Working face sensor dynamic activation management method and system based on digital twinning

Abstract

The invention provides a working face sensor dynamic activation management method and system based on digital twinning, and provides a life and death sensor mechanism for solving the problems of model updating lag, monitoring blind areas and insufficient geological adaptability caused by frequent manual movement of a physical sensor in the prior art. A virtual sensor array covering the full-length range of a working face is preset in a digital twin model, a virtual sensor activation / deactivation update rule is triggered in real time in combination with a propulsion distance, the sensor coverage range of the digital twin model and a physical scene is dynamically synchronized, and manual intervention delay is eliminated. The system automatically reduces the distance between the virtual sensors according to geological changes, predicts missing data through the GAN, and improves the monitoring density of a high-risk area. Kalman filtering calibration errors, isolation forest algorithm detection abnormity and reinforcement learning optimization threshold values are integrated, and monitoring continuity is guaranteed. According to the method, the model maintenance complexity can be remarkably reduced, the monitoring precision is improved, and key technical support is provided for intelligent mine construction.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent coal mine monitoring, and in particular to a method and system for dynamic activation management of working face sensors based on digital twins. Background Art

[0002] In the field of coal mining, working face monitoring is a core link in ensuring safe production. Currently, traditional monitoring systems rely on the deployment of physical sensors (such as pressure, displacement, and gas concentration sensors) to collect data in real time to feedback equipment status and environmental parameters. However, the inventors of this application have discovered through research that the existing technology has the following significant drawbacks:

[0003] (1) During the working face advancement process, physical sensors need to be frequently moved and adjusted manually, resulting in high installation costs, low maintenance efficiency, and the possibility of monitoring blind spots during movement. Especially in the longwall mining scenario of coal mines, the sensor coverage range is limited (usually only 300 meters), which makes it difficult to adapt to the dynamic advancement requirements, resulting in delayed model updates.

[0004] (2) Although existing digital twin technology can construct a virtual working surface model, it lacks the ability to simulate the dynamic behavior of sensors. Virtual sensors usually statically map physical devices and cannot automatically adjust their activation range as the working surface advances. This leads to the accumulation of synchronization errors between the model and the actual scene, affecting monitoring accuracy.

[0005] (3) Geologically abnormal areas such as faults and fracture zones require dense deployment of sensors. However, the traditional fixed-spacing arrangement makes it difficult to flexibly respond to geological changes. Furthermore, physical sensors are limited by hardware constraints such as explosion protection and power consumption, making it impossible to achieve high-density coverage. Furthermore, when sensors fail or data is missing, the system lacks an effective analog compensation mechanism, which can easily lead to monitoring interruptions or misjudgments.

[0006] (4) The interaction between physical sensor data and virtual models relies on manual configuration, the communication protocol is not unified, and the data processing chain is lengthy, making it difficult to meet real-time requirements.

[0007] Existing technologies have attempted to address these issues by optimizing sensor network topology or enhancing communication protocols, but none have fundamentally resolved the challenges of model synchronization and sensor management in dynamic advancement scenarios. Therefore, a method that integrates digital twin technology, dynamically simulates sensor states, and achieves high-precision data mapping is urgently needed to improve the intelligent monitoring of coal longwall mining under complex geological conditions. Summary of the Invention

[0008] In response to the technical problems in the existing coal mine working face monitoring system, such as the need for frequent manual movement and adjustment of physical sensors leading to monitoring blind spots, the accumulation of synchronization errors between the static mapping of digital twin models and the actual scene, the difficulty of fixed sensor arrangements to adapt to the intensive monitoring needs of geological anomaly areas, and the reliance on manual configuration of multi-source data interaction leading to insufficient real-time performance, the present invention provides a dynamic activation management method for working face sensors based on digital twins. The method is based on digital twin modeling of a dynamic "life and death sensor" mechanism, and achieves the following goals through adaptive activation rules of virtual sensor arrays, real-time data fusion algorithms and anomaly self-healing strategies: Eliminate manual intervention delays: automatically synchronize the sensor coverage of the model and the physical scene through the dynamic activation and deactivation logic of virtual sensors; enhance geological anomaly adaptability: dynamically adjust the virtual sensor spacing and monitoring frequency based on the fault and fracture zone detection results to solve the problem of insufficient physical deployment density; optimize data interaction efficiency: integrate real-time physical data, historical records and predicted trajectories, calibrate thresholds and synchronization errors through machine learning algorithms, and improve model accuracy; ensure monitoring continuity: introduce anomaly detection and virtual data coverage mechanisms to seamlessly switch data sources when physical sensors fail to ensure uninterrupted monitoring.

