Multi-source perception fusion driven intelligent control method and system for rapid tunneling
Patent Information
- Application Number
- CN202610842007.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-11
- Publication Date
- 2026-09-15
Smart Images
Figure CN122752046A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of rapid tunneling and intelligent mining technology, specifically relating to a multi-source sensing fusion-driven intelligent control method and system for rapid tunneling. Background Technology
[0002] With the continuous increase in the depth and intensity of coal mining in my country, rapid tunneling has become a core technological requirement for alleviating the tension in mining succession and ensuring safe and efficient mine production. Its tunneling efficiency directly determines the overall production capacity release level of the mine. Currently, coal mine tunneling generally faces prominent challenges such as complex and variable geological conditions, difficulty in controlling the stability of surrounding rock, and a high proportion of time spent on support operations, severely restricting further improvements in tunneling efficiency. To overcome these difficulties, China has developed and widely applied rapid tunneling systems centered on integrated tunneling, support, and transportation. These systems have achieved significant results in key aspects such as improving cutting power, optimizing rock-breaking performance, and realizing continuous transportation, effectively promoting the upgrading of tunneling operation modes. However, existing rapid tunneling systems still have significant shortcomings in intelligent control: the setting of tunneling speed and the design of anchor bolt support parameters largely rely on the static scheme formulation before tunneling or the experience judgment of on-site operators, lacking quantitative perception support of the real-time stability state of the surrounding rock, making it difficult to dynamically adapt and adjust according to the actual surrounding rock conditions revealed during tunneling. Once the support density is determined, it remains fixed, which easily leads to the dual problems of over-supporting when the surrounding rock conditions are good, wasting time and construction costs, and under-supporting when the surrounding rock conditions are poor, posing safety hazards. Therefore, how to dynamically adjust the support density based on the real-time condition of the surrounding rock and adaptively match the tunneling speed has become a core technical bottleneck that urgently needs to be solved to achieve intelligent and rapid tunneling in coal mine roadways.
[0003] In recent years, significant progress has been made in both intelligent tunneling technology and surrounding rock monitoring technology in coal mines. However, there are still significant shortcomings in their deep integration and application, failing to achieve a synergistic control effect. On the one hand, existing surrounding rock monitoring systems mostly focus on data acquisition and over-limit alarm functions. Although they can achieve real-time online monitoring of key parameters such as roof delamination, anchor bolt stress, surrounding rock stress, and deformation, the sensed data and the tunneling equipment control system are independent of each other, failing to establish a complete link between monitoring, analysis, and control. Even if some systems can issue early warnings of surrounding rock anomalies, they cannot automatically generate quantitative control commands. Adjustments to support parameters and changes in tunneling speed still rely on manual decision-making, resulting in not only delayed response but also control accuracy that cannot meet actual engineering needs. On the other hand, existing research on intelligent tunneling control focuses more on the equipment's own trajectory planning, cutting parameter optimization, and remote visual control, rarely using the real-time stable state of the surrounding rock as the core input variable of the control model. This leads to a disconnect between surrounding rock sensing and tunneling control at the information transmission level, preventing dynamic synergy between the two. The "Key Research and Development Catalog of Intelligent Robots for Mining" issued by the National Mine Safety Administration clearly requires that intelligent tunneling robots possess core functions such as intelligent perception through multi-sensor data fusion and dynamic optimization of operational parameters. Therefore, transforming multi-source surrounding rock perception data into executable quantitative control decisions, establishing a dynamic matching mechanism between surrounding rock condition, support density, and tunneling speed, and especially constructing a stability quantitative assessment model that comprehensively reflects the multiple effects of surrounding rock strength reserve, damage evolution, and creep aging, and based on this, forming deterministic control rules between safety factors and support parameters, has become a core technological direction urgently needing breakthroughs in the field of rapid tunneling in coal mine roadways.
[0004] To address the shortcomings of existing technologies, there is an urgent need for a multi-source sensing fusion-driven intelligent control method for rapid tunnel excavation, which will provide technical support for overcoming the core technological bottlenecks in intelligent rapid tunneling. Summary of the Invention
[0005] To address the problems existing in the prior art, this invention provides a multi-source sensing fusion-driven intelligent control method and system for rapid tunnel excavation. This method can simultaneously ensure the stability of the surrounding rock and the excavation efficiency, and can significantly improve the automation and adaptive control level of rapid tunnel excavation in coal mines. The system can realize real-time probabilistic evaluation of the stability of the surrounding rock during the excavation process, risk classification and early warning, autonomous optimization of support and propulsion parameters, and adaptive correction of the model. It has the advantages of high reliability, high efficiency, and strong robustness in complex rapid tunnel excavation scenarios in coal mines.
[0006] To achieve the above objectives, the present invention provides an intelligent control method for rapid tunnel excavation driven by multi-source sensing fusion, comprising the following steps: Step 1: Multi-source sensing and data acquisition, spatiotemporal fusion; Multi-source sensing mechanisms are deployed synchronously within a range of 5m to 30m behind the tunneling face; multi-source physical field data are collected through the multi-source sensing mechanisms, and a multi-dimensional feature matrix with a unified spatiotemporal reference is generated after data preprocessing; Step 2: Reliability assessment of disturbance coupling; Constructing a rock stability function that integrates strength theory, damage mechanics, and rheology This includes the introduction of additional disturbance stress calculated in real time from the cutting parameters of the tunneling machine. The fracture mechanics effect of mechanical rock breaking on the instantaneous stress field of the surrounding rock is characterized; based on the mean and standard deviation of the function within the sliding window, the real-time reliability index is calculated using the first second moment method. The measured variability of each parameter is uniformly quantified into a probabilistic stability criterion; Step 3: Risk classification and composite decision-making; Based on real-time reliability indicators and its least-squares linear fit slope in the most recent three complete tunneling cycles. The surrounding rock condition is divided into four levels: stable, concerning, warning, and dangerous. At the same time, the short-term sudden increase in the slope of the microseismic energy release rate is monitored to trigger feedforward warning. Based on different levels, anchor spacing control instructions and tunneling speed adaptive matching instructions are generated and issued according to standard industrial protocols.
[0007] As a preferred option, in step 1, the process of multi-source sensing, data acquisition, and spatiotemporal fusion is as follows: S11: Multi-source sensor deployment; Fiber grating strain sensor strings with a center wavelength demodulation accuracy of not less than 1 pm are deployed in the tunnel roof and boreholes on both sides, with an adjacent measuring point spacing of 0.5 m; a three-dimensional array of microseismic detectors with a sensitivity of not less than 100 V / g and a frequency band coverage of 10 Hz to 2000 Hz is deployed in the surrounding rock of the tunnel, covering an area 20 m in front of the face and 30 m behind; a point cloud density of not less than 10 is erected behind the tunneling machine. 4 A 3D laser scanner with a point-per-square-meter resolution; S12: Continuous acquisition of multi-source data; The data on tangential strain and axial displacement distribution on the surrounding rock surface are continuously collected using a fiber optic strain sensor at a sampling frequency of 50 Hz and converted into digital signals by a fiber optic demodulator. Based on the continuous acquisition of microseismic waveform signals generated by rock mass fracture using a microseismic detector, with a sampling frequency of not less than 2kHz, microseismic events are detected by triggering thresholds, and the source location, energy and occurrence time are recorded; Based on the 3D laser scanner, the tunnel cross-section point cloud is automatically collected once after each cutting cycle, and high-precision 3D spatial coordinate data of roof subsidence, sidewall movement and cross-section convergence deformation are obtained. All sensor data is transmitted in real time to edge computing nodes deployed near the tunneling face via the underground Ethernet ring network TCP / IP protocol. S13: Data preprocessing; Extended Kalman filtering is used to align timestamps and suppress random noise in multi-source data, eliminating sensor noise and transmission interference. S14: Spatiotemporal fusion and feature matrix generation; Adaptive weighted fusion based on sensor prior confidence is used to generate a multidimensional feature matrix under a unified spatiotemporal reference.
