A surrounding rock deformation scenario prediction method based on multi-point displacement meter anchorage section state

Abstract

The application provides a surrounding rock deformation scenario prediction method based on multi-point displacement meter anchoring section state, belongs to the technical field of surrounding rock internal deformation monitoring and prediction during underground engineering construction period, and obtains the cumulative displacement of the multi-point displacement meter, defines the rock mass section through adjacent anchoring points, and forms the anchoring section state with the unique identification of the rock mass section; according to the formed chamber axis side structure at the construction time, an effective connected graph is constructed, the path length of each accessible working face to the monitoring section is calculated, the excavation-supporting event coordinates with the working face identification are generated in combination with the working face track and the supporting event; the history condition is truncated and assembled at the current prediction time; the corresponding future condition is obtained according to the plan information of each construction scenario; the same history condition is respectively matched with each future condition to input the section prediction model, and the relative deformation prediction quantity is output. The application realizes the surrounding rock deformation prediction of the rock mass section dimension multiple construction scenarios, the result can be traced, and is suitable for the safety evaluation of the surrounding rock of underground engineering such as pumped storage and tunnel.

Description

Technical Field

[0001] This invention relates to the field of monitoring and predicting internal deformation of surrounding rock during the construction period of underground engineering, and in particular to a method for predicting the deformation scenario of surrounding rock based on the state of the anchorage section of a multi-point displacement meter. Background Technology

[0002] After the excavation and unloading of underground caverns, the stress in the surrounding rock redistributes, accompanied by the opening of joints and fissures and the continuous expansion of relaxation zones. Under the combined effects of excavation disturbance, groundwater, and support sequence, dynamic internal layered deformation occurs, varying with time and the location of the working face. Multi-point displacement gauges can acquire displacement data at different anchoring depths in the borehole. They are important monitoring instruments for identifying areas of concentrated deformation within the surrounding rock and determining the depth of relaxation, and are widely used in underground engineering projects such as pumped storage water conveyance caverns, traffic tunnels, and mine roadways.

[0003] Most existing surrounding rock deformation prediction schemes directly use each independent anchoring point of a multi-point displacement gauge as the prediction object, lacking directional characterization of rock mass sections between adjacent anchoring points. This makes it impossible to pinpoint the predicted results to the actual rock mass strata where deformation actually occurs inside the borehole, resulting in a one-to-one correspondence between the measurement point output and the actual rock mass deformation unit. For complex underground engineering projects with multiple working faces excavating in parallel, and with branch tunnels and multiple connecting chambers, existing technologies generally adopt a simplified approach of selecting the single working face closest to the monitoring section to represent all construction disturbances, discarding the disturbance contributions of other accessible working faces. This fails to reflect the superimposed disturbance effects of multiple working faces. Furthermore, most prediction methods do not strictly distinguish between historical construction events that have already occurred and future planned construction events that have not yet been implemented at the prediction cutoff point. This easily leads to the leakage of post-hoc actual construction information into the historical input, resulting in inflated prediction results and an inability to reproduce the actual prediction conditions at the engineering site.

[0004] Therefore, it is necessary to propose a method for predicting the deformation scenario of surrounding rock based on the state of the anchored section of a multi-point displacement meter in order to solve or at least alleviate the above-mentioned defects. Summary of the Invention

[0005] The main objective of this invention is to provide a method for predicting surrounding rock deformation scenarios based on the state of anchored sections using multi-point displacement gauges. This method addresses the problems of existing technologies that use independent measuring points as prediction objects, simplify the superposition of disturbances across multiple working faces, confuse the boundaries between historical and future construction conditions, have poor reproducibility of scenario simulations, and are difficult to achieve multi-scenario surrounding rock deformation prediction with traceable dimensions of rock mass sections.

[0006] To achieve the above objectives, this invention provides a method for predicting surrounding rock deformation scenarios based on the state of the anchorage section of a multi-point displacement meter, comprising the following steps: The cumulative displacement of the same multi-point displacement gauge is obtained, the rock mass section is defined by the adjacent anchoring points, and the synchronous displacement difference of the adjacent anchoring points is calculated under the unified displacement direction and cumulative benchmark to form the anchoring section state with a unique identifier of the rock mass section. Construct an effective connectivity diagram along the tunnel axis formed at the construction time, calculate the path length from each reachable working face to the monitoring section of the rock mass section along the effective connectivity diagram, and combine the working face trajectory and support events to form the excavation-support event coordinates with working face identification. The status of the anchorage section up to the current prediction time and the coordinates of the excavation-support event are combined to form historical conditions; Based on the planned edge formation time, planned working face trajectory, and planned support completion time for each construction scenario, update the future effective connectivity map and the coordinates of future excavation-support events for each predicted time interval to form future conditions. The historical conditions and each of the future conditions are input into the pre-trained segment prediction model to obtain the relative deformation prediction amount corresponding to the construction scenario, the prediction time interval, and the unique identifier of the rock mass segment.

[0007] Preferably, the step of acquiring the cumulative displacement of the same multi-point displacement gauge, defining the rock mass section by adjacent anchoring points, calculating the synchronous displacement difference of the adjacent anchoring points under a unified displacement direction and cumulative benchmark, and forming an anchoring section state with a unique identifier for the rock mass section, includes: The cumulative displacement and anchorage depth of each anchorage point of the same multi-point displacement gauge are obtained. Adjacent anchorage points are determined in ascending order of anchorage depth. The adjacent anchorage points define the rock mass section and establish a unique identifier for the rock mass section. Based on the borehole axial and zero-point adjustment records, the cumulative displacement is converted to a unified displacement direction and cumulative reference. The effective cumulative displacements of adjacent anchor points with the same observation time or an observation time difference not exceeding the preset pairing time difference are paired, and the cumulative displacement of the shallow anchor points is subtracted from the cumulative displacement of the deep anchor points to obtain the synchronous displacement difference. The average relative deformation rate of the section is calculated based on the synchronous displacement difference and the actual time difference between adjacent effective observations. The synchronous displacement difference, the average relative deformation rate of the section, and the unique identifier of the rock mass section are associated to form the anchorage section status.

[0008] Preferably, the step of constructing an effective connectivity map based on the tunnel axis edges formed at the construction time, calculating the path length from each reachable working face to the monitoring section where the rock mass section is located along the effective connectivity map, and forming excavation-support event coordinates with working face identifiers by combining the working face trajectory and support events, includes: Extract the actual tunnel axis edges formed up to the construction time, map the monitoring section where the rock mass section is located and each working face to the tunnel axis edges, and construct an effective connectivity diagram; The passage status of each reachable working face relative to the monitoring section is determined based on the working face trajectory, and the direction coefficient is determined based on the passage status; Based on the effective connectivity graph and the direction coefficient, the path length and oriented path coordinates from each reachable working face to the monitoring section are calculated using the following formula: in, Indicates working face To the monitoring section The path length of the section in which it is located; Indicates monitoring section In an effective connected graph The mapping position in Indicates working face At any moment Location, Indicates that only time is used The path length operator for active valid edges; Indicates working face At any moment directional coefficient, Indicates working face The assigned path coordinates; The working face identifier of each reachable working face, the path length, and the coordinates of the oriented path are respectively used to form the excavation event sub-coordinates; The support status acting on the monitoring section is determined based on the range of action of the support event, and the support status and the sub-coordinates of each excavation event are combined to form an excavation-support event coordinate system with the working face identifier.

