Deep coal mine upward mining overhead gob-side entry driving support and construction method

Abstract

The invention discloses a deep coal mine upward mining air-stepping gob-side entry driving support and construction method, and relates to the technical field of deep coal mine mining roadway support, the method comprises the following components: monitoring network deployment, data processing and prediction, intelligent early warning triggering, coordinated regulation execution and loop optimization; according to the invention, a three-dimensional and key node reinforced distributed monitoring network is deployed, and a fiber grating stress sensor, a micro-seismic monitoring probe and an infrared thermal imager are combined, so that real-time monitoring of multiple parameters such as surrounding rock stress, displacement, temperature and the like is realized, and through a 5G, wireless Mesh and wired backup redundancy transmission scheme, the real-time monitoring of the surrounding rock stress, displacement, temperature and the like is realized. According to the method, data are acquired, it is ensured that the data are uploaded to a cloud data center in real time, a high-precision surrounding rock deformation prediction model is established by means of a big data analysis algorithm and a geological parameter coupling-attention mechanism LSTM algorithm, the model can accurately pre-judge the abnormal deformation trend of surrounding rock 1-3 hours in advance, and through a double-threshold dynamic adjustment and risk superposition assessment mechanism, the abnormal deformation trend of the surrounding rock can be accurately predicted. And automatically triggering graded early warning.

Description

Technical Field

[0001] This invention relates to the field of deep coal mine roadway support technology, specifically a method for supporting and constructing roadways for upward mining in deep coal mines. Background Technology

[0002] With the increasing depth of coal mining, deep coal mining has become an inevitable trend in the industry. During the upward mining process in deep coal mines, the control of the surrounding rock in the roadway faces enormous challenges due to the increased ground stress and complex and variable surrounding rock conditions, especially when excavating roadways along the goaf.

[0003] Traditional support and construction methods for goaf excavation in deep coal mines have the following drawbacks: First, monitoring methods are outdated, relying heavily on manual periodic inspections, making it difficult to monitor multiple parameters such as surrounding rock stress, displacement, and temperature in real time, resulting in untimely and inaccurate data acquisition. Second, the early warning mechanism is imperfect, relying mainly on simple judgments based on fixed thresholds, unable to dynamically adjust according to roadway geological conditions and construction stages, easily leading to false alarms or missed alarms. Third, control measures are passive, mostly taking remedial measures after roadway deformation, rather than proactively controlling in advance based on predictive models, resulting in poor support effects. Finally, there is a lack of cyclic optimization mechanisms, making it impossible to iteratively optimize predictive model parameters and support control thresholds based on post-control monitoring data, making it difficult to adapt to dynamic changes in the surrounding rock.

[0004] In view of the shortcomings of traditional technologies, such as outdated monitoring methods, imperfect early warning mechanisms, passive control measures, and lack of cyclical optimization mechanisms, this invention proposes a method for support and construction of goaf-side excavation roadways in deep coal mines, which is of particular importance. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a method for support and construction of deep coal mine uphill mining roadways. This method utilizes a distributed monitoring network with enhanced key nodes in a three-dimensional configuration, combined with fiber optic stress sensors, microseismic monitoring probes, and infrared thermal imagers, to achieve real-time monitoring of multiple parameters such as surrounding rock stress, displacement, and temperature. It employs big data analysis algorithms and the geological parameter coupling-attention mechanism LSTM algorithm to establish a high-precision surrounding rock deformation prediction model. Through a dual-threshold dynamic adjustment and risk superposition assessment mechanism, it automatically triggers graded early warnings. After the early warning signal is triggered, it links the hydraulic support control system and the deep grouting system to achieve coordinated control and execution. Simultaneously, it employs a dynamic data window-multi-source verification-segmented retraining cyclic optimization technique to continuously collect monitoring data after control and iteratively optimize the prediction model parameters and support control thresholds, forming a dynamic closed loop.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for support and construction of goaf-side excavation roadways in deep coal mines, the specific steps of which are as follows:

[0007] Monitoring network deployment: A distributed layout scheme with three-dimensional positioning and key node reinforcement is adopted. Fiber optic stress sensors, microseismic monitoring probes, and infrared thermal imagers are deployed on the key sections of the roof, sides, and floor of the tunnel along the goaf to construct a real-time monitoring network for multiple parameters of surrounding rock stress, displacement, and temperature. The sensor data acquisition frequency is set to 1-5Hz. Data transmission adopts a redundant transmission scheme of 5G + wireless Mesh + wired backup to ensure real-time data upload.

[0008] Data processing and prediction: The monitoring data is transmitted to the cloud data center in real time. The multi-dimensional data is fused and processed through big data analysis algorithms. A surrounding rock deformation prediction model is established based on the geological parameter coupling-attention mechanism LSTM algorithm. Historical monitoring data and geological parameters are input to complete the model training. Dual threshold standards for deformation rate and stress peak are set.

