Large hydro-junction sluice deep foundation pit construction safety control method and system

Abstract

The invention provides a safety control method and system for construction of a deep foundation pit of a water drain gate of a large-scale hydro-junction, and is suitable for the field of construction of deep foundation pits. Comprising the steps of formulating a large hydro-junction water drain gate deep foundation pit construction monitoring scheme, arranging monitoring point positions, obtaining monitoring data, grouping the monitoring point positions, processing the monitoring data, constructing a prediction model, performing prediction model training and optimization adjustment, and obtaining a prediction value and performing early warning. According to the method, the types of the monitoring points are considered in a grouping mode, different data processing methods are adopted in a targeted mode, more accurate predicted values are obtained, the response capacity of the whole monitoring system is improved, and the method can be widely applied to the field of deep foundation pit construction.

Description

Technical Field

[0001] The invention relates to a safety control method and system for deep foundation pit construction of a sluice gate of a large-scale water conservancy hub, and is applicable to the field of deep foundation pit construction. Background Art

[0002] With the continuous development of large-scale water conservancy projects, especially in the construction of water conservancy hubs and sluice gates, the safety of deep foundation pit construction has received increasing attention. Deep foundation pit construction involves complex geological conditions and dynamic hydrological environment. Therefore, various safety hazards are prone to occur during the construction process, such as foundation pit deformation, soil sliding, water level abnormality, etc. If effective monitoring and control are not carried out, serious accidents may occur, and huge economic losses may be caused to the project. In this context, traditional monitoring methods have been unable to meet the high requirements of modern engineering construction for safety management. The rapid development of modern technology, especially in the fields of data acquisition, data processing and artificial intelligence, has provided new means and ideas for safety monitoring and early warning of foundation pit construction. Existing monitoring methods often rely on manual experience and regular inspections, lack real-time and systematic, and are prone to miss key change information. In recent years, the concept of introducing intelligent monitoring systems in deep foundation pit construction has gradually received attention. Through the combination of data acquisition module, data transmission module, data processing module and early warning module, a real-time and accurate monitoring and early warning system can be built. The system can acquire various environmental data of foundation pit construction in real time by arranging necessary monitoring points, and use advanced data processing and analysis techniques to intelligently evaluate construction status and predict potential risks, thereby providing strong protection for construction safety.

[0003] In terms of monitoring data processing, the following problems exist in the safety control system and method for deep foundation pit construction of sluice gates in large-scale water conservancy hubs: 1) Due to the variety of monitoring points, the monitoring data obtained from different types of monitoring points have different data change characteristics. For example, the data of some points show a relatively obvious random change trend, such as base uplift (rebound) monitoring points, soil nail internal force monitoring points, etc. The main reason is that they are affected by multiple uncertain factors; some points show obvious periodic change trends, such as water level monitoring, air temperature and water temperature monitoring, etc. The main reason is that they are affected by periodic factors (seasonal changes, daily operation cycles) Some points show an obvious trend of increasing or decreasing, such as foundation pit deformation monitoring points, anchor internal force monitoring points, etc.; and the current research results cannot carry out a unified classification analysis for a variety of monitoring points; 2) Lack of effective classification standards: Current research focuses on the change analysis of a single monitoring point, lacks a unified classification standard and analysis method for different monitoring points, especially for monitoring points whose data show random, periodic or tendency changes. The current tools and technologies are often unable to carry out effective grouping and processing, which limits the system's response ability in the face of complex real-time changes.

[0004] In order to overcome the above problems, it is urgent to develop a more flexible and intelligent safety control system and method for deep foundation pit construction of large-scale water conservancy hub sluice gates, which can identify and process data change characteristics for different types of monitoring points. At the same time, advanced technologies such as deep learning should be combined to explore multi-level monitoring data processing solutions based on feature extraction and prediction models. Summary of the invention

[0005] The purpose of the present invention is to provide a safety control method and system for deep foundation pit construction of a sluice gate of a large-scale water conservancy hub.