[0009] In order to solve the above technical problems, the following technical solutions are adopted:

[0010] The dynamic activation management method of working face sensors based on digital twins includes the following steps:

[0011] S1. Using 3D geological modeling software in the digital twin system, a high-precision digital twin model of the coal mine working face transport and return air lanes is generated based on the actual geological data of the coal mine. The model data sources include lane profiles, rock hardness distribution, and geological anomaly area data. A virtual sensor array is preset along the working face advancement direction. The virtual sensor array covers the entire length of the working face. The coordinate position of each virtual sensor is bound to the actual sensor of the physical working face through a unique identifier, and the virtual sensor receives monitoring data from the physical sensor in real time.

[0012] S2. Define the initial activation state of the virtual sensor array in the digital twin model. The state includes: virtual sensor number and grouping information, divided by level and tunnel segments, with only a predetermined number of sensors before the head end being in an active state, and the rest marked as dead; establish a mapping relationship between the virtual sensor coordinates and the longitude and latitude of the physical sensors; and define data synchronization mapping rules, including normalization, timestamp alignment, and conflict correction mechanisms for the monitoring data of the physical sensors.

[0013] S3. The shearer's cutting position and propulsion speed are acquired in real time through the propulsion system of the physical working face. The working face propulsion distance is calculated in combination with the inertial navigation unit data. The propulsion distance is synchronized with the digital twin model to dynamically update the distance between the virtual sensor and the propulsion end. When the distance between the propulsion end and the nearest currently activated virtual sensor reaches a preset threshold, the virtual sensor state switching update rule is triggered.

[0014] S4. According to the state switching update rule, the following operations are performed in the digital twin model: deactivate the virtual sensor in the nearest active state, freeze the data input and output channels, mark it as a black icon in the visual interface, and record the deactivation time and reason in the database; at the same time, activate the virtual sensor in the dead state at the back end, initialize the data mapping relationship of the activated virtual sensor, and start the simulation data generation module. The simulation data generation module trains a generative adversarial network based on the historical data of the physical sensor to predict missing values, marks it as a red flashing icon in the visual interface, and sends a status update notification to the third-party monitoring platform through the API interface of the digital twin system;

[0015] S5. Loop through steps S3 and S4 until the working face advances to the stop-mining line, whereupon the loop is stopped. This allows for dynamic adjustment of the virtual sensor's activation range, and calibration of the model synchronization error through the Kalman filter to ensure that the real-time mapping error is less than 5%.

[0016] Furthermore, the distribution of the virtual sensor array in step S1 is equidistant or unequally spaced, and the unequal spacing is dynamically adjusted based on the geological anomaly detection results: after the fault or fracture zone is identified by the microseismic monitoring system, the virtual sensor spacing in the high-risk area is automatically reduced to 10 to 15 meters and the monitoring frequency is increased to 1 time per second. The virtual sensor spacing in the normal rock formation area is expanded to 20 to 30 meters and the monitoring frequency is adjusted to 1 time every 5 seconds.

[0017] Furthermore, step S2 also includes real-time analysis of physical sensor data using an isolation forest algorithm. If an abnormality is detected in the physical sensor data and it persists for more than 30 seconds without recovery, the data stream is switched to the virtual sensor data stream and a maintenance alert is sent to the monitoring platform.

[0018] Furthermore, the preset threshold in step S3 is a dynamic distance value between 5 meters and 20 meters. The threshold optimization process includes calculating the single advancement step length based on the coal mining machine cutting depth, statistically analyzing the standard deviation of historical advancement data through a sliding window algorithm, constructing a topological network based on the distribution density of virtual sensors, using a graph convolutional neural network to predict the optimal trigger distance, and dynamically adjusting the threshold in combination with geological parameters.

[0019] Furthermore, the state switching update operation in step S4 is implemented through the state machine module of the digital twin system. The state machine module includes two states: sleep and death. The state migration is triggered by the threshold of the working surface distance to the nearest sensor; the death state executes to shut down the data acquisition thread, release memory resources and is marked as a black circular icon in the visual interface. The activation state starts the data simulation thread and binds the physical sensor data stream, which is marked as a red flashing icon in the interface and displays the real-time monitoring value.

