Shield tunnel underpass railway deformation subgrade backfilling method and system

Abstract

The application belongs to the technical field of underground engineering construction and relates to a method and system for backfilling of a deformed roadbed of a railway undercrossed by a shield tunnel.A method for backfilling of a deformed roadbed of a railway undercrossed by a shield tunnel, by collecting roadbed environment response data, shield tunneling parameters and geological condition data in real time, combining a dynamically calculated settlement prediction model, and realizing accurate control of railway roadbed settlement in shield tunnel construction.A system for backfilling of a deformed roadbed of a railway undercrossed by a shield tunnel, comprising a data collection module, a working distance calculation module, a weight coefficient calculation module, a settlement prediction module and an optimization decision module.The method and system for backfilling of a deformed roadbed of a railway undercrossed by a shield tunnel can dynamically adjust backfilling strategies and grouting parameters according to different construction stages and geological conditions, ensure that railway roadbed settlement is controlled within a safe range, and effectively avoid the influence of excessive settlement or uneven settlement on railway safety.

Description

Technical Field

[0001] This invention belongs to the field of underground engineering construction technology, specifically relating to a method and system for backfilling deformed roadbed when a shield tunnel passes under a railway. Background Technology

[0002] Shield tunneling technology is an underground construction method widely used in infrastructure construction such as urban rail transit, high-speed railways, and underground pipelines. Especially in complex geological conditions and densely populated urban areas, when shield tunnels pass under existing structures such as railways and roads, how to effectively control the impact on the surrounding environment during construction, especially the settlement and deformation of railway subgrades, has become an important technical problem.

[0003] In traditional shield tunnel construction, the tunnel boring machine's excavation process, especially when passing under a railway, can have a significant impact on the railway subgrade, potentially leading to subgrade settlement, tilting, and other deformations, which can affect the railway's operational safety and stability. Therefore, how to accurately predict and effectively control settlement issues during shield tunnel construction, and ensure that the railway subgrade does not experience excessive settlement during construction, has become an urgent technical challenge.

[0004] Currently, several technologies and methods have been proposed to reduce the settlement risk when shield tunnels pass under railways. For example, common backfilling methods typically involve injecting grouting material during tunnel excavation to support the surrounding soil and prevent settlement and collapse. However, existing technologies often face several problems in practical applications: a lack of precise settlement prediction and control mechanisms, insufficient adaptability to geological conditions, and unreasonable setting of grouting parameters. Existing grouting parameters are often based on experience and preset values, lacking real-time construction data feedback. Grouting volume and pressure are usually determined before construction begins, but inappropriate grouting volumes or pressures may occur during construction, affecting the backfilling effect and even leading to over-grouting (wasting materials) or under-grouting (causing settlement problems).

[0005] Therefore, this invention proposes a method and system for backfilling deformed roadbeds under a shield tunnel passing through a railway. The optimal grouting parameters are dynamically calculated through an optimization algorithm to ensure that settlement is controlled within an acceptable range while reducing the consumption of grouting materials. Summary of the Invention

[0006] This invention achieves precise control of railway subgrade settlement during shield tunnel construction by real-time acquisition of subgrade environmental response data, shield tunneling parameters, and geological condition data, combined with a dynamically calculated settlement prediction model. The system can dynamically adjust backfilling strategies and grouting parameters according to different construction stages and geological conditions to ensure that railway subgrade settlement is controlled within a safe range, thereby effectively avoiding the impact of excessive or uneven settlement on railway safety.

[0007] A method for backfilling deformed roadbed when a shield tunnel passes under a railway includes:

[0008] Controlling the tunnel boring machine to excavate a certain dynamic working distance, the time taken to complete this dynamic working distance is defined as a tunneling cycle; within this tunneling cycle, real-time data on roadbed environmental response, tunnel boring machine parameters, and geological conditions are collected. Roadbed environmental response data includes the settlement and horizontal displacement values ​​of the railway roadbed, tunnel boring machine parameters include the tunneling speed of the tunnel boring machine, cutterhead torque, and soil chamber pressure, and geological conditions data include geological type and soil mechanical parameters;

[0009] At the end of the current tunneling cycle, the construction stage is determined based on the position information of the tunnel boring machine relative to the railway subgrade projection line. The construction stage includes the approach stage, the tunneling stage, and the disengagement stage. Combined with the geological condition data of the subsequent tunneling area, the dynamic working distance of the next tunneling cycle is calculated.

[0010] Simultaneously, based on the geological condition data collected during the current tunneling cycle, the weighting coefficients of the shield tunneling parameters and geological condition data are calculated.

[0011] The roadbed environmental response data collected during the current tunneling cycle are sorted by time to form the roadbed environmental response time series data. The roadbed environmental response time series data, along with shield tunneling parameters and geological condition data with weighted coefficients, are input into the settlement prediction model to output the settlement prediction results of the railway roadbed.

[0012] Based on the optimization objectives set for the construction stage, the settlement prediction results and the optimization objectives are input into a multi-objective optimization algorithm to obtain a set of optimal grouting parameters, including grouting pressure and grouting volume.

