Deep and large foundation pit deformation monitoring and early warning method based on multi-source heterogeneous data fusion

Abstract

The invention discloses a deep and large foundation pit deformation monitoring and early warning method based on multi-source heterogeneous data fusion, which adopts a buffer area analysis and improved Stacking integrated learning method to analyze a more comprehensive change trend of a foundation pit at an early warning part, reduce false alarms, improve accuracy and reliability of foundation pit early warning and better guide site construction. Meanwhile, multi-source heterogeneous data can be automatically fused and analyzed, the data processing time is shortened, and the automation and intelligence level of deep and large foundation pit monitoring is further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of foundation pit monitoring, and in particular to a deep and large foundation pit deformation monitoring and early warning method based on multi-source heterogeneous data fusion. Background Art

[0002] The development of underground space (such as rail transit stations, large underground pedestrian streets, underground shopping malls, underground parking lots, etc.) has gradually become a hot spot in my country's current construction. An important part of underground space construction is foundation pit construction. With the advancement of technology and construction technology, the scale and excavation depth of foundation pits are getting larger and larger. However, due to the large scale and depth of foundation pits and complex geological conditions, the foundation pits are prone to deformation. Therefore, it is necessary to monitor the foundation pit regularly and provide timely feedback to ensure construction safety.

[0003] In the prior art, CN117249803A discloses a deep foundation pit support column settlement monitoring and early warning method, which formulates a column settlement monitoring plan, implements monitoring and obtains monitoring data, pre-processes monitoring data, analyzes monitoring data and realizes early warning effects, but this method only uses a single sensor to monitor the foundation pit, the data source is single, the credibility is not high, and it cannot reflect the local and comprehensive change trend of the foundation pit. CN118581930A discloses a building deep foundation pit remote monitoring method and processing system, which transmits multi-source sensors to the cloud through the network for display and analysis, and performs message recording and early warning management based on the results of data analysis. Although this method integrates multi-source sensor data, there is no association between multi-source heterogeneous data, and only data collection is performed, and early warning is performed with sensors as units. The uncertainty of information is high, the coverage is effective, and the credibility of the early warning needs to be further improved. Summary of the invention

[0004] The present application provides a deep and large foundation pit deformation monitoring and early warning method based on multi-source heterogeneous data fusion to solve the current problems of high false alarm rate and low data credibility of a single sensor in deep and large foundation pit monitoring.

[0005] According to the first aspect, an embodiment provides a method for monitoring and early warning of deformation of a deep and large foundation pit based on multi-source heterogeneous data fusion, the method comprising:

[0006] Deploy monitoring points and various sensors at the foundation pit site for monitoring;

[0007] When the monitoring data of a certain sensor triggers an early warning, a buffer area is established with the sensor that triggers the early warning as the center, and the monitoring data of all sensors in the buffer area at the early warning time and historical time are obtained and pre-processed;

[0008] The improved Stacking ensemble learning model is trained using the historical monitoring data of various sensors obtained through preprocessing;

[0009] The trained Stacking ensemble learning model is used to predict the monitoring data of various sensors obtained at the current warning moment to obtain the warning classification results. Based on the model prediction results, reasonable suggestions and follow-up measures are put forward to ensure the safety of the foundation pit.

[0010] Furthermore, monitoring points and multiple types of sensors are deployed at the foundation pit site for monitoring, including:

[0011] Before deployment, each category of sensors is numbered according to category, and the sensor coordinates are collected when the sensors are buried, including plane coordinates and elevation information.

[0012] Furthermore, monitoring points and multiple types of sensors are deployed at the foundation pit site for monitoring, including:

[0013] Various types of sensors include automatic inclinometers, wall vertical and horizontal displacement meters, support axial force meters, surface settlement observation meters, and water level meters.

[0014] Furthermore, when the monitoring data of a certain sensor triggers an early warning, a buffer area is established with the sensor that triggers the early warning as the center, specifically including:

[0015] According to the size and depth of the foundation pit, a spherical buffer area is established with the sensor that triggers the early warning as the center, and the radius of the buffer area is the depth of the foundation pit.

