cutting position partition and the current collaborative action stage are read from the pressure coordination relationship benchmark model. The setup process first determines the length of the sliding time window, ensuring that the window contains enough data points for statistical analysis without being too long and causing delays in anomaly detection. Then, the preset sliding step size is determined. A smaller step size results in more timely detection but a larger computational load, while a larger step size reduces computational load but may miss short-term anomalies. The window position is updated periodically with the sliding step size as the cycle. Within each time window, the corrected pressure timing signals of all adjacent support pairs are extracted. Simultaneously, the current positioning coordinates of the coal mining machine are read and mapped to the corresponding cutting position partition index. Then, the corresponding collaborative action stage at the current moment is queried from the corrected collaborative action timing. Finally, the corresponding allowable deviation upper and lower bounds are read from the pressure coordination relationship benchmark model using "partition index + support pair number + collaborative action stage" as the query key. These parameters provide the judgment criteria for subsequent over-limit event detection.

[0033] The pressure difference between adjacent supports within the current time window is calculated hourly. Time points where the pressure difference exceeds the upper or lower bound of the allowable deviation are recorded as deviation exceedance moments. A series of consecutive deviation exceedance moments within the time window are merged into a single pressure-coordinated deviation exceedance event, and the start time, duration, and maximum deviation amplitude of this event are recorded. The pressure difference calculation process first aligns the time-delay of the corrected pressure timing signals of adjacent supports (using the pressure transmission delay stored in the baseline model), then calculates the hourly pressure difference. Next, an exceedance judgment is made for the pressure difference at each moment: if the pressure difference is greater than the upper deviation bound or less than the lower bound... If the deviation falls below the lower bound, the corresponding moment is marked as a deviation exceeding the limit moment. Then, all exceeding moments are scanned, and temporally consecutive exceeding moments (with an interval less than a preset merging threshold, such as 2 seconds) are merged into a pressure coordination deviation exceeding event. Each event record includes: event start time (first exceeding moment), event end time (last exceeding moment), duration, maximum deviation amplitude, involved stent pair number, corresponding cutting position partition, and coordination action stage. These exceeding events constitute a preliminary anomaly candidate set, but may contain spurious deviations caused by gap measurement uncertainties, which require further verification.

[0034] The real-time gap expansion is obtained, and the residual of the real-time force transmission attenuation coefficient of each support at the current moment is recalculated based on it. This residual is then converted into an equivalent error band for pressure difference. If the maximum deviation amplitude of the pressure coordination deviation exceeding the limit event falls within the equivalent error band for pressure difference, it is marked as a pseudo-deviation event and discarded. If the maximum deviation amplitude exceeds the equivalent error band for pressure difference, it is retained as a valid pressure anomaly event and output to the valid pressure anomaly event sequence. The verification process first reads the real-time gap expansion of each hinge point of each support at the current moment from the wear gap distribution map. Then, based on the real-time gap expansion, the real-time force transmission attenuation coefficient of each support is recalculated, and the residual is calculated with the attenuation coefficient previously used for pressure correction. The residual reflects the real-time fluctuation or measurement uncertainty of the gap state. Then, the attenuation coefficient residual is converted into an equivalent error band for pressure difference. For adjacent supports n and n+1, the measurement uncertainty of their pressure difference is jointly determined by the attenuation coefficient residuals of the two supports. Next, each pressure coordination deviation exceeding the limit event is verified. Judgment: If the maximum deviation amplitude of the event falls within the equivalent error band, the event is determined to be caused by the uncertainty of gap measurement, marked as a false deviation event, and removed from the candidate set; if it exceeds the equivalent error band, it is confirmed as a real pressure anomaly event, retained, and output to the effective pressure anomaly event sequence; the final output effective pressure anomaly event sequence only contains real anomalies that have passed the secondary verification, providing highly reliable input data for subsequent leak diagnosis; this dual detection mechanism (benchmark model initial screening + gap residual verification) effectively balances detection sensitivity and false alarm rate, achieving robust anomaly detection.

