Shield tunnel dynamic settlement compensation construction method based on adaptive optimization algorithm

Abstract

The invention provides a shield tunnel dynamic settlement compensation construction method based on an adaptive optimization algorithm, and the method comprises the steps: collecting the ground surface settlement, soil stress and underground water level data in real time through an Internet of Things sensor, achieving the data preprocessing and feature extraction in combination with an edge calculation node, constructing a three-dimensional geologic model, integrating the historical engineering data through transfer learning, and achieving the dynamic settlement compensation of a shield tunnel. The method comprises the following steps: identifying a high-risk area by using a clustering algorithm, designing a hybrid adaptive optimization framework with fusion of random forest and incremental learning, dynamically adjusting shield tunneling speed and soil bin pressure construction parameters, introducing an adaptive step length mechanism to cope with geological complexity change, and identifying a settlement abnormal mode through Fourier transform. Precise compensation is achieved in combination with a layered grouting strategy, the pressure of a soil bin is dynamically adjusted based on a hydraulic system, a closed-loop feedback mechanism is established, the predicted deviation rate is compared with an actual monitoring value, model parameters and the compensation strategy are continuously optimized, the settlement control precision is improved, and the construction risk is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of shield tunnel construction, and particularly relates to a shield tunnel dynamic settlement compensation construction method based on an adaptive optimization algorithm. BACKGROUND

[0002] In current shield tunnel construction, the surface settlement control technology has significant limitations. Traditional methods rely on static geological models and empirical construction parameter adjustments such as fixed advancing speed and soil chamber pressure, which are difficult to cope with complex conditions such as sudden changes in soft strata and fluctuations in underground water levels in composite strata. Due to the lack of real-time monitoring and dynamic feedback mechanisms, such technologies are prone to risks such as over-excavation, segment misalignment, and even ground collapse. Although some projects have introduced Internet of Things sensors to collect settlement data, they are limited by data transmission delays, insufficient edge computing capabilities, and difficulties in fusing multi-source heterogeneous data, making it difficult to achieve real-time optimization of construction parameters. In addition, existing three-dimensional geological models rely on manual correction, have long update cycles, and cannot dynamically adapt to changes in strata during construction, resulting in a disconnect between the model and the actual working conditions.

[0003] In terms of settlement compensation strategies, traditional methods mostly use single-stage grouting without combining real-time settlement prediction with dynamic feedback from geological risk areas, resulting in limited compensation effects. For example, the amount of synchronous grouting is often calculated using a fixed formula, ignoring the different needs of soft strata and hard rock strata. Secondary grouting relies on manual experience to trigger, with significant lag. Meanwhile, soil chamber pressure regulation still relies on static water and soil pressure calculations, without considering the impact of dynamic settlement trends and construction parameter linkage.

[0004] The core problems that need to be broken through in existing technologies include: insufficient real-time fusion and intelligent analysis capabilities of multi-source heterogeneous data, inability to dynamically adjust construction parameters; three-dimensional geological models lack a closed-loop optimization mechanism, with low model update efficiency and difficulty in adapting to complex strata; grouting compensation strategies lack hierarchical collaborative control and are difficult to respond to sudden settlement risks. The shield tunnel dynamic settlement compensation construction method based on an adaptive optimization algorithm proposed in the present application realizes intelligent regulation and control of construction parameters and active intervention in settlement risks through real-time monitoring of the Internet of Things, hybrid machine learning models, and hierarchical grouting strategies, improving the safety and efficiency of construction in complex strata. SUMMARY

[0005] The purpose of the present application is to provide a shield tunnel dynamic settlement compensation construction method based on an adaptive optimization algorithm.

[0006] The problem to be solved by the present application is that the method addresses the difficulties in surface settlement control during shield tunnel construction and solves technical problems such as real-time monitoring lag, delayed geological model updates, passive parameter adjustments, and inaccurate risk identification in traditional construction.

[0007] The adaptive optimization algorithm-based shield tunnel dynamic settlement compensation construction method adopts the technical solutions as follows: S1: Deploy Internet of Things sensors to collect surface settlement, soil stress, and underground water level data in real time, and deploy edge computing nodes at the construction site to perform real-time preprocessing on the original data; S2: Construct a simplified three-dimensional geological model, update the model in real time through real-time monitoring data, migrate historical data of shield tunnel projects to the current project based on transfer learning technology, and use clustering analysis algorithms in machine learning to identify high-risk areas, including soft strata and underground cavities; S3: Design a hybrid adaptive optimization algorithm framework combining random forests and incremental learning, analyze features based on random forests, dynamically adjust shield construction parameters including advance speed and soil chamber pressure, update the model in real time using newly collected monitoring data, and introduce an adaptive step mechanism to automatically adjust the algorithm iteration step based on geological complexity; S4: Input real-time data into the model established based on the hybrid adaptive optimization algorithm framework to simulate the settlement trend in the next 30 minutes. If the predicted settlement is close to the threshold, identify abnormal settlement patterns including local mutations and periodic fluctuations, trigger an early warning, and generate a compensation strategy; S5: Based on the output of the hybrid adaptive optimization algorithm, generate a layered grouting compensation strategy. The first layer is synchronous grouting, which immediately grouts behind the shield machine. The second layer is secondary grouting, which supplements grouting based on settlement monitoring results, and dynamically adjusts soil chamber pressure using a hydraulic control system; S6: Based on the comparison of predicted settlement and actual monitoring values, calculate the deviation rate, feed it back to the hybrid adaptive optimization algorithm, update the model parameters, and select the most effective data for model optimization and record the compensation strategy.