[0016] Furthermore, the monitoring data of all sensors in the buffer area at the warning time and historical time are obtained and preprocessed, specifically including:

[0017] Obtain the warning time data of each type of sensor in the buffer area, as well as the historical data of N-1 periods before the warning time and the corresponding monitoring conclusions;

[0018] For the missing data, the polynomial fitting method is used to interpolate so that the time of each type of sensor data remains unified, and each type of sensor data is normalized to obtain the preprocessed data set:

[0019]

[0020] Where N means there are N periods of data, the Nth period is the current warning moment data, and the rest are past historical data. M means there are M types of sensors. Represents the data of the mth sensor in the nth period, It indicates the monitoring conclusion corresponding to the nth period of the mth type sensor.

[0021] Furthermore, the improved Stacking ensemble learning model is trained using the historical monitoring data of various sensors obtained through preprocessing, including:

[0022] The Stacking integrated learning model includes a multi-source data feature extraction module and a multi-source data feature fusion and classification module;

[0023] The multi-source data feature extraction module is used to extract sub-features of each type of sensor data through multiple parallel sub-feature extractors, and perform early warning classification prediction on each sub-feature through a classifier;

[0024] The multi-source data feature fusion and classification module is used to fuse the sub-feature warning classification results of each type of sensor data to obtain the final warning classification result.

[0025] Furthermore, the multi-source data feature extraction module specifically includes:

[0026] Each sub-feature extractor includes multiple different sub-models, which can extract historical data sets of various sensors. Input to the corresponding sub-feature extractor for K-fold cross validation, where N represents N periods of data, D p represents the p-th type of sensor data, represents the data of the nth period of the pth sensor, Indicates the monitoring conclusion corresponding to the nth period of the pth type sensor;

[0027] Each sub-model of each sub-feature extractor outputs prediction results and probabilities respectively, and the prediction results and probabilities of each sub-model are combined to obtain the sub-feature D ， 1;

[0028] Each sub-feature D ， 1 Input into the classifier to obtain the warning classification results and corresponding probability sets:

[0029]

[0030] Where G represents the type of classification results. Indicates the classification result of the P-th sensor into the g-th category, It represents the probability that the P-th sensor is classified into the g-th category.

[0031] Furthermore, each sub-feature extractor includes a long short-term memory network sub-model, a random forest sub-model and a support vector machine sub-model.

[0032] Furthermore, the multi-source data feature fusion and classification module specifically includes:

[0033] The average value of the warning classification results and corresponding probabilities of each parallel sub-feature extractor As the final prediction result, G represents the type of classification result and M represents the category of the sensor.

[0034] According to the second aspect, an embodiment provides a deep and large foundation pit deformation monitoring and early warning system based on multi-source heterogeneous data fusion, the system comprising:

[0035] The sensor deployment module is used to deploy monitoring points and multiple types of sensors at the foundation pit site for monitoring;

[0036] A monitoring data acquisition module is used to establish a buffer area with the sensor that triggers the warning as the center when the monitoring data of a certain sensor triggers an early warning, obtain the monitoring data of all sensors in the buffer area at the early warning time and historical time, and perform preprocessing;

[0037] Model training module, used to train the improved Stacking ensemble learning model using the historical monitoring data of various sensors obtained through preprocessing;

[0038] The model prediction module is used to use the trained Stacking ensemble learning model to predict the various sensor monitoring data obtained at the current warning moment to obtain the warning classification results, and to put forward reasonable suggestions and follow-up measures based on the model prediction results to ensure the safety of the foundation pit.

[0039] The present application provides a deep and large foundation pit deformation monitoring and early warning method based on multi-source heterogeneous data fusion. The method adopts buffer zone analysis to quickly obtain the monitoring point data of the early warning part. At the same time, combined with improved Stacking integrated learning, the fusion processing and intelligent analysis of multi-source heterogeneous data are performed to further improve the accuracy and stability of the early warning of the present invention. Specifically, it has the following beneficial effects:

[0040] 1. Buffer zone analysis and multi-source sensor fusion are used for comprehensive early warning analysis, which improves the accuracy and reliability of early warning and ensures the safety of foundation pit;

[0041] 2. Through the improved Stacking ensemble learning algorithm, the accuracy and intelligence level of multi-source sensor data processing are further improved, and the monitoring efficiency is improved BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A flowchart of a method for monitoring and early warning of deformation of a deep and large foundation pit based on multi-source heterogeneous data fusion provided by an embodiment of the present invention;

[0043] Figure 2 A diagram of the architecture of a Stacking integrated learning model in a deep and large foundation pit deformation monitoring and early warning method based on multi-source heterogeneous data fusion provided by an embodiment of the present invention;