[0013] Backfilling operations are carried out according to the grouting parameters.

[0014] Preferably, at the end of the current tunneling cycle, the construction stage is determined based on the position information of the tunnel boring machine relative to the railway subgrade projection line. The specific operation is as follows:

[0015] The projection line of the railway subgrade onto the vertical plane containing the tunnel axis is defined as the baseline;

[0016] The front face of the tunnel boring machine cutterhead is used as the positioning reference point;

[0017] Calculate the horizontal distance d from the positioning reference point to the baseline;

[0018] The symbolic distance D based on the tunneling direction of the tunnel boring machine is defined as follows: when the tunneling direction points to the baseline, D = d; when the tunneling direction deviates from the baseline, D = -d.

[0019] The construction stage is determined based on the symbol distance D:

[0020] When D>L1, it is determined to be in the approach stage, where L1 is a preset positive threshold, representing the boundary for starting to enter the high-precision control area;

[0021] When -L2≤D≤L1, it is determined to be the tunneling stage, where L2 is the length of the tunnel boring machine and -L2 indicates that the positioning reference point has crossed the reference line and the distance is L2.

[0022] When D < -L2, it is determined to be in the disengagement phase.

[0023] Preferably, based on the position information of the tunnel boring machine relative to the railway subgrade projection line and the geological condition data of the area to be excavated, the dynamic working distance of the next excavation cycle is calculated. The specific operation is as follows:

[0024] The stage adjustment coefficient K_phase is determined based on the current construction stage. The K_phase value corresponding to the approach stage is K_p1; the K_phase value corresponding to the through stage is K_p2; and the K_phase value corresponding to the departure stage is K_p3. K_p2 < K_p3 < K_p1 is satisfied, where K_p1, K_p2 and K_p3 are all real numbers in the range (0,1].

[0025] Based on the geological type and soil mechanical parameters in the geological condition data, the rock mass quality classification index RQD in the field of geotechnical engineering is adopted as the benchmark value G for geological stability evaluation.

[0026] Divide the RQD value by the upper limit of its grading system to obtain the standardized RQD ratio; then, using the first preset coefficient K_min as the benchmark value, add the product of the difference between the second preset coefficient K_max and K_min and the RQD ratio, and the result is the geological adjustment coefficient K_geo; where K_min and K_max are preset constant coefficients, and satisfy K_max is greater than K_min, and the values ​​of K_min and K_max are both greater than zero and less than or equal to one;

[0027] The average of the phase adjustment coefficient K_phase and the geological adjustment coefficient K_geo is used as the distance adjustment coefficient;

[0028] Set the maximum allowable working distance and the minimum safe distance. Multiply the maximum allowable working distance by the distance adjustment coefficient to obtain the preliminary working distance. Determine whether the preliminary working distance is greater than the minimum safe distance. If it is, use the preliminary working distance as the final dynamic working distance; otherwise, use the minimum safe distance as the final dynamic working distance.

[0029] Preferably, based on the geological condition data collected during the current tunneling cycle, weighting coefficients for the tunnel boring machine parameters and the geological condition data are calculated. The specific operation is as follows:

[0030] The initial weights W_p0 for the shield tunneling parameters and W_g0 for the geological condition data are both set to 0.5.

[0031] The weight adjustment amount Δ is calculated based on the geological stability evaluation benchmark value G. The weight adjustment amount Δ is determined in the following way: when G is greater than the preset stability threshold G0, Δ=k*(G-G0), where k is the preset adjustment coefficient; when G is less than or equal to G0, Δ=0.

[0032] The initial weights are adjusted, and the weight coefficients for the shield tunneling parameters are W_p=W_p0+Δ, and the weight coefficients for the geological condition data are W_g=W_g0–Δ.

[0033] Preferably, the settlement prediction model is based on a long short-term memory neural network, and the network structure includes an input layer, a feature fusion layer, a temporal feature extraction layer, and an output layer.

[0034] The input layer is used to receive preprocessed multi-source input data, including: roadbed environmental response time series data, shield tunneling parameters with weighted coefficients, and geological condition data with weighted coefficients.

[0035] The feature fusion layer includes a fully connected network used to encode the weighted shield tunneling parameters and geological condition data into features, and then splice and fuse them with the initial features of the roadbed environmental response time series data.

[0036] The temporal feature extraction layer is composed of multiple stacked LSTM units, used to extract time dependencies from the fused feature sequence and capture the dynamic laws of sedimentation changes.

[0037] The output layer adopts a fully connected network structure, which maps the final features extracted by the LSTM layer to the predicted settlement values ​​of key points of the railway subgrade within a certain time window in the future.

[0038] Preferably, the specific steps for training the settlement prediction model are as follows:

[0039] Several training samples are acquired, each including multi-source data extracted from a complete historical tunneling cycle, specifically including time-series data of roadbed environmental response, shield tunneling parameters, and geological conditions arranged chronologically within that cycle. All training samples are divided into a training set and a validation set. The settlement prediction model with initialized parameters is trained using the training set, and validated using the validation set to obtain the validation results. Training conditions are set, and it is determined whether the validation results meet the training conditions. If they do, the trained settlement prediction model is output; otherwise, the settlement prediction model is trained again using the training set.