[0006] The purpose of the present invention can be achieved by adopting the following technical solutions:

[0007] A safety control method for deep foundation pit construction of a large-scale water conservancy project sluice gate, comprising the following steps:

[0008] S101 Develop a monitoring plan for deep foundation pit construction of sluice gates in large-scale water conservancy projects;

[0009] The formulation of a monitoring plan for the deep foundation pit construction of a large-scale water conservancy project sluice gate includes, based on the project conditions and survey results, formulating a monitoring plan for the deep foundation pit construction of a large-scale water conservancy project sluice gate, wherein the monitoring plan for the deep foundation pit construction of a large-scale water conservancy project sluice gate includes monitoring content, monitoring items, monitoring equipment and instruments, monitoring point layout, monitoring frequency, early warning value setting, and monitoring data processing and analysis;

[0010] S102 Arrange monitoring points;

[0011] The arrangement of monitoring points includes arranging a total of m monitoring points for deep foundation pit construction, marked as D1 to D m The types of deep foundation pit monitoring points include foundation pit deformation monitoring points, stress monitoring points, water level monitoring points, and foundation pit base uplift monitoring points. The types of deep foundation pit monitoring points are divided into n categories, expressed as J1~J n kind;

[0012] S103 obtains monitoring data;

[0013] The monitoring data obtained includes the following: m Get the corresponding monitoring data Data1~Data n , the monitoring data includes a timestamp of obtaining the monitoring data and a specific monitoring value;

[0014] S104 Grouping of monitoring points;

[0015] The monitoring point grouping includes using expert judgment to classify the monitoring point categories J1 to J n The deep foundation pit construction monitoring points D1 to D m Apply label A or label B or label C, and finally sort out the monitoring points of label A and mark them as monitoring points G1~G x , the monitoring points of tag B are marked as H1~H y The monitoring points of label C are marked as K1~K z , where x+y+z=m;

[0016] S105 monitoring data processing;

[0017] The monitoring data processing includes the following steps: x 、H1~H y and K1~K z The monitoring data of G1 to G2 are used to check the missing values ​​and supplement the missing values. x The missing values ​​of the monitoring data are supplemented by the periodic moving average method. The monitoring points H1~H y The missing values ​​of the monitoring data are supplemented by linear interpolation method. The monitoring points K1~K z The missing values ​​of monitoring data were supplemented by multiple interpolation method;

[0018] S106 constructs a prediction model;

[0019] The prediction model is constructed, including the following steps: x The monitoring data is used to build a recurrent neural network prediction model M G , for monitoring points H1~H y The monitoring data is used to build a recurrent neural network prediction model M H , for monitoring points K1~K z The monitoring data is used to build a recurrent neural network prediction model M K ;

[0020] S107 prediction model training and optimization adjustment;

[0021] The prediction model training and optimization adjustment include the following steps: x The monitoring data adopts model M G Carry out model training and optimization, and finally obtain the optimized and adjusted model M G1 , for monitoring points H1~H y The monitoring data adopts model M H Carry out model training and optimization, and finally obtain the optimized and adjusted model M H1 , for monitoring points K1~K z The monitoring data adopts model M K Carry out model training and optimization, and finally obtain the optimized and adjusted model M K1 ;

[0022] S108 obtains prediction value and warning;

[0023] The obtained prediction value and warning include: G1 、M H1 and M K1 , analyze the latest data to obtain the predicted value for a period of time in the future, compare the predicted value with the early warning value in the deep foundation pit construction monitoring plan of the sluice gate of a large-scale water conservancy hub, and issue an early warning if it exceeds the early warning value.

[0024] Furthermore, in the above step S104, the grouping step is:

[0025] a) Obtain the type of monitoring point J i Class, i = 1~n;

[0026] b) Use expert judgment to evaluate J i The data change trend of the data of the monitoring point is judged, and the judgment result is i The data change trend of the monitoring point is one of the following: periodic change trend, tendency change trend and random change trend;

[0027] c) Get J iThe data change trend judgment results of the monitoring points are as follows: i When the data change trend of the monitoring point is judged as a periodic change trend, it is divided into Group I. i When the data change trend of the monitoring point is a tendency change trend, it is divided into Group II. i When the data change trend of the monitoring point is determined to be a random change trend, it is divided into Group III;

[0028] d) Repeat steps a to c until all monitoring point types are grouped.

