A tunnel health assessment method and system based on machine learning

Abstract

The present invention proposes a tunnel health assessment method and system based on machine learning. The method includes: constructing a tunnel health assessment hierarchical structure model; obtaining criterion layer data based on the current tunnel index layer data, where the criterion layer data includes lining crack, leakage, lining material deterioration, cavity behind the lining, lining delamination and spalling, and lining deformation data; calculating the superposition amount of lining deformation according to the lining crack data and the leakage data, and correcting the lining deformation data based on the superposition amount to obtain the true lining deformation data; inputting the true lining deformation data and other current tunnel criterion layer data into a machine learning model to obtain the current tunnel health assessment result; wherein, the machine learning model is constructed based on the criterion layer data and the target layer grade, and the machine learning model is trained by using the historical tunnel criterion layer data and the corresponding health grades. The present invention effectively improves the accuracy of tunnel health assessment by considering the superposition effect between leakage and lining cracks.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel health assessment, and particularly to a tunnel health assessment method and system based on machine learning. Background Art

[0002] With the rapid development of transportation infrastructure, the number of tunnels and their operation time have increased. At the same time, many operating tunnels have different types of diseases such as cracks, water leakage, voids behind linings, and lining deformation. The development of these diseases will not only damage the integrity of the tunnel structure but also may threaten the safety of vehicles and personnel. Conducting a health assessment of tunnels can ensure traffic safety and extend the service life of tunnels. Traditional tunnel health assessment methods rely on regular manual inspections, which are not only time-consuming and laborious but also difficult to predict the development of tunnel diseases.

[0003] Machine learning models can comprehensively consider various factors affecting tunnel health, capture the complex non-linear relationship between disease indicators and health levels through training the models, thereby predicting the health status of tunnels and improving the accuracy and efficiency of tunnel health assessment.

[0004] The inventor found that existing technologies mostly consider the relationship between each disease indicator and the tunnel health condition independently. However, lining cracks and water leakage often do not exist in isolation but are interrelated and interact with each other: Lining cracks and water leakage are two common diseases in tunnels. Lining cracks provide a channel for water, making water leakage more likely to occur; while water leakage will further exacerbate the corrosion and deterioration of the lining, promoting the expansion and increase of cracks. This superposition situation will significantly exacerbate the damage to the tunnel structure and reduce the service life and safety of the tunnel. Therefore, how to determine whether lining cracks and water leakage can produce a superposition effect and consider the superposition effect when evaluating tunnel health to improve the evaluation accuracy is an urgent problem to be solved. Summary of the Invention

[0005] To solve the above problems, the present invention proposes a tunnel health assessment method and system based on machine learning, which can more accurately reflect the actual health condition of the tunnel and improve the accuracy of tunnel health assessment by preferentially judging whether a superposition effect occurs and considering the superposition effect between water leakage and lining cracks.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] In the first aspect, the present invention provides a tunnel health assessment method based on machine learning, including:

[0008] Constructing a tunnel health assessment hierarchical structure model, and defining an index layer, a criterion layer, and a target layer;

[0009] Collect the current tunnel index layer data, and obtain the criterion layer data according to the index layer data; the criterion layer data includes lining crack, leakage, lining material deterioration, cavity behind the lining, lining delamination and lining deformation data;

[0010] Calculate the lining deformation superposition amount according to the lining crack data and the leakage data, and correct the lining deformation data based on the lining deformation superposition amount to obtain the true lining deformation data;

[0011] Input the true lining deformation data and other current tunnel criterion layer data into the trained machine learning model to obtain the current tunnel health evaluation result;

[0012] Among them, the machine learning model is constructed based on the criterion layer data and the target layer grade, and the machine learning model is trained by using the historical tunnel criterion layer data and the corresponding health grades; the target layer grades include four levels of health grades.

[0013] Preferably, the index layer data corresponding to the criterion layer includes:

[0014] Lining cracks include crack length, crack width and crack depth data;

[0015] Leakage includes leakage state, pH value and frost damage state data;

[0016] Lining material deterioration includes lining strength, lining thickness and steel bar corrosion data;

[0017] The cavity behind the lining includes cavity depth data;

[0018] Lining delamination includes the possibility of falling, delamination depth and delamination diameter data;

[0019] Lining deformation includes deformation amount and deformation speed data.