[0040] Preferably, based on the optimization objectives set for the construction stage, the settlement prediction results and the optimization objectives are input into a multi-objective optimization algorithm to obtain a set of optimal grouting parameters. The specific operation is as follows:

[0041] Step 1: Set differentiated multi-objective optimization functions based on the determined construction stages;

[0042] When in the approach stage, the primary objective of optimizing the objective function is to control the predicted settlement value to be slightly above zero, while the secondary objective is the amount of grouting material consumed.

[0043] When in the direct crossing phase, optimizing the objective function to minimize the predicted settlement value is the highest priority objective.

[0044] When in the separation stage, the main objective is to optimize the objective function to control the settlement stability and convergence, while prioritizing the reduction of grouting material consumption.

[0045] Step 2: Set the physical constraints of the grouting parameters, including: upper and lower limit constraints of grouting pressure, non-negativity constraints of grouting volume, and constraints on the coupling relationship between grouting pressure and grouting volume.

[0046] Step 3: Use a non-dominated sorting genetic algorithm with an elitist strategy to solve the established multi-objective optimization problem, and obtain a set of Pareto optimal solutions through population evolution;

[0047] Step 4: From the Pareto optimal solution set, select the final optimal combination of grouting parameters based on the safety requirements of the current construction stage: select the solution with the best settlement control effect in the penetration stage, and select the optimal compromise solution that takes into account settlement control and material consumption in the approach and separation stages.

[0048] A system for backfilling deformed roadbed under a railway in a shield tunnel, the system being used to implement the aforementioned method for backfilling deformed roadbed under a railway in a shield tunnel, comprising:

[0049] The data acquisition module is used to collect roadbed environmental response data, shield tunneling parameters and geological condition data in real time during the tunneling cycle. The roadbed environmental response data includes the settlement value and horizontal displacement value of the railway roadbed. The shield tunneling parameters include the tunneling speed of the shield machine, the cutterhead torque and the soil chamber pressure. The geological condition data includes the geological type and soil mechanical parameters.

[0050] The working distance calculation module is used to determine the construction stage at the end of the current tunneling cycle based on the position information of the tunnel boring machine relative to the railway subgrade projection line. The construction stage includes the approach stage, the crossing stage, and the separation stage. Combined with the geological condition data of the subsequent tunneling area, the module calculates the dynamic working distance for the next tunneling cycle.

[0051] The weighting coefficient calculation module is used to calculate the weighting coefficients of the shield tunneling parameters and geological condition data based on the geological condition data collected in the current tunneling cycle.

[0052] The settlement prediction module is used to sort the roadbed environmental response data collected in the current tunneling cycle into roadbed environmental response time series data according to time. The roadbed environmental response time series data, as well as shield tunneling parameters and geological condition data with weighted coefficients, are input into the settlement prediction model, and the settlement prediction results of the railway roadbed are output.

[0053] The optimization decision module is used to set optimization objectives based on the construction stage, input the settlement prediction results and the optimization objectives into a multi-objective optimization algorithm, and solve for a set of optimal grouting parameters, including grouting pressure and grouting volume; and to carry out backfilling operations based on the grouting parameters.

[0054] The present invention has the following advantages:

[0055] 1. This invention achieves precise control of railway subgrade settlement during shield tunnel construction by real-time acquisition of subgrade environmental response data, shield tunneling parameters, and geological condition data, combined with a dynamically calculated settlement prediction model. The system can dynamically adjust backfilling strategies and grouting parameters according to different construction stages and geological conditions to ensure that railway subgrade settlement is controlled within a safe range, thereby effectively avoiding the impact of excessive or uneven settlement on railway safety.

[0056] 2. The optimal grouting parameters calculated by the present invention through a multi-objective optimization algorithm can reasonably allocate grouting pressure and grouting volume at different construction stages. This method not only ensures the settlement control effect, but also adjusts the grouting volume according to the actual construction situation, avoiding the problems of excessive consumption or insufficient grouting of grouting materials in traditional methods. This achieves a more efficient and environmentally friendly construction method, reduces material waste, and lowers project costs. Attached Figure Description

[0057] Figure 1 This is a schematic diagram of the architecture of the shield tunnel under the deformed roadbed backfilling system used in an embodiment of the present invention. Detailed Implementation

[0058] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this invention.