[0035] In an embodiment of the present invention, the process of obtaining a leak candidate stent includes: The effective pressure anomaly event sequence was grouped by support number. For each support, the cumulative number of effective pressure anomaly events it participated in and the sum of the durations of each event were counted within the current coal mining cycle. The statistical process first traversed the effective pressure anomaly event sequence, recording the support pair number involved in each event. Then, groups were established by support number. Events involving support n included all events in which it was either a member of a support pair (i.e., events involved in "support n - support n+1" or "support n-1 - support n"). For each support, two key indicators were counted: the cumulative number of events (defined as the total number of effective pressure anomaly events the support participated in) and the total duration (defined as the sum of the durations of all events the support participated in). These two indicators quantified the severity of the anomaly from the dimensions of frequency and duration, respectively. Supports that frequently participated in anomaly events and had a long total duration were more likely to have a failure.

[0036] For stents whose cumulative number of events exceeds a preset abnormal frequency threshold and whose total duration exceeds a preset duration threshold, the spatial distribution characteristics of their effective pressure abnormal events are extracted: the number of times the corresponding stent participates in abnormal events as the low-pressure side and the number of times it participates in abnormal events as the high-pressure side are counted. If the proportion of the number of times the low-pressure side participates in the total number of events exceeds a preset low-pressure proportion threshold, the corresponding stent is marked as an active leakage candidate stent; if the proportion of the number of times the high-pressure side participates in the event exceeds a preset high-pressure proportion threshold, it is marked as a passive pressure compensation candidate stent. The threshold setting process first determines the preset anomaly frequency threshold, typically set at 5-10 times per cycle, representing the upper limit of the number of times the support participates in anomaly events within a coal mining cycle; exceeding this value is considered abnormally frequent. Then, a preset duration threshold is determined, typically set at 60-120 seconds per cycle, representing the upper limit of the total duration of the anomaly event. For supports that meet both the cumulative number of occurrences exceeding the preset anomaly frequency threshold and the total duration exceeding the preset duration threshold, further spatial distribution characteristic analysis is performed. This spatial distribution characteristic analysis focuses on the support's "role" in the anomaly event: if the support's own pressure is low, while the pressure of adjacent supports is normal or high, it indicates that the support has actively lost pressure, possibly due to leakage; if the support's own pressure is normal or high, while the pressure of adjacent supports is low, it indicates that the support is passively bearing pressure. Increased load is a passive response to leakage in adjacent supports. The determination process identifies the support role for each abnormal event involving support n based on the sign of the pressure difference: if the pressure of support n is less than that of the adjacent support n+1, support n is recorded as participating once as the low-pressure side; if it is greater, support n is recorded as participating once as the high-pressure side. The number of low-pressure and high-pressure side participations is counted, and the proportions of low-pressure and high-pressure sides are calculated. If the proportion of low-pressure sides is greater than a preset low-pressure proportion threshold, it is marked as an active leakage candidate support; if the proportion of high-pressure sides is greater than a preset high-pressure proportion threshold, it is marked as a passive pressure compensation candidate support. Passive compensation supports themselves are not faulty, but their presence indicates potential leakage in adjacent supports, requiring further inspection of the adjacent supports. This role differentiation significantly improves the accuracy of fault location.

[0037] Read the original pressure timing signal of the active leakage candidate stent within the same time period, calculate the pressure drop slope of the original pressure timing signal during the duration of the effective pressure anomaly event, and select stents with a pressure drop slope less than the preset leakage slope threshold as leakage candidate stents. The calculation process first reads the original pressure time-series signal of the active leakage candidate stent during the duration of the abnormal event; then, it performs linear regression fitting on the pressure data during this period to establish a pressure-time linear model, where the slope of the pressure-time linear model is the pressure drop slope; if the slope is less than zero, it indicates a pressure drop; if the slope is greater than zero, it indicates a pressure rise; if the slope is approximately equal to zero, it indicates stable pressure; for leakage faults, the slope should be significantly negative and have a large absolute value; a preset leakage slope threshold is set, and if the slope is less than the leakage slope threshold (note that the threshold is negative, so this inequality indicates that the absolute value of the slope is greater than the threshold), the stent is confirmed as a leakage candidate stent; otherwise, if the slope does not meet the condition, it may be a misjudgment or other non-leakage pressure fluctuation, and it is removed from the candidate set; this time-series slope-based verification significantly improves the accuracy of leakage diagnosis and avoids misjudging transient pressure changes as leaks.

