Building engineering foundation pit monitoring system

Abstract

The invention relates to the field of constructional engineering safety monitoring, and discloses a constructional engineering foundation pit monitoring system, which comprises a data acquisition module used for acquiring monitoring data of a foundation pit; the data storage and preprocessing module is used for storing the foundation pit monitoring data and carrying out denoising and preprocessing; the dynamic threshold calculation module is used for calculating a dynamic threshold of the monitoring data; the optimal early warning decision module is used for solving an optimal early warning strategy based on a Hamiltonian-Jacobian-Bellman equation; the visualization and monitoring management module is used for displaying the real-time state of the foundation pit monitoring data; and the remote early warning and notification module is used for sending a remote notification to a construction party or a management unit when the early warning trigger condition is met. According to the method, the dynamic threshold value of the monitoring data is calculated through variational Bayesian reasoning, the early warning standard can be adjusted in a self-adaptive mode, the method adapts to the complex and changeable foundation pit environment, the early warning accuracy is effectively improved, and the problem that a traditional method cannot adapt to environment changes is solved.

Description

Technical Field

[0001] The present invention relates to the field of construction engineering safety monitoring, and in particular to a system for monitoring foundation pits of construction engineering. Background Art

[0002] During construction, foundation pit safety monitoring is crucial. Excavation can trigger soil stress redistribution, groundwater level fluctuations, and surrounding building settlement. Failure to provide timely monitoring and early warning can lead to structural instability and even accidents. Therefore, construction and management organizations require efficient, accurate, and real-time monitoring systems to monitor the safety status of foundation pits, identify potential risks in advance, and prevent irreversible losses.

[0003] Currently, foundation pit monitoring primarily relies on methods such as fixed threshold alarms, single-sensor measurements, and empirical statistical analysis. These technologies are simple to operate, low-cost, and quickly deployable in practical applications, making them suitable for general construction environments. Traditional monitoring systems can provide data on foundation pit settlement, tilt, stress, and groundwater levels, and issue alarms based on fixed early warning thresholds, enabling construction teams to intervene before abnormal conditions occur. Furthermore, some systems now feature wireless transmission and remote monitoring capabilities, improving information transmission efficiency and reducing reliance on manual inspections.

[0004] However, with the increasing complexity of the construction environment, the limitations of traditional monitoring methods have gradually become apparent. Fixed threshold warnings are too rigid and cannot adapt to the dynamic changes in foundation pit conditions, resulting in frequent false alarms and missed alarms, which affect construction decisions. Warning strategies rely on manual experience and fail to introduce scientific optimization methods, making it difficult to strike a balance between false alarms and missed alarms. Single sensors have limited monitoring coverage and are subject to significant environmental interference, resulting in insufficient data reliability and some risks being overlooked. Outdated information display methods often require manual data analysis by construction parties, resulting in time lags in warning responses and missed opportunities for intervention. Therefore, those skilled in the art have proposed a system for monitoring construction foundation pits to address these issues. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a system for monitoring foundation pits in construction projects, which solves the problems of rigid thresholds, backward early warning strategies, one-sided data, and inefficient information transmission in existing foundation pit monitoring technologies.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A system for monitoring a foundation pit of a construction project, comprising: A data acquisition module is used to collect monitoring data of the foundation pit and transmit the data to the data storage and preprocessing module via wireless transmission; The data storage and preprocessing module is connected to the data acquisition module and is used to store the foundation pit monitoring data, and transmit the monitoring data to the dynamic threshold calculation module after denoising and preprocessing; A dynamic threshold calculation module, connected to the data storage and preprocessing module, is used to calculate the dynamic threshold of the monitoring data based on variational Bayesian reasoning and transmit the calculated threshold to the optimal warning decision module; The optimal warning decision module is connected to the dynamic threshold calculation module and is used to solve the optimal warning strategy based on the Hamilton-Jacobi-Bellman equation and determine whether to trigger the warning signal based on the calculation results; The visualization and monitoring management module is connected to the optimal early warning decision module to display the real-time status of foundation pit monitoring data and visualize early warning information when the early warning is triggered; The remote warning and notification module is connected to the optimal warning decision module and is used to send remote notifications to the construction party or management unit when the warning trigger conditions are met.

[0007] Preferably, the data acquisition module includes: The sensor subsystem includes: GNSS high-precision positioning equipment, MEMS tilt sensors, fiber Bragg grating sensors, and piezometers, which are used to collect foundation pit settlement, tilt, stress, and groundwater level data; A wireless data transmission subsystem, including: LoRa technology or a multi-hop LoRa network, for transmitting data collected by the sensor subsystem; The data acquisition terminal is used to receive the data transmitted by the wireless data transmission subsystem and send it to the data storage and preprocessing module.

[0008] Preferably, the data storage and preprocessing module includes: A data storage unit, used for storing foundation pit monitoring data; A data denoising unit, used for denoising monitoring data; The data smoothing unit is used to smooth the monitoring data and then transmit it to the dynamic threshold calculation module.

[0009] Preferably, the dynamic threshold calculation module includes: Variational Bayesian inference threshold estimation unit, used to calculate the posterior distribution of foundation pit monitoring data; The adaptive threshold updating unit is used to update the dynamic threshold of the monitoring data in real time based on the posterior distribution and transmit it to the optimal warning decision module.

[0010] Preferably, the optimal early warning decision module includes: False alarm and omission loss calculation unit, used to calculate the cost of false alarm and omission; Hamilton equation solving unit, used to solve the optimal early warning strategy based on Hamilton-Jacobi-Bellman equation; The early warning trigger unit is used to determine whether to trigger the early warning signal according to the optimal early warning strategy, and transmit the early warning signal to the visualization and monitoring management module and the remote early warning and notification module.

