Foundation pit deformation real-time monitoring method, device and system based on BIM technology

Abstract

The invention relates to the technical field of foundation pit detection, and discloses a foundation pit deformation real-time monitoring method, device and system based on a BIM technology. The method comprises the steps of sensor layout and data acquisition, data transmission and preprocessing, nonlinear dynamic model construction, data fusion and optimization, foundation pit deformation prediction, BIM visual presentation and data display, intelligent support optimization and construction scheme adjustment. The device comprises a sensor assembly, a data acquisition interface, a 5G communication module, a central processing unit, a fusion and prediction unit, a man-machine interaction interface and an execution mechanism interface. The system comprises a data acquisition module, a data transmission module, a data processing module, a deformation modeling module, a data fusion module, a deformation prediction module, a visualization module and an intelligent regulation and control module. By adopting the multi-type sensor fusion and data assimilation technology, the high-precision and low-error foundation pit deformation monitoring effect is achieved, the monitoring data is more stable, and the monitoring precision is ensured to meet the engineering safety management requirement.

Description

Technical Field

[0001] The present invention relates to the field of foundation pit detection technology, and specifically to a method, device and system for real-time monitoring of foundation pit deformation based on BIM technology. Background Art

[0002] Foundation pits are temporary or permanent structures used to support soil and ensure the safety of construction space during underground engineering construction. They are widely used in subway construction, underground parking lots, high-rise building foundation excavation and other fields. After the foundation pit is excavated, the stress state of the surrounding soil changes, resulting in varying degrees of deformation of the retaining structure and adjacent strata. This deformation includes horizontal displacement, vertical settlement, tilt deformation, changes in the internal force of the retaining structure, etc., which are affected by multiple factors such as soil characteristics, groundwater level, construction process, and support structure stiffness. Reasonable monitoring of foundation pit deformation and taking corresponding support measures are of great significance to prevent engineering accidents such as support instability, soil collapse, and surrounding building settlement.

[0003] Foundation pit deformation monitoring has been widely used in engineering construction safety management. Various sensors are mainly deployed around the foundation pit and on the surrounding structure to obtain the deformation status of the foundation pit in real time. Common monitoring methods include traditional manual measurement (such as level, total station), automated monitoring (such as GPS, displacement sensor, strain gauge) and new measurement methods such as ground-based radar and fiber grating.

[0004] However, the integrity and reliability of monitoring data of existing real-time monitoring technology for foundation pit deformation are low. It is difficult for a single sensor type to fully reflect the deformation of the foundation pit. In addition, unreasonable measurement point layout or sensor failure may lead to data loss, affecting the overall monitoring accuracy. The deformation prediction ability is limited. Traditional methods are mostly based on empirical formulas or simple trend extrapolation, which is difficult to accurately predict the nonlinear and time-varying deformation evolution process, resulting in delayed warning and affecting construction safety. In addition, the visualization ability of monitoring data is insufficient. The existing technology mostly uses two-dimensional data tables or simple curve graphs to display deformation trends, lacks intuitive presentation of three-dimensional deformation of the foundation pit, and it is difficult for engineering personnel to identify key risk areas in a timely manner. Therefore, the present invention provides a real-time monitoring method, device and system for foundation pit deformation based on BIM technology to solve the shortcomings of the existing technology. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a method, device and system for real-time monitoring of foundation pit deformation based on BIM technology, which solves the problems that the integrity and reliability of monitoring data of the existing real-time monitoring technology for foundation pit deformation are low, and a single sensor type is difficult to fully reflect the deformation of the foundation pit.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A real-time monitoring method for foundation pit deformation based on BIM technology comprises the following steps: Sensors are placed around and at the bottom of the foundation pit to collect displacement, inclination, strain, soil stress, temperature and humidity, and groundwater level data; the collected data are transmitted to the data processing unit, which performs noise filtering, outlier elimination, and data completion on the collected data; Based on the collected data obtained by the data processing unit, the predicted value of the nonlinear dynamic model of the foundation pit deformation is constructed; The variational data assimilation method is used to optimize the fusion of the predicted value of the nonlinear dynamic model and the data collected by the sensor to obtain the corrected foundation pit deformation state data; Based on the corrected foundation pit deformation state data, the extended Kalman filter method is used to predict and calculate the future foundation pit deformation state; Based on the prediction results of the extended Kalman filter method, the deep reinforcement learning method is used to optimize the foundation pit support measures and adjust the construction plan; Based on BIM technology, the deformation status data of the foundation pit is visualized, and the key parameters of the foundation pit deformation are displayed in real time in the three-dimensional digital model.

[0007] Preferably, the sensor layout is optimized using optimal observation theory to calculate the optimal sensor layout plan to maximize the information entropy of the collected data to improve the integrity and spatial resolution of the data.

[0008] Preferably, the sensor network includes an adaptive wireless network and a 5G communication module, and the data processing unit uses an edge computing method to perform preliminary data processing, including data interpolation, denoising and abnormal data removal.

[0009] Preferably, the nonlinear dynamic model uses fractional derivative form to describe the historical memory effect of foundation pit deformation, and the dynamic equation is: Among them, x(t) is the pit deformation state vector, which represents the strain state variable of the pit at time t, A is the pit system stiffness matrix, B is the support control influence matrix, u(t) is the external influencing factor, f(x,t) is the nonlinear disturbance term, and w(t) is the system random noise. It is the description of foundation pit deformation based on fractional derivatives, and α is the fractional order.

[0010] Preferably, the collected data is optimized and fused using a variational data assimilation method, and the optimization objective function is: Among them, J(x) is the objective function, x is the foundation pit deformation state vector to be optimized, and x b is the background field data, B is the background error covariance matrix, H(x) is the observation operator, y is the sensor measured data, and R is the observation error covariance matrix.

[0011] Preferably, the modified foundation pit deformation state is predicted by an extended Kalman filter method, and its state transfer equation is: x k+1 =A k x k +B k u k +w k ; The observation equation is: y k =H k x k +v k ; Among them, x k+1 is the state transfer equation, A k is the state transfer matrix, B k is the control input matrix, x k is the deformation state vector of the foundation pit at the current moment, w k and v k are system noise and measurement noise respectively, which obey Gaussian distribution, y k is the sensor observation data, H k is the observation matrix, u k For external input quantity.

[0012] Preferably, the BIM technology constructs a three-dimensional foundation pit monitoring model based on the digital twin technology, and displays the deformation state of the foundation pit in real time in a visual interface, including displacement, inclination, strain, soil stress, temperature and humidity, and groundwater level parameters.

[0013] Preferably, the foundation pit support measures are optimized by a deep reinforcement learning method, wherein the reinforcement learning model is constructed based on a Markov decision process, including a state space, an action space and a reward function, wherein the reward function is defined as: R 1 =-∥x t -x safe ∥ 2 ; Among them, x t is the current deformation state of the foundation pit, x safe is the safety threshold, R 1 is the reward value, ∥x t -x safe ∥ 2 is the Euclidean norm.

