Tunnel safety state evaluation and prediction method based on lining damage evolution

Abstract

The application relates to a tunnel safety state evaluation and prediction method based on lining damage evolution in the technical field of tunnel engineering. The method comprises the following steps: collecting tunnel lining crack and damage data, inputting the data into a TF-Transformer deep learning analysis model, outputting damage characteristics, summarizing the damage evolution mechanism and development law of the crack and damage tunnel, and establishing a crack and damage tunnel safety evaluation index system; constructing a PL-VIKOR comprehensive evaluation model based on the crack and damage tunnel safety evaluation index system and an improved subjective and objective combination weighting method based on the goodness of normal distribution fitting; establishing a multi-dimensional crack and damage lining monitoring system to monitor real-time tunnel section data; using an LSTM model to predict the development results of the tunnel section based on the real-time tunnel section data, and combining the PL-VIKOR comprehensive evaluation model to evaluate the future safety state of the tunnel. Real-time and objective evaluation of the safety state of the crack and damage tunnel structure is realized.

Description

Technical Field

[0001] This application relates to the field of tunnel engineering technology, and in particular to a method for predicting and assessing the safety status of tunnels based on the evolution of lining damage. Background Technology

[0002] Layered soft rock is a typical rock mass widely distributed in the rugged mountainous areas of western my country, exhibiting significant anisotropy. Under high geostress conditions, layered soft rock is highly susceptible to significant uneven deformation, which is one of the core causes of serious defects such as cracking, spalling, and even collapse in the lining structure of operating tunnels.

[0003] Currently, there is a lack of systematic and in-depth research on the damage evolution mechanism and development law of high-stress layered soft rock fracture tunnels in operation. In engineering practice, traditional treatment methods for existing lining fractures have significant limitations and cannot meet the long-term treatment needs under complex geological conditions. On the one hand, existing fracture identification and assessment rely heavily on periodic manual inspections. This method depends on human experience, is inefficient and lacks coverage, and is difficult to capture the early initiation and rapid development stages of damage, leading to delayed treatment decisions and a high risk of missing the best intervention opportunity. On the other hand, traditional treatment solutions such as grouting reinforcement, patching, and lining are often formulated based on fixed standards or empirical formulas, without fully considering the differences in key factors such as the stress state, bedding attitude, rock mechanical properties, damage evolution stage, and environmental loads (such as groundwater) of specific tunnel sections. This "one-size-fits-all" approach cannot accurately match the complex and ever-changing actual working conditions, resulting not only in unstable treatment effects and easy recurrence of defects, but also potential waste of resources.

[0004] Therefore, there is an urgent need in related technologies for a way to improve the accuracy and response efficiency of tunnel safety status assessment and prediction. Summary of the Invention

[0005] Therefore, it is necessary to provide a tunnel safety status assessment and prediction method based on lining damage evolution that can improve the accuracy and response efficiency of tunnel safety status evaluation and prediction, and address the above-mentioned technical problems.

[0006] Firstly, this application provides a method for predicting and assessing the safety status of tunnels based on the evolution of lining damage. The method includes:

[0007] Based on the simulation test system of lining structure-stratum complex, tunnel lining crack damage data is collected, and the data is input into the TF-Transformer deep learning analysis model to obtain damage characteristics. Based on the damage characteristics, the damage evolution mechanism and development law of cracked tunnels are summarized, and a safety evaluation index system for cracked tunnels is established.

[0008] A PL-VIKOR comprehensive evaluation model is constructed based on the safety evaluation index system for cracked tunnels and an improved subjective and objective combined weighting method based on the goodness of fit of normal distribution.

[0009] Establish a multi-dimensional cracked lining monitoring system to monitor real-time tunnel cross-section data;

[0010] Based on the real-time tunnel cross-section data, a pre-trained LSTM model is used to predict the tunnel cross-section development results. Based on the tunnel cross-section development results, the PL-VIKOR comprehensive evaluation model is used to assess the future safety status of the tunnel.

[0011] Optionally, in one embodiment of this application, the safety evaluation index system for cracked tunnels includes structural mechanical characteristic indexes, tunnel geological condition indexes, surrounding rock engineering characteristic indexes, and lining cracking characteristic indexes.