[0009] Intelligent early warning trigger: The system adopts a dual-threshold dynamic adjustment + risk superposition assessment mechanism. When the monitoring data exceeds the preset threshold, the system automatically triggers a graded early warning and pushes the early warning information synchronously through the audible and visual alarm and the downhole communication terminal, clearly identifying the location of the abnormal area and the risk level.

[0010] Coordinated control and execution: After the early warning signal is triggered, the hydraulic support control system and the deep grouting system are linked; the hydraulic support adopts a dual closed-loop control strategy of predicted deformation and real-time stress feedback, and adaptively adjusts the support resistance according to the output results of the prediction model, with an adjustment range of 10-30MPa; the deep grouting system adopts the technology of crack targeted positioning and precise matching of grouting parameters, and controls the grouting pressure at 2-5MPa and the grouting volume at 5-10m³ / segment based on the geological data of the abnormal area to complete the targeted reinforcement;

[0011] Cyclic optimization: The dynamic data window-multi-source verification-segmented retraining cyclic optimization technology is adopted to continuously collect monitoring data after regulation and control, and to iteratively optimize the prediction model parameters and support regulation thresholds in reverse to achieve a dynamic closed loop of monitoring-early warning-regulation; at the same time, an adaptation adjustment mechanism is preset for special geological conditions such as water-rich and rockburst.

[0012] Furthermore, the three-dimensional + key node reinforced distributed layout scheme in the monitoring network deployment is as follows: the spacing of the fiber optic grating sensors on the roof along the roadway axis is set to 1.5-2m, and one sensor is added on each side of the roadway centerline at a distance of 0.8-1m, forming a triangular monitoring array; the sidewall sensors are arranged in three layers: the upper layer is 0.5m from the roof, the lower layer is 0.8m from the floor, and the middle layer is located in the middle of the roadway side, with an interlayer spacing of ≤3m, and three sensors are arranged 5-8m ahead of the tunnel face. Force sensors are used to detect stress concentration signals in front of the working face in advance; bottom plate sensors are arranged at the center of the roadway and 1.5m on both sides, and the contact surface between the sensor and the surrounding rock adopts an arc-shaped fit design. The contact surface is coated with high thermal conductivity and electrical conductivity grease to reduce data transmission loss; all sensors are encapsulated with a double-layer stainless steel anti-smashing and wear-resistant shell, with a buffer rubber pad in the inner layer, and the protection level is ≥IP68. The sensor cables adopt an armored flame-retardant design with a tensile strength ≥50MPa, which is suitable for underground high humidity, high dust, impact and electromagnetic interference environments.

[0013] Furthermore, the 5G+wireless Mesh+wired backup redundant transmission scheme in the monitoring network deployment is as follows: Downhole sensor data is first aggregated to an edge computing node in the middle of the roadway via a wireless Mesh network, with an aggregation period of ≤1s. The edge computing node preprocesses the data, and the preprocessed data is transmitted to the cloud data center via 5G, with a transmission rate of ≥100Mbps and a latency of ≤20ms. Simultaneously, a wired optical fiber is deployed as a backup transmission channel. When the 5G signal is interrupted, it automatically switches to wired transmission, with a switching time of ≤1s. The cloud data center adopts a distributed storage architecture with a data storage capacity of ≥10TB and supports data encryption and access control. It uses the AES-256 encryption algorithm to encrypt the transmission of monitoring data and control commands to prevent data leakage and malicious tampering. In addition, the edge computing node has a built-in backup power supply that can provide continuous power for ≥4 hours, ensuring that data is not lost in the event of a sudden power outage.

[0014] Furthermore, the surrounding rock deformation prediction model constructed by the geological parameter coupling-attention mechanism LSTM algorithm in the data processing and prediction is as follows: ,in For the future Predicted deformation of the surrounding rock. , , These are the LSTM time-series feature weights, geological parameter coupling weights, and stress-strain synergistic weights, respectively. The initial values ​​were determined through multiple linear regression analysis of more than 100 sets of historical data from similar projects. For improved LSTM network output, This is the multi-parameter monitoring data matrix at time t. This is a coupling function for geological parameters. A vector of geological parameters. is the stress-strain collaborative term, is the surrounding rock strain value at time is the error correction term, with a value range of ±0.1 mm, determined by the sensor accuracy calibration data; this algorithm predicts with an accuracy error ≤ 3% by fusing time-series features and geological parameters, and can accurately predict the abnormal deformation trend of the surrounding rock 1 - 3 h in advance.