[0011] Preferably, the visualization and monitoring management module includes: BIM+GIS visualization unit, used to display the foundation pit monitoring status based on building information model and geographic information system; The real-time monitoring dashboard unit is used to display the real-time changes of foundation pit monitoring data and update the visual interface when the early warning is triggered.

[0012] Preferably, the remote warning and notification module includes: An alarm sending unit, used to send an alarm when the monitoring data exceeds the threshold or meets the optimal early warning strategy; A construction party notification unit, used to send early warning information to the construction party; The management unit notification unit is used to send early warning information to the management unit.

[0013] Preferably, the adaptive threshold updating method of the adaptive threshold updating unit comprises the following steps: The threshold estimation unit sets the prior distribution of the initial threshold based on the variational Bayesian inference and calculates the posterior distribution of the monitoring data; Then calculate the approximate distribution of variational inference; Finally, the dynamic threshold is updated by minimizing the Kullback-Leibler divergence.

[0014] Preferably, the optimal early warning strategy calculation method of the Hamiltonian equation solving unit includes the following steps: Calculating the losses of false alarms and missed alarms based on the false alarm and missed alarm loss calculation unit; Then calculate the Hamiltonian function; And calculate the partial derivatives of the value function; Finally, the optimal early warning strategy is obtained by solving the Hamilton-Jacobi-Bellman equation.

[0015] Preferably, the warning triggering method of the warning triggering unit includes the following steps: Determine whether the monitoring data exceeds the dynamic threshold; Determine whether the calculated Hamiltonian function value meets the trigger condition; If the warning triggering conditions are met, a warning signal is sent; If the warning trigger conditions are not met, continue to monitor the data and update the optimal warning strategy The present invention provides a system for monitoring foundation pits in construction projects. It has the following beneficial effects: 1. The present invention adopts variational Bayesian reasoning to calculate the dynamic threshold of monitoring data, which can adaptively adjust the early warning standard to adapt to the complex and changeable foundation pit environment. Compared with the traditional fixed threshold monitoring method, this method can dynamically correct the risk of false alarms and missed alarms, effectively improve the accuracy of early warning, and solve the problem that traditional methods cannot adapt to environmental changes.

[0016] 2. The present invention combines the Hamilton-Jacobi-Bellman equation (HJB) to calculate the optimal early warning strategy, ensuring a balance between false alarms and missed alarms. Traditional early warning systems often use empirical thresholds or simple statistical methods, resulting in excessively high false alarm rates or serious missed alarms. The present invention reduces unnecessary alarms through dynamic optimization, while ensuring that real risks are not ignored, thereby enhancing the decision-making ability of the early warning system.

[0017] 3. The present invention integrates multiple monitoring devices such as GNSS high-precision positioning, MEMS tilt sensors, fiber Bragg grating sensors and piezometers, and transmits data through LoRa or multi-hop LoRa networks. Compared with the single sensor monitoring solution, the multi-sensor fusion technology of the present invention makes the monitoring data more comprehensive and has stronger anti-interference ability, effectively solving the problems of traditional single sensor monitoring being greatly affected by environmental interference and one-sided data, and ensuring the stability and reliability of the monitoring system.

[0018] 4. The present invention combines BIM+GIS visualization technology to enable monitoring data to be presented intuitively in the form of a real-time dashboard. Construction parties and management units can obtain risk information in a timely manner through a remote early warning and notification system. Compared with traditional manual inspections or offline data analysis, this system makes monitoring information more intuitive, reduces human misjudgment, improves management efficiency, and makes decision-making more scientific. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 Schematic diagram of the system architecture of the present invention; Figure 2 This is a schematic diagram of the data acquisition module architecture of the present invention; Figure 3 This is a schematic diagram of the data storage and preprocessing module architecture of the present invention; Figure 4 This is a schematic diagram of the dynamic threshold calculation module architecture of the present invention; Figure 5 This is a schematic diagram of the optimal early warning decision module architecture of the present invention; Figure 6 This is a schematic diagram of the visualization and monitoring management module architecture of the present invention; Figure 7 This is a schematic diagram of the remote warning and notification module architecture of the present invention. DETAILED DESCRIPTION

[0020] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. 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 making creative efforts are within the scope of protection of the present invention.

[0021] Please see the attached Figure 1 -Attached Figure 7 The embodiment of the present invention provides a system for monitoring a foundation pit of a construction project, comprising: A data acquisition module is used to collect monitoring data of the foundation pit and transmit the data to the data storage and preprocessing module via wireless transmission; Specifically, the data acquisition module undertakes the core task of obtaining foundation pit status information. Generally, this module needs to collect multiple key parameters of the foundation pit, including but not limited to settlement, inclination, stress and groundwater level. These parameters are important bases for foundation pit stability assessment. In order to ensure the accuracy and real-time nature of the data, the data acquisition module should have high-precision sensing capabilities and combine wireless transmission technology to transmit data stably and efficiently to the data storage and preprocessing module to ensure the reliability of subsequent calculations.

[0022] In one possible implementation, the data acquisition module is mainly composed of a sensor subsystem, a wireless data transmission subsystem and a data acquisition terminal. Specifically, the sensor subsystem is responsible for collecting physical quantities, the wireless data transmission subsystem is used for long-distance transmission of data, and the data acquisition terminal is responsible for receiving and organizing data for subsequent processing.

[0023] In this embodiment, the sensor subsystem includes multiple types of sensors, which are used to measure different foundation pit parameters.

[0024] As an option, settlement monitoring uses GNSS high-precision positioning equipment, which obtains elevation information of benchmark points based on the global navigation satellite system, with measurement accuracy up to millimeter level. In some embodiments, the GNSS equipment uses RTK (real-time kinematic) technology to correct errors through the base station to improve positioning accuracy.

[0025] In another possible implementation, tilt monitoring uses a MEMS inclination sensor, which calculates the inclination by measuring changes in gravitational acceleration. Generally, the resolution of the MEMS sensor can reach 0.001°, which can meet the accuracy requirements of foundation pit monitoring. In order to reduce the error caused by temperature drift, a temperature compensation algorithm can be used to perform corrections at different ambient temperatures to improve measurement stability.