[0020] Furthermore, in step S1, efficient interaction between physical sensor data and the digital twin model is achieved through the OPC UA server and the Profinet / Wi-Fi 6 dual-channel communication protocol, and step S4 supports the RESTful API interface to access a third-party monitoring platform.

[0021] The present invention also provides a working face sensor dynamic activation management system based on digital twin, comprising:

[0022] The digital twin model module is used to use 3D geological modeling software in the digital twin system to generate high-precision digital twin models of the coal mine working face transport and return air lanes based on the actual geological data of the coal mine. The model data source includes the lane profile, rock hardness distribution, and geological anomaly area data. A virtual sensor array is preset along the working face advancement direction. The virtual sensor array covers the entire length of the working face. The coordinate position of each virtual sensor is bound to the actual sensor of the physical working face through a unique identifier, and the module receives monitoring data from the physical sensors in real time.

[0023] The sensor state definition module is used to define the initial activation state of the virtual sensor array in the digital twin model. The state includes: virtual sensor number and grouping information, divided by level and tunnel segments, only a predetermined number of sensors before the head end are in the active state, and the rest are marked as dead; establish a mapping relationship between the virtual sensor coordinate position and the longitude and latitude of the physical sensor; define data synchronization mapping rules, including normalization processing of the monitoring data of the physical sensor, time stamp alignment and conflict correction mechanism;

[0024] The sensor status monitoring module is used to obtain the shearer cutting position and propulsion speed in real time through the propulsion system of the physical working face, calculate the working face propulsion distance in combination with the inertial navigation unit data, and synchronize the propulsion distance to the digital twin model to dynamically update the distance between the virtual sensor and the propulsion end. When the distance between the propulsion end and the nearest currently activated virtual sensor reaches a preset threshold, the virtual sensor state switching update rule is triggered.

[0025] A sensor status update module is configured to perform the following operations in the digital twin model according to the state switching update rule: deactivate the virtual sensor in the nearest active state, freeze the data input and output channels, mark it as a black icon in the visual interface, and record the deactivation time and reason in the database; at the same time, activate the virtual sensor in the back-end dead state, initialize the data mapping relationship of the activated virtual sensor, and start the simulation data generation module. The simulation data generation module trains a generative adversarial network based on the historical data of the physical sensor to predict missing values, marks it as a red flashing icon in the visual interface, and sends a status update notification to the third-party monitoring platform through the API interface of the digital twin system;

[0026] The state machine module is used to cyclically execute the sensor status monitoring module and the sensor status update module until the advance length along the working face reaches the stop-mining line. In this way, the activation range of the virtual sensor is dynamically adjusted, and the model synchronization error is calibrated through the Kalman filter to ensure that the real-time mapping error is less than 5%.

[0027] Compared with the existing technology, the working face sensor dynamic activation management method and system based on digital twin provided by the present invention has the following advantages:

[0028] 1. The present invention uses a "life and death sensor" mechanism to preset a virtual sensor array and dynamically trigger activation / deactivation rules (such as advancing distance threshold triggering), automatically synchronizing the sensor coverage of the digital twin model and the physical scene, eliminating manual adjustment delays and achieving millisecond-level synchronization with an error of less than 5%. This overcomes the problem of traditional technology relying on manual movement of physical sensors, resulting in monitoring blind spots and model update lags.

[0029] 2. The present invention dynamically adjusts the virtual sensor spacing (reduced to 10-15 meters in high-risk areas) and monitoring frequency (increased to once per second) based on microseismic monitoring data, and generates missing data through GAN (generative adversarial network), solving the problem that physical sensors cannot be deployed in high density due to explosion-proof and power consumption limitations. It overcomes the difficulty of existing fixed-spacing sensor arrangements in responding to the intensive monitoring needs of high-risk areas such as faults and fracture zones.

[0030] 3. The present invention integrates Kalman filtering (real-time mapping error calibration), sliding window weighted averaging (eliminating instantaneous noise) and graph convolutional neural network (GCN dynamically optimizes the trigger threshold), combined with the coal mining machine cutting depth (0.8 to 1.5 meters) and geological parameters (reducing the threshold when the roof pressure is >25MPa), significantly improving the accuracy and stability of the digital twin model, overcoming the problems of traditional models lacking real-time data calibration capabilities and serious synchronization error accumulation.