[0008] Further, the deployment of Internet of Things sensors in S1 to collect surface settlement, soil stress, and underground water level data in real time includes: Multiple types of Internet of Things sensors are deployed at key positions on the surface and underground soil in the shield tunnel construction area, including the shield machine head, pipe joint, and soft stratum area. Based on NB-IoT, sensor data is transmitted over long distances; Deploy surface settlement sensor laser range finders to measure surface elevation changes with high precision and capture millimeter-level settlement data in real time. Deploy soil stress sensors, strain gauges, embedded in the shield machine cutterhead and pipe segments, to monitor stress changes including soil shear force and compression force. Deploy underground water level sensors, pressure type water level meters in the tunnel periphery aquifer, to calculate underground water level dynamics through pressure signals.

[0009] Further, the deployment of edge computing nodes in S1 at the construction site to perform real-time preprocessing on the original data includes: Based on the low-pass filter, the instantaneous abnormal value generated by the vibration of the sensor is eliminated, the sampling rate of the low-frequency change data including the stable period of groundwater level is reduced, and the time correlation of the sensor data is identified, including the sudden increase of the settlement rate in the shield construction area for 3 minutes. The sensor data in the same area is combined with 5 adjacent ground settlement points as a comprehensive index.

[0010] Further, the simplified three-dimensional geological model is constructed in S2, and the model is updated in real time by driving data, including: Combined with the ground settlement, soil stress, and groundwater level data collected by the Internet of Things sensors deployed in S1, combined with the geological exploration report before construction including drilling data and geological radar scanning results, as the basic data for three-dimensional modeling; Using differential modeling technology, the geological space is divided into different regions including soft stratum, hard rock stratum, and water-bearing stratum, each region is modeled independently, and based on Boolean operation including intersection and union, the stratum model is automatically aligned with the tunnel structure model including segment position and shield machine path to form a stratum-structure integrated model. The integrated model is simplified into a two-dimensional slice using the contour picking algorithm; Based on real-time data, the model parameters including soil elastic modulus and permeability coefficient are adjusted using NSGA-II algorithm, the update frequency is automatically adjusted according to the geological complexity, the soft stratum region is updated every 10 seconds, and the stable region is updated every hour. The distributed database SQLite is used to record the history version of each model update, and it supports rollback to any time point.

[0011] Further, the historical data of the shield tunnel project is migrated to the current project in S2, and the clustering analysis algorithm in machine learning is used to identify high-risk areas, including: The feature similarity index including ground settlement, soil stress, and groundwater level data is extracted from the historical database of the shield tunnel project, the matching project data is filtered, the common features including stratum shear modulus and grouting effect correlation are extracted, the distribution difference between historical data and the current project is adjusted, and the deviation caused by different geological conditions is eliminated. The expert experience of the historical project including the need to improve the soil pressure in soft stratum is converted into a rule base and embedded into the current model as prior knowledge; Based on the Euclidean distance, the region is divided into three categories: high-risk soft stratum, medium-risk ordinary sand, and low-risk hard rock stratum. Based on the DBSCAN algorithm, local density anomaly areas including underground cavities are identified, and the clustering results are marked with colors in the model.

[0012] Further, a hybrid adaptive optimization algorithm framework of random forest + incremental learning is designed in S3, and the shield construction parameters including the advance speed and soil pressure are dynamically adjusted, and the adaptive step mechanism is introduced. The algorithm iteration step is automatically adjusted according to the geological complexity, including: S31: Based on the real-time monitoring data provided by the Internet of Things sensors in S1, the ground settlement , soil stress , groundwater level and the clustering results of the three-dimensional geological model in S2 as input features, the feature data format is standardized to a time series matrix; S32: Based on the feature importance ranking of the random forest, determine the key driving factors, input the newly collected monitoring data into the model in batches, and realize dynamic updating of the model; S33: Based on the risk score output by the random forest , develop dynamic adjustment rules for shield construction parameters including advance speed , soil pressure , , where is the reference speed, is the upper limit of the risk score, is the reference pressure, is the adjustment coefficient, which is 0.1, is the safety stress threshold, is the current soil stress; S34: Combine the edge computing nodes in S1 to transmit the adjusted advance speed and soil pressure parameters to the shield control system, where is the response time of the shield machine after parameter adjustment, is the length of the shield machine cutter; S35: According to the clustering results in S2 and the sensor data in S1, calculate the geological complexity index , where is the risk score of the ith region, is the ground settlement value of the ith region, is the settlement safety threshold, set the algorithm iteration step , is the reference step, set to 0.1, is the complexity amplification coefficient, take 0.5, the iteration step does not exceed 1.0, in the random forest training process, adjust the model update frequency according to the current step.

