Method and system for monitoring offset of adjacent subway foundation pit support

Abstract

The invention relates to the technical field of building monitoring, and discloses a method and a system for monitoring support offset of an adjacent subway foundation pit, and the method comprises the following steps: firstly, finding a group of better monitoring parameters as an initial monitoring scheme according to larger redundancy setting by using a genetic algorithm in a foundation pit-subway tunnel-soil coupling model established before construction; according to the method, the monitoring parameters are manually adjusted according to field detection data, during monitoring in construction, risk warning is carried out, a prediction model is used, the support offset is predicted according to accumulated monitoring data, an optimal automatic monitoring scheme for the support offset of the adjacent subway foundation pit is automatically generated, automatic monitoring is carried out, and the construction efficiency is improved. The monitoring cost and the monitoring difficulty are reduced, and the technical problems that in the prior art, setting of a monitoring scheme depends on local engineering experience, and the requirement for the professional level of monitoring personnel is high are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of building monitoring, in particular to a method and system for monitoring the offset of foundation pit support near a subway. Background Art

[0002] A building foundation pit is a space excavated from the ground for the construction of the underground part of a building (structure), referred to as a foundation pit. The support structure of the foundation pit is a structural system that supports and reinforces the side walls of the foundation pit to ensure the safety of the foundation pit excavation and the construction of the underground structure, as well as to protect the surrounding environment of the foundation pit. It includes retaining walls and support (or anchor) systems. Foundation pits adjacent to subways are usually located in prosperous areas of the city, with dense surrounding buildings and complex underground pipelines. The impact of the surrounding environment on the stability of the foundation pit needs to be fully considered. At the same time, the construction period of foundation pit projects adjacent to subways is relatively long, generally taking several months or even more than a year. The offset of the support structure needs to be monitored for a long time at a certain monitoring frequency.

[0003] In existing technologies, the development of monitoring plans, such as setting early warning values for foundation pits and supporting structures, monitoring frequency, and the location and number of monitoring points, often relies on local engineering experience and places high demands on the professional expertise of monitoring personnel. While relevant regulations require that the builder commission a qualified third party to conduct on-site monitoring of the foundation pit prior to construction, third-party monitoring does not replace the necessary construction monitoring performed by the construction unit itself, which must continue to conduct necessary monitoring throughout the construction process.

[0004] Therefore, a method and system for monitoring the support offset of the adjacent subway foundation pit is needed, so that the construction party can efficiently and accurately monitor the support offset by accumulating local construction experience during the long construction period of the adjacent subway foundation pit project. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for monitoring the support offset of a foundation pit adjacent to a subway. Before construction, a genetic algorithm is used in a coupling model of foundation pit-subway tunnel-soil. According to a large redundancy setting, a set of better monitoring parameters is first found as an initial monitoring scheme, and then the monitoring parameters are manually adjusted according to the field detection data. When monitoring during construction, in addition to risk warnings, a prediction model is also used to predict the support offset based on the accumulated monitoring data. This realizes the automatic generation of a better automatic monitoring scheme for the support offset of the foundation pit adjacent to the subway and performs automatic monitoring, reduces the monitoring cost and difficulty, and solves the technical problem in the prior art that the setting of the monitoring scheme depends on local engineering experience and has high requirements on the professional level of the monitoring personnel.

[0006] The present invention is achieved through the following technical solutions:

[0007] In the first aspect, the present application provides a method for monitoring the displacement of a foundation pit support near a subway:

[0008] First, a coupled model of foundation pit, subway tunnel, and soil quality was established based on construction data, subway data, and environmental data. A genetic algorithm was used to search for the first monitoring point distribution, first monitoring frequency, and first offset warning value in the coupled model according to preset redundancy conditions.

[0009] Then, key point adjustment is performed on the first monitoring point distribution, the first monitoring frequency, and the first offset warning value to obtain a second monitoring point distribution, a second monitoring frequency, and a second offset warning value;

[0010] Next, the support offset monitoring data and the non-support offset monitoring data from each monitoring device are received and pre-processed;

[0011] Finally, when the support offset monitoring data reaches the second offset warning value, an alarm is issued according to the preset alarm settings. When the support offset monitoring data does not reach the second offset warning value, a data-driven prediction of the support offset is performed based on the coupling model, the accumulated support offset monitoring data and the non-support offset monitoring data, and the pit support offset prediction data is output.