[0009] Preferably, the step of extracting the actual tunnel axis edges formed up to the construction time, mapping the monitoring section where the rock mass section is located and each working face to the tunnel axis edges, and constructing an effective connectivity graph includes: The location of the working face at the construction time is determined based on the working face trajectory, and the monitoring section where the rock mass section is located is mapped to the tunnel axis model; When the working face is located inside the tunnel axis, a dynamic node is set at the working face to divide the tunnel axis into formed sub-edges and unformed sub-edges. Extract the formed sub-edges and the tunnel axis edges that have been fully formed up to the construction time as valid edges, and map each working face to the corresponding valid edge. Construct a valid connected graph based on the connection relationship of the valid edges.

[0010] Preferably, the step of assembling the anchorage section status up to the current prediction time and the excavation-support event coordinates into historical conditions includes: Using the current prediction time as the cutoff point, extract the segment records from the anchorage segment status that have been obtained up to the cutoff point and whose observation time is no later than the cutoff point; Extract event records from the excavation-support event coordinates that have been obtained up to the cutoff point and whose event occurrence time is no later than the cutoff point; According to the monitoring section corresponding to the unique identifier of the rock mass section, the section record is associated with the event record, and the working face identifier in the event record is retained; The associated segment records and event records, arranged according to the observation time and the event occurrence time, constitute historical conditions.

[0011] Preferably, the step of updating the future effective connectivity graph and future excavation-support event coordinates for each predicted time interval according to the planned edge formation time, planned working face trajectory, and planned support completion time for each construction scenario, to form future conditions, includes: Obtain the construction plan versions for each construction scenario frozen at the current prediction time, and determine the planned edge formation time, planned working face trajectory, and planned support completion time; At each prediction time interval, the planned edge whose formation time is no later than the prediction time is combined with the effective connected graph at the current prediction time to form a future effective connected graph. Along the future effective connectivity graph, the future excavation event sub-coordinates of each reachable working face are formed according to the planned working face trajectory, and the support events whose planned support completion time is no later than the predicted time are set to the completed state; The sub-coordinates of each future excavation event and the corresponding support status are combined to form the coordinates of the future excavation-support event, which, together with the future effective connectivity graph, form the future conditions of the corresponding construction scenario.

[0012] Preferably, the step of inputting the historical conditions and each of the future conditions into a pre-trained segment prediction model to obtain the relative deformation prediction corresponding to the construction scenario, the prediction time interval, and the unique identifier of the rock mass segment includes: Between each of the construction scenarios corresponding to the current prediction time, the parameters of the pre-trained segment prediction model and the historical conditions remain unchanged; Based on the unique identifier of the rock mass section, the historical conditions are paired with each of the future conditions to form a model input that retains the working face identifier; Based on the model input and the prediction time interval, the segment prediction model outputs the relative deformation prediction amount corresponding to the construction scenario, the prediction time interval, and the unique identifier of the rock mass segment. The calculation formula is as follows: in, Indicates the time interval of future forecasts The predicted deformation amount afterward, Indicate output conditions Quantile segmental deformation mapping, Indicates as of The historical status of the period and the events that have occurred. Describing a scenario Future events that will be activated at the planned times within the prediction interval. Indicates the index of rock mass sections. Indicates the predicted time interval. Indicates quantile level, satisfying ; This represents the model parameters obtained through training.

[0013] Preferably, after outputting the predicted relative deformation corresponding to the construction scenario, the predicted time interval, and the unique identifier of the rock mass segment through the segment prediction model based on the model input and the predicted time interval, the method further includes: Using the segment prediction model, the upper quantile of the absolute values ​​of the relative deformation, relative deformation rate, and relative deformation acceleration of the segment corresponding to each construction scenario and the prediction time interval is generated respectively. When the upper quantile of the predicted absolute value of the relative deformation amount or the relative deformation rate of the section exceeds the corresponding preset deformation control value or preset rate control value, a risk warning containing the unique identifier of the rock mass section, the construction scenario, and the expected time of exceeding the limit is generated.

[0014] Preferably, after generating the risk warning containing the unique identifier of the rock mass section, the construction scenario, and the expected time of exceeding the limit, the method further includes: Obtain the pre-support acceptance status. When the pre-support acceptance status is "acceptance completed", select a continuous preset number of prediction time periods in each of the construction scenarios. The prediction time periods all satisfy the condition that the upper quantile of the absolute values ​​of the relative deformation of the section, the relative deformation rate of the section, and the relative deformation acceleration of the section does not exceed the preset deformation control value, the preset stability rate threshold, and the preset stability acceleration threshold, respectively. The earliest start time among the selected continuous prediction time periods is determined as the start time of the candidate time window for the stability of closed support or secondary lining under the corresponding construction scenario.

[0015] Preferably, the segment prediction model adopts one of the following: state-space model, regularized regression model, tree model, or time-series network model.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a method for predicting surrounding rock deformation scenarios based on the anchorage section status of multi-point displacement gauges. This method utilizes adjacent anchorage points of the same multi-point displacement gauge to define rock mass sections and assigns each section a unique identifier. Under a unified displacement direction and cumulative displacement benchmark, the anchorage section status is constructed. The observation data from discrete anchorage points are transformed into the relative deformation status of rock mass layered units, ensuring that the model input and prediction output directly correspond to the actual rock mass sections inside the borehole, facilitating the identification of areas with concentrated deformation within the surrounding rock. An effective connectivity diagram is constructed based on the actual excavated tunnel axis edges at each construction moment. The path length from each accessible working face to the monitoring section is calculated along the actual tunnel pathway, fully preserving all working face identifiers. This method abandons the simplistic assumption of selecting only the nearest working face to represent disturbance and fully considers the parallel excavation of multiple working faces in bifurcated tunnels. The superimposed disturbances generated during the process are mitigated; the current prediction time is used as a strict time boundary to isolate historical conditions from future conditions of various construction scenarios, avoiding the leakage of post-construction information and restoring the actual working conditions on site where only historical monitoring data and frozen construction plans can be obtained; in the same round of prediction, the predicted model and parameters of the section completed by training are reused for different construction scenarios, only replacing the future construction conditions, so that the difference in prediction results between scenarios comes only from the excavation and support plan, ensuring the objectivity and reproducibility of multi-scheme comparison, and at the same time, the output results are synchronously bound to the construction scenario, prediction time interval, and unique identifier of the rock mass section, and the prediction results can be traced back to the source, which can provide standardized data support for risk assessment and support scheme evaluation, and is applicable to various underground cavern projects such as pumped storage water conveyance caverns, traffic tunnels, and mine roadways with multiple displacement gauges. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the prediction method in one embodiment of the present invention.