[0014] A real-time monitoring device for foundation pit deformation based on BIM technology is also provided, including: A sensor assembly for collecting physical parameters of foundation pit deformation; A data acquisition interface, used to receive data collected by the sensor assembly; 5G communication module, used to transmit the data acquired by the data acquisition component to the central processor; The central processing unit is used to filter noise, remove outliers, and complete data for the transmitted data; Fusion and prediction unit, used to optimize data fusion and calculate foundation pit deformation trends using variational data assimilation methods; Human-computer interaction interface, used to present a 3D visualization model of foundation pit deformation based on BIM technology and display deformation trend analysis results; The actuator interface is used to adjust the foundation pit support measures based on the foundation pit deformation trend and send control instructions to the actuator.

[0015] It also provides a real-time monitoring system for foundation pit deformation based on BIM technology, including the following modules: A data acquisition module, used for collecting foundation pit deformation data through a sensor unit; A data transmission module, used to transmit the data collected by the data acquisition module to the data processing module through the wireless sensor network and the 5G communication module; The data processing module is used to receive the data transmitted by the data transmission module, and perform format conversion, abnormal data removal, noise filtering and data completion on the data; A deformation modeling module is used to construct a nonlinear dynamic model of foundation pit deformation based on the data provided by the data processing module; The data fusion module is used to optimize the foundation pit deformation data by using the variational data assimilation method, fuse the sensor measured data and the dynamic model calculation results, and generate the corrected foundation pit deformation state data; The deformation prediction module is used to predict the future deformation trend of the foundation pit based on the corrected foundation pit deformation state data using the extended Kalman filter method; Visualization module, used to visualize the deformation state of the foundation pit in three dimensions based on BIM technology, and to display the key parameters of the foundation pit deformation in the digital model; The intelligent control module is used to optimize foundation pit support measures based on deformation prediction results and provide construction adjustment plans.

[0016] The present invention provides a method, device and system for real-time monitoring of foundation pit deformation based on BIM technology. It has the following beneficial effects: 1. The present invention adopts multi-type sensor fusion and data assimilation technology to achieve high-precision and low-error foundation pit deformation monitoring effect, and solves the problems of incomplete data, large measurement errors, and single-point failures affecting monitoring results. The present invention optimizes the sensor measured data and the dynamic model calculated value through multi-source data fusion and combined with the variational data assimilation method, making the monitoring data more stable and credible, ensuring that the monitoring accuracy meets the engineering safety management requirements.

[0017] 2. The present invention adopts the extended Kalman filter (EKF) prediction method to achieve the effect of accurately predicting the future deformation trend of the foundation pit, solving the problems of being unable to quantify future deformation, prediction lag, and lack of dynamic adjustment basis. The present invention constructs a nonlinear state space model based on real-time data, and combines time series analysis to calculate the deformation trend in the short term, providing a scientific basis for construction regulation and improving construction safety and the ability to cope with sudden deformation.

[0018] 3. The present invention adopts BIM visualization technology combined with a digital twin model to achieve the effect of three-dimensional dynamic display of foundation pit deformation data, solving the problems of poor data intuitiveness, lack of spatial information, and difficulty in quickly analyzing deformation areas. The present invention maps monitoring data on the BIM model in real time to make the overall deformation trend of the foundation pit clearly visible. Engineering personnel can intuitively view the deformation status of different areas and replay the deformation process, thereby improving the efficiency and accuracy of construction management.

[0019] 4. The present invention adopts deep reinforcement learning to optimize the foundation pit support scheme, achieves the effect of intelligently controlling construction measures and reducing deformation risks, and solves the problems of lack of adaptive ability, adjustment lag, and difficulty in dynamically responding to deformation changes in support scheme optimization. The present invention trains the optimization paths of different support strategies through reinforcement learning models, automatically calculates the optimal support adjustment plan, and corrects the strategy in combination with real-time construction data, making foundation pit deformation control more intelligent and scientific, and improving engineering safety and construction efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a flow chart of the method steps of the present invention; Figure 2 It is a schematic diagram of the device of the present invention; Figure 3 It is a system architecture diagram of the present invention. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] Embodiment 1: Please see attached Figure 1 , an embodiment of the present invention provides a real-time monitoring method for foundation pit deformation based on BIM technology, comprising the following steps: S1, placing sensors around and at the bottom of the foundation pit to collect displacement, inclination, strain, soil stress, temperature and humidity, and groundwater level data; S2, transmitting the collected data to a data processing unit, and filtering the collected data for noise, removing outliers, and completing data through the data processing unit; S3, constructing a predicted value of a nonlinear dynamic model of foundation pit deformation based on the collected data obtained by the data processing unit; S4. Using the variational data assimilation method, the predicted value of the nonlinear dynamics model is optimally integrated with the data collected by the sensor to obtain the corrected foundation pit deformation state data; S5. Based on the corrected foundation pit deformation state data, an extended Kalman filter method is used to predict and calculate the future foundation pit deformation state; S6. Based on the prediction results of the extended Kalman filter method, the deep reinforcement learning method is used to optimize the foundation pit support measures and adjust the construction plan; S7. Visualize the deformation status data of the foundation pit based on BIM technology, and display the key parameters of the foundation pit deformation in real time in the three-dimensional digital model.

[0023] For step S1, data collection is the basic link of the whole system, and its accuracy and completeness directly affect the effects of subsequent data analysis, modeling, prediction and regulation. Therefore, it is necessary to combine the monitoring characteristics of different sensors, the time synchronization of data acquisition, spatial distribution optimization and other aspects for detailed design to ensure efficient acquisition and high reliability of data. In general, the deformation of the foundation pit is affected by many factors, such as soil properties, groundwater level, changes in surrounding loads, etc. Therefore, the monitoring system needs to have strong environmental adaptability and real-time response capabilities. In this embodiment, multiple types of sensors are deployed according to the deformation characteristics of different parts of the foundation pit, and combined with wireless transmission technology, automatic data collection and real-time transmission are realized.

[0024] In this embodiment, the foundation pit deformation monitoring adopts multi-source sensor fusion technology, and sensors are deployed in key locations such as the foundation pit retaining structure, support system, and surrounding soil, including but not limited to: Displacement sensor: used to measure the horizontal and vertical displacement of the foundation pit retaining structure and obtain the deformation of the support structure.

[0025] Inclination sensor: installed on the foundation pit support structure to monitor the changes in its inclination angle and determine the stability of the support system.

[0026] Strain gauges: Installed on support beams, retaining piles and internal structures of foundation pits to monitor structural strain distribution.

[0027] Soil pressure gauge: used to measure the force exerted by the soil around the foundation pit on the surrounding structure and provide stress information of the support system.

[0028] Temperature and humidity sensor: collects environmental temperature and humidity data and analyzes the impact of environmental factors on foundation pit deformation.

[0029] Piezometer: Installed at the bottom of the foundation pit and in the surrounding soil layers to monitor changes in groundwater level and pore water pressure in real time.

