A method and system for adaptive adjustment of foundation pit support strength based on real-time monitoring data

Abstract

This invention discloses an adaptive adjustment method and system for foundation pit support strength based on real-time monitoring data, relating to the field of engineering control technology. The method includes the following steps: real-time acquisition of monitoring data of the foundation pit support structure; preprocessing the monitoring data to obtain standardized time series data; inputting the standardized time series data into a pre-trained LSTM prediction model to output a displacement prediction value sequence; calculating the predicted state index based on the displacement prediction value sequence; constructing a multi-objective optimization model with the objectives of minimizing the maximum predicted state index and minimizing the total cost of support strength adjustment, and solving for the optimal prestress adjustment amount of each intelligent support actuator; converting the optimal prestress adjustment amount into control commands to drive the corresponding intelligent support actuator to perform actions. This invention can automatically and accurately adjust the foundation pit support strength based on real-time monitoring data.

Description

Technical Field

[0001] This invention relates to the field of foundation pit data processing technology, specifically to a method and system for adaptive adjustment of foundation pit support strength based on real-time monitoring data. Background Technology

[0002] During the construction of the foundation pit, the complexity of geological conditions, changes in ambient temperature, and on-site construction errors often cause the axial force of the support or the horizontal displacement of the retaining structure within the foundation pit to reach or even exceed the warning values ​​in the specifications, resulting in significant construction risks.

[0003] Traditional foundation pit support design primarily relies on static geological survey reports and theoretical calculations to determine the initial parameters of the support structure. During construction, although monitoring points are set up for manual or semi-automated data collection, when the monitoring data exceeds warning values, engineers typically decide whether and how to adjust the support parameters based on their personal experience. This method has significant drawbacks: the cycle from risk identification to manual decision-making and on-site adjustment is lengthy; it relies on personal experience and cannot perform forward-looking risk prediction; and it often adopts conservative over-support measures to avoid risk.

[0004] Therefore, there is an urgent need for an intelligent method and system that can automatically, accurately, and proactively adjust the dynamic adaptive strength of foundation pit support based on real-time monitoring data, so as to optimize support costs while ensuring safety.

[0005] For example, Chinese patent CN120990664A discloses an intelligent support method and system for hydraulic supports, as well as electronic equipment. The method includes: real-time acquisition of roof pressure, support posture, and geological structure parameters through a multi-modal sensor array; spatiotemporal feature fusion using edge computing nodes to generate a three-dimensional support status map; construction of a dynamic weight adjustment strategy based on reinforcement learning algorithms to generate multi-objective optimized support parameters; adaptive adjustment of support strength by issuing control commands through a dual-mode communication network; and establishment of a closed-loop feedback verification mechanism by combining lidar scanning of roof subsidence. It integrates a digital twin model and energy consumption optimization functions. This invention solves the problems of single data dimension, rigid decision-making, and insufficient reliability in traditional methods, achieving dynamic optimization of support parameters, proactive risk prevention and control, and synergistic improvement of energy efficiency, significantly improving the safety and intelligence level of support under complex working conditions. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to address the shortcomings of the prior art by providing a method and system for adaptive adjustment of foundation pit support strength based on real-time monitoring data.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] The method for adaptive adjustment of foundation pit support strength based on real-time monitoring data includes the following steps:

[0009] Step S1: Real-time collection of monitoring data on the foundation pit support structure and surrounding environment;

[0010] Step S2: Preprocess the monitoring data to obtain standardized time series data;

[0011] Step S3: Input the standardized time series data into the pre-trained LSTM prediction model and output the displacement prediction value sequence of the foundation pit monitoring points within a preset time period in the future.

[0012] Step S4: Based on the displacement prediction value sequence, calculate the predicted state index reflecting the stability of the foundation pit at each future time.

[0013] Step S5: With the objectives of minimizing the maximum predicted state index and minimizing the total cost of support strength adjustment, construct a multi-objective optimization model and solve for the optimal prestress adjustment amount of each intelligent support actuator.

[0014] Step S6: Convert the optimal prestress adjustment amount into a control command to drive the corresponding intelligent support actuator to perform the action and complete the adaptive adjustment of the support strength.

[0015] Furthermore, in step S1, the monitoring data includes at least the deep and top horizontal displacement of the support structure, the stress of the support components, the surface settlement around the foundation pit, and the pore water pressure.

