Construction deformation prediction method and device, electronic equipment and storage medium

Abstract

The application provides a construction deformation prediction method and device, electronic equipment and storage medium, and belongs to the technical field of data prediction. The method comprises the following steps: acquiring historical working condition parameters of a subway tunnel construction process; constructing a regularization term about a sample point feature space; determining a decision tree split criterion based on the regularization term, recursively dividing each feature in the sample point feature space into mutually non-overlapping hyper-rectangular regions, and constructing a decision tree; calculating posterior probability distribution parameters of each leaf node in the decision tree based on a Bayesian inference update rule; defining a structural risk penalty function in combination with the posterior probability distribution parameters to perform post-pruning on the decision tree, and obtaining an optimal decision tree; collecting real-time geological exploration parameters and real-time shield tunneling parameters of the subway tunnel construction process; inputting the real-time geological exploration parameters and the real-time shield tunneling parameters into the optimal decision tree, and outputting a monitoring point ground settlement prediction interval, thereby effectively improving the deformation prediction accuracy under complex geological conditions.

Description

Technical Field

[0001] This application belongs to the field of data prediction technology, and more specifically, relates to a construction deformation prediction method and device, electronic equipment, and storage medium. Background Technology

[0002] Subway tunnel construction typically employs the shield tunneling method. During construction, the tunnel boring machine (TBM) excavates underground while simultaneously installing precast segments to form the tunnel lining. The process involves key procedures such as soil excavation, spoil transportation, synchronous grouting, and segment assembly. Construction parameters (such as thrust, cutterhead torque, and grouting pressure) and geological conditions (such as soil strength and groundwater) jointly affect the stress balance of the surrounding strata, potentially leading to surface subsidence, tunnel convergence, and other deformations.

[0003] Construction deformation prediction is crucial in subway tunnel construction because tunnel excavation causes deformation of the surrounding strata, which may affect groundwater flow, the safety of nearby buildings or pipelines, and even lead to ground subsidence. By accurately predicting deformation during construction, preventative measures can be taken in advance to reduce potential risks and losses, ensuring the smooth progress of tunnel construction and the safety of the surrounding environment.

[0004] However, existing methods for predicting construction deformation often ignore the complex nonlinear relationship between geological and construction parameters and lack the quantification of prediction uncertainties, resulting in a sharp drop in the accuracy of ground settlement prediction. Summary of the Invention

[0005] The purpose of this application is to provide a construction deformation prediction method, device, electronic equipment, and storage medium to solve the problem that existing construction deformation prediction methods often ignore the complex nonlinear relationship between geological and construction parameters and lack the quantification of prediction uncertainty, resulting in a sharp drop in the accuracy of ground settlement prediction.

[0006] A first aspect of this application provides a method for predicting construction deformation, the method comprising:

[0007] Historical working condition parameters of the subway tunnel construction process were obtained, including geological exploration parameters, tunnel boring machine excavation parameters, and ground settlement at monitoring points at different sample points.

[0008] Based on historical operating parameters, a regularization term is constructed for the feature space of sample points;

[0009] The decision tree splitting criterion is determined based on the regularization term. The feature space of the sample points is recursively divided into non-overlapping super-rectangular regions to construct the decision tree.

[0010] The posterior probability distribution parameters of each leaf node in the decision tree are calculated based on the Bayesian inference update rule. The posterior probability distribution parameters include the posterior mean and posterior variance of the ground settlement at the monitoring point.

[0011] A structural risk penalty function is defined by combining posterior probability distribution parameters to perform post-pruning on the decision tree and obtain the optimal decision tree;

[0012] Collect real-time geological exploration parameters and real-time tunnel boring machine excavation parameters during the construction of subway tunnels;

[0013] Real-time geological exploration parameters and real-time tunnel boring machine excavation parameters are input into the optimal decision tree, and the predicted range of ground settlement at the monitoring points is output.

[0014] A second aspect of this application provides a construction deformation prediction device, the device comprising:

[0015] The acquisition module is used to acquire historical working condition parameters of the subway tunnel construction process. These historical working condition parameters include geological exploration parameters, tunnel boring machine excavation parameters, and ground settlement at monitoring points for different sample points.

[0016] The module is used to construct a regularization term for the feature space of sample points based on historical operating parameters;

[0017] The partitioning module is used to determine the decision tree splitting criteria based on the regularization term, and recursively divide each feature in the feature space of the sample points into non-overlapping super-rectangular regions to construct the decision tree;

[0018] The calculation module is used to calculate the posterior probability distribution parameters of each leaf node in the decision tree based on the Bayesian inference update rule. The posterior probability distribution parameters include the posterior mean and posterior variance of the ground settlement at the monitoring point.

[0019] The definition module is used to define the structural risk penalty function in combination with the posterior probability distribution parameters, so as to perform post-pruning on the decision tree and obtain the optimal decision tree;

[0020] The data acquisition module is used to collect real-time geological exploration parameters and real-time tunnel boring machine excavation parameters during the construction of subway tunnels.

[0021] The output module is used to input real-time geological exploration parameters and real-time tunnel boring machine excavation parameters into the optimal decision tree and output the predicted range of ground settlement at the monitoring points.

[0022] A third aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the above-described construction deformation prediction method.

[0023] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described construction deformation prediction method.