[0013] Further, the S4 simulates the settlement trend in the next 30 minutes, identifies the settlement abnormal pattern including local mutation and periodic fluctuation, triggers the early warning and generates the compensation strategy, including: S41: input the current monitoring data, combine the geological complexity index and risk score to predict ground settlement, introduce an adaptive step mechanism in the simulation process, and determine whether to trigger an early warning based on the set settlement threshold. If the settlement value within 30 minutes is ≥ 90% of the settlement threshold, trigger the anomaly pattern recognition; S42: Based on the short-term change rate of the current settlement data, if the sudden increase exceeds 50% of the reference value, it is determined to be a local mutation. Combine the three-dimensional geological model in S2 to determine whether the mutation area corresponds to a soft stratum or a cavity. S43: Fourier transform analysis of the settlement time series to identify whether there is periodic fluctuation. If the fluctuation frequency is related to the shield machine propulsion cycle, it is determined to be a periodic anomaly. S44: The first level warning is that the settlement value within 30 minutes is ≥ 80% of the settlement threshold, triggering the LED screen prompt and on-site sound and light alarm, and manual review is taken. The second level warning is that the settlement value within 30 minutes is ≥ 90% of the settlement threshold, automatically suspending the shield machine propulsion, and adjusting the shield construction parameters based on the control system in S3.

[0014] Further, the output based on the hybrid adaptive optimization algorithm in S5 generates a hierarchical grouting compensation strategy, which dynamically adjusts the soil bin pressure using a hydraulic control system, including: S51: Synchronous grouting amount According to the shield machine propulsion speed And the geological complexity index Dynamic adjustment, , The basic grouting coefficient is 0.5, The propulsion time, The geological complexity correction coefficient, including 0.3 for soft stratum and 0.1 for hard rock stratum; S52: Control grouting pressure , Grouting pressure needs to be higher than static water and soil pressure , , The settlement prediction correction coefficient is 0.05, The settlement trend value within the next 30 minutes simulated in S4, automatically adjust the grouting pressure to suppress sudden settlement; S53: If the settlement deviation rate Exceeds the threshold of 15%, start secondary grouting, Where Is the true settlement value, the secondary grouting amount According to the settlement deviation rate And the affected area Calculate, Where The compensation coefficient is 2, The risk score correction coefficient includes 0.5 for high-risk areas, The risk score based on the random forest output, the grouting position is accurately positioned through the three-dimensional geological model in S2, and the local mutation area is preferentially filled; S54: soil bin pressure Dynamic balance of excavation face water and soil pressure And the settlement control requirement, Wherein The settlement correction coefficient is 0.1, The advance speed variation rate correction coefficient is 0.02, The advance speed variation rate is based on the calculated value, and the hydraulic control system is used to dynamically adjust the soil bin pressure.

[0015] Further, the feedback to the mixed adaptive optimization algorithm in S6 includes: The difference data between the predicted settlement and the actual monitoring value are returned, including timestamp, position, deviation rate D, and high deviation data is returned D>15%, if the deviation mainly comes from the prediction error of the soil stress The feature weight of Is reduced, and the model iteration step is reduced in the high complexity area C>0.8.

[0016] Further, the S6 filters the data most effective for model optimization and records the compensation strategy, including: From the historical data, the data most effective for model optimization are filtered, the deviation rate D>15% and located in the high-risk area >0.7, as an incremental training set, the feature importance is recalculated based on the random forest model, and the model parameters are updated; The synchronous grouting amount in S5 The secondary grouting amount The soil bin pressure adjustment range ΔP parameter is converted into structured data, and is recorded to the compensation strategy knowledge base, and the abnormal mode recognition result of S4 including local mutation and periodic fluctuation is combined, and the optimal compensation strategy is bound for each abnormal type.