[0029] Furthermore, in the above step S105, the step of supplementing missing values ​​by using the periodic moving average method is:

[0030] a) Obtain monitoring point G i , the monitoring data in i=1~x and the inspection of missing values;

[0031] b) Select the period length T based on the periodicity of the data and the location of the missing values;

[0032] c) according to the selected cycle length T, for each missing value position, calculate the average value of a certain number N of data points before and after it as the supplementary value, wherein the number N is calculated by formula (1);

[0033]

[0034] e) Repeat steps a to c until all monitoring points are at G i Missing values ​​were filled in.

[0035] Furthermore, in the above step S105, the step of using the linear interpolation method to supplement the missing values ​​is:

[0036] a) Obtain monitoring point H j , the monitoring data in j=1~y and the inspection of missing values;

[0037] b) For each missing value, get its previous value represented as y 前 , and the subsequent value is expressed as y 后 ;

[0038] c) Calculate the linear interpolation y as the supplementary value of the missing value. The calculation formula is shown in formula (2):

[0039]

[0040] In the formula, ω1 is the weight of the previous value, ω2 is the weight of the next value, x is the timestamp of the missing value, and x 前 is the timestamp of the previous value, x后 The timestamp of the latter value;

[0041] d) Repeat steps a to c until all monitoring points are at H j Missing values ​​were filled in.

[0042] Furthermore, in the above step S105, the step of using the multiple interpolation method to supplement the missing values ​​is:

[0043] a) Obtain monitoring point K q , the monitoring data in q=1~z and the inspection of missing values;

[0044] b) selecting an interpolation model based on the distribution and characteristics of the monitoring data, wherein the interpolation model includes linear regression, logistic regression and random forest;

[0045] c) Use the interpolation model to interpolate each missing value to generate n complete data sets, where n is 5 to 10;

[0046] d) Use Rubin's rule to merge multiple imputed values ​​and fill in missing values;

[0047] e) Repeat steps a to d until all monitoring points are K q Missing values ​​were filled in.

[0048] Furthermore, in the above step S106, the prediction model M G The activation function uses a combination of a hyperbolic tangent function and a linear rectifier function. The expression of the combination of a hyperbolic tangent function and a linear rectifier function is formula (3). The combination of a hyperbolic tangent function and a linear rectifier function can better target periodically changing data. Periodic time series data usually has a certain periodicity and regularity. The combination of a hyperbolic tangent function and a linear rectifier function can simultaneously learn the periodic patterns and other potential nonlinear features in the data, so that the model can more accurately capture the changing trend when making predictions.

[0049]

[0050] Where ω is the combination coefficient, and the optimal value is determined by learning data during model training.

[0051] Furthermore, in the above step S106, the prediction model M HThe activation function adopts a combination of polynomial and exponential functions. The expression of the combination of polynomial and exponential functions is formula (4). The combination of polynomial and exponential functions can make the network have flexible function approximation ability, which can not only handle the steady change of data (polynomial) but also capture the fast-growing behavior (exponential). The combination of polynomial and exponential functions can handle the changes of different features in the prediction process, and can also increase the expression ability of the network, so that the recurrent neural network can capture various potential dynamic features when facing complex time series data. Considering the above, choosing the combination of polynomial and exponential functions as the activation function can better meet the needs of modeling time series data with tendency changes and improve the prediction accuracy and robustness of the model.

[0052] f(x)=(az 2 +bz+c)exp(-dz 2 ) (4)

[0053] Where a, b, c and d are coefficients.

[0054] Furthermore, in the above step S106, the prediction model M K The activation function uses a sinusoidal amplification unit function, which is expressed as formula (5). The sinusoidal amplification unit function can generate outputs in both negative and positive regions, which enables the model to flexibly respond to different input conditions when processing random data with positive and negative fluctuations. Compared with traditional activation functions (such as ReLU or sigmoid), it can provide a wider output dynamic range.