[0024] The beneficial effects of the construction deformation prediction method and device, electronic equipment, and storage medium provided in this application embodiment are as follows:

[0025] In this embodiment, by fusing historical working condition parameters with real-time data, and employing regularized decision trees and Bayesian inference techniques, the complex nonlinear relationship between geological and construction parameters is fully considered, thereby improving prediction accuracy. When constructing the decision tree, a regularization term is used to optimize the feature space partitioning, and Bayesian updates the posterior probability distribution parameters to quantify prediction uncertainty, ensuring the reliability and accuracy of the prediction results. Simultaneously, post-pruning optimizes the decision tree to avoid overfitting, further enhancing the accuracy and stability of the prediction model. This dynamically adapts to changes during construction, providing real-time ground settlement prediction intervals, significantly reducing risks caused by construction deformation. The final output prediction interval not only provides settlement estimates but also includes confidence information, significantly improving prediction reliability under complex geological conditions. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 A schematic flowchart illustrating a construction deformation prediction method provided in an embodiment of this application;

[0028] Figure 2 A structural block diagram of a construction deformation prediction device provided in one embodiment of this application;

[0029] Figure 3 This is a schematic block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0030] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0031] It is understood that in the embodiments of this application, data such as user information are involved. When the embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with relevant laws, regulations and standards.

[0032] It should be noted that the terms "first," "second," etc., used in the specification, claims, and drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in sequences other than those illustrated or described herein.

[0033] Before providing a further detailed description of the embodiments of this application, the nouns and terms involved in the embodiments of this application will be explained, and the nouns and terms involved in the embodiments of this application shall be interpreted as follows.

[0034] To make the objectives, technical solutions, and advantages of this application clearer, the following description will be provided in conjunction with the accompanying drawings and specific embodiments.

[0035] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a construction deformation prediction method provided in an embodiment of this application.

[0036] like Figure 1 As shown, the construction deformation prediction method provided in this application embodiment may include:

[0037] Historical working condition parameters of the subway tunnel construction process were obtained, including geological exploration parameters, tunnel boring machine excavation parameters, and ground settlement at monitoring points at different sample points.

[0038] Geological exploration parameters refer to the physical and chemical properties of soil, such as soil type, strength, compressibility, porosity, and groundwater level, obtained through underground soil exploration. These parameters help assess soil stability and potential settlement risks during construction. Tunneling parameters include operational data such as thrust, rotational speed, cutterhead pressure, and excavated soil volume used by the tunnel boring machine (TBM) during construction. These affect the degree of soil disturbance, construction speed, and ground settlement around the tunnel. Ground settlement monitoring refers to the vertical displacement of the ground measured at a specific location (which can be in any direction around the construction path) during tunnel construction. Settlement changes at this location are monitored using settlement instruments to assess the impact of tunnel construction on the surrounding environment and safety risks.

[0039] It should be noted that by collecting historical working condition data from different locations during the subway tunnel construction process (each location's historical working condition data constitutes a sample point), including geological exploration parameters, tunnel boring machine parameters, and ground settlement data at monitoring points, fundamental information was provided for subsequent deformation prediction. This historical working condition data reflects the actual conditions under different construction stages and geological conditions, helping the model accurately identify the relationships between various parameters and improve prediction accuracy. By integrating multi-dimensional data, a comprehensive understanding of the impact of construction on the geological environment can be achieved, providing reliable data support for subsequent construction deformation prediction.

[0040] Based on historical operating parameters, a regularization term is constructed for the feature space of sample points.

[0041] Regularization is a machine learning technique used to prevent overfitting. In this method, regularization constrains model complexity by measuring the differences between sample points and the discrepancies between their predicted values, enabling the model to generalize better. By incorporating regularization, we can ensure that the decision tree not only fits the training data but also avoids over-reliance on certain local features, thereby improving the model's performance on unseen data.

[0042] In one possible implementation, the features in the sample point feature space include geological exploration parameters and tunnel boring machine (TBM) excavation parameters. Based on historical working condition parameters, a regularization term is constructed for the sample point feature space, specifically including:

[0043] Calculate the similarity weights of the feature data between each sample point.

[0044] Specifically, similarity weight refers to the similarity weight between any two sample points in terms of tunneling parameters and geological exploration parameters.

[0045] The specific formula for calculating similarity weight is as follows:

[0046] .

[0047] .

[0048] .

[0049] in, Represents sample points i With sample points j Similarity weights between them This represents the natural exponential function. and Representing sample points respectively i and sample points j Feature data, express and Mahalanobis distance between them Indicates bandwidth parameter, This represents the covariance matrix that describes the correlation of feature data among various sample points. This represents the total number of sample points. This represents the mean vector of feature data for all sample points, with the subscript T indicating transpose.

[0050] Optionally, the bandwidth parameter can be set to a value ranging from 0.1 to 2.0.

[0051] It should be noted that this step calculates similarity weights between sample points, measures the differences in sample features based on Mahalanobis distance, and controls the sensitivity of similarity through a bandwidth parameter. This method can assign different weights according to the similarity between samples, ensuring that similar samples have a greater impact on the prediction results, while distant samples have a smaller impact, thereby improving the accuracy and generalization ability of the model. In this way, it is possible to better capture the inherent patterns in the data and avoid model overfitting.

[0052] The regularization term is calculated by combining the similarity weights.