[0026] Fiber Bragg grating (FBG) sensors or electrical strain gauges are commonly used to monitor stress in foundation pits. FBG sensors measure stress changes by shifting the wavelength of light waves, offering high sensitivity and immunity to electromagnetic interference. As an alternative, electrical strain gauges measure stress conditions in foundation pits by analyzing the relationship between strain and resistance changes. In certain embodiments, to improve the reliability of stress measurement, multi-point distributed measurement can be employed, deploying multiple sensors at different measurement points to obtain more comprehensive stress distribution data.

[0027] For groundwater level monitoring, piezometers or water level gauges are generally used. Piezometers measure pore water pressure, thereby indirectly estimating groundwater levels. If the foundation pit soil is relatively uniform, a single-point piezometer can be used. Otherwise, multiple piezometers are deployed to obtain groundwater level data at different depths.

[0028] In this embodiment, the wireless data transmission subsystem is responsible for transmitting the data collected by the sensor to the data collection terminal.

[0029] In one possible implementation, data transmission uses LoRa (long-range wireless communication) technology, which is characterized by low power consumption and long transmission distance, and is suitable for the complex environment of the construction site. In some embodiments, if the foundation pit monitoring points are widely distributed, a multi-hop LoRa network can be used, that is, the transmission range is extended through multiple relay nodes.

[0030] As another option, if the foundation pit is located in an urban area, NB-IoT (narrowband Internet of Things) technology can be used. Its advantages are wide signal coverage and strong penetration capability, which can ensure stable data transmission at underground monitoring points. In some embodiments, to reduce data loss, a breakpoint resume mechanism can be added to the NB-IoT terminal. That is, when the network is interrupted, the device can continue to upload unsent data after the connection is restored.

[0031] In high-real-time scenarios, such as subway foundation pit monitoring, it is recommended to use 5G transmission technology. The low latency characteristics of 5G can achieve millisecond-level data synchronization, ensuring the real-time nature of monitoring data. In some embodiments, in order to reduce data traffic consumption, edge computing can be combined to process part of the data locally and only upload key data to the server.

[0032] In this embodiment, the data acquisition terminal is used to receive wirelessly transmitted data and perform preliminary processing on the data.

[0033] Generally, data acquisition terminals use industrial-grade embedded systems, such as ARM-based microcontrollers or industrial computers. As a possible implementation method, the data terminal includes a data buffer module for storing monitoring data in a short period of time to prevent data loss due to sudden network interruptions.

[0034] In some embodiments, in order to improve the anti-interference ability of data transmission, a CRC (cyclic redundancy check) algorithm can be used to check the integrity of the data. Specifically, each transmitted data packet is accompanied by a check code, and the receiving end verifies whether the data is correct. If an error is found, retransmission can be requested to ensure the reliability of data transmission.

[0035] In some special scenarios, such as strong electromagnetic interference around foundation pits, wireless data transmission can be affected. In such cases, wired transmission solutions such as fiber optic communication or RS485 bus can be used. Fiber optic communication has strong anti-interference capabilities and is suitable for long-distance transmission. RS485 is suitable for short-distance wired communication, with stable transmission and low cost.

[0036] As a possible option, if there are many monitoring points, a distributed data acquisition architecture can be adopted, that is, multiple data acquisition terminals work together, each terminal is responsible for data collection in one area, and the data is aggregated to the central processing unit via wired or wireless means. In some embodiments, the data acquisition terminal can also be equipped with edge computing capabilities to pre-process the monitoring data, such as data compression and denoising, to reduce the amount of transmitted data and improve system efficiency.

[0037] The data storage and preprocessing module is connected to the data acquisition module and is used to store the foundation pit monitoring data, and transmit the monitoring data to the dynamic threshold calculation module after denoising and preprocessing; Specifically, the data storage and preprocessing module is responsible for the functions of data storage, noise reduction, outlier detection and data smoothing. Generally, the raw data obtained by the data acquisition module will contain noise, missing values ​​or abnormal points, which are directly used to analyze the accuracy of the monitoring results. Therefore, the main task of this module is to reasonably process the data so that it meets the requirements of subsequent calculations while ensuring the integrity and stability of the data.

[0038] In one possible implementation, the module consists of a data storage unit, a data denoising unit, and a data smoothing unit. Specifically, the data storage unit is used to store and manage foundation pit monitoring data, the data denoising unit is used to remove interference signals in the collected data, and the data smoothing unit is used to improve the continuity and consistency of the data to facilitate subsequent dynamic threshold calculation and early warning decision-making.

[0039] In this embodiment, the data storage unit is responsible for storing and managing monitoring data to support efficient query and analysis.

[0040] Generally speaking, monitoring data has time series characteristics, that is, the data is continuously generated in chronological order. In order to optimize storage performance, a time series database (TimescaleDB, InfluxDB, etc.) can be used. This type of database is specifically used to store and query high-frequency, time-series monitoring data. In some embodiments, in order to increase the query speed, an index structure can be established for the data, such as an index method based on a B+ tree or LSM tree, which greatly improves the query efficiency.

[0041] As an option, in order to reduce storage space, the monitoring data can be compressed and stored. In some embodiments, differential encoding can be used, that is, only the changes between adjacent time points are stored, rather than the complete data value. For example, if the settlement value at a certain time point is , then only store: ; in: Indicates the increment of settlement value, For the moment The settlement data value; For the moment The settlement data value.

[0042] This method can effectively reduce storage costs when data changes are relatively stable.