[0017] The beneficial effects of the present application are: the mixed adaptive optimization algorithm based on random forest+incremental learning, combined with the geological complexity index and the risk score, dynamically adjusts the construction parameters of the advance speed and the soil bin pressure, significantly reduces the settlement risk caused by geological mutation or parameter mismatch, and ensures the safety of the construction process; Through the combination of historical data transfer learning and real-time monitoring data, a simplified three-dimensional geological model is constructed and continuously updated, and a clustering algorithm is used to identify high-risk areas such as soft stratum and underground cavity, to provide accurate geological basis for settlement prediction; The combination of random forest feature importance ranking and adaptive step mechanism enables the model to quickly respond to geological changes and dynamically adjust the iteration frequency, such as updating every 10 seconds in soft stratum areas, significantly improving the prediction accuracy and timeliness of compensation strategies. By Fourier transform to identify periodic fluctuations, short-time change rate to judge local mutations, combined with threshold triggering first-level (80% settlement threshold) and second-level (90% settlement threshold) early warning, realizing hierarchical response from manual review to automatic suspension of shield machine, minimizing the impact of settlement on surrounding buildings. Combined with synchronous grouting and secondary grouting, based on settlement deviation rate (D>15%), dynamically adjusting grouting amount and pressure, preferentially filling local mutation areas, through hydraulic control system to realize real-time balance of soil chamber pressure, forming a closed-loop control compensation system. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 Flow chart of adaptive optimization algorithm-based dynamic settlement compensation construction method for shield tunnel. DETAILED DESCRIPTION

[0019] The present application will be further clarified by the following detailed description, but the scope of protection of the present application is not limited thereto.

[0020] The adaptive optimization algorithm-based dynamic settlement compensation construction method for shield tunnel employs the following technical solutions: S1: Deploy Internet of Things sensors to collect real-time ground settlement, soil stress, and underground water level data, and deploy edge computing nodes on the construction site to pre-process the original data in real time; S2: Construct a simplified three-dimensional geological model, update the model based on real-time monitoring data, migrate historical data of shield tunnel projects to the current project based on transfer learning technology, and use clustering analysis algorithms in machine learning to identify high-risk areas, including soft strata and underground cavities; S3: Design a hybrid adaptive optimization algorithm framework based on random forest and incremental learning, dynamically adjust shield construction parameters including advance speed and soil chamber pressure based on random forest analysis features, update the model in real time with newly collected monitoring data, and introduce an adaptive step mechanism to automatically adjust the algorithm iteration step according to the complexity of the geology; S4: Input real-time data into the model established based on the hybrid adaptive optimization algorithm framework to simulate the settlement trend in the next 30 minutes, identify abnormal settlement patterns including local mutations and periodic fluctuations if the predicted settlement amount approaches the threshold, trigger an early warning and generate a compensation strategy; S5: Based on the output of the hybrid adaptive optimization algorithm, generate a hierarchical grouting compensation strategy, the first layer is synchronous grouting, grouting immediately behind the shield machine, the second layer is secondary grouting, according to the settlement monitoring results, supplement grouting, use hydraulic control system to dynamically adjust the soil chamber pressure; S6: Based on the comparison of predicted settlement and actual monitoring value, calculate the deviation rate, feedback to the hybrid adaptive optimization algorithm, update the model parameters, select the most effective data for model optimization and record the compensation strategy.

[0021] Reference Figure 1 As shown in the figure, it is a flow chart of the adaptive optimization algorithm-based dynamic settlement compensation construction method of shield tunnel.

[0022] Further, the S1 deployed Internet of Things sensors collect real-time ground settlement, soil stress, and underground water level data, including: In the ground surface and underground soil of the shield tunnel construction area, including the shield machine head, segment joint, and soft soil area, multiple types of Internet of Things sensors are deployed, and sensor data is transmitted remotely based on NB-IoT; Deploy ground settlement sensor laser range finder to measure ground elevation changes with high precision, capture millimeter-level settlement data in real time, deploy soil stress sensor strain gauge embedded in the shield machine cutterhead and segments to monitor stress changes including soil shear force and compression force, and deploy underground water level sensor as pressure type water level gauge in the tunnel surrounding water-bearing layer to calculate underground water level dynamically through pressure signal.

[0023] Further, the S1 deployed edge computing nodes in the construction site for real-time preprocessing of raw data, including: Based on the low-pass filter, eliminate transient abnormal values generated by sensor vibration, reduce the sampling rate of low-frequency change data including stable period underground water level, identify the time correlation of sensor data including sudden increase of settlement rate in shield construction area for 3 minutes, and combine sensor data in the same area, 5 adjacent ground settlement points into comprehensive indicators.

