Shield tunnel settlement prediction method and device, equipment and storage medium

Abstract

The invention discloses a shield tunnel settlement prediction method, device and equipment and a storage medium, and the method comprises the steps: obtaining various feature factor data related to shield tunnel settlement, and screening out at least one feature factor data from the feature factor data as target feature factor data; constructing a settlement prediction model, taking the target characteristic factor data as the input of the settlement prediction model, taking the error between the minimum predicted settlement value and the real settlement value as the optimization target of the settlement prediction model, and carrying out iterative updating on the settlement prediction model; and based on the updated settlement prediction model, obtaining a predicted settlement value corresponding to the current target characteristic factor data, and taking the predicted settlement value as a settlement prediction result. According to the technical scheme of the embodiment of the invention, accurate prediction of the tunnel settlement trend is realized by establishing the complex relation model between the tunnel settlement and the influence factors.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tunnel rail transit, in particular to a shield tunnel settlement prediction method, device, equipment and storage medium. BACKGROUND

[0002] The shield tunnel is a common tunnel construction method in subway construction. Due to the complexity and variability of the surrounding environment of the shield tunnel, uneven settlement often occurs in the operation stage of the shield tunnel, and secondary disasters such as segment misalignment, water leakage, bolt yield and the like caused by uneven settlement seriously threaten the safety of subway operation in major cities.

[0003] There are some tunnel settlement prediction methods in traditional technologies, but these methods often have deficiencies when dealing with numerous complex factors affecting tunnel settlement. Due to the numerous factors affecting tunnel settlement and the correlation between them, there is a collinearity problem, which leads to a decrease in the accuracy and stability of the prediction.

[0004] Shield tunnel settlement is a problem that cannot be ignored, which not only affects the normal operation of the subway, but also may cause harm to the surrounding environment. Therefore, accurately predicting the settlement of the operating subway shield tunnel is of great significance for ensuring the safe operation of the subway and taking appropriate maintenance measures in advance. SUMMARY

[0005] The present application provides a shield tunnel settlement prediction method, device, equipment and storage medium, which realizes efficient prediction of tunnel settlement while improving the accuracy and reliability of settlement prediction.

[0006] In a first aspect, the present application provides a shield tunnel settlement prediction method, comprising:

[0007] Obtaining a plurality of characteristic factor data associated with the settlement of the shield tunnel, and selecting at least one characteristic factor data therefrom as target characteristic factor data;

[0008] Building a settlement prediction model, taking the target characteristic factor data as the input of the settlement prediction model, taking the error between the predicted settlement value and the true settlement value as the optimization target of the settlement prediction model, and iteratively updating the settlement prediction model;

[0009] Based on the updated settlement prediction model, obtaining the predicted settlement value corresponding to the current target characteristic factor data as the settlement prediction result.

[0010] Further, the selection process of the target characteristic factor data comprises:

[0011] Data cleaning is performed on the plurality of characteristic factor data associated with the settlement of the shield tunnel;

[0012] constructing a covariance matrix of the characteristic factor data after cleaning by principal component analysis;

[0013] solving eigenvalues in the covariance matrix, and sequentially ranking each of the eigenvalues from large to small;

[0014] sequentially calculating the contribution rate of each of the eigenvalues, and accumulating the contribution value according to the ranking order of each of the eigenvalues;

[0015] when the contribution value is accumulated to a preset range, the corresponding characteristic factor data is taken as the target characteristic factor data.

[0016] Further, in the process of constructing and iteratively updating the settlement prediction model, it comprises:

[0017] selecting a training sample set and a test sample set based on the target characteristic factor data; wherein the training sample set is used to construct the settlement prediction model based on the extreme gradient boosting algorithm, and the test sample set is used to update the settlement prediction model based on the leave-one-out cross-validation algorithm;

[0018] finding the optimal hyperparameter combination of the settlement prediction model in the hyperparameter optimization space based on the Bayesian optimization algorithm, the optimal hyperparameter combination is used to determine the model structure of the settlement prediction model constructed based on the extreme gradient boosting algorithm; the optimal hyperparameter combination includes: the maximum depth of the tree, the learning rate, the proportion of the sub-sample, and the minimum leaf node weight.

[0019] Further, the training sample set and the test sample set belong to the same sample set, based on the model structure of the settlement prediction model, a plurality of groups of settlement prediction models are constructed, when constructing any group of settlement prediction models, a group of samples is selected from the sample set as the test sample set of the group of models, and the remaining samples are used as the training sample set of the group of models.

[0020] determining the objective function of the settlement prediction model based on the training sample set, and evaluating the model structure of the settlement prediction model based on the test sample set; the evaluation parameters include: the determination coefficient, the root mean square error, and the mean absolute percentage error.