[0038] The candidate leak support's support number, column number, leak confirmation time, pressure drop slope, and shield beam pitch angle are encapsulated into a support leak status data packet and reported to the digital twin monitoring platform. Data packet encapsulation is the process of integrating scattered diagnostic information into a structured message, facilitating network transmission and platform processing. The encapsulation process constructs a support leak status data packet containing the following fields: support number, column number, leak confirmation time, pressure drop slope, shield beam pitch angle, current column pressure value, and support coordinate position. The above information is serialized according to a predefined communication protocol format (such as JSON or ProtoBuf) to form a data packet. Then, the data packet is reported to the digital twin monitoring platform on the ground or in the central control center via the industrial Ethernet or wireless communication network of the working face. The reporting frequency is determined according to the severity of the anomaly; severe leaks (large absolute slope value) are reported immediately, while minor leaks can be reported in batches. After receiving the data packet, the digital twin monitoring platform triggers the visualization reproduction and diagnostic report generation process.

[0039] In embodiments of the present invention, the process of visualizing leak reproduction includes: In the digital twin monitoring platform, based on the factory-issued three-dimensional parametric model of all hydraulic supports in the fully mechanized mining face, an independent three-dimensional digital twin model is established for each support. The three-dimensional digital twin model includes the hinge motion constraint relationship of each link of the four-bar linkage, the parametric mapping of the column extension stroke, and the visualization parameter channel of the gap expansion at each hinge point. The three-dimensional digital twin model is the core of virtual-real fusion, and through high-fidelity geometric models and real-time data driving, it realizes the synchronous mapping between the physical world and the digital world. The modeling process begins by obtaining the factory-produced 3D CAD model (such as STEP, IGES, or proprietary formats) from the support manufacturer. This model contains the precise geometry, dimensional parameters, and assembly relationships of each component (top beam, shield beam, front link, rear link, column, base, etc.). The CAD model is then imported into a digital twin platform (such as Unity3D, UE4, or specialized industrial software) for geometric simplification and mesh optimization. This reduces model complexity while maintaining visual fidelity and ensuring real-time rendering performance. Next, kinematic constraints are added to the model, defining the hinge relationships between the links of the four-bar linkage: the top beam and shield beam are connected via revolute joints; the shield beam and front and rear links are connected via revolute joints; the links and base are connected via revolute joints; and the column is connected to the top beam and base via prismatic joints. These constraints form the input to the kinematics solver, supporting forward kinematics... The process involves calculating kinematics (finding the end-effector position given joint angles) and inverse kinematics (finding joint angles given end-effector positions). Then, a parametric mapping is established between the column's extension / retraction stroke and the 3D model: the real-time extension / retraction of the column (obtained from pressure and stroke sensors) directly drives the length change of the column components in the 3D model, achieving a smooth transition through linear interpolation. Simultaneously, visualization parameter channels are added to each hinge point: programmable rendering objects (such as dynamic textures, particle effects, or geometric annotations) are created at the 3D coordinates of the hinge points. The visual attributes of these objects (color, size, transparency, etc.) are bound to the gap expansion value, forming a data-driven visualization. Finally, independent 3D digital twin model instances are established for each of the 150 supports on the work surface. Each instance is associated with a corresponding real-time data stream based on the support number, forming a digital twin scene of the entire work surface.

[0040] The system receives data packets reporting the leakage status of candidate supports and drives the kinematic solver of the corresponding 3D digital twin model of the support based on these packets. It then solves the current spatial attitude of each member and drives the 3D digital twin model to move in real time to the spatial position corresponding to the moment of leakage. The kinematic solution is a calculation process that reconstructs the attitude of the support based on real-time data, ensuring that the virtual model is synchronized with the actual state of the site. The solution process first extracts key attitude parameters from the support leakage status data package: the pitch angle of the shield beam, the extension and retraction of the column, and the support height, etc. These parameters are then input into a kinematics solver. Based on the closed-loop vector equations and constraints of the four-bar linkage, the solver calculates the spatial attitude (including position coordinates and rotation angles) of each link. Specifically, for a planar four-bar linkage, a coordinate system is established with the base center as the origin. Based on the pitch angle of the shield beam and the link length parameters, the coordinates of each hinge point are solved using the cosine and sine theorems. The angle values ​​satisfying the closed-loop constraints are solved iteratively. After the solution is completed, the spatial attitude parameters (position, rotation) of each link are assigned to the transformation matrix of the corresponding component in the 3D digital twin model, driving the model to update in real time. The update uses an interpolation algorithm (such as spherical linear interpolation Slerp) to make the model's motion smooth and natural. Finally, each link of the 3D digital twin model moves precisely to its spatial position at the moment of leakage, achieving virtual-real synchronization and providing operators with a "what you see is what you get" on-site state reproduction.