[0019] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0020] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0022] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0023] Please see the appendix Figure 1 The present invention provides a method for predicting surrounding rock deformation scenarios based on the state of the anchorage section of a multi-point displacement meter, comprising the following steps: The cumulative displacement of the same multi-point displacement gauge is obtained, and rock mass sections are defined by adjacent anchor points. Under a unified displacement direction and cumulative benchmark, the synchronous displacement difference between the adjacent anchor points is calculated, forming an anchoring section state with a unique identifier for the rock mass section. For multiple anchoring measurement points inside the multi-point displacement gauge borehole, the corresponding rock mass sections are divided using the pairwise adjacent anchor points within the same multi-point displacement gauge as boundaries. Under the constraints of a unified positive displacement direction and a unified cumulative displacement benchmark, only observation data that meets the time synchronization condition is used to calculate the synchronous displacement difference between adjacent anchor points, and this is for… Each segment of the rock mass is assigned a globally unique identifier. This process transforms the scattered anchor point measurement data into the deformation state of the corresponding real rock mass layers inside the borehole. This solves the technical problem that directly using the absolute displacement of anchor points can only obtain the displacement of the measurement points and cannot characterize the relative compression or tension deformation of the rock mass between two anchor points. This allows all subsequent prediction inputs and outputs to be accurately located to the specific rock mass segment inside the borehole, eliminating the defect of mismatch between measurement points and rock mass objects, and providing the basic input of the segment dimension for the entire prediction process.

[0024] An effective connectivity graph is constructed based on the tunnel axis edges already formed at the construction time. The path length from each reachable working face to the monitoring section of the rock mass segment is calculated along the effective connectivity graph. Combined with the working face trajectory and support events, excavation-support event coordinates with working face identifiers are formed. In this way, the effective connectivity graph is constructed based on the tunnel axis edges that have been actually excavated and are ready for construction at the corresponding construction time. Unexcavated and unconnected planned tunnel segments are excluded from the path calculation. All excavation working faces that can reach the monitoring section through the effective connectivity graph are traversed. The path length from each excavation working face to the monitoring section is calculated along the actual tunnel connectivity path. Combined with the complete excavation trajectory of the working face and the log of support events executed on site, independent excavation-support event coordinates for each excavation working face are generated. The working face identifiers corresponding to each excavation working face are retained throughout the process, and multiple working faces are not merged or weighted. In this way, the surrounding rock disturbance information brought by all working faces under the condition of parallel excavation of multiple working faces is completely preserved, and the real spatial disturbance boundary conditions under the conditions of bifurcated tunnels and multiple working faces are restored.

[0025] It should be noted that prior to this, a documentation process is required to establish an effective connectivity diagram. This process begins with the documentation of multi-point displacement gauge projects: each physical multi-point displacement gauge is uniquely identified according to the hierarchy of water conveyance line—structure—tunnel section—monitoring section—borehole—anchor point, distinguishing between "number of physical instruments" and "number of anchorage depth measurement channels." The instrument file must include at least the tunnel name, section number, section station number or elevation, borehole 3D coordinates, borehole axial unit vector, number of anchor points, depth of each anchor point from the borehole opening, installation date, initial reading date, range, resolution, positive and negative directions, zero-point adjustment record, and failure status. Three-point, four-point, and other multi-point displacement gauges are uniformly represented as a variable-length anchor point sequence, with anchor points ordered from smallest to largest depth from the borehole opening. The actual number of anchor points and validity masks are retained, and three-point instruments are not forcibly extended to four-point instruments by copying measurement points. The positive displacement direction is also standardized based on the borehole axial unit vector. The preferred definition is that the relative displacement from the stable deep surrounding rock towards the tunnel wall is positive; when the original instrument definition is reversed, all observations of the instrument are multiplied by a direction correction factor. Data before and after zero-point reset retain the original values ​​and are converted to a unified cumulative displacement benchmark based on the reset records.

[0026] Then, files are created for the working faces and support objects: For each excavation face, a working face identifier, the corresponding chamber, the direction of advancement, the start and end station numbers or elevations for each cycle, the excavation start time, the blasting time, the muck removal end time, and the cycle advance are established. For bidirectional excavation upstream and downstream, multiple access tunnels, or vertical shaft layered excavation, separate working face trajectories are established. It is prohibited to use only one working face record for a chamber with the same name to represent multiple working faces. Initial support, reinforcement support, sealing support, and secondary lining records are converted into support events. Each support event includes the affected tunnel section, support type, start time, completion time, main parameters, and the working face location at the time of construction. For layered excavation or sectional support, the crown, sidewalls, floor slab, or zone identifiers should also be recorded.

[0027] The historical conditions are composed of the anchorage section status and the excavation-support event coordinates up to the current prediction time. Strictly using the current prediction time as the time cutoff boundary, only anchorage section status records and excavation-support event coordinate records whose observation time and event occurrence time are no later than the current prediction time are selected to assemble the historical conditions. Any observation data or construction events that have not yet been collected or occurred at the current prediction time are not allowed to be written into the historical conditions. Through this forced isolation of the time boundary, it is ensured that all historical conditions are real information that can actually be obtained from the site at the prediction time, completely avoiding the leakage of future known data into the historical input, preventing information crossover between model training and online prediction, and ensuring that the prediction logic conforms to the real-world usage scenario of the engineering site.

[0028] Based on the planned edge formation time, planned working face trajectory, and planned support completion time for each construction scenario, the future effective connectivity graph and future excavation-support event coordinates for each prediction time interval are updated to form future conditions. For each pre-set construction scenario, the construction plan version that has been frozen and locked at the current prediction time is read, and the planned edge formation time, planned working face trajectory, and planned support completion time are extracted. For each prediction time interval corresponding to the future prediction time, the future effective connectivity graph specific to that construction scenario is dynamically updated, and the future excavation-support event coordinates are updated synchronously. This generates multiple sets of independent future conditions corresponding to different excavation rhythms and different support sequences. Relying on the frozen plan versions, the conditions for different construction schemes are constructed differently, simulating the evolution process of gradual excavation and gradual support construction under different construction plans.

[0029] The historical conditions and future conditions are input into the pre-trained segment prediction model to obtain the relative deformation predictions corresponding to the construction scenario, the prediction time interval, and the unique identifier of the rock mass segment. During the calculation of all construction scenarios in the same round of prediction, all parameters of the pre-trained segment prediction model remain fixed. The same set of historical conditions is paired with future conditions corresponding to different construction scenarios and input into the segment prediction model for mapping calculation. Each set of relative deformation predictions output by the model is simultaneously bound to the corresponding construction scenario, prediction time interval, and unique identifier of the rock mass segment. This ensures that the differences between prediction results for different construction scenarios stem only from the differences in future construction conditions, and will not lead to result deviations due to model parameter readjustment under different scenarios. It can reproduce the rock mass segment deformation results corresponding to different excavation and support schemes under the same historical state of surrounding rock. The prediction results can be traced back to the corresponding rock mass segment, construction scenario, and prediction time, facilitating engineering auditing and risk tracing.