[0015] Furthermore, the double-threshold dynamic adjustment + risk superposition assessment warning mechanism in the intelligent warning trigger is specifically as follows: both the deformation rate threshold and the stress peak threshold are dynamically corrected based on the roadway geological conditions and construction stages. The correction formula is: , where is the dynamic threshold at time is the initial threshold. The initial deformation rate threshold is 3 - 5 mm / d, and the initial stress peak threshold is 20 - 25 MPa. is the cumulative driving depth of the roadway at time is the total designed depth of the roadway, is the driving speed at time is the maximum allowable driving speed, , are correction factors, iteratively optimized through construction process data, with a value range of 0.1 - 0.3; the warning levels are divided into three levels. A first-level warning triggers a red audible and visual alarm, synchronously cuts off the power supply of the driving equipment, and pushes it定向 to the construction team and the dispatching center through the underground communication terminal; a second-level warning triggers a yellow alarm, pushes the warning information and prompts to increase the monitoring frequency; a third-level warning triggers a blue warning, only pushing a prompt message; at the same time, the warning information includes the three-dimensional coordinates of the abnormal area, the risk level, the predicted deformation increment, and the recommended disposal measures, and supports linkage display with the underground GIS system.

[0016] Furthermore, the specific double-closed-loop control strategy of predicting the deformation amount of the hydraulic support - real-time stress feedback in the collaborative regulation execution is as follows: the support resistance adjustment is achieved through an electro-hydraulic proportional valve group, and the formula is: , where is the target support resistance, is the current support resistance, is the maximum adjustable resistance increment, is the predicted deformation amount, is the allowable maximum deformation amount; the telescopic amount of the hydraulic support is dynamically adjusted according to the data of the fiber Bragg grating displacement sensor, and the adjustment formula is: , where is the target telescopic amount, is the current telescopic amount, The displacement compensation coefficient is set at 1.1-1.3 to ensure a tight fit between the support and the surrounding rock. The support resistance adjustment response time is ≤5s, the maximum extension stroke of the support is ≥500mm, and the support is equipped with a pressure overload protection device. When the actual support resistance exceeds the target value by 10%, the pressure is automatically released to a safe range. At the same time, an angle sensor is installed at the connection between the support top beam and the shield beam to monitor the support posture in real time. When the posture deviation exceeds 5°, the extension and retraction of the support hydraulic cylinder is adjusted synchronously to prevent the support from tilting and causing support failure.

[0017] Furthermore, the collaborative control and execution of the deep grouting system's fracture-targeted positioning and precise matching technology for grouting parameters specifically involves: the grouting hole arrangement angle is determined based on the fracture orientation detected by ground-penetrating radar, with an angle ≤30° to the fracture orientation; the grouting hole depth is 2-5m, and the hole diameter is 42-50mm; adjacent grouting holes are staggered with a spacing of 1.5-2.5m, forming a three-dimensional grouting network; the grouting volume is precisely calculated using the following formula: ,in This refers to the grouting volume per section. The grouting filling coefficient is set to 1.2-1.5, depending on the degree of fissure development in the surrounding rock; 1.5 is used in areas with dense fissures, and 1.2 is used in areas with sparse fissures. The volume affected by grouting is determined by the depth, spacing, and diffusion radius of the grouting holes. The porosity of the surrounding rock was determined through borehole sampling tests. The shrinkage compensation coefficient for the grout is set at 0.1-0.2. The grout uses nano-silane modified cement-based material with a mix ratio of cement:sand:nanosilane:water = 1:2.5:0.03:0.45. The setting time is controlled at 15-30 min, the 28-day compressive strength is ≥40 MPa, and the bond strength is ≥3.5 MPa. The grouting process adopts a pulse pressurization-stabilization and grout retention process. The pulse pressure fluctuation range is ±0.5 MPa of the set pressure, and the grout retention time is ≥30 min to ensure that the grout fully penetrates into the micro-cracks and solidifies.

[0018] Furthermore, the aforementioned iterative optimization employs a dynamic data window-multi-source verification-segmented retraining iterative optimization technique. The specific implementation steps are as follows: First, data windows are dynamically divided based on the tunnel excavation progress and the surrounding rock deformation rate. When the deformation rate is ≤2mm / d, a fixed 72-hour window is used; when the deformation rate is >2mm / d, it automatically switches to a 24-hour sliding window. The data within the window includes real-time monitored stress, displacement, temperature values, and corresponding geological parameters. Second, multi-source verification is performed on the data within the window. By comparing the correlation between fiber optic grating sensor data and microseismic monitoring energy values, outliers are eliminated, and combined with underground geological data... The model input is corrected using high-quality sketch data. Next, the model is retrained in segments according to the tunneling stage, with each segment containing at least 300 training samples. A recent data weighting strategy is used: data from the last 7 days has a weight of 0.6, data from 7-30 days has a weight of 0.3, and data older than 30 days has a weight of 0.1. Key influencing parameters are selected using a random forest algorithm. Finally, after training, a dual-index evaluation is performed. When the model prediction error is ≤3% and the prediction trend is consistent for 5 consecutive times, the model parameters are updated and the support control threshold is adjusted simultaneously. If the target is not met, the above steps are repeated until the requirements are met, ensuring that the model always accurately matches the dynamic changes of the surrounding rock.