[0021] Further, the settlement prediction model is a stepwise additive model composed of a plurality of decision tree models; the objective function of the settlement prediction model is composed of a loss function and a decision tree function, the loss function is composed of the sum of the mean square error between the predicted settlement value and the true settlement value of each training set sample, and the decision tree function is composed of the sum of the complexity of each decision tree model.

[0022] Further, the construction process of the settlement prediction model comprises:

[0023] determine a model structure of the settlement prediction model based on the optimal hyperparameter combination, select each decision tree model based on the training set samples, correct the prediction error of a preceding decision tree model by a subsequent decision tree model in the selection process, and construct multiple groups of objective functions;

[0024] find the target function corresponding to the minimum error between the predicted settlement value and the actual settlement value through cyclic iteration, and use the target function as the target function of the settlement prediction model.

[0025] Further, the iterative updating process of the settlement prediction model comprises:

[0026] based on the evaluation results of each group of constructed settlement prediction models, calculate the average value of the evaluation results;

[0027] If the average value meets the preset threshold, the group of settlement prediction models with the optimal evaluation result is used as the updated settlement prediction model.

[0028] If the average value does not meet the preset threshold, the hyperparameter optimization space is adjusted, the model structure of the settlement prediction model constructed based on the extreme gradient boosting algorithm is re-determined, and the process is repeated until the average value of the evaluation results under the model structure meets the preset threshold.

[0029] In a second aspect, the present application provides a shield tunnel settlement prediction device, comprising:

[0030] a feature selection module configured to obtain multiple feature factor data associated with the settlement of the shield tunnel, and select at least one feature factor data as target feature factor data;

[0031] a model construction module configured to construct a settlement prediction model, use the target feature factor data as the input of the settlement prediction model, minimize the error between the predicted settlement value and the actual settlement value as the optimization target of the settlement prediction model, and iteratively update the settlement prediction model;

[0032] a settlement prediction module configured to obtain a predicted settlement value corresponding to the current target feature factor data based on the updated settlement prediction model, and use the predicted settlement value as the settlement prediction result.

[0033] In a third aspect, an electronic device is provided, and the electronic device comprises:

[0034] at least one processor; and a memory connected to the at least one processor in communication;

[0035] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the steps of the shield tunnel settlement prediction method of any one of the embodiments of the application.

[0036] In a fourth aspect, the embodiments of the application provide a computer readable storage medium storing computer instructions for enabling a processor to perform the steps of the shield tunnel settlement prediction method of any one of the embodiments of the application when the processor executes the computer instructions.

[0037] Compared with the prior art, the application has the following beneficial effects:

[0038] The technical solution in the embodiments of the application first acquires a plurality of characteristic factor data associated with the settlement of the shield tunnel, and screens at least one characteristic factor data from the plurality of characteristic factor data as target characteristic factor data; then constructs a settlement prediction model, takes the target characteristic factor data as the input of the settlement prediction model, takes the minimization of the error between the predicted settlement value and the actual settlement value as the optimization target of the settlement prediction model, and iteratively updates the settlement prediction model; and finally obtains the predicted settlement value corresponding to the current target characteristic factor data based on the updated settlement prediction model, and takes the predicted settlement value as the settlement prediction result. Through the technical solution in the embodiments of the application, the settlement of the operating subway shield tunnel can be accurately predicted. BRIEF DESCRIPTION OF DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only preferred embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0040] Figure 1 A flowchart of a shield tunnel settlement prediction method provided by the embodiments of the application is shown in the figure.

[0041] Figure 2 A structure diagram of a PCA-BOA-XGBoost algorithm provided by the embodiments of the application is shown in the figure.

[0042] Figure 3 A structure diagram of a leave-one-cross-validation algorithm provided by the embodiments of the application is shown in the figure.

[0043] Figure 4 A flowchart of a shield tunnel settlement prediction process provided by the embodiments of the application is shown in the figure.

[0044] Figure 5 A structure diagram of a shield tunnel settlement prediction device provided by the embodiments of the application is shown in the figure.

[0045] Figure 6 A structural schematic diagram of an electronic device is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0046] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0047] The present application provides a shield tunnel settlement prediction method, Figure 1 A flowchart of a shield tunnel settlement prediction method is provided for an embodiment of the present application. The embodiment can be applicable to real-time settlement prediction through tunnel settlement related characteristic factor data. The method can be executed by a shield tunnel settlement prediction device. The device can be realized in the form of software and / or hardware. The device can be configured in an electronic device.