[0016] Furthermore, step S2 specifically includes the following steps:

[0017] Step S2.1: Resample and align sensor data with different sampling frequencies and transmission delays to form data frames with consistent timestamps;

[0018] Step S2.2: For missing data points, construct a spatiotemporal kernel function based on the spatial topological relationship of the monitoring points and the temporal correlation of the data, and perform adaptive interpolation by weighted fusion of the observations of nearby sensors within the effective spatiotemporal window.

[0019] Step S2.3: Based on the statistical distribution characteristics and the principle of mechanical equilibrium, perform a joint test on the data to identify and correct real outliers.

[0020] Furthermore, step S2.3 specifically includes the following steps:

[0021] A sliding window-based statistical method was used for preliminary outlier screening.

[0022] Based on the principle of mechanical equilibrium of the foundation pit, the anomalies initially screened out were verified a second time to distinguish between real abnormal data and reasonable data mutations.

[0023] For data that is confirmed as a genuine anomaly through secondary verification, spatiotemporal interpolation values ​​are used for correction.

[0024] Furthermore, in step S3, the LSTM prediction model uses a risk-weighted average absolute error function as the training loss function; the risk-weighted average absolute error function is configured to dynamically adjust its weight in the loss function calculation based on the proximity of the predicted or actual displacement values ​​in the training samples to a preset engineering warning threshold.

[0025] Furthermore, step S4 specifically includes the following steps:

[0026] Step S4.1: Calculate the rate of change of the predicted displacement value corresponding to the predicted displacement value sequence;

[0027] Step S4.2: Calculate the displacement risk factor and the rate of change risk factor based on the predicted displacement value and the rate of change of the predicted displacement value;

[0028] Step S4.3: Using a weighted fusion method, the displacement risk factor and the rate of change risk factor are fused to generate a comprehensive predicted state index.

[0029] Furthermore, in step S5, the constraints of the multi-objective optimization model include equipment safety constraints, motion smoothness constraints, and force coordination constraints.

[0030] The equipment safety constraint is that the prestress value of each support unit after adjustment must be within the allowable working range of the mechanical design.

[0031] The motion smoothness constraint is that the absolute value of the adjustment of each unit in a single control cycle must not exceed the product of its maximum allowable adjustment rate and the cycle duration.

[0032] The force coordination constraint limits the total deviation between the adjustment amount of multiple support points located in the same force region or structural unit and the average adjustment amount of that region.

[0033] The multi-objective optimization model is solved using the NSGA-II algorithm.

[0034] An adaptive adjustment system for foundation pit support strength based on real-time monitoring data is used to implement any one of the aforementioned adaptive adjustment methods for foundation pit support strength based on real-time monitoring data, including:

[0035] The monitoring and sensing module consists of multiple types of sensor nodes and data acquisition units deployed at the foundation pit site, used to collect monitoring data in real time.

[0036] The data preprocessing and communication module is used to receive monitoring data, complete local preprocessing, and transmit it to the computing unit via wired or wireless network;

[0037] An edge computing platform is used to deploy and run the LSTM prediction model and the multi-objective optimization decision algorithm.

[0038] The intelligent support execution module includes several independent intelligent support units with force feedback and remote control functions;

[0039] The human-computer interaction and early warning module is used to provide a visual monitoring interface, early warning information dissemination, and manual parameter configuration.

[0040] Furthermore, the intelligent support unit in the intelligent support execution module is an intelligent prestressed anchor cable system, which specifically includes: anchor cable body, bearing plate and intelligent tensioning and locking device.

[0041] Furthermore, the intelligent tensioning and locking device integrates a servo hydraulic drive mechanism, a force sensor, a displacement sensor, and an embedded controller; wherein, the embedded controller is configured to receive remote adjustment commands and drive the servo hydraulic drive mechanism to achieve dynamic adjustment and locking of the anchor cable prestress.

[0042] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0043] 1. This invention, through a closed-loop control of monitoring-prediction-optimization-execution, can automatically make decisions and perform adjustments based on real-time data, thereby changing the problems of relying on human experience and delayed response and improving the risk prevention and control capabilities of foundation pit engineering.

[0044] 2. This invention uses an LSTM model to predict future displacement trends and calculates a predicted state index, enabling early warning and proactive adjustment before risks actually occur, thus transforming it into proactive prevention and control and improving engineering safety.