[0019] Furthermore, the special geological condition adaptation and adjustment mechanism in the cyclic optimization is as follows: When a water-rich area is detected in the hollow area, the support and construction parameters are automatically adjusted: the data acquisition frequency of the fiber optic grating sensor is increased to 3-5Hz, and a humidity sensor is added for synchronous monitoring; a water content correction term is added to the surrounding rock deformation prediction model to adjust the relevant weight coefficients; the lower limit of the hydraulic support resistance is increased to 15MPa, and a hydrophobic protective coating is used to treat the support surface; the deep grouting system is changed to hydrophilic epoxy resin grout, the grouting pressure is increased to 3-5MPa, the grouting volume is increased by 20%-30%, and a polyurethane water-stopping agent is injected before grouting to form a water-proof curtain; when encountering a precursor signal of rockburst, the emergency control mode is triggered, the hydraulic support resistance is instantly increased to the maximum threshold of 30MPa, and the buffer energy dissipation devices on both sides of the roadway are activated simultaneously. The impact energy is absorbed through the spring damping structure, and the grouting system adopts a high-pressure jet grouting process to quickly form a rigid bearing layer to resist the impact load.

[0020] Compared with existing technologies, this method for support and construction of goaf-side excavation roadways in deep coal mines has the following advantages:

[0021] I. This invention utilizes a distributed monitoring network with enhanced three-dimensional structure and key nodes, combined with fiber optic stress sensors, microseismic monitoring probes, and infrared thermal imagers, to achieve real-time monitoring of multiple parameters such as surrounding rock stress, displacement, and temperature. A redundant transmission scheme of 5G, wireless mesh, and wired backup ensures real-time data upload to the cloud data center. Furthermore, a high-precision surrounding rock deformation prediction model is established using big data analysis algorithms and the geological parameter coupling-attention mechanism LSTM algorithm. This model can accurately predict abnormal deformation trends of the surrounding rock 1-3 hours in advance. Through a dual-threshold dynamic adjustment and risk superposition assessment mechanism, it automatically triggers graded early warnings, significantly improving operational safety during the upward mining process in deep coal mines.

[0022] Second, after the early warning signal is triggered, this invention can link the hydraulic support control system and the deep grouting system to achieve coordinated control and execution. The hydraulic support adopts a dual closed-loop control strategy of predicted deformation and real-time stress feedback, and adaptively adjusts the support resistance according to the output results of the prediction model to ensure the optimal support effect. At the same time, the deep grouting system uses crack targeted positioning and precise matching technology of grouting parameters to accurately control the grouting pressure and grouting volume based on the geological data of abnormal areas to complete targeted reinforcement. In addition, through dynamic data window, multi-source verification and segmented retraining cyclic optimization technology, the monitoring data after control is continuously collected, and the prediction model parameters and support control threshold are iteratively optimized in reverse to form a dynamic closed loop, which significantly improves construction efficiency and quality.

[0023] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0025] Figure 1 This is a flowchart illustrating the operation of a method for supporting and constructing a goaf-side excavation tunnel in deep coal mines during upward mining.

[0026] Figure 2 This is a schematic diagram of a method for supporting and constructing a goaf-side excavation tunnel in deep coal mines during upward mining. Detailed Implementation

[0027] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0028] Example 1

[0029] A distributed layout scheme with three-dimensional positioning and key node reinforcement was adopted. Fiber optic stress sensors, microseismic monitoring probes, and infrared thermal imagers were deployed on key sections of the roof, sidewalls, and floor of the tunnel excavation along the goaf. Additionally, humidity sensors were added to construct a real-time monitoring network for multiple parameters of surrounding rock stress, displacement, temperature, and humidity. The roof fiber optic sensors were spaced 1.5m apart along the tunnel axis, with one additional sensor at 0.8m on each side of the tunnel centerline, forming a triangular monitoring array. The sidewall sensors were arranged in three layers: the upper layer 0.5m from the roof, the lower layer 0.8m from the floor, and the middle layer located in the middle of the sidewall, with a layer spacing of 2.5m. Three stress sensors were placed 5m ahead of the tunnel face to detect stress concentration ahead of the tunnel face in advance. Signal; the bottom plate sensors are arranged at the center of the roadway and 1.5m on both sides. The contact surface between the sensor and the surrounding rock adopts an arc-shaped fit design, and the fit surface is coated with a high thermal conductivity and electrical conductivity paste. Data transmission adopts a redundant transmission scheme of 5G + wireless Mesh + wired backup. The underground sensor data is first aggregated to the edge computing node in the middle of the roadway through the wireless Mesh network. After the edge computing node preprocesses the data, it is transmitted to the cloud data center via 5G. At the same time, wired optical fiber is deployed as a backup transmission channel. When the 5G signal is interrupted, it automatically switches to wired transmission. The cloud data center adopts a distributed storage architecture, supports data encryption and access control, and uses the AES-256 encryption algorithm to encrypt the transmission of monitoring data and control commands. The edge computing node has a built-in backup power supply.