[0030] As a preferred embodiment of the present invention, the step of obtaining the cumulative displacement of the same multi-point displacement gauge, defining the rock mass section by adjacent anchoring points, calculating the synchronous displacement difference of the adjacent anchoring points under a unified displacement direction and cumulative benchmark, and forming an anchoring section state with a unique identifier for the rock mass section includes: The cumulative displacement and anchorage depth of each anchorage point of the same multi-point displacement gauge are obtained. Adjacent anchorage points are determined in ascending order of anchorage depth, and these adjacent anchorage points define the rock mass segment and establish a unique identifier for the rock mass segment. By reading the archive records of the same multi-point displacement gauge, the cumulative displacement observation value and the anchorage depth of each anchorage point from the borehole opening are obtained. The anchorage points are sorted from shallow to deep according to their anchorage depth. After sorting, the two adjacent anchorage points in the sequence are used as the upper and lower boundaries to define the rock mass segments between them. The and the first The anchor points together define the rock mass section. , This indicates a monitoring section comprised of multiple physical displacement gauges within a cross-section. A globally unique identifier for each rock mass segment is generated. It should be noted that the anchoring points of the same multi-point displacement gauge are arranged in ascending order of distance from the borehole opening. Number, No. The and the first The anchoring points are the shallow and deep ends of this section, and the rock mass section between them is denoted as... Section numbering by Determined, i.e., segment index The value is taken from the sequence number of the shallow end anchorage point in this section. By sorting and segmenting, one-dimensional discrete anchoring measuring points are mapped to continuously layered rock mass objects inside the borehole, realizing the transformation of monitoring data from the measuring point dimension to the rock mass segment dimension. This ensures that all subsequent inputs and outputs can accurately point to the corresponding rock mass layers inside the borehole, and there will be no confusion between measuring points and rock mass segment objects.

[0031] Based on the borehole axial and zero-point adjustment records, the cumulative displacement is converted to a unified displacement direction and cumulative reference. By retrieving the borehole axial unit vector from the multi-point displacement gauge installation file, and simultaneously retrieving the zero-point reset and zero-point adjustment records throughout the instrument's lifecycle, the original cumulative displacement of each anchor point is corrected in direction and aligned with the zero-point reference. All observation values ​​with inconsistent original reading directions from different instruments and reference jumps caused by multiple zeroing are converted to the same positive displacement direction and the same cumulative displacement reference. This eliminates the data reference inconsistency caused by instrument factory settings and on-site zero-point resets, ensuring that the displacement observation values ​​between different anchor points can be directly used for differential calculations, and avoiding distortion in synchronous displacement difference calculations caused by reference errors.

[0032] The effective cumulative displacements of adjacent anchor points with the same observation time or an observation time difference not exceeding the preset pairing time difference are paired. The cumulative displacement of the shallow anchor point is subtracted from the cumulative displacement of the deep anchor point to obtain the synchronous displacement difference. Observation time pairing constraint rules are set, and two sets of observations are considered to be able to be paired only when the observation times of adjacent anchor points are exactly the same, or the time difference between the two observations is less than the preset pairing time difference. At the same time, invalid observation records due to instrument failure, exceeding the range, or missing measurements are eliminated. After pairing, the synchronous displacement difference is obtained by subtracting the cumulative displacement of the deep anchor point from the cumulative displacement of the shallow anchor point. Direct differential calculation on asynchronous observation values ​​is not allowed to avoid false deformation calculations due to asynchronous observation times, ensuring that the synchronous displacement difference can truly reflect the actual relative deformation of the rock mass section within the same time window.

[0033] Specifically, the two anchor points are observed at the same time. The cumulative displacements in the same positive direction and under the same cumulative displacement reference are respectively denoted as... and The anchoring depth is denoted as follows: and The relative deformation of the section (i.e., synchronous displacement difference) and the equivalent index normalized according to the section length are respectively: in, Indicates rock mass section At any moment The relative deformation; Used for comparison between different anchorage spacings, it is an equivalent index and does not replace continuous strain testing in the field. Only two valid observations paired within a uniform time synchronization or allowable time difference constitute a synchronous displacement difference; asynchronous observations are not directly used for subtraction. In this formula, the segment index... Corresponding to the The and the first The rock mass section between the anchoring points, and meets the requirements .

[0034] The average relative deformation rate of the section is calculated based on the synchronous displacement difference and the actual time difference between adjacent valid observations. The synchronous displacement difference, the average relative deformation rate of the section, and the unique identifier of the rock mass section are associated to form the anchored section state. It can be understood that the actual time interval between two valid observation records that meet the pairing conditions before and after the rock mass section is read, and the average relative deformation rate of the section is obtained by dividing the change in synchronous displacement difference by the actual time interval. Then, the synchronous displacement difference and the average relative deformation rate of the section are bound and encapsulated with the unique identifier of the rock mass section to form a complete anchored section state. In addition to the deformation amount, the deformation rate time series feature is added to fully characterize the deformation magnitude and deformation development speed of the rock mass section, enrich the model input feature dimension, and allow the model to learn the deformation degree and deformation development trend of the rock mass section at the same time.

[0035] In detail, the following formula is used for calculation: in, Indicates the time interval between adjacent valid observations of a single anchor point and The average displacement rate between Indicates the anchor point is at Cumulative displacement at time step Indicates the anchor point is at The cumulative displacement at any given time; This represents the average relative deformation rate of the rock mass segment itself. Indicates rock mass section At any moment The relative deformation.

[0036] It is worth noting that the anchorage section status should include at least the section's unique identifier, the anchorage point identifiers at both ends, the synchronous cumulative displacement difference, the section length, the relative deformation, the section's average relative deformation rate, the observation time interval, the directional validity, the missing mask, and the quality level. Cumulative displacement, single-point rate, robust acceleration, and equivalent section deformation rate can be used as preferred auxiliary features; three-point, four-point, or other multi-point instruments retain the actual number of anchorage points and do not supplement them through copy channels.

[0037] As a preferred embodiment of the present invention, the step of constructing an effective connectivity map based on the tunnel axis edges formed at the construction time, calculating the path length from each reachable working face to the monitoring section where the rock mass section is located along the effective connectivity map, and forming excavation-support event coordinates with working face identifiers by combining the working face trajectory and support events, includes: Extract the actual tunnel axis edges formed up to the construction time, map the monitoring sections and working faces of the rock mass section to the tunnel axis edges, and construct an effective connectivity graph; for any construction time... Only the axial sections that have been actually excavated and formed construction connections up to that point are extracted as valid connectivity maps. The effective connected graph is calculated using an undirected skeleton for path length calculation. Each edge records its geometric length, actual activation time, and source log. The working face advancement direction is determined separately by the corresponding working face trajectory. When the working face is located inside an axis edge, a dynamic node is set based on the current position of the working face, dividing the edge into formed sub-edges and unformed sub-edges. Only the formed sub-edges are added. The dynamic nodes are moved according to the planned working face position at each predicted time for future construction scenarios. Monitoring sections, excavation working faces, and support action areas are mapped to corresponding edges according to station number, elevation, or trajectory mileage. Edges that have not yet been formed at historical times are not used for path calculation at that time; future planned edges are only added to the effective connected graph of the corresponding construction scenario after the planned activation time is reached.