[0041] The current gap expansion amount of each hinge point of the leak candidate support is read from the wear gap distribution map. The gap expansion amount is then mapped to the visualization parameter channel of each hinge point. The wear state of each hinge point is presented in the 3D digital twin model by rendering a visual annotation ring for the gap expansion amount at each hinge point. The mapping process first queries the current gap expansion amount of each hinge point of the leak candidate support (such as the top beam-shield beam hinge point, the shield beam-front connecting rod hinge point, etc.) from the wear gap distribution map. Then, visualization mapping rules are designed: gap expansion amount less than 1mm (slight wear) is mapped to a green annotation ring; 1-3mm (moderate wear) is mapped to a yellow annotation ring; and greater than 3mm (severe wear) is mapped to a red annotation ring. The size of the annotation ring is proportional to the gap expansion amount. ,in To label the ring radius, Based on the radius, The scaling factor is used. Next, in the 3D digital twin model, the 3D coordinates of each hinge point are located (calculated by the kinematics solver). A torus or particle effect is created at this location, and its color, radius, and other visual attributes are set. The torus is rendered around the hinge pin axis, highlighting the wear status with a striking color and size. At the same time, numerical labels can be added to display the specific gap expansion value (e.g., "Gap: 2.3mm"). To enhance readability, dynamic effects (e.g., blinking torus, pulse amplification, etc.) are applied to severely worn hinge points to attract user attention. Finally, each hinge point in the 3D digital twin model presents a color-coded visual torus, allowing users to easily identify which hinge points are severely worn and which are close to normal, providing an intuitive basis for maintenance decisions.

[0042] Differential color rendering is applied to the columns confirmed to be leaking in the 3D digital twin model of the candidate leaking support, and a warning symbol layer that flashes over time is added to the surface of the 3D model of the leaking column. The rendering process first locates the corresponding column component (e.g., front column, rear column) in the 3D digital twin model based on the column number in the support leak status data package; then, the material of this column component is differentially rendered: the material color of normal columns is set to gray or metallic, while the material color of the leaking column is set to a striking red or orange, creating a strong visual contrast; simultaneously, the emission intensity of the leaking column's material is increased, making it self-illuminating in the scene and stand out even against complex backgrounds; next, a warning symbol layer is added to the surface of the 3D model of the leaking column: common warning symbols (such as triangle warning signs, exclamation marks, lightning bolt symbols, etc.) are selected to... Texture mapping or overlay geometry is used to render the column surface; the warning symbol uses a time-flashing animation effect, which is adjusted by modulating the transparency or color brightness to flash at a frequency of 1-2Hz, simulating the effect of on-site warning lights and enhancing the sense of urgency; the flashing animation is implemented through shader programs or animation controllers, and the time parameters are driven by the system clock; in addition, floating text labels can be displayed near the leaking column, such as indicating key information like "column leak"; finally, the leaking column is prominently presented in the 3D digital twin scene as a red self-illuminating, flashing warning symbol, and users can observe the leak status from any angle by rotating and zooming in a 3D view, achieving intuitive and efficient visualization and reproduction of the fault.

[0043] In an embodiment of the present invention, the process of generating a diagnostic report includes: Starting from the earliest effective pressure anomaly, the timeline extends back to find the turning point when the pressure coordination deviation first rises continuously from zero, which is taken as the leakage initiation time. Tracing back to the leakage initiation time is key to identifying the nascent stage of a leak, helping to assess the leakage development process and predict future trends. The backtracking process first extracts all events involving leak candidate stents from the effective pressure anomaly event sequence, sorts them by event start time, and determines the start time of the earliest event. Then, using the start time as the backtracking starting point, it queries the historical database for the corrected pressure timing signal and pressure coordination deviation data of the stent at an earlier time. Pressure coordination deviation is defined as the deviation of the real-time pressure difference from the mean of the baseline model. The search process scans the pressure coordination deviation curve second by second backward from the start time to identify the turning point when the deviation first rises continuously from zero: a deviation baseline threshold is set, and the time that meets the following conditions is searched backward and recorded as the first time: and in the period before the first time (e.g., the first 60 seconds), the deviation is consistently less than the deviation baseline threshold; in the period after the corresponding first time (e.g., the last 60 seconds), the deviation continues to rise and no longer falls back below the baseline; accordingly, the first time is the leakage start time, marking the first occurrence of the leakage failure. If historical data is insufficient or the deviation curve does not have a clear turning point, the start time is used as a conservative estimate of the leakage start time. The determination of the leakage start time provides a time benchmark for subsequent leakage development trend analysis and remaining life estimation.