[0048] As shown in the figure, the method specifically includes: Figure 1

[0049] S1, obtaining a plurality of characteristic factor data associated with shield tunnel settlement, and screening at least one characteristic factor data therefrom as target characteristic factor data.

[0050] It should be noted that the plurality of characteristic factor data associated with shield tunnel settlement at least includes train operation load, operation time, train running speed, train axle load, tunnel burial depth, soil body property, soil layer type, soil body parameter, underground water level, pore water pressure, and corresponding tunnel cumulative settlement.

[0051] Further, the screening process of the target characteristic factor data includes:

[0052] S101, data cleaning of the plurality of characteristic factor data associated with shield tunnel settlement.

[0053] S102, constructing a covariance matrix of the cleaned characteristic factor data by principal component analysis.

[0054] S103, solving the eigenvalues in the covariance matrix, and sequentially sorting each eigenvalue from large to small.

[0055] S104, sequentially calculating the contribution rate of each eigenvalue, and accumulating the contribution value according to the eigenvalue arrangement order.

[0056] ​S105, when the contribution value accumulates to a preset range, the corresponding characteristic factor data is taken as the target characteristic factor data.

[0057] Specifically, first, the acquired multiple characteristic factor data is removed of abnormal values and missing values; then the cleaned data is normalized; then the normalized characteristic factor data is zero-mean; then the covariance matrix of the characteristic factor data is constructed by principal component analysis, and the eigenvalues and eigenvectors of the covariance matrix are solved; finally, the eigenvalues are arranged in descending order, and the eigenvalues with a cumulative contribution rate reaching a preset range (90%) are selected, and the corresponding characteristic factor data is taken as the target characteristic factor data.

[0058] The normalization process can be represented as:

[0059]

[0060] In the formula, X is the normalized eigenvalue, x is the original value of the characteristic factor data, is the maximum value in the original value, is the minimum value in the original value.

[0061] The zero-mean process can be represented as:

[0062]

[0063] In the formula, is the characteristic mean, is the decentered eigenvalue.

[0064] For example, taking a subway line in a certain city as an example, first, the relevant data of multiple shield tunnel sections of the subway line in the city during the operation period are collected, 55 groups of long-term monitoring data of five sections are collected, including operation time, tunnel depth, soil properties, groundwater level, train operation load and corresponding tunnel cumulative settlement, etc. characteristic factor data, the data is cleaned, 1 group of abnormal value is removed, the remaining 54 groups of data sets are normalized and zero-mean processed, the processed 54 groups of data are input into the PCA model, and the contribution rate of each principal component is calculated, the contribution rate threshold is set to 90%, the cumulative contribution rate of the first three principal components reaches 95%, the characteristic factors corresponding to the three principal components, i.e. tunnel depth, operation time and groundwater level, are taken as the main characteristic factors for subsequent use. At this time, the characteristic data is reduced from the original 6 factors to 3 factors, effectively reducing the data dimension.

[0065] In the embodiment of the present application, the collected multiple characteristic factor data is first processed to remove outliers and missing values, and then standardized to make the data of different characteristic factors have the same dimension, facilitating subsequent analysis and modeling. The cleaned data is input into the PCA model for collinearity analysis of the characteristic factors affecting tunnel settlement. PCA converts the original multiple characteristics into a few uncorrelated principal component characteristics through linear transformation, which retains as much information of the original data as possible. By calculating the contribution rate of each principal component, the influence degree of each factor on settlement is determined according to the contribution rate. A contribution rate threshold range is preset, and the characteristic factors corresponding to the principal components with a contribution rate greater than the threshold are retained as the main characteristic factors for subsequent model training. While retaining the main information, the number of characteristic factors is effectively reduced, the dimension of data is reduced, and the collinearity problem is solved.

[0066] The technical solution in the embodiment of the present application processes the numerous characteristic factors affecting tunnel settlement through principal component analysis, effectively identifies and eliminates collinearity, reduces redundant information between characteristic factors, reduces computational complexity, reduces the demand for computing power, reduces the amount of characteristic data collection, speeds up the database collection speed, and improves the stability and modeling speed of the model.

[0067] S2, a settlement prediction model is constructed, the target characteristic factor data is used as the input of the settlement prediction model, the error between the predicted settlement value and the true settlement value is minimized as the optimization target of the settlement prediction model, and the settlement prediction model is iteratively updated.