[0030] In a possible implementation, in order to optimize the layout of sensors and improve the integrity and spatial resolution of data, this embodiment uses the optimal observation theory to calculate the layout position and number of sensors. Through information entropy analysis, the optimal sensor distribution is determined so that the acquired monitoring data has the maximum amount of information and reduces redundant measurement points. For a deep foundation pit project, sensors can be laid out according to the importance of key parts. The corners of the enclosure structure, the node positions of the support beams, and the central area of ​​the bottom of the foundation pit are all key monitoring points. A sensor array method can be used to form a distributed monitoring network on the enclosure structure to obtain more comprehensive deformation data.

[0031] Specifically, in terms of the time synchronization of the monitoring data, this embodiment uses high-precision clock synchronization technology to ensure that the data collection time of each sensor is unified, avoid data errors caused by time deviation, and use GPS timing or IEEE1588 time synchronization protocol to control the time synchronization error of all sensors within the millisecond range. In a possible implementation, a wireless synchronization trigger method can be used to enable all sensors to collect data at the same time point, further improving the timeliness of the data.

[0032] In this embodiment, in order to improve the stability and anti-interference ability of data transmission, a wireless sensor network (WSN) is used for data transmission, and a 5G communication module is combined to provide high-speed and low-latency data transmission capabilities. WSN can adopt an adaptive channel switching strategy to dynamically adjust the communication channel in an environment with large interference to ensure the reliability of data transmission. In a complex environment with many construction machines, WSN can automatically avoid interference channels and improve data transmission efficiency. For long-distance data transmission, it can be combined with the deployment of relay nodes to ensure that the data of remote sensors is stably transmitted to the data processing unit.

[0033] In one possible implementation, during data transmission, in order to reduce the data loss rate and improve the stability of data transmission, this embodiment uses adaptive error correction coding technology, such as convolutional code or low-density parity check code (LDPC), to encode the data and improve the anti-interference ability of the data during transmission. If the sensor in a certain area loses some data due to signal interference, it can be restored through the error correction algorithm at the receiving end to avoid affecting the overall monitoring accuracy due to the loss of single-point data.

[0034] During the data collection process, some sensors may be disturbed by environmental noise, external vibration and other factors, resulting in abnormal data. In this embodiment, the data collection unit uses an adaptive filtering algorithm to preprocess real-time data, including Kalman filtering, mean filtering and wavelet denoising. When Kalman filtering is used, the state estimation equation is as follows: k∣k-1 =A k x k-1∣k-1 +B k u k +w k ; Among them, x k∣k-1 is the predicted state variable, A k is the state transfer matrix, B k is the input control matrix, u k is the external input, w k is the process noise, P k∣k-1 is the state covariance matrix, Q k is the process noise covariance matrix.

[0035] This embodiment uses an adaptive threshold rejection strategy to remove abnormal data. In general, the change range of sensor data should meet the physical laws. If a data point deviates too much from the historical data and does not conform to the deformation trend of the foundation pit, it can be regarded as an abnormal value. For the inclination sensor, if the inclination change at a certain time point exceeds 5 times the normal deformation rate, the data can be preliminarily judged to be abnormal, and secondary confirmation is performed in combination with the surrounding sensor data to avoid misjudgment.

[0036] For step S2, after data collection is completed, the acquired data needs to be transmitted and preprocessed to ensure the reliability, integrity and consistency of the data. In general, the amount of data for foundation pit deformation monitoring is large, and the sensor data may be affected by factors such as noise interference, environmental changes, and transmission delays. Therefore, the key to this step is to adopt an efficient data transmission protocol, combined with an intelligent data processing algorithm, to denoise the original monitoring data, remove outliers, and complete the data to obtain accurate data that can be used for modeling and prediction. As an option, this step can use edge computing to perform partial data preprocessing at the data collection end to reduce the transmission burden and improve the response speed.

[0037] In this embodiment, the monitoring data is transmitted to the data processing unit through a wireless sensor network (WSN) or a 5G communication module. WSN adopts a multi-hop transmission mechanism to reduce the signal attenuation caused by long-distance transmission and ensure the integrity and real-time performance of the data. For a large amount of sensor data deployed around the foundation pit, the system can prioritize data aggregation through neighboring nodes, and then upload it to the data processing unit from the master node, thereby improving the stability of data transmission.

[0038] In one possible implementation, in order to reduce the bit error rate during data transmission, this embodiment adopts forward error correction coding (FEC) technology, such as low-density parity check code (LDPC) or convolutional code, to improve the anti-interference ability of data transmission. Specifically, at the data transmitting end, the transmission data ddd is processed by the encoder to obtain the encoded data ccc, which is then sent to the receiving end for decoding. If some bits are interfered during the transmission process, the receiving end can perform error correction through the decoder to improve data integrity.

[0039] In this embodiment, the data processing unit preprocesses the received raw data, including noise filtering, outlier removal, data completion and other operations. As an option, this embodiment uses an adaptive filtering algorithm to reduce the noise of the data to reduce the random error in the sensor measurement process. In a high noise environment, Kalman filtering can be used for dynamic data smoothing, and its prediction and update equations are as follows: x k∣k-1 =A k x k-1∣k-1 +B k u k +w k ; Among them, x k∣k-1 is the predicted state variable, x k-1∣k-1 is the foundation pit state estimated at the last moment k-1, A k is the state transfer matrix, B k is the input control matrix, u k is the external input, w k is the process noise, P k∣k-1 is the state covariance matrix, Q k is the process noise covariance matrix, P k-1∣k-1 is the error covariance matrix of the k-1 estimate at the previous moment, is the transpose of the state transfer matrix.

[0040] In some embodiments, in order to further improve the noise suppression effect, a wavelet denoising method can be combined to perform multi-scale signal analysis. Specifically, the monitoring data X(t) is decomposed into different frequency components through wavelet transform, where the high-frequency component mainly contains noise information. After removing it by setting a threshold, the inverse wavelet transform is performed to restore the signal, thereby achieving noise filtering. For example, a soft threshold denoising method can be used, and its processing formula is: Among them, sign(X) is the sign function, is the denoised signal, X is the original signal, λ is the threshold parameter, and max(|X|-λ,0) sets the signal whose absolute value is less than the threshold λ to zero.

[0041] In this embodiment, the outliers that may exist in the sensor data are eliminated by using statistical analysis methods. For example, the abnormal data screening criteria are set based on the 3σ principle, that is, if a data value deviates from the historical mean μ by more than 3 times the standard deviation σ, then the data point can be regarded as an outlier and eliminated, that is, the values ​​that meet the following relationship will be eliminated: |X i -μ|>3σ; Among them, X i is the current data value, μ is the mean of the data, and σ is the standard deviation.

