A tunnel collapse collaborative treatment system and method based on twin model prediction

Abstract

The application discloses a tunnel collapse collaborative treatment system and method based on a twin model prediction, and belongs to the technical field of underground engineering disaster prevention and treatment. The system comprises a multi-source fusion intelligent perception layer, a digital twin analysis and decision-making layer and a distributed collaborative execution layer. The perception layer collects data in real time; the decision-making layer constructs a dynamic digital twin and runs a model prediction control algorithm, performs simulation prediction and rolling optimization, and generates collaborative control instructions; and each terminal of the execution layer receives the instructions and collaborates. Through the "perception-simulation-prediction-execution" closed loop, the digital twin and advanced process control theory are deeply integrated, the problems of traditional treatment measures, such as dispersion, poor collaboration and decision lag, are solved, advanced simulation, dynamic optimization and adaptive collaborative control of the collapse treatment process are realized, and the safety, efficiency and structural reliability are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of underground engineering construction safety and geological disaster prevention technology, and in particular to a collaborative disposal system and method for tunnel collapse based on twin model prediction. Background Technology

[0002] Tunnel engineering projects often encounter frequent landslides when traversing fault fracture zones characterized by high-intensity disturbance, large deformation, and high water pressure, making response extremely difficult. Existing technologies typically employ a combination of measures such as counter-pressure backfilling, pipe roof support, and grouting consolidation. However, in actual construction, these measures often exhibit a "sequential construction, information silo" approach: the initiation timing and construction parameters (such as grouting pressure and concrete backfilling speed) of each measure rely heavily on engineers' on-site experience and remain fixed or undergo rough adjustments during construction. There is a lack of real-time information exchange and collaborative optimization between different construction processes.

[0003] The fundamental flaw in this model lies in the lack of a "decision-making center" capable of integrating multi-source real-time information, performing dynamic mechanical simulations of complex surrounding rock-support systems, and proactively optimizing parameters of various measures based on simulation predictions. For example, it cannot predict the impact of different pressures on the stability of adjacent counterweight bodies before grouting; nor can it dynamically optimize pumping strategies to prevent structural overload during concrete backfilling based on real-time feedback from surrounding rock deformation. This results in a degree of blindness in the treatment process, delayed response, and difficulty in achieving optimal spatiotemporal and mechanical synergy among various support structures, thus limiting overall safety and efficiency. Summary of the Invention

[0004] The purpose of this invention is to provide a collaborative handling system and method for tunnel collapse based on twin model prediction, aiming to solve the technical problems in the prior art, such as poor coordination of collapse handling measures, reliance on experience for parameter setting, and inability to proactively avoid risks, due to the lack of dynamic simulation and predictive decision-making capabilities.

[0005] To address the aforementioned problems, according to one aspect of this application, an embodiment of the present invention provides a collaborative handling system for tunnel collapse based on twin model prediction, characterized in that it includes: a multi-source fusion intelligent perception layer, a digital twin analysis and decision-making layer, and a distributed collaborative execution layer; The multi-source fusion intelligent sensing layer is deployed at the face of the collapsed section, the cavity, the supported section and the surrounding rock to collect geomechanical data, structural response data and environmental data in real time. The digital twin analysis and decision-making layer is communicatively connected to the perception layer. It internally constructs a dynamic digital twin that is synchronously updated with the physical collapse area and runs a model prediction and control algorithm. The distributed collaborative execution layer includes a slag back pressure execution terminal, a pipe roof support execution terminal, a cavity backfill execution terminal, a lock reinforcement execution terminal, a surrounding rock modification execution terminal, and a bottom reinforcement execution terminal. Each terminal is communicatively connected to the decision-making layer. The digital twin analysis and decision-making layer is configured to: fuse and process real-time data uploaded by the perception layer, drive the evolution of the dynamic digital twin to reflect the real state of the physical world; based on this, using the model predictive control algorithm, with the dynamic digital twin as the predictive model and surrounding rock stability and structural safety as the optimization objectives, continuously calculate the optimal control command sequence for each execution terminal within multiple future construction steps; and send the command sequence to the corresponding execution terminal in real time, thereby dynamically coordinating the start-up timing, operation parameters and linkage logic of each terminal.

[0006] Furthermore, the multi-source fusion intelligent sensing layer includes: The microseismic monitoring array is deployed deep within the surrounding rock of the landslide-affected area to capture microseismic event signals of the initiation and propagation of internal ruptures in the surrounding rock, thereby enabling early warning of disasters.

[0007] The composite sensor network includes a distributed fiber optic strain / pressure sensor array embedded in the backfill and existing support structure, a microelectromechanical system inclinometer and acoustic emission sensor installed at key nodes, for high-density capture of the stress, deformation and vibration state of the structure.

[0008] The three-dimensional laser scanning unit periodically scans the cavity and working face area to acquire dynamic change data of the cavity shape and excavation contour.

[0009] Furthermore, the digital twin analysis and decision-making layer specifically includes: The dynamic digital twin construction and update module, based on building information modeling and geographic information system technology, integrates the initial geological survey model, the support structure design model and the real-time input perception data to construct and continuously correct a computable engineering model containing geometry, attributes and mechanical state, namely the dynamic digital twin.

[0010] The multi-source data fusion and mechanical parameter inversion module adopts a fusion algorithm based on Bayesian theory to couple and analyze multi-source heterogeneous monitoring data such as microseismic, stress, and deformation with the dynamic digital twin, and invert the mechanical parameter field and damage evolution field of the surrounding rock in real time.

[0011] The model predictive control optimization engine uses the updated dynamic digital twin as a model to predict the future behavior of the system, uses the construction parameters of each execution terminal as optimization variables, and takes maintaining the stability of the surrounding rock and the safety of the support structure as the control objectives. It solves a finite-time domain optimization problem in each control cycle and outputs the optimal control command sequence for each execution terminal in the future.

[0012] Furthermore, the structure and control relationship of each terminal in the distributed collaborative execution layer are as follows: The tunnel slag counter-pressure execution terminal includes a mechanical counter-pressure body formed by the accumulation of tunnel slag and at least two concrete conveying pipes pre-embedded on its top. The decision-making layer, based on the three-dimensional stress field and surface displacement data inside the counter-pressure body fed back by the sensing layer, dynamically adjusts its accumulation height, slope angle, and accumulation rate through model prediction and control optimization engine calculations.