[0020] Preferably, the collecting the current tunnel index layer data and obtaining the criterion layer data according to the index layer data specifically includes:

[0021] According to the important influence relationship between the characteristic parameters in each layer, select the corresponding exponential scale, establish the judgment matrix between the index layer and the criterion layer, and if the consistency of the judgment matrix is less than the set value, calculate the weight value between the index layer and the criterion layer in the judgment matrix;

[0022] Calculate the criterion layer data according to the index layer data and the weight value.

[0023] Preferably, before calculating the lining deformation superposition amount according to the lining crack data and the leakage data, it also includes judging whether there is a superposition effect of lining cracks and leakage:

[0024] Obtain images of lining cracks and leakage water, identify the contours of lining cracks and leakage water, and establish detection frames based on the contours;

[0025] Obtain the midpoints of the detection frames. If the distance between the midpoints of the detection frames of adjacent lining cracks and leakage water is less than a preset length, it is determined that the lining cracks and leakage water within the detection frames have a superposition effect.

[0026] Preferably, calculating the superposition amount of lining deformation according to the lining crack data and leakage water data specifically includes:

[0027] Input the current lining crack data and leakage water data into a linear regression model to obtain the superposition amount of lining deformation;

[0028] Among them, the training process of the linear regression model is to use the lining crack data and leakage water data at the first moment in the historical period, and the lining crack data and leakage water data at the second moment as input values, and use the lining deformation amount within the corresponding period as the output value to train the linear regression model until the model loss function is minimized.

[0029] Preferably, correcting the lining deformation data based on the superposition amount of lining deformation to obtain the true lining deformation data specifically includes:

[0030] Set weights for the lining deformation data and the superposition amount of lining deformation respectively;

[0031] Multiply the lining deformation data and the superposition amount of lining deformation by their respective weights and then add them to obtain the true lining deformation data.

[0032] Preferably, the machine learning model includes: random forest, XGBOOST, decision tree, support vector machine and neural network.

[0033] In a second aspect, the present invention provides a tunnel health evaluation system based on machine learning, including:

[0034] A hierarchical model construction module for constructing a tunnel health evaluation hierarchical structure model and defining an index layer, a criterion layer and a target layer;

[0035] A criterion layer data acquisition module for collecting current tunnel index layer data and obtaining criterion layer data according to the index layer data; the criterion layer data includes lining crack, leakage water, lining material deterioration, cavity behind the lining, lining delamination and lining deformation data;

[0036] A deformation data correction module for calculating the superposition amount of lining deformation according to the lining crack data and leakage water data, and correcting the lining deformation data based on the superposition amount of lining deformation to obtain the true lining deformation data;

[0037] A health assessment module is configured to input the actual lining deformation data and other current tunnel criterion layer data into a trained machine learning model to obtain the current tunnel health assessment result. Among them, the machine learning model is constructed based on the criterion layer data and the target layer grades, and the machine learning model is trained using the historical tunnel criterion layer data and the corresponding health grades. The target layer grades include four levels of health grades.

[0038] In a third aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the steps in a tunnel health assessment method based on machine learning described in the first aspect are implemented.

[0039] In a fourth aspect, the present invention provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps in a tunnel health assessment method based on machine learning described in the first aspect are implemented.

[0040] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0041] This application takes into account that lining cracks and water seepage in tunnels are often not isolated diseases, and there may be interactions between them. Therefore, when assessing the tunnel health based on the collected disease feature data, considering this superposition effect and determining whether a superposition effect has occurred before considering the superposition effect can more accurately reflect the actual health status of the tunnel and avoid the one-sidedness that may be brought by single-factor evaluation.

[0042] The advantages of the additional aspects of the present invention will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The specification drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute a limitation to the present invention.

[0044] Figure 1 It is the main flowchart of a tunnel health assessment method based on machine learning provided by an embodiment of the present invention;

[0045] Figure 2 It is the health assessment index system provided by an embodiment of the present invention;

[0046] Figure 3 It is the iteration graph of 10-fold cross-validation of the machine learning model provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] The present invention will be further described below in conjunction with the drawings and embodiments.