[0024] Further, the S2 constructs a simplified three-dimensional geological model, and updates the model through real-time monitoring data, including: Combined with the ground settlement, soil stress, and underground water level data collected by the Internet of Things sensors deployed in S1, combined with the geological exploration report before construction including drilling data and geological radar scanning results as basic data for three-dimensional modeling; Adopting differential modeling technology, the geological space is divided into different regions including soft stratum, hard rock stratum and aquifer, each region is independently modeled, and the stratum model is automatically aligned with the tunnel structure model including segment position and shield machine path based on Boolean operation including intersection and union to form a stratum-structure integrated model, and the integrated model is simplified into a two-dimensional slice by using contour picking algorithm; Based on real-time data, the model parameters including soil elastic modulus and permeability coefficient are adjusted by combining NSGA-II algorithm, the update frequency is automatically adjusted according to the geological complexity, the soft stratum region is updated every 10 seconds, the stable region is updated every hour, the distributed database SQLite is used to record the historical version of each model update, and rollback to any time point is supported.

[0025] Further, the historical data of shield tunnel engineering in S2 is migrated to the current project, and the clustering analysis algorithm in machine learning is used to identify high-risk areas, including: The feature similarity index including ground settlement, soil stress and underground water level data is extracted from the historical database of shield tunnel engineering, the matching project data is screened, the common features including stratum shear modulus and grouting effect correlation are extracted, the distribution difference between historical data and the current project is adjusted, the deviation caused by different geological conditions is eliminated, the expert experience of historical projects including the need to improve the soil pressure of soft stratum is converted into a rule base, and embedded into the current model as prior knowledge; Based on Euclidean distance, the region is divided into three categories: high-risk soft stratum, medium-risk ordinary sand and low-risk hard rock, and based on DBSCAN algorithm, the local density anomaly area including underground cavity is identified, and the clustering results are marked with color in the model.

[0026] Further, a hybrid adaptive optimization algorithm framework of random forest + incremental learning is designed in S3, the shield construction parameters including advancing speed and soil pressure are dynamically adjusted, and an adaptive step mechanism is introduced, the algorithm iteration step is automatically adjusted according to the geological complexity, including: S31: Based on the real-time monitoring data provided by the Internet of Things sensors in S1, the ground settlement , soil stress , underground water level and the clustering results of the three-dimensional geological model in S2 are used as input features, and the feature data format is standardized as a time series matrix; S32: The key driving factors are determined based on the feature importance ranking of random forest, the newly collected monitoring data is input into the model in batches, and the model is dynamically updated; S33: Based on the risk score output by random forest, the shield construction parameters including advancing speed and soil pressure dynamic adjustment rules, wherein is the reference speed, is the upper limit of the risk score, is the reference pressure, is the adjustment coefficient taking 0.1, is the safety stress threshold, is the current soil stress; S34: The adjusted advancing speed and the soil chamber pressure parameters are transmitted to the shield control system in combination with the edge computing node in S1, is the response time of the shield machine after parameter adjustment, is the length of the cutter head of the shield machine; S35: According to the clustering results in S2 and the sensor data in S1, the geological complexity index is calculated is the risk score of the ith region, is the surface subsidence value of the ith region, is the subsidence safety threshold, and the algorithm iteration step is set is the reference step, set to 0.1, is the complexity amplification coefficient taking 0.5, and the iteration step is not more than 1.0. In the random forest training process, the model update frequency is adjusted according to the current step.

[0027] Further, the S4 simulates the subsidence trend in the next 30 minutes, identifies the subsidence abnormal pattern including local mutation and periodic fluctuation, triggers the early warning and generates the compensation strategy, including: S41: The current monitoring data is taken as the input, the geological complexity index and the risk score are combined to predict the surface subsidence, and an adaptive step mechanism is introduced in the simulation process. Based on the set subsidence threshold, it is judged whether the early warning is triggered. If the subsidence value in 30 minutes is ≥90% of the subsidence threshold, the abnormal pattern recognition is triggered; S42: Based on the short-time change rate of the current subsidence data, if the sudden increase exceeds 50% of the reference value, it is determined as a local mutation. In combination with the three-dimensional geological model in S2, it is located whether the mutation region corresponds to a soft stratum or a cavity; S43: The Fourier transform analysis is performed on the subsidence time series to identify whether there is periodic fluctuation. If the fluctuation frequency is related to the advancing period of the shield machine, it is determined as a periodic abnormality; ​​​​​S44: The first early warning is the settlement threshold value ≥ 80% of the settlement value within 30 minutes, triggering the LED screen prompt and on-site sound and light alarm, manual review is taken, the second early warning is the settlement threshold value ≥ 90% of the settlement value within 30 minutes, automatically suspending the shield machine advancing, and adjusting the shield construction parameters based on the control system of S3.

