Foundation pit deformation prediction and evaluation method based on correlation coefficient and improved salat swarm algorithm

Abstract

The invention discloses a foundation pit deformation prediction and evaluation method based on a correlation coefficient and an improved salat swarm algorithm, which comprises two steps of correlation analysis and prediction analysis and evaluation, and comprises the following steps of: establishing a correlation analysis model according to measured data of foundation pit deformation and characteristic parameters of a foundation pit; analyzing the correlation degree of the foundation pit characteristics and the foundation pit deformation according to the Pearson's correlation coefficient, selecting the first four groups of factors with higher correlation degree and the foundation pit deformation measured data to carry out data preprocessing, proposing a prediction precision requirement, taking the correlation influence factors in the processed data as an input layer, taking the foundation pit deformation as an output layer to carry out training, and obtaining a prediction result; according to the method, the error of the prediction result and the actual measurement result of the training set is analyzed, the deformation of the foundation pit is predicted according to the characteristics of the foundation pit, the safety of the supporting structure of the foundation pit is evaluated according to the prediction result, and the problems that an existing prediction model is large in error and cannot accurately predict the deformation development condition of the foundation pit can be effectively solved.

Description

Technical Field

[0001] The invention relates to the technical field of foundation pit deformation prediction, and in particular to a foundation pit deformation prediction and evaluation method based on a correlation coefficient and an improved sand cat group algorithm. Background Art

[0002] With the rapid development of the economy and society, underground space for various purposes has been developed and utilized in large and medium-sized cities, and the scale of foundation pits has become deep, large and complex.

[0003] However, the development of underground space often has a certain impact on the surrounding environment, and in extreme cases, it can cause great damage to the surrounding environment. Therefore, in the process of underground space development, intelligent monitoring, prediction and evaluation are of great significance for identifying hidden dangers in the construction process and guiding design and construction.

[0004] Foundation pit monitoring technology and deformation prediction are important measures to ensure the safety of foundation pit construction. At present, the most common foundation pit deformation monitoring and prediction methods use time series prediction methods, and establish training sets based on field measured data to generate prediction models to predict the subsequent deformation development of the foundation pit. The prediction model generated by this method is relatively simple. Since the training set is only learned from the existing deformation, the prediction model has a large error and cannot accurately predict the development of foundation pit deformation, which often cannot guide design and construction.

[0005] Based on the above reasons, the present invention discloses a foundation pit deformation prediction and evaluation method based on correlation coefficient and improved sand cat group algorithm Summary of the invention

[0006] The purpose of the present invention is to provide a foundation pit deformation prediction and evaluation method based on correlation coefficient and improved sand cat group algorithm, which can effectively solve the problem that the existing prediction model has large errors and cannot accurately predict the development of foundation pit deformation. Starting from the characteristics of the foundation pit itself, the inherent law between the foundation pit deformation and the foundation pit characteristics is analyzed, and the correlation of various factors causing the foundation pit deformation is studied to establish a suitable deformation prediction model to reflect the actual deformation of the foundation pit and predict the future deformation trend of the foundation pit, with high accuracy. The specific technical scheme is as follows:

[0007] The following steps are involved:

[0008] S1. Correlation analysis: the correlation analysis model is established based on the measured data of foundation pit deformation and the characteristic parameters of the foundation pit during foundation pit engineering monitoring. The correlation between the foundation pit characteristics and the foundation pit deformation is analyzed based on the Pearson correlation coefficient. The first four groups of factors with higher correlation are selected for data preprocessing with the measured data of foundation pit deformation, and prediction accuracy requirements are proposed.

[0009] S2. Prediction analysis and evaluation. The prediction analysis and evaluation uses the correlation influencing factors in the preprocessed data described in the above step S1 as the input layer and the foundation pit deformation as the output layer for training, analyzes the errors between the prediction results of the training set and the measured results, predicts the foundation pit deformation according to the foundation pit characteristics, and evaluates the foundation pit according to the prediction results.