[0012] Optionally, in one embodiment of this application, the improved subjective-objective combination weighting method based on the goodness of fit of the normal distribution includes:

[0013] The subjective weights are calculated using an improved analytic hierarchy process (AHP), and the objective weights are calculated using an improved entropy weighting method.

[0014] A combined weighting method is used based on the goodness of fit of the normal distribution of the sample evaluation result sequence.

[0015] Optionally, in one embodiment of this application, the step of calculating subjective weights using an improved analytic hierarchy process and objective weights using an improved entropy weight method includes:

[0016] An improved analytic hierarchy process (AHP) is based on a quasi-optimal transfer matrix, and an improved entropy weighting method is based on relative weights.

[0017] Optionally, in one embodiment of this application, the method of combining weights based on the goodness of fit of the normal distribution of the sample evaluation result sequence includes:

[0018] Calculate the goodness of fit of the normal distribution under different combination coefficients of each method, and determine the final combination weighting result based on the goodness of fit of the normal distribution.

[0019] Optionally, in one embodiment of this application, the PL-VIKOR comprehensive evaluation model includes:

[0020] Calculate the group utility value and individual regret value based on safety evaluation index data, index weights, positive ideal solutions, and negative ideal solutions;

[0021] Calculate the cosine angle value based on safety evaluation index data;

[0022] The benefit ratio is calculated based on the group utility value, individual regret value, and cosine angle value.

[0023] The safety level of the monitoring section is determined based on the aforementioned benefit ratio.

[0024] Optionally, in one embodiment of this application, the method further includes:

[0025] Based on the future safety status of the tunnel, a tiered intelligent response will be implemented.

[0026] The aforementioned tunnel safety status assessment and prediction method based on lining damage evolution firstly collects tunnel lining crack data using a lining structure-soil complex simulation test system. This data is then input into a TF-Transformer deep learning analysis model to obtain damage characteristics. Based on these characteristics, the damage evolution mechanism and development law of the cracked tunnel are summarized, and a safety evaluation index system for cracked tunnels is established. Next, a PL-VIKOR comprehensive evaluation model is constructed based on the cracked tunnel safety evaluation index system and an improved subjective-objective combination weighting method based on the goodness of fit of a normal distribution. Following this, a multi-dimensional cracked lining monitoring system is established to monitor real-time tunnel cross-sectional data. Finally, a pre-trained LSTM model is used to predict the tunnel cross-sectional development results based on the real-time tunnel cross-sectional data. The future safety status of the tunnel is then assessed based on the tunnel cross-sectional development results combined with the PL-VIKOR comprehensive evaluation model. In other words, by establishing a multi-dimensional cracked lining monitoring system, a cracked tunnel safety evaluation index system, and a cracked tunnel comprehensive evaluation method, and combining it with a prediction and early warning model calculated by LSTM, the evolution trend of the safety status of lining structures with potential risks can be predicted in a forward-looking manner based on monitoring data. This allows for timely and accurate understanding of the tunnel status at the monitoring section, and enables real-time and objective assessment of the safety status of cracked tunnel structures. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating a tunnel safety status assessment and prediction method based on lining damage evolution in one embodiment.

[0028] Figure 2 This is a schematic diagram of a simulation test system for the lining structure-soil complex in one embodiment;

[0029] Figure 3 This is a schematic diagram of the structure of the TF-Transformer deep learning analysis model in one embodiment;

[0030] Figure 4 This is a schematic diagram of the structure of an LSTM tunnel monitoring data prediction model in one embodiment;

[0031] Figure 5 This is a flowchart illustrating the specific steps of a tunnel safety status assessment and prediction method based on lining damage evolution in one embodiment. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0033] In one embodiment, such as Figure 1 As shown, a tunnel safety status assessment and prediction method based on lining damage evolution is provided, including the following steps:

[0034] S101: Based on the lining structure-stratum composite simulation test system, the tunnel lining crack damage data is collected, and the data is input into the TF-Transformer deep learning analysis model to obtain damage characteristics. Based on the damage characteristics, the damage evolution mechanism and development law of the cracked tunnel are summarized, and a safety evaluation index system for cracked tunnels is established.