[0059] Example 1: A method for backfilling deformed roadbed under a railway tunnel using a shield tunnel, comprising:

[0060] A tunneling cycle is defined as the time taken for a tunnel boring machine (TBM) to excavate a certain dynamic working distance. Within this cycle, real-time data on the roadbed environmental response, TBM excavation parameters, and geological conditions are collected. During implementation, the TBM is first controlled to excavate at a predetermined dynamic working distance. This dynamic working distance is dynamically adjusted based on the construction stage, geological conditions, and the TBM's excavation progress to ensure that the impact of the tunnel excavation on the railway roadbed is kept to a minimum. Roadbed environmental response data includes settlement and horizontal displacement values, primarily used to monitor roadbed deformation and ensure it does not exceed allowable settlement limits. TBM excavation parameters include the TBM's excavation speed, cutterhead torque, and soil chamber pressure, reflecting the TBM's operating status and its impact on the soil. Geological condition data includes geological type and soil mechanics parameters, which form the basis for calculating ground deformation during tunnel excavation and determining backfilling methods.

[0061] At the end of the current tunneling cycle, the construction stage is determined based on the position information of the tunnel boring machine relative to the railway subgrade projection line. The construction stage includes the approach stage, the tunneling stage, and the disengagement stage. Combined with the geological condition data of the subsequent tunneling area, the dynamic working distance of the next tunneling cycle is calculated.

[0062] In the implementation of this invention, at the end of each tunneling cycle, the system determines the current construction stage based on the location data of the tunnel boring machine (TBM). This location data is obtained through a real-time positioning system, specifically including the horizontal and vertical distances of the TBM cutterhead front face relative to the railway subgrade projection line. Based on this, the construction stage is divided into three main stages: the approach stage, the penetration stage, and the separation stage. When the TBM is far from the railway subgrade, the system determines it to be in the approach stage. In this stage, the TBM's working distance is relatively long, and the impact on the railway subgrade is small; therefore, a lower control frequency is typically used to improve construction efficiency. When the TBM approaches and enters beneath the railway subgrade, it enters the penetration stage. At this point, the impact of the TBM on the railway subgrade is greatest; therefore, the system needs higher-frequency monitoring and control to ensure settlement remains within the specified safe range. The tunneling distance is relatively short, and the construction process requires more precise control. When the TBM passes beneath the railway subgrade and gradually moves away from it, it enters the separation stage. At this point, as the distance between the TBM and the railway subgrade gradually increases, the impact on the subgrade gradually decreases, allowing for the restoration of higher work efficiency and a longer working distance.

[0063] Simultaneously, based on the geological condition data collected during the current tunneling cycle, weighting coefficients for the tunnel boring machine (TBM) parameters and geological condition data are calculated. TBM tunneling parameters include tunneling speed, cutterhead torque, and soil chamber pressure. Each parameter has a different impact on ground settlement, TBM efficiency, and soil disturbance during tunnel construction. For example, a higher tunneling speed may lead to greater settlement, while higher cutterhead torque and soil chamber pressure may have a greater impact on soil structure. Therefore, each parameter's weighting coefficient is calculated based on its specific impact on construction stability. Geological condition data includes soil type (such as soft soil, sand, and rock), soil density, soil friction coefficient, and groundwater level. Geological conditions are a key factor affecting TBM tunneling performance and ground settlement. For example, soft soil and sandy layers typically lead to greater settlement, while stable soil layers such as hard rock are more stable. Therefore, based on the geological conditions of the current construction area, the system assesses soil stability and calculates corresponding weighting coefficients.

[0064] The roadbed environmental response data collected during the current tunneling cycle are sorted by time to form the roadbed environmental response time series data. The roadbed environmental response time series data, along with shield tunneling parameters and geological condition data with weighted coefficients, are input into the settlement prediction model to output the settlement prediction results of the railway roadbed.

[0065] Based on the construction stages, optimization objectives are set. For example, for the approach stage, the main objective is to ensure that the settlement does not exceed the specified allowable range and that the consumption of grouting material is kept within a reasonable range. Therefore, the optimization objective is to control the settlement value close to zero and minimize the amount of grouting. For the through-pass stage, the focus is on minimizing the settlement of the railway subgrade and avoiding large settlement deformation. The system will prioritize optimizing the settlement control effect, setting the grouting pressure and grouting volume to the optimal values ​​to ensure sufficient support for the stability of the subgrade. For the separation stage, the optimization objective focuses on the stable convergence of settlement while also considering the consumption of grouting material. Therefore, the system will select a compromise optimal solution that balances settlement control and material usage. The settlement prediction results and the optimization objectives are input into a multi-objective optimization algorithm to obtain a set of optimal grouting parameters, including grouting pressure and grouting volume.

[0066] Backfilling operations are carried out according to the grouting parameters.

[0067] At the end of the current tunneling cycle, the construction stage is determined based on the position information of the tunnel boring machine relative to the railway subgrade projection line. The specific operation is as follows:

[0068] The projection line of the railway subgrade onto the vertical plane containing the tunnel axis is defined as the baseline;

[0069] The front face of the tunnel boring machine cutterhead is used as the positioning reference point;

[0070] Calculate the horizontal distance d from the positioning reference point to the baseline;

[0071] The symbolic distance D based on the tunneling direction of the tunnel boring machine is defined as follows: when the tunneling direction points to the baseline, D = d; when the tunneling direction deviates from the baseline, D = -d.