[0012] In order to better implement the present invention, a method for constructing a coupled model of foundation pit-subway tunnel-soil is further established based on construction data, subway data and environmental data, including:

[0013] Integrate construction data, subway data, and environmental data to obtain model parameters for building a coupled model;

[0014] The soil is set as the Mohr-Coulomb model, the foundation pit support structure is set as the elastic beam unit, the subway tunnel is set as the shell unit, and the subway vibration load is discretized through Fourier transform. The model parameters are used to construct a coupled model of foundation pit-subway tunnel-soil in the form of a finite element model.

[0015] To better implement the present invention, further, using a genetic algorithm, according to a preset redundancy condition, a method for searching for the first monitoring point distribution, the first monitoring frequency, and the first offset warning value in the coupling model includes:

[0016] Set corresponding decision variable codes for monitoring point distribution, monitoring frequency and offset warning value, and set layout constraints for monitoring point distribution;

[0017] The chromosomes were constructed using the distribution of monitoring points, monitoring frequency and offset warning values as genes;

[0018] Using monitoring coverage and prediction accuracy as safety indicators, and monitoring point cost and monitoring frequency cost as cost indicators, we assign corresponding weights to the safety and cost indicators according to the preset redundancy conditions and construct a fitness calculation formula.

[0019] Generate the initial population, set the crossover operation, mutation operation and selection operation according to the fitness, and set the maximum number of iterations and the convergence threshold according to the fitness as the termination condition;

[0020] The initial population is optimized according to the crossover operation, the mutation operation, the selection operation and the termination condition to obtain the first monitoring point distribution, the first monitoring frequency and the first offset warning value.

[0021] In order to better implement the present invention, the foundation pit area in the coupling model is further discretized into grids, each grid corresponds to a binary bit, and a binary decision variable encoding of the monitoring point distribution is constructed, where 1 means a point is deployed and 0 means no point is deployed;

[0022] Match an integer code to each monitoring frequency and construct the integer decision variable code of the monitoring frequency. The larger the integer code number, the higher the corresponding monitoring frequency. The monitoring frequency corresponding to the integer code with the maximum value is real-time monitoring.

[0023] According to the initial offset warning value, the offset warning value range is set, and the floating-point decision variable encoding of the offset warning value is constructed.

[0024] In order to better implement the present invention, further, the following is also included at the end:

[0025] According to the support offset monitoring data and the pit support offset prediction data, the coupling model is dynamically adjusted, and the second monitoring point distribution, the second monitoring frequency and the second offset warning value are synchronously and dynamically adjusted by using a genetic algorithm and finite element inversion according to preset adjustment rules.

[0026] In order to better implement the present invention, further, the adjustment rules include:

[0027] Rules for adding and deleting monitoring points, frequency adjustment rules, and offset warning value adjustment rules;

[0028] The rules for adding and deleting monitoring points include adding monitoring points to areas where the prediction error is greater than the target value and no monitoring points cover the area, and deleting monitoring points to areas where the migration velocity is less than the target value and the inversion sensitivity ranking is lower than the target value;

[0029] The frequency adjustment rules include increasing the monitoring frequency for monitoring points whose deviation rate is greater than the target value, and decreasing the monitoring frequency for monitoring points whose deviation rate is continuously less than the target value for a specified time;

[0030] The offset warning value adjustment rule includes obtaining a third offset warning value according to the emergency time and the offset rate. If the third offset warning value is less than the second offset warning value, the third offset warning value is used as the current offset warning value.

[0031] To better implement the present invention, further, when the support offset monitoring data does not reach the second offset warning value, a method for performing a data-driven prediction of the support offset based on the coupling model, the accumulated support offset monitoring data, and the non-support offset monitoring data, and outputting the pit support offset prediction data includes:

[0032] Preprocessing the coupling model parameters, accumulated support offset monitoring data, and unsupport offset monitoring data;

[0033] Perform correlation analysis and clustering on the preprocessed data to obtain the prediction parameters corresponding to each support offset and extract the characteristic values of each prediction parameter to construct a prediction data set;

[0034] The offset prediction model is trained using the prediction data set until the prediction accuracy of the offset prediction model reaches a preset target value;

[0035] The parameters of the coupling model, the support offset monitoring data in the first time period and the non-support offset monitoring data in the first time period are input into the trained offset prediction model, and the foundation pit support offset prediction data in the second time period is output.

[0036] In order to better implement the present invention, further, the pretreatment includes:

[0037] Noise filtering, missing value processing, abnormal data monitoring and multi-source data fusion.