[0030] The real-time data of multiple parameters, such as stress, displacement, temperature, and humidity, collected by the monitoring network are transmitted to the cloud data center. The multi-dimensional data is then fused and processed using big data analytics algorithms. A surrounding rock deformation prediction model is established based on the geological parameter coupling-attention mechanism LSTM algorithm, with the following formula: ,in For the future Predicted deformation of the surrounding rock. , , These are the LSTM time-series feature weights, geological parameter coupling weights, and stress-strain synergistic weights, respectively. For improved LSTM network output, This is the multi-parameter monitoring data matrix at time t. This is a coupling function for geological parameters. A vector of geological parameters. For stress-strain coordination terms, for The strain value of the surrounding rock at any time As an error correction term, a water content correction term is added to the model to adjust the relevant weight coefficients. Historical monitoring data and geological parameters including water content are input to complete model training. Deformation rate and peak stress are set as dual threshold standards.

[0031] A dual-threshold dynamic adjustment + risk superposition assessment mechanism is adopted. The deformation rate threshold and stress peak threshold are dynamically corrected based on the water-rich geological conditions of the tunnel and the construction stage. The correction formula is as follows: ,in for Real-time dynamic threshold As the initial threshold, for The cumulative excavation depth of the tunnel at any given time The total design depth of the alleyway, for The speed of tunneling at all times, To the maximum permissible tunneling speed, , As a correction factor, the initial threshold for deformation rate is set to 3 mm / d, and the initial threshold for peak stress is set to 20 MPa. When the monitored deformation rate and peak stress of the surrounding rock exceed the preset thresholds after dynamic correction, the system automatically triggers a graded early warning and pushes the early warning information synchronously through the audible and visual alarm and the downhole communication terminal to clarify the specific location and risk level of the abnormal area in the water-rich area.

[0032] After the warning signal is triggered, the hydraulic support control system and the deep grouting system are linked. The hydraulic support adopts a dual closed-loop control strategy of predicted deformation and real-time stress feedback, as shown in the formula: ,in For the target support resistance, Given the current support resistance, For the maximum adjustable resistance increment, To predict the amount of deformation, To allow the maximum deformation, the extension and retraction of the hydraulic support is dynamically adjusted based on data from the fiber optic displacement sensor. The adjustment formula is as follows: ,in For the target scaling amount, This is the current scaling amount. As the displacement compensation coefficient, the support resistance is adaptively adjusted based on the output of the prediction model, with an adjustment range of 15-30MPa. The surface of the hydraulic support is treated with a hydrophobic protective coating, and its expansion and contraction are dynamically adjusted based on the data from the fiber optic displacement sensor. The deep grouting system adopts the fissure-targeted positioning and grouting parameter precise matching technology. Based on the direction of the fissures in the water-rich area detected by ground radar, the angle of the grouting hole arrangement is determined, and the angle with the fissure direction is controlled at 25°. The grouting hole depth is 3m, the hole diameter is 42mm, and adjacent grouting holes are staggered with a spacing of 1.5m to form a three-dimensional grouting network. Before grouting, polyurethane water-stopping agent is injected to form a water-proof curtain, and then hydrophilic epoxy resin grout is used for grouting, controlling the grouting pressure at 3MPa. The grouting volume per section is increased by 20% according to the standard of 5-10m³ / section, completing the targeted reinforcement of the surrounding rock in the water-rich area.

[0033] A dynamic data window-multi-source verification-segmented retraining cyclic optimization technique is adopted to continuously collect monitoring data after regulation and control, and to iteratively optimize the prediction model parameters and support regulation thresholds. Data windows are dynamically divided based on the tunnel excavation progress and surrounding rock deformation rate. A fixed 72-hour window is used when the deformation rate is ≤2mm / d, and automatically switches to a 24-hour sliding window when the deformation rate is >2mm / d. The data within the window includes real-time monitored stress, displacement, temperature, humidity values, and corresponding geological parameters. Multi-source verification is performed on the data within the window. Outliers are eliminated by comparing the correlation between fiber optic grating sensor data and microseismic monitoring energy values, and the model input is corrected by combining downhole geological sketch data. The model is retrained in segments according to the tunneling stage, with 300 training samples per segment. A recent data weighting strategy is used, and key influencing parameters are selected using a random forest algorithm. After training, a dual-index evaluation is performed. When the model prediction error is ≤3% and the prediction trend is consistent for 5 consecutive times, the model parameters are updated and the support regulation threshold is adjusted simultaneously to adapt to the dynamic geological conditions of water-rich areas.

