Numerical simulation method for influence of multi-line stacked tunnel construction on tunnel deformation

Abstract

The application discloses a numerical simulation method for the influence of multi-line stacked tunnel construction on tunnel deformation, relates to the technical field of tunnel engineering monitoring, and is characterized in that the monitoring data and the construction log are matched on a unified mileage and time axis, the monitoring lag is corrected, and a staged excitation sequence is generated; a three-dimensional stratum-structure coupling model is constructed offline, unit stage disturbance is applied to each construction line in each stage, and a database influence kernel and an abnormal disturbance influence kernel template are extracted; a low-dimensional prediction expression is established by means of influence kernel superposition, an interaction coefficient diagram is recursively updated by means of monitoring feedback, and an abnormal kernel is switched and augmented when a residual shape anomaly occurs; the upper and lower bounds of the feasible region are predicted according to the updated coefficient, a construction parameter suggestion interval is output, and a closed loop is left. The method realizes rapid updating and early warning without rebuilding a high-fidelity model, improves the prediction consistency and abnormal explainability of multi-line stacked construction, and reduces the on-site calculation burden.

Description

Technical Field

[0001] This invention relates to the field of tunnel engineering monitoring technology, specifically a numerical simulation method for the impact of multi-line overlapping tunnel construction on tunnel deformation. Background Technology

[0002] With the construction of urban rail transit, there is an increasing number of new tunnels crossing, passing under, intersecting obliquely with, or closely paralleling existing operating tunnels. This often results in multiple new lines being constructed sequentially within the same spatial area, creating overlapping intersections. These projects are generally located within a built-up environment. Construction disturbances are transmitted and superimposed through stages such as soil unloading, shield tail gap formation, synchronous and compensating grouting, and material hardening, leading to settlement, convergence, and a redistribution of internal forces in the existing tunnel. Current engineering practice largely adopts a "numerical analysis-on-site monitoring-threshold early warning" technical approach: using methods such as three-dimensional finite element / finite difference to determine the impact range and deformation level of the construction, and setting up monitoring points for settlement, convergence, and strain for timely tracking and comparison to facilitate adjustments to construction parameters. This approach can serve as a reference for single disturbances or simple geometric relationships, but it still suffers from poor applicability and consistency in cases of multiple overlapping lines and continuous superposition of disturbances at multiple stages.

[0003] Specifically, the impact of multi-line overlapping construction is temporal and coupled: disturbances at different stages of different lines are intertwined in time, the response of existing tunnels exhibits gradual and abrupt changes, and is affected by the heterogeneity of the strata, groundwater conditions, the initial state of existing structures, and construction deviations. Existing numerical models need to be set based on the values ​​of some key parameters and stage boundaries, which are not easily identified before the start of the project. Monitoring data also suffers from problems such as large differences in sampling frequency, transmission delay, and insufficient representativeness of points, resulting in a certain time or space deviation between the monitoring response and construction events. Prediction results based on comparison and verification or parameter inversion fluctuate within a certain time range. High-fidelity three-dimensional simulation has a large computational load, making it difficult to continuously iterate and verify multiple disturbance combinations during the construction period, and unable to characterize the risk evolution of the continuous construction process. In addition, when atypical events such as grout leakage, over-excavation, or localized weak zones occur, the spatial distribution of deformation in existing tunnels may deviate from normal operating conditions. If the source of residuals cannot be explained in a timely manner, it may lead to delayed early warnings or incorrect judgments. If these issues are not addressed in time, they will increase the risk of excessive deformation, structural cracking, and leakage in existing tunnels, which may result in operational disruptions, construction delays, and compensation disputes.

[0004] Therefore, the current technical problem is that, under the condition of continuous construction disturbance of multi-line overlapping tunnels, existing technologies are unable to stably identify the impact contribution and make continuous and reliable predictions and judgments on the deformation of existing tunnels based on the strict spatiotemporal correspondence between monitoring data and construction stages. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] To address the shortcomings of existing technologies, this invention provides a numerical simulation method for the impact of multi-line overlapping tunnel construction on tunnel deformation. This method involves applying unit-stage perturbations to each stage of each construction line, extracting and storing influence kernels and anomalous perturbation influence kernel templates in a database, establishing a low-dimensional predictive representation using superimposed influence kernels, recursively updating the interaction coefficient graph using monitoring feedback, and switching to anomalous kernel augmentation when residual morphology is abnormal. Based on the updated coefficients, the method continuously predicts the upper and lower bounds of the feasible region, outputs suggested intervals for construction parameters, and leaves a closed-loop record. This method achieves rapid updates and early warnings without rebuilding a high-fidelity model, improves the consistency and anomaly interpretability of multi-line overlapping predictions, and solves the technical problems described in the background.

[0007] (II) Technical Solution

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] A numerical simulation method for the impact of multi-line overlapping tunnel construction on tunnel deformation includes: establishing a longitudinal coordinate axis and a time axis; mapping and registering existing tunnel monitoring data with construction log data; calculating the monitoring lag correction amount based on the process switching nodes and the inflection points of the monitoring curves; dividing the construction into stages and generating stage excitation sequences; establishing a three-dimensional stratum-structure coupled numerical model offline; applying unit stage disturbances to each construction stage of each construction line; calculating and solidifying the influence cores; forming an influence core library; and presetting abnormal disturbance influence core templates.

[0010] Based on the influence kernel library, a predictive expression is constructed and an interaction coefficient map is set. The interaction coefficients are recursively updated within a sliding time window or sliding ring number window using the aligned monitoring sequence. When the residual does not match the influence kernel, an abnormal disturbance influence kernel template is introduced and the abnormal kernel intensity coefficient is updated. The updated interaction coefficient map is used to perform rolling interval prediction, output the upper and lower bound curves of the feasible region, and generate suggested intervals for construction parameters and record keeping.

[0011] Furthermore, the monitoring lag correction is formed by resampling the monitoring data through conformal interpolation and together with the process switching event nodes to form an event-driven time grid. On this time grid, the alignment relationship that satisfies the monotonic constraint is obtained, and the aligned monitoring sequence is obtained by using the piecewise robust error criterion and outlier removal rule, which is used to generate the stage excitation sequence.

[0012] Furthermore, mapping and registration include establishing a monitoring point ledger and a construction event list, binding each monitoring point identifier to the mileage position, existing line identifier, and section identifier on the vertical coordinate axis, and extracting the construction log into a sequence of process switching event nodes with construction line identifiers according to the advancement ring number and timestamp, so as to form a unified field and a unified sampling rhythm.

[0013] Furthermore, the division of construction stages includes generating a sequence of stage labels for each construction line based on the process switching event nodes, and performing consistency verification at the stage boundaries in conjunction with the inflection points of the monitoring curves, so that only one stage label corresponds to the same moment; when the impact of shutdown overlaps with synchronous grouting or grout hardening, the stage label is determined first based on the stage of shutdown impact and the stage boundaries of the overlapping interval are sealed.

[0014] Furthermore, the stage excitation sequence is generated by fusing the surface pressure, synchronous grouting volume, synchronous grouting pressure, advance speed, shutdown and re-excavation time, and secondary compensation grouting records from the construction log segment by segment after boundary truncation and normalization. The original parameters before normalization are archived along with the stage excitation sequence for consistent use in offline modeling unit stage disturbance settings and kernel library indexes.

[0015] Furthermore, an index vector is established for the influence of the core library based on net distance, intersection angle, burial depth difference, soil cover thickness and stratum combination, and the model mesh version, boundary condition version and unit stage disturbance caliber are recorded for each index vector;

[0016] Weighted retrieval interpolation of the influence kernel is performed between adjacent samples of the index vector, and the influence kernels are all uniformly expressed using the vertical coordinate axis and include construction line and stage identifiers.