[0013] The pipe roof support execution terminal consists of large-diameter pipe roofs laid along the tunnel arch contour and small grouting guide pipes laid along the sidewall contour, together forming the pipe roof advanced support. The decision-making layer optimizes the circumferential spacing of the pipe roofs (preferably 300 mm in the arch area and 400 mm in the sidewall area), longitudinal overlap length (not less than the design value), and initial grouting pressure and grout mix ratio based on the load distribution simulated by the dynamic digital twin and the stress data of the pipe roofs monitored by the sensing layer.

[0014] Cavity backfilling execution terminal: Concrete is poured into the cavity through the pre-embedded concrete delivery conduit 8 to form cavity backfill concrete. The decision layer predictively controls the layer thickness of the concrete pouring (single layer controlled between 800 mm and 0 mm) and the real-time pumping pressure based on the cavity wall pressure distribution cloud map calculated by the dynamic digital twin and combined with the microseismic activity monitored by the sensing layer, to prevent secondary damage caused by excessive pressure.

[0015] Locking-in reinforcement execution terminal: Located behind the collapsed section, the already supported section is reinforced with grouting using small guide pipes with an outward insertion angle of 15 degrees, forming a locking-in reinforced support. The decision-making layer assesses the stability margin in real time based on the deformation synergy index and acoustic emission energy accumulation value of the locking area provided by the perception layer. This assessment result serves as a key logical permission condition, allowing the decision-making layer to determine whether to authorize subsequent high-risk operations such as pipe roof support execution terminals.

[0016] The surrounding rock modification execution terminal includes a deep-hole grouting group arranged in a spatial quincunx pattern within the surrounding rock of the collapse-affected area, i.e., uncovered re-grouting holes. Specifically, the grouting hole spacing in the top arch area is 1.25m × 1.25m, and in the sidewall area it is 2.0m × 2.0m, with a hole depth of 5m. The decision-making layer, based on the surrounding rock permeability coefficient field and damage field obtained by the data fusion inversion module, simulates the grout diffusion process in a dynamic digital twin, and then dynamically adjusts the grout water-cement ratio, gel time, and grouting pressure of the grouting holes in different zones to achieve differentiated and precise surrounding rock modification.

[0017] Bottom-reinforced execution terminal: This includes an I-beam support laterally connected to the bottom foot of the tunnel steel arch frame, forming a bottom transverse steel support. The decision-making layer, based on the bottom drum development trend predicted by the dynamic digital twin and combined with the stress and displacement data of the steel arch frame foot monitored in real time by the perception layer, uses a model predictive control optimization engine to determine the optimal installation time of this terminal, the preload value of the transverse support, and the stiffness requirements of connection node 9.

[0018] Compared with the prior art, the tunnel collapse collaborative handling system and method based on twin model prediction of the present invention has at least the following beneficial effects: It has achieved a shift in decision-making paradigm from "experience-driven" to "model and data hybrid-driven": by introducing dynamic digital twins and model predictive control algorithms, the system has the ability to simulate, evaluate and optimize construction plans in advance in virtual space, so that the decision-making process changes from relying on ex post experience to relying on ex pre simulation prediction, which significantly improves the scientific nature and foresight of the handling.

[0019] A deep synergy between various support measures in terms of time, space, and mechanics was achieved: the system no longer treats each measure as an independent process, but instead uses a global optimization algorithm to uniformly schedule them as interrelated and mutually restrictive control variables. This ensures that actions such as counterpressure, support, grouting, backfilling, and reinforcement are seamlessly connected in time, precisely matched in space, and mutually reinforced in mechanics, forming a synergistic effect.

[0020] A high-barrier interdisciplinary technology integration solution has been constructed: This invention deeply integrates multiple cutting-edge technologies such as geotechnical engineering, sensor technology, multi-source data fusion, digital twin simulation and advanced process control (model predictive control).

[0021] The system ensures the reliability of both the treatment process and the outcome: its adaptive closed-loop control mechanism dynamically adjusts the strategy based on the real-time response of the surrounding rock and structure, ensuring that the treatment process remains safe and controllable. Simultaneously, through full-cycle data-driven and model-verified processes, the resulting composite support structure exhibits a clear stress state, strong integrity, and fundamentally guaranteed long-term reliability.

[0022] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a schematic diagram illustrating the overall architecture and workflow of the adaptive collaborative processing system described in an embodiment of the present invention; Figure 2 This is a schematic diagram of a typical deployment of the multi-source fusion intelligent sensing layer on the tunnel cross section in an embodiment of the present invention; Figure 3 This is a schematic diagram showing the position and collaborative relationship of each terminal in the distributed collaborative execution layer in the longitudinal section of the tunnel in an embodiment of the present invention.

[0025] Explanation of reference numerals in the attached figures: 1. Mechanical counter-pressure body; 2. Advanced support for pipe roof; 3. Backfill the cavity with concrete; 4. Reinforce the lock opening; 5. Permeable steel pipe; 6. Bottom horizontal steel support; 7. Steel arch frame; 8. Concrete conveying conduit; 9. Connecting ribs; 10. Uncovered grouting holes. Detailed Implementation

[0026] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the specific embodiments, structures, features, and effects according to the present invention will be described in detail below with reference to the accompanying drawings and preferred embodiments. In the following description, different "embodiments" or "embodiments" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0027] In the description of this invention, it should be clearly stated that the terms "first," "second," etc., in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence; the terms "vertical," "lateral," "longitudinal," "front," "rear," "left," "right," "up," "down," "horizontal," etc., indicate orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, and are merely for the convenience of describing this invention, and do not mean that the device or element referred to must have a specific orientation or position, and therefore should not be construed as a limitation of this invention.

[0028] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0029] like Figures 1-3 As shown, this embodiment of the invention provides a collaborative handling system for tunnel collapse based on twin model prediction, including: a multi-source fusion intelligent perception layer, a digital twin analysis and decision-making layer, and a distributed collaborative execution layer; The multi-source fusion intelligent sensing layer is deployed in the collapse area to collect real-time monitoring data of the surrounding rock and structure; The digital twin analysis and decision-making layer is communicatively connected to the multi-source fusion intelligent perception layer, and a dynamic digital twin is constructed inside it and a model prediction and control algorithm is run. The distributed collaborative execution layer includes a slag back pressure execution terminal, a pipe roof support execution terminal, a cavity backfill execution terminal, a lock reinforcement execution terminal, a surrounding rock modification execution terminal, and a bottom reinforcement execution terminal. Each terminal is communicatively connected to the digital twin analysis and decision-making layer. The digital twin analysis and decision-making layer is configured to: integrate the monitoring data to drive the dynamic digital twin update, and use the model predictive control algorithm to continuously calculate the optimal control command sequence for each execution terminal in order to coordinate the construction of each terminal.