[0010] Furthermore, the parameters of the foundation pit characteristics in step S1 include excavation depth, excavation time, support axial force, internal friction angle of soil layer within the excavation depth range, cohesion, gravity, and water level difference inside and outside the foundation pit.

[0011] Furthermore, the specific steps of step S2 include:

[0012] S21, foundation pit measured data, foundation pit characteristic parameter analysis;

[0013] S22, analyzing the correlation between the measured data of foundation pit deformation and the characteristic parameters of the foundation pit according to the correlation analysis part;

[0014] S23, preprocessing the measured data of foundation pit deformation and foundation pit characteristic parameters, with the foundation pit characteristic parameters as the input layer and the measured data of foundation pit deformation as the output layer, and dividing the preprocessed data into a training set and an output set;

[0015] S24, establish a wandering strategy to improve the sand cat swarm algorithm model;

[0016] S25, performing comparative analysis based on the training set prediction model and the measured data of foundation pit deformation, analyzing the error, and determining whether the prediction accuracy requirement is met;

[0017] S26, analyzing the test set to predict the development of foundation pit deformation;

[0018] S27. Evaluate the foundation pit deformation based on the prediction results.

[0019] Furthermore, the step S22 defines the correlation between each characteristic parameter of the foundation pit and the measured data of the foundation pit deformation by using the Pearson correlation coefficient, by calculating the quotient of the covariance and standard deviation between each characteristic parameter of the foundation pit and the measured data of the foundation pit deformation:

[0020]

[0021] Where ρ(X,Y) is the overall Pearson correlation coefficient between the two variables, X is the characteristic parameter of the foundation pit, and Y is the measured data of the foundation pit deformation;

[0022] Determine the correlation between the foundation pit characteristic parameter X and the measured data Y of foundation pit deformation under the predicted working condition, and select the first four foundation pit characteristic parameters with the highest correlation.

[0023] Furthermore, the step S23 performs data preprocessing on the first four groups of factors with a high degree of correlation and the measured data of foundation pit deformation to generate an input layer matrix:

[0024]

[0025]

[0026] In the above formula: Fitness is the fitness function.

[0027] Furthermore, the above step S24 improves the sand cat swarm algorithm according to the wandering strategy, evaluates the fitness function of the input layer parameters, and obtains the optimal value of the parameters. ij The corresponding X1 will be output. When an iteration is completed, the best solution is selected. If there is no ideal model in the previous iteration, and other layer data in the next iteration, try to move towards the direction of the ideal model. The best solution in each iteration can represent the closer the predicted value is to the measured data of the foundation pit deformation, as follows:

[0028] Pos(t+1)=r(Pos bc (t)-rand(0,1)*Pos c (t))

[0029] In the above formula: Pos bc is the optimal position (1), Pos c is its current position, r is the sensitivity range, the input layer updates its position according to the optimal solution, current position and sensitivity range, searches for other possible best output layers, and finds a new local optimum in the new search area.

[0030] Furthermore, r is the sensitivity range, and the sensitivity ranges of foundation pit characteristic parameters are different. The specific algorithm is:

[0031]

[0032] R=2*r G *rand(0,1)-r G

[0033] r=r G *rand(0,1)

[0034] Where S M is the characteristic value of the foundation pit characteristic parameter, t is the current iteration number, T is the maximum iteration number, and R controls the transition between the exploration stage and the development stage. The characteristic value of the foundation pit characteristic parameter is summarized based on the field measured data.

[0035] Furthermore, Pos(t+1) represents the distance between the optimal position and the current position, as follows

[0036] Pos rnd =|rand(0,1)*Pos b (t)-Pos c (t)|

[0037] Pos(t+1)=Pos b (t)-r*Pos rnd *cos(α)

[0038] Where:Pos b is the optimal position (2); Pos c Current location; Pos rnd is a random position, α is a random angle and α is selected by roulette method.