[0017] Furthermore, the unit stage disturbance is applied in the three-dimensional stratum-structure coupled numerical model according to the stage. The excavation unloading stage adopts the stress release boundary, the shield tail gap formation stage adopts the equivalent volume loss boundary, the synchronous grouting stage adopts the equivalent pressure boundary, the grouting hardening stage adopts the material parameter time sequence switching boundary, the secondary compensation stage adopts the equivalent pressure boundary, and the shutdown influence stage adopts the step sequence maintenance and intermittent load application boundary, and the stage identifier solidification influence core is recorded.

[0018] Furthermore, the templates for the impact core of abnormal disturbances include templates for grout leakage, over-excavation, local voids, sudden surge, and shutdown expansion. Each template and the impact core have the same vertical coordinate representation and stage identification field. The template is called based on the similarity test between the residual morphology curve and the template morphology to meet the threshold, and the corresponding abnormal event node in the construction event list is verified to meet the time neighborhood condition as the trigger condition.

[0019] Furthermore, the interaction coefficient diagram is a set of coupled edges established according to "construction line-existing line-stage". An interaction coefficient is set for each coupled edge and updated synchronously within the sliding ring number window during the sliding time window. During the update process, upper and lower bound constraints, change range constraints, and shrinkage rules for coupled edges far from the overlapping area are applied to the interaction coefficient.

[0020] Furthermore, the residual is obtained by subtracting the corresponding output of the predicted expression from the aligned monitoring data, and the residual morphology curve is reconstructed along the vertical coordinate axis; when the residual morphology curve is inconsistent with the typical morphology of the kernel library, the abnormal perturbation affecting the kernel template is switched and the abnormal kernel strength coefficient is updated synchronously.

[0021] Furthermore, the rolling interval prediction uses the time period corresponding to several future rings as the prediction interval. Based on the interaction coefficient diagram and its upper and lower bounds, it generates the upper boundary curve and lower boundary curve of the feasible region. The construction event nodes within the prediction interval are included in the stage excitation sequence to complete the continuous calculation within the interval. At the same time, the mileage section sequence corresponding to the vertical coordinate axis is output as the result index.

[0022] Furthermore, the suggested intervals for construction parameters are obtained by conducting small-scale trial calculations on the surface pressure, synchronous grouting volume, synchronous grouting pressure, and advancement speed within the influence core superimposed frame. During the trial calculations, the stage label sequence is kept unchanged, and the input parameters of the stage excitation sequence are adjusted segment by segment to form intervalized suggested values ​​corresponding to the prediction intervals.

[0023] Furthermore, the entire process record includes the definition of the longitudinal coordinate axis, the definition of the time axis, the monitoring lag correction amount, the stage label sequence, the stage excitation sequence, the interaction coefficient diagram, the anomaly core strength coefficient and the suggested range of construction parameters, and triggers offline review to update the impact core library when entering critical overlapping mileage sections or when anomalies occur frequently.

[0024] (III) Beneficial Effects

[0025] This invention provides a numerical simulation method for the impact of multi-line overlapping tunnel construction on tunnel deformation, which has the following beneficial effects:

[0026] Monitoring data and construction log data are registered on the same vertical coordinate axis and time axis. Stage label sequences and stage excitation sequences are generated using monitoring lag correction. Process switching and deformation responses are clearly correlated, reducing spatiotemporal mismatch. A three-dimensional stratum-structure coupling model is established offline, and an influence kernel library is solidified. The distribution shape of unit stage disturbance responses for each stage of each construction line is stored and retrieved using an index, ensuring reusable simulation shape information and consistent calculation caliber. A low-dimensional predictive expression is obtained by overlaying influence kernels, and an interaction coefficient diagram is plotted. Boundary, sparsity, and smoothing constraints are applied to the interaction coefficients using a sliding time window or sliding ring number window, and updates are recursively applied. Uncertainty is centrally reflected in updatable state quantities, reducing constitutive recalibration. An anomalous disturbance influence kernel template is set, and anomaly kernel strength coefficients are introduced for augmented updates when residual shapes do not match. Anomalous contributions such as grout leakage, over-excavation, voids, sudden surges, and downtime amplification are distinguished from conventional stage contributions and are easily interpreted.

[0027] Rolling interval prediction is performed based on the updated interaction coefficient diagram to obtain the upper and lower bound curves of the feasible region. The suggested intervals of construction parameters are obtained according to the sensitivity of the influence kernel. It can map surface pressure, synchronous grouting volume, pressure, advance speed and compensation grouting trigger conditions. The whole process is recorded and offline review and update of the influence kernel library is triggered at the overlapping mileage section. The stage excitation sequence, influence kernel library and interaction coefficient diagram interact and iterate in the closed loop of suggestion, execution, monitoring feedback and re-update, which enhances traceability and reusability. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the numerical simulation method for the impact of multi-line overlapping tunnel construction on tunnel deformation according to the present invention;

[0029] Figure 2 This is a schematic diagram of the overall process structure of the present invention;

[0030] Figure 3 This is a schematic diagram illustrating the spatiotemporal alignment principle of monitoring data based on process nodes in this invention.

[0031] Figure 4 This is a schematic diagram illustrating the generation and normalization of the stage-specific excitation sequence in this invention.

[0032] Figure 5 This is a schematic diagram illustrating the principle of kernel extraction for unit-stage perturbation in this invention.

[0033] Figure 6 This is the interactive coefficient graph structure and edge weight update logic diagram of the present invention;

[0034] Figure 7 This is a flowchart of the abnormal disturbance identification based on residual shape matching in this invention;

[0035] Figure 8 This is a schematic diagram of the interval prediction propagation and suggested interval inversion of the present invention. Detailed Implementation

[0036] 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 some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] Please see Figures 1-8 This invention provides a numerical simulation method for the impact of multi-line overlapping tunnel construction on tunnel deformation, including step one: aligning existing tunnel monitoring data and construction log data on the same vertical coordinate. On the same timeline A traceable, consistent correspondence is formed, and a phased activation sequence is generated accordingly. This allows step two to be identified as a single stage. Define the unit stage perturbation and generate the influence kernel.

[0038] First, establish a unique mapping between the monitoring records, including who collected them, which cross-section they correspond to, which construction line they belong to, and which ring they belong to, before proceeding with the alignment calculation. Common on-site diameter differences, if directly incorporated into the alignment process, will be misinterpreted as lagging changes; therefore, verifiable locking must be completed before data entry.

[0039] During the period of multiple overlapping lines, there are many monitoring points and frequent team records. The names of monitoring points, ring numbers and process labels are prone to having the same name but different meanings or the same meaning but different names. This causes the same physical object to be split into multiple objects at the data layer, causing cross-line errors in subsequent alignment and stage incentive generation.

[0040] Assign a monitoring point identifier to each monitoring point It is formed by piecing together the existing tunnel mileage, cross-sectional circumferential location, and burial type; and a construction line marker is assigned to each construction line. Construction event identifiers are formed using ring numbers as indexes. And write shutdown and re-excavation as event states bound to the ring number. Each monitoring record must include a collection timestamp, monitoring point identifier, and measurement type; each construction log must include a construction line identifier, ring number, and process label. When fields are missing, a supplementary entry list is generated and sent back to the field recording end. Alignment only proceeds after all fields are completed. Simultaneously, the unit conversion and range boundaries of the monitored quantities are fixed upon data entry: unified units and symbols are assigned to the four types of measurements—settlement, convergence, strain, and internal force. A dimensional check is performed before data entry; if abnormal units are found, they are not included in the alignment process but are sent back for verification, preventing subsequent misinterpretation of unit differences as abrupt responses.

[0041] The location of monitoring points is written into the monitoring point mapping table using mileage + circumferential angle + radius, and the same monitoring point will not be repeated during on-site verification; the construction log is written into the construction event mapping table using ring number monotonically and non-repeating, and multiple process nodes of the same ring will not be split into different indexes.

[0042] Mark construction events The process nodes are divided into: excavation unloading start point, shield tail gap formation start point, synchronous grouting start and end point, grouting hardening start and end point, secondary compensation start and end point, and shutdown start and end point. A specified stage label sequence is also provided. For piecewise constant functions: at any time Only one stage marker is allowed. And the stage boundary must fall at the process node (i.e. The set of transition points is a subset of the set of process nodes.