[0044] Figure 3 A schematic diagram of the logical structure of a deep and large foundation pit deformation monitoring and early warning system based on multi-source heterogeneous data fusion provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The present invention is further described in detail below by specific embodiments in conjunction with the accompanying drawings. Wherein similar elements in different embodiments adopt associated similar element numbers. In the following embodiments, many detailed descriptions are for making the present application better understood. However, those skilled in the art can easily recognize that some features can be omitted in different situations, or can be replaced by other elements, materials, methods. In some cases, some operations related to the present application are not shown or described in the specification, this is to avoid the core part of the present application being overwhelmed by too much description, and for those skilled in the art, it is not necessary to describe these related operations in detail, and they can fully understand the related operations according to the description in the specification and the general technical knowledge in the art.

[0046] In addition, the features, operations or characteristics described in the specification can be combined in any appropriate manner to form various implementations. At the same time, the steps or actions in the method description can also be interchanged or adjusted in a manner that is obvious to those skilled in the art. Therefore, the various sequences in the specification and the drawings are only for the purpose of clearly describing a certain embodiment and are not meant to be a required sequence, unless otherwise specified that a certain sequence must be followed.

[0047] The first embodiment of the present invention provides a deep foundation pit deformation monitoring and early warning method based on multi-source heterogeneous data fusion. Figure 1 Provide detailed explanation.

[0048] like Figure 1 As shown, in step S100, monitoring points and multiple types of sensors are deployed at the foundation pit site for monitoring.

[0049] Specifically, bury sensors on site and establish a monitoring database. According to the design drawings and specifications, monitoring points and various sensors are arranged around the foundation pit, such as automatic inclinometers, wall vertical and horizontal displacement observation prisms, axial force meters, surface settlement observation points, water level meters, etc. Before deployment, each category of sensors is numbered according to the category, and the sensor coordinates are collected when the sensors are buried, including plane coordinates and approximate elevation information. The elevation information on the ground is at a uniform elevation, and the actual elevation inside the foundation pit is used, accurate to meters, and the plane coordinates are the actual measured coordinates. Specific examples are as follows:

[0050] (1) Deep horizontal displacement monitoring uses an ABS inclinometer tube embedded in the retaining wall or soil, and uses an automated inclinometer to observe the horizontal displacement at each depth. The monitoring point can be expressed as (TST001, x, y, h), where TST represents the abbreviation of deep horizontal displacement, 001 represents the serial number, and x, y, h represent the location information of the measuring point.

[0051] (2) The spacing between monitoring points for horizontal and vertical displacement of supporting piles (walls) is 10m, and monitoring points are set in the depth change area, middle part, positive corner part, and important environmental parts of adjacent buildings. The monitoring points can be expressed as (ZQC001, x, y, h) and (ZQS001, x, y, h), where ZQC and ZQS represent the abbreviations of the vertical displacement monitoring point and the horizontal displacement monitoring point of the wall, respectively, 001 represents the serial number, and x, y, h represent the location information of the measuring point.

[0052] (3) The supporting axial force is arranged in the middle of each side of the foundation pit, the depth change part, the part with large water and soil pressure, the part with complex stress conditions, etc. The monitoring point can be expressed as (ZCL001, x, y, h), where ZCL represents the abbreviation of supporting axial force, 001 represents the serial number; x, y, h represent the location information of the measuring point.

[0053] (4) Surface settlement observation: There should be no less than 2 rows of surface settlement monitoring points arranged parallel to the perimeter of the foundation pit, and the row spacing is 5m. The first row of monitoring points should be no more than 2m away from the edge of the foundation pit, and the spacing between each row of monitoring points should be 10m. The monitoring point can be expressed as (DBC001, x, y, h), where DBC represents the abbreviation of the surface settlement observation point, 001 represents the serial number, and x, y, h represent the location information of the measuring point.

[0054] (5) Water level observation points are arranged in the precipitation area and the influence area. The number of observation holes should meet the requirements of mastering the dynamic changes of groundwater levels in the precipitation area and the influence area. The monitoring point can be expressed as (DSW001, x, y, h), where DSW represents the abbreviation of the water level observation point, 001 represents the serial number, and x, y, h represent the location information of the measurement point.