[0053] The formula for calculating the regularization term is as follows:

[0054]

[0055] in, and These represent decision tree pairs. and The output is the predicted ground subsidence at the monitoring points. Representation and prediction function of decision trees f Related regularization terms.

[0056] It should be noted that by constructing a regularization term for the feature space of sample points based on historical working condition parameters, the similarity weights between sample points are first calculated to reflect the similarity of feature data. Then, the regularization term is calculated using these weights to balance the differences in predicted values ​​between sample points and avoid overfitting the model to individual samples. This process effectively improves the robustness of the model, ensuring that deformation prediction results maintain high accuracy and reliability even under different construction environments.

[0057] The decision tree splitting criterion is determined based on the regularization term. The feature space of the sample points is recursively divided into non-overlapping super-rectangular regions to construct the decision tree.

[0058] The decision tree splitting criterion is used to determine how to divide the feature space of sample points into different sub-regions. A hyperrectangular region refers to a multi-dimensional rectangular region in the feature space formed by the value ranges of each feature; these regions do not overlap, ensuring that each sample point is accurately classified. This step determines the splitting criterion based on a regularization term, recursively dividing sample points into different hyperrectangular regions, progressively constructing the decision tree. This process ensures that the model can efficiently learn the feature relationships between samples and reduces the risk of overfitting. Simultaneously, the introduction of a regularization term effectively balances model complexity and prediction accuracy, improving the model's generalization ability.

[0059] In one possible implementation, the decision tree splitting criterion is determined based on a regularization term, and each feature in the feature space of the sample points is recursively divided into non-overlapping hyperrectangular regions to construct a decision tree, specifically including:

[0060] At any split node in the decision tree, calculate the sum of the weighted variances of the features after the candidate split point divides them into the left and right child nodes, and obtain the global prediction variance term.

[0061] Determine the regularization terms at the candidate split points.

[0062] A split gain function is constructed by combining the regularization term and the global prediction variance term.

[0063] The specific formula for the split gain function is as follows:

[0064] .

[0065] in, Indicates in features j According to the alternative split points s The splitting gain function obtained during the splitting process. This represents the global predicted variance of ground subsidence at monitoring points before the split. Indicates the regularization weight. This represents the total number of sample points in the current node. I This represents the set of sample point indices in the current node. and These represent the decision tree after splitting according to the candidate split points, respectively. p sample points and the q sample points The output is the predicted ground subsidence at the monitoring points. Indicates the first in the current node p sample points and the q sample points The regularization term between them, and These represent the number of sample points for the left and right child nodes, respectively. This represents the variance of the predicted ground settlement at the monitoring point corresponding to the left child node. This represents the variance of the predicted ground settlement at the monitoring point corresponding to the right child node. This represents the global prediction variance term for ground subsidence at the monitoring points after the split. Indicates the alternative splitting points s The regularization term is as follows.

[0066] The global prediction variance of ground subsidence at monitoring points is calculated based on the ground subsidence at monitoring points predicted by the decision tree before splitting based on the characteristic data of sample points, and the ground subsidence at monitoring points in historical operating parameters.

[0067] The split gain function combines prediction variance and a regularization term to maximize prediction accuracy after each split while reducing the risk of overfitting. By calculating the difference between sample points and the regularization term at each split node, the model ensures a balance between accuracy and stability when partitioning the sample space. The split gain function not only enhances the predictive power of the decision tree but also effectively controls the model's complexity through the regularization term, thereby improving its generalization ability and enabling the model to maintain high prediction accuracy in various construction environments.

[0068] Optionally, the regularization weight can take values ​​in the range of 0.01 to 10.

[0069] Specifically, this step involves calculating the weighted variance and regularization term for each candidate split point at each split node of the decision tree, and then constructing a split gain function by combining the global prediction variance and the regularization term. The split gain function optimizes the choice of node splits, ensuring that each split effectively reduces prediction variance and balances model complexity. The introduction of the regularization term helps avoid overfitting, improves the model's generalization ability, and avoids excessive reliance on local data. Through this process, the decision tree can accurately partition the feature space of sample points, improving prediction accuracy and stability, especially when dealing with complex, nonlinear data, ensuring efficient and reliable prediction results.

[0070] At each feature, different features and their corresponding candidate split points are traversed. The feature and candidate split point that maximizes the split gain function are selected as the optimal splitting strategy until the preset stopping condition is met, and the resulting super-rectangular region is used as the decision tree.

[0071] The preset stopping condition can be set to the number of sample points in the current node being less than a preset number of sample points, or the growth depth of the decision tree reaching a preset growth depth. It should be noted that those skilled in the art can set the preset number of sample points and the preset growth depth according to actual needs, and this invention does not limit this. Optionally, the preset number of sample points can be set to 20, and the preset growth depth can be set to 10 layers.

[0072] The posterior probability distribution parameters of each leaf node in the decision tree are calculated based on the Bayesian inference update rule. The posterior probability distribution parameters include the posterior mean and posterior variance of the ground subsidence at the monitoring point.