[0038] The passage status of each reachable working face relative to the monitoring section is determined based on the working face trajectory, and the direction coefficient is determined based on the passage status. The complete tunneling trajectory log of each excavation working face is retrieved, and the sequential position relationship between the working face and the monitoring section on the tunnel axis is compared to determine whether the working face is in one of three states: approaching the monitoring section, having passed the monitoring section, or moving away from the monitoring section. The corresponding direction coefficient is assigned based on the determined passage status. In addition to the unsigned path length, a direction attribute is added to distinguish whether the working face is tunneling towards or away from the monitoring section. The path length alone can only express the distance, while the direction coefficient is used to mark the passage status of the working face relative to the monitoring section. The strength of the disturbance also needs to be judged in combination with the advance speed, blasting events, and monitoring response.

[0039] Specifically, based on the effective connectivity graph and the direction coefficient, the path length and oriented path coordinates from each reachable working face to the monitoring section are calculated using the following formula: in, Indicates working face To the monitoring section The path length of the section in which it is located; Indicates monitoring section In an effective connected graph The mapping position in Indicates working face At any moment Location, Indicates that only time is used The path length operator for active valid edges; Indicates working face At any moment The directional coefficient is determined based on the predetermined advancement path of the working face and its sequential position relative to the cross-section: +1 is taken before the working face passes the cross-section, -1 is taken after passing it, and 0 is taken only when the mapping positions coincide; the working face's shutdown status is recorded separately, without changing the directional coefficient determined before shutdown. , Indicates working face The path coordinates are assigned. It's important to note that multiple work surfaces retain their work surface identifiers and form independent event sub-coordinates; they are not merged into a single distance simply because they are equal or close in distance. When When empty, output the working surface event is unavailable; retain the value when the direction cannot be reliably determined. Set the direction validity mask to zero to prevent unknown directions from being mistakenly written as distances of zero.

[0040] Each reachable working face is assigned a working face identifier, path length, and directional path coordinates to form excavation event sub-coordinates. For each excavation working face accessible to the monitoring section via the effective connectivity graph, an independent set of excavation event sub-coordinates is generated. The sub-coordinates fully retain the working face identifier specific to that working face, without averaging or minimizing the paths and coordinates of multiple working faces. All disturbance information of reachable working faces is fully preserved, resolving the issue of selecting only the nearest working face and losing the disturbance contribution of other working faces in multi-working-face scenarios. For each working face… The path length, directional validity, oriented path coordinates, advance speed, duration after the most recent blast, and disturbance stage are combined to form event sub-coordinates with working face identification.

[0041] The support status acting on the monitoring section is determined according to the range of action of the support event. The support status and the sub-coordinates of each excavation event are combined to form the excavation-support event coordinates with the working face identifier. The support events are mapped to the section according to their range of action and start, completion or acceptance time. The initial support, reinforcement support, closure support or secondary lining events acting on the monitoring section are recorded separately for completion status and acceptance status. Finally, the sub-coordinates of all working face events of the same section and the support events together constitute the excavation-support event coordinates.

[0042] In a preferred embodiment of the present invention, the step of extracting the actual tunnel axis edge formed up to the construction time, mapping the monitoring section where the rock mass section is located and each working face to the tunnel axis edge, and constructing an effective connectivity graph includes: The working face position at the construction time is determined based on the working face trajectory. The monitoring section where the rock mass section is located is mapped to the tunnel axis model. The working face excavation trajectory is read and the real-time position of the working face on the tunnel axis at the current construction time is obtained. Then, the monitoring section where the rock mass section is located is mapped to the pre-established tunnel axis geometric model according to the station number. A spatial mapping benchmark is established between the monitoring section, the excavation working face and the tunnel topology to provide a spatial positioning basis for the subsequent splitting of axis edges and selection of effective edges.

[0043] When the working face is located inside the tunnel axis, a dynamic node is set at the working face location to divide the tunnel axis into formed sub-edges and unformed sub-edges. By judging the real-time position of the working face, if the working face has not yet reached the end point of the tunnel axis, that is, the working face falls within the interval of a single tunnel axis, a dynamic node is inserted at the current position of the working face. The original complete tunnel axis is divided into two parts by the dynamic node: the part of the dynamic node facing the already excavated side is recorded as the formed sub-edge, and the part of the dynamic node facing the rock mass to be excavated is recorded as the unformed sub-edge. This simulates the physical process of a single tunnel edge gradually forming as the tunnel face advances during the tunneling process, solving the problem that the entire axis edge is included in the topology at once and cannot reflect the local forming state of the tunnel during the tunneling process.

[0044] The formed sub-edges and the tunnel axis edges that have been fully formed up to the current construction time are extracted as valid edges, and each working face is mapped to the corresponding valid edge. Based on the connection relationship of the valid edges, a valid connected graph is constructed. Finally, the formed sub-edges obtained from the above splitting are collected, and the complete tunnel axis edges that have been fully excavated up to the current construction time are also collected. Both types of edges are used as valid edges, and each excavation working face is mapped to the corresponding valid edge. Based on the topological connection relationship between all valid edges, a valid connected graph is constructed. This enables the tunnel connection topology to be dynamically updated according to the excavation progress of the working face, ensuring that the valid connected graph corresponding to each construction time accurately reflects the actual excavation and connection status of the tunnel on site.

[0045] In a preferred embodiment of the present invention, the step of forming historical conditions from the state of the anchorage section up to the current prediction time and the coordinates of the excavation-support event includes: Using the current prediction time as the cutoff point, extract the segment records from the anchorage segment status that have been obtained up to the cutoff point and whose observation time is no later than the cutoff point; thus, using the current prediction time as a strict time cutoff point, traverse all anchorage segment status records, only filter segment records whose observation time is less than or equal to the cutoff point, and discard future observation data whose observation time is later than the cutoff point; ensure that all segment records entering the historical conditions are observations that have been collected on-site at the prediction time, and prevent future observation data from flowing into the historical input in advance.

[0046] Extract event records from the excavation-support event coordinates that have been obtained up to the cutoff point and whose event occurrence time is no later than the cutoff point; similarly, using the current prediction time as the time boundary, filter excavation-support event coordinate records whose event occurrence time is no later than the cutoff point, and exclude all excavation events and support events that have not yet occurred, and do not write them into the historical conditions; prevent future planned construction events from being mixed into historical input, and avoid the typical future information leakage problem in time series prediction.