[0044] The system reads continuously corrected pressure time-series signals from the leak initiation time to the current time, calculates the end-of-cycle pressure value of the corresponding support column pressure in each coal mining cycle, and forms a pressure decay curve. Based on the current slope of the pressure decay curve, combined with the lower limit of the rated working pressure of the support column and the current column pressure value, the estimated remaining working time of the support is calculated. This estimate is then compared with the shift schedule of the coal mining face to output a priority recommendation on whether the leak candidate support needs to be immediately shut down in the current shift. The pressure decay curve is the core data for quantifying the leak development speed and predicting the remaining life. The construction process first reads the leak candidate support from the historical database from the leak initiation time... Up to the current moment Continuous correction pressure timing signal The time span may cover several hours to several days; then, it is divided into segments according to the coal mining cycle. Each coal mining cycle corresponds to the coal mining machine traveling back and forth along the working face once. At the end of each cycle, the column pressure value is extracted and recorded as the cycle end pressure value. Where k1 is the cycle number; the pressure value at the end of the cycle is measured during the stable bearing stage of the support (such as when the coal mining machine is far from the support position) to avoid dynamic disturbances; the pressure values ​​at the end of the cycle are arranged in time series to form a pressure decay curve; the pressure decay curve is linearly or exponentially fitted to extract the decay slope: linear fitting ,in The slope represents the recession, with negative values ​​indicating a decrease in pressure; exponential fit. ,in The attenuation coefficient is used; then, based on the slope of the pressure decay curve at the current moment... (Obtained by fitting data from the most recent 3-5 cycles, reflecting the current leakage rate), calculate the remaining working time of the support; the calculation formula is based on the lower pressure limit constraint: the column pressure must not be lower than the lower limit of the rated working pressure. Otherwise, the support structure will lose its effective support capacity; the remaining working time is estimated as follows: In the formula, Indicates the remaining working time; This represents the current column pressure value. The current pressure decline slope is used; next, the shift schedule of the coal mining face is read and compared. And the remaining time of the current shift; if <Remaining time for the current shift indicates that the support structure may fail within the current shift; the recommended output priority is "Emergency - Immediate Shutdown". If the remaining time for the current shift is ≤ ≤2 × remaining time of the current shift indicates that the support structure can be maintained until the end of the current shift, but may fail in the next shift. The recommended output priority is "high - maintain immediately after shift ends"; if A value greater than 2 × the remaining time of the current shift indicates that the support structure is safe in the short term, and the recommended output priority is "medium-planned maintenance". This priority determination based on remaining lifespan and shift time provides scientific decision support for on-site managers, balancing safety and production continuity.

[0045] The structured diagnostic report is assembled from the support number, column number, leakage start time, pressure decay curve, estimated remaining working time, priority recommendations, and screenshots of the 3D digital twin model of the support corresponding to the leakage occurrence time of the candidate leak support. The assembly process follows the predefined report template, filling in the corresponding columns with the following information: (1) Report header, which includes metadata such as report generation time, working face number, report type (hydraulic support leakage diagnosis), and report number; (2) Fault summary, which lists the support number, column number, leakage start time, leakage confirmation time, current column pressure, pressure drop slope, remaining working time, and priority recommendations (such as "maintain immediately after the high-shift ends") in text and table format; (3) Pressure decay curve chart, which plots a line graph with time as the horizontal axis and cycle end pressure as the vertical axis, marking key points such as leakage start time, current time, and pressure lower limit line to intuitively show the pressure drop trend; the chart can be generated using a data visualization library; (4) Screenshots of the 3D digital twin model of the support. , capture three-dimensional scene images from multiple perspectives (such as front view, side view, top view), the leaking column in the screenshot is highlighted in red, the gap of the hinge point is marked with a color ring, and the warning symbol flashes (displayed as an overlay symbol in the static screenshot), providing maintenance personnel with an intuitive presentation of the fault location and status; (5) Cause analysis and suggestions, based on the pressure decay curve shape, gap wear status, historical maintenance records and other information, the expert system or manual analysis gives possible causes of leakage (such as aging of column seal ring, loosening of hydraulic pipeline joints, one-way valve failure, etc.), and provides targeted maintenance suggestions (such as replacing seals, tightening joints, checking hydraulic system, etc.); finally, the above content is assembled into a structured diagnostic report in PDF or HTML format, displayed through the user interface of the digital twin monitoring platform, and supports downloading, printing and distribution to relevant personnel; the diagnostic report realizes the whole process closed loop from data collection, feature extraction, anomaly detection, fault diagnosis to decision support, providing a complete technical solution for the intelligent operation and maintenance of coal mine fully mechanized mining faces.