[0035] In this embodiment of the application, firstly, based on as Figure 2 The lining structure-stratum complex simulation test system shown collects tunnel lining crack damage data. The monitoring instruments used include earth pressure cells, strain gauges, displacement gauges, crack rangefinders, acoustic emission testing devices, and piezoelectric ceramic equipment. The tunnel bedding angles are simulated by a polytetrafluoroethylene film, and the damage status is comprehensively determined by the acoustic emission testing device and the piezoelectric ceramic equipment.

[0036] Then, the data is input into the TF-Transformer deep learning analysis model to obtain damage features including the location and extent of damage. Based on the location and extent of damage, the damage evolution mechanism and development law of the fractured tunnel are summarized. For example... Figure 3 As shown, the structure of the TF-Transformer deep learning analysis model includes a feature segmentation module and a Transformer module. The feature segmentation module converts features into embeddings, which are then processed by the Transformer module. The Transformer module's process includes normalization, multi-head self-attention mechanism, and feedforward. Specifically, the feature segmentation module converts input features x into embeddings T∈Rk×d, given features... The embedding calculation is as follows:

[0037]

[0038] in, It is the j-th eigenvalue. It is the learnable bias vector corresponding to the j-th feature. These are transformation functions related to feature types.

[0039] In summary:

[0040]

[0041]

[0042]

[0043] in, It is a numerical eigenvalue. It is a learnable feature scaling vector. It is a learnable feature bias vector. It is the embedding result of numerical features; It is the embedding result of categorical features. It is a learnable feature bias vector. It is an embedded lookup operation, in which It is the one-hot vector of the corresponding category. It is a learnable embedding matrix; It is a token matrix. It represents the total number of features, and stack[⋅] represents vertical stacking, which means that all feature embeddings are sequentially concatenated into a matrix.

[0044] In practical applications, the TF-Transformer deep learning analysis model includes two steps: pre-training and application. The pre-trained model only needs to be input with bedding angle, strain data, crack width and depth, and initial crack type to determine the damage status of the tunnel lining, thereby better summarizing the damage evolution mechanism and development law of cracked tunnels, and thus establishing a safety evaluation index system for cracked tunnels.

[0045] In one embodiment of this application, the safety evaluation index system for cracked tunnels includes structural mechanical characteristic indexes, tunnel geological condition indexes, surrounding rock engineering characteristic indexes, and lining cracking characteristic indexes.

[0046] S103: Constructing a PL-VIKOR comprehensive evaluation model based on the safety evaluation index system for cracked tunnels and an improved subjective and objective weighting method based on the goodness of fit of normal distribution.

[0047] In this embodiment of the application, the weights of each safety evaluation index are determined by an improved subjective and objective weighting method based on the goodness of fit of the normal distribution, and a PL-VIKOR comprehensive evaluation model is constructed.

[0048] Specifically, in one embodiment of this application, the improved subjective-objective combination weighting method based on the goodness of fit of the normal distribution includes:

[0049] S201: The subjective weights are calculated using an improved analytic hierarchy process, and the objective weights are calculated using an improved entropy weighting method.

[0050] S203: Combination weighting based on the method of goodness of fit of the normal distribution of the sample evaluation result sequence.

[0051] In this embodiment, the improved Analytic Hierarchy Process (AHP) and the Entropy Weight Method (EWM) are used to calculate the subjective and objective weights, respectively. Based on these subjective and objective weights, a method using the goodness of fit of the normal distribution of the sample evaluation result sequence is employed for combined weighting. Optionally, the combined weighting method includes multiplicative addition, linear weighting, and game theory combined weighting.

[0052] Specifically, in one embodiment of this application, the step of calculating subjective weights using an improved analytic hierarchy process and objective weights using an improved entropy weight method includes:

[0053] An improved analytic hierarchy process (AHP) is based on a quasi-optimal transfer matrix, and an improved entropy weighting method is based on relative weights.