[0008] As a preferred embodiment, the reliability assessment process for the disturbance coupling in step 2 is as follows: S21: Function definition; Rock stability function Calculate using the following formula: ; In the formula, The uniaxial compressive strength of the surrounding rock; This represents the measured maximum tangential stress. This refers to the stress term caused by tunneling disturbance. This represents the rate of damage zone expansion. The critical spread rate; The strain gradient change rate of the top plate; For reference strain rate; and Two dynamic weighted indices; This is the lithological attenuation coefficient; This is the safety threshold constant; S22: Additional disturbance stress Calculation; Additional disturbance stress Calculate using the following formula: ; In the formula, To cut off the motor torque; This refers to the cutting head rotation speed; Where is the diameter of the cutting head; F is the propulsion force of the tunneling machine; The cross-sectional area at the heading; , The coefficient is a dimensionless coupling coefficient. S23: Damage zone expansion rate Damage zone expansion rate Calculate using the following formula: ; In the formula, For the first Micro-seismic events at time The source location vector; It is the Euclidean norm; S24: Rate of change of strain gradient in the top plate Calculation: The axial displacement field measured by the fiber Bragg grating sensor is first calculated, then the spatial strain is determined, and finally, the derivative with respect to time is obtained. As shown in the following formula: ; In the formula, This is the total length of the sensor array; The average value of the monitoring interval of the roof is taken as the final value. ; S25: Real-time Reliability Indicators Calculate the function within a sliding window of 10 consecutive sampling periods (corresponding to 10 data points). sample mean and sample variance Take the standard deviation The real-time reliability index As shown in the following formula: .
[0009] As a preferred option, the risk classification and composite decision-making process in step 3 is as follows: S31: Monitoring of short-term sudden increase in microseismic energy release rate; The slope of the short-term surge in energy release rate is defined as: the normalized ratio of the average energy release rate in the current 5 minutes to the average energy release rate in the previous 15 minutes; when this ratio exceeds 1.8, a feedforward warning signal is triggered, forcibly increasing the frequency of control command generation from once per cycle to twice per cycle, and prioritizing the execution of support enhancement operations; S32: Risk Classification Rules; Based on the following process and its trend slope Risk classification: when and When the situation is deemed stable, the spacing between anchor bolts is increased. when or When this occurs, it is classified as a concern level, the current spacing between rows is maintained, and the maximum advance per cycle is limited to 0.8 times the rated value; when Or the slope of the sudden increase in the energy release rate of microseismic events At 1.8, the alert level was set, and the spacing between rows was reduced. when If the situation is deemed dangerous, the spacing between rows should be immediately reduced to 0.8 times the thickness of the top slab layer, and an automatic shutdown should be triggered. S33: Anchor bolt spacing adjustment; Let the current row spacing be The thickness of the top strata is The rules for adjusting the spacing between anchor bolts are as follows: At a stable level, Calculate using the following formula: ; After magnification ; At the warning level, Calculate using the following formula: ; After shrinking ; When the danger level is reached, Calculate using the following formula: ; S34: Adaptive matching of tunneling speed; tunneling speed Calculate using the following formula: ; In the formula, This represents the current actual tunneling speed; For nonlinear velocity correction factors, when hour, ,when hour, ,when hour, ,when At that time, the machine had stopped and its speed was zero. The adjusted speed satisfies ; S35: Instruction issuance and execution; Control commands are sent to the electro-hydraulic proportional control system of the bolt drilling rig and the frequency conversion speed control system of the tunneling machine via Modbus TCP or PROFINET industrial protocols, and are executed automatically.
[0010] Furthermore, in order to achieve adaptive evolution of model parameters and feedforward enhancement of decision-making, step 4 is also included: parameter tuning and closed-loop evolution; After each regulation cycle, the equivalent safety factor observation is retrieved using newly acquired laser point cloud convergence data, and the lithology attenuation coefficient in the function is estimated using recursive Bayesian estimation. Online corrections are performed; meanwhile, the reduced-order digital twin model deployed in the edge computing nodes continuously inverts the stress field and damage field of the surrounding rock, and superimposes the change trend of the reliability index in the next 1 to 3 cycles as a feedforward component into the risk classification decision.
[0011] As a further optimization, the parameter adjustment and closed-loop evolution process in step 4 is as follows: S41: Equivalent safety factor inversion; The iterative nearest point algorithm was used to register the laser point clouds of two adjacent periods, calculate the roof subsidence rate and the approach rate of the two sides, and invert the equivalent safety factor observation value by combining the fiber optic strain data. ; S42: Recursive Bayesian Estimation Correction ; With lithological attenuation coefficient For the state variables, the state equation and observation equation are established as follows: In the formula, For process noise, To observe noise; The mapping relationship between the equivalent safety factor and the lithological attenuation coefficient pre-calculated by the digital twin model; Recursive update using unscented Kalman filtering The posterior mean and variance are used to output a correction value after each adjustment cycle. When the correction magnitude of three consecutive cycles At that time, the digital twin model parameter calibration is triggered; S43: Reduced-order digital twin model; A reduced-order digital twin model of the surrounding rock of the roadway is deployed within the edge computing node. The digital twin model uses real-time fiber optic strain data as boundary condition input, microseismic event rate distribution as the basis for updating the internal damage field, and laser convergence data as the displacement field verification benchmark. It continuously inverts the spatial distribution of the current stress field and damage field of the surrounding rock and outputs the predicted trend of reliability index changes in the next 1 to 3 cycles. This trend is then superimposed as a feedforward component into the risk classification decision in step three. S44: Closed-loop iteration; By running the process in a cyclical manner according to steps 1 to 4, the on-site perception, reliability assessment, risk classification, control execution and model correction are linked into a complete closed loop of iterative iteration, enabling the surrounding rock stability control during the tunnel excavation process to be dynamically adjusted based on real-time monitoring data, thus achieving adaptive management.