[0031] 4. This invention uses an isolation forest algorithm to analyze physical sensor data in real time. If a data anomaly is detected for more than 30 seconds, it automatically switches to the virtual sensor data stream and triggers a multi-level alarm (interface warning, voice prompt, model freeze). Simultaneously, it uses a reinforcement learning (RL) framework to optimize the switching (self-healing) strategy, reducing false alarm rates and improving system robustness. This ensures uninterrupted monitoring and supports historical data backtracking, overcoming the problem of monitoring interruptions that occur in existing systems when sensors fail.

[0032] 5. The present invention realizes efficient interaction between physical sensors and digital twin models through the OPC UA server and the Profinet / Wi-Fi 6 dual-channel communication protocol, and supports RESTful API interface to access third-party monitoring platforms. Data synchronization integrates the time series database (InfluxDB) and the message queue (RabbitMQ), and distributes instructions through the message queue to achieve seamless integration and low-latency transmission of multi-source data (real-time, historical, and predicted), overcoming the problem that traditional technologies rely on manual configuration of communication protocols and have low interaction efficiency.

[0033] 6. The present invention is applicable to the digital monitoring of longwall mining in coal mines. It synchronizes the digital twin model by monitoring the movement speed and position of the coal mining machine, and defines the status of the deployed virtual sensors through the movement of physical sensors in real scenes. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a flow chart of the dynamic activation management method of working surface sensors based on digital twins provided by the present invention.

[0035] Figure 2 It is a schematic diagram of the virtual sensor sequence layout provided by the present invention and the relationship between activation / deactivation state switching and working surface advancement position. DETAILED DESCRIPTION

[0036] In order to make the technical means, creative features, objectives and effects achieved by the present invention easier to understand, the present invention is further described below with reference to specific illustrations.

[0037] Please refer to Figure 1 and Figure 2 As shown, the present invention provides a method for dynamic activation management of working surface sensors based on digital twins, comprising the following steps:

[0038] S1. Three-dimensional geological modeling software is used in the digital twin system to generate high-precision digital twin models of the coal mine working face transport level tunnel and return air level tunnel based on the actual geological data of the coal mine. The model data source includes tunnel contours, rock hardness distribution and geological anomaly area (such as faults and fracture zones) data; a virtual sensor array is preset along the working face advancement direction to support the simulation of multiple types of sensors such as pressure sensors, gas concentration sensors and displacement sensors. The virtual sensor array covers the entire length of the working face. The coordinate position of each virtual sensor is bound to the actual sensor of the physical working face through a unique identifier (ID) to ensure the accuracy and traceability of data mapping, and receive monitoring data (such as pressure, displacement, and gas concentration) from the physical sensor in real time.

[0039] S2. Define the initial activation state of the virtual sensor array in the digital twin model. The state includes: virtual sensor number and grouping information, divided by level tunnel segment (such as transport level tunnel sensor group A1.A2.A3... and return air level tunnel sensor group B1.B2.B3...). During initialization, only a predetermined number of sensors before the head end (such as the first 20 sensors A1~A20, B1~B20) are in the activated state, and the rest are marked as dead; establish a mapping relationship between the virtual sensor coordinate position and the longitude and latitude of the physical sensor, specifically, the virtual sensor can be mapped to the physical sensor through the Gaussian projection algorithm. The coordinates of the virtual sensor are mapped to the actual longitude and latitude of the physical sensor to ensure that the spatial error tolerance is less than 0.5 meters. Data synchronization mapping rules are defined, including normalization of the monitoring data of the physical sensor, timestamp alignment and conflict correction mechanism. For example, normalization can convert pressure values 0-10MPa to 0-1. Timestamp alignment calibrates the timestamp through the NTP protocol to eliminate the time difference between systems (<1ms). The conflict correction mechanism gives priority to physical sensor data in the event of data conflict and records automatic correction logs in the MySQL database for subsequent analysis and auditing.

[0040] S3. The shearer's cutting position and propulsion speed are acquired in real time through the physical working face's propulsion system. The working face's propulsion distance is calculated using inertial navigation unit (IMU) data. This propulsion distance is synchronized to the digital twin model via the Profinet protocol to dynamically update the distance between the virtual sensor and the propulsion end. Specifically, the distance between the propulsion end and the nearest currently active virtual sensor is dynamically calculated. When the distance between the propulsion end and the nearest currently active virtual sensor reaches a preset threshold, the virtual sensor (life or death sensor) state switching update rule is triggered. For example, when the propulsion end coordinates are (x = 115m, y = 50m, z = -5m), the nearest activated sensor is transport level sensor A5 (x = 100m), and the real-time distance is 15 meters. The system continuously monitors this distance and compares it with a preset threshold (e.g., 20 meters) to provide a basis for state switching.