[0048] Example 1

[0049] As Figure 1 shown, this embodiment discloses a tunnel health evaluation method based on machine learning, including the following steps:

[0050] S1: Construct a tunnel health evaluation hierarchical model, and define the index layer, criterion layer, and target layer;

[0051] S2: Collect the current tunnel index layer data, and obtain the criterion layer data according to the index layer data; the criterion layer data includes lining crack, leakage, lining material deterioration, cavity behind the lining, lining delamination and spalling, and lining deformation data;

[0052] S3: Calculate the superposition amount of lining deformation according to the lining crack data and leakage data, and correct the lining deformation data based on the superposition amount of lining deformation to obtain the true lining deformation data;

[0053] S4: Input the true lining deformation data and other current tunnel criterion layer data into the trained machine learning model to obtain the current tunnel health evaluation result; wherein, the machine learning model is constructed based on the criterion layer data and the target layer grade, and the machine learning model is trained by using the historical tunnel criterion layer data and the corresponding health grades; the target layer grade includes four levels of health grades.

[0054] Specifically, first construct a tunnel health evaluation hierarchical model, and define the index layer, criterion layer, and target layer.

[0055] When establishing the highway tunnel health status index system, follow the principles of scientificity, completeness, simplicity, independence, hierarchy, and operability to determine the key indicators affecting the tunnel health status.

[0056] Establish a health evaluation index system for highway tunnels during the operation period as Figure 2 shown, and use the analytic hierarchy process to divide the highway tunnel health evaluation index system into three layers.

[0057] The first layer is the target layer, which contains one target object, that is, the health status of the highway tunnel, and is characterized by the health grade. When establishing the highway tunnel health grade, according to relevant domestic and foreign specifications, a four-level classification method is adopted, and the tunnel structure health status is divided into four grades: grade 1 (no disease or minor disease), grade 2 (general disease), grade 3 (relatively serious disease), and grade 4 (serious disease). The boundaries between each grade are obvious.

[0058] The second layer is the criterion layer, which consists of factors affecting the health status of the highway tunnel, including six aspects: lining crack, leakage, lining material deterioration, cavity behind the lining, lining delamination and spalling, and lining deformation, movement, and settlement.

[0059] The third layer is the index layer, which consists of various indexes affecting the factors in the criterion layer. The lining cracks include three indexes: crack length, width, and depth; the leakage includes three indexes: leakage state, pH value, and frost damage state; the deterioration of the lining material includes three indexes: lining strength ratio, lining thickness ratio, and steel bar corrosion; the cavity behind the lining includes one index: cavity depth; the delamination and spalling of the lining include three indexes: possibility of falling, depth, and diameter; the deformation, movement, and settlement of the lining include two indexes: internal limit ratio and deformation speed.

[0060] For the original dataset of tunnel diseases obtained from on-site detection, the box plot method is used to detect and remove data outliers, the interpolation method is used to fill in missing values, and the Z-score method is used to standardize the data to form a standardized evaluation dataset, providing a data basis for model training. At the same time, based on the standardized dataset, the Karl Pearson correlation coefficient calculation formula is used to conduct a correlation analysis between different characteristic indexes. The calculation formula of the Pearson correlation coefficient is:

[0061]

[0062] In the formula: x and y are the observed values of two variables, is the mean value of variables x and y.

[0063] Generally speaking, when the absolute value of the correlation coefficient is in the intervals of 0.0 - 0.2, 0.2 - 0.4, 0.4 - 0.6, 0.6 - 0.8, and 0.8 - 1.0, the correlations between different characteristic indexes are "extremely weak correlation or no correlation", "weak correlation", "moderate correlation", "strong correlation", and "extremely strong correlation" respectively. The closer the absolute value of the correlation coefficient is to 1, the stronger the correlation. By calculating the correlation coefficients between different indexes, a correlation coefficient heat matrix is constructed to verify the rationality of the selection of highway tunnel health evaluation indexes.

[0064] The processed data is stored in the database, and Python is used for data calling and further processing to ensure the integrity and availability of the dataset.

[0065] As a further implementation method, considering that lining cracks and leakage are two main diseases affecting tunnel health, their superposition will significantly exacerbate the corrosion of the lining, and further cause a more serious impact on the overall health of the tunnel. Therefore, this embodiment considers the superposition effect between lining cracks and leakage.