[0073] The Bayesian inference update rule is a method for calculating posterior probabilities based on prior information and new observation data. It progressively updates the model's estimates of unknown parameters by combining prior distributions and actual data. Leaf nodes are terminal nodes in the decision tree, used to store the final prediction results. Posterior probability distribution parameters include the posterior mean and variance of ground settlement at monitoring points, used to quantify the uncertainty of the prediction results. By applying the Bayesian inference rule to update the posterior probability distribution parameters of each leaf node in the decision tree, the prediction results become more accurate and reliable. By calculating the posterior mean and variance, not only can the predicted value be obtained, but the uncertainty of the prediction can also be assessed. This method effectively quantifies the model's prediction risk and can dynamically adjust to adapt to different construction environments, improving the model's accuracy and robustness.

[0074] In one possible implementation, the posterior probability distribution parameters of each leaf node in the decision tree are calculated based on the Bayesian inference update rule, specifically including:

[0075] The number of sample points falling into each leaf node of the decision tree and the average ground subsidence of the monitoring points corresponding to the sample points are counted.

[0076] By combining the number of sample points and the mean ground subsidence at the monitoring points, the posterior mean and posterior variance are calculated based on the Bayesian inference update rule.

[0077] The specific formula for calculating the posterior mean is as follows:

[0078] .

[0079] in, Leaf nodes m The posterior mean, Leaf nodes m The number of sample points, Leaf nodes m The average ground subsidence at the monitoring points and Let represent the prior mean and prior confidence strength, respectively.

[0080] The specific formula for calculating the posterior variance is as follows:

[0081] .

[0082] in, Leaf nodes m The posterior variance, and Let these represent the shape parameter and the scaling parameter of the prior variance, respectively. Indicates falling into a leaf node m sample points i The corresponding ground subsidence at the monitoring point.

[0083] The prior mean is a preliminary estimate of the target variable based on previous experience or assumptions, representing a prediction of ground settlement at the monitoring point without any new data. Prior confidence strength indicates the degree of trust in the prior mean; higher confidence strength means a greater impact. The prior variance shape parameter and prior variance scale parameter control the shape and scale of the prior variance, respectively, reflecting the expected uncertainty regarding data changes. The posterior mean and variance are updated using Bayesian inference rules combined with the number of sample points and the mean settlement, providing more accurate predictions. This method allows for adjustments to previous predictions based on actual data, making the model more adaptable to new data. Simultaneously, by calculating the posterior variance, uncertainty is quantified, ensuring that the prediction results not only possess accuracy but also reflect reliability. This update mechanism improves the accuracy and robustness of predictions, especially in complex construction environments, enabling dynamic adjustments and reducing prediction errors.

[0084] Specifically, in this process, the Bayesian inference update rule progressively optimizes the model's prediction results by combining prior information with new observational data. First, the number of sample points falling into each leaf node and the mean settlement of their corresponding monitoring points are calculated as the basis for the update. Then, the posterior mean and posterior variance are calculated using Bayes' theorem. These two parameters effectively reflect the reliability and uncertainty of the prediction results for each leaf node. The posterior mean provides the best estimate of the settlement, while the posterior variance quantifies the uncertainty of the prediction, helping the model adjust its prediction strategy under different construction environments. This method not only enhances the accuracy of the prediction but also improves the model's ability to cope with uncertainties, ensuring high prediction accuracy and robustness even under complex geological conditions. By continuously updating the parameters of the leaf nodes, the model can flexibly adapt to changes that may occur during construction, improving the overall prediction performance.

[0085] By combining the posterior probability distribution parameters, a structural risk penalty function is defined to perform post-pruning on the decision tree, thereby obtaining the optimal decision tree.

[0086] In one possible implementation, a structural risk penalty function is defined in conjunction with posterior probability distribution parameters to perform post-pruning on the decision tree, thereby obtaining the optimal decision tree. Specifically, this includes:

[0087] By combining the posterior variance in the posterior probability distribution parameters, the sum of the differential entropy of all leaf nodes in the decision tree is calculated to quantify the prediction uncertainty of the decision tree.

[0088] The total number of leaf nodes in the decision tree is counted, and the decision tree complexity penalty term is calculated based on the total number of sample points.

[0089] The sum of differential entropy measures the uncertainty of all leaf nodes in the decision tree, reflecting the degree of uncertainty in the predicted value of each leaf node. The decision tree complexity penalty term controls the complexity of the decision tree model, preventing overfitting during training. The more leaf nodes, the higher the model complexity. Calculating the complexity penalty term in conjunction with the total number of sample points limits tree overgrowth, preventing overfitting of the model to the training data and ensuring the model's generalization ability in real-world applications. The combination of the sum of differential entropy and the complexity penalty term balances the model's predictive accuracy and complexity. The sum of differential entropy quantifies the model's uncertainty, while the complexity penalty term controls the tree's size and complexity. This balance allows for the construction of a decision tree that accurately predicts without easily overfitting, thereby enhancing the model's stability and generalization ability in real-world applications and ensuring its adaptability to new data.

[0090] By combining the sum of differential entropy and the decision tree complexity penalty term, a structural risk penalty function is established.

[0091] The structural risk penalty function is as follows:

[0092] .

[0093] in, Let T represent the set of leaf nodes of decision tree T. This represents the structural risk penalty function value of decision tree T. Represents pi (π). Represents the natural logarithm. Represents the natural constant. This represents the total number of leaf nodes in the decision tree. This represents the total number of sample points. Indicates the penalty coefficient. This represents the sum of differential entropy. This represents the penalty term for the complexity of the decision tree.