[0043] In another possible implementation, in order to improve the security of data storage, a distributed storage architecture can be adopted. Specifically, data can be stored in multiple server nodes at the same time to form replica storage to ensure that the data can still be recovered in the event of a single point failure. In some embodiments, to prevent data tampering, blockchain evidence technology can be used, that is, the hash value of key monitoring data is stored in the blockchain to ensure that the data cannot be tampered with, thereby improving the credibility of the monitoring data.

[0044] In this embodiment, the data denoising unit is used to remove interference signals in the monitoring data to improve the reliability of the data.

[0045] As a possible option, data denoising can be performed using wavelet transforms. Specifically, the signal can be decomposed into multiple frequency components using wavelet decomposition. High-frequency noise components are then removed before signal reconstruction. In some embodiments, the Daubechies wavelet (db4) can be used to optimize denoising, as it exhibits superior smoothness and boundary preservation.

[0046] In another possible implementation, Kalman filtering can be used for data denoising. Kalman filtering is a recursive estimation method suitable for real-time signal processing. Its state update equation is as follows: ; in: is the estimated value of the current state; For the moment The estimated value of the state; is the Kalman gain, which indicates the influence of new data on the estimated value; is the observed value; is the measurement matrix, which is used to convert the true state into the observed value.

[0047] In some embodiments, the Kalman filter parameters can be adjusted according to the characteristics of foundation pit monitoring data to make it applicable to different types of data such as settlement, inclination, and stress.

[0048] In this embodiment, the data smoothing unit is used to improve the continuity of the data so that the data is more consistent with physical laws.

[0049] In one possible implementation, Savitzky-Golay filtering can be used for smoothing. This method uses local polynomial fitting to reduce short-term fluctuations while maintaining the trend characteristics of the data. The calculation formula is as follows: ; in: is the smoothed data value; is the original data value; is the smoothing weight, calculated by the least squares method; is the window size, which determines the degree of smoothing.

[0050] In some embodiments, the window size can be adjusted dynamically based on the data characteristics. For example, if the data changes drastically (such as due to an earthquake), the window size can be increased appropriately. To obtain a more stable curve; if the data is relatively stable, you can reduce To retain more detailed information.

[0051] In another possible implementation, in order to avoid data distortion caused by over-smoothing, the exponentially weighted moving average (EWMA) method can be combined: ; in: is the smoothed data; is the current data point; is the smoothing factor, which usually ranges from ; For the previous moment smoothed data.

[0052] In some embodiments, if the fluctuation of foundation pit monitoring data is large, a lower value (such as 0.1) to retain more historical information; if the data trend is relatively stable, you can increase A value (such as 0.8) makes the smoothing result closer to the current data.

[0053] In some special cases, such as sensor failure or abnormal environmental interference, the data may have a large deviation. Therefore, an outlier detection mechanism may be added in this embodiment to ensure data quality.

[0054] As a possible option, a box plot method (BoxPlot) can be used to detect abnormal data. Specifically, the interquartile range (IQR) of the data is calculated: ; in: is the first quartile; is the third quartile; The interquartile range represents the middle 50% of the data distribution range and is used to detect abnormal data.

[0055] If a data point satisfy: or ; It is considered as an abnormal point.

[0056] A dynamic threshold calculation module, connected to the data storage and preprocessing module, is used to calculate the dynamic threshold of the monitoring data based on variational Bayesian reasoning and transmit the calculated threshold to the optimal warning decision module; Specifically, the dynamic threshold calculation module is used to adaptively adjust the warning threshold according to the characteristics of the monitoring data to improve the accuracy and flexibility of the warning. Generally, fixed thresholds are difficult to adapt to changes in different foundation pit environments and construction stages, which may lead to false alarms or missed alarms. Therefore, the core task of this module is to dynamically optimize the threshold through the variational Bayesian inference threshold estimation unit and the adaptive threshold update unit, combined with real-time monitoring data.

[0057] In one possible implementation, the dynamic threshold calculation module uses the Bayesian inference method to estimate the threshold and minimizes the Kullback-Leibler (KL) divergence through variational inference optimization, thereby adaptively adjusting the monitoring and warning threshold. Specifically, the module first sets the prior distribution of the initial threshold, then updates the posterior distribution based on the newly collected monitoring data, and uses variational inference to obtain an approximate distribution, and finally calculates the dynamic threshold.

[0058] In this embodiment, the variational Bayesian inference threshold estimation unit is responsible for constructing and updating the statistical distribution of the threshold to achieve dynamic estimation of the threshold.

[0059] Generally, setting a threshold The prior distribution of can be set as normal distribution: ; in: is the prior distribution; is the mean of the prior distribution, indicating the initial preset threshold; is the variance, which indicates the uncertainty about the initial threshold; is the set threshold.

[0060] In some embodiments, if the initial knowledge of the threshold is relatively vague, a generalized normal distribution or a gamma distribution may be selected as a priori to give the threshold a greater uncertainty, thereby enhancing the adaptive capability of the system.

[0061] In the new monitoring data After arrival, the posterior distribution can be calculated by Bayesian update: ; in: is the posterior distribution, which means that when the data is observed Post-parameter estimates; is the likelihood function, which means that at a given threshold The probability distribution of the following data; is the marginal distribution of the data, used for normalization.

[0062] As a possible option, the likelihood function can be modeled as a Gaussian distribution: ; in: For the Monitoring data; is the variance of the data; is the set threshold; is a Gaussian distribution; is the total number of monitoring data.

[0063] In some embodiments, in order to adapt to the characteristics of different construction stages, dynamic likelihood estimation can be used, that is, adaptive adjustment based on historical data , thereby optimizing the estimation accuracy of the threshold.

[0064] In this embodiment, the adaptive threshold updating unit is used to adjust the threshold according to the posterior distribution and optimize it through variational inference to improve computational efficiency.

[0065] In one possible implementation, in order to avoid directly calculating the complex posterior distribution, variational inference approximation can be used. Specifically, set the variational distribution As an approximation to the posterior distribution: ; in: is the variational distribution, which is an approximation of the posterior distribution; is the posterior distribution.