[0045] 3. This invention constructs a multi-objective optimization model with the objectives of minimizing the maximum predicted state index and minimizing the adjustment cost, and uses NSGA-II to solve it. The output adjustment scheme is the optimal trade-off between safety and economy under the current conditions, avoiding the waste caused by over-adjustment or the risk caused by under-adjustment.

[0046] 4. By introducing force coordination constraints, this invention forces the support units within the same force area to coordinate their actions, preventing uneven redistribution of forces within the support system due to excessive local adjustments, and ensuring the overall collaborative performance of the support structure. Attached Figure Description

[0047] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0048] Figure 1 This is a flowchart illustrating an embodiment of the present invention;

[0049] Figure 2 This is a schematic diagram of the system architecture according to an embodiment of the present invention;

[0050] Figure 3 This is a flowchart of LSTM prediction and risk assessment according to an embodiment of the present invention;

[0051] Figure 4 This is a diagram of an intelligent prestressed anchor cable system according to an embodiment of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0053] like Figure 1 As shown, the adaptive adjustment method for foundation pit support strength based on real-time monitoring data includes the following steps:

[0054] Step S1: Real-time collection of monitoring data on the foundation pit support structure and surrounding environment;

[0055] Step S2: Preprocess the monitoring data to obtain standardized time series data;

[0056] Step S3: Input the standardized time series data into the pre-trained LSTM prediction model and output the displacement prediction value sequence of the foundation pit monitoring points within a preset time period in the future.

[0057] Step S4: Based on the displacement prediction value sequence, calculate the predicted state index reflecting the stability of the foundation pit at each future time.

[0058] Step S5: With the objectives of minimizing the maximum predicted state index and minimizing the total cost of support strength adjustment, construct a multi-objective optimization model and solve for the optimal prestress adjustment amount of each intelligent support actuator.

[0059] Step S6: Convert the optimal prestress adjustment amount into a control command to drive the corresponding intelligent support actuator to perform the action and complete the adaptive adjustment of the support strength.

[0060] In step S1, the monitoring data includes at least the deep and top horizontal displacement of the support structure, the stress of the support components, the surface settlement around the foundation pit, and the pore water pressure.

[0061] Step S2 specifically includes the following steps:

[0062] Step S2.1: Resample and align sensor data with different sampling frequencies and transmission delays to form data frames with consistent timestamps;

[0063] Step S2.2: For missing data points, based on the spatial topological relationship of the monitoring points and the temporal correlation of the data, a spatiotemporal kernel function is constructed. The observations from neighboring sensors within the effective spatiotemporal window are then weighted and fused for adaptive interpolation. The specific formula is as follows:

[0064]

[0065] in, This represents the interpolated estimate of sensor i at time t. Describes the set of spatially adjacent sensors of sensor i, satisfying... Less than or equal to 30m This represents the set of effective observation time points of sensor j within the time window [t−ΔT, t+ΔT]. Indicates that sensor j is in time The original observations, This represents the spatial Euclidean distance between sensors i and j. This represents the absolute value of the time difference, i.e. The absolute value, Represents the spacetime kernel function;

[0066] The formula for calculating the spatiotemporal kernel function is as follows:

[0067]

[0068] in, and The kernel function scaling parameter controls the decay rate of spatial and temporal correlations, and is typically set to a value that represents the kernel function scaling parameter. =10m, =1800s;

[0069] =10m is determined based on the conventional layout spacing of monitoring points in the foundation pit. According to engineering practice, the horizontal spacing of monitoring points around the foundation pit is usually between 10m and 20m. Taking 10m can make the spatial correlation decrease to about 0.61 between adjacent monitoring points and decrease to 0.14 at an interval of 20m, which is in line with the physical law that the spatial correlation weakens with the increase of distance. =1800s is determined based on the time evolution characteristics of foundation pit deformation. Foundation pit deformation typically exhibits gradual changes on an hourly scale. Using 1800s ensures that the time correlation decays to 0.61 within 30 minutes and to 0.14 within 60 minutes, thus ensuring smooth handling of short-term fluctuations while retaining effective trend information. The above parameters can be appropriately adjusted according to the actual monitoring frequency and site geological conditions. The adjustment range is typically [not specified in the original text]. ∈[5m,20m], ∈[900s,3600s];

[0070] Step S2.3: Based on the statistical distribution characteristics and the principle of mechanical equilibrium, perform a joint test on the data to identify and correct real outliers.