[0012] This invention breaks through the traditional passive mode of speed-based support, addressing the problem of disconnect between surrounding rock perception and control and the lack of real-time quantitative basis for regulation during tunneling. It proposes a multi-source sensing fusion-driven intelligent control method for rapid tunneling. First, a multi-source sensing mechanism is simultaneously deployed behind the tunneling face to comprehensively acquire continuous multi-physics field data on internal rock strain, micro-fracture events, and surface convergence deformation. After data preprocessing, a multi-dimensional feature matrix with a unified spatiotemporal reference is generated, providing a high-precision, spatiotemporally aligned unified data foundation for subsequent evaluation. Second, for the first time, a surrounding rock stability function integrating strength theory, damage mechanics, and rheology is constructed, innovatively calculating the tunneling machine cutting parameters (torque, speed, and thrust) as additional disturbance stress in real time. Furthermore, a function was introduced to quantify the deterioration effect of mechanical rock breaking, overcoming the limitation of traditional methods that ignore the dynamic disturbance of mechanical rock breaking. Real-time reliability indices were calculated based on the first-order second-moment method. The measured variability of parameters such as surrounding rock strength, damage propagation rate, and creep strain rate was uniformly quantified into probabilistic criteria, significantly improving the scientific rigor and reliability of the assessment. Subsequently, based on reliability indicators and their changing trends, and integrating short-term surge detection of microseismic energy release rate, a four-level risk classification rule (stable, watchful, early warning, and hazardous) was constructed, achieving multi-dimensional and highly sensitive identification of the surrounding rock condition of the roadway. An adaptive control strategy for anchor bolt spacing and tunneling speed linked to the risk level was formulated, and a nonlinear speed correction factor was introduced to maximize tunneling efficiency while ensuring safety.
[0013] This method transforms the traditional passive mode of speed-based support into an active intelligent control mode of perception, evaluation, and decision-making. Under strong disturbance and complex geological conditions, it can simultaneously ensure the stability of the surrounding rock and the tunneling efficiency, significantly improving the automation and adaptive control level of rapid tunneling in coal mine roadways.
[0014] This invention also provides a multi-source sensing fusion-driven intelligent control system for rapid tunnel excavation, used to implement a multi-source sensing fusion-driven intelligent control method for rapid tunnel excavation, comprising: The multi-source sensing and edge computing layer includes a fiber optic demodulator, a microseismic acquisition station, a laser scanning host, and edge computing nodes. The edge computing nodes have built-in data spatiotemporal integration modules and real-time reliability calculation modules. The multi-source sensing and edge computing layer is responsible for the continuous acquisition, spatiotemporal fusion, and real-time calculation of reliability indicators of multi-source physical field data, providing unified and high-precision state input for upper-level decision-making. The digital twin and model evolution layer includes a reduced-order digital twin model of the tunnel and an online parameter correction module based on unscented Kalman filtering. The digital twin and model evolution layer is used to invert the surrounding rock stress / damage field online through the reduced-order digital twin model and recursively correct the model parameters using unscented Kalman filtering to achieve adaptive evolution of the system to geological conditions. The risk decision and control command generation layer includes a four-level risk classification module, a feedforward early warning trigger module, a spacing and speed matching control module, and an output module. The risk decision and control command generation layer is used to complete the four-level risk classification based on reliability indicators and their trends, combined with microseismic feedforward early warning, and generate differentiated control commands for anchor spacing and tunneling speed, forming a composite decision of feedback and early warning. The equipment execution and visualization layer includes an equipment execution linkage module and a visualization module. The equipment execution and visualization layer is used to send control commands to execution equipment such as tunneling machines and anchor drilling rigs through standard industrial protocols, and to visualize the surrounding rock status, risk level and control effect in real time, completing the end execution and human-machine interaction of the closed-loop control chain.
[0015] In this invention, the multi-source sensing and edge computing layer integrates a fiber optic demodulator, a microseismic acquisition station, a laser scanning host, and edge computing nodes. It incorporates a built-in data spatiotemporal fusion and real-time reliability calculation module, facilitating rapid on-site processing and feature extraction of multi-physics data. This significantly reduces communication latency when uploading data to the ground, providing technical support for real-time control. The digital twin and model evolution layer deploys a roadway-reduced digital twin model and an online parameter correction module based on unscented Kalman filtering. This enables the inversion of the surrounding rock stress field and damage field at extremely low computational cost, and dynamically updates key model parameters (such as the lithological attenuation coefficient), giving the system self-learning and evolutionary capabilities, effectively combating geological uncertainties. The risk decision-making and control command generation layer incorporates a four-level risk classification module, a feedforward early warning trigger module, and a spacing and speed matching control module. This automatically generates precise support and advance commands based on changes in real-time reliability indicators, achieving differentiated risk classification and avoiding a one-size-fits-all conservative control approach. The equipment execution and visualization layer is linked with the electro-hydraulic proportional system of the anchor drilling rig and the frequency conversion speed control system of the tunneling machine through the industrial control network, ensuring the reliable issuance and rapid execution of control commands. The augmented reality color bands (green / yellow / orange / red) are displayed on the cockpit screen in real time to show the risk level and key parameters, which greatly improves the operator's situational awareness and emergency response efficiency.
[0016] The system has a simple structure and a high degree of intelligence. It realizes real-time probabilistic evaluation of the stability of the surrounding rock during the tunneling process, risk classification and early warning, autonomous optimization of support and propulsion parameters, and adaptive correction of the model. It has high reliability, high efficiency and strong robustness in the scenario of rapid tunneling in complex coal mine roadways. Attached Figure Description
[0017] Figure 1 This is a flowchart of the control method section of this invention; Figure 2 This is a schematic diagram of the control system part of the present invention. Detailed Implementation
[0018] like Figure 1 As shown, this invention provides an intelligent control method for rapid tunnel excavation driven by multi-source sensing fusion, comprising the following steps: Step 1: Multi-source sensing and data acquisition, spatiotemporal fusion; Multi-source sensing mechanisms (including a high-precision fiber optic grating sensor array, a broadband micro-vibration detector array, and a high-density three-dimensional laser scanning device) are simultaneously deployed within a range of 5m to 30m behind the tunneling face. These mechanisms are connected to edge computing nodes via an underground Ethernet ring network. Multi-source physical field data (including the internal strain distribution of the surrounding rock, the spatiotemporal distribution of micro-fracture events, and the convergence deformation of the roadway surface) are collected through the multi-source sensing mechanisms and transmitted to the edge computing nodes via an underground Ethernet ring network. After data preprocessing (timestamp alignment and noise reduction via extended Kalman filtering, followed by adaptive weighted fusion), a multi-dimensional feature matrix with a unified spatiotemporal reference is generated. As a preferred approach, the process of multi-source sensing, data acquisition, and spatiotemporal fusion is as follows: S11: Multi-source sensor deployment; Fiber grating strain sensor strings with a center wavelength demodulation accuracy of not less than 1 pm are deployed in the tunnel roof and boreholes on both sides, with an adjacent measuring point spacing of 0.5 m; a three-dimensional array of microseismic detectors with a sensitivity of not less than 100 V / g and a frequency band coverage of 10 Hz to 2000 Hz is deployed in the surrounding rock of the tunnel, covering an area 20 m in front of the face and 30 m behind; a point cloud density of not less than 10 is erected behind the tunneling machine. 