[0066] To optimize , using the Kullback-Leibler (KL) divergence to minimize: ; in: is the Kullback-Leibler (KL) divergence; is the logarithmic ratio, which measures the deviation of the variational distribution from the true posterior distribution; is the integration variable This optimization objective can be approximated by the variational lower bound (ELBO): ; in: is the variational lower bound; is the expected log-likelihood, which indicates the data under the current variational distribution possibility.

[0067] The first term is the likelihood expectation, which indicates the possibility of data under the current distribution; the second term is the KL divergence of the prior-variational distribution, which controls the deviation of the distribution.

[0068] In some embodiments, the variational distribution Gaussian distribution with optional mean-variance parameterization: ; in: is the mean of the variational distribution, which represents the optimized threshold; is the variance, which controls the uncertainty of the threshold; is a normal distribution.

[0069] In another possible implementation, in order to improve computational efficiency, the gradient descent method can be used to update the variational distribution parameters: ; ; in: is the number of iteration steps; is the learning rate; is the variational distribution in The mean parameter at the iteration; is the variational distribution in The standard deviation parameter at the iteration; The variation distribution Mean parameter after iterations; is the variational distribution in The standard deviation parameter after iterations; For ELBO The gradient of , which indicates how to adjust the mean parameter to increase the ELBO; For ELBO The gradient of , which indicates how to adjust the standard deviation parameter to increase the ELBO.

[0070] In some embodiments, an adaptive learning rate adjustment strategy, such as the Adam optimization algorithm, can be used to make parameter convergence more stable.

[0071] In practical applications, in order to ensure the stability of the dynamic threshold, a smooth update mechanism can be further introduced. That is, after each update, the new threshold and the historical threshold are weighted averaged: ; in: is the updated threshold; is the threshold value at the previous moment; is the smoothing factor, which usually ranges from ; is the mean of the variational distribution.

[0072] In some embodiments, in order to avoid severe fluctuations in the threshold, an exponentially weighted moving average (EWMA) method may be used to smooth the threshold update to improve the robustness of the system.

[0073] The optimal warning decision module is connected to the dynamic threshold calculation module and is used to solve the optimal warning strategy based on the Hamilton-Jacobi-Bellman equation and determine whether to trigger the warning signal based on the calculation results; Specifically, the core function of the optimal early warning decision module is to determine the optimal early warning strategy based on the real-time monitoring data provided by the dynamic threshold calculation module, combined with the calculation of false alarm and missed alarm losses, and then issue an early warning signal when the trigger conditions are met. Traditional early warning methods are generally limited by fixed thresholds and empirical rules, which can lead to false alarms or missed alarms. Therefore, this module, based on optimal control theory, establishes the Hamilton-Jacobi-Bellman (HJB) equation through a false alarm and missed alarm loss calculation unit, a Hamiltonian equation solver unit, and an early warning trigger unit. This calculates the optimal early warning strategy and ensures the scientific nature and stability of early warning decisions.

[0074] In one possible implementation, the false alarm and missed alarm loss calculation unit first quantifies the losses caused by false alarms (FalseAlarm) and missed alarms (MissedAlarm), and then constructs a Hamiltonian function; then, the Hamiltonian equation solving unit calculates the value function based on the dynamic programming method and solves the HJB equation to obtain the optimal early warning strategy; finally, the early warning triggering unit analyzes the monitoring data in real time and determines whether to trigger the early warning signal based on the calculated Hamiltonian function value.

[0075] In this embodiment, the false alarm and missed alarm loss calculation unit is used to perform quantitative analysis on different types of early warning errors to optimize the early warning strategy.

[0076] Generally speaking, false alarm losses and underreporting losses Modeling needs to be done based on the actual project situation. False positives can lead to unnecessary project downtime, waste of resources, and other economic losses, while missed positives can cause foundation pit instability and lead to serious accidents. Therefore, the following loss function can be used for modeling: ; ; in: Unit economic loss caused by false alarms; is the probability of a false alarm; Unit economic loss caused by omission; is the probability of underreporting; for misreporting losses; For underreported losses.

[0077] In some embodiments, the false positive and false negative probabilities can be calculated using Bayesian decision theory: ; ; in: Indicates that the real state is safe, Indicates that the system incorrectly issues an alert; Indicates a real state of danger, Indicates that the system failed to issue an early warning correctly.

[0078] In another possible implementation, historical monitoring data can be combined to use maximum likelihood estimation (MLE) or Markov chain Monte Carlo (MCMC) methods to estimate the probability distribution of false positives and false negatives, thereby optimizing the accuracy of loss calculation.

[0079] In this embodiment, the Hamiltonian equation solving unit is used to calculate the value function and obtain the optimal early warning strategy by solving the Hamilton-Jacobi-Bellman (HJB) equation.

[0080] In one possible implementation, the optimization of the early warning strategy can be solved by optimal control theory. Define the Hamiltonian function: ; in: is the Hamiltonian function; is the current monitoring status; is a control variable, indicating whether the warning is triggered; is the Lagrange multiplier; It is an instantaneous loss function, usually provided by the false positive and false negative loss calculation unit; is the state transfer equation, which represents the evolution law of the system state.

[0081] In some embodiments, the value function can be calculated using a partial differential equation solution method. : ; in: Value function over time The partial derivative of is the Hamiltonian function; is the current monitoring status; is a control variable, indicating whether the warning is triggered; The gradient of the value function; For all possible control variables Under the given value, the optimal decision that minimizes the Hamiltonian function is selected.

[0082] This equation, the Hamilton-Jacobi-Bellman (HJB) equation, is used to determine the optimal early warning strategy.