[0068] Figure 2 A structure diagram of a PCA-BOA-XGBoost algorithm provided by the embodiment of the present application is shown in FIG. 1. Figure 2 As shown in FIG. 1, a few uncorrelated principal component characteristics are first selected from the multiple characteristic factor data through principal component analysis (PCA), then the optimal hyperparameter combination is found through Bayesian optimization algorithm (BOA) according to the principal component characteristics and the corresponding true settlement value, and finally the settlement prediction model distributed and superimposed by multiple decision tree models is constructed through extreme gradient boosting algorithm (XGBoost) according to the optimal hyperparameter combination, and the decision tree models in the settlement prediction model are iteratively updated.

[0069] In the process of constructing and iteratively updating the settlement prediction model, it further includes:

[0070] The training sample set and the test sample set are selected based on the target characteristic factor data; the training sample set is used to construct the settlement prediction model based on the extreme gradient boosting algorithm; and the test sample set is used to update the settlement prediction model based on the leave-one-out cross-validation algorithm.

[0071] The Bayesian optimization algorithm is used to search for an optimal hyperparameter combination in a hyperparameter optimization space of the settlement prediction model, and the optimal hyperparameter combination is used to determine a model structure of the settlement prediction model constructed based on the extreme gradient boosting algorithm; the optimal hyperparameter combination includes a maximum depth of a tree, a learning rate, a proportion of a subsample, and a minimum leaf node weight.

[0072] Further, the training sample set and the test sample set belong to the same sample set, and based on the model structure of the settlement prediction model, a plurality of groups of settlement prediction models are constructed, and when any group of settlement prediction models is constructed, a group of samples is selected from the sample set as a test sample set for construction of the group of models, and the remaining samples are used as a training sample set for construction of the group of models.

[0073] Based on the training sample set, a target function of the settlement prediction model is determined, and based on the test sample set, the settlement prediction model is evaluated; the evaluation parameters include a coefficient of determination, a root mean square error, and a mean absolute percentage error.

[0074] It should be noted that the settlement prediction model is a stepwise additive model composed of a plurality of decision tree models; the target function of the settlement prediction model is composed of a loss function and a decision tree function, the loss function is composed of a sum of mean square errors between predicted settlement values and actual settlement values of each training set sample, and the decision tree function is composed of a sum of complexities of each decision tree model.

[0075] Further, the construction process of the settlement prediction model includes:

[0076] S201, a model structure of the settlement prediction model is determined based on the optimal hyperparameter combination, each decision tree model is selected based on the training set sample, and in the selection process, the prediction error of a previous decision tree model is corrected by a subsequent decision tree model, and a plurality of target functions are constructed;

[0077] S202, the target function corresponding to a minimum error between a predicted settlement value and an actual settlement value is searched for through a cyclic iteration manner, and is used as the target function of the settlement prediction model.

[0078] It should be noted that the Bayesian optimization algorithm (BOA) is a machine learning algorithm for global optimization, which predicts a target function by constructing a probability model and finds an optimal parameter. Firstly, a Gaussian process (GP) is used to describe the uncertainty of the target function, that is, the hyperparameters are fitted to the target function f by Gaussian process regression, and the optimal hyperparameter combination is found, and the process is as follows:

[0079]

[0080] In the formula, The optimal hyperparameter set is represented as X, and f(x) is the target function of the optimization.

[0081] For example, the BOA algorithm is used to optimize the hyperparameters of the XGBoost algorithm, and the hyperparameter optimization space X to be optimized is defined as: the maximum depth of the tree is 1-10, the learning rate is 0.01-0.5, the sub-sample ratio is 0.5-1, and the minimum leaf node weight is 1-10. Through iterative optimization, after 500 iterations, the optimal hyperparameter combination is found : the maximum depth of the tree is 2, the learning rate is 0.21, the sub-sample ratio is 0.8, and the minimum leaf node weight is 5.

[0082] Since the objective function satisfies the normal distribution, is the mean, and K is the covariance function. When given a data set D{(x1, y1), …, (xi, yi)}, the predicted value also follows a normal distribution, i.e.

[0083]

[0084]

[0085] )

[0086] The above formula is the core formula of Gaussian process regression. Gaussian process regression is a non-parametric regression method based on Bayesian theory, which is used to model and predict unknown functions.

[0087] where, indicates that the output y follows a normal distribution with mean and variance when given the input sample x and the data set D.

[0088] where, is the expression of the predicted mean, where is the kernel function between the input sample x and the training data input, K is the kernel matrix between the training data inputs, is the noise variance, l is the length scale, and y is the output of the training data.

[0089] where, ) is the expression of the predicted variance, which measures the uncertainty of the predicted result, is the kernel function value of the input x itself.

[0090] In the posterior probability iterative optimization process , the objective function and the optimization function error are minimized to obtain the optimal hyperparameter set. Through the BOA algorithm, the automatic optimization of the hyperparameters of the XGBoost algorithm can be realized, and the most suitable prediction hyperparameters under the current situation can be obtained.