[0028] Further, the S5 based on the hybrid adaptive optimization algorithm outputs generates a hierarchical grouting compensation strategy, and a hydraulic control system is used to dynamically adjust the soil bin pressure, including: S51: Synchronous grouting amount According to the shield machine advancing speed And the geological complexity index Dynamically adjust, , The basic grouting coefficient is 0.5, The advancing time, The geological complexity correction coefficient, including 0.3 for soft stratum and 0.1 for hard rock stratum; S52: Control grouting pressure , The grouting pressure needs to be higher than the static water and soil pressure , , The settlement prediction correction coefficient is 0.05, The settlement trend value within the next 30 minutes simulated in S4, automatically adjusting the grouting pressure to suppress sudden settlement; S53: If the settlement deviation rate Exceeds the threshold value 15%, start secondary grouting, Wherein The real settlement value, the secondary grouting amount According to the settlement deviation rate And the affected area Calculate, Wherein The compensation coefficient is 2, The risk score correction coefficient includes 0.5 for high-risk areas, The risk score based on the random forest output, the grouting position is accurately positioned through the three-dimensional geological model in S2, and the local mutation area is preferentially filled; S54: The soil bin pressure Dynamically balances the water and soil pressure of the excavation face And the settlement control demand, Wherein The settlement correction coefficient is 0.1, The advancing speed change rate correction coefficient is 0.02, The advancing speed change rate, based on the calculated value, the hydraulic control system is used to dynamically adjust the soil bin pressure.

[0029] Further, the feedback in S6 to the hybrid adaptive optimization algorithm includes: The difference data between the predicted settlement and the actual monitoring value are returned, including time stamp, position, deviation rate D, and high deviation data D>15% is returned. If the deviation mainly comes from the prediction error of the soil stress, the feature weight of the model is reduced, and the model iteration step is reduced in the high complexity area C>0.8.

[0030] Further, the S6 screens the data most effective for model optimization and records the compensation strategy, including: The data most effective for model optimization are screened from historical data, the deviation rate D>15% and located in the high-risk area C>0.7, as an incremental training set, the feature importance is recalculated based on the random forest model, and the model parameters are updated. The synchronous grouting amount in S5 The secondary grouting amount The soil chamber pressure adjustment range ΔP parameter is converted into structured data and recorded to the compensation strategy knowledge base, and combined with the abnormal pattern recognition result of S4 including local mutation and periodic fluctuation, the optimal compensation strategy is bound for each abnormal type.

[0031] The present application provides a shield tunnel dynamic settlement compensation construction method based on an adaptive optimization algorithm, which collects ground settlement, soil stress and underground water level data in real time through Internet of Things sensors, realizes data preprocessing and feature extraction combined with edge computing nodes, constructs a three-dimensional geological model and integrates historical engineering data using transfer learning, identifies high-risk areas using clustering algorithm, designs a hybrid adaptive optimization framework combining random forest and incremental learning, dynamically adjusts the shield advancing speed and soil chamber pressure construction parameters, introduces an adaptive step mechanism to cope with the change of geological complexity, identifies settlement abnormal patterns through Fourier transform, realizes accurate compensation combined with layered grouting strategy, and dynamically adjusts the soil chamber pressure based on the hydraulic system. A closed-loop feedback mechanism is established to compare the predicted deviation rate with the actual monitoring value, continuously optimize the model parameters and compensation strategy, improve the settlement control accuracy, and reduce the construction risk.​​

Claims

1. A shield tunnel dynamic settlement compensation construction method based on an adaptive optimization algorithm, characterized in that: include: S1: Deploy IoT sensors to collect real-time data on surface settlement, soil stress, and groundwater levels. Deploy edge computing nodes at the construction site to pre-process the raw data in real time. S2: Build a simplified 3D geological model and use real-time monitoring data to drive model updates. Based on transfer learning technology, historical data from shield tunneling projects will be migrated to the current project. Cluster analysis algorithms in machine learning will be used to identify high-risk areas, such as weak strata and underground cavities. S3: Design a hybrid adaptive optimization algorithm framework combining random forest and incremental learning. Based on the analysis characteristics of random forest, dynamically adjust shield construction parameters including propulsion speed and soil bin pressure. Use newly collected monitoring data to update the model in real time. Introduce an adaptive step size mechanism to automatically adjust the algorithm iteration step size according to geological complexity. S4: Input real-time data into a model built based on a hybrid adaptive optimization algorithm framework to simulate the settlement trend within the next 30 minutes. If the predicted settlement amount approaches the threshold, the model identifies abnormal settlement patterns, including local mutations and periodic fluctuations, triggers an early warning, and generates a compensation strategy. S5: Based on the output of the hybrid adaptive optimization algorithm, a layered grouting compensation strategy is generated. The first layer is synchronous grouting, with grouting immediately behind the shield machine. The second layer is secondary grouting, with additional grouting based on settlement monitoring results. The hydraulic control system is used to dynamically adjust the soil bin pressure. S6: Based on the comparison between the predicted settlement and the actual monitoring value, the deviation rate is calculated and fed back to the hybrid adaptive optimization algorithm to update the model parameters, select the most effective data for model optimization and record the compensation strategy.