[0083] In another possible implementation, in order to improve computational efficiency, numerical methods (such as finite difference method and dynamic programming method) can be used to solve the HJB equation. For example, the finite difference method can be used to discretize the partial derivatives: ; in: is the value function; Discrete points of state variables; Adjacent state points of the state variable on the discretized grid; is a state variable The length of the discrete steps; is the Hamiltonian function; is a control variable, indicating whether the warning is triggered; For all possible control variables Under the value, choose the optimal decision that minimizes the Hamiltonian function; Finite difference approximation of the value function with respect to the state variables.

[0084] In some embodiments, in order to improve calculation accuracy, deep reinforcement learning (DRL) can be combined to train a neural network to approximate the optimal value function, thereby obtaining the optimal early warning strategy.

[0085] In this embodiment, the early warning triggering unit is used to monitor data in real time and determine whether to trigger an early warning signal based on the optimal early warning strategy.

[0086] In one possible implementation, the warning triggering process includes the following steps: 1. Determine whether the monitoring data exceeds the dynamic threshold: If the current monitoring data satisfy: ; in: For the current moment monitoring data; is the dynamic threshold.

[0087] Then go to the next step, otherwise continue monitoring.

[0088] 2. Calculate the Hamiltonian function value and determine whether the trigger condition is met: Calculate the Hamiltonian function value under the current state: ; in: is the current system status; is the control variable; is the Lagrange multiplier; is the Hamiltonian function; Instant loss function; is the state transfer equation.

[0089] like: ; in: To select the optimal strategy that minimizes the Hamiltonian function among all possible control variables; If it is the trigger threshold, go to the next step, otherwise continue monitoring.

[0090] 3. Trigger the early warning signal: If the early warning conditions are met, a warning signal is sent to the control system and the current status is recorded; if the early warning trigger conditions are not met, the data continues to be monitored.

[0091] In another possible implementation, in order to reduce the false alarm rate, adaptive threshold adjustment can be combined, that is: ; in: is the dynamic threshold; For the current moment monitoring data; To adjust the factors, the threshold value can be adjusted over time to improve the flexibility of early warning decision-making; is the threshold at the next time step The value of .

[0092] In some embodiments, in order to enhance the reliability of the warning signal, multi-sensor fusion can be used, that is, the weighted warning probability is calculated by integrating the data of multiple sensors: ; in: is the weighted warning probability; For the The warning probability of each sensor; is the weighting coefficient, which is dynamically adjusted according to the reliability of the sensor; is the total number of monitoring data.

[0093] The visualization and monitoring management module is connected to the optimal early warning decision module to display the real-time status of the foundation pit monitoring data and visualize the early warning information when the early warning is triggered; Specifically, the visualization and monitoring management module is used to intuitively display monitoring data and foundation pit status to support monitoring analysis and early warning response. Generally, foundation pit monitoring data includes multiple variables such as settlement, inclination, stress and groundwater level. Its temporal and spatial variation characteristics are relatively complex and difficult to be clearly presented through traditional tables or static graphics. Therefore, this module uses the BIM+GIS visualization unit and the real-time monitoring dashboard unit, combined with the building information model (BIM) and the geographic information system (GIS), to achieve three-dimensional dynamic visualization, while providing real-time data display and early warning information updates to improve the intuitiveness and efficiency of monitoring management.

[0094] In one possible implementation, the BIM+GIS visualization unit constructs the three-dimensional structure of the foundation pit based on the BIM model and uses GIS data to integrate surrounding environmental information to form a complete monitoring scenario. Subsequently, the real-time monitoring dashboard unit analyzes the sensor data and displays data changes in the form of dynamic charts, heat maps, etc., enabling monitoring personnel to quickly identify abnormal situations and make corresponding decisions when early warnings are triggered.

[0095] In this embodiment, the BIM+GIS visualization unit is used to integrate the building information model (BIM) and geographic information system (GIS) of the foundation pit to achieve three-dimensional visualization of the monitoring data.

[0096] Generally speaking, BIM can be used to accurately model the geometry, structural components and construction stage information of foundation pits, while GIS can provide data support for geographical environment, geological characteristics and external influencing factors. In order to achieve the integration of the two, IFC (Industry Foundation Classes) format can be used as the standard data format of BIM models, and GIS data can be stored in GeoJSON or Shapefile format, so as to display the building structure and geographical environment in the same system.

[0097] As an option, the three-dimensional visualization of the foundation pit can be rendered using Cesium.js or Unity3D. Specifically, Cesium.js is based on WebGL technology, which can efficiently load BIM+GIS data on the Web side and support dynamic interaction such as rotation, zooming and section analysis. In some embodiments, in order to improve rendering efficiency, LOD (Level of Detail) technology can be used, that is, automatically adjusting the detail level of the BIM model according to the viewing distance to optimize performance.

[0098] In another possible implementation method, BIM+GIS visualization can be combined with time-series monitoring data, that is, dynamic monitoring information can be superimposed on the three-dimensional model. For example, a heat map (Heatmap) or contour line (ContourLine) can be used on the surface of the foundation pit structure to display the distribution of data such as settlement and tilt, so that monitoring personnel can intuitively judge abnormal areas.

[0099] In some embodiments, to enhance data interactivity, a data query function may be integrated, allowing users to click on a specific area of ​​the foundation pit structure to obtain historical monitoring data for that location. Specifically, this can be achieved through the following query methods: ; in: is the spatial coordinate in the BIM model; For this location at time Monitoring data from the.

[0100] In this embodiment, the real-time monitoring dashboard unit is used to display the real-time changes in the foundation pit monitoring data and update the visual interface when the early warning is triggered.

[0101] In one possible implementation, the dashboard interface can be designed using ECharts.js or D3.js for visualization, supporting multiple data display methods, including: Time series curve: used to show the changing trend of monitoring parameters such as settlement, tilt, and stress. The time resolution can be adjusted dynamically. Bar charts and pie charts: used to show the status distribution of different monitoring points, such as the ranking of settlement at different measuring points; Geographic coordinate heat map: used to present the risk level of each monitoring point based on GIS data.