[0091] After obtaining the optimal hyperparameter set based on the BOA algorithm for a given data set D{(x1, y1), …, (xi, yi)}, the target function corresponding to the settlement prediction model needs to be iteratively updated. The target function can be expressed as:

[0092] wherein, is the target function, is the loss function, is the function expression of the decision tree, i is the ith data sample, is the predicted value of the ith sample xi.

[0093] Further, the predicted value of the ith sample xi can be expressed as:

[0094]

[0095] Then the target function can be expressed as:

[0096]

[0097] Let be the second-order Taylor expansion at x0, and assume is also the second-order Taylor expansion at x0.

[0098] Let x= +ft(xi), the first derivative is denoted as ), and the second derivative is denoted as . The second-order Taylor expansion of f(x) and at x0 is brought into the target function, which can be expressed as:

[0099]

[0100] Since is a constant term, and since the preceding function t-1 tree structure is determined, it can also be denoted as a constant, and after removing and combining, the target function can be further simplified and expressed as:

[0101]

[0102] The XGBoost algorithm learner can be expressed as a CART tree:

[0103]

[0104] Wherein, xi is the i th sample, yi is the value corresponding to the i th sample, the i th tree is divided into leaf node q of the tree and corresponding weight part omega, is the weight of the leaf node q where the sample xi is located, is the position of the leaf node where the sample xi is located, that is, for any one sample xi, finally falls on a certain leaf node of the tree, and the value is .

[0105] Regularization part of the objective function All samples xi belonging to the j th leaf node are divided into a leaf node sample set, and can be represented as:

[0106]

[0107] The objective function can be represented as:

[0108]

[0109] The coefficients of the linear term and the quadratic term are combined, that is, let 、 The objective function can be represented as:

[0110]

[0111] At the minimum value of the target value, the decision tree structure reaches the optimum, and at this time, it is the optimal solution of the objective function.

[0112] In the embodiment of the application, the BOA algorithm is used to optimize the hyperparameters of the XGBoost algorithm. The BOA algorithm predicts the combination of hyperparameters that is most likely to make the objective function optimal by constructing a Gaussian process of the objective function according to the existing observation data. XGBoost is an ensemble learning algorithm based on gradient boosting decision trees, which has the advantages of high efficiency and accuracy. However, the performance of the XGBoost algorithm is sensitive to the setting of its hyperparameters, and the traditional hyperparameter optimization method is time-consuming and laborious and has poor effect. The XGBoost hyperparameters include but are not limited to the maximum depth of the tree, the learning rate, the proportion of the child sample, the minimum leaf node weight, etc. In each iteration, the BOA selects a combination of hyperparameters for evaluation according to the current probability model, and then updates the probability model using the evaluation result, and iterates until the optimal combination of hyperparameters is found.

[0113] The technical solution in the embodiment of the application optimizes the hyperparameters of the XGBoost algorithm by using the BOA algorithm, automatically optimizes through the Gaussian process, can find the optimal combination of hyperparameters in a short time, improves the performance of the XGBoost model, and thus improves the accuracy of the tunnel settlement prediction.

[0114] Further, the iterative updating process of the settlement prediction model comprises:

[0115] Based on the evaluation results of the settlement prediction model constructed for each group, an average value of the evaluation results is calculated;

[0116] If the average value meets a preset threshold, the settlement prediction model with the optimal evaluation result is taken as the updated settlement prediction model;

[0117] If the average value does not meet the preset threshold, the hyperparameter optimization space is adjusted, the model structure of the settlement prediction model constructed based on the extreme gradient boosting algorithm is re-determined, and the average value of the evaluation results under the model structure is adjusted until the average value meets the preset threshold.

[0118] Figure 3 A structure diagram of a leave-one-out cross-validation algorithm provided by an embodiment of the present application is shown in FIG. 1. Figure 3 For the divided K groups of sample data, K-1 groups of sample data are selected as a training set each time, one group of sample data that is not selected is taken as a test set, and the model is verified.

[0119] The leave-one-out method is a special case of the cross-validation method. Since it is difficult to collect the settlement data of the subway tunnel in operation, it is necessary to improve the data utilization rate, and the leave-one-out method can maximize the utilization of the data set and is suitable for the case where the sample quantity is small. The core idea is that each data is taken as a test set, and the data other than the test set is taken as a training set. The verification output quantity is the same as the training quantity, and therefore the leave-one-out method is not affected by the subjective division of samples. Since the K samples can only be divided into K subsets in a unique way, each subset contains one sample, the training set used by the leave-one-out method is only one sample less than the initial data set. A total of K rounds of circulation are performed to obtain K training results. Then, the average training error of the K results is calculated to evaluate the generalization error of the model. This makes the model actually evaluated in the leave-one-out method most similar to the model expected to be evaluated and trained in most cases.