[0094] Optionally, the penalty coefficient can be set to a value between 0.01 and 1.

[0095] It should be noted that the structural risk penalty function is established by combining differential entropy and a decision tree complexity penalty term, aiming to balance the model's prediction accuracy and complexity. Differential entropy reflects the uncertainty of prediction, while the complexity penalty term constrains the growth of the decision tree, preventing overfitting. In this way, it can be ensured that the decision tree improves accuracy without reducing its generalization ability due to overly complex structure, thereby enhancing the model's stability and reliability.

[0096] Traverse the non-leaf nodes of the decision tree from bottom to top, calculate the structural risk penalty function value of the child node state of each non-leaf node before pre-merging, and perform pre-merging if the structural risk penalty function value is less than the structural risk penalty function value before pre-merging, thus completing the pruning operation of the decision tree.

[0097] The purpose of pre-merging is to reduce the complexity of the decision tree by merging subtrees, thereby reducing the risk of overfitting. By calculating and comparing the structural risk penalty function values ​​before and after merging, a decision is made whether to merge, thus simplifying the decision tree structure and improving model performance.

[0098] Specifically, this step optimizes the decision tree structure by defining a structural risk penalty function and improves the model's accuracy and robustness during post-pruning. First, by combining posterior variance with the differential entropy of leaf nodes, the uncertainty of prediction is quantified, ensuring the decision tree can make more stable predictions when faced with uncertain data. Next, the total number of leaf nodes in the decision tree is counted, and a penalty term for model complexity is calculated based on the total number of sample points to prevent the decision tree from becoming overly complex. Then, a structural risk penalty function is established to balance prediction accuracy and model complexity, ensuring the decision tree does not overfit. Finally, by traversing non-leaf nodes in a bottom-up manner, the risk penalty function values ​​before and after pruning are calculated, and pruning is performed as necessary to obtain the optimal decision tree. This method, by combining the penalty of prediction uncertainty with the penalty of model complexity, effectively avoids overfitting, improves the generalization ability of the decision tree, and enables the model to better adapt to different construction conditions and provide reliable prediction results.

[0099] Real-time geological exploration parameters and real-time tunnel boring machine excavation parameters are collected during the construction of subway tunnels.

[0100] Real-time geological exploration parameters and real-time tunnel boring machine excavation parameters are input into the optimal decision tree, and the predicted range of ground settlement at the monitoring points is output.

[0101] In one possible implementation, real-time geological exploration parameters and real-time tunnel boring machine excavation parameters are input into the optimal decision tree, and the predicted range of ground settlement at the monitoring points is output, specifically including:

[0102] Real-time geological exploration parameters and real-time tunnel boring machine excavation parameters are input into the optimal decision tree to obtain the predicted value of the optimal decision tree.

[0103] Calculate the posterior mean and posterior variance of the predicted values.

[0104] Marginal error is calculated by combining posterior variance.

[0105] Marginal error, calculated in conjunction with posterior variance, quantifies the uncertainty of a model's predictions. It reflects the degree of uncertainty the model has regarding a specific predicted value and is typically related to the model's confidence level. A larger marginal error indicates lower confidence in the prediction, suggesting a need for more information to reduce uncertainty. Marginal error, calculated based on posterior variance, quantifies the reliability and uncertainty of the predicted value. In regression analysis, a larger marginal error usually indicates a significant bias in the model's predictions of certain input data, requiring further model optimization or the acquisition of more sample data to improve prediction accuracy.

[0106] By correcting the posterior mean for marginal error, the predicted range of ground subsidence at the monitoring point is obtained.

[0107] The specific formula for calculating the predicted range of ground subsidence at monitoring points is as follows:

[0108] .

[0109]

[0110] in, This represents real-time operating data, including real-time geological exploration parameters and real-time tunnel boring machine excavation parameters. x The predicted range of ground subsidence at the monitoring points and These represent the target leaf nodes in the optimal decision tree. Output of real-time operating condition data x The posterior mean and posterior variance, Represents the target leaf node The number of historical sample points Indicates the significance level. Indicates the relationship with the target leaf node Number of historical sample points The relevant degrees of freedom Degrees of freedom v The lower confidence level is Error amplification factor, This represents the marginal error.

[0111] The specific value of the error amplification factor can be obtained from the t-distribution table. The target leaf node is the terminal node in the decision tree, representing the final prediction result for a specific sample. The number of historical sample points refers to the number of past sample data stored in this leaf node, used to calculate the posterior mean and variance, reflecting the distribution of the sample data. The degrees of freedom represent statistics related to the number of historical sample points of the leaf node, reflecting the uncertainty of the sample data. The ground settlement prediction interval for monitoring points, by combining the posterior mean and variance with the error amplification factor, provides not only an estimate of the settlement but also its possible error range. In this way, the prediction interval can quantify the uncertainty of the prediction, ensuring higher reliability and transparency in the prediction of settlement during construction. This method can cope with changes in different construction environments, ensuring that the prediction results have stronger adaptability and accuracy.

[0112] Specifically, this step first obtains the predicted values, then calculates the posterior mean and variance to quantify the uncertainty of the prediction. Next, the marginal error is calculated using the posterior variance to further refine the predicted values, resulting in the final predicted range for ground settlement at the monitoring points. This prediction range not only provides an estimate of the settlement but also includes the error range, enabling the model to provide confident predictions under different construction conditions. In this way, the model not only provides accurate predictions but also quantifies the reliability of the prediction results.