[0047] According to the monitoring section corresponding to the unique identifier of the rock mass section, the section record is associated with the event record, and the working face identifier in the event record is retained; using the monitoring section number corresponding to the unique identifier of the rock mass section as the association key value, the filtered section record is associated and matched with the event record. The matching process fully retains the working face identifier of each excavation working face within the event record; so that each deformation record of the rock mass section is matched with the construction disturbance information of all accessible working faces at the corresponding time, maintaining a data organization structure of independent and non-merged multiple working faces.

[0048] The associated segment records and event records are arranged according to the observation time and the event occurrence time to form historical conditions. Finally, the associated and matched segment records and event records are sorted chronologically according to the observation time and the event occurrence time. After sorting, they are packaged and assembled to obtain complete historical conditions. It is ensured that the data inside the historical conditions are arranged strictly according to the on-site time evolution order, which conforms to the input logic of the time series model and restores the complete real on-site historical state that can be obtained at the prediction time.

[0049] Furthermore, the future effective connectivity graph and future excavation-support event coordinates are updated according to the planned edge formation time, planned working face trajectory, and planned support completion time for each construction scenario, forming future conditions, including: Obtain the construction plan versions of each construction scenario frozen at the current prediction time, and determine the planned edge formation time, planned working face trajectory, and planned support completion time; by reading the construction plan versions corresponding to each set of construction scenarios that have been frozen and locked at the current prediction time, extract the planned edge formation time, planned working face tunneling trajectory, and planned support completion time for each set of construction scenarios; use the plan version frozen at the prediction time, rather than the construction records after actual changes in the later stages of the project, so that when conducting predictions on the engineering site, only the real scenario of the established construction plan can be obtained, avoiding information leakage caused by using data that actually occurred afterward.

[0050] At each prediction time interval, the planned edges whose formation time is no later than the prediction time are combined with the effective connected graph at the current prediction time to form a future effective connected graph; traversing each prediction time interval For a given construction scenario, only the planned tunnel edges whose formation time is less than or equal to the predicted future time and the current corresponding support status are combined with the effective connectivity graph to generate a future effective connectivity graph specific to that scenario and that predicted time. .

[0051] Along the future effective connectivity graph, future excavation event sub-coordinates for each reachable working face are formed based on the planned working face trajectory, and support events with planned support completion times no later than the predicted time are marked as completed. Based on the aforementioned generated future effective connectivity graph, the planned working face trajectory is read, and future excavation event sub-coordinates corresponding to each reachable excavation working face at the future predicted time are calculated and generated. At the same time, the planned support completion time is compared, and support events with planned support completion times no later than the future predicted time are marked as completed. The excavation positions of different working faces and the completion status of support construction are simulated at the corresponding future time.

[0052] The sub-coordinates of each future excavation event and the corresponding support status are combined to form the coordinates of the future excavation-support event, which are then combined with the future effective connectivity graph to form the future conditions of the corresponding construction scenario; thus, a complete set of future conditions for the construction scenario is obtained; the future conditions between different construction scenarios are independent of each other, and the planned edges, working face trajectories, and support statuses are not mixed, thereby completing the construction of multiple sets of differentiated future inputs.

[0053] Further, the step of inputting the historical conditions and each of the future conditions into the pre-trained segment prediction model to obtain the relative deformation prediction corresponding to the construction scenario, the prediction time interval, and the unique identifier of the rock mass segment includes: Among the various construction scenarios corresponding to the current prediction time, the parameters of the pre-trained segment prediction model and the historical conditions remain unchanged. In the same complete prediction task, the same pre-trained segment prediction model is called for all construction scenarios, and all weight parameters of the model remain fixed. At the same time, all scenarios reuse the same set of assembled historical conditions, and the model parameters are not retrained or tuned for a particular construction scenario. This ensures that the differences in prediction results of different scenarios can only come from the differences in future conditions, eliminates the interference of model parameter changes on the comparison results of multiple scenarios, and ensures that the deformation prediction results of multiple scenarios are horizontally comparable.

[0054] According to the unique identifier of the rock mass section, the historical conditions are paired with each of the future conditions to form a model input that retains the working face identifier; using the unique identifier of the rock mass section as the matching basis, the same set of historical conditions is paired with the future conditions corresponding to each set of construction scenarios one by one. In the process of generating the model input, the working face identifier of each excavation working face is completely preserved, and the input structure of independent sub-coordinates of multiple working faces is maintained; it is ensured that the input sent to the model carries the identity information of the rock mass section, the historical deformation state, and the future construction disturbance information of all working faces, and maintains the input structure corresponding to the complete technical logic of the present invention.

[0055] Based on the model input and the prediction time interval, the segment prediction model outputs the relative deformation prediction amount corresponding to the construction scenario, the prediction time interval, and the unique identifier of the rock mass segment. The calculation formula is as follows: in, Indicates the time interval of future forecasts The predicted deformation amount afterward, Indicate output conditions Quantile segmental deformation mapping, Indicates as of The historical status of the period and the events that have occurred. Describing a scenario Future events that will be activated at the planned times within the prediction interval. Indicates the index of rock mass sections. Indicates the predicted time interval. Indicates quantile level, satisfying ; This represents the model parameters obtained through training.

[0056] In another preferred embodiment of the present invention, as a preferred spatial correction, the monitoring section is used as the node; a candidate association is established only when two sections are reachable through the effective connectivity graph at the corresponding time and the path distance does not exceed the influence range confirmed by the engineering. The weights simultaneously consider the path distance, the standardized disturbance stage difference between the two sections, geological similarity, and support stage similarity: in, Indicates the cross-sectional correlation weight. Indicates whether the two cross sections are at time 10:00. The conditions for effective connectivity and influence range are met. Indicates the effective path length. This represents the positive path distance attenuation scale, using the same units as the effective path scale. This indicates the attenuation scale during the disturbance phase. , Represents the coordinates of comparable standardized perturbation stages. Indicating geological similarity, This indicates the similarity of the support stages.

[0057] Then, the topologically weighted neighbor section features are calculated based on the section association weights: in, This represents the topologically weighted adjacent section characteristics. Indicates the recent trend of adjacent cross sections. This indicates a positive number that prevents the denominator from being zero.

[0058] When there are no qualified adjacent sections, let By setting the vector to zero and the availability mask of adjacent sections to zero, the model degenerates into single-section prediction. Given the current limited number of sections, deterministic weighted features are preferred; graph message passing is only used when the validity is verified by checking the reserved sections.

[0059] Preferably, This represents the predicted relative deformation of the segment after incorporating topologically weighted adjacent section features.

[0060] Furthermore, after outputting the predicted relative deformation corresponding to the construction scenario, the predicted time interval, and the unique identifier of the rock mass segment through the segment prediction model based on the model input and the predicted time interval, the method further includes: Using the segment prediction model, the upper quantiles of the absolute values ​​of the relative deformation, relative deformation rate, and relative deformation acceleration of the segment are generated for each construction scenario and the corresponding prediction time interval. The predicted distributions of the relative deformation, relative deformation rate, and relative deformation acceleration of the segment are obtained using the output of the segment prediction model, and the upper quantiles of the absolute values ​​of the three are extracted as the basis for risk assessment. The upper quantiles are selected to conduct risk discrimination, which is used to assess the higher level of deformation risk under a given construction scenario, but is not used as the upper bound of the worst-case scenario.

