Tunnel disease structure deformation prediction method and system based on fusion model

Abstract

The application discloses a tunnel disease structure deformation prediction method and system based on a fusion model. The method first establishes a disease structure deformation prediction model for each kind of shield tunnel disease by using multiple different neural network models. Then, a fusion model of multiple disease structure deformation prediction models is established based on a weight analysis method, and the fusion model is used to predict the structure deformation of all shield tunnel diseases. In the fusion model, the prediction results are obtained by weighted summation of the prediction values according to the errors of each disease structure deformation prediction model, and the fusion prediction results are obtained by weighted summation of the prediction results. The fusion model established by the application is suitable for structure deformation prediction of multiple diseases of a tunnel, has stronger universality, and has higher prediction accuracy.

Description

Technical Field

[0001] This invention relates to the field of tunnel defect prediction, and in particular to a method and system for predicting structural deformation of tunnel defects based on a fusion model. Background Technology

[0002] With the advancement and increasing maturity of tunnel boring machine (TBM) technology, more and more tunnels are being constructed using this method. Currently, a large number of TBM tunnels are in operation. However, as their service life increases, various degrees and types of structural defects have emerged. Therefore, the detection, identification, and treatment of these defects have become urgent problems to be solved. Because tunnels occupy a crucial position in transportation routes, their downtime is relatively short, limiting the time available for maintenance and inspection. For example, in urban rail transit (metro) tunnels constructed using the TBM method, defect detection and treatment must be carried out after midnight, when the environment is relatively harsh. Most river-crossing tunnels constructed using the TBM method are often important hubs connecting two banks, with high traffic volume. Their maintenance must be carried out during periods of lower traffic volume, and efficiency is also crucial. Due to the vital role of TBM tunnels, there are high requirements for the time and efficiency of their defect maintenance, and traditional methods are insufficient to meet these practical needs.

[0003] Currently, with the development of detection technology, different tunnel defects can be predicted by combining different defect information obtained from detection with artificial intelligence algorithms. However, research usually uses different methods to conduct prediction studies based on different defect characteristics, which makes it difficult for prediction methods to be universally applicable under different defect conditions. Summary of the Invention

[0004] Purpose of the invention: The purpose of this invention is to provide a method and system for high-precision prediction of structural deformation of tunnels under various diseases using a single model, thereby solving the problem in the background technology that different disease characteristics require different methods for prediction research, resulting in poor model universality.

[0005] Technical solution: The tunnel structural deformation prediction method based on a fusion model described in this invention includes the following steps:

[0006] For each type of shield tunnel defect, multiple different neural network models are used to establish a defect structural deformation prediction model;

[0007] A fusion model based on weighted analysis is established to predict the structural deformation of multiple defects in shield tunnels. The fusion model is then used to predict the structural deformation of all defects in shield tunnels.

[0008] In the fusion model, the predicted values ​​are weighted and summed based on the error of each disease structure deformation prediction model to obtain the prediction result, and the prediction results are weighted and summed to obtain the fusion prediction result.

[0009] Furthermore, the fusion model based on weighted analysis to establish multiple disease structure deformation prediction models includes:

[0010] From multiple disease structure deformation prediction models, a single optimal prediction model is obtained using the K-fold cross-validation method.

[0011] The prediction results of the fusion model are compared with the prediction results of the single optimal prediction model for each shield tunnel defect to determine the value of the fusion weight in the fusion model.

[0012] Furthermore, among the plurality of disease structure deformation prediction models, obtaining a single optimal prediction model according to the K-fold cross-validation method includes: for each disease structure deformation prediction model, determining a single optimal prediction model by calculating the mean square error and mean absolute error between the predicted value and the measured value on the dataset.

[0013] Furthermore, the formulas for calculating the mean square error (MSE) and the mean absolute error (MAE) are as follows:

[0014]

[0015]

[0016] Where y j y represents the predicted value of the structural deformation prediction model for disease. j t represents the measured value, and t represents the number of data points in the dataset.

[0017] Furthermore, in the fusion model, the predicted values ​​are weighted and summed based on the errors of each disease structure deformation prediction model to obtain the prediction result. The weighted summation of the prediction results to obtain the fusion prediction result includes: calculating a first prediction result based on the mean square error between the predicted value and the measured value of each disease structure deformation prediction model; calculating a second prediction result based on the mean absolute error between the predicted value and the measured value of each disease structure deformation prediction model; and weighting and summing the first prediction result and the second prediction result according to the fusion weights to obtain the fusion model.