[0072] The construction stage is determined based on the symbol distance D:

[0073] When D>L1, it is determined to be in the approach stage, where L1 is a preset positive threshold, representing the boundary for starting to enter the high-precision control area;

[0074] When -L2≤D≤L1, it is determined to be the tunneling stage, where L2 is the length of the tunnel boring machine and -L2 indicates that the positioning reference point has crossed the reference line and the distance is L2.

[0075] When D < -L2, it is determined to be in the disengagement phase.

[0076] Based on the position information of the tunnel boring machine relative to the railway subgrade projection line and the geological condition data of the area to be excavated, the dynamic working distance of the next excavation cycle is calculated. The specific operation is as follows:

[0077] The stage adjustment coefficient K_phase is determined based on the current construction stage. The K_phase value corresponding to the approach stage is K_p1; the K_phase value corresponding to the through stage is K_p2; and the K_phase value corresponding to the departure stage is K_p3. K_p2 < K_p3 < K_p1 is satisfied, where K_p1, K_p2 and K_p3 are all real numbers in the range (0,1].

[0078] Based on the geological type and soil mechanical parameters in the geological condition data, the rock mass quality classification index RQD in the field of geotechnical engineering is adopted as the benchmark value G for geological stability evaluation.

[0079] Divide the RQD value by the upper limit of its grading system to obtain the standardized RQD ratio; then, using the first preset coefficient K_min as the benchmark value, add the product of the difference between the second preset coefficient K_max and K_min and the RQD ratio, and the result is the geological adjustment coefficient K_geo; where K_min and K_max are preset constant coefficients, and satisfy K_max is greater than K_min, and the values ​​of K_min and K_max are both greater than zero and less than or equal to one;

[0080] The specific calculation formula is: Geological adjustment coefficient K_geo = K_min + (K_max - K_min) * (G / G_max); where G_max is the upper limit of the evaluation benchmark value in its classification system, and K_min and K_max are preset coefficients that satisfy 0. <K_min<K_max≤1;

[0081] The average of the phase adjustment coefficient K_phase and the geological adjustment coefficient K_geo is used as the distance adjustment coefficient;

[0082] Set the maximum allowable working distance and the minimum safe distance. Multiply the maximum allowable working distance by the distance adjustment coefficient to obtain the preliminary working distance. Determine whether the preliminary working distance is greater than the minimum safe distance. If it is, use the preliminary working distance as the final dynamic working distance; otherwise, use the minimum safe distance as the final dynamic working distance.

[0083] Based on the geological condition data collected during the current tunneling cycle, the weighting coefficients of the tunnel boring machine (TBM) parameters and the geological condition data are calculated. The specific steps are as follows:

[0084] The initial weights W_p0 for the shield tunneling parameters and W_g0 for the geological condition data are both set to 0.5.

[0085] The weight adjustment amount Δ is calculated based on the geological stability evaluation benchmark value G. The weight adjustment amount Δ is determined in the following way: when G is greater than the preset stability threshold G0, Δ=k*(G-G0), where k is the preset adjustment coefficient; when G is less than or equal to G0, Δ=0.

[0086] The initial weights are adjusted, and the weight coefficients for the shield tunneling parameters are W_p=W_p0+Δ, and the weight coefficients for the geological condition data are W_g=W_g0–Δ.

[0087] The settlement prediction model is based on a long short-term memory neural network, which includes an input layer, a feature fusion layer, a temporal feature extraction layer, and an output layer.

[0088] The input layer is used to receive preprocessed multi-source input data, including: roadbed environmental response time series data, shield tunneling parameters with weighted coefficients, and geological condition data with weighted coefficients.

[0089] The feature fusion layer includes a fully connected network used to encode the weighted shield tunneling parameters and geological condition data into features, and then splice and fuse them with the initial features of the roadbed environmental response time series data.

[0090] The temporal feature extraction layer is composed of multiple stacked LSTM units, used to extract time dependencies from the fused feature sequence and capture the dynamic patterns of settlement changes. The temporal feature extraction layer has 3 layers of LSTM units, with the number of hidden nodes in each layer decreasing from 128 to 64 to 32, used to extract settlement features at different time scales step by step. Each layer contains forward and backward units, and the dimensions are doubled after splicing, used to simultaneously capture the "precursor" and "lag" effects of settlement trends.

[0091] The output layer adopts a fully connected network structure, which maps the final features extracted by the LSTM layer to the predicted settlement values ​​of key points of the railway subgrade within a certain time window in the future.

[0092] The specific steps for training the settlement prediction model are as follows:

[0093] Several training samples are acquired, each including multi-source data extracted from a complete historical tunneling cycle, specifically including time-series data of roadbed environmental response, shield tunneling parameters, and geological conditions arranged chronologically within that cycle. All training samples are divided into a training set and a validation set. The settlement prediction model with initialized parameters is trained using the training set, and validated using the validation set to obtain the validation results. Training conditions are set, and it is determined whether the validation results meet the training conditions. If they do, the trained settlement prediction model is output; otherwise, the settlement prediction model is trained again using the training set.