[0030] In this embodiment, the system integrates discrete construction units into an organic whole by constructing a three-tiered intelligent architecture. The multi-source fusion intelligent sensing layer, acting as the system's "nerve endings," is widely deployed in the landslide-affected area, responsible for collecting multi-dimensional data such as surrounding rock deformation, structural stress, and microseismic activity. This real-time data is transmitted to the digital twin analysis and decision-making layer, which is the system's "intelligent hub." Internally, it first constructs a dynamic digital twin that strictly corresponds to the physical landslide area. This model can fuse sensing data to achieve synchronous evolution. More importantly, this layer runs a model predictive control algorithm. This algorithm uses the dynamic digital twin as the predictive model, and within each control cycle, it continuously simulates multiple future construction steps in digital space and solves the optimal control problem, thereby generating a series of forward-looking optimization instructions. Finally, the distributed collaborative execution layer, acting as the system's "executive limbs," receives these instructions and drives the coordinated actions of six terminals: muck backpressure, pipe roof support, cavity backfilling, interlock reinforcement, surrounding rock modification, and bottom reinforcement. The execution effects of each terminal are captured by the sensing layer, forming continuous feedback. The entire process transforms the traditional static construction that relies on experience into an intelligent adaptive process driven by data and models, capable of predictive planning and dynamic adjustment. This achieves deep synergy among various support measures in terms of time, space, and mechanical parameters, fundamentally improving the efficiency and safety of the treatment.

[0031] Lock stability evaluation model: The "lock stability index" Slock is calculated using the following formula: Where: θi is the tilt angle change monitored by each inclinometer; EAE is the cumulative acoustic emission energy; σmax is the maximum stress in the interlocking area; α, β, γ are weighting coefficients calibrated according to geological conditions and support type.

[0032] When Slock < 1.0, the system automatically locks subsequent high-risk procedures; when Slock ≥ 1.5, it automatically unlocks and allows construction to proceed.

[0033] Multi-terminal collaborative scheduling logic: The system adopts a priority-based dynamic scheduling algorithm. In the control instruction sequence generated under the MPC framework, a time-series dependency relationship and resource conflict detection mechanism are embedded to ensure that each terminal works collaboratively in time, space and mechanics, and avoid mutual interference.

[0034] In some embodiments, the multi-source fusion intelligent sensing layer includes a microseismic monitoring array deployed deep within the surrounding rock, a composite sensing network deployed on the surface of the structure, and a three-dimensional laser scanning unit for scanning cavities.

[0035] In this embodiment, the sensing layer is the foundation for achieving precise sensing, and it consists of three types of core monitoring units. A microseismic monitoring array deployed deep within the surrounding rock can capture elastic wave signals generated by rock fracturing. By analyzing the spatiotemporal evolution and energy characteristics of microseismic events, it can proactively identify the initiation and expansion trends of damage zones within the surrounding rock, achieving early warning of disasters. A composite sensing network installed in the existing support steel arch frame 7, the constructed pipe roof, and the interlocking area mainly includes distributed fiber optic strain sensors and MEMS tilt sensors. The fiber optic sensors can continuously measure the strain distribution of the structure, accurately reflecting its stress state; the tilt sensors monitor the overall or local rotational changes of the structure. The combination of these two achieves millimeter-level, high-density sensing of the health status of the support system. Furthermore, a three-dimensional laser scanning unit deployed inside the tunnel periodically scans the tunnel face and cavities, rapidly generating a high-precision three-dimensional point cloud model. This accurately quantifies the spatiotemporal changes in the cavity's morphology, volume, and contour, providing direct evidence for assessing the scale of the collapse and the amount of backfilling work. This multi-source heterogeneous data acquisition system, which integrates air and ground and macro and micro data, together forms a comprehensive, three-dimensional sensing network.

[0036] In some implementations, the digital twin analysis and decision-making layer includes: a dynamic digital twin construction and update module, a multi-source data fusion and mechanical parameter inversion module, and a model prediction control optimization engine.

[0037] In this embodiment, the decision-making layer operates as a multi-module collaborative data processing and optimization process. The dynamic digital twin construction and update module first connects the initial geological model, design drawings, and real-time sensing data to create a high-fidelity virtual tunnel model integrating geometric, physical, and mechanical properties. This digital twin is not static; it dynamically updates as construction progresses and sensing data is continuously input. The multi-source data fusion and mechanical parameter inversion module is responsible for processing the influx of heterogeneous data. It employs advanced algorithms such as Kalman filtering and Bayesian inversion to fuse and analyze multi-source information such as microseismic event location, structural strain, and surface displacement. It then reverse-calculates key mechanical parameter fields such as the real-time elastic modulus, cohesion, and damage factor of the surrounding rock, and corrects the material properties in the digital twin in real time to make them infinitely close to the real state of the physical world. The model predictive control optimization engine is the final decision-maker. It uses the currently updated high-fidelity digital twin as the prediction model, with the surrounding rock displacement and support stress not exceeding the safety threshold as control objectives, and the adjustable parameters of each execution terminal (such as grouting pressure and pumping speed) as control variables. In each decision cycle, the engine simulates all possible combinations of control strategies over a future period (prediction time domain) in virtual space. By solving a constrained optimization problem, it continuously outputs a sequence of control commands that optimizes the future state of the overall system. This achieves a fundamental shift from "monitoring-response" to "simulation-prediction-optimization".

[0038] The construction and updating of the dynamic digital twin employs a multi-source heterogeneous data fusion mechanism and parameter inversion-model coupling technology. Specifically, it includes: Data fusion mechanism: A fusion algorithm based on Kalman filtering and Bayesian update is adopted to perform spatiotemporal alignment and uncertainty quantification of multi-source information such as microseismic event location data, fiber strain distribution, and surface displacement monitoring, so as to achieve high confidence reconstruction of the stress field, displacement field and damage field of the surrounding rock.

[0039] Mechanical model selection: Based on the type of surrounding rock and the construction stage, a finite element-discrete element coupled model (FEM-DEM) or an elastoplastic-damage mechanical model is selected to realize multi-physics field coupled simulation of progressive failure of surrounding rock, interaction of support structure, and grout diffusion in a digital twin.

[0040] Dynamic update mechanism: Each time a batch of sensing data is received, the system automatically triggers the parameter inversion process, and corrects the mechanical parameters (such as elastic modulus, cohesion, and permeability coefficient) in the model through sensitivity analysis and optimization algorithms (such as genetic algorithm and gradient descent method) to ensure that the error between the twin state and the physical entity is controlled within 5%.