2. The shield tunnel dynamic settlement compensation construction method based on the adaptive optimization algorithm according to claim 1 is characterized in that: IoT sensors deployed in S1 collect real-time data on surface settlement, soil stress, and groundwater level, including: Multiple IoT sensors are deployed at key locations on the surface and underground soil within the shield tunnel construction area, including the shield machine head, segment joints, and areas of weak strata. NB-IoT is used to enable long-distance transmission of sensor data. Surface settlement sensor laser rangefinders are deployed to measure surface elevation changes with high precision and capture millimeter-level settlement data in real time. Soil stress sensor strain gauges are deployed and embedded in the cutterhead and segments of the shield machine to monitor stress changes, including soil shear and compression forces. Groundwater level sensors are pressure-type water level gauges deployed in the aquifers around the tunnel to infer groundwater level dynamics through pressure signals.

3. The shield tunnel dynamic settlement compensation construction method based on the adaptive optimization algorithm according to claim 1 is characterized in that: Deploy edge computing nodes at the construction site to perform real-time pre-processing of raw data, including: Based on a low-pass filter, instantaneous outliers caused by sensor vibration are eliminated, the sampling rate of low-frequency changing data, including the groundwater level during the stable period, is reduced, and the temporal correlation of sensor data, including a sudden increase in settlement rate for three consecutive minutes in the shield construction area, is identified. The sensor data of five adjacent surface settlement points in the same area are merged into a comprehensive indicator.

4. The shield tunnel dynamic settlement compensation construction method based on the adaptive optimization algorithm according to claim 1 is characterized in that: S2 constructs a simplified 3D geological model and drives model updates through real-time monitoring data, including: Combine the surface settlement, soil stress, and groundwater level data collected by the IoT sensors deployed as described in S1 with the pre-construction geological survey report including drilling data and geological radar scanning results as the basic data for 3D modeling; Using differentiated modeling technology, the geological space is divided into different regions, including weak strata, hard rock strata, and aquifers. Each region is modeled independently. Boolean operations, including intersection and union, are used to automatically align the stratum model with the tunnel structure model, including segment positions and shield machine paths, to form an integrated stratum-structure model. A contour picking algorithm is used to simplify the integrated model into two-dimensional slices. Model parameters including soil elastic modulus and permeability are adjusted based on real-time data combined with the NSGA-II algorithm. The update frequency is automatically adjusted according to geological complexity, with updates every 10 seconds in weak strata areas and every hour in stable areas. The distributed database SQLite is used to record the historical version of each model update, supporting rollback to any point in time.

5. The shield tunnel dynamic settlement compensation construction method based on the adaptive optimization algorithm according to claim 1 is characterized in that: Migrate historical data from shield tunneling projects to the current project and use cluster analysis algorithms in machine learning to identify high-risk areas, including: Feature similarity indicators, including surface settlement, soil stress, and groundwater level data, were extracted from the historical database of shield tunnel projects. Matching project data was screened and common features, including stratum shear modulus and correlation with grouting effects, were extracted. Distribution differences between historical data and current projects were adjusted to eliminate deviations caused by different geological conditions. Expert experience from historical projects, including the need to increase soil bin pressure in weak strata, was converted into a rule base and embedded into the current model as prior knowledge. Based on Euclidean distance, the area is divided into three categories: high risk (soft strata), medium risk (ordinary sand and soil), and low risk (hard rock strata). Based on the DBSCAN algorithm, local density anomaly areas including underground cavities are identified, and clustering results are marked with colors in the model.

6. The shield tunnel dynamic settlement compensation construction method based on the adaptive optimization algorithm according to claim 1 is characterized in that: The S3 framework is a hybrid adaptive optimization algorithm framework combining random forest and incremental learning. It dynamically adjusts shield construction parameters including propulsion speed and soil bin pressure. It also introduces an adaptive step size mechanism to automatically adjust the algorithm iteration step size according to geological complexity, including: S31: Based on the real-time monitoring data provided by the IoT sensors described in S1, surface subsidence , soil stress , groundwater level The clustering results of the 3D geological model described in S2 are used as input features, and the feature data format is standardized into a time series matrix; S32: Determine the key driving factors based on the feature importance ranking of random forest, input the newly collected monitoring data into the model in batches, and realize dynamic model update; S33: Risk scoring based on random forest output , formulate shield construction parameters including advancement speed , soil bin pressure Dynamic adjustment rules, , ,in is the base speed, is the upper limit of risk score, is the base pressure, The adjustment coefficient is 0.