[0102] In another possible implementation, in order to enhance data readability, data aggregation and statistical analysis can be combined, that is, the mean, variance and number of abnormal points of the monitoring data can be automatically calculated. For example, in some embodiments, a sliding window method can be used to calculate the most recent The mean of the data points: ; in: is the sliding window mean; For the The data value of each time step; is the sliding window size.

[0103] In some embodiments, to improve the speed of early warning response, an anomaly detection algorithm can be integrated into the dashboard, such as detecting abnormal data based on the Z-score method: ; in: is the Z-score (standard score); is the standard deviation; is the mean of historical data; Indicates the monitoring data at the current moment.

[0104] when When an abnormal data point is detected, it can highlight the abnormal data point on the dashboard and trigger the update of the early warning interface.

[0105] When an alert is triggered, the dashboard needs to automatically update the interface, including: Warning sign highlight: If the foundation pit tilt exceeds the safe range, the warning symbol will flash on the visual interface; Risk level display: based on the risk assessment results provided by the optimal early warning decision module, the risk level color is updated, such as green (safe), yellow (warning), and red (dangerous); Notification push mechanism: Automatically send SMS or email to monitoring personnel to remind them of abnormal conditions.

[0106] In some embodiments, to improve data access efficiency, WebSocket technology can be used to achieve two-way real-time communication between the server and the front end, so that the dashboard data can be synchronized in real time.

[0107] In actual applications, in order to further improve the intelligent level of monitoring and management, an automatic report generation function can be integrated, that is, foundation pit monitoring and analysis reports are generated regularly and distributed to relevant personnel via email or cloud storage. The report content may include: Overview of monitoring data; Warning event records; Analysis of changing trends of monitoring parameters; Future risk forecast.

[0108] In some embodiments, risk prediction can be performed in conjunction with machine learning methods, such as a time series prediction model based on a long short-term memory network (LSTM): ; ; in: is the hidden state of the current time step; is the hidden state of the previous time step; Input for current monitoring data; , , is the model weight; , is the bias term; is the activation function; is the output value.

[0109] This method can predict future monitoring trends based on historical data, thereby identifying potential risks in advance.

[0110] The remote warning and notification module is connected to the optimal warning decision module and is used to send remote notifications to the construction party or management unit when the warning trigger conditions are met.

[0111] Specifically, the remote early warning and notification module is used to send early warning information to the construction party and the management unit when the monitoring data exceeds the threshold or meets the optimal early warning strategy. Generally, during the foundation pit monitoring process, when the monitoring data is abnormal or meets the early warning conditions, if the alarm is not issued to the relevant parties in time, it may cause delayed on-site response and increase safety risks. Therefore, this module ensures that the early warning information is conveyed in the shortest time through the alarm sending unit, the construction party notification unit and the management unit notification unit to support construction decision-making and emergency management.

[0112] In one possible implementation, the alarm sending unit is used to identify over-limit monitoring data and decide whether to trigger an alarm based on the calculation results of the optimal early warning decision module. Subsequently, the construction party notification unit and the management unit notification unit send notifications to the construction party and the management unit respectively according to the urgency of the alarm information. The notification methods include SMS, email, telephone alarm or mobile application push, etc., to ensure the timely delivery of the early warning information.

[0113] In this embodiment, the alarm sending unit is used to trigger the alarm mechanism when the monitoring data exceeds the threshold or meets the optimal early warning strategy, and transmit the alarm information to the notification unit.

[0114] Generally, the alarm triggering is based on a combination of fixed threshold rules and dynamic optimization strategies. The fixed threshold can be set by engineering safety standards, for example: or ; in: For the current moment monitoring data; is the upper threshold; is the lower threshold.

[0115] In some embodiments, if the rate of change of the monitoring data is too fast, that is, if: ; in: is the monitoring data of the previous moment; is the sampling time interval; is the rate threshold.

[0116] The alarm can be triggered in advance to prevent instability caused by drastic changes in a short period of time.

[0117] Alternatively, the alarm triggering can also be based on the calculation of the optimal warning strategy. Specifically, if the solution of the Hamilton-Jacobi-Bellman (HJB) equation calculated by the optimal control satisfies: ; in: is the Hamiltonian function value of the current state; is the trigger threshold.

[0118] The alarm sending unit immediately starts the early warning process and sends alarm information to the relevant notification units.

[0119] In this embodiment, the construction party notification unit is used to send early warning information to the construction party after the alarm is triggered so that the construction site can take timely response measures.

[0120] In one possible implementation, the unit uses a multi-channel notification mechanism to ensure reliable information delivery. Specifically, one or a combination of the following methods can be used: SMS notification: Use GSM SMS gateway to send alarm information to the construction manager; Email: Send detailed early warning reports to construction management personnel via SMTP server; Telephone alarm: Based on the automated voice system (IVR), the preset emergency contact number is dialed; Mobile app push: If the construction party uses a dedicated monitoring application, WebSocket or Firebase will be used to push alert messages.

[0121] In some embodiments, in order to distinguish the alarm levels, the notification unit may add the alarm type in the text message or application push, for example: General alert (low risk): monitoring is recommended, but no intervention is recommended; Warning Alert (Medium Risk): It is recommended to strengthen on-site inspections and adjust construction measures if necessary; Emergency Alert (High Risk): Stop construction immediately and activate the emergency plan.

[0122] As an option, the construction party notification unit can also be connected to the BIM+GIS visual system, that is, the monitoring point where the alarm occurs is directly highlighted in the construction party's monitoring interface, and the corresponding risk assessment data is provided to enhance readability.

[0123] In this embodiment, the management unit notification unit is used to send early warning information to the project management unit or government regulatory agency after the alarm is triggered, so as to facilitate superior decision-making and safety supervision.