[0094] Based on the optimization objectives set for the construction stages, the settlement prediction results and the optimization objectives are input into a multi-objective optimization algorithm to obtain a set of optimal grouting parameters. The specific operation is as follows:

[0095] Step 1: Set differentiated multi-objective optimization functions based on the determined construction stages;

[0096] When in the approach stage, the primary objective of optimizing the objective function is to control the predicted settlement value to be slightly above zero, while the secondary objective is the amount of grouting material consumed.

[0097] When in the direct crossing phase, optimizing the objective function to minimize the predicted settlement value is the highest priority objective.

[0098] When in the separation stage, the main objective is to optimize the objective function to control the settlement stability and convergence, while prioritizing the reduction of grouting material consumption.

[0099] Step 2: Set the physical constraints of the grouting parameters, including: upper and lower limit constraints of grouting pressure, non-negativity constraints of grouting volume, and constraints on the coupling relationship between grouting pressure and grouting volume.

[0100] Step 3: Use a non-dominated sorting genetic algorithm with an elitist strategy to solve the established multi-objective optimization problem, and obtain a set of Pareto optimal solutions through population evolution;

[0101] Step 4: From the Pareto optimal solution set, select the final optimal combination of grouting parameters based on the safety requirements of the current construction stage: select the solution with the best settlement control effect in the penetration stage, and select the optimal compromise solution that takes into account settlement control and material consumption in the approach and separation stages.

[0102] Example 2: To more clearly illustrate the specific implementation of the present invention, a hypothetical engineering scenario is presented below. In this example, the shield tunnel needs to pass under an operating high-speed railway. Below the railway subgrade, there is a composite stratum consisting of soft clay and sand layers. This stratum has poor stability and is prone to settlement.

[0103] Step 1: Initial setup and the first tunneling cycle.

[0104] System initialization: Set the maximum allowable working distance to 10 meters and the minimum safe distance to 1 meter.

[0105] The tunnel boring machine (TBM) begins excavation. Based on the initial positioning, the TBM cutterhead is still far from the railway projection line (D>L1), and the system determines it to be in the approaching stage.

[0106] Based on the information and geological forecasts at this stage, the system calculates the dynamic working distance for the first tunneling cycle to be 8 meters, and the tunnel boring machine begins tunneling accordingly.

[0107] Step 2: Data collection and phase transformation.

[0108] During this 8-meter tunneling cycle, the system collects data in real time through a sensor network:

[0109] Roadbed environmental response data: The initial settlement of roadbed monitoring point A is -0.5mm (micro-settlement);

[0110] Tunneling parameters: average tunneling speed 35 mm / min, cutterhead torque 3200 kN·m, soil chamber pressure 1.8 bar;

[0111] Geological data: The current and subsequent strata are confirmed to be soft clay with a low rock mass quality classification index (RQD) value, and the calculated value is G=0.25 (not belonging to stable strata).

[0112] At the end of the cycle, the calculation showed that the tunnel boring machine had entered the range of -L2≤D≤L1, and the system determined that it had entered the main tunneling stage;

[0113] Step 3: Dynamic decision-making and weight calculation.

[0114] Dynamic working distance adjustment: Due to entering the main traverse stage (the stage coefficient K_phase takes a small value K_p2) and poor geological conditions (the geological adjustment coefficient K_geo is small due to the low G value), the system calculates that the dynamic working distance for the next cycle will be sharply reduced to 3 meters in order to achieve high-frequency control.

[0115] Weight coefficient calculation: Based on the geological stability evaluation benchmark value G=0.25 (low), the system calculates the weight adjustment amount Δ; assuming G0=0.5, k=0.5, then Δ=0; therefore, the weight W_g of geological condition data remains at 0.5, and the weight W_p of shield tunneling parameters also remains at 0.5, which reflects that under the current adverse geological conditions, the impact of both on settlement is considered equally important;

[0116] Step 4: Settlement prediction and optimization decision-making.

[0117] The system inputs the time-series settlement data, weighted shield tunneling parameters, and geological data collected in this cycle into the trained LSTM settlement prediction model.

[0118] The model outputs the following prediction: If no intervention is taken, the settlement of monitoring point A will reach -4.8 mm in the next hour, exceeding the safety threshold of -3.0 mm.

[0119] The system initiates a multi-objective optimization algorithm (NSGA-II) based on the optimization objective of the forward crossing stage (minimizing settlement first).

[0120] The algorithm solves the problem under the constraints of grouting pressure and grouting volume, and obtains a set of Pareto optimal solutions. The system finally selects the solution with the best settlement control effect and determines the optimal grouting parameters as follows: grouting pressure 3.8 bar, grouting volume 6.5 m³.

[0121] Step 5: Backfill execution.

[0122] Based on the above parameters, the control system starts the synchronous grouting system for precise backfilling.