[0046] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0047] All formulas in this manual are dimensionless and calculated numerically. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0048] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A digital twin monitoring system for fully mechanized coal mining faces based on virtual-real fusion, characterized in that, include: The data acquisition module is used to acquire real-time operating status data of the hydraulic support group in the fully mechanized mining face during the coal mining cycle. The real-time operating status data includes the pressure timing signal of each support column, the triaxial vibration acceleration signal of the hinge point of the four-bar linkage, the pitch angle data of the shield beam, and the sequence of coordinated action commands between adjacent supports. The parameter generation module extracts the impact response characteristics of each hinge point based on the triaxial vibration acceleration signal, and constructs a wear gap distribution map using the pitch angle data of the shield beam as a geometric constraint; and calculates the displacement deviation value at the end of the shield beam and the force transmission attenuation coefficient of the support column based on it. The benchmark modeling module corrects the gap error of each pressure timing signal based on the force transmission attenuation coefficient of the support column to obtain the corrected pressure timing signal. It also aligns and corrects the action time in the coordinated action command according to the displacement deviation value at the end of the shield beam to obtain the corrected coordinated action timing. A benchmark model of pressure coordination relationship is constructed based on the modified pressure time series signal and the modified coordination time series. The dual verification module is used to continuously collect the corrected pressure time sequence signal of each support in a sliding time window of preset duration, and extract pressure coordination deviation exceeding the limit event by combining the pressure coordination relationship benchmark model. Based on the wear gap distribution map, the pressure coordination deviation exceeding the limit event is verified a second time to obtain the effective pressure anomaly time sequence. The diagnostic generation module performs anomaly screening on each stent based on valid abnormal time series to obtain leakage candidate stents. It synchronously reports the column number of the leakage candidate stent and the stent attitude data at the corresponding time to the digital twin monitoring platform. Based on the column number of the leakage candidate stent and the stent attitude data at the corresponding time, it performs visual leakage reproduction and generates a diagnostic report.

2. The digital twin monitoring system for fully mechanized coal mining faces based on virtual-real fusion as described in claim 1, characterized in that, The process of constructing a wear gap distribution map includes: Vector synthesis is performed on the triaxial vibration acceleration signals of each hinge point, and peak detection is performed on the vector synthesis results to extract the impact peak point and its corresponding timestamp at the moment of action switching. Signal segments of preset duration before and after the impact peak point are extracted as the time domain waveform of a single impact response. Based on wavelet packet decomposition, the frequency range of the time-domain waveform of the impact response is divided into several sub-bands, and the center frequency, energy proportion and decay time constant of the sub-band with energy proportion greater than the preset energy threshold are encapsulated as the impact response feature vector of the hinge point. The impact response feature vector is compared parameter by parameter with the baseline feature vectors of the corresponding hinge point and the corresponding action type in the pre-constructed impact response baseline feature library to obtain the impact response energy offset. Obtain the contact surface material parameters of the hinge point, and combine them with the impact response energy offset to perform gap inversion and obtain the gap expansion amount; The collected pitch angle data of the shield beam is converted into the geometric position coordinates of each hinge point of the four-bar linkage in the corresponding attitude; and with the geometric position coordinates of each hinge point as constraints, the expansion of the gap between each hinge point is projected to the displacement space of the end of the shield beam, and the comprehensive displacement contribution of the gap between each hinge point to the end of the shield beam is calculated under the current attitude of the shield beam. The expansion amount of the gap at each hinge point within each coal mining cycle is stored using a two-dimensional index based on the support number and the hinge point number. At the same time, the corresponding shield beam attitude interval identifier is recorded to form a wear gap distribution map.