[0055] like Figure 1 As shown, in step S200, when the monitoring data of a certain sensor triggers an early warning, a buffer area is established with the sensor that triggers the early warning as the center, and the monitoring data of all sensors in the buffer area at the early warning time and historical time are obtained and preprocessed.

[0056] Specifically, after receiving an early warning of a certain monitoring data, a buffer area is established with the sensor as the center based on the size and depth of the foundation pit. The radius of the buffer zone is the depth of the foundation pit, and the data of all sensors in the buffer area are obtained. The data obtained this time include deep horizontal displacement, wall vertical displacement, wall horizontal displacement, support axial force, surface settlement, and groundwater level data, a total of 6 categories, which are summarized in the data center.

[0057] Obtain the warning time data of each type of sensor in the area, as well as the historical data of the 100 periods before the time and the corresponding monitoring conclusions. Preprocess the data of each type of sensor:

[0058] For missing data, Newton polynomial fitting is optionally used to interpolate so that the sensor times remain uniform:

[0059] f(x)=b1+b2(x-x0)+…+b n (x-x0)(x-x1)…(x n-1 )

[0060] in:

[0061] b1=f(x0)

[0062]

[0063] n represents the number of monitoring data periods.

[0064] And normalize the data of each type of sensor, the formula is as follows:

[0065]

[0066] Get the preprocessed dataset in Shows the data of the first period of the first type of sensor, Indicates the monitoring conclusion corresponding to the first period of the first type of sensor.

[0067] like Figure 1 As shown, in step S300, the improved Stacking ensemble learning model is trained using the historical monitoring data of various sensors obtained through preprocessing.

[0068] In this embodiment, Figure 2 As shown, the improved Stacking ensemble learning model includes a first part: a multi-source data feature extraction module and a second part: a multi-source data feature fusion and classification module.

[0069] The multi-source data feature extraction module consists of 6 parallel sub-feature extractors. Since the data dimensions of each type of sensor are not uniform and the health status of the foundation pit reflected is also different, the sub-feature extractor can fully learn the data features of each type of sensor, so as to more comprehensively reflect the health status of the foundation pit.

[0070] The sub-feature extractor is composed of three common machine learning methods. Optionally, the present invention adopts long short-term memory network, random forest and support vector machine.

[0071] The steps are as follows: Perform 5-fold cross validation, where D1 represents the first type of sensor. That is, randomly divide the training set D1 into 5 equal parts, and use 1 of them as the 5-fold test set for each sub-feature extraction part, and the remaining 4 parts as the training set. Repeat this training step to obtain the prediction conclusion and corresponding probability of each sub-feature extraction part, and superimpose the features to obtain the corresponding probability to the sub-feature D ， 1.

[0072] Then, the sub-feature D ， 1 is input into the classifier. Optionally, the classifier uses a simple multi-layer perceptron (MLP) to obtain the classification result of a single sensor and the corresponding probability set The classification results are divided into 5 categories, mainly divided into no warning, yellow warning, orange warning and red warning. Indicates that the first type of sensor is classified as no warning, Indicates the probability that the first type of sensor is classified as no warning. For example, if the control value of the surface settlement observation point is ±40mm, then its five categories of boundaries are no warning: data ≤ ±28mm; yellow warning: ±28mm≤data≤±32mm; orange warning: ±32mm≤data≤±36mm; red warning: data ≥36mm.

[0073] The multi-source data feature fusion and classification module is mainly used to efficiently fuse and robustly classify the multi-model features extracted in the first part. In order to reduce the impact of single sensor noise and machine learning models on the entire classification result and improve the robustness of the classification result, the present invention uses the average value of the classification results and corresponding probabilities of each parallel sub-feature extractor as the final result. g=1,2,3,4,5}.

[0074] like Figure 1 As shown, in step S400, the trained Stacking ensemble learning model is used to predict the various sensor monitoring data acquired at the current warning moment to obtain a warning classification result, and reasonable suggestions and subsequent measures are put forward based on the model prediction results to ensure the safety of the foundation pit.

[0075] Specifically, after training, the data of this warning is input into the improved Stacking ensemble learning method to obtain the classification results and probabilities. Combined with the model classification results, reasonable suggestions and subsequent measures are proposed.