[0066] If the location of the lining crack is far from the area where leakage occurs, there is no possibility of interaction. Therefore, before calculating the superposition effect, it is necessary to confirm whether the superposition effect has occurred. The specific steps are as follows:

[0067] S301: Obtain the images of lining cracks and seepage water, identify the contours of lining cracks and seepage water, and establish detection frames based on the contours.

[0068] S302: Obtain the midpoints of the detection frames. If the distance between the midpoints of the detection frames of adjacent lining cracks and seepage water is less than the preset length, it is determined that superposition effects occur within the detection frames for the lining cracks and seepage water.

[0069] After determining that superposition effects occur, calculate the lining deformation amount generated by the two due to the superposition effects. The specific steps are as follows:

[0070] S311: Input the current lining crack data and seepage water data into the linear regression model to obtain the lining deformation superposition amount.

[0071] Among them, the training process of the linear regression model is to use the lining crack data and seepage water data at the first moment in the historical period, as well as the lining crack data and seepage water data at the second moment, as input values, and use the lining deformation amount within the corresponding period as the output value to train the linear regression model until the model loss function is minimized.

[0072] After obtaining the lining deformation superposition amount, correct the initial lining deformation data. The specific steps are as follows:

[0073] S321: Set weights for the lining deformation data and the lining deformation superposition amount respectively.

[0074] S322: Multiply the lining deformation data and the lining deformation superposition amount by their respective weights and then add them together to obtain the true lining deformation data.

[0075] In S4, obtain the current tunnel health evaluation result based on machine learning.

[0076] In the Python environment, use the scikit-learn library to implement the random forest and XGBOOST algorithms.

[0077] First, perform stratified sampling according to a ratio of 7:3 to divide the dataset into a training set and a test set. The training set is used to fit the model and adjust the model parameters, and the test set is used to evaluate the final model.

[0078] The selection of hyperparameters is carried out by combining manual tuning and the grid search method. In manual tuning, first determine the parameters that have a greater impact on the model performance, and then refine them based on these parameters through grid search. Use 10-fold cross-validation to select the optimal hyperparameter combination, as Figure 3 shown. Through the traversal of grid search and the evaluation of cross-validation, it helps to find the model configuration that best represents the data characteristics, thereby improving the prediction accuracy and generalization ability of the model.

[0079] The model is evaluated using 10-fold cross-validation to accurately estimate the model's performance in actual applications. First, the training set is divided into 10 mutually exclusive subsets of similar size and distribution. Then, it loops 10 times. Each time, 9 of these subsets are combined into a new training set, and the remaining 1 subset is used as the validation set. In each loop, the model is trained using the training set, and the performance of the model is evaluated using the validation set. Finally, the mean of the 10 training and validation results is taken as the actual performance of the model.

[0080] Hyperparameter optimization of the machine learning model is performed based on the combination of manual tuning and grid search. Before performing grid search tuning, it is necessary to determine, as much as possible through manual tuning, the hyperparameters that have a greater impact on the model performance, thereby reducing the computational amount in the grid search process. In grid search, evaluating the performance of each hyperparameter combination through 10-fold cross-validation helps to mitigate the risk of overfitting. Through the traversal of grid search and the evaluation of cross-validation, the optimal hyperparameter combination is finally selected to ensure the maximization of the model performance.

[0081] After determining the optimal hyperparameter combination, the performance of the machine learning model is verified on the test set, and the machine learning model is finally evaluated using 4 evaluation metrics: accuracy, precision, recall, and F1-score. The calculation methods of the evaluation metrics are as follows:

[0082] 1. Accuracy

[0083] Accuracy refers to the proportion of the number of correctly predicted samples to the total number of samples.

[0084]

[0085] 2. Precision

[0086] Precision refers to the proportion of the samples actually being positive samples among the samples predicted as positive samples.

[0087]

[0088] 3. Recall

[0089] Recall refers to the proportion of the samples correctly predicted as positive samples among the samples actually being positive samples.

[0090]

[0091] 4. F1-Score

[0092] F1-Score is the harmonic mean of precision and recall.

[0093]

[0094] In the formula, TP represents the positive samples predicted as positive by the model; FP represents the negative samples predicted as positive by the model; TN represents the negative samples predicted as negative by the model; FN represents the positive samples predicted as negative by the model.