[0039] Furthermore, in order to better expand the prediction range of the model training set and improve the convergence speed, the walking strategy is introduced, including lens imaging reverse learning, weighted walking, and triangle walking;

[0040] Among them, the lens imaging reverse learning generates a reverse position through the principle of convex lens imaging to expand the search range, avoid falling into the local optimum, expand the search range, and improve the population diversity. Specifically, in the two-dimensional coordinate, the value range of x is (a, b), y represents the convex lens, and A(X, Y) is known, where X is the characteristic parameter of the foundation pit and Y is the measured data of the foundation pit deformation.

[0041] The projection on the x-axis is x, and the projection on the y-axis is h. Through reverse learning of lens imaging, A is obtained. * (X′,Y′), where the projection on the x-axis is x * , the projection on the y-axis is h * ;

[0042] According to the lens imaging principle:

[0043]

[0044] make We can get:

[0045]

[0046] in is the position of the individual in the j-dimensional space, For x j is the reverse solution of j and b j is the maximum and minimum bounds of the j-th dimension in the search space;

[0047] The weighted walk can specify the weighted value of the next iteration position to control the convergence speed. Specifically, the fitness functions of the first four foundation pit characteristic parameters are f1, f2, f3, and f4 respectively.

[0048] In order to speed up the overall convergence of the prediction model, dynamic weights w1, w2, and w3 are introduced.

[0049]

[0050]

[0051]

[0052] τ randomly takes values ​​of f1, f2, f3, f4 and τ does not take MAX(fi) or MIN(fi).

[0053]

[0054] The introduction of weighted walk increases the role of the parameters with greater influence on the fitness of the foundation pit deformation in the iteration process.

[0055] Posnew=Pos b (t)+(Pos b (t)-Posc(t))*w

[0056] Pos b is the optimal position; Pos c is the current position; Posnew is the optimal position after introducing weights;

[0057] The triangle walk is performed around the triangle during the iterative position update process.

[0058] L1=Pos b (t)-Pos c (t)

[0059] L2=L2*rand(0,1)

[0060] Among them, L1 is the distance between X and Y, X is the characteristic parameter of the foundation pit, Y is the measured data of the foundation pit deformation, and L2 is the walking step length.

[0061] Determine the direction of travel The walking direction applicable to this project

[0062] After walking through the defined triangle

[0063] Posnew=Pos b (t)+P*r

[0064] Among them, P is the distance between X and Y after walking through the triangle, X is the characteristic parameter of the foundation pit, and Y is the measured data of the foundation pit deformation, which is defined as

[0065] Furthermore, the optimization termination condition is determined to determine whether the number of iterations reaches the maximum or whether the fitness value of the global optimal position meets the set value of the fitness value. If it is satisfied, the parameter optimization algorithm ends, otherwise, the training optimization is restarted;

[0066] Furthermore, the prediction results of the test set are judged according to the Technical Code for Construction Foundation Pit Support, the Technical Standard for Monitoring Construction Foundation Pit Engineering and local regulations to determine the deformation of the foundation pit in the future, thereby evaluating the safety of the foundation pit support structure.

[0067] Beneficial effects of the present invention: This method establishes a correlation analysis model based on the measured data of foundation pit deformation and the characteristic parameters of the foundation pit, analyzes the degree of correlation between the foundation pit characteristics and the foundation pit deformation based on the Pearson correlation coefficient, selects the first four groups of factors with higher correlation and the measured data of the foundation pit deformation for data preprocessing, and proposes prediction accuracy requirements. The correlation influencing factors in the processed data are used as the input layer, and the foundation pit deformation is used as the output layer for training, and the errors between the prediction results of the training set and the measured results are analyzed. The foundation pit deformation is predicted based on the characteristics of the foundation pit, and the safety of the foundation pit support structure is evaluated based on the prediction results, which can effectively solve the problem that the existing prediction model has large errors and cannot accurately predict the development of foundation pit deformation. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a flow chart of the foundation pit deformation prediction and evaluation method based on correlation coefficient and improved sand cat group algorithm of the present invention;