[0042] In a possible implementation, in order to avoid the impact of missing data on modeling and analysis, this embodiment uses an interpolation algorithm to complete the data. Common interpolation methods include linear interpolation, spline interpolation, and Kalman predictive interpolation. The calculation formula for linear interpolation is as follows: Among them, X(t) is the interpolation result at time t, which represents the estimated value of the foundation pit deformation, X(t 1 ) and X(t 2 ) is the known time t 1 and t 2 The foundation pit deformation data, t is the time point to be estimated, To calculate t 1 and t 2 The rate of change between 1 ) is the time difference. This method is suitable for situations where the data changes relatively slowly. For monitoring data with nonlinear changes, the spline interpolation method can be used to improve the interpolation accuracy.

[0043] In some embodiments, the data processing unit adopts an edge computing architecture for distributed data processing, that is, part of the data preprocessing is completed on the sensor node or local computing device close to the data source, reducing the computing burden of the central server and improving the real-time performance of the system. Some sensors can have built-in embedded computing units to directly perform data filtering and anomaly detection, and then only upload the processed data to the server, reducing unnecessary data transmission.

[0044] This embodiment also involves an optimization management strategy for data storage. Generally, the data of foundation pit deformation monitoring needs to be stored for a long time to facilitate subsequent trend analysis and model correction. Therefore, this embodiment adopts a hierarchical storage strategy, that is, short-term data is stored in a cache, and long-term data is stored in a database or cloud server. Specifically, the data storage hierarchy includes: Real-time cache: stores the latest monitoring data for quick access and processing; Local database: stores monitoring data for the past week and supports query and analysis; Remote cloud storage: Stores long-term historical data and provides data backup and security management.

[0045] In one possible implementation, data storage is managed using a NoSQL database (such as MongoDB) to improve the reading and writing efficiency of large-scale monitoring data. Sensor data can be stored in JSON format and queried by time index, making data calls more flexible.

[0046] For step S3, based on the results of data collection and preprocessing, it is necessary to further establish a mathematical model of foundation pit deformation in order to accurately describe the deformation law and provide theoretical support for subsequent state prediction and regulation. In general, foundation pit deformation is a nonlinear, time-varying, multi-factor coupled dynamic process. It is difficult to accurately reflect the complexity of foundation pit stress and deformation by relying solely on linear models. Therefore, the present invention adopts a nonlinear dynamic model for modeling, and combines the fractional-order differential method to describe the historical memory effect and time-lag characteristics of foundation pit deformation. As an option, the sensor measured data and the finite element analysis results can be fused in the modeling process to improve the adaptability and accuracy of the model.

[0047] In this embodiment, the mathematical modeling of foundation pit deformation is based on the fractional derivative dynamics model, and its mathematical expression is as follows: Among them, x(t) is the state vector of foundation pit deformation, including horizontal displacement, vertical displacement, inclination, strain and other variables. is a fractional derivative, 0<α<1, which represents the historical dependence characteristics and long-term memory effect of foundation pit deformation. Compared with integer-order derivatives, it can more realistically describe the creep and long-term deformation characteristics of soil. A is the system stiffness matrix, which describes the stiffness characteristics of the retaining structure and the foundation pit support system and determines the deformation response of the foundation pit. B is the control input matrix, which represents the influence of external regulation (such as support stiffness adjustment and dewatering measures) on the deformation of the foundation pit. u(t) is the external excitation vector, including environmental factors such as construction load changes and groundwater level changes. f(x, t) is a nonlinear disturbance term, which represents the nonlinear dynamic influence caused by factors such as nonlinear characteristics of soil and deformation of retaining structure. w(t) is a random noise term. Considering measurement errors, environmental interference and unmodeled factors, it is usually assumed to obey Gaussian white noise distribution.

[0048] In one possible implementation, the fractional derivative is calculated using the Caputo definition, which is expressed as follows: in, is the Caputo fractional derivative, which means the α-order differentiation of x(t), 0<α<1, Γ(1-α) is the Gamma function, x(τ) is the deformation state at the past time τ, (t-τ) -α is a weight function, which indicates the dependence of fractional derivatives on historical data, and dx(τ) / dτ is the integer derivative operation on x(τ). This integral form shows that the current state of foundation pit deformation is affected by historical deformation data, which is consistent with the characteristics of soil creep and long-term deformation.

[0049] Specifically, in this embodiment, the dynamic characteristics of foundation pit deformation can be divided into elastic response stage, plastic development stage and long-term creep stage. Among them, the elastic response stage is mainly controlled by the stiffness matrix A, the plastic development stage is mainly affected by the external load matrix B, and the long-term creep stage is dominated by the fractional order term.

[0050] For step S4, the variational data assimilation method is used to jointly optimize the model calculation results with the observed data to make the final deformation state closer to the actual situation and improve the reliability and consistency of the monitoring data. In general, the core of data assimilation is to balance the weights of measured data and model prediction data to ensure that the final deformation state is consistent with actual observations while maintaining the rationality of the model.

[0051] In this embodiment, the main goal of the variational data assimilation method is to correct the deformation state calculated by the dynamic model so that it is as close as possible to the real data measured by the sensor. In the process of data assimilation, the system first uses the data collected by the sensor as input, and combines it with the calculated value of the foundation pit deformation model to determine the difference between the data. Subsequently, based on error analysis, the deformation state is optimized so that the model calculation results can more accurately reflect the deformation trend of the monitored area while maintaining physical rationality.

[0052] Specifically, the data assimilation process involves two main parts: first, the observed data and the model prediction data are compared to calculate the error between the two. The error calculation needs to take into account the sensor measurement error, environmental interference, and uncertainty in model calculation. In the error analysis stage, the system will evaluate the measurement deviations of various sensors based on historical monitoring data and assign different weights. If the measurement error of a sensor is large, the influence weight of the sensor data in data assimilation will be appropriately reduced, and the sensor data with higher measurement accuracy will receive a higher weight.

[0053] In the optimization stage of data assimilation, the system adjusts the deformation state of the foundation pit so that it conforms to the observed data and satisfies the constraints of the dynamic model. The optimization process is usually carried out in an iterative manner. Each adjustment combines the model prediction data with the corrected observed data to gradually reduce the gap between the two. Finally, when the error between the corrected deformation state and the observed data is reduced to a reasonable range, the data assimilation process ends and the optimized deformation state data is output.

[0054] In some embodiments, in order to further improve the accuracy of data assimilation, the system will consider the spatial correlation of sensor data. The deformation of different areas of the foundation pit does not occur in isolation, but affects each other. Therefore, during the data assimilation process, the system will analyze the data correlation between different sensors and adjust the deformation state of adjacent areas. If the sensor in a certain area measures a large deformation, but the sensor data in the adjacent area does not show abnormalities, the system may make reasonable corrections to the abnormal data to ensure data consistency.

[0055] In another possible implementation, the results of data assimilation can also be used to optimize the parameters of the dynamic model. Since the deformation of the foundation pit is affected by external factors such as groundwater level changes and soil pressure adjustments, the stiffness parameters and damping parameters of the dynamic model may need to be adjusted according to real-time monitoring data. Through data assimilation, the model parameters can be dynamically corrected based on the latest monitoring data to make the model calculation results more consistent with the current state of the foundation pit. In soft soil areas, due to the low deformation modulus of the soil, the system can adjust the stiffness parameters in the model to make the deformation predicted by the model more consistent with the actual monitoring data.