[0071] Step S2.3 specifically includes the following steps:

[0072] A sliding window-based statistical method was used for preliminary outlier screening. The standardized Z-score of the data within the sliding window was calculated. If the standardized Z-score of the data within the sliding window was greater than 3, the data was marked as a suspected outlier.

[0073] The Z-score threshold is set to 3, which is determined based on the statistical characteristics of the normal distribution. According to the normal distribution theory, the probability that the data falls within the range of mean ± 3 standard deviations is 99.73%. Therefore, setting the threshold to 3 means that the possibility of the data exceeding the normal fluctuation range is extremely low, which can be used as a statistical basis for preliminary screening of outliers.

[0074] Based on the principle of mechanical equilibrium in foundation pits, the initially screened anomalies are subjected to secondary verification to distinguish between real anomalies and reasonable data abrupt changes. Specifically, this includes calculating the residuals between the displacement of three consecutive points on the support pile and the theoretical value of beam bending. The specific formula is as follows:

[0075]

[0076] in, Indicates mechanical compatibility deviation. , and These are the horizontal displacement observations at three consecutive monitoring points a, a-1, and a+1 at the same time. Indicates the horizontal distance between adjacent monitoring points. This represents the estimated bending moment borne by the support structure section at monitoring point a. This represents the elastic modulus of the supporting structure material. It represents the moment of inertia of the cross section of the supporting structure about the neutral axis;

[0077] The calculated mechanical compatibility deviation is compared with the preset physical residual threshold. If the mechanical compatibility deviation is greater than the preset physical residual threshold, the point is determined to be real abnormal data; otherwise, it is considered to be a reasonable data mutation within the scope of mechanical principles and is retained.

[0078] The preset physical residual threshold is typically set to 1.0mm to 3.0mm. This range is determined based on the combined result of the displacement sensor measurement accuracy and the theoretical error of the mechanical model. Taking a typical high-precision inclinometer as an example, the second-order difference combination error of three consecutive point displacements is approximately 0.3mm. The theoretical simplification error of the elastic foundation beam model is typically controlled within the range of 0.5mm to 1.0mm. By summing the squares of the errors and taking the square root, the threshold range can be calculated to be approximately 0.58mm to 1.04mm. Multiplying the threshold range by a safety factor of 2 or 3 and rounding it down, we obtain the engineering threshold of 1.0mm to 3.0mm. For scenarios using higher precision sensors or more refined mechanical models, the lower limit can be used; for projects with complex geological conditions and large model errors, the upper limit can be used.

[0079] For data confirmed as genuine anomalies through secondary verification, spatiotemporal interpolation values ​​are used for correction; that is, for genuine anomaly data, the aforementioned interpolation values ​​are applied. Replace it.

[0080] In step S3, the LSTM prediction model uses the risk-weighted average absolute error function as the training loss function. The risk-weighted average absolute error function is configured to dynamically adjust its weight in the loss function calculation based on the proximity of the predicted or actual displacement values ​​in the training samples to a preset engineering warning threshold.

[0081] The specific formula for the risk-weighted average absolute error function is as follows:

[0082]

[0083] in, This represents the risk-weighted average absolute error function. Indicates the batch size of the training samples. Indicates the index of the sample within the batch. This represents the predicted displacement value of the nth sample at time t+m. This represents the true displacement value of the nth sample at time t+m in the future. This represents the total number of time steps in the model's future predictions. Indices representing future time steps. This represents the risk perception weighting function;

[0084] The formula for calculating the risk perception weight function is as follows:

[0085]

[0086] in, This represents the risk penalty intensity coefficient, set to 4.0, meaning that when the displacement approaches the threshold, the weight of the prediction error in the loss function can increase by up to 5 times. This represents the threshold transition steepness coefficient, set to 0.1, which controls the rate of smooth change of the weights near the threshold. This represents the activation function. This indicates the first-level alarm threshold for engineering displacement, which can be set to 0.3% of the pit depth.

[0087] The risk penalty intensity coefficient is set to 4.0. This value determines that when the displacement approaches the warning threshold, the weight of the prediction error in the loss function can be increased to up to 5 times the original weight, i.e., 1+α. Taking α=4.0 can make high-risk samples receive a significantly higher penalty intensity than normal samples, effectively guiding the model to focus on dangerous working conditions, while avoiding excessive weight leading to training instability.