[0113] In one possible implementation, it also includes:

[0114] If the upper limit of the predicted ground subsidence range at the monitoring point exceeds the preset upper limit, an early warning signal will be output.

[0115] It should be noted that those skilled in the art can set the upper limit of the preset interval according to actual needs, and this invention does not limit this. Specifically, when the upper limit of the predicted ground settlement interval at the monitoring point exceeds the set safety threshold, i.e., the upper limit of the preset interval, the system will automatically issue an early warning signal to remind relevant personnel to take preventive measures. This is to avoid potential risks and ensure that ground settlement does not exceed the safe range during construction, thereby ensuring construction safety and the stability of the surrounding environment.

[0116] In practical applications, the construction of the optimal decision tree involves pre-operation based on historical data and does not require repetition each time. During actual execution, the process only requires following these steps: collecting real-time geological exploration parameters and real-time tunnel boring machine (TBM) excavation parameters during subway tunnel construction; inputting these parameters into the optimal decision tree; and outputting the predicted ground settlement range for the monitoring points. By integrating historical working condition data from subway tunnel construction, including geological exploration, TBM excavation, and ground settlement information, a precise construction deformation prediction model is constructed. Through regularized decision trees combined with Bayesian inference update rules, the model can not only efficiently identify key feature relationships during construction but also dynamically adjust prediction results, taking into account uncertainties in actual construction. When making predictions, the model not only outputs estimated ground settlement values ​​but also quantifies the uncertainty of the prediction by calculating the posterior mean and variance, making the prediction results more reliable. This method effectively avoids overfitting, improves the model's generalization ability, and allows for timely adjustments to deformations under different construction environments based on real-time data, thereby achieving more accurate and stable predictions and ensuring safety and controllability during the construction process.

[0117] In this embodiment, by fusing historical working condition parameters with real-time data, and employing regularized decision trees and Bayesian inference techniques, the complex nonlinear relationship between geological and construction parameters is fully considered, thereby improving prediction accuracy. When constructing the decision tree, a regularization term is used to optimize the feature space partitioning, and Bayesian updates the posterior probability distribution parameters to quantify prediction uncertainty, ensuring the reliability and accuracy of the prediction results. Simultaneously, post-pruning optimizes the decision tree to avoid overfitting, further enhancing the accuracy and stability of the prediction model. This dynamically adapts to changes during construction, providing real-time ground settlement prediction intervals, significantly reducing risks caused by construction deformation. The final output prediction interval not only provides settlement estimates but also includes confidence information, significantly improving prediction reliability under complex geological conditions.

[0118] Based on the same inventive concept, this application also provides a construction deformation prediction device for implementing the construction deformation prediction method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more embodiments of the construction deformation prediction device provided below can be found in the limitations of the construction deformation prediction method described above, and will not be repeated here.

[0119] This application provides a construction deformation prediction device, such as... Figure 2 As shown, the construction deformation prediction device 20 includes:

[0120] The acquisition module 201 is used to acquire historical working condition parameters of the subway tunnel construction process. These historical working condition parameters include geological exploration parameters, tunnel boring machine excavation parameters, and ground settlement at monitoring points for different sample points.

[0121] Module 202 is used to construct a regularization term for the feature space of sample points based on historical operating parameters.

[0122] The partitioning module 203 is used to determine the decision tree splitting criteria based on the regularization term, and recursively divide each feature in the feature space of the sample points into non-overlapping super-rectangular regions to construct a decision tree.

[0123] The calculation module 204 is used to calculate the posterior probability distribution parameters of each leaf node in the decision tree based on the Bayesian inference update rule. The posterior probability distribution parameters include the posterior mean and posterior variance of the ground settlement at the monitoring point.

[0124] Module 205 is defined to combine posterior probability distribution parameters to define a structural risk penalty function, so as to perform post-pruning on the decision tree and obtain the optimal decision tree.

[0125] The data acquisition module 206 is used to acquire real-time geological exploration parameters and real-time tunnel boring machine excavation parameters during the construction of the subway tunnel.

[0126] The output module 207 is used to input real-time geological exploration parameters and real-time tunnel boring machine excavation parameters into the optimal decision tree and output the predicted range of ground settlement at the monitoring point.

[0127] See Figure 3 , Figure 3 This is a schematic block diagram of an electronic device provided according to an embodiment of this application. Figure 3 The electronic device 300 in this embodiment may include one or more processors 301, one or more input devices 302, one or more output devices 303, and one or more memories 304. The processors 301, input devices 302, output devices 303, and memories 304 communicate with each other via a communication bus 305. The memories 304 store computer programs, including program instructions. The processors 301 execute the program instructions stored in the memories 304. Specifically, the processors 301 are configured to invoke the program instructions to perform the functions of each module / unit in the above-described device embodiments, for example... Figure 3 The functions of the acquisition module 201, construction module 202, partitioning module 203, calculation module 204, definition module 205, acquisition module 206, and output module 207 are shown.

[0128] It should be understood that, in the embodiments of this application, the processor 301 may be a central processing unit (CPU), but it may also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0129] Input device 302 may include a touchpad, a fingerprint sensor (for collecting the user's fingerprint information and fingerprint orientation information), a microphone, etc., and output device 303 may include a display (LCD, etc.), a speaker, etc.