[0034] Example 2

[0035] A distributed layout scheme with three-dimensional positioning and key node reinforcement was adopted. Fiber optic stress sensors, microseismic monitoring probes, and infrared thermal imagers were deployed on key sections of the roof, sidewalls, and floor of the tunnel excavation along the goaf to construct a real-time monitoring network for multiple parameters of surrounding rock stress, displacement, and temperature. The roof fiber optic sensors were installed at 2m intervals along the tunnel axis, with one additional sensor installed 1m on each side of the tunnel centerline, forming a triangular monitoring array. Sidewall sensors were arranged in three layers: the upper layer 0.5m from the roof, the lower layer 0.8m from the floor, and the middle layer located in the middle of the sidewalls, with a 3m interval between layers. Three stress sensors were deployed 8m ahead of the tunnel face to capture stress concentration signals in advance. Floor sensors were also deployed... At the center of the tunnel and 1.5m on both sides, the sensor and the surrounding rock contact surface adopt an arc-shaped fitting design, and the fitting surface is coated with a high thermal conductivity and electrical conductivity paste. Data transmission adopts a redundant transmission scheme of 5G + wireless Mesh + wired backup. The underground sensor data is first aggregated to the edge computing node in the middle of the tunnel through the wireless Mesh network. After the edge computing node preprocesses the data, it is transmitted to the cloud data center via 5G. At the same time, wired optical fiber is deployed as a backup transmission channel. When the 5G signal is interrupted, it automatically switches to wired transmission. The cloud data center adopts a distributed storage architecture, supports data encryption and access control, and uses the AES-256 encryption algorithm to encrypt the transmission of monitoring data and control commands. The edge computing node has a built-in backup power supply.

[0036] The monitoring network collects real-time data on multiple parameters such as stress, displacement, and temperature, which are then transmitted to a cloud data center. The multi-dimensional data is fused and processed using big data analysis algorithms. A surrounding rock deformation prediction model is established based on the geological parameter coupling-attention mechanism LSTM algorithm. Historical monitoring data and geological parameters including risk parameters related to rockburst are input to complete model training. Dual threshold standards for deformation rate and peak stress are set.

[0037] A dual-threshold dynamic adjustment + risk superposition assessment mechanism is adopted. The deformation rate threshold and the peak stress threshold are dynamically corrected based on the geological conditions of the roadway rockburst risk and the construction stage. The initial threshold for deformation rate is set to 5 mm / d, and the initial threshold for peak stress is set to 25 MPa. When a precursor signal of rockburst is detected or the monitoring data exceeds the preset threshold, the system automatically triggers a graded early warning and pushes the early warning information synchronously through the audible and visual alarm and the underground communication terminal, clarifying the specific location and risk level of the abnormal area within the rockburst risk area.

[0038] After the early warning signal is triggered, the hydraulic support control system and the deep grouting system are linked. The hydraulic support adopts a dual closed-loop control strategy of predicted deformation and real-time stress feedback. It adaptively adjusts the support resistance according to the output of the prediction model, with an adjustment range of 10-30MPa. When encountering a precursor signal of rockburst, the emergency control mode is triggered, and the hydraulic support resistance instantly increases to the maximum threshold of 30MPa. At the same time, the buffer energy dissipation devices on both sides of the roadway are activated to absorb the impact energy through the spring damping structure. The extension and retraction of the hydraulic support are dynamically adjusted according to the data of the fiber optic displacement sensor. The deep grouting system adopts the technology of crack targeted positioning and precise matching of grouting parameters. Based on the crack direction detected by ground radar, the angle of the grouting hole arrangement is determined, and the angle with the crack direction is controlled at 30°. The grouting hole depth is 5m and the hole diameter is 50mm. Adjacent grouting holes are staggered with a spacing of 2.5m to form a three-dimensional grouting network. The high-pressure jet grouting process is adopted, controlling the grouting pressure at 5MPa and the grouting volume of 10m³ / segment to complete the targeted reinforcement of the surrounding rock in the rockburst risk area.