[0055] f(x)=xsin(x) (5)

[0056] Where sin(x) is the sine function of the variable x.

[0057] A safety control system for deep foundation pit construction of a large-scale water conservancy hub sluice gate, specifically includes a data acquisition module, a data transmission module, a data processing module, and an early warning module. The data acquisition module is used to obtain monitoring data, the data transmission module is used to transmit the monitoring data to the data processing module, the transmission module is also used to transmit the early warning result of the data processing module to the early warning module, the data processing module is used to process the monitoring data to obtain the early warning result, and the early warning module is used to make corresponding operations according to the early warning result after obtaining the early warning result.

[0058] The present invention has the following beneficial effects: 1) Intelligent classification analysis: by classifying the changing trends of data at different monitoring points (periodicity, tendency, randomness, etc.), more targeted analysis and processing can be achieved. This intelligent classification method can make the system more flexible and adaptable when facing complex construction environments, thereby improving the responsiveness of the overall monitoring system; 2) Diversified activation function design: based on the data classification results, different activation functions (such as hyperbolic tangent function and linear rectification function, and a combination of polynomial and exponential function) are applied in the prediction model, thereby improving the model's targeted fitting ability for the data. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a flow chart of a method for safety control of deep foundation pit construction of a large-scale water conservancy project sluice gate of the present invention;

[0060] Figure 2 A typical diagram of monitoring point data changes showing a periodic change trend according to an embodiment of the present invention;

[0061] Figure 3 A typical diagram of monitoring point data changes showing a tendency change trend in an embodiment of the present invention;

[0062] Figure 4 This is a typical diagram of monitoring point data changes with a random change trend according to an embodiment of the present invention. DETAILED DESCRIPTION

[0063] The specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings; it should be understood that the specific embodiments given here are only used to illustrate and explain the present invention and cannot be used to limit the present invention.

[0064] The following is a specific embodiment of a method and system for controlling the safety of deep foundation pit construction of a sluice gate in a large-scale water conservancy project.

[0065] A method for safety control of deep foundation pit construction of a sluice gate of a large-scale water conservancy hub, characterized in that the method comprises the following steps:

[0066] S101 Develop a monitoring plan for deep foundation pit construction of large-scale water conservancy project sluice gates

[0067] The formulation of a monitoring plan for the deep foundation pit construction of a large-scale water conservancy project sluice gate includes, based on the project conditions and survey results, formulating a monitoring plan for the deep foundation pit construction of a large-scale water conservancy project sluice gate, wherein the monitoring plan for the deep foundation pit construction of a large-scale water conservancy project sluice gate includes monitoring content, monitoring items, monitoring equipment and instruments, monitoring point layout, monitoring frequency, early warning value setting, and monitoring data processing and analysis;

[0068] S102 Arrange monitoring points;

[0069] The arrangement of monitoring points includes arranging a total of m monitoring points for deep foundation pit construction, marked as D1 to D m The types of deep foundation pit monitoring points include foundation pit deformation monitoring points, stress monitoring points, water level monitoring points, and foundation pit base uplift monitoring points. The types of deep foundation pit monitoring points are divided into n categories, expressed as J1~J n kind;

[0070] S103 obtains monitoring data;

[0071] The monitoring data obtained includes the following: m Get the corresponding monitoring data Data1~Data n , the monitoring data includes a timestamp of obtaining the monitoring data and a specific monitoring value;

[0072] S104 Grouping of monitoring points

[0073] The monitoring point grouping includes using expert judgment to classify the monitoring point categories J1 to J n The deep foundation pit construction monitoring points D1 to D m Apply label A or label B or label C, and finally sort out the monitoring points of label A and mark them as monitoring points G1~G x , the monitoring points of tag B are marked as H1~H y The monitoring points of label C are marked as K1~K z , where x+y+z=m;

[0074] Furthermore, in the above step S104, the grouping step is:

[0075] a) Obtain the type of monitoring point J i Class, i = 1~n;

[0076] b) Use expert judgment to evaluate J i The data change trend of the data of the monitoring point is judged, and the judgment result is i The data change trend of the monitoring point is one of the following: periodic change trend, tendency change trend and random change trend;

[0077] c) Get J i The data change trend judgment results of the monitoring points are as follows: i When the data change trend of the monitoring point is judged as a periodic change trend, it is divided into Group I. i When the data change trend of the monitoring point is a tendency change trend, it is divided into Group II.i When the data change trend of the monitoring point is determined to be a random change trend, it is divided into Group III;

[0078] d) Repeat steps a to c until all monitoring point types are grouped.