[0120] For example, 45 groups of data are randomly selected from the 54 groups of data processed by PCA as a training set, and 9 groups of data are taken as a test set. The training set data are used to train the settlement prediction model by combining the XGBoost algorithm optimized by BOA. The trained model is verified on the training set by using the leave-one-out cross-validation method, and the determination coefficient, the root mean square error, and the average absolute percentage error are 0.929, 5.346, and 6.942%, respectively. Then, the test set data are used to test the settlement prediction model, and the root mean square error of the test set is 5.3 mm. According to the verification and test results, the model is fine-tuned, and finally the optimal prediction model is obtained.

[0121] Specifically, the PCA-BOA-XGBoost tunnel settlement prediction model is written in a computer using Pythom, the trained prediction model is saved in a mobile electronic device, current target characteristic factor data in the field is input, the operation shield tunnel settlement value is predicted, and tunnel settlement values in future time periods are obtained. After a period of time, the real value of tunnel settlement is measured by automatic monitoring of a total station or leveling, which is compared with the predicted value, and the actual value is input as feedback into the database of the prediction model, the model is continuously updated and improved, and a more accurate prediction model can be obtained.

[0122] In the embodiment of the application, the main characteristic factor data after PCA processing is divided into a training set and a test set. The training set data is used to train the model by combining the XGBoost algorithm optimized by BOA, and a model between tunnel settlement and main characteristic factors is established. The trained model is verified by leave-one-out cross-validation. Leave-one-out cross-validation is a special cross-validation method, which divides the data set into K samples, takes one sample as the test set each time, and the remaining K-1 samples as the training set, performs K times of training and testing, and finally takes the average value of K times of test results as the evaluation index of the model. Through leave-one-out cross-validation, the performance of the model can be more accurately evaluated by fully utilizing the limited data, and the problem of overfitting can be avoided. According to the verification result, the model parameters and the training process are continuously adjusted until the optimal prediction model is obtained.

[0123] The technical solution in the embodiment of the application combines the PCA, BOA and XGBoost algorithm to establish a model, fully considers the complex influence factors of tunnel settlement, further ensures the reliability of the model through leave-one-out cross-validation, maximizes the generalization ability of the algorithm model, and can more accurately predict the settlement of the operation subway shield tunnel, thereby providing strong support for subway operation and maintenance.

[0124] S3, based on the updated settlement prediction model, a predicted settlement value corresponding to the current target characteristic factor data is obtained and taken as a settlement prediction result.

[0125] Specifically, the present embodiment relates to a data storage for storing a database and a prediction model; and a computing device comprising a processor and a touch display screen, the computing device being coupled with the data storage and configured to execute a program stored in the data storage to implement steps in the PCA-BOA-XGBoost settlement prediction method. The device storing the prediction model can be used for subway shield tunnels in other areas to collect relevant feature factor data, and after pre-processing and PCA processing, the data is input into the finally obtained optimal model, for example: the predicted settlement of a tunnel after a period of operation is 8.1 mm. The subway operation management department can make corresponding maintenance measures in advance according to the prediction result.

[0126] Figure 4 A flowchart of a shield tunnel settlement prediction process provided by the embodiment of the present application is shown in Figure 4 The flowchart shows that a plurality of feature factor data associated with shield tunnel settlement is retrieved from a database, the data is normalized and principal component analysis is performed to extract main features as target feature factor data, the optimal hyperparameter combination of the target function is found according to the target feature data and the true settlement value and through a Bayesian optimization algorithm, an initial settlement prediction model is established according to the optimal hyperparameter combination, the decision tree in the prediction model is iteratively updated through an extreme gradient boosting tree algorithm, and then the error of the prediction model is predicted by leave-one-out cross-validation, the optimal prediction model is selected under the condition of limited sample quantity, and the output of the tunnel settlement prediction value is completed.

[0127] In the embodiment of the present application, the new operation subway shield tunnel related feature factor data (also pre-processed and PCA processed) is input into the finally obtained optimal prediction model, and the model can predict the settlement of the tunnel. The prediction result can be used by the subway operation management department to make maintenance plans in advance and take corresponding measures to ensure the safe operation of the subway.