[0124] In one possible implementation, this unit connects to a cloud-based early warning management system, allowing management units to access alert information via a web or mobile device and conduct remote control. Specifically, the management unit's notification unit can automatically generate early warning reports and deliver them to relevant management personnel via email or cloud storage sharing.

[0125] In some embodiments, in order to improve the response speed of the management unit, an intelligent notification classification mechanism can be adopted. Specifically: If the warning level is low, the report will only be sent by email; If the warning level is high, the responsible person will be notified by more direct means such as text messages or phone calls; If the monitoring data exceeds the limit continuously, a red alarm will be triggered and emergency plan suggestions will be pushed to the management platform.

[0126] In another possible implementation, the unit could be connected to a government emergency response system. For example, at a certain risk level, the system could automatically send an alert to municipal authorities or safety regulators to ensure that foundation pit safety issues receive regulatory attention.

[0127] As an option, the management unit notification unit can also integrate remote video monitoring. That is, when sending an alarm, it provides a real-time video stream to the management unit to facilitate remote judgment of the on-site situation. For example, if abnormal settlement occurs in the foundation pit, the system can automatically capture the video image at the corresponding time point and attach a video link in the notification content to improve decision-making efficiency.

[0128] While 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 these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A system for monitoring foundation pits in construction projects, characterized in that: include: A data acquisition module is used to collect monitoring data of the foundation pit and transmit the data to the data storage and preprocessing module via wireless transmission; The data storage and preprocessing module is connected to the data acquisition module and is used to store the foundation pit monitoring data, and transmit the monitoring data to the dynamic threshold calculation module after denoising and preprocessing; A dynamic threshold calculation module, connected to the data storage and preprocessing module, is used to calculate the dynamic threshold of the monitoring data based on variational Bayesian reasoning and transmit the calculated threshold to the optimal warning decision module; The optimal warning decision module is connected to the dynamic threshold calculation module and is used to solve the optimal warning strategy based on the Hamilton-Jacobi-Bellman equation and determine whether to trigger the warning signal based on the calculation results; The visualization and monitoring management module is connected to the optimal early warning decision module to display the real-time status of the foundation pit monitoring data and visualize the early warning information when the early warning is triggered; The remote warning and notification module is connected to the optimal warning decision module and is used to send remote notifications to the construction party or management unit when the warning trigger conditions are met.

2. A system for monitoring a construction pit according to claim 1, characterized in that: The data acquisition module includes: The sensor subsystem includes: GNSS high-precision positioning equipment, MEMS tilt sensors, fiber Bragg grating sensors, and piezometers, which are used to collect foundation pit settlement, tilt, stress, and groundwater level data; A wireless data transmission subsystem, including: LoRa technology or a multi-hop LoRa network, for transmitting data collected by the sensor subsystem; The data acquisition terminal is used to receive the data transmitted by the wireless data transmission subsystem and send it to the data storage and preprocessing module.

3. A system for monitoring a construction pit according to claim 1, characterized in that: The data storage and preprocessing module includes: A data storage unit, used for storing foundation pit monitoring data; A data denoising unit, used for denoising monitoring data; The data smoothing unit is used to smooth the monitoring data and then transmit it to the dynamic threshold calculation module.

4. A system for monitoring a construction pit according to claim 1, characterized in that: The dynamic threshold calculation module includes: Variational Bayesian inference threshold estimation unit, used to calculate the posterior distribution of foundation pit monitoring data; The adaptive threshold updating unit is used to update the dynamic threshold of the monitoring data in real time based on the posterior distribution and transmit it to the optimal warning decision module.

5. The system for monitoring a construction pit according to claim 1, characterized in that: The optimal early warning decision module includes: False alarm and omission loss calculation unit, used to calculate the cost of false alarm and omission; Hamilton equation solving unit, used to solve the optimal early warning strategy based on Hamilton-Jacobi-Bellman equation; The early warning trigger unit is used to determine whether to trigger the early warning signal according to the optimal early warning strategy, and transmit the early warning signal to the visualization and monitoring management module and the remote early warning and notification module.

6. A system for monitoring a construction pit according to claim 1, characterized in that: The visualization and monitoring management module includes: BIM+GIS visualization unit, used to display the foundation pit monitoring status based on building information model and geographic information system; The real-time monitoring dashboard unit is used to display the real-time changes of foundation pit monitoring data and update the visual interface when the early warning is triggered.

7. A system for monitoring a construction pit according to claim 1, characterized in that: The remote warning and notification module includes: An alarm sending unit, used to send an alarm when the monitoring data exceeds the threshold or meets the optimal early warning strategy; A construction party notification unit, used to send early warning information to the construction party; The management unit notification unit is used to send early warning information to the management unit.

8. A system for monitoring a construction pit according to claim 4, characterized in that: The adaptive threshold updating method of the adaptive threshold updating unit comprises the following steps: The threshold estimation unit sets the prior distribution of the initial threshold based on the variational Bayesian inference and calculates the posterior distribution of the monitoring data; Then calculate the approximate distribution of variational inference; Finally, the dynamic threshold is updated by minimizing the Kullback-Leibler divergence.

9. A system for monitoring a construction pit according to claim 5, characterized in that: The optimal early warning strategy calculation method of the Hamiltonian equation solving unit includes the following steps: Calculating the losses of false alarms and missed alarms based on the false alarm and missed alarm loss calculation unit; Then calculate the Hamiltonian function; And calculate the partial derivatives of the value function; Finally, the optimal early warning strategy is obtained by solving the Hamilton-Jacobi-Bellman equation.

10. A system for monitoring a construction pit according to claim 5, characterized in that: The early warning triggering method of the early warning triggering unit comprises the following steps: Determine whether the monitoring data exceeds the dynamic threshold; Determine whether the calculated Hamiltonian function value meets the trigger condition; If the warning triggering conditions are met, a warning signal is sent; If the warning trigger conditions are not met, continue to monitor the data and update the optimal warning strategy.

Images