[0076] This embodiment provides a deformation monitoring and early warning method for deep and large foundation pits based on the fusion of multi-source heterogeneous data. It adopts buffer analysis and improved Stacking ensemble learning method to analyze the more comprehensive change trend of the foundation pit at the warning location, reduce false alarms, improve the accuracy and reliability of foundation pit early warning, and better guide on-site construction. At the same time, the method can automatically fuse and analyze multi-source heterogeneous data, reduce data processing time, and further improve the automation and intelligence level of deep and large foundation pit monitoring.

[0077] Corresponding to the above-disclosed method for monitoring and early warning of deep and large foundation pit deformation based on multi-source heterogeneous data fusion, the embodiment of the present invention further discloses a system for monitoring and early warning of deep and large foundation pit deformation based on multi-source heterogeneous data fusion, such as Figure 3 As shown, it specifically includes:

[0078] The sensor deployment module is used to deploy monitoring points and multiple types of sensors at the foundation pit site for monitoring;

[0079] A monitoring data acquisition module is used to establish a buffer area with the sensor that triggers the warning as the center when the monitoring data of a certain sensor triggers an early warning, obtain the monitoring data of all sensors in the buffer area at the early warning time and historical time, and perform preprocessing;

[0080] Model training module, used to train the improved Stacking ensemble learning model using the historical monitoring data of various sensors obtained through preprocessing;

[0081] The model prediction module is used to use the trained Stacking ensemble learning model to predict the various sensor monitoring data obtained at the current warning moment to obtain the warning classification results, and to put forward reasonable suggestions and follow-up measures based on the model prediction results to ensure the safety of the foundation pit.

[0082] It should be noted that, for the detailed description of a deep and large foundation pit deformation monitoring and early warning system based on multi-source heterogeneous data fusion provided in an embodiment of the present invention, reference can be made to the relevant description of a deep and large foundation pit deformation monitoring and early warning method based on multi-source heterogeneous data fusion provided in an embodiment of the present application, which will not be repeated here.

[0083] Those skilled in the art will appreciate that all or part of the functions of the various methods in the above-mentioned embodiments can be implemented by hardware or by computer programs. When all or part of the functions in the above-mentioned embodiments are implemented by computer programs, the program can be stored in a computer-readable storage medium, and the storage medium can include: read-only memory, random access memory, disk, optical disk, hard disk, etc., and the program is executed by a computer to implement the above-mentioned functions. For example, the program is stored in the memory of the device, and when the program in the memory is executed by the processor, all or part of the above-mentioned functions can be implemented. In addition, when all or part of the functions in the above-mentioned embodiments are implemented by computer programs, the program can also be stored in a storage medium such as a server, another computer, disk, optical disk, flash disk or mobile hard disk, and can be downloaded or copied and saved in the memory of the local device, or the system of the local device is updated, and when the program in the memory is executed by the processor, all or part of the functions in the above-mentioned embodiments can be implemented.

[0084] The above specific examples are used to illustrate the present invention, which is only used to help understand the present invention and is not intended to limit the present invention. For those skilled in the art, according to the concept of the present invention, some simple deductions, modifications or substitutions can be made.

Claims

1. A method for monitoring and early warning of deep and large foundation pit deformation based on multi-source heterogeneous data fusion, characterized in that: The method comprises: Deploy monitoring points and various sensors at the foundation pit site for monitoring; When the monitoring data of a certain sensor triggers an early warning, a buffer area is established with the sensor that triggers the early warning as the center, and the monitoring data of all sensors in the buffer area at the early warning time and historical time are obtained and pre-processed; The improved Stacking ensemble learning model is trained using the historical monitoring data of various sensors obtained through preprocessing; The trained Stacking ensemble learning model is used to predict the monitoring data of various sensors obtained at the current warning moment to obtain the warning classification results. Based on the model prediction results, reasonable suggestions and follow-up measures are put forward to ensure the safety of the foundation pit.

2. The method for monitoring and early warning of deep and large foundation pit deformation based on multi-source heterogeneous data fusion according to claim 1, characterized in that: Monitoring points and various sensors are deployed at the foundation pit site for monitoring, including: Before deployment, each category of sensors is numbered according to category, and the sensor coordinates are collected when the sensors are buried, including plane coordinates and elevation information.

3. The method for monitoring and early warning of deep and large foundation pit deformation based on multi-source heterogeneous data fusion according to claim 1, characterized in that: Monitoring points and various sensors are deployed at the foundation pit site for monitoring, including: Various types of sensors include automatic inclinometers, wall vertical and horizontal displacement meters, support axial force meters, surface settlement observation meters, and water level meters.