[0038] In order to better implement the present invention, further, the pretreatment includes:

[0039] According to the subway data, a band-stop filter with corresponding frequency is used to filter the subway vibration noise;

[0040] The semivariogram model is Gaussian kriging interpolation to handle spatial missing values, and cubic spline interpolation to handle temporal missing values.

[0041] Use the isolation forest algorithm to detect abnormal data based on data recorded over a period of time or engineering data with similar working conditions;

[0042] Multi-source data fusion is achieved through time synchronization, coordinate unification and offset data normalization.

[0043] In a second aspect, the present application provides a system for monitoring the displacement of a foundation pit support near a subway station, comprising:

[0044] A search module is configured to establish a coupled model of foundation pit-subway tunnel-soil properties based on construction data, subway data, and environmental data, and to search the coupled model for a first monitoring point distribution, a first monitoring frequency, and a first offset warning value using a genetic algorithm and in accordance with a preset redundancy condition;

[0045] An adjustment module, configured to perform key point adjustment on the first monitoring point distribution, the first monitoring frequency, and the first offset warning value to obtain a second monitoring point distribution, a second monitoring frequency, and a second offset warning value;

[0046] A preprocessing module, the preprocessing module is used to receive the support offset monitoring data and the non-support offset monitoring data from each monitoring device and perform preprocessing;

[0047] The prediction alarm module is used to issue an alarm according to the preset alarm setting when the support offset monitoring data reaches the second offset warning value. When the support offset monitoring data does not reach the second offset warning value, the support offset is predicted based on data-driven prediction based on the accumulated support offset monitoring data and non-support offset monitoring data, and the pit support offset prediction data is output.

[0048] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0049] By building a coupled model and combining genetic algorithm and finite element inversion, a better monitoring plan for the displacement of temporary subway foundation pit support can be automatically generated, which improves the efficiency of monitoring plan formulation and reduces the difficulty of generating monitoring plans.

[0050] Adjust key points of the generated monitoring plan to further ensure the security of the monitoring results;

[0051] By predicting the changes in the pit support offset through accumulated monitoring data, we can know the possible risks in advance, facilitate advance preparation, and deal with the support structure in advance to prevent accidents during construction. We can also combine the prediction results with the actual monitoring results to adjust the coupling model and monitoring plan to further ensure monitoring accuracy and reduce monitoring costs. It is suitable for monitoring temporary subway foundation pit projects with a long construction period. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The present invention is further described in conjunction with the following drawings and embodiments, and all concepts and innovations of the present invention should be regarded as disclosed contents and the protection scope of the present invention.

[0053] Figure 1 This is an embodiment of a method for monitoring the displacement of support for a subway foundation pit in the present application.

[0054] Figure 2 This is an embodiment of a system for monitoring the displacement of support for a subway foundation pit in the present application. DETAILED DESCRIPTION

[0055] Example 1:

[0056] See also Figure 1 , a method for monitoring the displacement of the support of the subway foundation pit:

[0057] First, a coupled model of foundation pit, subway tunnel, and soil quality was established based on construction data, subway data, and environmental data. A genetic algorithm was used to search for the first monitoring point distribution, first monitoring frequency, and first offset warning value in the coupled model according to preset redundancy conditions.

[0058] Then, key point adjustment is performed on the first monitoring point distribution, the first monitoring frequency, and the first offset warning value to obtain a second monitoring point distribution, a second monitoring frequency, and a second offset warning value;

[0059] Next, the support offset monitoring data and the non-support offset monitoring data from each monitoring device are received and pre-processed;

[0060] Finally, when the support offset monitoring data reaches the second offset warning value, an alarm is issued according to the preset alarm settings. When the support offset monitoring data does not reach the second offset warning value, a data-driven prediction of the support offset is performed based on the coupling model, the accumulated support offset monitoring data and the non-support offset monitoring data, and the pit support offset prediction data is output.

[0061] In this embodiment, the construction data may include excavation data, support structure data and dewatering phase plan, etc. The excavation data may include the excavation depth, area, start and end time, machinery configuration, earthwork transportation route, etc. of each layer; the support structure data may include support type, installation position (coordinates), prestressing value, retaining wall size, concrete strength, reinforcement ratio, support stiffness, connection node structure, etc. The dewatering phase plan may include well point layout diagram, daily dewatering rate, water level control target, etc.; subway data may include subway tunnel distribution, subway tunnel burial depth, track settlement threshold, subway operation timetable, subway vibration frequency, etc.; environmental data may include geological drilling data, geotechnical test parameters, geophysical exploration data, groundwater level monitoring data, permeability coefficient, vibration propagation attenuation model, etc.; PLAXIS or MIDAS GTS NX may be used to construct a three-dimensional finite element coupling model.