[0128] The technical solution in the embodiment of the present application does not depend on specific subway lines or geological conditions, and can establish a complex relationship model between tunnel settlement and influencing factors through the algorithm model to predict the settlement trend of the future tunnel, and provide more scientific and accurate decision support for the design, construction and maintenance of the tunnel. The present application has good applicability to the settlement prediction of different types of operation subway shield tunnels in different regions, and has broad application prospects.

[0129] In summary, the technical scheme in the embodiment of the present application proposes an operation subway shield tunnel settlement prediction method based on a PCA-BOA-XGBoost algorithm. By combining principal component analysis, a Bayesian optimization algorithm and an extreme gradient boosting algorithm, the PCA effectively solves the collinearity problem of factors affecting tunnel settlement, reduces the data dimension, optimizes the XGBoost model hyperparameters by using the BOA algorithm, obtains an optimal prediction model, reduces the calculation resources while improving the efficiency of tunnel settlement prediction, and improves the accuracy and reliability of operation subway shield tunnel settlement prediction. The prediction model is packaged and installed in a portable electronic device. By inputting relevant data in the device, the tunnel settlement prediction result is obtained, which provides more scientific and accurate decision support for the design, construction and maintenance of the tunnel. Meanwhile, subsequent tunnel settlement data is continuously collected, returned to the recording module and added to the database for optimization of the prediction model.

[0130] Under the same inventive concept, a shield tunnel settlement prediction device is also involved. Figure 5 A structural diagram of a shield tunnel settlement prediction device provided by the embodiment of the present application is shown in Figure 5 The device specifically includes:

[0131] A feature screening module 100 is configured to acquire a plurality of feature factor data associated with shield tunnel settlement, and screen at least one feature factor data therefrom as target feature factor data.

[0132] A model construction module 200 is configured to construct a settlement prediction model, take the target feature factor data as the input of the settlement prediction model, take the error minimization between the predicted settlement value and the actual settlement value as the optimization target of the settlement prediction model, and iteratively update the settlement prediction model.

[0133] A settlement prediction module 300 is configured to obtain a predicted settlement value corresponding to the current target feature factor data based on the updated settlement prediction model, and take the predicted settlement value as the settlement prediction result.

[0134] The technical scheme in the embodiment of the present application fully considers the complex influence factors of tunnel settlement, iteratively updates the prediction model, ensures the reliability of the model, maximizes the generalization ability of the algorithm model, and can more accurately predict the settlement of the operation subway shield tunnel, thereby providing strong support for the operation and maintenance of the subway.

[0135] Under the same inventive concept, an electronic device is also involved. Figure 6This is a schematic diagram of the structure of an electronic device implementing the shield tunnel settlement prediction method according to embodiments of the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0136] like Figure 6 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0137] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0138] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as shield tunnel settlement prediction methods.

[0139] In some embodiments, the shield tunnel settlement prediction method can be implemented as a computer program tangibly embodied in a computer readable storage medium, e.g., storage unit 18. In some embodiments, parts or all of the computer program can be loaded and / or installed onto electronic device 10 via, e.g., ROM 12 and / or communication unit 19. When the computer program is loaded onto RAM 13 and executed by processor 11, one or more steps of the shield tunnel settlement prediction method described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to perform the shield tunnel settlement prediction method by other any suitable means, e.g., with the aid of firmware.

[0140] Various implementations of the systems and techniques described above can be realized in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a programmable logic device (PLD), a computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0141] Computer programs used to implement the processes of the present application can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the computer program, when executed, implements the functions / acts specified in the flowcharts and / or block diagrams. The computer program can be executed entirely on a machine, partially on a machine, partially on a machine as a stand-alone software package, partially on a machine and partially on a remote machine or entirely on a remote machine or server.

[0142] In the context of the present application, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. A computer-readable storage medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium can be a machine-readable signal medium. More specific examples of a machine-readable storage medium will include one or more lines of a program of instructions in a transitory signal, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0143] To provide for interaction with a user, the systems and techniques described here can be implemented on an electronic device having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the electronic device. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0144] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0145] The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system, to solve the defects of large management difficulty and weak business expansion in traditional physical host and VPS service.

[0146] It should be understood that the various forms of flow shown above can be reordered, added to, or have steps deleted, for example. The steps recited in the present disclosure can be performed in parallel, in series, or in a different order, as long as the desired results of the technical solutions of the present disclosure are achieved, and this is not limited herein.

[0147] The above only is the preferred embodiment of the present application, and does not limit the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method of predicting settlement of a shield tunnel, characterized by, The method comprises the following steps: obtaining a plurality of characteristic factor data associated with the settlement of a shield tunnel, and selecting at least one characteristic factor data as target characteristic factor data from the plurality of characteristic factor data; constructing a settlement prediction model, taking the target characteristic factor data as the input of the settlement prediction model, minimizing the error between the predicted settlement value and the actual settlement value as the optimization target of the settlement prediction model, and iteratively updating the settlement prediction model; based on the updated settlement prediction model, obtaining the predicted settlement value corresponding to the current target characteristic factor data as the settlement prediction result.