4 A 3D laser scanner with a point / square meter capability; three types of sensors are connected to edge computing nodes via a downhole Ethernet ring network, and all of them operate at a fixed sampling / acquisition frequency. The acquired data is connected to the edge computing nodes via the downhole Ethernet ring network. S12: Continuous acquisition of multi-source data; The data on tangential strain and axial displacement distribution on the surrounding rock surface are continuously collected using a fiber optic strain sensor at a sampling frequency of 50 Hz and converted into digital signals by a fiber optic demodulator. Based on the continuous acquisition of microseismic waveform signals generated by rock mass fracture using a microseismic detector, with a sampling frequency of not less than 2kHz, microseismic events are detected by triggering thresholds, and the source location (wave velocity model and multi-point time difference of arrival location), energy and occurrence time are recorded. Based on the 3D laser scanner, the tunnel cross-section point cloud is automatically collected once after each cutting cycle, and high-precision 3D spatial coordinate data of roof subsidence, sidewall movement and cross-section convergence deformation are obtained. All sensor data is transmitted in real time to edge computing nodes deployed near the tunneling face via the underground Ethernet ring network TCP / IP protocol. S13: Data preprocessing; Extended Kalman filtering is used to perform timestamp alignment (synchronization accuracy better than 1ms) and random noise suppression on multi-source data, eliminating sensor noise and transmission interference. S14: Spatiotemporal fusion and feature matrix generation; Adaptive weighted fusion is performed based on the prior confidence of the sensors (0.4 for fiber optic grating sensors, 0.3 for micro-vibration detectors, and 0.3 for laser scanning) to generate a multi-dimensional feature matrix under a unified spatiotemporal reference, providing a unified high-quality input for subsequent reliability assessment. In this technical solution, by simultaneously deploying high-precision fiber optic strain sensors, broadband microseismic detector arrays, and high-density 3D laser scanners, and utilizing downhole Ethernet ring networks to transmit data in real time, continuous multi-physics sensing of internal strain, micro-fracture events, and surface convergence deformation of the surrounding rock is achieved. Extended Kalman filtering is employed to achieve millisecond-level timestamp alignment and noise suppression, effectively eliminating environmental and transmission interference. Adaptive weighted fusion based on prior confidence further generates a multi-dimensional feature matrix with a unified spatiotemporal benchmark, providing a high-quality, synchronous data foundation for subsequent reliability assessment, and significantly improving the comprehensiveness, accuracy, and real-time performance of surrounding rock condition sensing.
[0019] Step 2: Reliability assessment of disturbance coupling; Constructing a rock stability function that integrates strength theory, damage mechanics, and rheology This includes the introduction of additional disturbance stress calculated in real time from the cutting parameters of the tunneling machine. The fracture mechanics effect of mechanical rock breaking on the instantaneous stress field of the surrounding rock is characterized; based on the mean and standard deviation of the function within the sliding window, the real-time reliability index is calculated using the first second moment method. The measured variability of each parameter is uniformly quantified into a probabilistic stability criterion; As a preferred option, the reliability assessment process for perturbation coupling is as follows: S21: Definition of the function; Constructing a function for surrounding rock stability that integrates strength theory, damage mechanics, and rheology. ,Should An additional disturbance stress is introduced, calculated in real time from the torque, speed, and thrust of the tunneling machine's cutting motor. To quantitatively characterize the fracture mechanics effect of mechanical rock breaking on the instantaneous stress field of the surrounding rock, specifically Calculate using the following formula: ; In the formula, The uniaxial compressive strength of the surrounding rock is expressed in MPa and is obtained by point load tests at no less than 3 measuring points every 50m of excavation on site. The measured maximum tangential stress, in MPa, is the surface tangential strain obtained by demodulation from a fiber optic strain sensor. According to Hooke's Law Conversion, including elastic modulus Indoor uniaxial compression test using core samples taken from boreholes in the same geological area; This is the stress term for tunneling disturbance, in MPa. The rate of damage expansion is expressed in m / s, based on a unit time window. Internal microseismic events are determined by calculating the spatial centroid migration rate using a source energy weighting method. The critical propagation rate, in m / s, is the statistical lower quartile of the propagation rate corresponding to the damage acceleration initiation point in at least 5 historical instability cases in the same mining area. The strain gradient change rate of the top plate is expressed in seconds (S). -1 The displacement field is calculated from the axial displacement field of the fiber optic grating sensor. Reference strain rate, in units of S -1 The value is taken as the upper limit of the steady-state creep rate corresponding to the design service life of the tunnel; if no measured data is available, then [the value is taken as follows]. ; and These are two dynamic weighted indices, when the surrounding rock geological strength index... hour , ,when hour , ,when hour, and Determined by linear interpolation: , ; The lithological attenuation coefficient is 1.2 for mudstone, 1.5 for sandstone, and 1.8 for composite rock layers. This is a safety threshold constant, determined by back-calculation based on historical stable section data, with a default value of 0.5. After the system has accumulated stable state data for no less than 10 complete tunneling cycles, this value will be corrected through maximum likelihood estimation. when At that time, it was determined that the damage to the surrounding rock had entered the accelerated instability stage, directly causing... It skips subsequent sliding window statistical calculations and immediately triggers a danger level alarm; S22: Additional disturbance stress Calculation; Additional disturbance stress Calculate using the following formula: ; In the formula, The torque of the motor is measured in N·m and is obtained in real time from the inverter current feedback and converted by the speed constant. This refers to the cutting head rotation speed, in r / min. The cutting head diameter is in meters (m); F is the tunneling machine propulsion force in kN, measured by a hydraulic cylinder pressure sensor. The cross-sectional area is measured in square meters (m²). , The dimensionless coupling coefficient was obtained by synchronously recording at least 20 sets of cutting parameters and fiber strain response data during the first tunneling cycle and then calibrating it using least squares multivariate linear regression. S23: Damage zone expansion rate Damage zone expansion rate Calculate using the following formula: ; In the formula, For the first Micro-seismic events at time The focal location vector, in meters; The value is a Euclidean norm. If the number of valid events in the window is less than 3, the sliding average of the first three windows is used. S24: Rate of change of strain gradient in the top plate Calculation: The axial displacement field measured by the fiber Bragg grating sensor is first calculated, then the spatial strain is determined, and finally, the derivative with respect to time is obtained. As shown in the following formula: ; In the formula, This represents the total length of the sensor array, in meters (m). The average value of the monitoring interval of the roof is taken as the final value. ; S25: Real-time Reliability Indicators Calculate the function within a sliding window of 10 consecutive sampling periods (corresponding to 10 data points). sample mean and sample variance Take the standard deviation The real-time reliability index As shown in the following formula: ; This index quantifies the measured variability of parameters such as uniaxial compressive strength, elastic modulus, and damage propagation rate of surrounding rock into a probabilistic stability criterion. In this technical solution, a rock stability function integrating strength theory, damage mechanics, and rheology is constructed. For the first time, additional disturbance stress calculated in real time by the cutting parameters of the tunnel boring machine is introduced to quantitatively characterize the fracture mechanics effect of mechanical rock breaking on the stress field of the surrounding rock. Combining the damage zone propagation rate calculated by energy weighting of microseismic events, the roof strain gradient change rate obtained by fiber optic strain, and the lithological attenuation coefficient, the real-time reliability index is calculated using the first second-order moment method. The measured variability of parameters such as rock strength, elastic modulus, and damage propagation rate is uniformly quantified into a probabilistic stability criterion, which significantly improves the scientificity and reliability of the assessment and provides a quantitative basis for the risk of rock instability under disturbance conditions.