[0041] In some embodiments, the tunnel muck counter-pressure execution terminal includes a mechanical counter-pressure body 1 formed by the tunnel muck and a pre-embedded concrete conveying conduit 8; the digital twin analysis and decision-making layer dynamically adjusts its accumulation parameters based on the sensing data of the mechanical counter-pressure body 1 through the model prediction control algorithm.

[0042] In this embodiment, the tunnel muck counterpressure execution terminal utilizes excavated muck to form a mechanical counterpressure body 1, but its construction parameters are dynamically determined by the decision-making layer. A network of earth pressure sensors embedded within the mechanical counterpressure body 1 transmits stress values ​​from different locations in real time. The decision-making layer's model predictive control optimization engine analyzes this data and, combined with a digital twin simulation of the evolution of the plastic zone at the tunnel face, issues control commands after comprehensive calculation. For example, when the system detects that the pressure on the left side of the mechanical counterpressure body 1 is significantly lower than that on the right, it may instruct an increase in the left-side filling height to balance the bias; when it predicts an acceleration in the extrusion rate at the tunnel face, it may instruct an acceleration of the overall filling speed. Simultaneously, a concrete delivery conduit 8, pre-embedded at the top of the filling body and leading to the cavity, pre-sets a precise path for subsequent grouting. This dynamic parameter adjustment based on real-time feedback and predictive models allows the mechanical counterpressure body 1 to form at the optimal shape and rate, quickly stabilizing the tunnel face while avoiding material waste or local instability.

[0043] The model predictive control optimization engine employs a nonlinear model predictive control framework. Its core is to solve a finite-time domain optimization problem using an updated dynamic digital twin as the predictive model in each control cycle (e.g., every 10 minutes). The optimization objective is to minimize the deviation between the system's critical states (e.g., surrounding rock displacement, support stress) and the safety target over a future period (prediction time domain, typically 4–8 construction steps), while also ensuring the smoothness of changes in control variables (e.g., grouting pressure, pumping speed). The solution process must satisfy multiple constraints, including equipment capacity (e.g., grouting pressure ≤ 3 MPa), construction specifications (e.g., backfill layer thickness ≤ 1.0 m), and safety thresholds (e.g., surrounding rock displacement ≤ 30 mm). Algorithms such as sequential quadratic programming or interior point methods are used to continuously output the optimal control command sequence for each execution terminal over the next 30–60 minutes.

[0044] The constraints include: Equipment capacity limitations (e.g., grouting pressure ≤3MPa); Construction specifications may restrict (e.g., backfill layer thickness ≤ 1.0m); Safety threshold (e.g., surrounding rock displacement ≤ 30mm).

[0045] The solution algorithm uses Sequential Quadratic Programming (SQP) or interior point method, and performs a rolling solution every 10 minutes to output the optimal control sequence for each execution terminal in the next 30 to 60 minutes.

[0046] In some embodiments, the pipe roof support execution terminal includes a pipe roof advance support 2 consisting of a large-diameter pipe roof laid along the tunnel outline and a grouting small pipe; the digital twin analysis and decision layer optimizes the circumferential spacing, overlap length and grouting parameters of the pipe roof advance support 2 based on the load distribution simulated by the dynamic digital twin.

[0047] In this embodiment, the bottom-strengthened execution terminal is responsible for constructing the core advanced protection structure—the pipe roof advanced support 2. Its design and construction process have been optimized throughout. Before construction, the decision-making level optimizes the pipe roof layout parameters based on the three-dimensional distribution of surrounding rock loosening pressure analyzed by the dynamic digital twin. For example, in the high-stress zone of the arch crown, the system specifies a close-packed scheme with a circumferential spacing of 300 mm; on the sidewalls, a spacing of 400 mm is used. During the grouting process, the system does not use constant pressure, but dynamically adjusts the grouting pressure and grout viscosity based on the initial strain fed back by fiber optic sensors on the pipe roof. If the strain of a certain section of the pipe roof does not meet expectations after grouting, the system automatically increases the subsequent grouting pressure in that area to ensure the formation of a complete shell. This load-guided, feedback-corrected, and refined construction ensures that the pipe roof structure is in its optimal load-bearing state from the outset.

[0048] In some embodiments, the cavity backfilling execution terminal injects concrete through the concrete delivery conduit 8 to form cavity backfilling concrete 3; the digital twin analysis and decision layer controls the layer thickness and pumping pressure of the injection based on the cavity wall pressure calculated by the dynamic digital twin and the sensed microseismic activity.

[0049] In this embodiment, the cavity backfilling execution terminal pumps concrete through the concrete delivery conduit 8 to fill and form the cavity backfill concrete 3. Its core control logic is to prevent overpressure and secondary disturbances. The decision-making layer monitors two key indicators in real time: one is the theoretical pressure value at each point on the cavity wall calculated by the digital twin, and the other is the acoustic emission activity rate of the surrounding rock captured by the microseismic monitoring array. When concrete is pumped, if the system predicts that the wall pressure at a certain point will exceed the safety threshold, or detects an abnormal increase in the frequency of microseismic events in that area, it will immediately reduce the pumping speed or suspend grouting, and may trigger a local reinforcement grouting command. Simultaneously, the system strictly controls the single-layer pouring thickness between 800 and 0 mm, implementing layered and windowed operations. This safety closed loop of pressure prediction and rock mass response dual monitoring minimizes the risk of new collapses caused by backfilling construction.

[0050] In some implementations, the lock reinforcement execution terminal uses small-diameter pipes for grouting to form lock reinforcement support 4; the digital twin analysis and decision layer uses logic to control whether to authorize the construction of subsequent terminals based on the real-time assessment results of the stability of the area where the lock reinforcement support 4 is located.

[0051] In this embodiment, the reinforced support 4 implemented by the lock reinforcement execution terminal is not only a physical barrier but also a safety switch embedded in the system process. Tilt sensors and acoustic emission probes deployed in this area continuously monitor its deformation coordination and energy release. The multi-source data fusion module of the decision-making layer integrates this information and calculates a "lock stability index" in real time. This index is written into the system's central control logic. Only when the index consistently exceeds the set safety line will the system automatically unlock, allowing construction instructions to be sent to the pipe roof support terminal and other components ahead. If the index abnormally drops, all forward-moving procedures will be automatically locked, and an alarm will be issued. This is equivalent to installing an intelligent, mandatory "safety valve" for the entire construction process, eliminating the possibility of recklessly advancing under unstable conditions.