[0018] Furthermore, in the fusion model, the predicted values ​​are weighted and summed based on the errors of each disease structure deformation prediction model to obtain the prediction result. The weighted summation of the prediction results to obtain the fusion prediction result includes:

[0019] The first weight of each disease structure deformation prediction model is calculated based on the mean square error between the predicted value and the measured value. The predicted values ​​of each disease structure deformation prediction model are then weighted and summed based on the first weight to obtain the first prediction result.

[0020] The second weight of each disease structure deformation prediction model is calculated based on the average absolute error between the predicted value and the measured value. The predicted values ​​of each disease structure deformation prediction model are then weighted and summed according to the second weight to obtain the second prediction result.

[0021] Furthermore, in the fusion model, the predicted values ​​are weighted and summed based on the errors of each disease structure deformation prediction model to obtain the prediction result. The weighted summation of the prediction results to obtain the fusion prediction result includes:

[0022] The fusion prediction result Y = βY1 + (1-β)Y2, where β is the fusion weight, with a value of [0,1]; Y1 is the first prediction result, and Y2 is the second prediction result;

[0023]

[0024]

[0025]

[0026]

[0027] Among them, P m As the first weight, Q m As the second weight; This represents the average absolute error of the prediction model for the i-th structural deformation of the disease on the prediction set. Let represent the MAE of the i-th disease structure deformation prediction model on the training set. The mean square error of the i-th disease structural deformation prediction model on the prediction set is represented by ; i,m=1,2,...,N represent N disease structural deformation prediction models.

[0028] The tunnel structural deformation prediction system based on a fusion model described in this invention includes:

[0029] A single prediction model building unit is used to build a prediction model for the structural deformation of each shield tunnel defect using multiple different neural network models.

[0030] The fusion prediction model establishment unit is used to establish a fusion model of multiple disease structure deformation prediction models based on the weighted analysis method. In the fusion model, the predicted values ​​are weighted and summed according to the error of each disease structure deformation prediction model to obtain the prediction result, and the prediction results are weighted and summed to obtain the fusion prediction result.

[0031] The prediction unit is used to predict the structural deformation of all shield tunnel defects using the fusion model.

[0032] The electronic device of the present invention includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor. When the computer program is loaded onto the processor, it implements the tunnel defect structural deformation prediction method based on the fusion model.

[0033] The computer-readable storage medium of the present invention stores a computer program, which, when executed by a processor, implements the tunnel defect structural deformation prediction method based on a fusion model.

[0034] Beneficial effects: Compared with the prior art, the advantages of the present invention are: (1) The present invention establishes a multi-model fusion prediction model based on the weight analysis method, which is applicable to the prediction of structural deformation of various tunnel diseases. Compared with establishing a prediction model for each disease separately, it is more adaptable and has higher prediction accuracy, which has guiding significance for the maintenance and management of operating tunnels; (2) The present invention uses tunnel disease information and geological information as input data to predict the structural deformation of tunnels, which can improve the accuracy of prediction; (3) When establishing the fusion model, the present invention fully considers the prediction deviation of each single model, and combines the mean square error to obtain the prediction value obtained by weighting according to the superiority or inferiority of different models, and combines the mean absolute error to obtain the prediction value obtained by weighting according to the prediction error of different models; (4) The present invention compares the prediction value of the fusion model with different fusion weights with the single optimal model, thereby determining the optimal fusion weight value. Attached Figure Description

[0035] Figure 1 This is a flowchart of the tunnel defect structural deformation prediction method of the present invention;

[0036] Figure 2 This is a schematic diagram of the ER model of the tunnel defect database according to an embodiment of the present invention;

[0037] Figure 3 This is a diagram showing the prediction results of the fusion prediction model for misalignment defects in an embodiment of the present invention.

[0038] Figure 4 The image shows the prediction results of the fusion prediction model for crack defects according to an embodiment of the present invention.