[0020] Step 3: Risk classification and composite decision-making; Based on real-time reliability indicators and its least-squares linear fit slope in the most recent three complete tunneling cycles. The surrounding rock condition is divided into four levels: stable, concerning, warning, and dangerous. At the same time, the short-term sudden increase in the slope of the microseismic energy release rate is monitored to trigger feedforward warning. Based on different levels, anchor spacing control instructions and tunneling speed adaptive matching instructions are generated and issued according to standard industrial protocols.
[0021] As a preferred approach, the process of risk stratification and composite decision-making is as follows: S31: Monitoring of short-term sudden increase in microseismic energy release rate; The slope of the short-term surge in energy release rate is defined as the normalized ratio of the average energy release rate in the current 5 minutes to the average energy release rate in the previous 15 minutes. When this ratio exceeds 1.8, a feedforward warning signal is triggered, forcibly increasing the frequency of control command generation from once per cycle to twice per cycle, and prioritizing the execution of support enhancement operations (reducing the spacing between rows and decreasing the advance speed). S32: Risk Classification Rules; Based on the following process and its trend slope Risk classification: when and When the condition is met, it is determined to be a stable level, the spacing between anchor bolts is increased, and the threshold of 2.5 corresponds to the target reliability index for ductile failure of the first-level safety level in the "Unified Standard for Reliability Design of Building Structures" (GB 50068); when or When this occurs, it is classified as a concern level, the current spacing between rows is maintained, and the maximum advance per cycle is limited to 0.8 times the rated value; when Or the slope of the sudden increase in the energy release rate of microseismic events When the value is 1.8, it is determined to be a warning level, and the spacing between rows is reduced. The thresholds of 1.8 and 1.2 are determined by referring to the relevant clauses of the "Technical Specification for Anchor Bolt Support in Coal Mine Roadways" (GB / T 35056) and combining them with the engineering experience of this mining area. when If the situation is deemed dangerous, the spacing between rows should be immediately reduced to 0.8 times the thickness of the top slab layer, and an automatic shutdown should be triggered. S33: Anchor bolt spacing adjustment; Let the current anchor spacing be... The unit is mm. If the spacing and row spacing are not equal, the smaller value of the two shall be taken. The thickness of the top strata is... The unit is meters (m), obtained from geological sketches and then converted; the spacing between anchor bolts. The regulatory rules are as follows: At a stable level, Calculate using the following formula: ; After magnification ; At the warning level, Calculate using the following formula: ; After shrinking ; When the danger level is reached, Calculate using the following formula: ; S34: Adaptive matching of tunneling speed; tunneling speed (Unit: m³ / h) Calculate using the following formula: ; In the formula, The current actual tunneling speed is expressed in m / h. For nonlinear velocity correction factors, when hour, ,when hour, ,when hour, ,when At that time, the machine had stopped and its speed was zero. The adjusted speed satisfies ( The minimum stable operating speed for the tunneling machine, (Rated maximum feed rate); S35: Instruction issuance and execution; For different risk levels, control commands for the spacing between anchor bolts and adaptive matching commands for tunneling speed are generated. The control results are constrained by a lower limit of 0.8 times the thickness of the roof rock strata and an upper limit of 1.5 times. The control commands are sent to the electro-hydraulic proportional control system of the anchor bolt drilling rig (for adjusting the spacing between anchor bolts) and the variable frequency speed control system of the tunneling machine (for adjusting the speed) via ModbusTCP or PROFINET industrial protocols, and are executed automatically.
[0022] This technical solution integrates reliability indicators and their evolution trends with short-term surge feedforward early warning of microseismic energy release rates to construct a four-level risk classification rule (stable, watchful, warning, and hazardous). Based on this, a differentiated adaptive control strategy for anchor bolt spacing and tunneling speed is formulated. The spacing control incorporates a nonlinear scaling factor related to the reliability indicators, while speed control uses a stepped correction coefficient to maximize tunneling efficiency while ensuring surrounding rock stability. Finally, the control commands are automatically executed by the anchor bolt drilling rig and tunneling machine via a standard industrial protocol. This solution forms a composite decision-making closed loop combining feedback evaluation and feedforward early warning, enabling accurate identification and rapid response to gradual and sudden instability of the surrounding rock, significantly improving the safety, intelligence, and efficiency of the tunneling process.
[0023] As another specific embodiment of the present invention, in order to achieve adaptive evolution of model parameters and feedforward enhancement of decision, step 4 is also included: parameter tuning and closed-loop evolution. After each regulation cycle, the equivalent safety factor observation is retrieved using newly acquired laser point cloud convergence data, and the lithology attenuation coefficient in the function is estimated using recursive Bayesian estimation. Online corrections are performed; meanwhile, the reduced-order digital twin model deployed in the edge computing nodes continuously inverts the stress field and damage field of the surrounding rock, and superimposes the change trend of the reliability index in the next 1 to 3 cycles as a feedforward component into the risk classification decision.
[0024] As a preferred option, the process of parameter tuning and closed-loop evolution is as follows: S41: Equivalent safety factor inversion; The Iterative Closest Point (ICP) algorithm was used to register the laser point clouds of two adjacent periods, calculate the roof subsidence rate and the approach rates of the two side walls, and then use fiber optic strain data to derive the equivalent safety factor observation value. ; S42: Recursive Bayesian Estimation Correction ; With lithological attenuation coefficient For the state variables, the state equation and observation equation are established as follows: ; ; In the formula, For process noise, To observe noise; The mapping relationship between the equivalent safety factor and the lithological attenuation coefficient pre-calculated by the digital twin model (linear approximation is...) , (The safety factor is calibrated using historical data). Unscented Kalman filter (UKF) is used for recursive updates. The posterior mean and variance are used to output a correction value after each adjustment cycle. When the correction magnitude of three consecutive cycles At that time, the digital twin model parameter calibration is triggered; S43: Reduced-order digital twin model; A reduced-order digital twin model of the surrounding rock of the tunnel is deployed within the edge computing node. This model is based on the equivalent continuous medium elastoplastic damage constitutive model. It uses the intrinsic orthogonal decomposition (POD) and discrete empirical interpolation (DEIM) methods to compress the degrees of freedom of the full-order finite element model to less than 5% of the original scale. The single solution time does not exceed 30 seconds. The digital twin model uses real-time fiber optic strain data as boundary condition input, microseismic event rate distribution as the basis for updating the internal damage field, and laser convergence data as the displacement field verification benchmark. It continuously inverts the spatial distribution of the current stress field and damage field of the surrounding rock and outputs the predicted change trend of reliability index in the next 1 to 3 cycles. This trend is then superimposed as a feedforward component into the risk classification decision in step three. S44: Closed-loop iteration; By running the process in a cyclical manner according to steps 1 to 4, the on-site perception, reliability assessment, risk classification, control execution and model correction are linked into a complete closed loop of iterative iteration, enabling the surrounding rock stability control during the tunnel excavation process to be dynamically adjusted based on real-time monitoring data, thus achieving adaptive management.