[0095] Based on the above model evaluation results, the machine learning model with the best performance in terms of generalization ability, prediction accuracy, etc. is selected. When new collected tunnel data is input into the model, the model will use the learned rules and patterns according to the input feature data to predict the health level of each tunnel section. Compared with the traditional tunnel health assessment method, the present invention can achieve higher assessment accuracy and efficiency, timely capture the subtle changes in the tunnel health condition, effectively predict and manage the development of tunnel diseases, thereby greatly improving the operation safety and maintenance efficiency of the tunnel.

[0096] The purpose of this specific embodiment is to improve the accuracy and efficiency of tunnel health status assessment and at the same time achieve early prediction of the development of tunnel diseases. First, a highway tunnel health evaluation index system covering the target layer, criterion layer and index layer is established, and multiple aspects such as lining cracks, water leakage, lining material deterioration, voids behind the lining, lining delamination and spalling, and lining deformation and settlement are considered in detail. The health condition of the tunnel structure is classified according to the four-level classification method, and the boundaries between each level are clarified. On this basis, the tunnel disease data obtained through on-site detection is preprocessed to form a standardized evaluation data set. Machine learning algorithms such as random forest and XGBOOST are used to train this data set, the optimal model parameters are selected, and continuous iteration and optimization are carried out to construct a machine learning model for predicting the tunnel health level. Finally, the model performance is verified on the test set, and the model is comprehensively evaluated to ensure high prediction accuracy and reliability. The present invention not only improves the scientificity and accuracy of highway tunnel health assessment and disease prediction, but also helps to optimize the allocation of tunnel maintenance resources and improve the tunnel operation management level.

[0097] Embodiment Two

[0098] This embodiment provides a tunnel health evaluation system based on machine learning, including:

[0099] A hierarchical model construction module, used to construct a tunnel health evaluation hierarchical structure model, and define the index layer, criterion layer and target layer;

[0100] A criterion layer data acquisition module, used to collect the current tunnel index layer data and obtain the criterion layer data according to the index layer data; the criterion layer data includes lining crack, water leakage, lining material deterioration, voids behind the lining, lining delamination and spalling, and lining deformation data;

[0101] The deformation data correction module is used to calculate the lining deformation superposition amount according to the lining crack data and the leakage data, and correct the lining deformation data based on the lining deformation superposition amount to obtain the true lining deformation data;

[0102] The health assessment module is used to input the true lining deformation data and other current tunnel criterion layer data into the trained machine learning model to obtain the current tunnel health assessment result; wherein, the machine learning model is constructed based on the criterion layer data and the target layer grades, and the machine learning model is trained by using the historical tunnel criterion layer data and the corresponding health grades; the target layer grades include four levels of health grades.

[0103] Embodiment III

[0104] This embodiment provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the steps in a tunnel health assessment method based on machine learning as described in Embodiment I above.

[0105] Embodiment IV

[0106] This embodiment provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the steps in a tunnel health assessment method based on machine learning as described in Embodiment I above.

[0107] The steps or modules involved in Embodiments II to IV above correspond to those in Embodiment I, and the specific implementation manners can be referred to the relevant description part of Embodiment I. The term "computer-readable storage medium" should be understood to include a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode, or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0108] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A tunnel health assessment method based on machine learning, characterized in that, Including: Construct a hierarchical model for tunnel health assessment, and define the index layer, criterion layer, and target layer; Collect the current tunnel index layer data, and obtain the criterion layer data according to the index layer data; The criterion layer data includes lining crack, leakage, lining material deterioration, cavity behind the lining, lining delamination and spalling, and lining deformation data; Calculate the superposition amount of lining deformation based on the lining crack data and leakage data, and correct the lining deformation data based on the superposition amount of lining deformation to obtain the true lining deformation data; Before calculating the superposition amount of lining deformation based on the lining crack data and leakage data, it also includes judging whether there is a superposition effect of lining crack and leakage: obtain the lining crack and leakage images, identify the contours of the lining crack and leakage, and establish a detection frame based on the contours; obtain the midpoint of the detection frame. If the distance between the midpoints of the adjacent lining crack and leakage detection frames is less than the preset length, it is judged that the lining crack and leakage within the detection frame have a superposition effect; The calculation of the superposition amount of lining deformation based on the lining crack data and leakage data specifically includes: input the current lining crack data and leakage data into a linear regression model to obtain the superposition amount of lining deformation; among them, the training process of the linear regression model is to use the lining crack data and leakage data at the first moment in the historical period, and the lining crack data and leakage data at the second moment as input values, and use the lining deformation amount within the corresponding period as the output value to train the linear regression model until the model loss function is minimized; Input the true lining deformation data and other current tunnel criterion layer data into the trained machine learning model to obtain the current tunnel health assessment result; Among them, the machine learning model is constructed based on the criterion layer data and the target layer grades, and the machine learning model is trained using the historical tunnel criterion layer data and the corresponding health grades; the target layer grades include four levels of health grades.