[0056] For step S5, it is necessary to further predict the optimized deformation state data in order to give early warning of the foundation pit deformation trend and provide a decision-making basis for the optimization of support measures. In general, foundation pit deformation is a complex nonlinear dynamic process, which is affected by factors such as soil creep, groundwater level changes, and construction load adjustments. Therefore, the prediction method needs to have strong anti-interference ability and time-varying adaptability. The present invention adopts the extended Kalman filter (EKF) method, combined with the foundation pit deformation state data after data assimilation, to estimate the future deformation trend. As an option, the time series analysis method can be combined to perform trend correction on the short-term prediction results to improve the prediction accuracy.

[0057] In this embodiment, the extended Kalman filter method predicts deformation based on a nonlinear state space model, and the prediction process is divided into two stages: state prediction and measurement update. In the state prediction stage, the system uses the deformation state data of the previous moment and combines it with the foundation pit deformation dynamics model to calculate the deformation state of the next moment. Specifically, the system uses a nonlinear state transfer equation to describe the evolution of foundation pit deformation, which can be expressed as: xk+1 =f(x k ,u k )+w k ; Among them, x k is the deformation state of the foundation pit at the current moment, including variables such as horizontal displacement, vertical displacement, inclination, strain, etc., f(x k ,u k ) is a nonlinear state transfer function, which indicates the dynamic evolution law of foundation pit deformation affected by factors such as support stiffness, soil pressure, and groundwater level. k External inputs include construction load changes, rainfall effects and other environmental factors. k is the system noise, which considers the random influence of unmodeled factors on the deformation state.

[0058] Specifically, the prediction correction process of the extended Kalman filter includes the following steps: State prediction: Based on the deformation state at the previous moment, the predicted state at the next moment is calculated using the dynamic equation; Error covariance prediction: calculate the uncertainty of the predicted state and evaluate the credibility of the prediction results; Measurement update: Combine sensor observation data to correct the predicted state to make it closer to the actual deformation; Update the error covariance matrix: adjust the uncertainty of the state estimate to improve the stability of subsequent predictions.

[0059] In some embodiments, in order to improve the short-term accuracy of deformation prediction, the system can combine time series analysis methods for trend correction, use sliding window methods to analyze recent deformation data, and establish a short-term trend model to optimize the prediction results of EKF. Common time series analysis methods include autoregressive integrated moving average model (ARIMA) and long short-term memory network (LSTM), among which the LSTM method is suitable for nonlinear time series prediction and can be used to capture the long-term trend of foundation pit deformation.

[0060] In another possible implementation, the system can perform confidence analysis on the prediction results to evaluate the credibility of the prediction data. In general, the uncertainty of the prediction is affected by sensor measurement errors, external interference factors, etc. Therefore, the standard deviation of the prediction error can be calculated, and the confidence interval can be set. The prediction interval can be calculated based on the historical error distribution. If the predicted value exceeds the normal deformation range, an early warning signal will be issued to remind managers to take measures.

[0061] In some application scenarios, in order to improve computing efficiency, the EKF prediction process can be optimized in combination with parallel computing technology. In a multi-measurement point foundation pit monitoring system, the deformation state of each measurement point can be predicted independently. Therefore, a multi-threaded parallel computing mode can be adopted to process data from multiple measurement points at the same time to improve the overall prediction efficiency. Under the cloud computing architecture, the prediction calculation tasks can be distributed to multiple computing nodes to achieve efficient distributed computing.

[0062] For step S6, after predicting the deformation state of the foundation pit, it is necessary to visualize the prediction results so that engineering personnel can intuitively obtain the current and future deformation information of the foundation pit. In general, foundation pit deformation monitoring involves multidimensional data, including displacement, inclination, strain, soil pressure and groundwater level changes at different measuring points. A single numerical data is difficult to fully display the overall deformation trend of the foundation pit. Therefore, the present invention is based on BIM (Building Information Modeling) technology to construct a three-dimensional digital model of the foundation pit, and maps the deformation state data in the model in real time to achieve dynamic visualization of the foundation pit deformation. As an option, digital twin technology can be combined in the visualization process to synchronize computer simulation data with measured data to improve the real-time and accuracy of visualization.

[0063] In this embodiment, the BIM visualization platform takes the three-dimensional digital model as the core, integrates the monitoring data with the prediction data, and presents the deformation state of the foundation pit in three-dimensional space. Specifically, the system first constructs the BIM three-dimensional structure based on the foundation pit construction design model, including the enclosure structure, support system, and the internal topography of the foundation pit, and defines the spatial position of each monitoring point in the model.

[0064] In the data mapping stage, the system associates the deformation monitoring data with the BIM model and adjusts the model's geometry or color identification in real time according to the changes in different deformation parameters. For example: The displacement data can be represented by color gradient. The area with greater displacement has darker color, which directly reflects the deformation trend of the foundation pit.

[0065] The inclination angle change can dynamically adjust the inclination angle of the enclosure structure in the model to display the stress state of the structure.

[0066] The strain information is displayed on the surface of the structure in the form of a thermal map, identifying areas of strain concentration in order to analyze changes in structural stress.

[0067] The changes in groundwater levels are represented by transparent blue areas at different heights, and the trend of groundwater level changes is updated dynamically.

[0068] In one possible implementation, in order to improve the real-time interactive capabilities of the BIM model, the system uses WebGL rendering technology, allowing engineers to directly view the three-dimensional visualization results through computers or mobile devices, and perform interactive operations such as rotating, zooming, and selecting specific measurement points to view detailed data.

[0069] Generally speaking, the visualization of foundation pit deformation needs to take into account the timeliness and continuity of data. This embodiment uses time series animation technology to enable the visualization system to replay the historical deformation process and predict future deformation trends. The system can dynamically adjust the shape of the BIM model based on the future deformation data calculated by the prediction model, so that engineering personnel can identify potential deformation risks in advance and formulate corresponding measures.

[0070] In some embodiments, the visualization system supports multi-level data presentation, that is, the user can choose to view data of different granularities according to needs: The global perspective shows the overall deformation trend of the foundation pit, such as overall inclination and changes in the force of the support system.

[0071] Local details allow users to zoom in and view detailed data of specific monitoring points, such as the strain curve of a support beam, the displacement trend of a measuring point, etc.

[0072] Section analysis cuts the BIM model to check the deformation of the underground structure and intuitively analyze the correlation between retaining piles, support beams, and groundwater levels.

[0073] In terms of data storage and management, this embodiment adopts a cloud storage architecture to store BIM visualization data, historical monitoring data, and predictive data in the cloud to support cross-device access and remote data analysis. Engineering management personnel can access the visualization results of foundation pit deformation on different devices through the cloud platform and conduct comparative analysis of historical data.