[0062] In this embodiment, the preset redundancy condition is set based on the safety considerations of the specific project. Under this condition, a search is performed to obtain a relatively optimal first monitoring point distribution, first monitoring frequency, and first offset warning value. The key point adjustment of the first monitoring point distribution, first monitoring frequency, and first offset warning value is mainly a manual secondary confirmation to ensure that monitoring points have been set in some areas that must be monitored.

[0063] Specifically, the preset redundant conditions can be to evenly distribute monitoring points at a certain interval on each side of the target foundation pit, reduce the spacing between monitoring points on the subway side and arrange them more densely, and forcibly arrange monitoring points at positive corners and support nodes.

[0064] In this embodiment, the support offset monitoring data is the project in the foundation pit monitoring project, in which the supporting structure and the supported structure can produce obvious displacement in the horizontal and vertical directions. For earth foundation pits, it includes the horizontal displacement of the top of the retaining wall (slope), the vertical displacement of the top of the retaining wall (slope), the deep horizontal displacement and the vertical displacement of the column. Other monitoring items are non-support offset monitoring data, which may include the internal force of the retaining wall, the axial force of the support, the internal force of the column, the axial force of the anchor rod, the uplift of the pit bottom, the lateral earth pressure of the retaining wall, the groundwater level, the vertical displacement of the soil layer and the vertical displacement of the surrounding surface. It may also include the vertical displacement, inclination and horizontal displacement of the surrounding buildings, cracks in the surrounding buildings and surface cracks, the vertical displacement and horizontal displacement of the surrounding pipelines, and the vertical displacement of the surrounding roads.

[0065] In this embodiment, data-driven prediction refers to predicting the changes in the pit support offset within a certain period of time in the future based on the data accumulated and obtained by the monitoring equipment within a certain period of time. When the accumulated data is insufficient, the data itself can be enhanced. For example, a DTW-based time series clustering algorithm is first used to process the data to obtain multiple clusters, and then Gaussian perturbation is performed on each cluster center to generate new samples. It is also possible to obtain engineering data corresponding to projects with high similarity in the database.

[0066] Example 2

[0067] This embodiment further optimizes the above-mentioned embodiment 1 and provides a preferred implementation of a method for monitoring the offset of support for a foundation pit adjacent to a subway. In this embodiment, a method for constructing a coupled model of foundation pit-subway tunnel-soil is established based on construction data, subway data, and environmental data, including:

[0068] Integrate construction data, subway data, and environmental data to obtain model parameters for building a coupled model;

[0069] The soil is set as the Mohr-Coulomb model, the foundation pit support structure is set as the elastic beam unit, the subway tunnel is set as the shell unit, and the subway vibration load is discretized through Fourier transform. The model parameters are used to construct a coupled model of foundation pit-subway tunnel-soil in the form of a finite element model.

[0070] By adopting this embodiment, setting the soil as the Mohr-Coulomb model can accurately simulate the nonlinear behavior of the soil, setting the foundation pit support structure as the elastic beam unit can simulate the bending deformation, axial compression and node stress state of the support structure with linear elastic deformation, and accurately capturing the key mechanical behaviors while simplifying the modeling. Setting the subway tunnel as the shell unit can improve the adaptability of the model to complex loads, and discretizing the subway vibration load through Fourier transform can decompose the time domain vibration load into frequency domain components, which is convenient for identifying the structural resonance frequency, so that the foundation pit-subway tunnel-soil coupling model can more accurately simulate the actual situation of the target foundation pit, the surrounding environment and the subway tunnel.

[0071] Optionally, the method of searching for the first monitoring point distribution, the first monitoring frequency, and the first offset warning value in the coupling model using a genetic algorithm according to a preset redundancy condition includes:

[0072] Set corresponding decision variable codes for monitoring point distribution, monitoring frequency and offset warning value, and set layout constraints for monitoring point distribution;

[0073] The chromosomes were constructed using the distribution of monitoring points, monitoring frequency and offset warning values as genes;

[0074] Using monitoring coverage and prediction accuracy as safety indicators, and monitoring point cost and monitoring frequency cost as cost indicators, we assign corresponding weights to the safety and cost indicators according to the preset redundancy conditions and construct a fitness calculation formula.