[0041] S4. According to the state switching update rule, the following operations are performed in the digital twin model: deactivate the virtual sensor in the nearest active state (such as the transport level sensor A5), freeze the data input and output channels, mark it as a black icon in the visual interface, and record the deactivation time and reason to the database; at the same time, activate the virtual sensor in the back end (the farthest end) that is in the dead state (such as the transport level sensor A25), initialize the data mapping relationship of the activated virtual sensor, and start the simulation data generation module. The simulation data generation module trains a generative adversarial network (GAN) based on the historical data of the physical sensor to predict missing values (prediction mean square error <10%), marks it as a red flashing icon in the visual interface, and sends a status update notification to the third-party monitoring platform through the API interface of the digital twin system;

[0042] S5. Loop through steps S3 and S4 until the working face advances to the stop-mining line, whereupon the loop is stopped. This allows for dynamic adjustment of the virtual sensor's activation range, and calibration of the model synchronization error through the Kalman filter to ensure that the real-time mapping error is less than 5%. The error calculation is based on the difference in monitoring data between the virtual sensor and the physical sensor, as well as the advancement position deviation.

[0043] As a specific embodiment, the distribution of the virtual sensor array in step S1 is equidistant (the default spacing is 20 meters) or unequal spacing, and the unequal spacing is dynamically adjusted based on the geological anomaly detection results: after the fault or fracture zone is identified by the microseismic monitoring system, the virtual sensor spacing in the high-risk area (such as the fault zone) is automatically reduced to 10 to 15 meters and the monitoring frequency is increased to 1 time per second to improve the monitoring density; the virtual sensor spacing in the normal rock formation area is expanded to 20 to 30 meters and the monitoring frequency is adjusted to 1 time every 5 seconds, meeting the deployment density while complying with the monitoring requirements.

[0044] As a specific embodiment, step S2 also includes real-time analysis of physical sensor data through the isolation forest algorithm. If an abnormality in the physical sensor data (such as a sudden change in gas concentration, etc.) is detected and continues for more than 30 seconds without recovery, it switches to the virtual sensor data stream and sends a maintenance alarm to the monitoring platform (red warning on the interface, SMS notification and voice prompt).

[0045] As a specific embodiment, the preset threshold in step S3 is a dynamic distance value between 5 meters and 20 meters. The threshold optimization process includes calculating the single advancement step length based on the coal mining machine cutting depth (0.8 to 1.5 meters), statistically analyzing the standard deviation of historical advancement data through a sliding window algorithm, constructing a topological network based on the virtual sensor distribution density (one node every 10 to 30 meters), using a graph convolutional neural network (GCN) to predict the optimal trigger distance, and dynamically adjusting the threshold in combination with geological parameters (such as roof pressure and gas concentration gradient). For example, when the roof pressure exceeds 25 MPa, the threshold is reduced to 5 to 10 meters.

[0046] As a specific embodiment, the state switching update operation in step S4 is implemented through the state machine module of the digital twin system. The state machine module includes two states: sleep and death. The state migration is triggered by the threshold of the working surface distance to the nearest sensor; the death state executes to close the data acquisition thread, release memory resources and is marked as a black circular icon in the visual interface; the activation state starts the data simulation thread and binds the physical sensor data stream, which is marked as a red flashing icon in the interface and displays the real-time monitoring value.

[0047] As a specific embodiment, the efficient interaction between the physical sensor data and the digital twin model in step S1 is achieved through the OPC UA server and the Profinet / Wi-Fi 6 dual-channel communication protocol. In step S4, the RESTful API interface is supported to access the third-party monitoring platform, thereby achieving efficient interaction between the physical sensor data, the digital twin model and the third-party monitoring platform, and supporting external systems to call real-time status, alarm information and risk assessment reports. Specifically, the physical sensor data is connected to the system in real time through the OPC UA protocol, and the environmental parameters (temperature, humidity) are transmitted via Wi-Fi 6 to ensure that the delay is less than 50ms; the interaction between the digital twin model and the third-party monitoring platform is achieved through the RESTful API interface, supporting external systems to call real-time status (such as GET / api / sensor_status to obtain the sensor activation list) and subscribe to alarm information; the risk assessment report is pushed to the intelligent mine management platform in JSON format, including safety index, deviation value and recommended measures, to facilitate efficient operation and maintenance decision-making.