[0052] In some embodiments, the surrounding rock modification execution terminal includes uncovered grouting holes 10 arranged in a quincunx pattern; the digital twin analysis and decision layer simulates grout diffusion in the dynamic digital twin based on the inverted surrounding rock parameter field, and adjusts the construction parameters of each grouting hole in a differentiated manner.

[0053] In this embodiment, the surrounding rock modification execution terminal reinforces the surrounding rock through the uncovered re-grouting hole 10, employing a "one-site-one-policy" strategy. The decision-making layer, based on the inverted surrounding rock permeability coefficient zoning map and damage cloud map, pre-simulates the grout diffusion process in each type of rock mass within a digital twin. Based on the simulation results, the system generates differentiated grouting prescriptions. For example, for loose, high-permeability zones, a "high-flow-rate, low-pressure, thin grout" permeation grouting command is issued; for low-permeability fractured rock masses, a "high-pressure, slow-injection, thick grout" fracturing grouting command is issued. During the grouting process, the system dynamically determines the grouting endpoint based on the real-time flow-pressure curve and may adjust the grout mix ratio. This precise grouting based on geological transparency greatly improves the effective diffusion radius of the grout and the uniformity of the consolidated body, maximizing material efficiency.

[0054] In some embodiments, the bottom reinforcement execution terminal includes a bottom transverse steel support 6 connected to the bottom foot of the steel arch frame 7; the digital twin analysis and decision layer determines the installation timing and preload of the bottom transverse steel support 6 based on the predicted bottom bulge trend and real-time stress data.

[0055] In this embodiment, a bottom-strengthened execution terminal is installed with a bottom transverse steel support 6 to close the support ring. Its innovation lies in transforming passive support into active pre-tensioning. The decision-making layer continuously analyzes the bottom heave trend predicted by the digital twin and the real-time data from the stress sensors at the foot of the steel arch frame 7. When the system predicts an accelerated risk of bottom heave deformation or detects an unfavorable stress distribution at the foot, it immediately calculates the optimal pre-tensioning force required and instructs the intelligent tensioning device to precisely apply this pre-tensioning force during support installation. This allows the bottom transverse steel support 6 to participate in the operation before significant heave in the surrounding rock, actively improving the stress state of the support ring. Simultaneously, the system proposes design or construction requirements for the node stiffness of the connecting ribs 9 to ensure effective transmission and long-term maintenance of the pre-tensioning force. This prediction-based active intervention significantly enhances the overall stiffness and anti-bottom heave capability of the support ring.

[0056] In some embodiments, an intelligent drainage execution terminal is also included; the intelligent drainage execution terminal includes a permeable steel pipe 5 pre-embedded at the bottom of the mechanical counterpressure body 1 formed by the tunnel muck counterpressure execution terminal, a geotextile filter layer wrapped around the permeable steel pipe 5, and a variable frequency high-pressure drainage device connected to the permeable steel pipe 5; the intelligent drainage execution terminal is communicatively connected to the digital twin analysis and decision-making layer; the digital twin analysis and decision-making layer dynamically controls the start-up and shutdown and drainage power of the variable frequency high-pressure drainage device based on the pore water pressure sensor data set in the tunnel face and surrounding rock.

[0057] In some preferred embodiments, to enhance the system's instantaneous response capability to drastic changes in local operating conditions, an edge intelligent control node is added between the distributed collaborative execution layer and the multi-source fusion intelligent sensing layer. This edge intelligent control node is deployed within the tunnel near the working face and connected to the intelligent sensing execution unit and execution terminal controller of its assigned area via a local network. The node is loaded with a lightweight artificial intelligence inference model and preset control strategies trained and distributed by the cloud-based digital twin analysis and decision-making layer. It is configured to: perform millisecond-level identification and judgment of sudden abnormal data reported by sensors in its assigned area (such as a sudden increase in stress or a sudden change in seepage); if judged as an emergency risk, it will directly drive the corresponding execution terminal (such as controlling the bottom support hydraulic cylinder to apply compensating preload or adjusting local grouting parameters) for rapid intervention without relying on cloud commands, achieving initial stabilization; simultaneously, it will upload the intervention event and data snapshot to the cloud-based decision-making layer for updating the global model and calibrating subsequent optimization strategies. This hybrid architecture, which combines "rapid edge autonomy" with "global optimization in the cloud," significantly enhances the overall robustness and security of the system when facing extreme conditions such as strong uncertainty in fault zones and rapid development of local risks.

[0058] A method for collaborative handling of tunnel collapses, using the system described in any of the preceding claims, the method comprising: Real-time monitoring is established through the multi-source fusion intelligent sensing layer; The digital twin analysis and decision-making layer updates the dynamic digital twin based on monitoring data and uses model predictive control algorithms to generate a sequence of collaborative control commands in a rolling manner. Each terminal in the distributed collaborative execution layer executes the instruction sequence, and the execution effect data is fed back through the multi-source fusion intelligent perception layer to form a closed-loop control until the processing is completed.

[0059] In this embodiment, the method embodies the application process of the aforementioned system, which is a continuous intelligent optimization cycle. It begins by deploying a monitoring network using a multi-source fusion intelligent sensing layer to achieve all-weather perception of the landslide area's condition. Subsequently, the digital twin analysis and decision-making layer begins its core work: it continuously integrates new data, driving dynamic updates of the digital twin to match reality; and utilizes model predictive control algorithms to perform rolling optimization in virtual space, generating the optimal sequence of collaborative instructions for each execution terminal in future time periods. Then, each terminal in the distributed collaborative execution layer precisely executes these instructions, completing a series of physical operations from backpressure to grouting. Each terminal's action changes the on-site state, and this change is immediately captured by the sensing layer, forming a new data stream fed back to the decision-making layer, thus initiating the next "perception-simulation-optimization-execution" cycle. This method transforms landslide handling from a fixed-sequence construction project into an autonomous adaptive process driven by an intelligent closed loop, capable of dynamically adjusting the path based on real-time feedback and continuously approaching the optimal goal, achieving a comprehensive improvement in safety, economy, and efficiency.

[0060] The system also includes a trusted interaction and evidence storage module. Based on blockchain technology, this module encrypts and stores key decision-making node information, including instructions received by each execution terminal, execution feedback data, and instruction sequences generated by the decision-making layer. This ensures the immutability and traceability of critical data during the handling process. In engineering security operations with extremely high requirements for cross-departmental collaboration and accountability, it solves the problem of potential single-point tampering or repudiation of data in traditional centralized systems. The distributed ledger and encrypted hashing characteristics of blockchain ensure the immutability and auditability of information throughout the entire chain from decision-making and instruction issuance to execution feedback. This provides technically reliable "ironclad evidence" for accident analysis, liability determination, and insurance claims—a security and trust dimension lacking in traditional control systems.