[0039] Figure 5 This is a graph showing the prediction results of the fusion prediction model for water leakage problems in an embodiment of the present invention. Detailed Implementation

[0040] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0041] Regardless of the type of defect, their occurrence typically leads to significant deformation of the tunnel structure, and the magnitude of this deformation is often closely related to the severity of the defect. To enable the use of the same model for predicting different defects, this invention proposes a method for establishing a predictive model for shield tunnel structural defects using artificial neural network technology, addressing the processing of on-site information on structural deformation and influencing factors at different defect locations. For example... Figure 1 As shown, this embodiment takes three defects—misalignment, cracks, and water leakage—as examples, and establishes three models: BP (Back Propagation), DBN (Deep Belief Network), and GRNN (Generalized Regression Neural Network). Then, the optimal fusion weights are obtained through weight analysis. The fusion model can provide better prediction results for each defect, including the following steps:

[0042] (1) Based on tunnel inspection information and basic engineering information, a tunnel defect database was constructed using MySQL. The database contains geological information on the location of defects, defect characteristic information, and information related to tunnel structural deformation. It includes three main types of defects: misalignment, cracks, and water leakage.

[0043] Table 1. Original data related to misalignment defects after linear processing.

[0044]

[0045]

[0046] Table 2. Original data related to crack defects after linear processing.

[0047]

[0048] Table 3. Raw data related to leakage problems after linear processing.

[0049]

[0050] As shown in Tables 1-3, the disease database contains geological information, disease characteristic information, and tunnel structural deformation-related data for the diseased locations in the tunnel. The difference between different disease data lies in the information describing the disease characteristics. For example, information describing misalignment includes misalignment length and maximum misalignment; crack length, width, and depth are selected as information describing crack disease characteristics; and leakage area, seepage level, and relative pH value are selected as information describing seepage disease characteristics.

[0051] Import the above information into the local MySQL database. This database uses the ER (Entity-Relationship) data model to design its table structure. Specifically, the ER model for segment cracks, misalignments, water leakage, cross-sectional deformation, and tunnel information in the shield tunnel defect database is shown below. Figure 2 As shown.

[0052] (2) BP, DBN and GRNN network models were established using MATLAB. The disease database data in step (1) were trained respectively, and the single optimal prediction model for different diseases was obtained by K-fold cross-validation.

[0053] (2.1) Establish prediction models for different diseases. The parameter values ​​of each prediction model are shown in Tables 4 to 6.

[0054] Table 4. Parameters of the Tunnel Misalignment Disease Prediction Model

[0055]

[0056] Table 5 Parameters of the Tunnel Crack Prediction Model

[0057]

[0058] Table 6 Parameters of the Tunnel Leakage Prediction Model

[0059]

[0060]

[0061] (2.2) The above prediction model is used to predict each disease. The following example is the prediction of misalignment disease.

[0062] First, the data in Table 1 was randomly shuffled using a random function, and 16 sets were selected as the training set and 4 sets as the prediction set. Ensuring that the training and prediction sets for the three models were identical, the training set data was divided into four equal parts (K=4). One part was used as the validation set for K-fold cross-validation, and the remaining three parts were used as training data. The training set was then input into the BP, DBN, and GRNN models for training, and the optimal model for each model was selected using K-fold cross-validation. The prediction set was then input into the optimal model to obtain the prediction results. The mean squared error (MSE) and mean absolute error (MAE) of the three prediction models are summarized in Table 7.

[0063] Table 7. Prediction and calculation errors of misalignment defects.

[0064]

[0065] Similarly, predictive studies on tunnel cracking and water leakage can be conducted, and the calculation results are summarized in Tables 8-9.

[0066] Table 8. Prediction and calculation errors for crack defects.

[0067]

[0068] Table 9. Prediction and Calculation Errors for Water Leakage Problems

[0069]

[0070]

[0071] (2.3) Analyzing the prediction errors in Tables 7 to 9, we can find that the optimal model for predicting structural misalignment is GRNN, while the optimal model for predicting cracks and water leakage is DBN.

[0072] (3) Establish a multi-model fusion prediction model based on weight analysis. Different weight values ​​β are selected to predict the structural deformation caused by tunnel defects, and the optimal β value is determined as the fusion weight of the fusion model.

[0073] Y = βY1 + (1-β)Y2

[0074]

[0075]

[0076]

[0077]

[0078] Where i,m=1,2,3 represent the BP model, DBN model and GRNN model respectively.