3. The digital twin monitoring system for fully mechanized coal mining faces based on virtual-real fusion as described in claim 1, characterized in that, The calculation process for the displacement deviation at the end of the protective beam and the force transmission attenuation coefficient of the support column includes: The hinge points in the four-bar linkage are arranged into an ordered node chain according to the kinematic transmission path. The gap expansion of the first node in the ordered node chain is used as the starting input. The gap expansion is converted into the transmission error contribution to the end displacement of the subsequent nodes through the Jacobian matrix according to the node sequence. The error is accumulated node by node, and the end displacement deviation value of the shield beam is output. The hydraulic pressure transmission path of the column at each hinge point in the four-bar linkage is analyzed. The correlation between the gap expansion at each hinge point and the increase in frictional dissipation during hydraulic pressure transmission is established. The total attenuation of the overall force transmission is calculated by summing the frictional dissipation increment at each hinge point along the force transmission path of the column. The force transmission attenuation coefficient of the support column is output with the current rated bearing capacity of the column as the normalization benchmark.

4. The digital twin monitoring system for fully mechanized coal mining faces based on virtual-real fusion as described in claim 1, characterized in that, The process of correcting gap errors includes: The effectiveness of the force transmission attenuation coefficient of the current support column is verified to determine whether it is within the preset reasonable range. If it is not, node processing is performed on it and it is restricted to the boundary of the preset reasonable range. The reciprocal of the force transmission attenuation coefficient of the support column is used as the pressure compensation gain, and the data of each sampling point in the pressure time sequence signal corresponding to each support column is multiplied with the corresponding pressure compensation gain to obtain the corrected pressure time sequence signal.

5. The digital twin monitoring system for fully mechanized coal mining faces based on virtual-real fusion as described in claim 1, characterized in that, The process of alignment correction includes: Obtain the displacement deviation vector of the shield beam end under the current gap state, and calculate the displacement compensation coefficient of the frame movement based on the target displacement of the frame movement and the magnitude of the displacement deviation vector of the shield beam end. Read the theoretical action time of adjacent supports in the coordinated action instruction sequence, calculate the action delay time caused by the gap based on the support moving speed parameters and displacement compensation coefficient, and add it to the theoretical action time to obtain the corrected actual action completion time. Arrange the actual completion times of the corrected actions according to the support number and action type to obtain the corrected coordinated action sequence.

6. The digital twin monitoring system for fully mechanized coal mining faces based on virtual-real fusion as described in claim 1, characterized in that, The process of constructing a benchmark model for the pressure coordination relationship between adjacent stents includes: Based on the direction of the coal mining machine's movement along the working face and the length of the working face, the fully mechanized mining face is divided into several cutting position zones, and a corresponding zone index is set for each cutting position zone. When the coal mining machine is in the corresponding cutting position zone, the corrected pressure timing signal in the corresponding time period is obtained, and the synchronous timing segment of the pressure timing signal corresponding to the adjacent support column is extracted according to the support number. Based on this, the cross-correlation coefficient sequence is obtained, and the pressure transmission delay between adjacent supports is determined by the peak position of the cross-correlation coefficient sequence. The pressure transmission delay is used to align the corresponding pressure time sequence signal with time delay, and the time-by-time pressure difference sequence is calculated. The mean and standard deviation of the pressure difference sequence after multiple coal mining cycles within the current cutting position zone are statistically analyzed, and the upper and lower limits of the allowable deviation of the pressure coordination between adjacent supports under the cutting position zone are set based on these. Align and label the stent action event time in the corrected coordinated action timeline with the pressure difference sequence, identify the stage change pattern of the pressure difference sequence in different coordinated action stages, and establish independent upper and lower bounds of allowable deviation for each type of coordinated action stage. The partition index, the coordinated action stage, and the corresponding upper and lower bounds of the allowable deviation together constitute a benchmark record of the benchmark model of the pressure coordination relationship between adjacent stents.

7. The digital twin monitoring system for fully mechanized coal mining faces based on virtual-real fusion as described in claim 1, characterized in that, The process of obtaining the effective pressure anomaly time series includes: The sliding time window is advanced with a preset sliding step size. Within each time window, the corrected pressure timing signal corresponding to all adjacent supports is extracted. The corresponding cutting position partition index is determined according to the current cutting position of the coal mining machine. The upper and lower bounds of the allowable deviation corresponding to the current cutting position partition and the current collaborative action stage are read from the pressure coordination relationship benchmark model. Calculate the pressure difference between adjacent supports at each time step within the current time window. Record the time when the pressure difference exceeds the upper and lower limits of the allowable deviation as a time when the deviation exceeds the limit. Merge the sequence of consecutive time when the deviation exceeds the limit within the time window into a pressure-coordinated deviation exceeding event, and record the start time, duration and maximum deviation amplitude of the event. The real-time gap expansion is obtained, and the residual of the real-time force transmission attenuation coefficient of each support at the current moment is recalculated based on it. This residual is then converted into an equivalent error band of pressure difference. If the maximum deviation amplitude of the pressure coordination deviation exceeding the limit event falls within the equivalent error band of pressure difference, it is marked as a pseudo-deviation event and removed. If the maximum deviation amplitude exceeds the equivalent error band of pressure difference, it is retained as a valid pressure anomaly event and output to the valid pressure anomaly event sequence.