[0043] When two process nodes overlap (e.g., downtime and hardening periods overlap), a priority is defined: the downtime impact phase covers other phases, and the hardening phase covers the non-downtime section after synchronous grouting is completed, to avoid duplicate counting. Monitoring point identification is used. Construction event sign The unique mapping provides input boundaries for subsequent alignment, preventing caliber differences from being treated as physical lags; and ensures that the stage stimulus input has a stable field source, thereby guaranteeing semantic consistency in subsequent impacts on the core index.

[0044] By fixing the spatial and temporal reference system to a single caliber and establishing an alignment relationship between process nodes and monitoring response characteristics, the monitoring sequence can be rearranged to the process-driven time. This alignment emphasizes monotonicity and verifiability, avoiding the use of subjective time windows to replace process boundaries. Specifically, construction logs are organized by ring numbers while monitoring is organized by timestamps. Furthermore, existing tunnel responses exhibit physical lag and acquisition delays. Directly comparing at the same moment would misjudge the lag as parameter drift, thereby affecting the stability of subsequent recursive updates.

[0045] Establish longitudinal coordinates based on existing tunnel mileage Calculate the longitudinal coordinates of each monitoring point. And write it into the monitoring point mapping table; establish a timeline with a unified timestamp. Identify each construction event Assign time to events The five process nodes—start of tunneling, start of synchronous grouting, end of synchronous grouting, start of shutdown, and end of shutdown—were solidified into node sets. Subsequently, a local cubic polynomial fitting was performed on the monitoring sequence within a sliding time window, and the derivative was calculated to extract the slope abrupt change points as response feature moments. The time window for the local cubic polynomial fitting was determined by covering at least one complete process segment, meaning the window spans the interval from the start to the end of a single synchronous grouting operation. This ensures that the calculation of the rate of change is consistent with the slope change at the process boundary, reducing derivative oscillations caused by sparse sampling. Finally, a monotonic alignment mapping function was constructed based on the principle of aligning process node moments to response feature moments. , and by Obtain the lag function; finally press The monitoring sequence is resampled to obtain an aligned monitoring sequence. .

[0046] Among them, the alignment mapping function Discrete monotonic path solution is adopted: the time window is discretized into nodes, and the mapping path is restricted to only forward and not backward. Huber loss is used to evaluate the consistency between the monitored value and the process template after mapping. Huber loss is penalized with squares in the small error region and with linear penalty in the large error region. After the discrete mapping points, piecewise cubic Hermite interpolation is used to restore the continuous mapping.

[0047] Alignment mapping function The specific format is as follows: First, obtain discrete alignment point pairs; then, take the event-driven mesh on a unified time axis. (Including all process switching event nodes and supplementary points), retrieve the corresponding aligned time from the original timestamp system being monitored. (Satisfies monotonicity and boundary constraints). Among them... It is obtained through "monotone minimum cost registration". This connects discrete point pairs into a continuous mapping: for any Define the alignment mapping as a piecewise linear function.

[0048]

[0049]

[0050] And stipulate It is monotonically non-decreasing and its value falls within the monitoring timestamp range.

[0051] To ensure on-site operability, both monitoring and construction ends record the same time marker as an anchor point during shift handover. This anchor point is used for manual verification of the alignment results and does not participate in the mapping solution. Example implementation: After the night shift ends, the recorder exports the shift's construction log and completes the five types of node times in the construction event mapping table; the monitor reads the mileage markers of the monitoring point sections along the existing tunnel and corrects the monitoring point mapping table. Technicians then generate... Subsequently, it becomes possible to directly view on the same graph whether the synchronous grouting start node of a certain ring is consistent with the inflection point of the settlement curve of the corresponding section, thus transforming subsequent judgments from guessing the correspondence to verifying the alignment relationship. To ensure that subsequent alignment mapping can be repeatedly executed, a ring event-driven grid is adopted for the unified sampling grid: sampling points are forcibly retained at each construction event node, and points are added between adjacent nodes according to the minimum sampling interval of the monitoring system; the point addition adopts two paths: linear interpolation or piecewise cubic Hermite interpolation. The former is used for stable segments with fewer missing points, and the latter is used near inflection points to maintain slope continuity. The interpolation selection rules and alignment results are recorded together for verification.

[0052] When used, a unified reference frame and monotonic alignment mapping explicitly write the physical hysteresis. , to align the monitoring sequence It can be directly used as the observation input for subsequent recursive updates; at the same time, the alignment process remains monotonic and verifiable, avoiding confusion between recording delay and structural response.

[0053] Based on the aligned monitoring sequence, the controllable parameters in the construction log are converted into stage identifiers. A one-to-one corresponding stage excitation sequence enables the intensity input of the same stage to be compared between different construction lines and ensures that the stage boundaries remain consistent throughout the entire process. The impact of multi-line overlap comes from the superposition of multiple processes, and the main control parameters are different in different stages; if the stage boundaries drift or the excitation is unbounded, it will lead to a mismatch between the unit stage disturbance definition in step two and the interaction coefficient update in step three.

[0054] Phase markers The entire process is finalized before construction begins according to the construction plan, covering six stages: excavation and unloading, shield tail gap formation, synchronous grouting, grout hardening, secondary compensation, and downtime impact. It is stipulated that at any given time, only [the following steps are permitted]. Choose one stage to avoid duplicate incentive calculations caused by concurrent stages. Based on construction event identifiers. The process nodes, the timeline The stage labels are written as a sequence of stage labels. Select surface pressure sequence Grouting volume sequence Propulsion speed sequence With stop flag sequence As the stage input, the input is truncated according to the upper and lower bounds of the device's allowable range and then monotonically normalized; the normalized input is then fused according to the stage weights to obtain a bounded stage excitation sequence. And use indicator functions to confine the excitation within the corresponding stage; check before and after each stage switching point. The continuity and traceability of fields are ensured, and in case of anomalies, a retrospective check is performed and supplementary entries are triggered.

[0055] The following incentive methods can be implemented directly:

[0056]

[0057] Where: Stage excitation sequence Construction line markings In the stage marker The excitation intensity, with a value range of [value range missing]. Timeline : A time variable formed by a unified timestamp, whose value is the valid period for monitoring and logging; construction line number. Construction line markings The index number, whose value is a discrete set; stage identifier. : Pre-cured construction stage set elements, with value range being a finite discrete set; surface pressure sequence : Surface pressure or soil chamber pressure recording sequence, with values ​​taken as positive values ​​within the allowable range of the equipment;

[0058] Grouting volume sequence : Synchronous grouting volume recording sequence, with values ​​taking non-negative values ​​within the allowable range of the grouting system, representing the grout compensation input; Advancement speed sequence : Propulsion speed recording sequence, with values ​​taking non-negative values ​​within the equipment's allowable range, used to characterize the time density of disturbance input; Stop flag sequence : Stop state sequence, with values ​​of Its function is to introduce the shutdown persistence effect into the incentive;

[0059] Stage weight parameters : Corresponding stage identifier The fixed weights, taking values ​​of finite real numbers and fixed before construction begins, serve to reflect the differences in main control inputs at different stages; compression function Mapping real numbers to The function, specifically in the form of Its function is to generate bounded stimuli to facilitate subsequent constraints; indicator function : When the stage label sequence Take stage marker The value is 1 if the incentive is active and 0 otherwise, ensuring that the incentive only takes effect in this phase.

[0060] Phase label sequence Construction line markings On the timeline The stage labels on the page have a value range that is the set of stage identifiers. It is a piecewise constant function, and any Take only one stage identifier The stage boundary set is a subset of the event node set. When the impact of shutdown overlaps with other stages, the following priorities are specified: the stage with the shutdown impact takes precedence; the hardening stage takes effect after synchronous grouting is completed and in non-shutdown sections; the secondary compensation stage takes effect between compensation grouting event nodes.