4. The method for monitoring and early warning of deep and large foundation pit deformation based on multi-source heterogeneous data fusion according to claim 1, characterized in that: When the monitoring data of a certain sensor triggers an early warning, a buffer area is established with the sensor that triggers the early warning as the center, specifically including: According to the size and depth of the foundation pit, a spherical buffer area is established with the sensor that triggers the early warning as the center, and the radius of the buffer area is the depth of the foundation pit.

5. The method for monitoring and early warning of deep and large foundation pit deformation based on multi-source heterogeneous data fusion according to claim 1, characterized in that: Acquire the monitoring data of all sensors in the buffer area at the warning time and historical time and perform preprocessing, specifically including: Obtain the warning time data of each type of sensor in the buffer area, as well as the historical data of N-1 periods before the warning time and the corresponding monitoring conclusions; For the missing data, the polynomial fitting method is used to interpolate so that the time of each type of sensor data remains unified, and each type of sensor data is normalized to obtain the preprocessed data set: Where N means there are N periods of data, the Nth period is the current warning moment data, and the rest are past historical data. M means there are M types of sensors. Represents the data of the mth sensor in the nth period, It indicates the monitoring conclusion corresponding to the nth period of the mth type sensor.

6. The method for monitoring and early warning of deep and large foundation pit deformation based on multi-source heterogeneous data fusion according to claim 1, characterized in that: The improved Stacking ensemble learning model is trained using the historical monitoring data of various sensors obtained through preprocessing, including: The Stacking integrated learning model includes a multi-source data feature extraction module and a multi-source data feature fusion and classification module; The multi-source data feature extraction module is used to extract sub-features of each type of sensor data through multiple parallel sub-feature extractors, and perform early warning classification prediction on each sub-feature through a classifier; The multi-source data feature fusion and classification module is used to fuse the sub-feature warning classification results of each type of sensor data to obtain the final warning classification result.

7. The method for monitoring and early warning of deep and large foundation pit deformation based on multi-source heterogeneous data fusion according to claim 6, characterized in that: The multi-source data feature extraction module specifically includes: Each sub-feature extractor includes multiple different sub-models, which can extract historical data sets of various sensors. Input to the corresponding sub-feature extractor for K-fold cross validation, where N represents N periods of data, D p represents the p-th type of sensor data, represents the data of the nth period of the pth sensor, Indicates the monitoring conclusion corresponding to the nth period of the pth type sensor; Each sub-model of each sub-feature extractor outputs prediction results and probabilities respectively, and the prediction results and probabilities of each sub-model are combined to obtain the sub-feature D ， 1; Each sub-feature D ， 1 Input into the classifier to obtain the warning classification results and corresponding probability sets: Where G represents the type of classification results. Indicates the classification result of the P-th sensor into the g-th category, It represents the probability that the P-th sensor is classified into the g-th category.

8. The method for monitoring and early warning of deep and large foundation pit deformation based on multi-source heterogeneous data fusion according to claim 7, characterized in that: Each sub-feature extractor includes a long short-term memory network sub-model, a random forest sub-model, and a support vector machine sub-model.

9. The method for monitoring and early warning of deep and large foundation pit deformation based on multi-source heterogeneous data fusion according to claim 7, characterized in that: The multi-source data feature fusion and classification module specifically includes: The average value of the warning classification results and corresponding probabilities of each parallel sub-feature extractor As the final prediction result, G represents the type of classification result and M represents the category of the sensor.

10. A deep and large foundation pit deformation monitoring and early warning system based on multi-source heterogeneous data fusion, characterized in that: The system comprises: The sensor deployment module is used to deploy monitoring points and multiple types of sensors at the foundation pit site for monitoring; A monitoring data acquisition module is used to establish a buffer area with the sensor that triggers the warning as the center when the monitoring data of a certain sensor triggers an early warning, obtain the monitoring data of all sensors in the buffer area at the early warning time and historical time, and perform preprocessing; Model training module, used to train the improved Stacking ensemble learning model using the historical monitoring data of various sensors obtained through preprocessing; The model prediction module is used to use the trained Stacking ensemble learning model to predict the various sensor monitoring data obtained at the current warning moment to obtain the warning classification results, and to put forward reasonable suggestions and follow-up measures based on the model prediction results to ensure the safety of the foundation pit.

Images