[0048] As a specific embodiment, the working face safety risk assessment report is generated by combining real-time data, historical data and predicted trajectory, performing multi-scale data fusion and visualizing the output in the form of a three-dimensional heat map. Among them, the real-time data uses a sliding window weighted average algorithm (window size 10 seconds) to eliminate instantaneous noise, the historical data is analyzed through the ARIMA time series model to analyze the advancement law, and the predicted trajectory is based on the Kalman filter to dynamically correct the error. The working face safety index (0-100 points) is divided into "safe" (≥80 points), "warning" (60-79 points), and "dangerous" (<60 points), and high-risk areas are marked on the three-dimensional heat map (red flashes when the deviation is >10%). The predicted trajectory is superimposed on the interface with a translucent green dotted line and compared with the actual advancement end position in real time to provide decision support for coal mining machine speed regulation and roof support.

[0049] The present invention also provides a working face sensor dynamic activation management system based on digital twin, comprising:

[0050] The digital twin model module is used to use 3D geological modeling software in the digital twin system to generate high-precision digital twin models of the coal mine working face transport and return air lanes based on the actual geological data of the coal mine. The model data source includes the lane profile, rock hardness distribution, and geological anomaly area data. A virtual sensor array is preset along the working face advancement direction. The virtual sensor array covers the entire length of the working face. The coordinate position of each virtual sensor is bound to the actual sensor of the physical working face through a unique identifier, and the module receives monitoring data from the physical sensors in real time.

[0051] The sensor state definition module is used to define the initial activation state of the virtual sensor array in the digital twin model. The state includes: virtual sensor number and grouping information, divided by level and tunnel segments, only a predetermined number of sensors before the head end are in the active state, and the rest are marked as dead; establish a mapping relationship between the virtual sensor coordinate position and the longitude and latitude of the physical sensor; define data synchronization mapping rules, including normalization processing of the monitoring data of the physical sensor, time stamp alignment and conflict correction mechanism;

[0052] The sensor status monitoring module is used to obtain the shearer cutting position and propulsion speed in real time through the propulsion system of the physical working face, calculate the working face propulsion distance in combination with the inertial navigation unit data, and synchronize the propulsion distance to the digital twin model to dynamically update the distance between the virtual sensor and the propulsion end. When the distance between the propulsion end and the nearest currently activated virtual sensor reaches a preset threshold, the virtual sensor state switching update rule is triggered.

[0053] A sensor status update module is configured to perform the following operations in the digital twin model according to the state switching update rule: deactivate the virtual sensor in the nearest active state, freeze the data input and output channels, mark it as a black icon in the visual interface, and record the deactivation time and reason in the database; at the same time, activate the virtual sensor in the back-end dead state, initialize the data mapping relationship of the activated virtual sensor, and start the simulation data generation module. The simulation data generation module trains a generative adversarial network based on the historical data of the physical sensor to predict missing values, marks it as a red flashing icon in the visual interface, and sends a status update notification to the third-party monitoring platform through the API interface of the digital twin system;

[0054] The state machine module is used to cyclically execute the sensor status monitoring module and the sensor status update module until the advance length along the working face reaches the stop-mining line. In this way, the activation range of the virtual sensor is dynamically adjusted, and the model synchronization error is calibrated through the Kalman filter to ensure that the real-time mapping error is less than 5%.

[0055] Compared with the existing technology, the working face sensor dynamic activation management method and system based on digital twin provided by the present invention has the following advantages:

[0056] 1. The present invention uses a "life and death sensor" mechanism to preset a virtual sensor array and dynamically trigger activation / deactivation rules (such as advancing distance threshold triggering), automatically synchronizing the sensor coverage of the digital twin model and the physical scene, eliminating manual adjustment delays and achieving millisecond-level synchronization with an error of less than 5%. This overcomes the problem of traditional technology relying on manual movement of physical sensors, resulting in monitoring blind spots and model update lags.

[0057] 2. The present invention dynamically adjusts the virtual sensor spacing (reduced to 10-15 meters in high-risk areas) and monitoring frequency (increased to once per second) based on microseismic monitoring data, and generates missing data through GAN (generative adversarial network), solving the problem that physical sensors cannot be deployed in high density due to explosion-proof and power consumption limitations. It overcomes the difficulty of existing fixed-spacing sensor arrangements in responding to the intensive monitoring needs of high-risk areas such as faults and fracture zones.