[0061] Stage weight parameters The given parameters can be set in two ways: one is by directly setting the priority of the control parameters in the construction organization design and having it approved by the chief engineer; the other is by using the aligned monitoring sequence in the trial excavation section. For reference, a least squares solver with boundaries is used to calibrate and freeze the weights in one go. The solver can be a sequential quadratic programming method or an interior point method, and it is only executed once in the trial excavation section to avoid overlapping meanings with the subsequent interaction coefficient diagram.

[0062] An alternative equivalent implementation is as follows: when the existing tunnel is primarily measured using convergence meters, Selected as a convergent sequence and put The normalized upper bound is solidified separately for convergence-sensitive sections; when strain gauges are the primary measurement, Select as strain sequence and put The weighting coefficients are increased to enhance the stage differences in grouting input.

[0063] When used, the phase label sequence With stage incentive sequence The construction log is rewritten into a computable input and kept bounded, so that subsequent stage indexing, unit perturbation definition and recursive update can run under the same semantics; and the incentives are strictly bound to the stage to avoid different processes being repeatedly counted into the same stage input.

[0064] Step 2: Using the vertical coordinates already unified in Step 1 Timeline Phase identifier Under the specified caliber, the unit disturbance response shape of the construction line-existing line-stage is calculated and solidified offline, enabling step three to use the same stage excitation sequence. The magnitude of the influence kernel is superimposed without introducing semantic drift.

[0065] By defining geometric relationships, stratigraphic layers, structural features, and stage boundaries as verifiable computational premises, each subsequent unit stage disturbance starts from the same reference state. In multi-line overlapping projects, if different examples use different mileage zero points, different stage divisions, or different structural contact assumptions, it will lead to the incomparability between influence cores, resulting in a lack of uniform scale in the interaction coefficient diagram of step three. The monitoring point identification has already been provided in step one. vertical coordinate Aligned monitoring sequence Stage label sequence With stage incentive sequence These outputs actually form the benchmark for how engineering events are segmented and responses are aligned; if the model is not built based on this benchmark in step two, the meaning of each stage will be misaligned, causing the disturbance of a unit stage to no longer correspond to the same construction segment. First, using the longitudinal coordinates of the existing line... As the principal axis of the model, the existing tunnel axis is projected onto... It also sets the zero point of mileage and direction, allowing any node to pass through. Unique location; then construction line marking. The axes are also projected onto the same coordinate reference to ensure that the net distance and intersection angle are obtained through geometric calculations within the model rather than manual input; then, the stratigraphic layers are written into the model according to the geological survey data, and the layer interfaces and groundwater boundaries are solidified into traceable input files; finally, the stage markers are added. The set and phase switching rules completely reference the phase label sequence in step one. The start and end event nodes of each stage will be identified by construction event identifiers. The process nodes are derived to ensure that the stage boundaries remain consistent between two steps.

[0066] The three-dimensional coupled model is constructed using a combination of stratum continuum elements and tunnel structure shell elements or beam-shell composite elements. Contact relationships are established at the tunnel-stratum interface to allow shear transfer and normal compaction. For the grouting hardening stage, an equivalent stiffness sequence evolving over time is set in the material properties, with the time marker using the time axis from step one. This binds the hardening process to the same event sequence. The solver type can be an implicit static solver to achieve stable convergence, or an explicit dynamic solver to express the transient response caused by short-term process switching. When using an explicit dynamic solver, the time integration step size is determined by the relationship between the grid scale and the wave velocity, and the state variables are forcibly output at the process node at the stage switching point so that the influence kernel can be extracted later.

[0067] When used, the three-dimensional coupled model is as follows: , , To ensure a common standard, the stage definition and data alignment results from step one are directly mapped to the offline calculation boundary, making the influence kernels comparable under different geometric indices; and enabling subsequent kernel library retrieval to be aligned with monitoring point identifiers. of Direct correspondence.

[0068] Furthermore, the complex construction process is broken down into repeatable unit-stage disturbances, and the existing tunnel is moved along the longitudinal coordinate. The response shape is solidified into an influence kernel, enabling it to interact with the stage excitation sequence of step one. In step three, they are multiplied directly. The shape of the nucleus must simultaneously satisfy stage specificity and spatial sampleability; otherwise, nuclei from different stages may contaminate each other or fail to correspond to the monitoring section.

[0069] In the case of multiple overlapping lines, the response of the existing tunnel comes from the superposition of multiple stages. If the offline calculation directly outputs a single response curve as a full-process simulation, it will not be able to be decomposed by the interaction coefficient diagram in step three. Therefore, it is necessary to extract the typical response shape of each stage from the high-fidelity model and give it a clear normalization scale.

[0070] Therefore, a baseline state is first established, defined as the state after the existing tunnel structure and the strata have achieved initial stress equilibrium, and before any construction lines are applied. The state during any stage of disturbance; and then the marking of a certain construction line. A certain stage identifier The unit stage disturbance is constructed. The unit stage disturbance is applied with a stage excitation amplitude of 1. That is, the input of the stage is normalized to the unit amplitude according to its physical meaning. For example, the excavation unloading stage is represented as the unit unloading ratio, the synchronous grouting stage is represented as the unit equivalent pressure or unit equivalent volume, and the shield tail gap stage is represented as the unit gap volume release.

[0071] Then, the existing line number is obtained by solving. The longitudinal response field is obtained, and longitudinal response is sampled along the existing axis at preset sampling intervals. Simultaneously extract the longitudinal response under the baseline state. Finally, the influence kernels are solidified according to the normalized scale and written into the kernel library record. The influence kernels are normalized using difference quotients to maintain unit consistency.

[0072]

[0073] In the formula: Influence kernel Construction line markings In the stage marker Below are the existing line numbers The shape of the unit disturbance response, with values ​​in the real number range; the vertical axis... The coordinates of the mileage along the existing tunnel are taken as the range of the existing tunnel mileage. They serve as the spatial independent variable defining the kernel and are related to the longitudinal coordinates of the monitoring points. Alignment; Construction line number Construction line markings The index number, with a value range that is a discrete set; existing line numbers : Existing tunnel number index, with values ​​taken as a discrete set;

[0074] Phase markers Step 1 solidifies the set of elements for the construction stage, whose value range is a finite discrete set; Longitudinal response : Applying construction line markings Stage markers After the unit stage disturbance, the existing line number The longitudinal response is a real number; the reference longitudinal response... : Existing line number when no stage of disturbance is applied The longitudinal response, with values ​​ranging from real numbers, serves to provide a repeatable reference baseline to eliminate differences in initial states;

[0075] Return to a value Construction line markings In the stage marker The unit-stage perturbation normalization scale, with a range of positive real numbers and fixed in the kernel library, serves to unify the amplitude caliber of perturbations at different stages and ensure... Comparability;

[0076] Among them, the method of applying the per unit stage disturbance is solidified in the core library as a disturbance load script: for the unit unloading ratio, the load script acts on the initial stress of the excavation zone unit in the form of a release coefficient; for the unit equivalent pressure, the load script acts on the equivalent boundary of the shield tail grouting influence zone as normal pressure; for the unit equivalent volume, the load script acts on the grouting influence zone unit in the form of equivalent volume strain.

[0077] For example, in the construction organization of a subway section that passes under an existing operating tunnel, the technicians first establish a system based on the mileage of the existing tunnel. And mark the monitoring points of Write the data into the mapping table, and then select the synchronous grouting stage as the stage identifier in the offline calculation. The unit stage disturbance of this stage is applied to the shield tail grouting influence zone, and the existing line number is obtained by solving. The longitudinal settlement response field; the visible result on site is that a longitudinal settlement response field can be extracted along the direction of the existing tunnel. The changing settlement shape curve exhibits a distinct peak-valley transition near the overlapping intersection point, gradually diminishing further away. Based on this, technicians solidified this shape as an impact nucleus and registered its corresponding construction line marker. Phase identifier This allows for adjustments to the stage excitation sequence only during subsequent construction. The amplitude can reproduce the shape of the spatial effect on the existing line at this stage.