[0123] Example 3: A backfilling system for deformed roadbed when a shield tunnel passes under a railway, such as... Figure 1 As shown, it includes:

[0124] The data acquisition module is used to collect roadbed environmental response data, shield tunneling parameters and geological condition data in real time during the tunneling cycle. The roadbed environmental response data includes the settlement value and horizontal displacement value of the railway roadbed. The shield tunneling parameters include the tunneling speed of the shield machine, the cutterhead torque and the soil chamber pressure. The geological condition data includes the geological type and soil mechanical parameters.

[0125] The working distance calculation module is used to determine the construction stage at the end of the current tunneling cycle based on the position information of the tunnel boring machine relative to the railway subgrade projection line. The construction stage includes the approach stage, the crossing stage, and the separation stage. Combined with the geological condition data of the subsequent tunneling area, the module calculates the dynamic working distance for the next tunneling cycle.

[0126] The weighting coefficient calculation module is used to calculate the weighting coefficients of the shield tunneling parameters and geological condition data based on the geological condition data collected in the current tunneling cycle.

[0127] The settlement prediction module is used to sort the roadbed environmental response data collected in the current tunneling cycle into roadbed environmental response time series data according to time. The roadbed environmental response time series data, as well as shield tunneling parameters and geological condition data with weighted coefficients, are input into the settlement prediction model, and the settlement prediction results of the railway roadbed are output.

[0128] The optimization decision module is used to set optimization objectives based on the construction stage, input the settlement prediction results and the optimization objectives into a multi-objective optimization algorithm, and solve for a set of optimal grouting parameters, including grouting pressure and grouting volume; and to carry out backfilling operations based on the grouting parameters.

[0129] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims. Parts not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A method for backfilling deformed roadbed when a shield tunnel passes under a railway, characterized in that, include: Controlling the tunnel boring machine to excavate a certain dynamic working distance, the time taken to complete this dynamic working distance is defined as a tunneling cycle; within this tunneling cycle, real-time data on roadbed environmental response, tunnel boring machine parameters, and geological conditions are collected. Roadbed environmental response data includes the settlement and horizontal displacement values ​​of the railway roadbed, tunnel boring machine parameters include the tunneling speed of the tunnel boring machine, cutterhead torque, and soil chamber pressure, and geological conditions data include geological type and soil mechanical parameters; At the end of the current tunneling cycle, the construction stage is determined based on the position information of the tunnel boring machine relative to the railway subgrade projection line. The construction stage includes the approach stage, the tunneling stage, and the disengagement stage. Combined with the geological condition data of the subsequent tunneling area, the dynamic working distance of the next tunneling cycle is calculated. Simultaneously, based on the geological condition data collected during the current tunneling cycle, the weighting coefficients of the shield tunneling parameters and geological condition data are calculated. The roadbed environmental response data collected during the current tunneling cycle are sorted by time to form the roadbed environmental response time series data. The roadbed environmental response time series data, shield tunneling parameters with weighted coefficients, and geological condition data with weighted coefficients are input into the settlement prediction model, and the settlement prediction results of the railway roadbed are output. Based on the optimization objectives set for the construction stage, the settlement prediction results and the optimization objectives are input into a multi-objective optimization algorithm to obtain a set of optimal grouting parameters, including grouting pressure and grouting volume. Backfilling operations shall be carried out according to the grouting parameters; Based on the position information of the tunnel boring machine relative to the railway subgrade projection line and the geological condition data of the area to be excavated, the dynamic working distance of the next excavation cycle is calculated. The specific operation is as follows: The stage adjustment coefficient K_phase is determined based on the current construction stage. The K_phase value corresponding to the approach stage is K_p1; the K_phase value corresponding to the through stage is K_p2; and the K_phase value corresponding to the departure stage is K_p3. K_p2 < K_p3 < K_p1 is satisfied, where K_p1, K_p2 and K_p3 are all real numbers in the range (0,1). Based on the geological type and soil mechanical parameters in the geological condition data, the rock mass quality classification index RQD in the field of geotechnical engineering is adopted as the benchmark value G for geological stability evaluation. Divide the RQD value by the upper limit of its grading system to obtain the standardized RQD ratio; then, using the first preset coefficient K_min as the benchmark value, add the product of the difference between the second preset coefficient K_max and K_min and the RQD ratio, and the result is the geological adjustment coefficient K_geo; where K_min and K_max are preset constant coefficients, and satisfy K_max is greater than K_min, and the values ​​of K_min and K_max are both greater than zero and less than or equal to one; The average of the phase adjustment coefficient K_phase and the geological adjustment coefficient K_geo is used as the distance adjustment coefficient; Set the maximum allowable working distance and the minimum safe distance. Multiply the maximum allowable working distance by the distance adjustment coefficient to obtain the preliminary working distance. Determine whether the preliminary working distance is greater than the minimum safe distance. If it is, use the preliminary working distance as the final dynamic working distance; otherwise, use the minimum safe distance as the final dynamic working distance.