[0130] The memory 304 may include read-only memory and random access memory, and provides instructions and data to the processor 301. A portion of the memory 304 may also include non-volatile random access memory. For example, the memory 304 may also store device type information.

[0131] In specific implementations, the processor 301, input device 302, and output device 303 described in the embodiments of this application can execute the implementation method described in the construction deformation prediction method provided in the embodiments of this application, or they can execute the implementation method of the electronic device described in the embodiments of this application, which will not be repeated here.

[0132] In another embodiment of this application, a computer-readable storage medium is provided. This computer-readable storage medium stores a computer program, which includes program instructions. When executed by a processor, the program instructions implement all or part of the processes in the methods described above. Alternatively, the computer program can instruct related hardware to complete the process. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include any entity or device capable of carrying computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0133] The computer-readable storage medium can be an internal storage unit of the electronic device in any of the foregoing embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., provided on the electronic device. Furthermore, the computer-readable storage medium can include both internal and external storage units of the electronic device. The computer-readable storage medium is used to store computer programs and other programs and data required by the electronic device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0134] Those skilled in the art will recognize that the modules / units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0135] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the electronic devices and units described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0136] In the several embodiments provided in this application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of modules / units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules, units, or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces or modules / units, or it may be an electrical, mechanical, or other form of connection.

[0137] The modules / units described as separate components may or may not be physically separate. Similarly, the components shown as modules / units may or may not be physical modules / units; they may be located in one place or distributed across multiple network modules / units. Some or all of the modules / units can be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0138] Furthermore, the functional modules / units in the various embodiments of this application can be integrated into one processing module / unit, or each module / unit can exist physically separately, or two or more modules / units can be integrated into one module / unit. The integrated modules / units described above can be implemented in hardware or in the form of software functional modules / units.

[0139] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for predicting construction deformation, characterized in that, The method includes: Historical working condition parameters of the subway tunnel construction process are obtained, including geological exploration parameters, tunnel boring machine excavation parameters and ground settlement at monitoring points at different sample points; Based on the historical operating parameters, a regularization term is constructed for the feature space of the sample points; Based on the regularization term, the decision tree splitting criterion is determined, and each feature in the feature space of the sample points is recursively divided into non-overlapping super-rectangular regions to construct a decision tree. The posterior probability distribution parameters of each leaf node in the decision tree are calculated based on the Bayesian inference update rule, wherein the posterior probability distribution parameters include the posterior mean and posterior variance of the ground subsidence at the monitoring point. A structural risk penalty function is defined based on the posterior probability distribution parameters to perform post-pruning on the decision tree and obtain the optimal decision tree; Collect real-time geological exploration parameters and real-time tunnel boring machine excavation parameters during the construction of subway tunnels; The real-time geological exploration parameters and the real-time tunnel boring machine excavation parameters are input into the optimal decision tree, and the predicted range of ground settlement at the monitoring point is output. Specifically, the step of determining the decision tree splitting criterion based on the regularization term, recursively dividing each feature in the feature space of the sample points into non-overlapping super-rectangular regions, and constructing a decision tree, includes: At any split node in the decision tree, calculate the sum of the weighted variances of the features after the candidate split point divides them into the left and right child nodes, and obtain the global prediction variance term. Determine the regularization terms at the candidate split points; By combining the regularization term and the global prediction variance term, a split gain function is constructed; The specific formula for the split gain function is as follows: ； in, Indicates in features j According to the alternative split points s The splitting gain function obtained during the splitting process. This represents the global predicted variance of ground subsidence at monitoring points before the split. Indicates the regularization weight. This represents the total number of sample points in the current node. I This represents the set of sample point indices in the current node. and These represent the decision tree after splitting according to the candidate split points, respectively. p sample points and the q sample points The output is the predicted ground subsidence at the monitoring points. Indicates the first in the current node p sample points and the q sample points The regularization term between them, and These represent the number of sample points for the left and right child nodes, respectively. This represents the variance of the predicted ground settlement at the monitoring point corresponding to the left child node. This represents the variance of the predicted ground settlement at the monitoring point corresponding to the right child node. This represents the global prediction variance term for ground subsidence at the monitoring points after the split. Indicates the alternative splitting points s The regularization term below; At each of the aforementioned features, different features and their corresponding candidate splitting points are traversed. The features and candidate splitting points that maximize the splitting gain function are selected as the optimal splitting strategy until a preset stopping condition is met, and the resulting super-rectangular region is used as the decision tree. Specifically, defining a structural risk penalty function based on the posterior probability distribution parameters to perform post-pruning on the decision tree to obtain the optimal decision tree includes: By combining the posterior variance in the posterior probability distribution parameters, the sum of the differential entropy of all leaf nodes in the decision tree is calculated to quantify the prediction uncertainty of the decision tree. Count the total number of leaf nodes in the decision tree, and calculate the decision tree complexity penalty term based on the total number of sample points; The structural risk penalty function is established by combining the sum of the differential entropy and the decision tree complexity penalty term; Traverse the non-leaf nodes of the decision tree from bottom to top, calculate the structural risk penalty function value of the child node state of each non-leaf node before pre-merging, and if the structural risk penalty function value is less than the structural risk penalty function value before pre-merging, then perform pre-merging to complete the pruning operation of the decision tree.