[0078] As a supplement: regarding the excavation and unloading phase... The unit unloading ratio is set to 1; that is, the proportion by which the disturbance script directly releases the initial stress in the specified excavation area. For the equivalent pressure diameter of synchronous grouting, The unit pressure is set to 1 (using a consistent unit system); the disturbance script applies a unit normal pressure to the boundary of the grouting influence zone. For the equivalent volume diameter of synchronous grouting, The strain per unit volume is taken as 1; the disturbance script applies a strain source term per unit volume to the elements in the grouting influence zone. For the shield tail gap stage, The unit gap release amount is taken as 1; the disturbance script is applied according to the equivalent volume loss.

[0079] In practice, the unit-stage perturbation breaks down the entire process into reusable shape inputs specific to each stage, enabling step three to differentiate the contributions of different stages using an interaction coefficient graph; and by normalizing the values... Standardizing the kernel scale allows kernels from different construction lines and stages to be compared within the same calculation framework.

[0080] Furthermore, the impact kernel is transformed from a single calculation result into a searchable and interpolable library. This allows step three to call upon the closest kernel when faced with different net distances, intersection angles, burial depth differences, and overburden conditions, avoiding repeated high-fidelity recalculations during construction. The indexes and interpolations must be publicly disclosed in their specific forms; otherwise, the kernel library will be difficult to implement.

[0081] Among them, the multi-line overlapping working conditions vary significantly in spatial relationships, which affects the verification of geometric parameters and geological conditions. If only discrete cases are stored without providing interpolation rules, the kernel library can only cover a small number of working conditions and cannot support continuous changes in engineering.

[0082] The net distance, intersection angle, burial depth difference, overburden thickness, and stratigraphic type are expressed as parameter vectors. And record the parameter vector for each discrete case. With corresponding influence core When it is necessary to specify a parameter vector When calling the kernel, the distances to several nearest neighbor instances are first calculated in the parameter space, then weights are generated based on these distances, and finally, the nearest neighbor kernels are weighted and reconstructed to obtain the interpolation kernel. The interpolation form is as follows:

[0083]

[0084]

[0085] In the formula: interpolation influence kernel In the parameter vector The influence kernel obtained from the reconstruction has a value range of real numbers and serves as the shape input in step three to cover continuous operating conditions; parameter vector : A vector describing geometric and geological conditions, with its value range limited by engineering conditions; Discrete example parameter vector : No. The parameter vector of each discrete example takes the value of the kernel library coverage range and the number of nearest neighbors. The number of nearest neighbor examples participating in the interpolation, a positive integer, is fixed in the kernel library rules and serves to balance interpolation smoothness and locality; interpolation weights. : No. The weights of the nearest neighbor examples, with values ​​ranging from 1 to 2. And the sum of the weights is 1;

[0086] distance Within the parameter space arrive The scaled Euclidean distance, a positive real number; power exponent. Distance weight decay exponent, a real number greater than 0, is fixed in the kernel library rules; scale matrix. A diagonal matrix with standardized dimensions for the parameter dimensions, consisting of diagonal elements; kernel sample. : No. The influence kernel corresponding to each discrete case is a real number and serves as the basis function for interpolation reconstruction.

[0087] Where the parameter vector The composition is fixed in the kernel library record as Net distance The intersection angle is a positive real number. for Inner curvature value, burial depth difference For real numbers, the thickness of the soil cover is... A positive real number, stratigraphic type index The values ​​are discrete integers and are given by the stratigraphic grouping rules; the scale matrix. The diagonal elements are given by the reciprocal of the allowable range of each dimension parameter, thus avoiding the dominance of a distance by a single parameter due to its large dimension. An optional engineering implementation path is: when the kernel library examples form a regular grid in the parameter space, nearest neighbor selection uses a method surrounding... The vertices of the hyperrectangular cells are identified, and their weights are calculated using multilinear interpolation. When the distribution of kernel library instances is irregular, nearest neighbor selection is performed using KD-tree retrieval and the weights are calculated according to the above weight formula.

[0088] In use, the index structure and interpolation rules extend the discrete kernel library to continuous operating conditions, ensuring that the influence kernel called in step three is consistent with the actual geometric conditions of the project; furthermore, this is achieved through the scaling matrix. The unification of dimensions ensures that weight calculation is not dominated by the dimension of a single parameter.

[0089] Furthermore, pre-configured kernel templates for anomalous perturbations that can be invoked in step three are included in the kernel library. This allows for rapid selection of an interpretation path and augmentation when the monitored residual shape does not match that of a typical kernel. The key to anomalous perturbation templates is that their shapes are identifiable, their triggers are verifiable, and their registration is traceable; otherwise, anomalous kernels will be confused with typical kernels.

[0090] Events such as grout leakage, over-excavation, local cavities, sudden surges, and shutdown amplification can alter soil constraints and load transfer paths, causing local peaks, double peaks, or long-tail attenuation in the deformation shape of existing tunnels. If only typical stage kernels are relied upon, abnormal responses will be misassigned to the interaction coefficient map, leading to interaction coefficient drift and weakening the interpretability of subsequent predictions.

[0091] First, write the abnormal scenarios as feasible equivalent disturbances. For example, the grout leakage scenario is written as an instantaneous decrease in stiffness and volume loss in the grouting affected area; the over-excavation scenario is written as an expansion of the excavation section and an increase in the unloading range; the local void scenario is written as a local volume void and a weakening of the support; the sudden surge scenario is written as a sudden change in the pore pressure boundary and a decrease in the effective stress; and the shutdown and scale-up scenario is written as an interruption of the stage excitation and an extension of the hardening sequence.

[0092] Then, in the high-fidelity model, a unit anomalous perturbation is applied to each anomalous scenario and solved. The anomalous response shape is extracted along the existing line axis and solidified into an anomalous perturbation influence kernel template according to the normalized amplitude value. Finally, each template is registered with its trigger condition description. The trigger condition adopts the residual shape features that can be directly calculated in step three, such as local kurtosis, peak position shift and long tail attenuation direction, so as to ensure that step three can call the template without introducing undefined information.

[0093] As a supplement, at a certain moment in the window segment Calculate the residual for each monitoring point. The residuals are sorted according to the existing line number. Aggregate and interpolate on the vertical axis to form a continuous residual shape. The interpolation method is the same as in step one, and piecewise cubic Hermite interpolation can be used with the longitudinal coordinates of the monitoring points as the interpolation method. For each node, number it according to the exception type. Calculate the normalized correlation similarity:

[0094]

[0095] Similarity The range of values ​​is Used to measure the consistency between the residual shape and the template shape; residual shape : Real number function; derived from monitoring residuals along Interpolation is used to match input; exception kernel template. : Real number function; Preset by step two; Used for matching the benchmark; Integration variable : Vertical coordinate; Domain is the kernel coverage area; Used for shape inner product calculation; Triggering rule: When ( The preset threshold value range is: When the configuration is solidified in the engineering setup, a second verification is performed; the second verification involves checking... Is there an anomaly type number nearby? The corresponding construction event nodes (such as shutdown start / stop, pressure interruption, grouting anomaly record, over-excavation alarm, etc.) are on the timeline. Within the neighborhood of the above; if it exists, the introduction of the exception template is allowed.

[0096] In this context, the abnormal kernel template and the typical kernel are on the same vertical coordinate. On the same sampling grid, and record the corresponding anomaly type number. And return a picture value When an anomaly occurs on site, technicians can modify the time of the anomaly in the construction log as a manual anchor point for verification.

[0097] When in use, abnormal perturbations affect the kernel template in the same way. Mesh solidification enables step three to introduce anomaly interpretation paths through residual shape matching without damaging typical kernel scales; trigger condition registration provides verifiable evidence for anomaly kernel calls, preventing anomaly responses from being assigned to the interaction coefficient graph.

[0098] Step 3: In the phased stimulus sequence and impact nuclear Under a common caliber, based on aligned monitoring sequences Recursively update the edge weights of the interaction coefficient graph When the residual shape deviates, an abnormal kernel strength coefficient is introduced. The stage disturbance contribution of multi-line overlapping construction is divided into traceable edge weights and anomalies.