2. The method for backfilling deformed roadbed under a railway tunnel according to claim 1, characterized in that, At the end of the current tunneling cycle, the construction stage is determined based on the position information of the tunnel boring machine relative to the railway subgrade projection line. The specific operation is as follows: The projection line of the railway subgrade onto the vertical plane containing the tunnel axis is defined as the baseline; The front face of the tunnel boring machine cutterhead is used as the positioning reference point; Calculate the horizontal distance d from the positioning reference point to the baseline; The symbolic distance D based on the tunneling direction of the tunnel boring machine is defined as follows: when the tunneling direction points to the baseline, D = d; when the tunneling direction deviates from the baseline, D = -d. The construction stage is determined based on the symbol distance D: When D>L1, it is determined to be in the approach stage, where L1 is a preset positive threshold, representing the boundary for starting to enter the high-precision control area; When -L2≤D≤L1, it is determined to be the tunneling stage, where L2 is the length of the tunnel boring machine and -L2 indicates that the positioning reference point has crossed the reference line and the distance is L2. When D < -L2, it is determined to be in the disengagement phase.

3. A method for backfilling deformed roadbed under a railway tunnel according to claim 2, characterized in that, Based on the geological condition data collected during the current tunneling cycle, the weighting coefficients of the tunnel boring machine (TBM) parameters and the geological condition data are calculated. The specific steps are as follows: The initial weights W_p0 for the shield tunneling parameters and W_g0 for the geological condition data are both set to 0.

5. The weight adjustment amount Δ is calculated based on the geological stability evaluation benchmark value G. The weight adjustment amount Δ is determined in the following way: when G is greater than the preset stability threshold G0, Δ=k*(G-G0), where k is the preset adjustment coefficient; when G is less than or equal to G0, Δ=0. The initial weights are adjusted, and the weight coefficients for the shield tunneling parameters are W_p=W_p0+Δ, and the weight coefficients for the geological condition data are W_g=W_g0–Δ.

4. A method for backfilling deformed roadbed under a railway tunnel according to claim 3, characterized in that, The settlement prediction model is based on a long short-term memory neural network. The network structure includes an input layer, a feature fusion layer, a temporal feature extraction layer, and an output layer. The input layer is used to receive preprocessed multi-source input data, including: roadbed environmental response time series data, shield tunneling parameters with weighted coefficients, and geological condition data with weighted coefficients. The feature fusion layer includes a fully connected network used to encode the weighted shield tunneling parameters and weighted geological condition data, and then splice and fuse them with the initial features of the roadbed environmental response time series data. The temporal feature extraction layer is composed of multiple stacked LSTM units, used to extract time dependencies from the fused feature sequence and capture the dynamic laws of sedimentation changes. The output layer adopts a fully connected network structure, which maps the final features extracted by the LSTM layer to the predicted settlement values ​​of key points of the railway subgrade within a certain time window in the future.

5. A method for backfilling deformed roadbed under a railway tunnel according to claim 4, characterized in that, The specific steps for training the settlement prediction model are as follows: Several training samples are obtained. Each training sample includes multi-source data extracted from a complete historical tunneling cycle, specifically including time-series data of roadbed environmental response, shield tunneling parameters, and geological conditions arranged in chronological order within that cycle. All training samples are divided into training set and validation set. The settlement prediction model with parameter initialization is trained using the training set and validated using the validation set to obtain the validation results. Set training conditions, determine whether the verification results meet the training conditions, and output the trained settlement prediction model if they meet the conditions; otherwise, continue to train the settlement prediction model using the training set.

6. A backfilling system for deformed roadbed when a shield tunnel passes under a railway, characterized in that, The system is used to implement the method for backfilling deformed roadbed under a railway via a shield tunnel as described in any one of claims 1-5, comprising: The data acquisition module is used to collect roadbed environmental response data, shield tunneling parameters and geological condition data in real time during the tunneling cycle. The roadbed environmental response data includes the settlement value and horizontal displacement value of the railway roadbed. The shield tunneling parameters include the tunneling speed of the shield machine, the cutterhead torque and the soil chamber pressure. The geological condition data includes the geological type and soil mechanical parameters. The working distance calculation module is used to determine the construction stage at the end of the current tunneling cycle based on the position information of the tunnel boring machine relative to the railway subgrade projection line. The construction stage includes the approach stage, the crossing stage, and the separation stage. Combined with the geological condition data of the subsequent tunneling area, the module calculates the dynamic working distance for the next tunneling cycle. The weighting coefficient calculation module is used to calculate the weighting coefficients of the shield tunneling parameters and geological condition data based on the geological condition data collected in the current tunneling cycle. The settlement prediction module is used to sort the roadbed environmental response data collected in the current tunneling cycle into roadbed environmental response time series data according to time. The roadbed environmental response time series data, shield tunneling parameters with weight coefficients, and geological condition data with weight coefficients are input into the settlement prediction model, and the settlement prediction results of the railway roadbed are output. The optimization decision module is used to set optimization objectives based on the construction stage, input the settlement prediction results and the optimization objectives into a multi-objective optimization algorithm, and solve for a set of optimal grouting parameters, including grouting pressure and grouting volume; and to carry out backfilling operations based on the grouting parameters.

Images