[0088] The threshold transition steepness coefficient is set to 0.1 to control the transition rate of the risk weight from 1 to 1+α smoothly. Taking β=0.1 means that when the displacement gradually approaches the threshold from 0.5 times the threshold, the weight slowly increases from about 1.01 to about 1.5; when the displacement exceeds the threshold, the weight rapidly increases to close to the maximum value.

[0089] The LSTM prediction model takes a multidimensional feature sequence of length A as input, such as A=144, corresponding to the past 24 hours. The feature vector of the multidimensional feature sequence includes: normalized displacement, stress, settlement, pore water pressure and their first difference, as well as construction stage coding. The output is a displacement prediction sequence for the next M time steps, such as M=36, corresponding to the next 6 hours. Each hour corresponds to 6 time steps, that is, every ten minutes is one time step.

[0090] Step S4 specifically includes the following steps:

[0091] Step S4.1: Calculate the rate of change of the predicted displacement value corresponding to the predicted displacement value sequence. The specific formula is as follows:

[0092]

[0093] in, This represents the rate of change of the predicted displacement value. Indicates the prediction time step. and This represents the displacement prediction values ​​at future times t+m and t+m-1, output by the trained LSTM prediction model.

[0094] Step S4.2: Calculate the displacement risk factor and the rate of change risk factor based on the predicted displacement value and the rate of change of the predicted displacement value. The specific formula is as follows:

[0095]

[0096] in, Indicates displacement risk factor, Indicates the rate of change risk factor. This indicates the design-allowed limit displacement value. This represents the design's allowable limit rate of change.

[0097] Step S4.3: Using a weighted fusion method, the displacement risk factor and the rate of change risk factor are fused to generate a comprehensive predicted state index. The specific formula is as follows:

[0098]

[0099] in, This represents the predicted state index at time t+m. A value greater than 1 indicates that the predicted state will exceed the safe range. This represents the cumulative risk weight of displacement. The default value is 0.7, which means that in risk assessment, the contribution of cumulative displacement accounts for 70%, and the instantaneous rate of change accounts for 30%. This parameter can be adjusted according to the importance of the project.

[0100] The system executes a tiered response based on the calculated predicted state index sequence and its maximum value: if the maximum value is less than 0.7, only monitoring is recorded; if the maximum value is greater than 0.7 but less than 1, the multi-objective optimization process is triggered to generate adjustment instructions; if the maximum value is greater than 1, while performing optimization adjustment, the highest level alarm is sent to the human-machine interaction module, and an emergency support plan is recommended or automatically activated.

[0101] In step S5, the constraints of the multi-objective optimization model include equipment safety constraints, motion stability constraints, and force coordination constraints.

[0102] The equipment safety constraint is that the prestress value of each support unit after adjustment must be within the allowable working range of the mechanical design.

[0103] The motion smoothness constraint is that the absolute value of the adjustment of each unit in a single control cycle must not exceed the product of its maximum allowable adjustment rate and the cycle duration.

[0104] The force coordination constraint limits the total deviation between the adjustment amount of multiple support points located in the same force region or structural unit and the average adjustment amount of that region.

[0105] The multi-objective optimization model is solved using the NSGA-II algorithm.

[0106] The specific formula for the multi-objective optimization model is as follows:

[0107]

[0108] in, This represents the decision variable vector, specifically the prestress adjustment amount for the K support units. This represents the minimization operator for the vector of decision variables. This represents the maximum predicted state index over the next M time points. This represents the target adjustment amount of prestress for the k-th support unit during this control cycle. This represents the cost weighting coefficient. Indicates the number of support units;

[0109] The cost weighting coefficient is used to quantify the difference in adjustment costs of different support units in the optimization objective. Its value is related to the type of support unit, the duration of use, and the current remaining design tensioning times, so as to encourage the system to prioritize the adjustment of units with better status and lower cost.

[0110] The constraint formulas for the multi-objective optimization model are as follows:

[0111]

[0112] in, and These represent the minimum and maximum prestress values ​​allowed by the mechanical design of the equipment, respectively. This represents the measured prestress before adjustment of the k-th support unit. This represents the maximum allowable adjustment amount per unit time for the k-th unit. This indicates the time interval between two adjustment commands. This represents the q-th force-coordination group. This represents the total number of support units in group q. Indicates the tolerance threshold for collaborative deviation. This represents the adjustment amount of unit x within the force coordination group. When k traverses each unit within the group, x also traverses all units within the same group. In order to limit the adjustment actions of all units within the same force coordination group, they should not differ too much and should be relatively uniform. That is, k is the outer loop index, representing the unit whose adjustment amount is currently being evaluated for coordination, and x is the inner loop index, used to traverse all units within the group to calculate the key average value.