[0099] First, set the existing line numbers The predicted deformation field is expressed as a spatial shape multiplied by its amplitude, with the spatial shape borne by the influence kernel and the amplitude by the edge weights of the interaction coefficient graph, and its temporal variation driven by the stage excitation sequence. However, under multi-line overlay conditions, directly inverting the stratigraphic constitutive parameters will result in multiple parameters jointly interpreting the same monitored change, leading to instability in the update; therefore, the variable component is limited to the edge weights of the interaction coefficient graph. It also requires that each edge weight simultaneously points to the construction line number. Existing line number with stage identifier This allows it to be traced back to the process nodes in the construction log.

[0100] First, on the vertical axis Upward call affects the core As a stage-specific spatial shape, then using the stage-specific stimulus sequence As a time-driven process, the predicted deformation field is obtained, and the longitudinal coordinates of the monitoring point are then used to obtain the deformation field. The sampled and aligned monitoring sequences were compared. The processing method used the following prediction expression:

[0101]

[0102] Where: Predicted deformation field Predicted response, range in real, function is to form residuals with observations; existing line number : Number index, ranging from a discrete set, used to distinguish existing tunnels; Vertical coordinate Mileage coordinates, with a range of mileage.

[0103] Timeline : A unified time variable, covering the valid construction period; construction line number : Number index, ranging from discrete sets; stage identifier Stage index, ranging from a finite discrete set; interaction coefficient graph edge weights. Stage amplitude coefficient, ranging from real numbers that meet boundary constraints, absorbs stratigraphic differences and construction deviations and uses them as the updating object; influence core : The shape of the unit stage disturbance response, in the range of real numbers; the stage excitation sequence Stage intensity input, range is ;

[0104] In practice, the shape of the fixed space is affected by the kernel, the stage excitation is driven by a fixed time, and the edge weights of the interaction coefficient graph become the only interpretable quantities that need to be recursively calculated; the edge weight index contains... , , This allows contributions to be traced back to specific construction phases. This enables subsequent sparsity penalties to apply the shrinkage effect to clearly defined construction line-phase edge weights.

[0105] The edge weights of the interaction coefficient graph are updated within a sliding time window, so that each update only explains the observation differences within the current window segment, and the edge weight migration is constrained by physical boundaries, sparsity, and step size continuity.

[0106] The aligned monitoring sequence exhibits uneven sampling and local outliers, which, without constraints, would cause multiple edge weights to drift synchronously. Therefore, the edge weights are set at a lower bound coefficient. With upper bound coefficient Between these values, sparse weights and step size weights are used to suppress the diffusion of irrelevant edge weights. After aligning the monitored and predicted values ​​within the window segment, residuals are constructed and robust losses are formed. Sparse penalties and step size penalties are then superimposed, and finally, the updated edge weights are obtained through constrained solution and boundary projection is performed.

[0107] The update objective is as follows, where the robust loss function is in log-hyperbolic-cosine form:

[0108]

[0109] Where: set of interaction coefficients : Set of variables, the range of which is the set of real numbers that satisfy the boundary constraints; Monitoring point number : Index, ranging from discrete sets; time axis : Unify the time variable, with the range being the current window segment; aligned monitoring values : Observation sequence, in the range of real numbers; existing line mapping : Mapping index, ranging from a discrete set; vertical coordinates of the monitoring point Location parameters, ranging from mileage;

[0110] Predicting deformation fields : Predicted sample values, in the range of real numbers; residual scale : Scale parameter, in the range of positive real numbers, used to set the residual normalization scale; sparse weights : Penalty coefficient, a non-negative real number, used to induce the shrinkage of irrelevant edge weights; Step size weight : Penalty coefficient, in the range of non-negative real numbers, which serves to restrict edge weights from jumping across the window;

[0111] Previous window segment coefficient The reference edge weight is a real number that satisfies the boundary constraints and serves as a reference for the continuity of the step size.

[0112] Construction line number Stage identifiers The index dimension, with ranges of discrete sets and finite discrete sets respectively, serves to characterize the decomposition of the penalty on the construction line-stage; lower bound coefficient. : Lower bound of edge weights; range is real numbers; function is to impose physical boundary constraints on the edge weights of the interaction coefficient graph.

[0113] Upper bound coefficient Upper bound on edge weights; range is real numbers; function is to impose physical boundary constraints on the edge weights of the interaction coefficient graph. For the boundary projection operator, for any scalar... , Weights on opposite sides: For anomalous nuclear strength: .

[0114] The solution can be obtained using an interior-point method solver or a sequential quadratic programming solver, and the edge weights are projected onto the interval after each iteration.

[0115] The derivative can be calculated according to the influence of the kernel. Multi-stage activation sequence The chained rules are written into the solver.

[0116] Sliding time windows control error propagation; robust loss suppresses outlier pull; sparse penalties and boundary projections suppress irrelevant edge weight drift; step size penalties ensure continuous edge weight evolution. Furthermore, after edge weight updates, it is determined whether typical influence kernels can explain the observations. If insufficient, anomaly perturbation influence kernel templates and anomaly kernel intensity coefficients are introduced to ensure that the anomaly contribution is carried independently without crowding out the interaction coefficient map. The responses caused by grout leakage, over-excavation, local voids, sudden surges, and downtime amplification are shown on the vertical coordinate. The peak position often shifts or unilateral attenuation. If it is forcibly absorbed by the edge weights, it will lead to synchronous shift of the edge weights in multiple stages.

[0117] First use the updated interaction coefficient graph edge weights The residuals at the monitoring points are calculated and rearranged into shape vectors according to the vertical coordinates, and then compared with the kernel template affected by the anomalous disturbances. Perform similarity matching; when the similarity exceeds a threshold, perform secondary verification using construction event nodes and trigger anomaly kernel switching; after triggering, only update the anomaly kernel strength coefficient. and number the exception type Write the version record as a discrete index; exception type number The scope is the set of template library numbers, and its function is to uniquely point to the kernel template affected by the abnormal disturbance that is being invoked.

[0118] In this method, the prediction after the anomaly kernel switch still uses the shape overlay framework, only superimposing the anomaly component on the predicted value, and using the anomaly kernel intensity coefficient as the only variable amplitude. Example description: When the construction line number... Upon entering the overlapping mileage section, the on-site recorder registered the event nodes of the start and end of the shutdown in the construction log. The monitor exported the aligned monitoring sequence and verified that the longitudinal coordinates of the monitoring points had not changed. After loading the version package of the current window segment into the computing terminal and completing the edge weight update, the technician found that the residual shape was consistent with the attenuation side of the shutdown magnification template. Subsequently, only the abnormal core strength coefficient was solved and written into the version record. Based on this, the on-site management personnel reviewed the time relationship between the shield tail grouting record and the shutdown re-excavation record.

[0119] In practice, residual shape triggering transforms anomaly identification from amplitude thresholding to shape-verifiable criteria; the anomaly kernel strength coefficient carries the anomaly contribution and the edge weights have stable meanings; secondary verification ensures that anomaly judgments correspond to construction event nodes. The interaction coefficient graph edge weights and anomaly kernel strength coefficients are organized as versioned outputs, and step four calls them in window segment order, using the stage excitation sequence and influence kernel in the rolling prediction while only propagating the amplitude.

[0120] If the start and end times of the window segment and the template number used are not recorded, step four cannot determine which set of coefficients to use, and the basis for the suggested construction parameter intervals cannot be traced on-site. The start and end times of the window segment, the set of monitoring points used, the set of construction event nodes used, the boundary interval, the updated edge weight discrete sequence, and the anomaly kernel strength coefficient discrete sequence are all written into the version package and indexed by the version number. After the version package is generated, a consistency check is performed to verify whether the edge weights are within the specified range. And whether the abnormal template number is unique.