[0061] The composite sensor network integrates a miniaturized atomic magnetometer based on quantum sensing principles. This magnetometer provides highly sensitive monitoring of the magnetoelastic effect caused by the slight stress redistribution within the support structure, even in environments with strong electromagnetic interference within tunnels. The operation of large electromechanical equipment within tunnels generates strong electromagnetic interference, severely degrading the signal-to-noise ratio of traditional electrical sensors. The atomic magnetometer, based on quantum hyperfine energy level transitions, exhibits extremely high sensitivity to magnetoelastic effects (stress-induced changes in the magnetic properties of materials) and strong resistance to electromagnetic interference. It can achieve "quiet" monitoring of early stress concentration zones within the support structure in harsh environments, providing early warning of microscopic damage that might be missed by traditional sensors.

[0062] The dynamic digital twin construction and update module employs digital thread technology to establish a time-stamped logical link for every change operation, data inflow, and decision command throughout the entire lifecycle of landslide response, enabling end-to-end traceability and version management from perceived data to execution feedback. This solves the problems of discrete data and command flows and difficulty in post-event review during complex response processes. Digital threads establish a "digital footprint" with strict temporal and causal relationships for each data update, model correction, command generation, and action execution, making the entire response process like a "digital movie" that can be played back and analyzed frame by frame, greatly improving process transparency, analyzability, and the convenience of subsequent optimization and learning.

[0063] The caving muck of the mechanical counterweight body 1 is premixed with microcapsule-encapsulated phase change material (PCM). The digital twin analysis and decision-making layer, based on perceived ambient temperature and internal temperature data of the counterweight body, simulates the impact of the phase change process on the overall stiffness of the muck pile within the dynamic digital twin, and optimizes control commands accordingly. The mechanical parameters of the caving muck counterweight body, such as internal friction angle and cohesion, may change due to the influence of ambient temperature. After incorporating PCM, the PCM absorbs or releases a large amount of latent heat near the phase change temperature, which can buffer the impact of temperature fluctuations on the overall mechanical properties of the muck pile. By simulating this process, the digital twin can more accurately predict the behavior of the counterweight body and actively adjust construction parameters to compensate for or utilize temperature effects, thus improving robustness to environmental disturbances.

[0064] The grouting conduit is integrally formed using additive manufacturing technology. Its inner wall is pre-installed with a micro-sized slurry vortex generator at a specific angle to the conduit axis. The digital twin analysis and decision-making layer optimizes the selection and combination of conduits with different vortex angles based on the inverted characteristics of the surrounding rock fracture network, thereby directionally enhancing the slurry's diffusion capability in specific fracture directions. Traditional grouting conduits have smooth inner walls, resulting in a predominantly plunger flow of slurry with poor controllability of diffusion direction. By using an integrally formed conduit with a built-in micro-vortex generator at a specific angle through additive manufacturing, controllable rotating or secondary flows can be induced in the slurry during grouting, thus physically guiding the slurry to preferentially penetrate in a specific direction (such as perpendicular to the main fracture). Combining the selection and combination of conduits with a geological model achieves "active guidance" of the slurry diffusion path, improving the accuracy and efficiency of grouting.

[0065] The cavity backfill concrete 3 is a self-healing concrete with internally incorporated microbial mineralizer. The digital twin analysis and decision-making layer simulates the microbial-induced calcium carbonate precipitation repair process after concrete cracking in the long-term prediction model, and incorporates it as an optimization factor affecting long-term stability into the control objective. Conventional concrete backfill may develop microcracks under long-term loads. After incorporating specific microorganisms (such as Bacillus pasteurellii) and their nutrients, when water penetrates the cracks, the microorganisms are activated and metabolize to produce calcium carbonate precipitation, automatically sealing the cracks. The digital twin considers this self-healing mechanism in the long-term prediction model, extending the optimization objective of the treatment from "state at construction completion" to "long-term service performance," making the decision based not only on immediate safety but also on long-term reliability, reflecting a higher level of intelligence.

[0066] The logic control is implemented through a hardware programmable logic array. This array directly receives out-of-range signals from key sensors in the lock area. It can lock the enable signal of the subsequent terminal drive circuit within milliseconds, providing the highest priority mandatory safety interlock, without relying on upper-level decision software cycles. Software-level logic control is susceptible to failure due to system latency, crashes, or network interruptions. By employing hardware PLA / FPGA to implement the core safety interlock logic, the most critical safety judgments (such as lock instability signals) are directly connected to the execution terminal enable signal through hardware circuitry. This achieves an extreme response speed in the nanosecond to microsecond range and absolute reliability independent of software / network failures, providing a fundamental, hardware-based security "moat" for the entire system.

[0067] The surrounding rock modification execution terminal also includes a microbial grouting subsystem for injecting a bacterial solution containing Bacillus pasteurellii and nutrient solution into specific zones. The digital twin analysis and decision-making layer, based on the inverted surrounding rock permeability and pH field, couples and simulates the microbial-induced calcium carbonate precipitation process within the twin, and collaboratively optimizes the construction sequence and parameters of chemical grouting and microbial grouting. Chemical grouting has limited effectiveness in some low-permeability or chemically sensitive formations. Microbial-induced calcium carbonate precipitation (MICP) technology utilizes natural biological processes to generate mineral cement, offering advantages such as environmental friendliness and particle-level sealing. By synergistically combining MIP and chemical grouting, with the digital twin simulating their interaction, strategies such as "biological first, chemical later" or "mixed injection" can be formulated for different formation characteristics (e.g., permeability, pH), forming a composite reinforcement system and broadening the applicable boundaries and technical dimensions of surrounding rock modification.

[0068] At least a portion of the bottom transverse steel support 6 is made of shape memory alloy. The digital twin analysis and decision-making layer, based on predicted temperature variations and stress states, decides whether and when to apply specific temperature excitation to the shape memory alloy component to actively adjust its length and thus change the preload on the steel arch 7. Chemical grouting has limited effectiveness in some low-permeability or chemically sensitive formations. Microbial-induced calcium carbonate precipitation (MICP) technology utilizes natural biological processes to generate mineral cement, offering advantages such as environmental friendliness and particle-level sealing. By synergizing MIP with chemical grouting, and using a digital twin to simulate their interaction, strategies such as "biological first, chemical later" or "mixed injection" can be developed for different formation characteristics (e.g., permeability, pH), forming a composite reinforcement system and broadening the applicable boundaries and technical dimensions of surrounding rock modification.