[0069] Figure 2 This is the wandering strategy of the present invention to improve the optimization algorithm process of the sand cat group;

[0070] Figure 3 is the fitness curve of the implementation case of the present invention;

[0071] Figure 4 is a comparison chart of the training set prediction value and the true value of the implementation case of the present invention;

[0072] Figure 5 is a comparison chart of the predicted values ​​and true values ​​of the test set of the implementation case of the present invention;

[0073] Figure 6 is a test set prediction error sample of an implementation case of the present invention;

[0074] Figure 7It is a comparison chart of predicted values ​​and true values ​​of the neural network test samples of the improved sand cat swarm algorithm of the wandering strategy of the implementation case of the present invention;

[0075] Figure 8 It is the error of the test sample of the neural network of the improved sand cat swarm algorithm by the wandering strategy of the implementation case of the present invention. DETAILED DESCRIPTION

[0076] Embodiment 1:

[0077] Combine the following Figure 1-Figure 8 To further clarify and fully describe the present invention, an embodiment of the present invention provides an improved sand cat group algorithm based on the Pearson correlation coefficient and the wandering strategy, including:

[0078] S1. Correlation analysis: the correlation analysis model is established based on the measured data of foundation pit deformation and the characteristic parameters of the foundation pit during foundation pit engineering monitoring. The correlation between the foundation pit characteristics and the foundation pit deformation is analyzed based on the Pearson correlation coefficient. The first four groups of factors with higher correlation are selected for data preprocessing with the measured data of foundation pit deformation, and prediction accuracy requirements are proposed.

[0079] S2. Prediction analysis and evaluation. The prediction analysis and evaluation uses the correlation influencing factors in the preprocessed data described in the above step S1 as the input layer and the foundation pit deformation as the output layer for training, analyzes the errors between the prediction results of the training set and the measured results, predicts the foundation pit deformation according to the foundation pit characteristics, and evaluates the foundation pit according to the prediction results.

[0080] Preferably, step S1 correlation analysis is a preliminary data processing stage, in which a correlation analysis model is established based on the measured data of foundation pit deformation monitored by the foundation pit engineering and the foundation pit characteristic parameters, the degree of correlation between the foundation pit characteristics and the foundation pit deformation is analyzed based on the Pearson correlation coefficient, the first four groups of factors with higher correlation degrees are selected for data preprocessing with the measured data of foundation pit deformation, and prediction accuracy requirements are proposed.

[0081] Preferably, the prediction analysis and evaluation are trained based on the correlation influencing factors in the processed data described in the correlation analysis as the input layer and the foundation pit deformation as the output layer, the errors between the prediction results of the training set and the measured results are analyzed, the foundation pit deformation is predicted according to the foundation pit characteristics, and the foundation pit is evaluated based on the prediction results.

[0082] Preferably: the foundation pit characteristic parameters include excavation depth, excavation time, support axial force, internal friction angle of soil layer within the excavation depth range, cohesion, gravity, and water level difference inside and outside the foundation pit.

[0083] Preferably, step S2 prediction analysis and evaluation specifically includes:

[0084] S21, foundation pit measured data, foundation pit characteristic parameter analysis;

[0085] S22, analyzing the correlation between the measured data of foundation pit deformation and the characteristic parameters of the foundation pit according to the correlation analysis part;

[0086] S23, preprocessing the measured data of foundation pit deformation and foundation pit characteristic parameters, with the foundation pit characteristic parameters as the input layer and the measured data of foundation pit deformation as the output layer, and dividing the preprocessed data into a training set and an output set;

[0087] S24, establish a wandering strategy to improve the sand cat swarm algorithm model;

[0088] S25, performing comparative analysis based on the training set prediction model and the measured data of foundation pit deformation, analyzing the error, and determining whether the prediction accuracy requirement is met;

[0089] S26, analyzing the test set to predict the development of foundation pit deformation;

[0090] S27. Evaluate the foundation pit deformation based on the prediction results.