2. The method of claim 1, wherein, The selection process of the target characteristic factor data comprises: data cleaning of the plurality of characteristic factor data associated with the settlement of the shield tunnel; constructing a covariance matrix of the cleaned characteristic factor data by principal component analysis; solving the eigenvalues in the covariance matrix, and sequentially sorting each eigenvalue from large to small; sequentially calculating the contribution rate of each eigenvalue, and accumulating the contribution value according to the sorting order of each eigenvalue; when the contribution value is accumulated to a preset range, the corresponding characteristic factor data is selected as the target characteristic factor data.

3. The method of claim 1, wherein, In the process of constructing and iteratively updating the settlement prediction model, the following steps are included: selecting a training sample set and a test sample set based on the target characteristic factor data; wherein the training sample set is used to construct the settlement prediction model based on the extreme gradient boosting algorithm, and the test sample set is used to update the settlement prediction model based on the leave-one-out cross-validation algorithm; finding the optimal hyperparameter combination of the settlement prediction model in the hyperparameter optimization space based on the Bayesian optimization algorithm, which is used to determine the model structure of the settlement prediction model constructed based on the extreme gradient boosting algorithm; the optimal hyperparameter combination includes the maximum depth of the tree, the learning rate, the proportion of the sub-sample, and the minimum leaf node weight.

4. The method of claim 3, wherein, The training sample set and the test sample set belong to the same sample set, based on the model structure of the settlement prediction model, a plurality of settlement prediction models are constructed, in the construction of any one of the settlement prediction models, a group of samples are selected from the sample set as the test sample set for the construction of the model, and the remaining samples are used as the training sample set for the construction of the model; determining the objective function of the settlement prediction model based on the training sample set, and evaluating the settlement prediction model based on the test sample set; the evaluation parameters include the coefficient of determination, the root mean square error, and the mean absolute percentage error.

5. The method of claim 3, wherein, The settlement prediction model is a stepwise additive model composed of a plurality of decision tree models; the objective function of the settlement prediction model is composed of a loss function and a decision tree function, the loss function is composed of the sum of the mean square error between the predicted settlement value and the actual settlement value of each training set sample, and the decision tree function is composed of the sum of the complexity of each decision tree model.

6. The method of claim 5, wherein, The construction process of the settlement prediction model comprises: determining the model structure of the settlement prediction model based on the optimal hyperparameter combination, selecting each decision tree model based on the training set sample, and correcting the prediction error of the previous decision tree model by the subsequent decision tree model in the selection process to construct a plurality of objective functions; The target function corresponding to the minimum error between the predicted settlement value and the real settlement value is found through cyclic iteration, and is taken as the target function of the settlement prediction model.

7. The method of claim 4, wherein, The iterative updating process of the settlement prediction model comprises: calculating the average value of the evaluation results based on each group of constructed settlement prediction models; if the average value meets the preset threshold, the group of settlement prediction models with the optimal evaluation result is taken as the updated settlement prediction model; if the average value does not meet the preset threshold, the hyperparameter optimization space is adjusted, the model structure of the settlement prediction model constructed based on the extreme gradient boosting algorithm is re-determined, and the process is repeated until the average value of the evaluation results under the model structure meets the preset threshold.

8. A shield tunnel settlement prediction device characterized by, The device is configured to implement the method of any one of claims 1-7, and the device comprises: a feature screening module configured to obtain a plurality of feature factor data associated with the settlement of the shield tunnel, and screen at least one feature factor data therefrom as target feature factor data; a model construction module configured to construct a settlement prediction model, take the target feature factor data as the input of the settlement prediction model, minimize the error between the predicted settlement value and the real settlement value as the optimization target of the settlement prediction model, and iteratively update the settlement prediction model; a settlement prediction module configured to obtain the predicted settlement value corresponding to the current target feature factor data based on the updated settlement prediction model, and take the predicted settlement value as the settlement prediction result.

9. An electronic device, comprising: The electronic device comprises: at least one processor; and a memory connected in communication with the at least one processor; the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the steps of the shield tunnel settlement prediction method of any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions for enabling the processor to implement the steps of the shield tunnel settlement prediction method of any one of claims 1-7 when executed by the processor. The computer readable storage medium stores computer instructions for enabling the processor to implement the steps of the shield tunnel settlement prediction method of any one of claims 1-7 when executed by the processor.

Images