[0079] In this embodiment, the expert judgment method is applied to J i The data change trend of the data of the monitoring point is judged, and the judgment result is i The data change trend of the monitoring point is one of the following: periodic change trend, tendency change trend and random change trend. The typical graph of the monitoring point data change with periodic change trend is as follows: Figure 2 As shown in the figure, the typical graph of the change of monitoring point data with tendency change trend is as follows Figure 3 As shown in the figure, the typical graph of random change trend of monitoring point data is as follows Figure 4 shown.

[0080] S105 monitoring data processing;

[0081] The monitoring data processing includes the following steps: x 、H1~H y and K1~K z The monitoring data of G1 to G2 are used to check and supplement the missing values ​​of the data. x The missing values ​​of the monitoring data are supplemented by the periodic moving average method. The monitoring points H1~H y The missing values ​​of the monitoring data are supplemented by linear interpolation method. The monitoring points K1~K z The missing values ​​of monitoring data were supplemented by multiple interpolation method;

[0082] S106 constructs a prediction model;

[0083] The prediction model is constructed, including the following steps: x The monitoring data is used to build a recurrent neural network prediction model M G , for monitoring points H1~H y The monitoring data is used to build a recurrent neural network prediction model M H , for monitoring points K1~K z The monitoring data is used to build a recurrent neural network prediction model M K ;

[0084] S107 prediction model training and optimization adjustment;

[0085] The prediction model training and optimization adjustment include the following steps: x The monitoring data adopts model MG Carry out model training and optimization, and finally obtain the optimized and adjusted model M G1 , for monitoring points H1~H y The monitoring data adopts model M H Carry out model training and optimization, and finally obtain the optimized and adjusted model M H1 , for monitoring points K1~K z The monitoring data adopts model M K Carry out model training and optimization, and finally obtain the optimized and adjusted model M K1 ;

[0086] S108 obtains prediction value and warning;

[0087] The obtained prediction value and warning include: G1 、M H1 and M K1 , analyze the latest data to obtain the predicted value for a period of time in the future, compare the predicted value with the early warning value in the deep foundation pit construction monitoring plan of the sluice gate of a large-scale water conservancy hub, and issue an early warning if it exceeds the early warning value.

[0088] Furthermore, in the above step S105, the step of supplementing missing values ​​by using the periodic moving average method is:

[0089] a) Obtain monitoring point G i , the monitoring data in i=1~x and the inspection of missing values;

[0090] b) Select the period length T based on the periodicity of the data and the location of the missing values;

[0091] c) according to the selected cycle length T, for each missing value position, calculate the average value of a certain number N of data points before and after it as the supplementary value, wherein the number N is calculated by formula (1);

[0092]

[0093] e) Repeat steps a to c until all monitoring points are at G i Missing values ​​were filled in.

[0094] Furthermore, in the above step S105, the step of using the linear interpolation method to supplement the missing values ​​is:

[0095] a) Obtain monitoring point H j , the monitoring data in j=1~y and the inspection of missing values;

[0096] b) For each missing value, get its previous value represented as y前 , and the subsequent value is expressed as y 后 ;

[0097] c) Calculate the linear interpolation y as the supplementary value of the missing value. The calculation formula is shown in formula (2):

[0098]

[0099] In the formula, ω1 is the weight of the previous value, ω2 is the weight of the next value, x is the timestamp of the missing value, and x 前 is the timestamp of the previous value, x 后 The timestamp of the latter value;

[0100] d) Repeat steps a to c until all monitoring points are at H j Missing values ​​were filled in.