[0091] Preferably, the frequency of measured data on foundation pit deformation is 1 time / h, including excavation depth, excavation time, support axial force, internal friction angle of soil layer within the excavation depth range, cohesion, gravity, and water level difference inside and outside the foundation pit.

[0092] Preferably, the Pearson correlation coefficient is calculated to determine the correlation between the foundation pit excavation depth, excavation time, support axial force, friction angle of soil layer within the excavation depth range, cohesion, gravity, water level difference inside and outside the foundation pit and the measured data of foundation pit deformation under the predicted working conditions, and the top 4 foundation pit characteristic parameters with higher correlation are selected. The calculation results of this implementation are shown in Table 1, and the calculation results show that the top 4 foundation pit characteristic parameters with higher correlation are support axial force, excavation depth, excavation depth, and foundation pit water level difference.

[0093] Table 1 Pearson correlation coefficient calculation results

[0094] Support shaft force Excavation time Excavation depth Water level difference inside and outside the foundation pit Correlation coefficient 0.995 0.997 0.977 0.966

[0095] Preferably, the first four groups of factors with higher correlation are preprocessed with the measured data of foundation pit deformation to generate an input layer matrix, and some of the selected results are shown in Table 2.

[0096] Table 2 Measured data of foundation pit deformation

[0097]

[0098] Preferably, the sand cat swarm algorithm is improved according to the wandering strategy, and the fitness function of the input layer parameters is evaluated to obtain the optimal value of the parameters. ijThe corresponding X1 will be output. When an iteration is completed, the best solution is selected. If there is no ideal model in the previous iteration, and other layer data in the next iteration, try to move towards the ideal model. The best solution in each iteration can represent the closer the predicted value is to the measured data of foundation pit deformation.

[0099] Preferably, according to the sand cat swarm algorithm, the input layer updates its position according to the optimal solution, the current position and the sensitivity range, searches for other possible best output layers, and finds a new local optimum in the new search area.

[0100] Pos(t+1)=r(Pos bc (t)-rand(0,1)*Pos c (t))

[0101] Where:Pos bc is the optimal position (1), Pos c is its current position, and r is the sensitivity range. The input layer updates its position according to the optimal solution, current position, and sensitivity range, searches for other possible best output layers, and finds a new local optimum in the new search area.

[0102] Preferably, r is a sensitivity range. The sensitivity ranges of the foundation pit characteristic parameters are different.

[0103]

[0104] R=2*r G *rand(0,1)-r G

[0105] r=r G *rand(0,1)

[0106] Where S M is the characteristic value of the foundation pit characteristic parameter, t is the current iteration number, T is the maximum iteration number, and R controls the transition between the exploration stage and the development stage. The characteristic value of the foundation pit characteristic parameter is summarized based on the field measured data.

[0107] Preferably, the distance between the optimal position and the current position is expressed as follows

[0108] Pos rnd =|rand(0,1)*Pos b (t)-Pos c (t)|

[0109] Pos(t+1)=Pos b (t)-r*Pos rnd *cos(α)

[0110] Where:Pos b is the optimal position (2); Pos c Current location; Pos rnd is a random position, and α is a random angle. Random angles can prevent the algorithm from falling into local optimality.

[0111] Preferably, in order to better expand the prediction range of the model training set and improve the convergence speed, a walking strategy is introduced, and the walking strategy introduced includes lens imaging reverse learning, weighted walking, and triangle walking;

[0112] Among them, the lens imaging reverse learning generates a reverse position through the principle of convex lens imaging to expand the search range, avoid falling into the local optimum, expand the search range, and improve the population diversity. Specifically, in the two-dimensional coordinate, the value range of x is (a, b), y represents the convex lens, and A(X, Y) is known, where X is the characteristic parameter of the foundation pit and Y is the measured data of the foundation pit deformation.