[0075] Generate the initial population, set the crossover operation, mutation operation and selection operation according to the fitness, and set the maximum number of iterations and the convergence threshold according to the fitness as the termination condition;

[0076] The initial population is optimized according to the crossover operation, the mutation operation, the selection operation and the termination condition to obtain the first monitoring point distribution, the first monitoring frequency and the first offset warning value.

[0077] Further optionally, the foundation pit area in the coupling model is discretized into grids, each grid corresponds to a binary bit, and a binary decision variable encoding of the monitoring point distribution is constructed, where 1 means a point is deployed and 0 means no point is deployed;

[0078] Match an integer code to each monitoring frequency and construct the integer decision variable code of the monitoring frequency. The larger the integer code number, the higher the corresponding monitoring frequency. The monitoring frequency corresponding to the integer code with the maximum value is real-time monitoring.

[0079] According to the initial offset warning value, the offset warning value range is set, and the floating-point decision variable encoding of the offset warning value is constructed.

[0080] Specifically, the layout constraints can be areas where monitoring points are mandatory, such as the subway side, and areas where monitoring points cannot be deleted, such as positive corners and support nodes;

[0081] For example, in a specific embodiment, the foundation pit area is discretized into 1m×1m grids, each grid corresponds to a binary bit (1=point distribution, 0=no point distribution), the monitoring frequency is 1-4, respectively representing: 1=once every 4 hours, 2=once every 2 hours, 3=once every 1 hour, 4=real-time monitoring, the warning value range is 50%-100% of the standard value, the step size is 1mm, and the chromosome can be expressed as: chromosome = [1,0,1,...,1|3,2,4,1|15.0,18.0,20.0,15.0], the total number of 1s in the first paragraph represents The table shows the number of monitoring points. The middle section shows the frequency of the excavation stage = 3 (once per hour), the support installation stage = 2 (once every 2 hours), the precipitation stage = 4 (real time), and the trenchless stage = 1 (once every 4 hours). The latter section shows the warning value of the horizontal displacement of the top of the retaining wall (slope) = 15 mm (50% of the design value of 30 mm), the vertical displacement of the top of the retaining wall (slope) = 18 mm (60% of the design value of 30 mm), the deep horizontal displacement warning value = 20 (50% of the design value of 40 mm), and the column vertical displacement warning value = 15 (50% of the design value of 30 mm).

[0082] Assign a weight of 70% to the safety index and a weight of 30% to the cost index. Furthermore, the weight of monitoring coverage is 40%, the weight of prediction accuracy is 30%, the weight of monitoring point cost is 15%, and the weight of monitoring frequency cost is 15%. The expression of monitoring coverage is: The expression of prediction accuracy is The expression of monitoring point cost is: The expression of monitoring frequency cost is Where n is the number of high-risk areas covered, N is the total number of high-risk areas, including the subway side area and the sunny corner area, RMSE is the root mean square error between the coupling model and the data of all monitoring points, m is the number of retained monitoring points, M is the number of initial monitoring points, p is the monitoring frequency level, ω is the time proportion corresponding to the monitoring frequency level, and the fitness expression is Fitness = 0.4 Coverage + 0.3 Precision + 0.15 (100% Cost_sensor) + 0.15 (100% Cost_freq);

[0083] Specifically, the generation rules for the initial population include a subway side monitoring point retention rate ≥ 80%, a random non-subway side monitoring point retention rate, and a random allocation of the initial value of the monitoring frequency. The selection operation can be a roulette wheel selection, the crossover operation is a single-point crossover, and the mutation operations include random flipping of monitoring points, random monitoring frequency ±1 level, and random warning value ±1 mm. The termination condition is a maximum number of iterations of 100 generations or a fitness improvement of less than 1% for 10 consecutive generations.

[0084] By adopting this embodiment, a better automatic monitoring scheme for the displacement of the foundation pit support of the subway can be automatically generated more accurately.

[0085] Example 3:

[0086] This embodiment is further optimized based on the above embodiment 2. In this embodiment, the following is further included at the end:

[0087] According to the support offset monitoring data and the pit support offset prediction data, the coupling model is dynamically adjusted, and the second monitoring point distribution, the second monitoring frequency and the second offset warning value are synchronously and dynamically adjusted by using a genetic algorithm and finite element inversion according to preset adjustment rules.