2. The construction deformation prediction method as described in claim 1, characterized in that, The features in the sample point feature space include geological exploration parameters and tunnel boring machine excavation parameters. The construction of a regularization term for the sample point feature space based on the historical working condition parameters specifically includes: Calculate the similarity weights of the feature data among the various sample points; The regularization term is calculated by combining the similarity weights.

3. The construction deformation prediction method as described in claim 1, characterized in that, The calculation of the posterior probability distribution parameters of each leaf node in the decision tree based on the Bayesian inference update rule specifically includes: The number of sample points falling into each leaf node of the decision tree and the average ground subsidence of the monitoring points corresponding to the sample points are counted. Based on the number of sample points and the average ground subsidence at the monitoring points, the posterior mean and posterior variance are calculated using the Bayesian inference update rule.

4. The construction deformation prediction method as described in claim 1, characterized in that, The process of inputting the real-time geological exploration parameters and the real-time tunnel boring machine excavation parameters into the optimal decision tree and outputting the predicted range of ground settlement at the monitoring points specifically includes: The real-time geological exploration parameters and the real-time tunnel boring machine excavation parameters are input into the optimal decision tree to obtain the predicted value of the optimal decision tree; Calculate the posterior mean and posterior variance of the predicted values; The marginal error is calculated by combining the posterior variance. The predicted range of ground subsidence at the monitoring point is obtained by correcting the posterior mean using the marginal error.

5. The construction deformation prediction method as described in claim 1, characterized in that, Also includes: If the upper limit of the predicted ground subsidence range at the monitoring point exceeds the preset upper limit, an early warning signal will be output.

6. A construction deformation prediction device, characterized in that, The device includes: The acquisition module is used to acquire historical working condition parameters of the subway tunnel construction process, wherein the historical working condition parameters include geological exploration parameters, tunnel boring machine excavation parameters and ground settlement at monitoring points at different sample points; The construction module is used to construct a regularization term for the feature space of sample points based on the historical operating condition parameters. The partitioning module is used to determine the decision tree splitting criterion based on the regularization term, and recursively divide each feature in the feature space of the sample points into non-overlapping super-rectangular regions to construct a decision tree; The calculation module is used to calculate the posterior probability distribution parameters of each leaf node in the decision tree based on the Bayesian inference update rule, wherein the posterior probability distribution parameters include the posterior mean and posterior variance of the ground subsidence at the monitoring point; The definition module is used to define a structural risk penalty function in combination with the posterior probability distribution parameters, so as to perform post-pruning on the decision tree and obtain the optimal decision tree; The data acquisition module is used to collect real-time geological exploration parameters and real-time tunnel boring machine excavation parameters during the construction of subway tunnels. The output module is used to input the real-time geological exploration parameters and the real-time tunnel boring machine excavation parameters into the optimal decision tree and output the predicted range of ground settlement at the monitoring point. Specifically, when the partitioning module determines the decision tree splitting criterion based on the regularization term and recursively divides each feature in the feature space of the sample points into non-overlapping hyperrectangular regions to construct the decision tree, it is used for: At any split node in the decision tree, calculate the sum of the weighted variances of the features after the candidate split point divides them into the left and right child nodes, and obtain the global prediction variance term. Determine the regularization terms at the candidate split points; By combining the regularization term and the global prediction variance term, a split gain function is constructed; The specific formula for the split gain function is as follows: ； in, Indicates in features j According to the alternative split points s The splitting gain function obtained during the splitting process. This represents the global predicted variance of ground subsidence at monitoring points before the split. Indicates the regularization weight. This represents the total number of sample points in the current node. I This represents the set of sample point indices in the current node. and These represent the decision tree after splitting according to the candidate split points, respectively. p sample points and the q sample points The output is the predicted ground subsidence at the monitoring points. Indicates the first in the current node p sample points and the q sample points The regularization term between them, and These represent the number of sample points for the left and right child nodes, respectively. This represents the variance of the predicted ground settlement at the monitoring point corresponding to the left child node. This represents the variance of the predicted ground settlement at the monitoring point corresponding to the right child node. This represents the global prediction variance term for ground subsidence at the monitoring points after the split. Indicates the alternative splitting points s The regularization term below; At each of the aforementioned features, different features and their corresponding candidate splitting points are traversed. The features and candidate splitting points that maximize the splitting gain function are selected as the optimal splitting strategy until a preset stopping condition is met, and the resulting super-rectangular region is used as the decision tree. Specifically, when the definition module defines the structural risk penalty function in conjunction with the posterior probability distribution parameters to perform post-pruning on the decision tree and obtain the optimal decision tree, it is used for: By combining the posterior variance in the posterior probability distribution parameters, the sum of the differential entropy of all leaf nodes in the decision tree is calculated to quantify the prediction uncertainty of the decision tree. Count the total number of leaf nodes in the decision tree, and calculate the decision tree complexity penalty term based on the total number of sample points; The structural risk penalty function is established by combining the sum of the differential entropy and the decision tree complexity penalty term; Traverse the non-leaf nodes of the decision tree from bottom to top, calculate the structural risk penalty function value of the child node state of each non-leaf node before pre-merging, and if the structural risk penalty function value is less than the structural risk penalty function value before pre-merging, then perform pre-merging to complete the pruning operation of the decision tree.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 5.

Images