[0113] The projection on the x-axis is x, and the projection on the y-axis is h. Through reverse learning of lens imaging, A is obtained. * (X′,Y′), where the projection on the x-axis is x * , the projection on the y-axis is h * ;

[0114] According to the lens imaging principle:

[0115]

[0116] make We can get:

[0117]

[0118] where x j is the position of the individual in the j-dimensional space, For x j is the reverse solution of j and b j is the maximum and minimum bounds of the j-th dimension in the search space;

[0119] The weighted walk can specify the weighted value of the next iteration position to control the convergence speed. Specifically, the fitness functions of the first four foundation pit characteristic parameters are f1, f2, f3, and f4 respectively.

[0120] In order to speed up the overall convergence of the prediction model, dynamic weights w1, w2, w3 are introduced, where

[0121]

[0122]

[0123]

[0124] τ randomly takes values ​​of f1, f2, f3, f4 and τ does not take MAX(fi) or MIN(fi). Specifically:

[0125]

[0126] The introduction of weighted walk increases the role of the parameters with greater influence on the fitness of the foundation pit deformation in the iteration process. After adjustment:

[0127] Posnew=Pos b (t)+(Pos b (t)-Posc(t))*w

[0128] Pos b is the optimal position; Pos c is the current position; Posnew is the optimal position after introducing weights;

[0129] The triangle walk is to walk around in the process of iteratively updating the position, specifically:

[0130] L1=Pos b (t)-Pos c (t)

[0131] L2=L1*rand(0,1)

[0132] Among them, L1 is the distance between X and Y, X is the characteristic parameter of the foundation pit, Y is the measured data of the foundation pit deformation, and L2 is the walking step length.

[0133] Determine the direction of travel The walking direction applicable to this project

[0134] After walking through the defined triangle:

[0135] Posnew=Pos b (t)+P*r

[0136] Among them, P is the distance between X and Y after walking through the triangle, X is the characteristic parameter of the foundation pit, and Y is the measured data of the foundation pit deformation, which is defined as

[0137] Preferably, the optimization termination condition is determined to determine whether the number of iterations reaches the maximum or whether the fitness value of the global optimal position meets the set value of the fitness value. If so, the parameter optimization algorithm ends, otherwise, training and optimization are performed again.

[0138] Preferably, the prediction results of the test set are determined according to the "Technical Code for Building Foundation Pit Support", "Technical Standards for Building Foundation Pit Engineering Monitoring" and local specifications to judge the deformation of the foundation pit in the future, thereby evaluating the safety of the foundation pit support structure.

[0139] Preferably, the prediction effect of this implementation is as follows Figure 3-8 The mean square error of the training set is 0.19938. The results of 60 groups (hours) are predicted in advance. By comparison, it is found that the prediction results of the sand cat group prediction model improved by the wandering strategy are closer to the true value than the common BP neural network prediction model for the next 60 hours. The prediction error is smaller than the common BP neural network prediction model, and the prediction effect is better.

[0140] The embodiments of the present invention disclose preferred embodiments, but are not limited thereto. A person skilled in the art can easily understand the spirit of the present invention based on the above embodiments and make different extensions and changes. However, as long as they do not deviate from the spirit of the present invention, they are all within the protection scope of the present invention.

Claims

1. A foundation pit deformation prediction and evaluation method based on correlation coefficient and improved sand cat group algorithm is characterized by: The following steps are involved: S1. Correlation analysis: the correlation analysis model is established based on the measured data of foundation pit deformation and the characteristic parameters of the foundation pit during foundation pit engineering monitoring. The correlation between the foundation pit characteristics and the foundation pit deformation is analyzed based on the Pearson correlation coefficient. The first four groups of factors with higher correlation are selected for data preprocessing with the measured data of foundation pit deformation, and prediction accuracy requirements are proposed. S2. Prediction analysis and evaluation. The prediction analysis and evaluation uses the correlation influencing factors in the preprocessed data described in the above step S1 as the input layer and the foundation pit deformation as the output layer for training, analyzes the errors between the prediction results of the training set and the measured results, predicts the foundation pit deformation according to the foundation pit characteristics, and evaluates the foundation pit according to the prediction results.