8. The digital twin monitoring system for fully mechanized coal mining faces based on virtual-real fusion as described in claim 1, characterized in that, The process of obtaining leak-prone candidate stents includes: The effective pressure anomaly event sequence is grouped by support number, and the cumulative number of effective pressure anomaly events in which each support participates and the sum of the duration of each event are counted within the current coal mining cycle. For stents whose cumulative number of events exceeds a preset abnormal frequency threshold and whose total duration exceeds a preset duration threshold, the spatial distribution characteristics of their effective pressure abnormal events are extracted: the number of times the corresponding stent participates in abnormal events as the low-pressure side and the number of times it participates in abnormal events as the high-pressure side are counted. If the proportion of the number of times the low-pressure side participates in the total number of events exceeds a preset low-pressure proportion threshold, the corresponding stent is marked as an active leakage candidate stent; if the proportion of the number of times the high-pressure side participates in the event exceeds a preset high-pressure proportion threshold, it is marked as a passive pressure compensation candidate stent. Read the original pressure timing signal of the active leakage candidate stent in the same time period, calculate the pressure drop slope of the original pressure timing signal during the duration of the effective pressure abnormal event, and take the stent with a lower than the preset leakage slope threshold as the leakage candidate stent. The candidate leak support's support number, column number, leak confirmation time, pressure drop slope, and shield beam pitch angle are packaged into a support leak status data packet and reported to the digital twin monitoring platform.

9. The digital twin monitoring system for fully mechanized coal mining faces based on virtual-real fusion as described in claim 1, characterized in that, The process of visualizing and reproducing a leak includes: In the digital twin monitoring platform, based on the factory-issued three-dimensional parametric model of all hydraulic supports in the fully mechanized mining face, an independent three-dimensional digital twin model is established for each support. The three-dimensional digital twin model includes the hinge motion constraint relationship of each link of the four-bar linkage, the parametric mapping of the column extension stroke, and the visualization parameter channel of the gap expansion at each hinge point. Receive the data packet of the leakage status of the candidate leakage support and drive the kinematic solver of the corresponding three-dimensional digital twin model of the support based on it to solve the current spatial attitude of each link and drive each link of the three-dimensional digital twin model to move in real time to the spatial position corresponding to the moment of leakage. The current gap expansion amount of each hinge point of the leak candidate support is read from the wear gap distribution map. The gap expansion amount value is mapped to the visualization parameter channel of each hinge point. The wear state of each hinge point is presented in the three-dimensional digital twin model by rendering a visual annotation ring of the gap expansion amount at the hinge point. Differential color rendering is applied to the column confirmed to have leaked in the three-dimensional digital twin model of the leak candidate support, and a warning symbol layer that flashes over time is added to the surface of the three-dimensional model of the leaking column.

10. The digital twin monitoring system for fully mechanized coal mining faces based on virtual-real fusion as described in claim 1, characterized in that, The process of generating a diagnostic report includes: Using the start time of the earliest effective pressure anomaly as the starting point for retrospection, we extend back to find the turning point when the pressure coordination deviation first rises continuously from zero, which is taken as the leakage start time. Read the continuously corrected pressure timing signal from the leak initiation time to the current time, calculate the end-of-cycle pressure value of the column pressure corresponding to the leak candidate support in each coal mining cycle, and form a pressure decay curve; based on the current slope value of the pressure decay curve, combined with the lower limit of the rated working pressure of the support column and the current column pressure value, calculate the estimated remaining working time of the support; compare it with the shift schedule of the coal mining face, and output a priority suggestion on whether the leak candidate support needs to be shut down immediately in the current shift; The structured diagnostic report is assembled from the stent number, column number, leakage initiation time, pressure decay curve, estimated remaining working time, priority recommendations, and a screenshot of the 3D digital twin model of the stent corresponding to the leakage occurrence time of the candidate stent.

Images