[0074] For step S7, the BIM visualization system can intuitively present the current and predicted deformation status, but it is difficult to adjust the construction plan in time and reduce the risk of deformation by relying solely on passive visual feedback. In general, the deformation trend of the foundation pit is not only affected by the characteristics of the soil itself, but is also closely related to the support measures, precipitation schemes and external load adjustments during the construction process. Therefore, after completing the deformation prediction, it is necessary to optimize and adjust the foundation pit support plan based on the prediction results to ensure the safety and stability of the foundation pit structure. The present invention adopts a deep reinforcement learning method, combined with the deformation prediction results of the extended Kalman filter, to automatically optimize the foundation pit support measures and dynamically adjust the construction plan. As an option, a Markov decision process can be introduced in the intelligent control process to systematically evaluate the long-term effects of different support strategies and improve the automatic adjustment capability of the construction process.

[0075] In this embodiment, the intelligent control process adopts the deep reinforcement learning (DRL) method to optimize the support measures of the foundation pit. Specifically, the system models the foundation pit deformation control problem as a Markov decision process (MDP), including state space, action space and reward function.

[0076] The state space is defined as the current deformation state of the foundation pit, including: Horizontal displacement and vertical displacement are used to characterize the deformation degree of foundation pit retaining structure; The change of inclination angle reflects the stability of the enclosure structure; Strain and earth pressure distribution, used to analyze the stress conditions of the support system; Height of groundwater level, taking into account the effect of seepage on foundation pit deformation; Construction load information, including prestressing adjustments of the support system, changes in excavation depth, etc.

[0077] In one possible implementation, the system constructs a high-dimensional state space based on historical monitoring data and simulation analysis results so that the reinforcement learning model can more comprehensively understand the deformation laws of the foundation pit.

[0078] The action space is defined as possible support adjustment strategies, including: Adjustment of the stiffness of the support structure, such as increasing or decreasing the prestress of the support system; Groundwater level control, such as adjusting the pumping rate of precipitation wells to reduce groundwater infiltration pressure; Optimize the construction process, such as adjusting the excavation sequence to reduce the lateral deformation of the foundation pit; Soil pressure adjustment, such as using backfill or pull-out piles to balance soil stress.

[0079] In another possible implementation, the action space can be constrained in combination with expert experience rules to avoid non-engineering feasible support adjustment strategies. For deep foundation pit support, the support stiffness should not be significantly reduced on one side to prevent structural instability.

[0080] In another possible implementation, the action space can be constrained by combining expert experience rules to avoid non-engineering feasible support adjustment strategies. For example, for deep foundation pit support, the support stiffness should not be greatly reduced on one side to prevent structural instability. The reward function is used to measure the effectiveness of the support strategy. Its core goal is to control the deformation of the foundation pit to keep it within the safety threshold while optimizing the labor cost. In general, the reward function can be defined as follows: R=-∥x t -x safe ∥ 2 -C(a t ); Among them, xt is the current deformation state of the foundation pit, x safe is the safety threshold, C(a t ) is the cost function of support adjustment, which prevents the system from frequently adjusting support measures and causing excessive construction costs.

[0081] In this embodiment, the deep reinforcement learning model is trained using a deep Q-network (DQN). DQN approximates the Q function through a deep neural network and learns the mapping relationship between the deformation state of the foundation pit and the optimal support strategy. During the training process, the system simulates a variety of possible construction plans, evaluates the impact of different support measures on the deformation of the foundation pit, and continuously optimizes the decision-making strategy.

[0082] In another possible implementation, the proximal policy optimization (PPO) can be used to train the reinforcement learning model to improve the stability of the policy optimization. PPO limits the policy update step size to avoid excessive adjustments during model training and improve the feasibility of the support optimization scheme.

[0083] In general, the output of the intelligent control system can be used to directly control the foundation pit support equipment, such as adjusting the prestressing parameters of the automatic tensioning system or optimizing the operation strategy of the dewatering well. In one possible implementation, the intelligent control system can be combined with the engineering management platform to provide visual support optimization suggestions for engineering personnel to review and implement. When the system detects that the lateral displacement of the foundation pit exceeds the set threshold, it can automatically recommend increasing the support stiffness and provide the expected deformation recovery to assist in decision-making.

[0084] Embodiment 2: Please refer to the attached Figure 2 , a real-time monitoring device for foundation pit deformation based on BIM technology, including: A sensor assembly for collecting physical parameters of foundation pit deformation; A data acquisition interface, used to receive data collected by the sensor assembly; 5G communication module, used to transmit the data acquired by the data acquisition component to the central processor; The central processing unit is used to filter noise, remove outliers, and complete data for the transmitted data; Fusion and prediction unit, used to optimize data fusion and calculate foundation pit deformation trends using variational data assimilation methods; Human-computer interaction interface, used to present a 3D visualization model of foundation pit deformation based on BIM technology and display deformation trend analysis results; The actuator interface is used to adjust the foundation pit support measures based on the foundation pit deformation trend and send control instructions to the actuator.

[0085] Specifically, in this embodiment, the sensor assembly is used to collect physical parameters of foundation pit deformation, including but not limited to displacement sensors, inclination sensors, strain gauges, earth pressure gauges, temperature and humidity sensors, and piezometers. The sensor assembly is distributed in the foundation pit retaining structure, support system, and surrounding soil mass to monitor parameters such as horizontal displacement, vertical displacement, inclination change, strain distribution, earth pressure change, groundwater level, and environmental temperature and humidity in real time. Some sensors use wireless transmission methods to improve the flexibility of data collection.

[0086] The data acquisition interface is used to receive the data collected by the sensor assembly and perform formatting conversion on the data to ensure compatibility for subsequent processing. Generally, the data acquisition interface includes an analog signal acquisition channel, a digital signal input port, and a data buffer unit. For analog signal sensors, the data acquisition interface includes an analog-to-digital conversion (ADC) module to convert analog signals into digital signals; for digital signal sensors, the data acquisition interface can directly read the collected data and store it according to the time stamp.

[0087] The 5G communication module is used to transmit the data obtained by the data acquisition component to the central processor, supporting high-bandwidth and low-latency data transmission. As an option, the 5G communication module adopts the stand-alone networking (SA) mode to provide a more stable real-time data transmission capability. The communication module supports the automatic channel switching function and can dynamically adjust the communication frequency band in an environment with strong signal interference to ensure the stability of data transmission. In remote monitoring applications, the 5G communication module can also interact with the cloud server to achieve remote storage and analysis of data.

[0088] The central processor is used to filter out noise, eliminate outliers, and complete data for the transmitted data. Generally, the central processor adopts a multi-core architecture to support high-concurrency data processing. Noise filtering uses adaptive filtering algorithms such as Kalman filtering, mean filtering, or wavelet denoising to reduce the impact of environmental noise on monitoring data. Outlier elimination is based on statistical analysis methods, such as eliminating abnormal data points based on the 3σ principle of standard deviation. Data completion uses interpolation algorithms or time series-based prediction methods to fill in the missing monitoring data within a short period and improve data continuity.