2. The foundation pit deformation prediction and evaluation method based on correlation coefficient and improved sand cat group algorithm according to claim 1 is characterized in that: The parameters of the foundation pit characteristics in step S1 include excavation depth, excavation time, support axial force, internal friction angle of soil layer within the excavation depth range, cohesion, gravity, and water level difference inside and outside the foundation pit.

3. The foundation pit deformation prediction and evaluation method based on correlation coefficient and improved sand cat group algorithm as claimed in claim 2 is characterized in that: The specific steps of step S2 include: S21, foundation pit measured data, foundation pit characteristic parameter analysis; S22, analyzing the correlation between the measured data of foundation pit deformation and the characteristic parameters of the foundation pit according to the correlation analysis part; S23, preprocessing the measured data of foundation pit deformation and foundation pit characteristic parameters, with the foundation pit characteristic parameters as the input layer and the measured data of foundation pit deformation as the output layer, and dividing the preprocessed data into a training set and an output set; S24, establish a wandering strategy to improve the sand cat swarm algorithm model; S25, performing comparative analysis based on the training set prediction model and the measured data of foundation pit deformation, analyzing the error, and determining whether the prediction accuracy requirement is met; S26, analyzing the test set to predict the development of foundation pit deformation; S27. Evaluate the foundation pit deformation based on the prediction results.

4. The foundation pit deformation prediction and evaluation method based on correlation coefficient and improved sand cat swarm algorithm as claimed in claim 3 is characterized in that: The step S22 defines the correlation between each characteristic parameter of the foundation pit and the measured data of the foundation pit deformation by using the Pearson correlation coefficient, and calculates the quotient of the covariance and standard deviation between each characteristic parameter of the foundation pit and the measured data of the foundation pit deformation: Where ρ(X,Y) is the overall Pearson correlation coefficient between the two variables, X is the characteristic parameter of the foundation pit, and Y is the measured data of the foundation pit deformation; Determine the correlation between the foundation pit characteristic parameter X and the measured data Y of foundation pit deformation under the predicted working condition, and select the first four foundation pit characteristic parameters with the highest correlation.

5. The foundation pit deformation prediction and evaluation method based on correlation coefficient and improved sand cat swarm algorithm according to claim 4 is characterized in that: The step S23 performs data preprocessing on the first four groups of factors with higher correlation and the measured data of foundation pit deformation to generate an input layer matrix: In the above formula: Fitness is the fitness function.

6. The foundation pit deformation prediction and evaluation method based on correlation coefficient and improved sand cat swarm algorithm according to claim 5 is characterized in that: The above step S24 improves the sand cat swarm algorithm according to the wandering strategy, evaluates the fitness function of the input layer parameters, and obtains the optimal value of the parameters. ij The corresponding X1 will be output. When an iteration is completed, the best solution is selected. If there is no ideal model in the previous iteration, and other layer data in the next iteration, try to move towards the direction of the ideal model. The best solution in each iteration can represent the closer the predicted value is to the measured data of the foundation pit deformation, as follows: Pos(t+1)=r(Pos bc (t)-rand(0,1)*Pos c (t)) In the above formula: Pos bc is the optimal position (1), Pos c is its current position, r is the sensitivity range, the input layer updates its position according to the optimal solution, current position and sensitivity range, searches for other possible best output layers, and finds a new local optimum in the new search area.