[0061] When the upper quantile of the predicted absolute value of the relative deformation amount or the relative deformation rate of the section exceeds the corresponding preset deformation control value or preset rate control value, a risk warning is generated, including the unique identifier of the rock mass section, the construction scenario, and the expected time of exceeding the limit. A dynamic early warning threshold for the section rate is also set, and its calculation formula is: in, The dynamic early warning threshold representing the relative deformation rate of a section. This represents the hard control value of the relative deformation rate of the design section. This represents a dimensionless, non-negative coefficient. This indicates the width of the interval for predicting the absolute value of the segment rate. This represents a nonnegative coefficient with the same dimension as the rate. This indicates data quality. It should be noted that, within the same section, under the same construction scenario, and at the same prediction time, an early warning is generated when the upper quantile of the absolute value of the predicted rate exceeds this early warning threshold; the determination of exceeding the limits of the design hard control value is reserved separately.

[0062] Data quality can be calculated using the following formula: in, Indicates the effective channel rate. Indicates the time elapsed since the most recent valid observation. This represents the maximum allowable interval from the nearest valid observation. Indicates the recent anomaly labeling rate. , , These represent the non-negative weights of the data quality score. It is worth noting that a 30-day window is used when there are at least 10 valid observations within the most recent 30 days; if there are fewer than 10, the window is extended backward to the time range covered by the most recent 10 valid observations; if there are still fewer than 10 historical observations, existing records are used and the data quality score is lowered. The worse the data quality or the wider the prediction interval, the lower the early warning line, but it must not replace the designed hard control value.

[0063] Furthermore, after generating the risk warning containing the unique identifier of the rock mass section, the construction scenario, and the expected time of exceeding the limit, the process also includes: The pre-support acceptance status is obtained. When the pre-support acceptance status is "acceptance completed," a continuous preset number of prediction time periods are selected within each construction scenario. Each prediction time period satisfies the condition that the upper quantile of the absolute values ​​of the relative deformation, relative deformation rate, and relative deformation acceleration of the segment does not exceed the preset deformation control value, preset stability rate threshold, and preset stability acceleration threshold, respectively. First, the pre-support engineering acceptance log is read. Rock stability condition assessment is only allowed if the pre-support has been accepted. Then, within all prediction time periods of the corresponding construction scenario, windows satisfying the continuous preset number of prediction time periods are searched. Within these windows, the upper quantile of the absolute values ​​of the relative deformation, relative deformation rate, and relative deformation acceleration of the segment must simultaneously be less than or equal to the preset deformation control value, preset stability rate threshold, and preset stability acceleration threshold, respectively. By setting rigid pre-conditions for pre-support acceptance and simultaneously using multiple indicators (deformation, rate, and acceleration) for continuous stability verification, the stability of the surrounding rock is avoided before the support is completed, and the misjudgment of the surrounding rock's stability based on a single indicator is also avoided.

[0064] Right now: ; ; ; in, The upper quantile of the predicted absolute value of the relative deformation rate of the control section. This represents the predicted upper quantile of the absolute value of the cumulative relative deformation in the control section. This represents the upper quantile of the predicted absolute value of the relative deformation acceleration in the control section. This indicates the preset stable rate threshold. This indicates the preset deformation control value. This indicates the preset stable acceleration threshold. This indicates whether the prescribed pre-support measures have been completed and accepted.

[0065] The earliest start time among the selected continuous prediction time periods is determined as the start time of the candidate time window for the stability of closed support or secondary lining under the corresponding construction scenario; when the above conditions are met, the start time is defined as follows: When there is no time that satisfies the condition within the prediction interval, only output the candidate window that has not yet formed a stable window, without forcing the date to be given.

[0066] Furthermore, the segment prediction model adopts one of the following: state-space model, regularized regression model, tree model, or time-series network model. When the number of available monitoring samples for the project is small, a state-space model, regularized regression model, or tree model with controllable parameters is selected to avoid overfitting under a large model with a small sample size. When sufficient monitoring samples have been accumulated, a time-series network model can be selected to improve the fitting ability. The core protected object of this invention is the complete technical processing link. The model is only used as the implementation carrier for segment deformation mapping. The scope of protection is not limited by a specific model structure or model name, thereby improving the adaptability and implementation capability of this method in underground engineering under different data conditions.

[0067] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for predicting surrounding rock deformation scenarios based on the state of the anchored section of a multi-point displacement gauge, characterized in that, Includes the following steps: The cumulative displacement of the same multi-point displacement gauge is obtained, the rock mass section is defined by the adjacent anchoring points, and the synchronous displacement difference of the adjacent anchoring points is calculated under the unified displacement direction and cumulative benchmark to form the anchoring section state with a unique identifier of the rock mass section. Construct an effective connectivity diagram along the tunnel axis formed at the construction time, calculate the path length from each reachable working face to the monitoring section of the rock mass section along the effective connectivity diagram, and combine the working face trajectory and support events to form the excavation-support event coordinates with working face identification. The status of the anchorage section up to the current prediction time and the coordinates of the excavation-support event are combined to form historical conditions; Based on the planned edge formation time, planned working face trajectory, and planned support completion time for each construction scenario, update the future effective connectivity map and the coordinates of future excavation-support events for each predicted time interval to form future conditions. The historical conditions and each of the future conditions are input into the pre-trained segment prediction model to obtain the relative deformation prediction amount corresponding to the construction scenario, the prediction time interval, and the unique identifier of the rock mass segment.

2. The method for predicting surrounding rock deformation scenarios based on the state of the anchored section of a multi-point displacement meter according to claim 1, characterized in that, The process of acquiring the cumulative displacement of the same multi-point displacement gauge, defining the rock mass section by adjacent anchoring points, calculating the synchronous displacement difference of the adjacent anchoring points under a unified displacement direction and cumulative benchmark, and forming an anchoring section state with a unique identifier for the rock mass section includes: The cumulative displacement and anchorage depth of each anchorage point of the same multi-point displacement gauge are obtained. Adjacent anchorage points are determined in ascending order of anchorage depth. The adjacent anchorage points define the rock mass section and establish a unique identifier for the rock mass section. Based on the borehole axial and zero-point adjustment records, the cumulative displacement is converted to a unified displacement direction and cumulative reference. The effective cumulative displacements of adjacent anchor points with the same observation time or an observation time difference not exceeding the preset pairing time difference are paired, and the cumulative displacement of the shallow anchor points is subtracted from the cumulative displacement of the deep anchor points to obtain the synchronous displacement difference. The average relative deformation rate of the section is calculated based on the synchronous displacement difference and the actual time difference between adjacent effective observations. The synchronous displacement difference, the average relative deformation rate of the section, and the unique identifier of the rock mass section are associated to form the anchorage section status.