[0101] Furthermore, in the above step S105, the step of using the multiple interpolation method to supplement the missing values ​​is:

[0102] a) Obtain monitoring point K q , the monitoring data in q=1~z and the inspection of missing values;

[0103] b) selecting an interpolation model based on the distribution and characteristics of the monitoring data, wherein the interpolation model includes linear regression, logistic regression and random forest;

[0104] c) Use the interpolation model to interpolate each missing value to generate n complete data sets, where n is 5 to 10;

[0105] d) Use Rubin's rule to merge multiple imputed values ​​and fill in missing values;

[0106] e) Repeat steps a to d until all monitoring points are K q Missing values ​​were filled in.

[0107] Furthermore, in the above step S106, the prediction model M G The activation function uses a combination of a hyperbolic tangent function and a linear rectifier function. The expression of the combination of a hyperbolic tangent function and a linear rectifier function is formula (3). The combination of a hyperbolic tangent function and a linear rectifier function can better target periodically changing data. Periodic time series data usually has a certain periodicity and regularity. The combination of a hyperbolic tangent function and a linear rectifier function can simultaneously learn the periodic patterns and other potential nonlinear features in the data, so that the model can more accurately capture the changing trend when making predictions.

[0108]

[0109] Where ω is the combination coefficient, and the optimal value is determined by learning data during model training.

[0110] Furthermore, in the above step S106, the prediction model M H The activation function adopts a combination of polynomial and exponential functions. The expression of the combination of polynomial and exponential functions is formula (4). The combination of polynomial and exponential functions can make the network have flexible function approximation ability, which can not only handle the steady change of data (polynomial) but also capture the fast-growing behavior (exponential). The combination of polynomial and exponential functions can handle the changes of different features in the prediction process, and can also increase the expression ability of the network, so that the recurrent neural network can capture various potential dynamic features when facing complex time series data. Considering the above, choosing the combination of polynomial and exponential functions as the activation function can better meet the needs of modeling time series data with tendency changes and improve the prediction accuracy and robustness of the model.

[0111] f(x)=(az 2 +bz+c)exp(-dz 2 ) (4)

[0112] Where a, b, c and d are coefficients.

[0113] Furthermore, in the above step S106, the prediction model M K The activation function uses a sinusoidal amplification unit function, which is expressed as formula (5). The sinusoidal amplification unit function can generate outputs in both negative and positive regions, which enables the model to flexibly respond to different input conditions when processing random data with positive and negative fluctuations. Compared with traditional activation functions (such as ReLU or sigmoid), it can provide a wider output dynamic range.

[0114] f(x)=xsin(x) (5)

[0115] Where sin(x) is the sine function of the variable x.

[0116] A safety control system for deep foundation pit construction of a large-scale water conservancy hub sluice gate, specifically includes a data acquisition module, a data transmission module, a data processing module, and an early warning module. The data acquisition module is used to obtain monitoring data, the data transmission module is used to transmit the monitoring data to the data processing module, the transmission module is also used to transmit the early warning result of the data processing module to the early warning module, the data processing module is used to process the monitoring data to obtain the early warning result, and the early warning module is used to make corresponding operations according to the early warning result after obtaining the early warning result.

[0117] In the above embodiments, the present invention discloses a method and system for safety control of deep foundation pit construction of a sluice gate of a large-scale water conservancy hub. A safety control system for deep foundation pit construction of a sluice gate of a large-scale water conservancy hub includes a data acquisition module, a data transmission module, a data processing module, and an early warning module. A method for safety control of deep foundation pit construction of a sluice gate of a large-scale water conservancy hub includes formulating a monitoring plan for deep foundation pit construction of a sluice gate of a large-scale water conservancy hub, arranging monitoring points, obtaining monitoring data, grouping monitoring points, processing monitoring data, building a prediction model, training and optimizing the prediction model, and obtaining prediction values ​​and early warnings; the method obtains more accurate prediction values ​​by grouping the types of monitoring points and adopting different data processing methods in a targeted manner, thereby improving the response capability of the overall monitoring system and can be widely used in the field of deep foundation pit construction.