[0039] A dynamic data window-multi-source verification-segmented retraining cyclic optimization technique is adopted to continuously collect monitoring data after regulation and control, and to iteratively optimize the prediction model parameters and support regulation thresholds. Data windows are dynamically divided based on the tunnel excavation progress and surrounding rock deformation rate. A fixed 72-hour window is used when the deformation rate is ≤2mm / d, and automatically switches to a 24-hour sliding window when the deformation rate is >2mm / d. The data within the window includes real-time monitored stress, displacement, temperature values, and corresponding geological parameters. Multi-source verification is performed on the data within the window. Outliers are eliminated by comparing the correlation between fiber optic grating sensor data and microseismic monitoring energy values, and the model input is corrected by combining downhole geological sketch data. The model is retrained in segments according to the tunneling stage, with 350 training samples per segment. A recent data weighting strategy is used, and key influencing parameters are selected using a random forest algorithm. After training, a dual-index evaluation is performed. When the model prediction error is ≤3% and the prediction trend is consistent for 5 consecutive times, the model parameters are updated and the support regulation threshold is adjusted synchronously to adapt to the dynamic geological conditions of the rockburst risk area.

[0040] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for support and construction of goaf excavation roadways in deep coal mines during upward mining, characterized in that, The specific steps of this method are as follows: Monitoring network deployment: A distributed layout scheme with three-dimensional depth and key node reinforcement is adopted. Fiber optic stress sensors, microseismic monitoring probes and infrared thermal imagers are deployed on the key sections of the roof, sides and bottom of the tunnel along the goaf to construct a real-time monitoring network for multiple parameters of surrounding rock stress, displacement and temperature. Data transmission adopts a redundant transmission scheme of 5G + wireless Mesh + wired backup. Data processing and prediction: The monitoring data is transmitted to the cloud data center in real time. The multi-dimensional data is fused and processed through big data analysis algorithms. A surrounding rock deformation prediction model is established based on the geological parameter coupling-attention mechanism LSTM algorithm. Historical monitoring data and geological parameters are input to complete the model training. Dual threshold standards for deformation rate and stress peak are set. Intelligent early warning trigger: The system adopts a dual-threshold dynamic adjustment + risk superposition assessment mechanism. When the monitoring data exceeds the preset threshold, the system automatically triggers a graded early warning and pushes the early warning information synchronously through the audible and visual alarm and the downhole communication terminal, clearly identifying the location of the abnormal area and the risk level. Coordinated control and execution: After the early warning signal is triggered, the hydraulic support control system and the deep grouting system are linked; the hydraulic support adopts a dual closed-loop control strategy of predicted deformation and real-time stress feedback, and adaptively adjusts the support resistance according to the output results of the prediction model, with an adjustment range of 10-30MPa; the deep grouting system adopts the technology of crack targeted positioning and precise matching of grouting parameters, and controls the grouting pressure at 2-5MPa and the grouting volume at 5-10m³ / segment based on the geological data of the abnormal area to complete the targeted reinforcement; Cyclic optimization: The dynamic data window-multi-source verification-segmented retraining cyclic optimization technology is adopted to continuously collect monitoring data after regulation and to iteratively optimize the prediction model parameters and support regulation thresholds in reverse; at the same time, an adaptation adjustment mechanism is preset for special geological conditions such as water-rich and rockburst.

2. The method for support and construction of goaf excavation roadways in deep coal mines as described in claim 1, characterized in that, The three-dimensional + key node reinforced distributed layout scheme in the monitoring network deployment is as follows: the spacing of the roof fiber optic grating sensors along the roadway axis is set to 1.5-2m, and one sensor is added at 0.8-1m on each side of the roadway centerline to form a triangular monitoring array; the sidewall sensors are arranged in three layers: the upper layer is 0.5m from the roof, the lower layer is 0.8m from the floor, and the middle layer is located in the middle of the roadway sidewall, with a layer spacing of ≤3m. In addition, three stress sensors are arranged 5-8m ahead of the tunnel face to capture stress concentration signals in advance; the floor sensors are arranged at the center of the roadway and 1.5m on both sides, and the contact surface between the sensor and the surrounding rock adopts an arc-shaped fitting design, with high thermal conductivity and electrical conductivity paste applied to the fitting surface.

3. The method for support and construction of goaf excavation roadways in deep coal mines as described in claim 1, characterized in that, The 5G+wireless Mesh+wired backup redundant transmission scheme described in the monitoring network deployment is as follows: Downhole sensor data is first aggregated to an edge computing node in the middle of the roadway via a wireless Mesh network. The edge computing node preprocesses the data, and the preprocessed data is transmitted to the cloud data center via 5G. Simultaneously, a wired optical fiber is deployed as a backup transmission channel, automatically switching to wired transmission when the 5G signal is interrupted. The cloud data center adopts a distributed storage architecture and supports data encryption and access control, using the AES-256 encryption algorithm to encrypt the transmission of monitoring data and control commands. Furthermore, the edge computing nodes have built-in backup power supplies.