[0113] The tolerance threshold for coordinated deviation is determined based on the difference coefficient of the design bearing capacity of the support units within the force coordination group and the standard deviation of the force value fluctuation in the historical adjustment of the group. The specific calculation formula is as follows:

[0114]

[0115] in, This represents the ratio of the maximum to the minimum design bearing capacity of the unit within the group, used to reflect the differences between components. This represents the standard deviation of the force fluctuation in the historical adjustment of this group. This represents the preset basic threshold, a benchmark value set according to the support structure design specifications and engineering experience, typically ranging from 50kN to 100kN, corresponding to the allowable adjustment deviation range of a single anchor cable or support under normal working conditions.

[0116] like Figure 3 As shown, the standardized historical monitoring sequence obtained after preprocessing is input into a pre-trained LSTM model, which outputs a sequence of predicted displacement values ​​for key points of the foundation pit within a future period (e.g., the next 6 hours). Based on this displacement prediction sequence, the corresponding displacement change rate sequence is calculated. The predicted displacement values ​​and their change rates are compared and normalized with the displacement limit values ​​and change rate limit values ​​specified in the design safety standards, respectively, to calculate factors reflecting the risk of exceeding displacement limits and factors reflecting the risk of excessive change. Finally, the risk factors are weighted and fused using preset weights to generate a comprehensive, dimensionless prediction state index sequence. This index intuitively quantifies the overall risk level of the foundation pit at various future times. When the FSI is greater than 1, it indicates that the predicted state will exceed the safe range, thereby triggering subsequent optimization and control of the system.

[0117] like Figure 2 As shown, the foundation pit support strength adaptive adjustment system based on real-time monitoring data is used to implement any of the foundation pit support strength adaptive adjustment methods based on real-time monitoring data, including:

[0118] The monitoring and sensing module consists of multiple types of sensor nodes and data acquisition units deployed at the foundation pit site, used to collect monitoring data in real time.

[0119] The data preprocessing and communication module is used to receive monitoring data, complete local preprocessing, and transmit it to the computing unit via wired or wireless network;

[0120] An edge computing platform is used to deploy and run the LSTM prediction model and the multi-objective optimization decision algorithm.

[0121] The intelligent support execution module includes several independent intelligent support units with force feedback and remote control functions;

[0122] The human-computer interaction and early warning module is used to provide a visual monitoring interface, early warning information dissemination, and manual parameter configuration.

[0123] The system adopts a cloud-edge collaborative architecture. The edge computing platform is deployed on the construction site and is responsible for running the forward inference of the LSTM prediction model, real-time risk assessment, and fast multi-objective optimization solution to ensure low-latency generation of control commands. The cloud server is responsible for the long-term training of the LSTM prediction model, parameter optimization, storage and macro-analysis of the entire project data, and regularly sends the updated model parameters to the edge computing platform.

[0124] like Figure 4 As shown, the intelligent support unit in the intelligent support execution module is an intelligent prestressed anchor cable system, which specifically includes: anchor cable body, bearing plate and intelligent tensioning and locking device.

[0125] The intelligent tensioning and locking device integrates a servo hydraulic drive mechanism, a force sensor, a displacement sensor, and an embedded controller; wherein, the embedded controller is configured to receive remote adjustment commands and drive the servo hydraulic drive mechanism to achieve dynamic adjustment and locking of the anchor cable prestress.

[0126] Any combination of one or more computer-readable media may be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0127] The examples described herein are merely preferred embodiments of the invention and are not intended to limit the concept and scope of the invention. Any modifications and improvements made by those skilled in the art to the technical solutions of the invention without departing from the design concept of the invention should fall within the protection scope of the invention.