[0079] Based on the calculation formula of the fusion model, it is not difficult to find that the final result of the fusion model is related to the predicted value of each model, the calculation error MSE and MAE of each model, and the model importance factor β. Among them, when the dataset and model parameters are the same, the prediction errors (MSE and MAE) of each model will not change, and the final prediction result of the fusion model is only related to β.

[0080] Select β values ​​of 0, 0.5, 0.6, 0.7 and 1 respectively. Under the condition that other conditions are the same as in step (2), input the data in (1) into the fusion model for training to obtain prediction results under different β values ​​and calculate the prediction error. The calculation results are shown in Tables 10 to 12. For comparison, the calculation error of the single optimal model in step (2) is also included in the table.

[0081] Table 10. Prediction results of misaligned terrace defects based on fusion model and single model

[0082]

[0083] Table 11. Crack disease prediction results based on fusion model and single model

[0084]

[0085] Table 12 Prediction results of leakage problems based on fusion model and single model

[0086]

[0087] Analysis of the prediction results reveals that the fusion model performs better than the single optimal model in predicting tunnel misalignment and cracking defects, while its prediction effect on water leakage is somewhat reduced, but the overall prediction error remains within an acceptable range. A comprehensive comparison of the prediction results of the fusion model and the single optimal model shows that when β is set to 0.7, the fusion model performs well in predicting various tunnel defects.

[0088] (4) Use the determined fusion model to predict the structural deformation at different disease locations, and compare the prediction results with the measured values.

[0089] Step (3) established a fusion model with a weight value β of 0.7. This model was used to predict different tunnel defects. To more intuitively verify the prediction effect of the fusion model, the prediction results were plotted as shown in the attached figure. Figures 3-5 As shown, Figure 3 It is the prediction result of the fusion model for misalignment defects. Figure 3 (a) and (b) in the figure represent the prediction results for lateral deformation and ellipticity, respectively. Figure 4 It is the prediction result of the fusion model for crack defects. Figure 4 In the middle (a) and (b), the prediction results for lateral deformation and ellipticity are respectively; Figure 5 It is the prediction result of the fusion prediction model for water leakage problems. Figure 5 Figures (a) and (b) show the predicted results for lateral deformation and ellipticity, respectively. As can be seen from the figures, the predicted results of the fusion model are in excellent agreement with the measured values, the prediction error can be controlled within a very small range, and the model's computational performance is good.

[0090] The method for predicting structural deformation of shield tunnel defects based on a fusion neural network model has good prediction results for different defects in tunnels and can provide a reference for defect management and maintenance of shield tunnels during operation.

[0091] The tunnel structural deformation prediction system based on a fusion model described in this invention includes:

[0092] A single prediction model building unit is used to build a prediction model for the structural deformation of each shield tunnel defect using multiple different neural network models.

[0093] The fusion prediction model establishment unit is used to establish a fusion model of multiple disease structure deformation prediction models based on the weighted analysis method. In the fusion model, the predicted values ​​are weighted and summed according to the error of each disease structure deformation prediction model to obtain the prediction result, and the prediction results are weighted and summed to obtain the fusion prediction result.

[0094] The prediction unit is used to predict the structural deformation of all shield tunnel defects using the fusion model.

[0095] The electronic device of the present invention includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor. When the computer program is loaded onto the processor, it implements the tunnel defect structural deformation prediction method based on the fusion model.

[0096] The computer-readable storage medium of the present invention stores a computer program, which, when executed by a processor, implements the tunnel defect structural deformation prediction method based on a fusion model.

[0097] The computer-readable storage medium may include RAM, ROM, EEPROM, CD-ROM or other optical disc storage devices, magnetic disk storage devices or other magnetic storage devices, flash memory, or any other media that can be used to store desired program code in the form of instructions or data structures and is accessible by a computer.

[0098] The processor is used to execute computer programs stored in memory to implement the various steps in the methods described in the above embodiments.