[0121] The version package simultaneously writes the contribution component curves for each stage, allowing step four to directly superimpose components when calculating the prediction mean without repeated calculations. Versioned output resolves the input ambiguity issue in step four; parameter threading ensures that step four only propagates edge weights and anomaly strengths under fixed influence kernels and fixed stage stimuli; consistency checks prevent missing data from entering subsequent prediction chains, ultimately enabling step four to directly call and complete the input concatenation for rolling interval prediction according to window version.

[0122] Step 4: On the existing line number vertical coordinate Above, the edge weights of the interaction coefficient graph output in step three are... With the anomalous kernel strength coefficient As the amplitude input, the influence kernel from step two is reused. Affecting the nuclear template by anomalous perturbations As shape input, it is combined with the stage excitation sequence formed in step one. This yields interval prediction results for several future cycles or time periods, and generates surface pressure sequences accordingly. Grouting volume sequence Propulsion speed sequence The suggested intervals are ultimately used to form a traceable closed-loop record for the next window segment to be updated recursively.

[0123] By prioritizing the standardization of the tunnel diameter, subsequent predictions are only allowed to vary the amplitude within the established diameter, without altering the shape. This avoids the risk of misjudgment due to changes in the displayed diameter on-site. In multi-track intersection scenarios, mixing the predicted diameter with the monitoring diameter dilutes the meaning of the boundary weights. Therefore, the prediction object is fixed as the existing tunnel in the longitudinal coordinate system. The response curve is obtained, and the sampling points of the curve are consistent with the location of the monitoring section.

[0124] First, read the end of the current window segment from step three. and Therefore, the predicted amplitude source follows the same interpretation path as the observed data; thus, the stage excitation sequence of step one is interpreted as follows. Extending to the forecast period, so that future periods are still marked by stages. The process switching rules drive the process; thus, step two is invoked to affect the core. Affecting the nuclear template by anomalous perturbations Generate a response curve and use this curve as the center line for calculating the interval boundary.

[0125] The centerline calculation uses the same kernel superposition expression as in step three, with only the following updates. , and The timing values ​​are determined; when the construction organization plan changes, only the stage excitation sequence of the current segment is updated, while the version package number is retained. The centerline and confidence boundary share the same spatial shape caliber, allowing for on-site verification of the prediction and monitoring relationship at the same monitoring section; when the work process plan changes, only the timing values ​​are updated. In the future segment, the meaning of the edge rights is continuous and traceable.

[0126] Confidence boundaries are constructed using feasible region propagation of coefficients, ensuring that the source of the confidence boundaries is clear and verifiable. However, if the interval is expanded using empirical margins, it becomes difficult to explain which construction line or stage the interval expansion originates from; therefore, the upper and lower bound coefficients of the boundary weights given in step three are used... , As a basis for interval propagation, and setting upper and lower bounds for the anomaly kernel strength coefficient. , This ensures that the expansion of outliers does not encroach on the explanatory space of the typical stage boundary weights.

[0127] First, determine the upper and lower bounds of the weight of each edge, and then number the outliers according to their outlier type. Take its upper and lower bounds; therefore, in the vertical coordinate... The influence kernel and the anomaly kernel are superimposed according to the stage-based excitation sequence to obtain the upper and lower bounds of the feasible region, respectively. The upper and lower bounds are then output as confidence boundaries and compared with the control threshold. The position and time period of the upper bound approaching the threshold are then output.

[0128] The following interval propagation expression is used:

[0129]

[0130]

[0131] Where: the upper bound prediction value of the feasible region : Real number, used for threshold comparison and triggering suggestion interval generation; lower bound prediction value of feasible region : Real number, used to specify the outer lower boundary and shape reference; existing line number : Discrete index, used to indicate the existing tunnel being predicted; vertical coordinate Mileage coordinates, taken from the existing tunnel mileage range, used as a reference for aligning the kernel function's independent variables with monitoring data; Time axis : Time variable, the value is the prediction period, used to drive the stage excitation sequence and output the rolling result;

[0132] Construction line number Discrete index, used to accumulate contributions from multiple construction lines; stage identifier. Discrete index, used to distinguish between stage kernel and stage excitation; upper bound edge weight coefficient. : Real number and satisfy , used for upper bound propagation amplitude; lower bound boundary weight coefficient : Real number, used for the lower bound of propagation amplitude; affects the kernel : Real number function used to provide the shape of the vertical distribution;

[0133] Stage incentive sequence : Values Used to transcribe construction log information into phase-driven processes; exception type number. : Discrete index, used to uniquely point to the anomaly kernel template; upper bound anomaly strength coefficient : Real number and satisfy Used for the upper bound of the propagation amplitude of anomalies; lower bound of the anomaly intensity coefficient. : Real number, used for the lower bound propagation amplitude of anomalies; the kernel template affected by anomalies. Real-valued functions are used to carry out the shape of abnormal contributions and to verify their reproducibility.

[0134] The optional project implementation path is: for Pre-compute and cache on the sample point set; interval propagation only involves coefficient multiplication and addition; the threshold intersection location is determined using binary search. Confidence bounds are obtained through coefficient feasible region propagation, and their origin can be traced back to... , and , ; Exceptions passed When carried alone, the interpretation space of edge weights in typical stages remains stable.

[0135] The positions and time periods approaching the upper bound threshold are converted into executable construction parameter suggestion ranges, creating a direct correspondence between on-site actions and trigger points. However, if only risk warnings are output, it's difficult for work teams to map these warnings to specific settings for surface pressure, grouting, and advancement speed; therefore, risk contribution is incorporated into the stage incentive sequence. The upper positioning is then used to map the contraction requirements of the stage excitation sequence to... , , The range.

[0136] First, on the vertical axis Upper bound prediction value of feasible region The section closest to the threshold is used to determine the construction line number that makes the major contribution. with stage identifier Thus maintaining Under the premise that the shape remains unchanged, for this stage Perform a finite set of trial calculations and back-substitute the interval propagation expression for verification to obtain the stage excitation allowable interval that causes the upper bound to regress; then, following the normalization fusion path defined in step one, decompose the stage excitation allowable interval in reverse into pairs of... , , The suggested range is then checked for consistency with equipment limits and construction organization constraints.

[0137] The proposed interval solution uses a progressive approach: first single-parameter, then two-parameter, and finally three-parameter combined. For single-parameter solutions, the remaining parameters are fixed, and the endpoints are obtained using binary search. For two-parameter solutions, the endpoints are obtained using two-dimensional discrete search. For three-parameter solutions, the endpoints are obtained using three-dimensional discrete search, and then the results are substituted back into the original solution. Verify threshold satisfaction; an optional engineering implementation path is to use a sequential quadratic programming solver to find the interval endpoints within the equipment limits, and then back-substitute to verify that the endpoints do not violate the threshold constraints.

[0138] An alternative implementation method is: when the surface pressure is not adjustable, and The two-parameter linkage gives the range; when the propulsion speed is fixed, it is based on... and The dual-parameter linkage provides the range; when compensation grouting needs to be introduced, the corresponding stage of compensation grouting is incorporated. After updating the caliber, the same solution is executed again. Example description: During the pre-shift briefing, the technician prints out the confidence boundary and suggested interval of the trigger section and delivers them to the shift team; the tunnel boring machine operator adjusts the surface pressure setting according to the suggested interval, the grouting worker adjusts the grouting volume setting according to the suggested interval, and the tunneling team adjusts the advance speed setting according to the suggested interval; after completing the next cross-section acquisition, the monitor overlays and verifies the monitoring curve with the confidence boundary, and the recorder archives the execution trajectory and version package number.

[0139] The recommended interval is triggered by confidence boundaries and the endpoints are verified by back substitution. The endpoints of the interval have an interpretable source. The progressive solution from single-parameter to three-parameter linkage allows the work team to obtain an executable interval under the constraint conditions. The intention is to transform risk warnings into executable intervals and maintain a consistent closed-loop approach.