7. The foundation pit deformation prediction and evaluation method based on correlation coefficient and improved sand cat group algorithm according to claim 6 is characterized in that: r is the sensitivity range. The sensitivity ranges of foundation pit characteristic parameters are different. The specific algorithm is: R=2*r G *rand(0,1)-r G r=r G *rand(0,1) Where S M is the characteristic value of the foundation pit characteristic parameter, t is the current iteration number, T is the maximum iteration number, R controls the conversion between the exploration stage and the development stage, and the characteristic value of the foundation pit characteristic parameter is summarized based on the field measured data.

8. The foundation pit deformation prediction and evaluation method based on correlation coefficient and improved sand cat swarm algorithm according to claim 7 is characterized in that: Pos(t+1) represents the distance between the optimal position and the current position, as follows: Pos rnd =|rand(0,1)*Pos b (t)-Pos c (t)| Pos(t+1)=Pos b (t)-r*Pos rnd *cos(α) Where:Pos b is the optimal position (2); Pos c is the current position; Pos rnd is a random position, α is a random angle and α is selected by roulette method.

9. The foundation pit deformation prediction and evaluation method based on correlation coefficient and improved sand cat swarm algorithm according to claim 8, characterized in that: In order to better expand the prediction range of the model training set and improve the convergence speed, the walking strategy is introduced, including lens imaging reverse learning, weighted walking, and triangle walking; Among them, the lens imaging reverse learning generates a reverse position through the principle of convex lens imaging to expand the search range, avoid falling into the local optimum, expand the search range, and improve the population diversity. Specifically, in the two-dimensional coordinate, the value range of x is (a, b), y represents the convex lens, and A(X, Y) is known, where X is the characteristic parameter of the foundation pit and Y is the measured data of the foundation pit deformation. The projection on the x-axis is x, and the projection on the y-axis is h. Through reverse learning of lens imaging, A is obtained. * (X′,Y′), where the projection on the x-axis is x * , the projection on the y-axis is h * ; According to the lens imaging principle: make We can get: where x j is the position of the individual in the j-dimensional space, For x j is the reverse solution of j and b j is the maximum and minimum bounds of the j-th dimension in the search space; The weighted walk can specify the weighted value of the next iteration position to control the convergence speed. Specifically, the fitness functions of the first four foundation pit characteristic parameters are f1, f2, f3, and f4 respectively. In order to speed up the overall convergence of the prediction model, dynamic weights w1, w2, w3 are introduced, where τ randomly takes values ​​of f1, f2, f3, f4 and τ does not take MAX(fi) or MIN(fi). Specifically: The introduction of weighted walk increases the role of the parameters with greater influence on the fitness of the foundation pit deformation in the iteration process. After adjustment: Posnew=Pos b (t)+(Pos b (t)-Posc(t))*w Pos b is the optimal position; Pos c is the current position; Posnew is the optimal position after introducing weights; The triangle walk is to walk around in the process of iteratively updating the position, specifically: L1=Pos b (t)-Pos c (t) L2=L1*rand(0,1) Among them, L1 is the distance between X and Y, X is the characteristic parameter of the foundation pit, Y is the measured data of the foundation pit deformation, and L2 is the walking step length. Determine the direction of travel The walking direction applicable to this project After walking through the defined triangle: Posnew=Pos b (t)+P*r Among them, P is the distance between X and Y after walking through the triangle, X is the characteristic parameter of the foundation pit, and Y is the measured data of the foundation pit deformation, which is defined as 10. The foundation pit deformation prediction and evaluation method based on correlation coefficient and improved sand cat swarm algorithm as claimed in claim 9 is characterized by: Determine the termination condition of optimization, whether the number of iterations reaches the maximum or whether the fitness value of the global optimal position meets the set value of the fitness value. If it meets the set value, the parameter optimization algorithm ends, otherwise, retraining and optimization are performed; The prediction results of the test set are determined according to the Technical Code for Construction Foundation Pit Support, the Technical Standard for Construction Foundation Pit Engineering Monitoring and local specifications to judge the deformation of the foundation pit in the future, thereby evaluating the safety of the foundation pit support structure.

Images