[0089] The fusion and prediction unit is used to optimize data fusion using the variational data assimilation method and calculate the deformation trend of the foundation pit. During the data fusion process, the system performs error analysis based on the sensor measurement values and the calculated values of the nonlinear dynamics model, and uses an optimization algorithm to adjust the data weights to improve the accuracy of the fused data. The prediction calculation uses the extended Kalman filter (EKF) method, combining historical data and real-time monitoring data to calculate the future deformation state of the foundation pit. In some embodiments, the fusion and prediction unit can combine deep learning methods to improve the deformation prediction ability in complex environments.

[0090] The human-computer interaction interface is used to present a three-dimensional visualization model of foundation pit deformation based on BIM technology and display deformation trend analysis results. The BIM visualization system constructs a three-dimensional digital model of the foundation pit and dynamically maps the deformation monitoring data to the model structure to intuitively display the deformation trend. The interaction methods include touch screen operation, mouse and keyboard interaction, and remote monitoring terminal access. The system supports multi-level data display, including global deformation trends, local measurement point data, historical trend analysis, etc., and provides user-defined view adjustment functions to meet different engineering management needs.

[0091] The actuator interface is used to adjust the foundation pit support measures based on the foundation pit deformation trend and send control instructions to the actuator. The actuator includes the anchor tensioning system, the dewatering well pump control unit, the foundation pit support stiffness adjustment device, etc. The system can dynamically adjust the support parameters according to the deformation prediction results, such as increasing the anchor prestress, optimizing the dewatering rate, adjusting the support stiffness, etc. In some implementations, the actuator interface can be combined with the automatic control system to directly adjust the support equipment parameters to improve the control efficiency and reduce the response time of manual intervention.

[0092] Embodiment three: Please see attached Figure 3 , the real-time monitoring system for foundation pit deformation based on BIM technology includes: A data acquisition module, used for collecting foundation pit deformation data through a sensor unit; A data transmission module, used to transmit the data collected by the data acquisition module to the data processing module through the wireless sensor network and the 5G communication module; The data processing module is used to receive the data transmitted by the data transmission module, and perform format conversion, abnormal data removal, noise filtering and data completion on the data; A deformation modeling module is used to construct a nonlinear dynamic model of foundation pit deformation based on the data provided by the data processing module; The data fusion module is used to optimize the foundation pit deformation data by using the variational data assimilation method, fuse the sensor measured data and the dynamic model calculation results, and generate the corrected foundation pit deformation state data; The deformation prediction module is used to predict the future deformation trend of the foundation pit based on the corrected foundation pit deformation state data using the extended Kalman filter method; Visualization module, used to visualize the deformation state of the foundation pit in three dimensions based on BIM technology, and to display the key parameters of the foundation pit deformation in the digital model; The intelligent control module is used to optimize foundation pit support measures based on deformation prediction results and provide construction adjustment plans.

[0093] The data acquisition module is used to collect foundation pit deformation data through sensor units, including displacement sensors, inclination sensors, strain gauges, soil pressure gauges, temperature and humidity sensors, and piezometers. The sensor units are distributed in the foundation pit retaining structure, support system, and surrounding soil to monitor horizontal displacement, vertical displacement, inclination change, strain distribution, soil pressure fluctuation, groundwater level change, and environmental temperature and humidity in real time. In some implementations, the data acquisition module supports automatic sampling and time synchronization mechanisms to ensure the temporal consistency of multi-point measurement data.

[0094] The data transmission module is used to transmit the data acquired by the data acquisition module to the data processing module through the wireless sensor network (WSN) and the 5G communication module. Generally, the data transmission module supports two modes: short-distance wireless transmission and long-distance high-speed transmission. The wireless sensor network is responsible for the aggregation of short-distance sensor data and reduces the burden of long-distance communication, while the 5G communication module ensures low latency and high bandwidth for long-distance data transmission. As an option, the data transmission module supports data encryption and error correction functions to improve the security and reliability of data transmission. In the remote monitoring scenario, the data transmission module can be combined with the edge computing architecture to pre-process the data locally before uploading it to the central server to reduce bandwidth usage.

[0095] The data processing module is used to receive the data transmitted by the data transmission module, and perform format conversion, abnormal data removal, noise filtering and data completion on the data. The format conversion function ensures that the data collected by different types of sensors can be processed uniformly, such as converting analog signals into digital signals and standardizing multiple data formats. Statistical analysis methods are used to remove abnormal data, such as removing abnormal values ​​based on dynamic thresholds of historical data. Kalman filtering, adaptive mean filtering or wavelet transform denoising are used for noise filtering to reduce the impact of environmental interference. Data completion methods include interpolation algorithms, short-term prediction corrections, etc., to fill in the missing data collected in a short period of time and improve the continuity and integrity of the data.

[0096] The deformation modeling module is used to construct a nonlinear dynamic model of foundation pit deformation based on the data provided by the data processing module to describe the time evolution of foundation pit deformation. In general, this module is based on a fractional differential model to reflect the historical memory effect of foundation pit deformation, and combines influencing factors such as construction load and groundwater level changes to construct a state equation. In some implementations, this module supports parameter adaptive optimization, uses historical data to invert model parameters, and improves the prediction accuracy of the model.

[0097] The data fusion module is used to optimize the foundation pit deformation data using the variational data assimilation method (4D-Var), fuse the sensor measured data with the calculation results of the nonlinear dynamic model, and generate the corrected foundation pit deformation state data. During the data fusion process, the system adjusts the weights of different sensor data to increase the influence weight of data with higher measurement accuracy and reduce the influence of data with larger errors on the results. In some implementations, this module supports spatiotemporal correlation analysis, that is, based on the spatial location of the monitoring point and the historical deformation trend, the data fusion algorithm is optimized to make the fusion result more reasonable.

[0098] The deformation prediction module is used to predict the future deformation trend of the foundation pit based on the corrected foundation pit deformation state data using the extended Kalman filter (EKF) method. During the prediction process, the system calculates the possible deformation value in the future according to the current deformation state, and makes corrections based on the real-time measurement values ​​of the sensor to improve the prediction accuracy. As an option, this module can be combined with time series analysis methods such as the autoregressive moving average model (ARMA) or the long short-term memory neural network (LSTM) to make short-term corrections to the deformation trend and improve the prediction stability within a short time range.

[0099] The visualization module is used to visualize the deformation state of the foundation pit in three dimensions based on BIM technology, and to display the key parameters of the foundation pit deformation in the digital model. This module constructs a three-dimensional digital twin model of the foundation pit, and dynamically maps the real-time monitoring data to the model structure to provide an intuitive deformation trend analysis. In general, the visualization interface supports multi-level data display, including overall deformation trends, key monitoring point data, historical trend curves, etc. In some implementations, the module supports augmented reality (AR) or virtual reality (VR) interaction, and AR devices can be used at the construction site to view the real-time deformation of the foundation pit, thereby improving the visualization capabilities of on-site monitoring.