[0069] The method also includes: recording and associating all data and events throughout the entire process using digital thread technology; and utilizing blockchain for distributed evidence storage of key decision instructions, execution confirmations, and abnormal events. This method not only controls construction at the physical level but also constructs a complete and reliable process archive at the information level. Digital threads ensure process traceability and analyzability, while blockchain ensures tamper-proof and trustworthy key records. Together, these constitute the methodological foundation supporting intelligent construction, digital delivery, and reliable operation and maintenance, enhancing the method's completeness and commercial application value.

[0070] In terms of data trustworthiness and process traceability, the system integrates a blockchain-based trusted interaction and evidence storage module. This module acts as a distributed, tamper-proof "engineering black box," automatically performing encrypted hashing and on-chain evidence storage on every key control command issued by the digital twin analysis and decision-making layer, confirmation feedback from each execution terminal, and all anomaly events marked during the handling process. This ensures clear responsibilities and auditable processes for all key operations in complex multi-party collaborative work, providing a solid technical trust foundation for engineering safety management and quality traceability. Simultaneously, digital thread technology is applied to the entire lifecycle management of the digital twin. From initial model construction to each data-driven update, each simulation prediction, and decision optimization, rigorous time stamps and logical connections are established through digital threads. This not only makes the handling process completely transparent and traceable but also provides a structured, high-quality data source for subsequent evaluation of handling effects and iterative optimization of model algorithms.

[0071] In terms of enhancing sensing capabilities, to address the interference of the complex electromagnetic environment within tunnels on precision measurements, a miniaturized atomic magnetometer based on quantum sensing principles has been introduced into the composite sensor network. This sensor is extremely sensitive to the weak magnetoelastic effect caused by stress changes in the supporting steel structure and is naturally resistant to electromagnetic interference. It can achieve "quiet" monitoring of early, microscopic stress concentration phenomena that are difficult for traditional electrical sensors to detect, even in the context of large-scale equipment operation, significantly improving early warning capabilities.

[0072] At the level of intelligent implementation of terminal materials and structures, this invention integrates multiple intelligent materials and advanced manufacturing technologies. For example, premixing microcapsule phase change materials in the backfill of the slag muck gives the accumulation a certain degree of temperature self-adaptation performance; using additive manufacturing technology to integrally form grouting conduits with embedded micro-vortex generators achieves physical guidance of grout diffusion direction; using self-healing concrete with internally incorporated microbial mineralizers for cavity backfilling gives the structure long-term self-repair potential; applying shape memory alloy components in the bottom transverse steel supports enables the support rings to actively adjust preload during service life; in addition, a microbial grouting subsystem is introduced, working in conjunction with chemical grouting to form a bio-chemical composite surrounding rock modification system. The digital twin, by coupling and simulating the behavior of these intelligent materials (such as phase change, self-healing, shape memory effect, and microbial mineralization), allows the model predictive control algorithm to run and optimize in a richer, more realistic virtual space, thereby making more forward-looking and adaptive decisions.

[0073] At the underlying reliability level of security control, to address potential delays or failures in the software system, the logic control authority for lock stability is assigned to an independent security circuit based on a hardware programmable logic array. This circuit directly receives hard-wired signals from key sensors, enabling millisecond-level forced security interlocking in extreme cases, independent of upper-layer software, providing the highest level of security redundancy for the entire system.

[0074] These technological features from various fields such as information technology, quantum technology, intelligent manufacturing, materials science, biotechnology, and digital circuits are deeply integrated and collaboratively optimized through the "intelligent hub" of digital twin models and model predictive control frameworks. Together, they expand the system's perception dimension, execution capability, reliability level, and security boundaries, solving deep-seated problems that traditional geotechnical engineering treatment methods cannot solve or have not considered (such as long-term performance adaptation, reliable process traceability, extreme environment monitoring, and underlying hardware security).

[0075] Quantum sensing and data fusion mechanism: The miniaturized atomic magnetometer, based on the principle of optically pumped magnetic resonance, detects subtle changes in the magnetoelastic effect caused by stress in the support structure (sensitivity up to 1 pT / √Hz), enabling early identification of stress concentration areas even under strong electromagnetic interference. Its output data, after wavelet denoising and feature extraction, is fused with fiber optic strain data to improve the confidence level of damage warning.

[0076] Self-healing concrete performance prediction model: In the long-term prediction module of the digital twin, a coupled model of microbial activity-crack width-calcium carbonate precipitation rate is embedded to simulate the impact of crack self-healing process on structural stiffness recovery, and this model is incorporated as a long-term reliability indicator into the optimization objective of MPC.

[0077] Shape memory alloy (SMA) active control mechanism: The SMA component can generate a preset strain (maximum recovery strain of about 6%) under temperature excitation. The system triggers its phase transformation through resistance heating or ambient temperature control to achieve dynamic fine adjustment of the support preload, with a response time of less than 5 minutes.

[0078] The following are specific embodiments of the present invention: like Figures 1 to 3 As shown, the present invention provides a collaborative handling system for tunnel collapse based on twin model prediction, the physical deployment and workflow of which are as follows: Step 1: System Deployment and Initialization A field monitoring station was established in the stable section behind the landslide area, and a digital twin analysis and decision-making hardware platform, including a high-performance computing server, was deployed.

[0079] Deploy a multi-source fusion intelligent sensing layer: install microseismic monitoring sensors within the impact range of the collapse to form a deep monitoring array; attach distributed fiber optic strain sensors to the surface of the existing initial support (steel arch frame 7); install MEMS inclinometers in the lock area; deploy laser targets in key parts of the cavity for three-dimensional laser scanning; and pre-embed earth pressure cells in the mechanical counterpressure body 1 area to be backfilled.

[0080] Prepare all terminals of the distributed collaborative execution layer: prepare grouting equipment, concrete pumping equipment, intelligent tensioning equipment, etc., and ensure that their controllers have interfaces to receive remote commands.

[0081] Step 2: Digital Twin Construction and Model Prediction Control Parameter Setting In the decision-making software platform, the tunnel design BIM model and geological survey report data are imported, and real-time data streams from all sensing layer sensors are connected. The dynamic digital twin construction and update module is activated to generate an initial three-dimensional computable model containing the collapsed body, surrounding rock, existing support, and the support structure to be constructed. In the model predictive control optimization engine, control objectives (e.g., maximum displacement of surrounding rock is less than a threshold, stress of support structure is within a safe range), control variables (construction parameters of each execution terminal), constraints (equipment capacity limitations, construction specifications), and prediction time domain (e.g., the next four construction steps) are set.