[0058] 3. The present invention integrates Kalman filtering (real-time mapping error calibration), sliding window weighted averaging (eliminating instantaneous noise) and graph convolutional neural network (GCN dynamically optimizes the trigger threshold), combined with the coal mining machine cutting depth (0.8 to 1.5 meters) and geological parameters (reducing the threshold when the roof pressure is >25MPa), significantly improving the accuracy and stability of the digital twin model, overcoming the problems of traditional models lacking real-time data calibration capabilities and serious synchronization error accumulation.

[0059] 4. This invention uses an isolation forest algorithm to analyze physical sensor data in real time. If a data anomaly is detected for more than 30 seconds, it automatically switches to the virtual sensor data stream and triggers a multi-level alarm (interface warning, voice prompt, model freeze). Simultaneously, it uses a reinforcement learning (RL) framework to optimize the switching (self-healing) strategy, reducing false alarm rates and improving system robustness. This ensures uninterrupted monitoring and supports historical data backtracking, overcoming the problem of monitoring interruptions that occur in existing systems when sensors fail.

[0060] 5. The present invention realizes efficient interaction between physical sensors and digital twin models through the OPC UA server and the Profinet / Wi-Fi 6 dual-channel communication protocol, and supports RESTful API interface to access third-party monitoring platforms. Data synchronization integrates the time series database (InfluxDB) and the message queue (RabbitMQ), and distributes instructions through the message queue to achieve seamless integration and low-latency transmission of multi-source data (real-time, historical, and predicted), overcoming the problem that traditional technologies rely on manual configuration of communication protocols and have low interaction efficiency.

[0061] 6. The present invention is applicable to the digital monitoring of longwall mining in coal mines. It synchronizes the digital twin model by monitoring the movement speed and position of the coal mining machine, and defines the status of the deployed virtual sensors through the movement of physical sensors in real scenes.

[0062] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A dynamic activation management method for working face sensors based on digital twins, characterized in that: The following steps are involved: S1. Using 3D geological modeling software in the digital twin system, a high-precision digital twin model of the coal mine working face transport and return air lanes is generated based on the actual geological data of the coal mine. The model data sources include lane profiles, rock hardness distribution, and geological anomaly area data. A virtual sensor array is preset along the advancing direction of the working face, and the virtual sensor array covers the entire length of the working face. The coordinate position of each virtual sensor is bound to the actual sensor of the physical working face through a unique identifier, and the monitoring data of the physical sensor is received in real time; S2. Define the initial activation state of the virtual sensor array in the digital twin model. The state includes: virtual sensor number and grouping information, divided by level and tunnel segments, with only a predetermined number of sensors before the head end being in an active state, and the rest marked as dead; establish a mapping relationship between the virtual sensor coordinates and the longitude and latitude of the physical sensors; and define data synchronization mapping rules, including normalization, timestamp alignment, and conflict correction mechanisms for the monitoring data of the physical sensors. S3. The shearer's cutting position and propulsion speed are acquired in real time through the propulsion system of the physical working face. The working face propulsion distance is calculated in combination with the inertial navigation unit data. The propulsion distance is synchronized with the digital twin model to dynamically update the distance between the virtual sensor and the propulsion end. When the distance between the propulsion end and the nearest currently activated virtual sensor reaches a preset threshold, the virtual sensor state switching update rule is triggered. S4. According to the state switching update rule, the following operations are performed in the digital twin model: deactivate the virtual sensor in the nearest active state, freeze the data input and output channels, mark it as a black icon in the visual interface, and record the deactivation time and reason in the database; at the same time, activate the virtual sensor in the dead state at the back end, initialize the data mapping relationship of the activated virtual sensor, and start the simulation data generation module. The simulation data generation module trains a generative adversarial network based on the historical data of the physical sensor to predict missing values, marks it as a red flashing icon in the visual interface, and sends a status update notification to the third-party monitoring platform through the API interface of the digital twin system; S5. Loop through steps S3 and S4 until the working face advances to the stop-mining line, whereupon the loop is stopped. This allows for dynamic adjustment of the virtual sensor's activation range, and calibration of the model synchronization error through the Kalman filter to ensure that the real-time mapping error is less than 5%.