[0100] The intelligent control module is used to optimize the foundation pit support measures based on the deformation prediction results and provide construction adjustment plans. In general, the module uses the deep reinforcement learning (DRL) method to automatically learn the best foundation pit support strategy to reduce the risk of foundation pit deformation. Support optimization measures include but are not limited to adjusting support stiffness, optimizing dewatering rate, changing excavation sequence, increasing anchor prestressing, etc. As an option, the intelligent control module can be combined with an expert system to provide support optimization suggestions based on engineering experience, which will be reviewed and implemented by engineering personnel. In some embodiments, the module supports automated control and can directly send control instructions to the actuator, such as adjusting the anchor tension, optimizing the operating parameters of the dewatering equipment, etc., to achieve fully automatic deformation control and improve construction safety and management efficiency.

[0101] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A real-time monitoring method for foundation pit deformation based on BIM technology, characterized in that: The following steps are involved: Sensors are placed around and at the bottom of the foundation pit to collect displacement, inclination, strain, soil stress, temperature and humidity, and groundwater level data; the collected data are transmitted to the data processing unit, which performs noise filtering, outlier elimination, and data completion on the collected data; Based on the collected data obtained by the data processing unit, the predicted value of the nonlinear dynamic model of the foundation pit deformation is constructed; The variational data assimilation method is used to optimize the fusion of the predicted value of the nonlinear dynamic model and the data collected by the sensor to obtain the corrected foundation pit deformation state data; Based on the corrected foundation pit deformation state data, the extended Kalman filter method is used to predict and calculate the future foundation pit deformation state; Based on the prediction results of the extended Kalman filter method, the deep reinforcement learning method is used to optimize the foundation pit support measures and adjust the construction plan; Based on BIM technology, the deformation status data of the foundation pit is visualized, and the key parameters of the foundation pit deformation are displayed in real time in the three-dimensional digital model.

2. The real-time monitoring method for foundation pit deformation based on BIM technology according to claim 1 is characterized in that: The sensor layout is optimized using the optimal observation theory to calculate the optimal sensor layout plan to maximize the information entropy of the collected data to improve the integrity and spatial resolution of the data.

3. The real-time monitoring method for foundation pit deformation based on BIM technology according to claim 1 is characterized in that: The sensor network includes an adaptive wireless network and a 5G communication module, and the data processing unit uses an edge computing method to perform preliminary data processing, including data interpolation, denoising and abnormal data elimination.

4. The method for real-time monitoring of foundation pit deformation based on BIM technology according to claim 1 is characterized in that: The nonlinear dynamic model uses fractional derivative form to describe the historical memory effect of foundation pit deformation, and the dynamic equation is: Among them, x(t) is the pit deformation state vector, which represents the strain state variable of the pit at time t, A is the pit system stiffness matrix, B is the support control influence matrix, u(t) is the external influencing factor, f(x,t) is the nonlinear disturbance term, and w(t) is the system random noise. It is the description of foundation pit deformation based on fractional derivatives, and α is the fractional order.

5. The real-time monitoring method for foundation pit deformation based on BIM technology according to claim 1 is characterized in that: The collected data are optimized and fused using the variational data assimilation method, and the optimization objective function is: Among them, J(x) is the objective function, x is the foundation pit deformation state vector to be optimized, and x b is the background field data, B is the background error covariance matrix, H(x) is the observation operator, y is the sensor measured data, and R is the observation error covariance matrix.

6. The method for real-time monitoring of foundation pit deformation based on BIM technology according to claim 1 is characterized in that: The modified foundation pit deformation state is predicted by the extended Kalman filter method, and its state transfer equation is: x k+1 =A k x k +B k u k +w k ; The observation equation is: y k =H k x k +v k ; Among them, x k+1 is the state transfer equation, A k is the state transfer matrix, B k is the control input matrix, x k is the deformation state vector of the foundation pit at the current moment, w k and v k are system noise and measurement noise respectively, which obey Gaussian distribution, y k is the sensor observation data, H k is the observation matrix, u k For external input quantity.

7. The method for real-time monitoring of foundation pit deformation based on BIM technology according to claim 1 is characterized in that: The BIM technology constructs a three-dimensional foundation pit monitoring model based on digital twin technology, and displays the deformation status of the foundation pit in real time in a visual interface, including displacement, inclination, strain, soil stress, temperature and humidity, and groundwater level parameters.

8. The method for real-time monitoring of foundation pit deformation based on BIM technology according to claim 1 is characterized in that: The foundation pit support measures are optimized by a deep reinforcement learning method. The reinforcement learning model is constructed based on a Markov decision process, including a state space, an action space and a reward function. The reward function is defined as: R1=-∥x t -x safe ∥ 2 ; Among them, x t is the current deformation state of the foundation pit, x safe is the safety threshold, R1 is the reward value, ∥x t -x safe ∥ 2 is the Euclidean norm.

9. A real-time monitoring device for foundation pit deformation based on BIM technology, applied to a real-time monitoring method for foundation pit deformation based on BIM technology according to any one of claims 1 to 8, characterized in that: include: A sensor assembly for collecting physical parameters of foundation pit deformation; A data acquisition interface, used to receive data collected by the sensor assembly; 5G communication module, used to transmit the data acquired by the data acquisition component to the central processor; The central processing unit is used to filter noise, remove outliers, and complete data for the transmitted data; Fusion and prediction unit, used to optimize data fusion and calculate foundation pit deformation trends using variational data assimilation methods; Human-computer interaction interface, used to present a 3D visualization model of foundation pit deformation based on BIM technology and display deformation trend analysis results; The actuator interface is used to adjust the foundation pit support measures based on the foundation pit deformation trend and send control instructions to the actuator.

10. A real-time monitoring system for foundation pit deformation based on BIM technology, applied to a real-time monitoring method for foundation pit deformation based on BIM technology according to any one of claims 1 to 8, characterized in that: Includes the following modules: A data acquisition module, used for collecting foundation pit deformation data through a sensor unit; A data transmission module, used to transmit the data collected by the data acquisition module to the data processing module through the wireless sensor network and the 5G communication module; The data processing module is used to receive the data transmitted by the data transmission module, and perform format conversion, abnormal data removal, noise filtering and data completion on the data; A deformation modeling module is used to construct a nonlinear dynamic model of foundation pit deformation based on the data provided by the data processing module; The data fusion module is used to optimize the foundation pit deformation data by using the variational data assimilation method, fuse the sensor measured data and the dynamic model calculation results, and generate the corrected foundation pit deformation state data; The deformation prediction module is used to predict the future deformation trend of the foundation pit based on the corrected foundation pit deformation state data using the extended Kalman filter method; Visualization module, used to visualize the deformation state of the foundation pit in three dimensions based on BIM technology, and to display the key parameters of the foundation pit deformation in the digital model; The intelligent control module is used to optimize foundation pit support measures based on deformation prediction results and provide construction adjustment plans.

Images