[0088] Optionally, the adjustment rules include:

[0089] Rules for adding and deleting monitoring points, frequency adjustment rules, and offset warning value adjustment rules;

[0090] The rules for adding and deleting monitoring points include adding monitoring points to areas where the prediction error is greater than the target value and no monitoring points cover the area, and deleting monitoring points to areas where the migration velocity is less than the target value and the inversion sensitivity ranking is lower than the target value;

[0091] The frequency adjustment rules include increasing the monitoring frequency for monitoring points whose deviation rate is greater than the target value, and decreasing the monitoring frequency for monitoring points whose deviation rate is continuously less than the target value for a specified time;

[0092] The offset warning value adjustment rule includes obtaining a third offset warning value according to the emergency time and the offset rate. If the third offset warning value is less than the second offset warning value, the third offset warning value is used as the current offset warning value.

[0093] Specifically, the inversion sensitivity is obtained by calculating the impact of each monitoring point on the overall RMSE through finite element inversion. The pre-processed displacement data can be input and the soil parameters, such as the elastic modulus of the soil and the friction coefficient of the support structure-soil interface, can be inverted using the Levenberg-Marquardt algorithm. The corrected displacement field is then output to identify high-risk areas that have not been monitored. If the inversion shows that the prediction error in a certain area is greater than 1mm, 1 to 2 monitoring points are added there.

[0094] The target values in the adjustment rules can be set according to actual conditions. For example, the offset speed less than the target value can be a offset rate of <0.1 mm / h for 5 consecutive days. The third offset warning value is the ratio of the emergency time to the offset rate.

[0095] In this embodiment, the monitoring plan is further automatically modified by using the acquired actual monitoring data and performing finite element inversion using the genetic algorithm again.

[0096] Example 4:

[0097] This embodiment is further optimized based on the above-mentioned embodiment 1, 2 or 3. In this embodiment, when the support offset monitoring data does not reach the second offset warning value, a method for performing a data-driven prediction of the support offset based on the coupling model, the accumulated support offset monitoring data and the non-support offset monitoring data, and outputting the pit support offset prediction data includes:

[0098] Preprocessing the coupling model parameters, accumulated support offset monitoring data, and unsupport offset monitoring data;

[0099] Perform correlation analysis and clustering on the preprocessed data to obtain the prediction parameters corresponding to each support offset and extract the characteristic values of each prediction parameter to construct a prediction data set;

[0100] The offset prediction model is trained using the prediction data set until the prediction accuracy of the offset prediction model reaches a preset target value;

[0101] The parameters of the coupling model, the support offset monitoring data in the first time period and the non-support offset monitoring data in the first time period are input into the trained offset prediction model, and the foundation pit support offset prediction data in the second time period is output.

[0102] Optionally, preprocessing includes: noise filtering, missing value processing, abnormal data monitoring and multi-source data fusion.

[0103] Further optional preprocessing includes:

[0104] According to the subway data, a band-stop filter with corresponding frequency is used to filter the subway vibration noise;

[0105] The semivariogram model is Gaussian kriging interpolation to handle spatial missing values, and cubic spline interpolation to handle temporal missing values.

[0106] Use the isolation forest algorithm to detect abnormal data based on data recorded over a period of time or engineering data with similar working conditions;

[0107] Multi-source data fusion is achieved through time synchronization, coordinate unification and offset data normalization.

[0108] Specifically, correlation analysis can include parameter correlation calculation and time lag effect analysis. For example, the Pearson correlation coefficient between the offset and the coupled model parameters / unsupported offset monitoring data is calculated, and the lag time between the parameters is determined through the cross-correlation function (CCF). Target feature screening can also be performed on each support offset to obtain the strongly correlated parameters of each support offset. Through K-means / DBSCAN clustering, the monitoring points are divided into high-risk, medium-risk, and low-risk groups according to the displacement rate and internal force fluctuation characteristics, and features are extracted separately for different risk groups.

[0109] The first time period is a certain period in the past, and the second time period is a certain period in the future. For example, using the data of the past 24 hours, the offset in the next 6 hours is predicted;

[0110] In terms of model selection, we can also perform sub-target modeling according to different risk groups and train models for each offset separately. Since the construction period of subway foundation pit projects is long, we can choose prediction models such as LSTM models or Transformer models that are suitable for long-term time series dependencies.

[0111] Furthermore, after obtaining new data, incremental learning is performed to retrain the model with a sliding window (such as the latest 30 days of data) to retain long-term patterns, and parameter correlations are recalculated regularly to eliminate invalid features and continuously optimize the prediction model.