4. The method for support and construction of goaf excavation roadways in deep coal mines as described in claim 1, characterized in that, The surrounding rock deformation prediction model constructed by the LSTM algorithm, which uses a geological parameter coupling-attention mechanism in the data processing and prediction process, is as follows: ,in For the future Predicted deformation of the surrounding rock. , , These are the LSTM time-series feature weights, geological parameter coupling weights, and stress-strain synergistic weights, respectively. For improved LSTM network output, This is the multi-parameter monitoring data matrix at time t. This is a coupling function for geological parameters. A vector of geological parameters. For stress-strain coordination terms, for The strain value of the surrounding rock at any time This is the error correction term.

5. The method for support and construction of goaf excavation roadways in deep coal mines as described in claim 1, characterized in that, The intelligent early warning triggering mechanism, which combines dynamic adjustment of dual thresholds with risk superposition assessment, specifically involves dynamically adjusting both the deformation rate threshold and the peak stress threshold based on the tunnel geological conditions and construction stage. The adjustment formula is as follows: ,in for Real-time dynamic threshold As the initial threshold, for The cumulative excavation depth of the tunnel at any given time The total design depth of the alleyway, for The speed of tunneling at all times, To the maximum permissible tunneling speed, , This is a correction factor.

6. The method for support and construction of goaf excavation roadways in deep coal mines as described in claim 1, characterized in that, The specific formula for the dual closed-loop control strategy of predicting hydraulic support deformation and real-time stress feedback during the coordinated regulation execution is as follows: ,in For the target support resistance, Given the current support resistance, For the maximum adjustable resistance increment, To predict the amount of deformation, To allow the maximum deformation, the extension and retraction of the hydraulic support is dynamically adjusted based on data from the fiber optic displacement sensor. The adjustment formula is as follows: ,in For the target scaling amount, This is the current scaling amount. This is the displacement compensation coefficient.

7. The method for support and construction of goaf excavation roadways in deep coal mines as described in claim 1, characterized in that, The specific details of the collaborative control and execution of the deep grouting system's fracture-targeted positioning and precise matching of grouting parameters are as follows: the grouting hole arrangement angle is determined based on the fracture orientation detected by ground-penetrating radar, with the angle between the grouting hole and the fracture orientation ≤30°; the grouting hole depth is 2-5m, the hole diameter is 42-50mm, and adjacent grouting holes are staggered with a spacing of 1.5-2.5m to form a three-dimensional grouting network; the grouting volume is precisely calculated using the following formula: ,in This refers to the grouting volume per section. The grouting filling coefficient is... To prevent grouting from affecting volume, Porosity of the surrounding rock This is the slurry shrinkage compensation coefficient.

8. The method for support and construction of goaf excavation roadways in deep coal mines as described in claim 1, characterized in that, The iterative optimization employs a dynamic data window-multi-source verification-segmented retraining iterative optimization technique. The specific implementation steps are as follows: First, data windows are dynamically divided based on the tunnel excavation progress and the surrounding rock deformation rate. When the deformation rate is ≤2mm / d, a fixed 72-hour window is used; when the deformation rate is >2mm / d, it automatically switches to a 24-hour sliding window. The data within the window includes real-time monitored stress, displacement, temperature values, and corresponding geological parameters. Second, multi-source verification is performed on the data within the window. By comparing the correlation between fiber optic grating sensor data and microseismic monitoring energy values, outliers are eliminated, and the model input is corrected in conjunction with underground geological sketch data. Third, the model is retrained segmentally according to the tunneling stage, with each segment having ≥300 training samples. A recent data weighting strategy is used, and key influencing parameters are selected using a random forest algorithm. Finally, after training, a dual-index evaluation is performed. When the model prediction error is ≤3% and the prediction trend is consistent for five consecutive times, the model parameters are updated, and the support control threshold is adjusted simultaneously.

9. A method for supporting and constructing goaf-side excavation roadways in deep coal mines as described in claim 1, characterized in that, The specific adjustment mechanism for special geological conditions in the cyclic optimization is as follows: When a water-rich area is detected in the hollow area, the support and construction parameters are automatically adjusted: the data acquisition frequency of the fiber optic grating sensor is increased to 3-5Hz, and a humidity sensor is added for synchronous monitoring; a water content correction term is added to the surrounding rock deformation prediction model to adjust the relevant weight coefficients; the lower limit of hydraulic support resistance is increased to 15MPa, and a hydrophobic protective coating is used to treat the support surface; the deep grouting system is changed to hydrophilic epoxy resin grout, the grouting pressure is increased to 3-5MPa, the grouting volume is increased by 20%-30%, and a polyurethane water-stopping agent is injected before grouting to form a water-proof curtain; when encountering a precursor signal of rockburst, the emergency control mode is triggered, the hydraulic support resistance is instantly increased to the maximum threshold of 30MPa, and the buffer energy dissipation devices on both sides of the roadway are activated simultaneously to absorb the impact energy through the spring damping structure, and the grouting system adopts a high-pressure jet grouting process.

Images