[0118] The above description is a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A safety control method for deep foundation pit construction of a large-scale water conservancy project sluice gate, characterized in that: The steps of the safety control method for deep foundation pit construction of a large-scale water conservancy project sluice gate are as follows: S101 Develop a monitoring plan for deep foundation pit construction of sluice gates in large-scale water conservancy projects; S102 Arrange monitoring points; S103 obtains monitoring data; S104 Grouping of monitoring points; S105 monitoring data processing; S106 constructs a prediction model; S107 prediction model training and optimization adjustment; S108 obtains prediction value and warning; The formulation of a monitoring plan for the deep foundation pit construction of a large-scale water conservancy project sluice gate includes, based on the project conditions and survey results, formulating a monitoring plan for the deep foundation pit construction of a large-scale water conservancy project sluice gate, wherein the monitoring plan for the deep foundation pit construction of a large-scale water conservancy project sluice gate includes monitoring content, monitoring items, monitoring equipment and instruments, monitoring point layout, monitoring frequency, early warning value setting, and monitoring data processing and analysis; The arrangement of monitoring points includes arranging a total of m monitoring points for deep foundation pit construction, marked as D1 to D m The types of deep foundation pit construction monitoring points include but are not limited to foundation pit deformation monitoring points, stress monitoring points, water level monitoring points, and foundation pit base uplift monitoring points. The types of deep foundation pit construction monitoring points are divided into n categories, expressed as J1~J n kind; The monitoring data obtained includes the following: m Get the corresponding monitoring data Data1~Data n , the monitoring data includes a timestamp of obtaining the monitoring data and a specific monitoring value; The monitoring point grouping includes using expert judgment to classify the monitoring point categories J1 to J n The deep foundation pit construction monitoring points D1 to D m Apply label A or label B or label C, and finally sort out the monitoring points of label A and mark them as monitoring points G1~G x , the monitoring points of tag B are marked as H1~H y The monitoring points of label C are marked as K1~K z , where x+y+z=m; The monitoring data processing includes the following steps: x 、H1~H y and K1~K z The monitoring data of G1 to G2 are used to check the missing values ​​and supplement the missing values. x The missing values ​​of the monitoring data are supplemented by the periodic moving average method. The monitoring points H1~H y The missing values ​​of the monitoring data are supplemented by linear interpolation method. The monitoring points K1~K z The missing values ​​of monitoring data were supplemented by multiple interpolation method; The prediction model is constructed, including the following steps: x The monitoring data is used to build a recurrent neural network prediction model M G , for monitoring points H1~H y The monitoring data is used to build a recurrent neural network prediction model M H , for monitoring points K1~K z The monitoring data is used to build a recurrent neural network prediction model M K ; The prediction model training and optimization adjustment include the following steps: x The monitoring data adopts model M G Carry out model training and optimization, and finally obtain the optimized and adjusted model M G1 , for monitoring points H1~H y The monitoring data adopts model M H Carry out model training and optimization, and finally obtain the optimized and adjusted model M H1 , for monitoring points K1~K z The monitoring data adopts model M K Carry out model training and optimization, and finally obtain the optimized and adjusted model M K1 ; The obtained prediction value and warning include: G1 、M H1 and M K1 , analyze the latest data to obtain the predicted value for a period of time in the future, compare the predicted value with the early warning value in the deep foundation pit construction monitoring plan of the sluice gate of a large-scale water conservancy hub, and issue an early warning if it exceeds the early warning value.

2. A method for controlling the safety of deep foundation pit construction of a large-scale water conservancy project sluice gate according to claim 1, characterized in that: In step S104, the monitoring point types J1 to J n The specific steps for grouping classes are: a) Obtain the type of monitoring point J i Class, i = 1~n; b) Use expert judgment to evaluate J i The data change trend of the data of the monitoring point is judged, and the judgment result is i The data change trend of the monitoring point is one of the following: periodic change trend, tendency change trend and random change trend; c) Get J i The data change trend judgment results of the monitoring points are as follows: i When the data change trend of the monitoring point is judged as a periodic change trend, it is divided into Group I. i When the data change trend of the monitoring point is a tendency change trend, it is divided into Group II. i When the data change trend of the monitoring point is determined to be a random change trend, it is divided into Group III; d) Repeat steps a to c until all monitoring point types are grouped.