[0054] In one embodiment of this application, for the calculation of subjective weights, firstly, a judgment matrix is ​​constructed. The judgment matrix is ​​obtained by each expert through pairwise comparisons of the evaluation indicators using the 1-9 scaling method; wherein The number of evaluation indicators Indicates the first The first indicator is relative to the first The importance of each indicator is determined. Then, an antisymmetric matrix is ​​constructed based on the reciprocal matrix A. ,in Then, construct the optimal transfer matrix based on the antisymmetric matrix B. ,in Finally, the near-optimal transfer matrix of matrix A is calculated based on the optimal transfer matrix C. ,in This matrix can obtain the weight values ​​at once without needing a consistency check, effectively avoiding the blindness of adjusting the judgment matrix.

[0055] For the calculation of objective weights, firstly, data standardization is performed to eliminate the influence of differences in the units of measurement of different indicators. This involves standardizing positive indicators (larger is better) and negative indicators (smaller is better). Let be the value of the i-th evaluation indicator for the j-th objective. The formula for calculating the standardized value is as follows:

[0056] Positive indicators:

[0057] Negative indicators:

[0058] in, These are the original indicator values, representing the j-th indicator at the 1st rank. The values ​​of each sample; The minimum value of the indicator; This represents the maximum value of the indicator.

[0059] Next, the entropy value of each evaluation index is calculated. The proportion of the indicator among all indicators The calculation formula is as follows:

[0060] ,

[0061]

[0062] in, Let be the sample weight, representing the weight of the i-th sample in the i-th sample. ; The entropy value is used to reflect the degree of dispersion of the indicator data.

[0063] Next, the entropy weights of each evaluation index are calculated using the following formula:

[0064]

[0065] in, It is the average of all entropy values ​​that are not equal to 1; N is the precision coefficient, which is taken as N=41.27 here.

[0066] In one embodiment of this application, the method of combining weights based on the goodness of fit of the normal distribution of the sample evaluation result sequence includes:

[0067] Calculate the goodness of fit of the normal distribution under different combination coefficients of each method, and determine the final combination weighting result based on the goodness of fit of the normal distribution.

[0068] In one embodiment of this application, the goodness of fit of the normal distribution under different combination coefficients of each method is calculated by enumeration under the condition of varying combination coefficients (with a variation interval of 0.001). The goodness of fit of different weighting methods is compared and selected through goodness of fit calculation to obtain the final combined weighting result. Specifically, the weighting values ​​assigned when the goodness of fit of the multiplicative addition method, linear weighting method, and game theory combined weighting method is maximized are first selected. Then, the three are compared, and the maximum value among them is selected to determine the final combined weighting result. The calculation method of the goodness of fit of the normal distribution is as follows: First, for n samples... Arranged in ascending order, we get Based on this, standardization is performed, and the formula is as follows:

[0069]

[0070] in The sample mean. S is the sample standard deviation.

[0071] Next, calculate the empirical distribution function of the sample. The formula is as follows:

[0072]

[0073] Then, based on the empirical distribution function of the sample... With the standard normal distribution function Construct a statistic based on the maximum deviation between them. And the statistics It follows a Kolmogorov distribution, and its distribution function is expressed as:

[0074]

[0075] Finally, the expression value of goodness of fit is calculated. This expression represents and Under approximate conditions, the probability of an outcome that is more extreme than the obtained sample observations (i.e., less likely to prove the conclusion that the two distribution functions are similar) occurs.

[0076] In one embodiment of this application, the PL-VIKOR comprehensive evaluation model includes:

[0077] S301: Calculate the group utility value and individual regret value based on safety evaluation index data, index weights, positive ideal solutions, and negative ideal solutions.

[0078] S303: Calculate the cosine angle value based on safety evaluation index data.

[0079] S305: Calculate the benefit ratio based on the group utility value, individual regret value, and cosine angle value.

[0080] S307: Determine the safety level of the monitoring section based on the aforementioned benefit ratio.