Claims

1. A method for predicting structural deformation of tunnel defects based on a fusion model, characterized in that, Includes the following steps: For each type of shield tunnel defect, multiple different neural network models are used to establish a defect structural deformation prediction model; A fusion model based on weighted analysis is established to predict the structural deformation of multiple defects in shield tunnels. The fusion model is then used to predict the structural deformation of all defects in shield tunnels. In the fusion model, the predicted values ​​are weighted and summed according to the error of each disease structure deformation prediction model to obtain the prediction result, and the prediction results are weighted and summed to obtain the fusion prediction result. In the fusion model, the predicted values ​​are weighted and summed based on the errors of each disease structure deformation prediction model to obtain the prediction result. The weighted summation of the prediction results to obtain the fusion prediction result includes: The first weight of each disease structure deformation prediction model is calculated based on the mean square error between the predicted value and the measured value. The predicted values ​​of each disease structure deformation prediction model are then weighted and summed based on the first weight to obtain the first prediction result. The second weight of each disease structure deformation prediction model is calculated based on the average absolute error between the predicted value and the measured value. The predicted values ​​of each disease structure deformation prediction model are then weighted and summed according to the second weight to obtain the second prediction result.

2. The method for predicting tunnel structural deformation based on a fusion model according to claim 1, characterized in that, The fusion model based on weighted analysis to establish multiple disease structure deformation prediction models includes: From multiple disease structure deformation prediction models, a single optimal prediction model is obtained using the K-fold cross-validation method. The prediction results of the fusion model are compared with the prediction results of the single optimal prediction model for each shield tunnel defect to determine the value of the fusion weight in the fusion model.

3. The method for predicting tunnel structural deformation based on a fusion model according to claim 1, characterized in that, Among the multiple disease structure deformation prediction models, the single optimal prediction model obtained by K-fold cross-validation includes: For each of the diseased structural deformation prediction models, a single optimal prediction model is determined by calculating the mean square error and mean absolute error between the predicted and measured values ​​on the dataset.

4. The method for predicting tunnel structural deformation based on a fusion model according to claim 3, characterized in that, The formulas for calculating the mean square error (MSE) and the mean absolute error (MAE) are as follows: ； ； in These are the predicted values ​​from the disease structure deformation prediction model. These are measured values. t This represents the number of data points in the dataset.

5. The method for predicting tunnel structural deformation based on a fusion model according to claim 1, characterized in that, In the fusion model, the predicted values ​​are weighted and summed based on the errors of each disease structure deformation prediction model to obtain the prediction result. The weighted summation of the prediction results to obtain the fusion prediction result includes: Fusion prediction results ,in The fusion weights have values ​​of [0,1]. This is the first prediction result. This is the second prediction result; ； ； ； ； in, As the first weight, As the second weight; Indicates the first The mean absolute error of the disease structure deformation prediction model on the prediction set. Indicates the first MAE of a disease structural deformation prediction model on the training set Indicates the first The mean square error of a disease structural deformation prediction model on the prediction set; This represents N disease structure deformation prediction models.

6. A tunnel structural deformation prediction system based on a fusion model, characterized in that, include: A single prediction model building unit is used to build a prediction model for the structural deformation of each shield tunnel defect using multiple different neural network models. The fusion prediction model establishment unit is used to establish a fusion model of multiple disease structure deformation prediction models based on the weight analysis method. In the fusion model, the predicted values ​​are weighted and summed according to the error of each disease structure deformation prediction model to obtain the prediction result, and the prediction results are weighted and summed to obtain the fusion prediction result. The prediction unit is used to predict the structural deformation of all shield tunnel defects using the fusion model; In the fusion model, the predicted values ​​are weighted and summed based on the errors of each disease structure deformation prediction model to obtain the prediction result. The weighted summation of the prediction results to obtain the fusion prediction result includes: The first weight of each disease structure deformation prediction model is calculated based on the mean square error between the predicted value and the measured value. The predicted values ​​of each disease structure deformation prediction model are then weighted and summed based on the first weight to obtain the first prediction result. The second weight of each disease structure deformation prediction model is calculated based on the average absolute error between the predicted value and the measured value. The predicted values ​​of each disease structure deformation prediction model are then weighted and summed according to the second weight to obtain the second prediction result.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is loaded into the processor, it implements the tunnel defect structural deformation prediction method based on the fusion model according to any one of claims 1-6.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the tunnel defect structural deformation prediction method based on the fusion model according to any one of claims 1-6.

Images