[0112] In this embodiment, various parameters are first preprocessed to ensure prediction accuracy. Then, through correlation analysis and clustering, the influence of each parameter on the offset is analyzed, and appropriate parameters are selected to predict different offsets, thereby further improving prediction accuracy and efficiency.

[0113] Example 5:

[0114] See also Figure 2 , an embodiment of a system for monitoring the displacement of a subway foundation pit support, comprising:

[0115] A search module is configured to establish a coupled model of foundation pit-subway tunnel-soil properties based on construction data, subway data, and environmental data, and to search the coupled model for a first monitoring point distribution, a first monitoring frequency, and a first offset warning value using a genetic algorithm and in accordance with a preset redundancy condition;

[0116] An adjustment module, configured to perform key point adjustment on the first monitoring point distribution, the first monitoring frequency, and the first offset warning value to obtain a second monitoring point distribution, a second monitoring frequency, and a second offset warning value;

[0117] A preprocessing module, the preprocessing module is used to receive the support offset monitoring data and the non-support offset monitoring data from each monitoring device and perform preprocessing;

[0118] The prediction alarm module is used to issue an alarm according to the preset alarm setting when the support offset monitoring data reaches the second offset warning value. When the support offset monitoring data does not reach the second offset warning value, the support offset is predicted based on data-driven prediction based on the accumulated support offset monitoring data and non-support offset monitoring data, and the pit support offset prediction data is output.

[0119] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any simple modification or equivalent change made to the above embodiment based on the technical essence of the present invention falls within the scope of protection of the present invention.

Claims

1. A method for monitoring the displacement of foundation pit support for a subway station, characterized by: First, a coupled model of foundation pit, subway tunnel, and soil quality was established based on construction data, subway data, and environmental data. A genetic algorithm was used to search for the first monitoring point distribution, first monitoring frequency, and first offset warning value in the coupled model according to preset redundancy conditions. Then, key point adjustment is performed on the first monitoring point distribution, the first monitoring frequency, and the first offset warning value to obtain a second monitoring point distribution, a second monitoring frequency, and a second offset warning value; Next, the support offset monitoring data and the non-support offset monitoring data from each monitoring device are received and pre-processed; Finally, when the support offset monitoring data reaches the second offset warning value, an alarm is issued according to the preset alarm settings. When the support offset monitoring data does not reach the second offset warning value, a data-driven prediction of the support offset is performed based on the coupling model, the accumulated support offset monitoring data and the non-support offset monitoring data, and the pit support offset prediction data is output.

2. The method for monitoring the displacement of the foundation pit support of a subway according to claim 1 is characterized in that: Based on construction data, subway data, and environmental data, a method for constructing a coupled model of foundation pit-subway tunnel-soil is established, including: Integrate construction data, subway data, and environmental data to obtain model parameters for building a coupled model; The soil is set as the Mohr-Coulomb model, the foundation pit support structure is set as the elastic beam unit, the subway tunnel is set as the shell unit, and the subway vibration load is discretized through Fourier transform. The model parameters are used to construct a coupled model of foundation pit-subway tunnel-soil in the form of a finite element model.

3. The method for monitoring the displacement of foundation pit support of a subway according to claim 1, characterized in that: The method of searching for the first monitoring point distribution, the first monitoring frequency, and the first offset warning value in the coupling model using a genetic algorithm according to a preset redundancy condition includes: Set corresponding decision variable codes for monitoring point distribution, monitoring frequency and offset warning value, and set layout constraints for monitoring point distribution; The chromosomes were constructed using the distribution of monitoring points, monitoring frequency and offset warning values as genes; Using monitoring coverage and prediction accuracy as safety indicators, and monitoring point cost and monitoring frequency cost as cost indicators, we assign corresponding weights to the safety and cost indicators according to the preset redundancy conditions and construct a fitness calculation formula. Generate the initial population, set the crossover operation, mutation operation and selection operation according to the fitness, and set the maximum number of iterations and the convergence threshold according to the fitness as the termination condition; The initial population is optimized according to the crossover operation, the mutation operation, the selection operation and the termination condition to obtain the first monitoring point distribution, the first monitoring frequency and the first offset warning value.