Claims

1. A method for adaptive adjustment of foundation pit support strength based on real-time monitoring data, characterized in that, Includes the following steps: Step S1: Real-time collection of monitoring data on the foundation pit support structure and surrounding environment; Step S2: Preprocess the monitoring data to obtain standardized time series data; Step S3: Input the standardized time series data into the pre-trained LSTM prediction model and output the displacement prediction value sequence of the foundation pit monitoring points within a preset time period in the future. Step S4: Based on the displacement prediction value sequence, calculate the predicted state index reflecting the stability of the foundation pit at each future time. Step S5: With the objectives of minimizing the maximum predicted state index and minimizing the total cost of support strength adjustment, construct a multi-objective optimization model and solve for the optimal prestress adjustment amount of each intelligent support actuator. Step S6: Convert the optimal prestress adjustment amount into a control command to drive the corresponding intelligent support actuator to perform the action and complete the adaptive adjustment of the support strength. Specifically, step S2 includes the following steps: Step S2.1: Resample and align sensor data with different sampling frequencies and transmission delays to form data frames with consistent timestamps; Step S2.2: For missing data points, based on the spatial topological relationship of the monitoring points and the temporal correlation of the data, construct a spatiotemporal kernel function, and perform adaptive interpolation by weighted fusion of the observations of nearby sensors within the effective spatiotemporal window. Step S2.3: Perform a joint test on the data based on statistical distribution characteristics and the principle of mechanical equilibrium to identify and correct true outliers; In step S3, the LSTM prediction model uses the risk-weighted average absolute error function as the training loss function. The risk-weighted average absolute error function is configured to dynamically adjust its weight in the loss function calculation based on the proximity of the predicted or actual displacement values ​​in the training samples to a preset engineering warning threshold. Step S4 specifically includes the following steps: Step S4.1: Calculate the rate of change of the predicted displacement value corresponding to the predicted displacement value sequence; Step S4.2: Calculate the displacement risk factor and the rate of change risk factor based on the predicted displacement value and the rate of change of the predicted displacement value; Step S4.3: Using a weighted fusion method, the displacement risk factor and the rate of change risk factor are fused to generate a comprehensive predicted state index; In step S5, the constraints of the multi-objective optimization model include equipment safety constraints, motion stability constraints, and force coordination constraints. The equipment safety constraint is that the prestress value of each support unit after adjustment must be within the allowable working range of the mechanical design. The motion smoothness constraint is that the absolute value of the adjustment of each unit in a single control cycle must not exceed the product of its maximum allowable adjustment rate and the cycle duration. The force coordination constraint limits the total deviation between the adjustment amount of multiple support points located in the same force area and the average adjustment amount of support points in the force area; The multi-objective optimization model is solved using the NSGA-II algorithm.

2. The method according to claim 1, characterized in that, In step S1, the monitoring data includes at least the deep and top horizontal displacement of the support structure, the stress of the support components, the surface settlement around the foundation pit, and the pore water pressure.

3. The method according to claim 2, characterized in that, Step S2.3 specifically includes the following steps: A sliding window-based statistical method was used for preliminary outlier screening. Based on the principle of mechanical equilibrium of the foundation pit, the anomalies initially screened out were verified a second time to distinguish between real abnormal data and reasonable data mutations. For data that is confirmed as a genuine anomaly through secondary verification, spatiotemporal interpolation values ​​are used for correction.

4. An adaptive adjustment system for foundation pit support strength based on real-time monitoring data, used to implement the adaptive adjustment method for foundation pit support strength based on real-time monitoring data as described in any one of claims 1-3, characterized in that, include: The monitoring and sensing module consists of multiple types of sensor nodes and data acquisition units deployed at the foundation pit site, used to collect monitoring data in real time. The data preprocessing and communication module is used to receive monitoring data, complete local preprocessing, and transmit it to the computing unit via wired or wireless network; An edge computing platform is used to deploy and run the LSTM prediction model and the multi-objective optimization decision algorithm. The intelligent support execution module includes several independent intelligent support units with force feedback and remote control functions; The human-computer interaction and early warning module is used to provide a visual monitoring interface, early warning information dissemination, and manual parameter configuration.

5. The system according to claim 4, characterized in that, The intelligent support unit in the intelligent support execution module is an intelligent prestressed anchor cable system, which specifically includes: anchor cable body, bearing plate and intelligent tensioning and locking device.

6. The system according to claim 5, characterized in that, The intelligent tensioning and locking device integrates a servo hydraulic drive mechanism, a force sensor, a displacement sensor, and an embedded controller; wherein, the embedded controller is configured to receive remote adjustment commands and drive the servo hydraulic drive mechanism to achieve dynamic adjustment and locking of the anchor cable prestress.

Images