[0082] Step 3: Adaptive and Collaborative Closed-Loop Operation The system enters the automated operation phase, which is a continuous closed loop of "perception-simulation-decision-execution": Data upload: The perception layer continuously collects data and uploads it to the decision-making layer.

[0083] Data fusion and model updating: The multi-source data fusion and mechanical parameter inversion module processes data, for example, using microseismic event location results and deformation data to invert and update the damage zone and mechanical parameters of the surrounding rock in the digital twin. The dynamic digital twin thus achieves synchronous evolution with the physical site.

[0084] Model Predictive Control Optimization: At each decision cycle (e.g., every 10 minutes), the model predictive control optimization engine is activated. Starting with the currently updated digital twin state, it rapidly simulates in virtual space the possible outcomes of different control strategies (e.g., a combination of "increasing grouting pressure by 10%" and "reducing concrete backfilling speed by 20%) over multiple future steps. By solving the optimization problem, the engine selects the sequence of future control instructions that optimally achieves the preset control objective.

[0085] Command Issuance and Collaborative Execution: The decision-making layer issues optimized real-time control commands to the corresponding execution terminals. For example, a command might include: "Rock modification execution terminal, use grout with a water-cement ratio of 0.8 in borehole 3 of area B, and grout at a pressure of 0.5 MPa for 15 minutes"; simultaneously, "Cavity backfilling execution terminal, control the thickness of the next layer of concrete pouring to 900 mm, and set the upper limit of pump pressure to 2.2 MPa." Each terminal receives and executes the commands, and its execution status (such as actual grouting volume and pump pressure) is captured by the sensing layer, forming a closed loop.

[0086] Logical coordination safeguards: In particular, the stability assessment of the interlock reinforcement execution terminal is encoded as a logical lock. This lock is only released when the decision-making level determines, based on the perception data, that the interlock stability meets the safety threshold, allowing construction instructions to be sent to the pipe roof support execution terminal, thereby ensuring the safety and rigidity of the construction sequence.

[0087] Through the above closed-loop process, the system of this invention realizes adaptive, intelligent and collaborative control of the entire process of handling complex landslides, transforming the traditional static construction plan into a dynamically optimized intelligent process, which greatly improves the safety, economy and reliability of the handling.

[0088] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0089] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A tunnel collapse collaborative treatment system based on twin model prediction, characterized in that, include: Multi-source fusion intelligent perception layer, digital twin analysis and decision-making layer, and distributed collaborative execution layer; The multi-source fusion intelligent sensing layer is deployed in the collapse area to collect real-time monitoring data of the surrounding rock and structure; The digital twin analysis and decision-making layer is communicatively connected to the multi-source fusion intelligent perception layer, and a dynamic digital twin is constructed inside it and a model prediction and control algorithm is run. The distributed collaborative execution layer includes a slag back pressure execution terminal, a pipe roof support execution terminal, a cavity backfill execution terminal, a lock reinforcement execution terminal, a surrounding rock modification execution terminal, and a bottom reinforcement execution terminal. Each terminal is communicatively connected to the digital twin analysis and decision-making layer. The digital twin analysis and decision-making layer is configured to: integrate the monitoring data to drive the dynamic digital twin update, and use the model predictive control algorithm to continuously calculate the optimal control command sequence for each execution terminal in order to coordinate the construction of each terminal.

2. The system according to claim 1, characterized in that, The multi-source fusion intelligent sensing layer includes a microseismic monitoring array deployed deep within the surrounding rock, a composite sensing network deployed on the surface of the structure, and a three-dimensional laser scanning unit for scanning cavities.

3. The system according to claim 1, characterized in that, The digital twin analysis and decision-making layer includes: a dynamic digital twin construction and update module, a multi-source data fusion and mechanical parameter inversion module, and a model prediction control optimization engine.

4. The system according to claim 1, characterized in that, The tunnel slag back pressure execution terminal includes a mechanical back pressure body (1) formed by the tunnel slag and a pre-embedded concrete conveying pipe (8); the digital twin analysis and decision layer dynamically adjusts its accumulation parameters according to the sensing data of the mechanical back pressure body (1) through the model prediction control algorithm.

5. The system according to claim 1 or 4, characterized in that, The pipe roof support execution terminal includes a pipe roof advanced support (2) consisting of a large-diameter pipe roof and a grouting small pipe installed along the tunnel outline; the digital twin analysis and decision layer optimizes the circumferential spacing, overlap length and grouting parameters of the pipe roof advanced support (2) based on the load distribution simulated by the dynamic digital twin.

6. The system according to claim 4, characterized in that, The cavity backfilling execution terminal injects concrete through the concrete delivery conduit (8) to form cavity backfilling concrete (3); the digital twin analysis and decision layer controls the layer thickness and pumping pressure of the injection based on the cavity wall pressure calculated by the dynamic digital twin and the sensed micro-vibration activity.

7. The system according to claim 1, characterized in that, The lock reinforcement execution terminal uses small pipe grouting to form lock reinforcement support (4); the digital twin analysis and decision layer uses logic to control whether to authorize the construction of subsequent terminals based on the real-time evaluation results of the stability of the area where the lock reinforcement support (4) is located.

8. The system according to claim 1, characterized in that, The surrounding rock modification execution terminal includes uncovered grouting holes (10) arranged in a quincunx pattern; the digital twin analysis and decision layer simulates grout diffusion in the dynamic digital twin based on the inverted surrounding rock parameter field, and adjusts the construction parameters of each grouting hole in a differentiated manner.

9. The system according to claim 1, characterized in that, The bottom reinforcement execution terminal includes a bottom transverse steel support (6) connected to the bottom foot of the steel arch frame (7); the digital twin analysis and decision layer determines the installation timing and preload of the bottom transverse steel support (6) based on the predicted bottom drum trend and real-time stress data.

10. A method for collaborative handling of tunnel collapses, characterized in that, Using the system as described in any one of claims 1 to 9, the method comprises: Real-time monitoring is established through the multi-source fusion intelligent sensing layer; The digital twin analysis and decision-making layer updates the dynamic digital twin based on monitoring data and uses model predictive control algorithms to generate a sequence of collaborative control commands in a rolling manner. Each terminal in the distributed collaborative execution layer executes the instruction sequence, and the execution effect data is fed back through the multi-source fusion intelligent perception layer to form a closed-loop control until the processing is completed.

Images