4. The method for monitoring the displacement of foundation pit support of a subway according to claim 3, characterized in that: The foundation pit area in the coupling model is discretized into grids, each grid corresponds to a binary bit, and the binary decision variable encoding of the monitoring point distribution is constructed, where 1 means a point is deployed and 0 means no point is deployed; Match an integer code to each monitoring frequency and construct the integer decision variable code of the monitoring frequency. The larger the integer code number, the higher the corresponding monitoring frequency. The integer code with the maximum value corresponds to the monitoring frequency of real-time monitoring. According to the initial offset warning value, the offset warning value range is set, and the floating-point decision variable encoding of the offset warning value is constructed.

5. The method for monitoring the displacement of foundation pit support of a subway according to claim 1, characterized in that: Also included at the end: According to the support offset monitoring data and the pit support offset prediction data, the coupling model is dynamically adjusted, and the second monitoring point distribution, the second monitoring frequency and the second offset warning value are synchronously and dynamically adjusted by using a genetic algorithm and finite element inversion according to preset adjustment rules.

6. The method for monitoring the displacement of foundation pit support of a subway according to claim 5, characterized in that: Adjustment rules include: Rules for adding and deleting monitoring points, frequency adjustment rules, and offset warning value adjustment rules; The rules for adding and deleting monitoring points include adding monitoring points to areas where the prediction error is greater than the target value and no monitoring points cover the area, and deleting monitoring points to areas where the migration velocity is less than the target value and the inversion sensitivity ranking is lower than the target value; The frequency adjustment rules include increasing the monitoring frequency for monitoring points whose deviation rate is greater than the target value, and decreasing the monitoring frequency for monitoring points whose deviation rate is continuously less than the target value for a specified time; The offset warning value adjustment rule includes obtaining a third offset warning value according to the emergency time and the offset rate. If the third offset warning value is less than the second offset warning value, the third offset warning value is used as the current offset warning value.

7. The method for monitoring the displacement of foundation pit support of a subway according to claim 1, characterized in that: When the support offset monitoring data does not reach the second offset warning value, a method for performing a data-driven prediction of the support offset based on the coupling model, the accumulated support offset monitoring data, and the non-support offset monitoring data, and outputting the pit support offset prediction data includes: Preprocessing the coupling model parameters, accumulated support offset monitoring data, and unsupport offset monitoring data; Perform correlation analysis and clustering on the preprocessed data to obtain the prediction parameters corresponding to each support offset and extract the characteristic values of each prediction parameter to construct a prediction data set; The offset prediction model is trained using the prediction data set until the prediction accuracy of the offset prediction model reaches a preset target value; The parameters of the coupling model, the support offset monitoring data in the first time period and the non-support offset monitoring data in the first time period are input into the trained offset prediction model, and the foundation pit support offset prediction data in the second time period is output.

8. A method for monitoring the displacement of foundation pit support for a subway according to claim 1 or 7, characterized in that: Preprocessing includes: Noise filtering, missing value processing, abnormal data monitoring and multi-source data fusion.

9. The method for monitoring the displacement of foundation pit support of a subway according to claim 8 is characterized in that the pre-processing include: According to the subway data, a band-stop filter with corresponding frequency is used to filter the subway vibration noise; The semivariogram model is Gaussian kriging interpolation to handle spatial missing values, and cubic spline interpolation to handle temporal missing values. Use the isolation forest algorithm to detect abnormal data based on data recorded over a period of time or engineering data with similar working conditions; Multi-source data fusion is achieved through time synchronization, coordinate unification and offset data normalization.

10. A system for monitoring the displacement of foundation pit support for a temporary subway, characterized in that: include: A search module is configured to establish a coupled model of foundation pit-subway tunnel-soil properties based on construction data, subway data, and environmental data, and to search the coupled model for a first monitoring point distribution, a first monitoring frequency, and a first offset warning value using a genetic algorithm and in accordance with a preset redundancy condition; An adjustment module, configured to perform key point adjustment on the first monitoring point distribution, the first monitoring frequency, and the first offset warning value to obtain a second monitoring point distribution, a second monitoring frequency, and a second offset warning value; A preprocessing module, the preprocessing module is used to receive the support offset monitoring data and the non-support offset monitoring data from each monitoring device and perform preprocessing; The prediction alarm module is used to issue an alarm according to the preset alarm setting when the support offset monitoring data reaches the second offset warning value. When the support offset monitoring data does not reach the second offset warning value, the support offset is predicted based on data-driven prediction based on the accumulated support offset monitoring data and non-support offset monitoring data, and the pit support offset prediction data is output.

Images