[0025] In this technical solution, the equivalent safety factor is inverted by laser point cloud and fiber strain data, the lithology attenuation coefficient is recursively corrected by unscented Kalman filtering, and the surrounding rock stress field and damage field are inverted in real time by combining reduced-order digital twin models (POD and DEIM) to predict the reliability change trend, thus forming an adaptive evolution mechanism. This mechanism enables the model parameters and decision basis to be dynamically updated with field data, which significantly improves the adaptability to complex geological conditions and the foresight of the control strategy, and realizes the continuous evolution and intelligent management of the stability control of the surrounding rock in tunnel excavation.
[0026] This invention breaks through the traditional passive mode of speed-based support, addressing the problem of disconnect between surrounding rock perception and control and the lack of real-time quantitative basis for regulation during tunneling. It proposes a multi-source sensing fusion-driven intelligent control method for rapid tunneling. First, a multi-source sensing mechanism is simultaneously deployed behind the tunneling face to comprehensively acquire continuous multi-physics field data on internal rock strain, micro-fracture events, and surface convergence deformation. After data preprocessing, a multi-dimensional feature matrix with a unified spatiotemporal reference is generated, providing a high-precision, spatiotemporally aligned unified data foundation for subsequent evaluation. Second, for the first time, a surrounding rock stability function integrating strength theory, damage mechanics, and rheology is constructed, innovatively calculating the tunneling machine cutting parameters (torque, speed, and thrust) as additional disturbance stress in real time. Furthermore, a function was introduced to quantify the deterioration effect of mechanical rock breaking, overcoming the limitation of traditional methods that ignore the dynamic disturbance of mechanical rock breaking. Real-time reliability indices were calculated based on the first-order second-moment method. The measured variability of parameters such as surrounding rock strength, damage propagation rate, and creep strain rate was uniformly quantified into probabilistic criteria, significantly improving the scientific rigor and reliability of the assessment. Subsequently, based on reliability indicators and their changing trends, and integrating short-term surge detection of microseismic energy release rate, a four-level risk classification rule (stable, watchful, early warning, and hazardous) was constructed, achieving multi-dimensional and highly sensitive identification of the surrounding rock condition of the roadway. An adaptive control strategy for anchor bolt spacing and tunneling speed linked to the risk level was formulated, and a nonlinear speed correction factor was introduced to maximize tunneling efficiency while ensuring safety.
[0027] This method transforms the traditional passive mode of speed-based support into an active intelligent control mode of perception, evaluation, and decision-making, realizing the intelligent transformation from passive experience-based judgment to active reliability-driven control. Under strong disturbance and complex geological conditions, it can simultaneously ensure the stability of the surrounding rock and the tunneling efficiency, significantly improving the automation and adaptive control level of rapid tunneling in coal mine roadways.
[0028] like Figure 2 As shown, the present invention also provides a multi-source sensing fusion-driven intelligent control system for rapid tunnel excavation, used to implement a multi-source sensing fusion-driven intelligent control method for rapid tunnel excavation, comprising: The multi-source sensing and edge computing layer includes a fiber optic demodulator, a microseismic acquisition station, a laser scanning host, and edge computing nodes. This layer is responsible for the continuous acquisition, spatiotemporal fusion, and real-time calculation of reliability indicators of multi-source physical field data, providing unified and high-precision state input for upper-level decision-making. The fiber optic demodulator is connected in series with the fiber optic strain sensors installed in the boreholes on the roadway roof and both sides. The acquisition frequency is 50Hz and the wavelength demodulation accuracy is not less than 1pm. It is used to continuously acquire tangential strain and axial displacement distribution data of the surrounding rock surface. The microseismic acquisition station is connected to a three-dimensional array of microseismic detectors (sensitivity not less than 100V / g, frequency band 10Hz~2000Hz), with a sampling frequency not less than 2kHz, to acquire microseismic waveform signals generated by rock mass fracturing in real time, and to extract microseismic events (source location, energy, and occurrence time) by triggering threshold detection. The laser scanning host is connected to a 3D laser scanner mounted behind the tunneling machine, with a point cloud density of not less than 10. 4 Points / square meter; automatically acquire point cloud data of the tunnel cross-section after each cutting cycle is completed. The edge computing node is deployed near the tunneling face and has a built-in high-performance embedded processor and large-capacity storage. The edge computing node has a built-in data spatiotemporal integration module and a real-time reliability calculation module, and receives the above three types of sensor data through the underground Ethernet ring network. The data spatiotemporal fusion module uses extended Kalman filtering to perform timestamp alignment (synchronization accuracy better than 1ms) and random noise suppression on multi-source sensor data; then, based on the prior confidence of the sensors (0.4 for fiber optic grating, 0.3 for micro-vibration, and 0.3 for laser), it performs adaptive weighted fusion to generate a multi-dimensional feature matrix under a unified spatiotemporal reference, which is then output to the real-time reliability calculation module.
[0029] The real-time reliability calculation module calculates the surrounding rock stability function in real time based on the disturbance coupling reliability evaluation model described in step 2. and its reliability index within the first and second moment sliding windows It also simultaneously calculates the short-term surge slope of the microseismic energy release rate. The module's output includes real-time reliability indicators, values of each component of the function, and feedforward early warning indicators; The digital twin and model evolution layer (deployed on the edge server) includes a reduced-order digital twin model of the tunnel and an online parameter correction module based on unscented Kalman filtering. The digital twin and model evolution layer is used to invert the surrounding rock stress / damage field online through the reduced-order digital twin model and recursively correct the model parameters using unscented Kalman filtering to achieve adaptive evolution of the system to geological conditions. The reduced-order digital twin model of the tunnel is based on the equivalent continuous medium elastoplastic damage constitutive model. It employs intrinsic orthogonal decomposition (POD) and discrete empirical interpolation (DEIM) methods to compress the degrees of freedom of the full-order finite element model to less than 5% of its original size, with a single solution time not exceeding 30 seconds. The model uses real-time fiber optic strain data as boundary conditions, microseismic event rate distribution as the basis for updating the internal damage field, and laser convergence data as the displacement field verification benchmark. It continuously inverts the spatial distribution of the current stress field and damage field of the surrounding rock and predicts the changing trends of reliability indicators within the next 1-3 tunneling cycles. The prediction results are then used as feedforward components to push to the risk decision-making layer. The online parameter correction module based on unscented Kalman filtering uses the lithology attenuation coefficient. Using the state variable and the observed equivalent safety factor after regulation and inversion as the observation, state equations and observation equations are established, and unscented Kalman filtering is used for recursive updates. The posterior mean and variance are calculated. A correction value is output after each adjustment cycle. When the correction magnitude exceeds 0.05 for three consecutive cycles, automatic recalibration of the digital twin model parameters is triggered, and the corrected values are... The control algorithm is updated synchronously to the risk decision-making layer.