1. is the safety stress threshold, is the current soil stress; S34: Combined with the edge computing node in S1, the adjusted propulsion speed and soil bin pressure parameters are transmitted to the shield control system. ,in is the response time of the shield machine after parameter adjustment, is the cutterhead length of the shield machine; S35: Calculate the geological complexity index based on the clustering results described in S2 and the sensor data described in S1 , ,in is the risk score of the i-th region, is the surface settlement value of the ith region, Set the algorithm iteration step length as the settlement safety threshold , is the base step size, set to 0.1, The complexity amplification factor is set to 0.5, and the iteration step size does not exceed 1.

0. During the random forest training process, the model update frequency is adjusted according to the current step size.

7. The shield tunnel dynamic settlement compensation construction method based on the adaptive optimization algorithm according to claim 1 is characterized in that: S4 simulates the settlement trend in the next 30 minutes, identifies abnormal settlement patterns including local mutations and periodic fluctuations, triggers early warnings, and generates compensation strategies, including: S41: Using current monitoring data as input, combined with the geological complexity index and risk score, surface settlement is predicted. An adaptive step-size mechanism is introduced into the simulation process to determine whether to trigger an early warning based on the set settlement threshold. If the settlement value within 30 minutes is ≥90% of the settlement threshold, abnormal pattern recognition is triggered. S42: Based on the short-term change rate of the current settlement data, if the sudden increase exceeds 50% of the baseline value, it is determined to be a local mutation. Combined with the 3D geological model in S2, it is determined whether the mutation area corresponds to a weak stratum or a cavity. S43: Perform Fourier transform analysis on the settlement time series to identify whether there is periodic fluctuation. If the fluctuation frequency is related to the shield machine propulsion cycle, it is determined to be a periodic anomaly; S44: The first-level warning is when the settlement value within 30 minutes is ≥80% of the settlement threshold, triggering the LED screen prompt and on-site sound and light alarm, and manual review is required. The second-level warning is when the settlement value within 30 minutes is ≥90% of the settlement threshold, automatically suspending the shield machine advance and adjusting the shield construction parameters based on the S3 control system.

8. The shield tunnel dynamic settlement compensation construction method based on the adaptive optimization algorithm according to claim 1 is characterized in that: In S5, based on the output of the hybrid adaptive optimization algorithm, a layered grouting compensation strategy is generated, and a hydraulic control system is used to dynamically adjust the soil bin pressure, including: S51: Synchronous grouting volume According to the shield machine's advancing speed and geological complexity index Dynamic adjustment, , The basic grouting coefficient is taken as 0.5, To advance time, is the correction factor for geological complexity, including 0.3 for soft strata and 0.1 for hard rock strata; S52: Control grouting pressure , grouting pressure needs to be higher than static water and soil pressure , , The settlement prediction correction coefficient is taken as 0.05, The settlement trend value within the next 30 minutes simulated in S4 is used to automatically adjust the grouting pressure to suppress sudden settlement; S53: If the settlement deviation rate When the threshold exceeds 15%, secondary grouting is started. ,in is the true settlement value, the secondary grouting volume According to the settlement deviation rate and affected area calculate, ,in The compensation coefficient is 2. The risk score correction factor includes high-risk areas and is taken as 0.

5. To obtain the risk score based on the random forest output, the grouting locations are precisely located using the 3D geological model in S2, with local mutation areas being filled first; S54: Soil Bin Pressure Dynamically balance water and soil pressure on the excavation surface and settlement control needs, ,in The settlement correction coefficient is taken as 0.1, The correction coefficient for the propulsion speed change rate is 0.

02. In order to adjust the speed change rate, the hydraulic control system is used to dynamically adjust the soil bin pressure based on the calculated value.

9. The shield tunnel dynamic settlement compensation construction method based on the adaptive optimization algorithm according to claim 1, characterized in that: The feedback to the hybrid adaptive optimization algorithm in S6 includes: The difference data between the predicted settlement and the actual monitoring value is returned, including the timestamp, location, and deviation rate D. The high deviation data returned is D>15%. If the deviation is mainly due to soil stress The prediction error is reduced In the high complexity area C>0.8, the model iteration step is reduced.

10. The shield tunnel dynamic settlement compensation construction method based on the adaptive optimization algorithm according to claim 1, characterized in that: Screen the data that is most effective for model optimization and record compensation strategies, including: Filter the most effective data for model optimization from historical data, with a deviation rate D>15% and in the high-risk area >0.7, used as incremental training set, recalculate feature importance based on random forest model, and update model parameters; The synchronous grouting volume in S5 , secondary grouting volume The soil bin pressure adjustment amplitude ΔP parameter is converted into structured data and recorded in the compensation strategy knowledge base. Combined with the abnormal pattern recognition results of S4, including local mutations and periodic fluctuations, the optimal compensation strategy is bound to each abnormality type.

Images