2. The method for dynamic activation management of working face sensors based on digital twins according to claim 1 is characterized in that: In step S1, the virtual sensor array is distributed in an equidistant or unequal manner, and the unequal spacing is dynamically adjusted based on the geological anomaly detection result: after the fault or fracture zone is identified by the microseismic monitoring system, the virtual sensor spacing in the high-risk area is automatically reduced to 10-15 meters and the monitoring frequency is increased to 1 time per second; the virtual sensor spacing in the normal rock formation area is expanded to 20-30 meters and the monitoring frequency is adjusted to 1 time every 5 seconds.

3. The method for dynamic activation management of working face sensors based on digital twins according to claim 1 is characterized in that: The step S2 also includes real-time analysis of the data of the physical sensor using the isolation forest algorithm. If the physical sensor data is detected to be abnormal and does not recover for more than 30 seconds, the data stream is switched to the virtual sensor data stream and a maintenance alert is sent to the monitoring platform.

4. The method for dynamic activation management of working face sensors based on digital twins according to claim 1, characterized in that: The preset threshold in step S3 is a dynamic distance value between 5 meters and 20 meters. The threshold optimization process includes calculating the single advancement step length based on the coal mining machine cutting depth, statistically analyzing the standard deviation of historical advancement data through a sliding window algorithm, constructing a topological network based on the distribution density of virtual sensors, using a graph convolutional neural network to predict the optimal trigger distance, and dynamically adjusting the threshold in combination with geological parameters.

5. The method for dynamic activation management of working face sensors based on digital twins according to claim 1, characterized in that: The state switching and updating operation in step S4 is implemented by the state machine module of the digital twin system. The state machine module includes two states: sleep and death. The state transition is triggered by the threshold of the working surface distance to the nearest sensor; The death state closes the data acquisition thread, releases memory resources, and is marked as a black circular icon in the visual interface. The activation state starts the data simulation thread and binds the physical sensor data stream. It is marked as a red flashing icon in the interface and displays real-time monitoring values.

6. The method for dynamic activation management of working face sensors based on digital twins according to claim 1, characterized in that: In step S1, the efficient interaction between the physical sensor data and the digital twin model is achieved through the OPC UA server and the Profinet / Wi-Fi 6 dual-channel communication protocol, and in step S4, the RESTful API interface is supported to access the third-party monitoring platform.

7. The working face sensor dynamic activation management system based on digital twin is characterized by: include: The digital twin model module uses 3D geological modeling software within the digital twin system to generate high-precision digital twin models of the coal mine's working face haulage and return air lanes based on the mine's actual geological data. The model data sources include lane profiles, rock hardness distribution, and geological anomaly area data. A virtual sensor array is preset along the advancing direction of the working face, and the virtual sensor array covers the entire length of the working face. The coordinate position of each virtual sensor is bound to the actual sensor of the physical working face through a unique identifier, and the monitoring data of the physical sensor is received in real time; The sensor state definition module is used to define the initial activation state of the virtual sensor array in the digital twin model. The state includes: virtual sensor number and grouping information, divided by level and tunnel segments, only a predetermined number of sensors before the head end are in the active state, and the rest are marked as dead; establish a mapping relationship between the virtual sensor coordinate position and the longitude and latitude of the physical sensor; define data synchronization mapping rules, including normalization processing of the monitoring data of the physical sensor, time stamp alignment and conflict correction mechanism; The sensor status monitoring module is used to obtain the shearer cutting position and propulsion speed in real time through the propulsion system of the physical working face, calculate the working face propulsion distance in combination with the inertial navigation unit data, and synchronize the propulsion distance to the digital twin model to dynamically update the distance between the virtual sensor and the propulsion end. When the distance between the propulsion end and the nearest currently activated virtual sensor reaches a preset threshold, the virtual sensor state switching update rule is triggered. A sensor status update module is configured to perform the following operations in the digital twin model according to the state switching update rule: deactivate the virtual sensor in the nearest active state, freeze the data input and output channels, mark it as a black icon in the visual interface, and record the deactivation time and reason in the database; at the same time, activate the virtual sensor in the back-end dead state, initialize the data mapping relationship of the activated virtual sensor, and start the simulation data generation module. The simulation data generation module trains a generative adversarial network based on the historical data of the physical sensor to predict missing values, marks it as a red flashing icon in the visual interface, and sends a status update notification to the third-party monitoring platform through the API interface of the digital twin system; The state machine module is used to cyclically execute the sensor status monitoring module and the sensor status update module until the advance length along the working face reaches the stop-mining line. In this way, the activation range of the virtual sensor is dynamically adjusted, and the model synchronization error is calibrated through the Kalman filter to ensure that the real-time mapping error is less than 5%.

Images