[0030] The risk decision and control command generation layer includes a four-level risk classification module, a feedforward early warning trigger module, a spacing and speed matching control module, and an output module. The risk decision and control command generation layer is used to complete the four-level risk classification based on reliability indicators and their trends, combined with microseismic feedforward early warning, and generate differentiated control commands for anchor spacing and tunneling speed, forming a composite decision of feedback and early warning. The four-level risk classification module is based on reliability indicators. Based on the least-squares linear fitting slope of the rock mass in the most recent three complete tunneling cycles, the surrounding rock condition is classified into four levels: stable, concerning, warning, and dangerous. The feedforward warning trigger module is used to immediately generate a feedforward warning signal when the slope of the microseismic energy release rate (the normalized ratio of the current 5-minute average to the previous 15-minute average) exceeds 1.8. This forces the frequency of control command generation to be increased from once per cycle to twice per cycle, and prioritizes the execution of support reinforcement operations (reducing the spacing between rows and decreasing the advance speed). The spacing and speed matching control module automatically calculates the anchor bolt spacing based on the risk level. and tunneling speed ; The output module is used to generate specific control commands (target spacing, target tunneling speed), and encapsulates them into Modbus TCP or PROFINET messages through industrial protocols, and sends them to the equipment execution layer.
[0031] The equipment execution and visualization layer includes an equipment execution linkage module (used to coordinate the adjustment of the electro-hydraulic proportional control system of the bolt drilling rig and the frequency conversion speed control system of the tunneling machine) and a visualization module. The equipment execution and visualization layer is used to send control commands to the execution equipment such as the tunneling machine and bolt drilling rig through standard industrial protocols, and to visualize the surrounding rock status, risk level and control effect in real time, completing the end execution of the closed-loop control chain and human-machine interaction.
[0032] The electro-hydraulic proportional control system of the anchor bolt drilling rig is used to receive the spacing control command, automatically adjust the anchor bolt spacing and support spacing of the drilling rig, and perform support operations; the frequency conversion speed control system of the tunneling machine is used to receive the tunneling speed command, adjust the cutting head speed and the feed speed of the propulsion cylinder in real time, and realize the adaptive matching of the tunneling speed; as a preferred embodiment, both types of execution systems communicate with the risk decision-making layer through an industrial control network (PROFINET / Modbus TCP), and the command response time is no more than 2 seconds.
[0033] The visualization module is equipped with industrial control touch screens in the tunneling machine's cab and the ground control center, dynamically displaying the current risk level (green / yellow / orange / red corresponding to stable / concern / warning / danger) in augmented reality color bands, while simultaneously updating the reliability indicators in real time. Key parameters such as anchor bolt spacing, tunneling speed, and micro-vibration energy release rate trend curve are displayed. When a feedforward warning or danger level occurs, a red warning box will automatically pop up on the screen and provide suggested operation items ("Stop immediately", "Strengthen support", etc.).
[0034] In this invention, the multi-source sensing and edge computing layer integrates a fiber optic demodulator, a microseismic acquisition station, a laser scanning host, and edge computing nodes. It incorporates a built-in data spatiotemporal fusion and real-time reliability calculation module, facilitating rapid on-site processing and feature extraction of multi-physics data. This significantly reduces communication latency when uploading data to the ground, providing technical support for real-time control. The digital twin and model evolution layer deploys a roadway-reduced digital twin model and an online parameter correction module based on unscented Kalman filtering. This enables the inversion of the surrounding rock stress field and damage field at extremely low computational cost, and dynamically updates key model parameters (such as the lithological attenuation coefficient), giving the system self-learning and evolutionary capabilities, effectively combating geological uncertainties. The risk decision-making and control command generation layer incorporates a four-level risk classification module, a feedforward early warning trigger module, and a spacing and speed matching control module. This automatically generates precise support and advance commands based on changes in real-time reliability indicators, achieving differentiated risk classification and avoiding a one-size-fits-all conservative control approach. The equipment execution and visualization layer is linked with the electro-hydraulic proportional system of the anchor drilling rig and the frequency conversion speed control system of the tunneling machine through the industrial control network, ensuring the reliable issuance and rapid execution of control commands. The augmented reality color bands (green / yellow / orange / red) are displayed on the cockpit screen in real time to show the risk level and key parameters, which greatly improves the operator's situational awareness and emergency response efficiency.
[0035] The system has a simple structure and a high degree of intelligence. It realizes real-time probabilistic evaluation of the stability of the surrounding rock during the tunneling process, risk classification and early warning, autonomous optimization of support and propulsion parameters, and adaptive correction of the model. It has high reliability, high efficiency and strong robustness in the scenario of rapid tunneling in complex coal mine roadways.
Claims
1. A method for intelligent control of rapid tunnel excavation driven by multi-source sensing fusion, characterized in that, Includes the following steps: Step 1: Multi-source sensing and data acquisition, spatiotemporal fusion; Multi-source sensing mechanisms are deployed synchronously within a range of 5m to 30m behind the tunneling face; multi-source physical field data are collected through the multi-source sensing mechanisms, and a multi-dimensional feature matrix with a unified spatiotemporal reference is generated after data preprocessing; Step 2: Reliability assessment of disturbance coupling; Constructing a rock stability function that integrates strength theory, damage mechanics, and rheology This includes the introduction of additional disturbance stress calculated in real time from the cutting parameters of the tunneling machine. The fracture mechanics effect of mechanical rock breaking on the instantaneous stress field of the surrounding rock is characterized; based on the mean and standard deviation of the function within the sliding window, the real-time reliability index is calculated using the first second moment method. The measured variability of each parameter is uniformly quantified into a probabilistic stability criterion; Step 3: Risk classification and composite decision-making; Based on real-time reliability indicators and its least-squares linear fit slope in the most recent three complete tunneling cycles. The surrounding rock condition is divided into four levels: stable, concerning, warning, and dangerous. At the same time, the short-term sudden increase in the slope of the microseismic energy release rate is monitored to trigger feedforward warning. Based on different levels, anchor spacing control instructions and tunneling speed adaptive matching instructions are generated and issued according to standard industrial protocols.
2. The intelligent control method for rapid tunnel excavation driven by multi-source sensing fusion as described in claim 1, characterized in that, In step 1, the process of multi-source sensing, data acquisition, and spatiotemporal fusion is as follows: S11: Install fiber optic strain sensor strings with a center wavelength demodulation accuracy of not less than 1 pm in the tunnel roof and boreholes on both sides, with an adjacent measuring point spacing of 0.5 m; install a three-dimensional array of microseismic detectors with a sensitivity of not less than 100 V / g and a frequency band coverage of 10 Hz to 2000 Hz in the surrounding rock of the tunnel, covering an area 20 m in front of the face and 30 m behind; set up a point cloud density of not less than 10 behind the tunneling machine. 4 A 3D laser scanner with a point-per-square-meter resolution; S12: Based on the fiber optic strain sensor, the tangential strain and axial displacement distribution data of the surrounding rock surface are continuously collected at a sampling frequency of 50Hz and converted into digital signals by the fiber optic demodulator. Based on the continuous acquisition of microseismic waveform signals generated by rock mass fracture using a microseismic detector, with a sampling frequency of not less than 2kHz, microseismic events are detected by triggering thresholds, and the source location, energy and occurrence time are recorded; Based on the 3D laser scanner, the tunnel cross-section point cloud is automatically collected once after each cutting cycle, and high-precision 3D spatial coordinate data of roof subsidence, sidewall movement and cross-section convergence deformation are obtained. All sensor data is transmitted in real time to edge computing nodes deployed near the tunneling face via the underground Ethernet ring network TCP / IP protocol. S13: Extended Kalman filtering is used to align timestamps and suppress random noise in multi-source data, eliminating sensor noise and transmission interference. S14: Adaptive weighted fusion based on sensor prior confidence to generate a multidimensional feature matrix under a unified spatiotemporal reference.