[0140] The content and feedback method of closed-loop recording are specified to ensure that the output of step four becomes a reliable input for the recursive update of the next window segment in step three. However, if the implementation status and reasons for deviations are not recorded, step three may mistakenly interpret deviations as changes in strata or edge weights, thus weakening the meaning of the interaction coefficient diagram.

[0141] First, combine the confidence boundary curve, trigger segment, and suggested interval with the construction line number. Stage identifiers Write it into a record; therefore, after the work team executes it, the actual... , , Set trajectory and stop flag sequence Write the same record and indicate the reason for the deviation; thus, when the exception type number is... When called, the corresponding The changes and on-site working condition verification conclusions are recorded together, thus providing a chain of evidence for the abnormal kernel switching and edge weight recursion in step three.

[0142] Among them, online layer interval propagation and finite group trial calculations are mainly used to adapt to the construction progress; for offline layer key overlapping mileages or frequent anomaly triggers, records are extracted and aligned to monitor sequences. The kernel library entries are reviewed, and the impact kernels for local operating conditions are recalculated and written back with the new version package. Commonly used tools include binary search, discrete mesh search, implicit static solver, and piecewise cubic interpolator. Execution trajectories and review conclusions are versioned and traced, minimizing the risk of misattribution in the next window update; online and offline tasks are separated, interval output is continuous, key milestones are verifiable, intent descriptions are linked to suggestions, execution, and review through traceability, and the recursive update feedback loop is complete and verifiable.

[0143] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0144] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0145] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0146] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0147] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A numerical simulation method for the influence of multi-line overlapping tunnel construction on tunnel deformation, characterized in that: include, Establish a vertical coordinate axis and a time axis, map and register existing tunnel monitoring data and construction log data, calculate the monitoring lag correction amount based on the process switching nodes and the inflection points of the monitoring curve, divide the construction stages and generate stage excitation sequences. A three-dimensional stratum-structure coupled numerical model was established offline. Unit stage disturbances were applied to each construction line and each construction stage. Influence cores were calculated and solidified to form an influence core library and pre-set abnormal disturbance influence core templates. Based on the influence kernel library, a predictive expression is constructed and an interaction coefficient map is set. The interaction coefficient is recursively updated within a sliding time window or a sliding ring number window using the aligned monitoring sequence. When the residual does not match the influence kernel, an abnormal perturbation influence kernel template is introduced and the abnormal kernel intensity coefficient is updated. The updated interaction coefficient graph is used to predict the rolling interval, output the upper and lower bound curves of the feasible region, and generate the suggested interval of construction parameters and the record.

2. The numerical simulation method for the influence of tunnel deformation according to claim 1, characterized in that: The monitoring lag correction is obtained by resampling the monitoring data through conformal interpolation. This data, together with the process switching event nodes, forms an event-driven time grid. On this time grid, the alignment relationship that satisfies the monotonic constraint is obtained. The aligned monitoring sequence is then obtained by using the piecewise robust error criterion and outlier removal rules, which are used to generate the stage excitation sequence.

3. The numerical simulation method for the influence of tunnel deformation according to claim 2, characterized in that: Mapping and registration include establishing a monitoring point ledger and a construction event list, binding each monitoring point identifier to the mileage position, existing line identifier, and section identifier on the vertical coordinate axis, and extracting the construction log into a sequence of process switching event nodes with construction line identifiers according to the advancement ring number and timestamp, so as to form a unified field and a unified sampling rhythm.

4. The numerical simulation method for the influence of tunnel deformation according to claim 1, characterized in that: The division of construction stages includes generating a sequence of stage labels for each construction line based on the process switching event nodes, and performing consistency checks at the stage boundaries in conjunction with the inflection points of the monitoring curves, so that only one stage label corresponds to the same moment. When the impact of shutdown overlaps with synchronous grouting or grout hardening, the stage label is determined first based on the stage of shutdown impact, and the stage boundaries of the overlapping area are sealed.

5. The numerical simulation method for the influence of tunnel deformation according to claim 4, characterized in that: The stage excitation sequence is generated by fusing the surface pressure, synchronous grouting volume, synchronous grouting pressure, advance speed, shutdown and re-excavation time, and secondary compensation grouting records from the construction log segment by segment after boundary truncation and normalization. The original parameters before normalization are archived along with the stage excitation sequence for consistent use in offline modeling unit stage disturbance settings and kernel library indexes.

6. The numerical simulation method for the influence of tunnel deformation according to claim 1, characterized in that: The influence core library is indexed by net distance, intersection angle, burial depth difference, soil cover thickness and stratum combination, and the model mesh version, boundary condition version and unit stage disturbance caliber are recorded for each index vector; Weighted retrieval interpolation of the influence kernel is performed between adjacent samples of the index vector, and the influence kernels are all uniformly expressed using the vertical coordinate axis and include construction line and stage identifiers.

7. The numerical simulation method for the influence of tunnel deformation according to claim 6, characterized in that: Unit stage disturbances are applied in the three-dimensional stratum-structure coupled numerical model according to the corresponding stages. Among them, the excavation unloading stage adopts stress release boundary, the shield tail gap formation stage adopts equivalent volume loss boundary, the synchronous grouting stage adopts equivalent pressure boundary, the grouting hardening stage adopts material parameter time sequence switching boundary, the secondary compensation stage adopts equivalent pressure boundary, and the shutdown influence stage adopts step sequence maintenance and intermittent load application boundary, and the stage identifier solidification influence core is recorded.

8. The numerical simulation method for the influence of tunnel deformation according to claim 7, characterized in that: The templates affected by abnormal disturbances include templates for grout leakage, over-excavation, local voids, sudden surge, and shutdown expansion. Each template and the affected core have the same longitudinal coordinate representation and stage identifier field. The template is called based on the similarity between the residual morphology curve and the template morphology to meet the threshold, and the corresponding abnormal event node in the construction event list is verified to meet the time neighborhood condition as the trigger condition.

9. The numerical simulation method for the influence of tunnel deformation according to claim 1, characterized in that: The interaction coefficient diagram is a set of coupled edges established according to "construction line-existing line-stage". An interaction coefficient is set for each coupled edge and updated synchronously in the sliding ring number window within the sliding time window. During the update process, upper and lower bound constraints, change range constraints, and shrinkage rules for coupled edges far from the overlapping area are applied to the interaction coefficient.

10. The numerical simulation method for the influence of tunnel deformation according to claim 1, characterized in that: The residual is obtained by subtracting the corresponding output of the predicted expression from the aligned monitoring data, and the residual shape curve is reconstructed along the vertical coordinate axis. When the residual shape curve is inconsistent with the typical shape of the kernel library, the abnormal perturbation is introduced to affect the kernel template and the abnormal kernel strength coefficient is updated synchronously.

11. The numerical simulation method for the influence of tunnel deformation according to claim 1, characterized in that: The rolling interval prediction uses the time period corresponding to several future rings as the prediction interval. Based on the interaction coefficient diagram and its upper and lower bounds, it generates the upper boundary curve and lower boundary curve of the feasible region. The construction event nodes within the prediction interval are included in the stage excitation sequence to complete the continuous calculation within the interval. At the same time, the mileage section sequence corresponding to the vertical coordinate axis is output as the result index.

12. The numerical simulation method for the influence of tunnel deformation according to claim 11, characterized in that: The recommended range of construction parameters was obtained by conducting small-scale trial calculations of surface pressure, synchronous grouting volume, synchronous grouting pressure and advancing speed within the superimposed frame of the influence core. During the trial calculation, the stage label sequence was kept unchanged and the input parameters of the stage excitation sequence were adjusted segment by segment to form the interval-based recommended values ​​corresponding to the prediction interval.

13. The numerical simulation method for the influence of tunnel deformation according to claim 12, characterized in that: The entire process record includes the definition of the longitudinal coordinate axis, the definition of the time axis, the monitoring lag correction amount, the stage label sequence, the stage excitation sequence, the interaction coefficient diagram, the anomaly core strength coefficient and the suggested range of construction parameters, and triggers offline review to update the impact core library when entering critical overlapping mileage sections or when anomalies occur frequently.

Images