3. A method for controlling the safety of deep foundation pit construction of a large-scale water conservancy project sluice gate according to claim 1, characterized in that: In step S105, the step of supplementing missing values ​​using the periodic moving average method is: a) Obtain monitoring point G i , the monitoring data in i=1~x and the inspection of missing values; b) Select the period length T based on the periodicity of the data and the location of the missing values; c) according to the selected cycle length T, for each missing value position, calculate the average value of a certain number N of data points before and after it as the supplementary value, wherein the number N is calculated by formula (1); e) Repeat steps a to c until all monitoring points are at G i Missing values ​​were filled in.

4. A method for controlling the safety of deep foundation pit construction of a large-scale water conservancy project sluice gate according to claim 1, characterized in that: In step S105, the step of using the linear interpolation method to supplement the missing values ​​is: a) Obtain monitoring point H j , the monitoring data in j=1~y and the inspection of missing values; b) For each missing value, get its previous value represented as y 前 , and the subsequent value is expressed as y 后 ; c) Calculate the linear interpolation y as the supplementary value of the missing value. The calculation formula is shown in formula (2): In the formula, ω1 is the weight of the previous value, ω2 is the weight of the next value, x is the timestamp of the missing value, and x 前 is the timestamp of the previous value, x 后 The timestamp of the latter value; d) Repeat steps a to c until all monitoring points are at H j Missing values ​​were filled in.

5. A method for controlling the safety of deep foundation pit construction of a large-scale water conservancy project sluice gate according to claim 1, characterized in that: In step S105, the step of using multiple interpolation method to supplement missing values ​​is: a) Obtain monitoring point K q , the monitoring data in q=1~z and the inspection of missing values; b) selecting an interpolation model based on the distribution and characteristics of the monitoring data, wherein the interpolation model includes linear regression, logistic regression and random forest; c) Use the interpolation model to interpolate each missing value to generate n complete data sets, where n is 5 to 10; d) Use Rubin's rule to merge multiple imputed values ​​and fill in missing values; e) Repeat steps a to d until all monitoring points are K q Missing values ​​were filled in.

6. A method for controlling the safety of deep foundation pit construction of a large-scale water conservancy project sluice gate according to claim 1, characterized in that: In step S106, the activation function of the prediction model MG adopts a combination of a hyperbolic tangent function and a linear rectification function. The expression of the combination of the hyperbolic tangent function and the linear rectification function is formula (3), Where ω is the combination coefficient, and the optimal value is determined by learning data during model training.

7. A method for controlling the safety of deep foundation pit construction of a large-scale water conservancy project sluice gate according to claim 1, characterized in that: In step S106, the activation function of the prediction model MH adopts a combination of a polynomial and an exponential function. The expression of the combination of a polynomial and an exponential function is formula (4): f(x)=(az 2 +bz+c)exp(-dz 2 ) (4) Where a, b, c and d are coefficients.

8. A method for controlling the safety of deep foundation pit construction of a large-scale water conservancy project sluice gate according to claim 1, characterized in that: In step S106, the activation function of the prediction model MK adopts a sinusoidal amplification unit function, and the expression of the sinusoidal amplification unit function is formula (5), f(x)=xsin(x) (5) Where sin(x) is the sine function of the variable x.

9. A safety control system for deep foundation pit construction of a large-scale water conservancy project sluice gate, used in the safety control method for deep foundation pit construction of a large-scale water conservancy project sluice gate as claimed in any one of claims 1 to 8, characterized in that: Specifically, it includes a data acquisition module, a data transmission module, a data processing module and an early warning module; the data acquisition module is used to obtain monitoring data, the data transmission module is used to transmit the monitoring data to the data processing module, the transmission module is also used to transmit the early warning result of the data processing module to the early warning module, the data processing module is used to process the monitoring data to obtain the early warning result, and the early warning module is used to make corresponding operations according to the early warning result after obtaining the early warning result.

Images