3. The intelligent control method for rapid tunnel excavation driven by multi-source sensing fusion as described in claim 1, characterized in that, In step 2, the reliability assessment process for the disturbance coupling is as follows: S21: Calculate the surrounding rock stability function using the following formula. : ； In the formula, The uniaxial compressive strength of the surrounding rock; This represents the measured maximum tangential stress. This refers to the stress term caused by tunneling disturbance. This represents the rate of damage zone expansion. The critical spread rate; The strain gradient change rate of the top plate; For reference strain rate; and Two dynamic weighted indices; This is the lithological attenuation coefficient; This is the safety threshold constant; S22: Calculate the additional disturbance stress using the following formula. : ； In the formula, To cut off the motor torque; This refers to the cutting head rotation speed; Where is the diameter of the cutting head; F is the propulsion force of the tunneling machine; The cross-sectional area at the heading; , The coefficient is a dimensionless coupling coefficient. S23: Calculate the damage zone expansion rate using the following formula. : ； In the formula, For the first Micro-seismic events at time The source location vector; It is the Euclidean norm; S24: The axial displacement field measured by the fiber Bragg grating sensor is obtained by first calculating the spatial strain and then differentiating it with respect to time. As shown in the following formula: ； In the formula, This is the total length of the sensor array; The average value of the monitoring interval of the roof is taken as the final value. ; S25: Calculate the function within a sliding window of 10 consecutive sampling periods (corresponding to 10 data points). sample mean and sample variance Take the standard deviation The real-time reliability index As shown in the following formula: 。 4. The intelligent control method for rapid tunnel excavation driven by multi-source sensing fusion according to claim 3, characterized in that, In step 3, the risk classification and composite decision-making process is as follows: S31: Define the slope of the short-term surge in energy release rate as: the normalized ratio of the average energy release rate in the current 5 minutes to the average in the previous 15 minutes; when this ratio exceeds 1.8, trigger the feedforward warning signal, forcibly increase the frequency of control command generation from once per cycle to twice per cycle, and prioritize the execution of support enhancement operations; S32: Based on the following process and its trend slope Risk classification: when and When the situation is deemed stable, the spacing between anchor bolts is increased. when or When this occurs, it is classified as a concern level, the current spacing between rows is maintained, and the maximum advance per cycle is limited to 0.8 times the rated value; when Or the slope of the sudden increase in the energy release rate of microseismic events At 1.8, the alert level was set, and the spacing between rows was reduced. when If the situation is deemed dangerous, the spacing between rows should be immediately reduced to 0.8 times the thickness of the top slab layer, and an automatic shutdown should be triggered. S33: Let the current row spacing be... The thickness of the top strata is The rules for adjusting the spacing between anchor bolts are as follows: At a stable level, Calculate using the following formula: ； After magnification ; At the warning level, Calculate using the following formula: ； After shrinking ; When the danger level is reached, Calculate using the following formula: ； S34: Calculate the tunneling speed using the following formula : ； In the formula, This represents the current actual tunneling speed; For nonlinear velocity correction factors, when hour, ,when hour, ,when hour, ,when At that time, the machine had stopped and its speed was zero. The adjusted speed satisfies ; S35: Control commands are sent to the electro-hydraulic proportional control system of the bolt drilling rig and the frequency conversion speed control system of the tunneling machine via Modbus TCP or PROFINET industrial protocols, and are executed automatically.
5. The intelligent control method for rapid tunnel excavation driven by multi-source sensing fusion according to claim 1, characterized in that, It also includes step 4: parameter regulation and closed-loop evolution; After each regulation cycle, the equivalent safety factor observation is retrieved using newly acquired laser point cloud convergence data, and the lithology attenuation coefficient in the function is estimated using recursive Bayesian estimation. Online corrections are performed; meanwhile, the reduced-order digital twin model deployed in the edge computing nodes continuously inverts the stress field and damage field of the surrounding rock, and superimposes the change trend of the reliability index in the next 1 to 3 cycles as a feedforward component into the risk classification decision.
6. The intelligent control method for rapid tunnel excavation driven by multi-source sensing fusion according to claim 5, characterized in that, In step 4, the parameter tuning and closed-loop evolution process is as follows: S41: The iterative nearest-point algorithm is used to register the laser point clouds of two adjacent periods, calculate the roof subsidence rate and the approach rate of the two side walls, and combine the fiber optic strain data to derive the observed value of the equivalent safety factor. ; S42: Lithology attenuation coefficient For the state variables, the state equation and observation equation are established as follows: ； ； In the formula, For process noise, To observe noise; The mapping relationship between the equivalent safety factor and the lithological attenuation coefficient pre-calculated by the digital twin model; Recursive update using unscented Kalman filtering The posterior mean and variance are used to output a correction value after each adjustment cycle. ; When the correction amplitude of three consecutive cycles At that time, the digital twin model parameter calibration is triggered; S43: Deploy a reduced-order digital twin model of the surrounding rock of the roadway within the edge computing node. The digital twin model uses real-time fiber optic strain data as the boundary condition input, microseismic event rate distribution as the basis for updating the internal damage field, and laser convergence data as the displacement field verification benchmark. It continuously inverts the spatial distribution of the current stress field and damage field of the surrounding rock and outputs the predicted trend of reliability index changes in the next 1 to 3 cycles. This trend is then superimposed as a feedforward component into the risk classification decision in step three. S44: The process runs in a loop according to the order of steps 1 to 4, connecting on-site perception, reliability assessment, risk classification, control execution and model correction into a complete closed loop of iterative iteration, so that the surrounding rock stability control during the tunnel excavation process can be dynamically adjusted according to real-time monitoring data to achieve adaptive management.
7. A multi-source sensing fusion-driven intelligent control system for rapid tunnel excavation, used to implement the multi-source sensing fusion-driven intelligent control method for rapid tunnel excavation as described in claims 1 to 6, characterized in that, include: The multi-source sensing and edge computing layer includes an optical fiber demodulator, a microseismic acquisition station, a laser scanning host, and an edge computing node. The edge computing node has a built-in data spatiotemporal integration module and a real-time reliability calculation module. The digital twin and model evolution layer includes a roadway reduced-order digital twin model and an online parameter correction module based on unscented Kalman filtering; The risk decision-making and control instruction generation layer includes a four-level risk classification module, a feedforward early warning trigger module, a spacing and speed matching control module, and an output module. The equipment execution and visualization layer includes the equipment execution linkage module and the visualization module.