2. The tunnel health assessment method based on machine learning according to claim 1, characterized in that, The index layer data corresponding to the criterion layer includes: Lining cracks include crack length, crack width, and crack depth data; Leakage includes leakage status, pH value, and frost damage status data; Lining material deterioration includes lining strength, lining thickness, and steel bar corrosion data; Cavity behind the lining includes cavity depth data; Lining delamination and spalling include the possibility of falling, spalling depth, and spalling diameter data; Lining deformation includes deformation amount and deformation speed data.

3. A tunnel health assessment method based on machine learning according to claim 1, characterized in that The collection of the current tunnel index layer data and obtaining the criterion layer data according to the index layer data specifically includes: According to the important influence relationship between the characteristic parameters in each level, select the corresponding exponential scale, establish a judgment matrix between the index layer and the criterion layer. If the consistency of the judgment matrix is less than the set value, calculate the weight value between the index layer and the criterion layer in the judgment matrix; Calculate the criterion layer data according to the index layer data and the weight value.

4. The tunnel health assessment method based on machine learning according to claim 1, wherein, The correction of the lining deformation data based on the superposition amount of lining deformation to obtain the true lining deformation data specifically includes: Set weights for the lining deformation data and the superposition amount of lining deformation respectively; Multiply the lining deformation data and the superposition amount of lining deformation by their respective weights and then add them to obtain the true lining deformation data.

5. The tunnel health assessment method based on machine learning according to claim 1, characterized in that The machine learning model includes: random forest, XGBOOST, decision tree, support vector machine, and neural network.

6. A tunnel health evaluation system based on machine learning, characterized in that, including: A hierarchical model construction module for constructing a tunnel health evaluation hierarchical structure model and defining an index layer, a criterion layer, and a target layer; A criterion layer data acquisition module for collecting current tunnel index layer data and obtaining criterion layer data based on the index layer data; the criterion layer data includes lining crack, leakage, lining material deterioration, cavity behind the lining, lining delamination and spalling, and lining deformation data; A deformation data correction module for calculating the lining deformation superposition amount according to the lining crack data and the leakage data, and correcting the lining deformation data based on the lining deformation superposition amount to obtain the true lining deformation data; Before calculating the lining deformation superposition amount according to the lining crack data and the leakage data, it also includes judging whether there is a superposition effect of lining cracks and leakage: obtaining lining crack and leakage images, identifying the contours of the lining cracks and leakage, and establishing a detection frame based on the contours; obtaining the midpoint of the detection frame, if the distance between the midpoints of the detection frames of adjacent lining cracks and leakage is less than a preset length, it is determined that there is a superposition effect of the lining cracks and leakage within the detection frame; The calculation of the lining deformation superposition amount according to the lining crack data and the leakage data specifically includes: inputting the current lining crack data and the leakage data into a linear regression model to obtain the lining deformation superposition amount; wherein, the training process of the linear regression model is to use the lining crack data and the leakage data at the first moment in the historical period, and the lining crack data and the leakage data at the second moment as input values, and the lining deformation amount within the corresponding period as the output value to train the linear regression model until the model loss function is minimized; A health evaluation module for inputting the true lining deformation data and other current tunnel criterion layer data into a trained machine learning model to obtain the current tunnel health evaluation result; wherein, the machine learning model is constructed based on the criterion layer data and the target layer level, and the machine learning model is trained using the historical tunnel criterion layer data and the corresponding health levels; the target layer level includes four levels of health levels.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by a processor, it implements the steps in a machine learning-based tunnel health evaluation method as described in any one of claims 1-5.

8. A computer device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in a machine learning-based tunnel health evaluation method as described in any one of claims 1-5.

Images