3. The method for predicting surrounding rock deformation scenarios based on the state of the anchored section of a multi-point displacement meter according to claim 1, characterized in that, The effective connectivity map is constructed based on the tunnel axis edges formed at the construction time. The path length from each reachable working face to the monitoring section of the rock mass section is calculated along the effective connectivity map. Combined with the working face trajectory and support events, excavation-support event coordinates with working face identifiers are formed, including: Extract the actual tunnel axis edges formed up to the construction time, map the monitoring section where the rock mass section is located and each working face to the tunnel axis edges, and construct an effective connectivity diagram; The passage status of each reachable working face relative to the monitoring section is determined based on the working face trajectory, and the direction coefficient is determined based on the passage status; Based on the effective connectivity graph and the direction coefficient, the path length and oriented path coordinates from each reachable working face to the monitoring section are calculated using the following formula: in, Indicates working face To the monitoring section The path length of the section in which it is located; Indicates monitoring section In an effective connected graph The mapping position in Indicates working face At any moment Location, Indicates that only time is used The path length operator for active valid edges; Indicates working face At any moment directional coefficient, Indicates working face The assigned path coordinates; The working face identifier of each reachable working face, the path length, and the coordinates of the oriented path are respectively used to form the excavation event sub-coordinates; The support status acting on the monitoring section is determined based on the range of action of the support event, and the support status and the sub-coordinates of each excavation event are combined to form an excavation-support event coordinate system with the working face identifier.

4. The method for predicting surrounding rock deformation scenarios based on the state of the anchored section of a multi-point displacement meter according to claim 3, characterized in that, The extraction of the actual tunnel axis edges formed up to the construction time, mapping the monitoring section where the rock mass section is located and each working face to the tunnel axis edges, and constructing an effective connectivity graph includes: The location of the working face at the construction time is determined based on the working face trajectory, and the monitoring section where the rock mass section is located is mapped to the tunnel axis model; When the working face is located inside the tunnel axis, a dynamic node is set at the working face to divide the tunnel axis into formed sub-edges and unformed sub-edges. Extract the formed sub-edges and the tunnel axis edges that have been fully formed up to the construction time as valid edges, and map each working face to the corresponding valid edge. Construct a valid connected graph based on the connection relationship of the valid edges.

5. The method for predicting surrounding rock deformation scenarios based on the state of the anchored section of a multi-point displacement meter according to claim 1, characterized in that, The process of constructing historical conditions from the state of the anchorage section up to the current prediction time and the coordinates of the excavation-support event includes: Using the current prediction time as the cutoff point, extract the segment records from the anchorage segment status that have been obtained up to the cutoff point and whose observation time is no later than the cutoff point; Extract event records from the excavation-support event coordinates that have been obtained up to the cutoff point and whose event occurrence time is no later than the cutoff point; According to the monitoring section corresponding to the unique identifier of the rock mass section, the section record is associated with the event record, and the working face identifier in the event record is retained; The associated segment records and event records, arranged according to the observation time and the event occurrence time, constitute historical conditions.

6. The method for predicting surrounding rock deformation scenarios based on the state of the anchored section of a multi-point displacement meter according to claim 1, characterized in that, The future effective connectivity graph and future excavation-support event coordinates are updated according to the planned edge formation time, planned working face trajectory, and planned support completion time for each construction scenario, forming future conditions, including: Obtain the construction plan versions for each construction scenario frozen at the current prediction time, and determine the planned edge formation time, planned working face trajectory, and planned support completion time; At each prediction time interval, the planned edge whose formation time is no later than the prediction time is combined with the effective connected graph at the current prediction time to form a future effective connected graph. Along the future effective connectivity graph, the future excavation event sub-coordinates of each reachable working face are formed according to the planned working face trajectory, and the support events whose planned support completion time is no later than the predicted time are set to the completed state; The sub-coordinates of each future excavation event and the corresponding support status are combined to form the coordinates of the future excavation-support event, which, together with the future effective connectivity graph, form the future conditions of the corresponding construction scenario.

7. The method for predicting surrounding rock deformation scenarios based on the state of the anchored section of a multi-point displacement meter according to claim 1, characterized in that, The step of inputting the historical conditions and each of the future conditions into a pre-trained segment prediction model to obtain the relative deformation prediction corresponding to the construction scenario, the prediction time interval, and the unique identifier of the rock mass segment includes: Between each of the construction scenarios corresponding to the current prediction time, the parameters of the pre-trained segment prediction model and the historical conditions remain unchanged; Based on the unique identifier of the rock mass section, the historical conditions are paired with each of the future conditions to form a model input that retains the working face identifier; Based on the model input and the prediction time interval, the segment prediction model outputs the relative deformation prediction amount corresponding to the construction scenario, the prediction time interval, and the unique identifier of the rock mass segment. The calculation formula is as follows: in, Indicates the time interval of future forecasts The predicted deformation amount afterward, Indicate output conditions Quantile segmental deformation mapping, Indicates as of The historical status of the period and the events that have occurred. Describing a scenario Future events that will be activated at the planned times within the prediction interval. Indicates the index of rock mass sections. Indicates the predicted time interval. Indicates quantile level, satisfying ; This represents the model parameters obtained through training.

8. The method for predicting surrounding rock deformation scenarios based on the state of the anchored section of a multi-point displacement meter according to claim 7, characterized in that, After the step of outputting the predicted relative deformation corresponding to the construction scenario, the predicted time interval, and the unique identifier of the rock mass segment through the segment prediction model based on the model input and the predicted time interval, the step further includes: Using the segment prediction model, the upper quantile of the absolute values ​​of the relative deformation, relative deformation rate, and relative deformation acceleration of the segment corresponding to each construction scenario and the prediction time interval is generated respectively. When the upper quantile of the predicted absolute value of the relative deformation amount or the relative deformation rate of the section exceeds the corresponding preset deformation control value or preset rate control value, a risk warning containing the unique identifier of the rock mass section, the construction scenario, and the expected time of exceeding the limit is generated.

9. The method for predicting surrounding rock deformation scenarios based on the state of the anchored section of a multi-point displacement meter according to claim 8, characterized in that, After generating the risk warning containing the unique identifier of the rock mass section, the construction scenario, and the expected time of exceeding the limit, the method further includes: Obtain the pre-support acceptance status. When the pre-support acceptance status is "acceptance completed", select a continuous preset number of prediction time periods in each of the construction scenarios. The prediction time periods all satisfy the condition that the upper quantile of the absolute values ​​of the relative deformation of the section, the relative deformation rate of the section, and the relative deformation acceleration of the section does not exceed the preset deformation control value, the preset stability rate threshold, and the preset stability acceleration threshold, respectively. The earliest start time among the selected continuous prediction time periods is determined as the start time of the candidate time window for the stability of closed support or secondary lining under the corresponding construction scenario.

10. The method for predicting surrounding rock deformation scenarios based on the state of the anchorage section of a multi-point displacement meter according to claim 1, characterized in that, The segment prediction model adopts one of the following: state-space model, regularized regression model, tree model, or time-series network model.

Images