[0081] In one embodiment of this application, the PL-VIKOR comprehensive evaluation method is used to establish a scientific and quantitative safety evaluation index system based on structural mechanical characteristics, tunnel geological conditions, surrounding rock engineering characteristics, and lining cracking characteristics. This system establishes both structural samples and measured samples to achieve real-time and objective assessment of the safety status of cracked tunnel structures. Specifically, the measured samples are obtained from monitoring data, while the structural samples are established based on the index system and its benchmarks. The benchmarks are proportionally divided, and these divided values ​​are used as input values ​​to form the structural samples. For example, if the baseline for index 1 is [0-0.25], [0.25-0.5], [0.5-0.75], [0.75-1.0], and the baseline for index 2 is [0-0.3], [0.3-0.5], [0.5-0.8], [0.8-1.0], then if the constructed samples are divided into 9 categories, the values ​​of these 9 constructed samples are: (0, 0), (0.125, 0.15), (0.25, 0.25)... The calculation formula is as follows: First, determine the positive ideal solution and the negative ideal solution.

[0082] Using the ideal solution as a reference:

[0083] Using the negative ideal solution as a reference:

[0084] in, , Let be the group utility value and individual regret value of the i-th sample with reference to the positive ideal solution; , The group utility value and individual regret value of the i-th sample with reference to the negative ideal solution; It is the positive ideal solution for the j-th evaluation index. It is the negative ideal solution for the j-th evaluation index. This represents the standardized indicator data.

[0085] Calculate the group utility value and individual regret value based on safety evaluation index data, index weights, positive ideal solutions, and negative ideal solutions:

[0086] based on Aggregate functions:

[0087] Take p=1 and calculate the cross section to be evaluated. Group utility under n indicators:

[0088]

[0089] in, It is the group utility value; The indicator weights are obtained by combining AHP and EWM weights. The normalized difference represents the relative distance between the current value and the optimal value.

[0090] Take p=∞, and calculate the cross section to be evaluated. Individual regrets under n indicators:

[0091]

[0092] in, It is an individual's regret value.

[0093] Then, the cosine angle value is calculated based on the safety evaluation index data:

[0094]

[0095] in, It is the index value of the i-th sample. It is the positive ideal solution vector.

[0096] Finally, the benefit ratio is calculated based on the group utility value, individual regret value, and cosine angle value, and the safety level of the monitoring section is determined based on the benefit ratio.

[0097]

[0098] in, Indicates the weight of group benefits. Indicates the individual regret weight, The similarity weights are adjusted based on the different levels of importance the organizers place on each component. The safety level of each monitoring section is determined by ranking the constructed and measured samples using Q-values.

[0099] S105: Establish a multi-dimensional crack damage lining monitoring system to monitor real-time tunnel cross-section data.

[0100] In this embodiment, a multi-dimensional cracked lining monitoring system is established. Surface strain gauges and surface crack gauges are installed at the cracked tunnel locations to construct a comprehensive monitoring network covering the appearance and internal state of the lining structure. This enables dynamic and comprehensive perception of the crack development process, providing real-time and reliable data support for structural safety assessment and treatment decisions.

[0101] S107: Based on the real-time tunnel cross-section data, a pre-trained LSTM model is used to predict the tunnel cross-section development results, and the future safety status of the tunnel is evaluated based on the tunnel cross-section development results combined with the PL-VIKOR comprehensive evaluation model.

[0102] In this embodiment, a pre-trained LSTM model is used to predict the tunnel cross-section development based on real-time tunnel cross-section data. The tunnel cross-section development results output by the LSTM model are then combined with the PL-VIKOR comprehensive evaluation model to assess the future safety status of the tunnel. This proactively predicts the evolution trend of the safety status of the lining structure with potential risks, thereby achieving early warning and early intervention.

[0103] LSTM models possess excellent memory capabilities, automatically extracting deep feature information from time series data, demonstrating good adaptability in prediction tasks involving complex time-varying data, such as tunnel engineering. In Long Short-Term Memory (LSTM) networks, the structure of the repeating modules is even more complex, consisting of four neural network layers that interact in a specific way. For example... Figure 4 As shown, the computation of a single neuron in LSTM consists of two parts: updating the neural network state and calculating the output value. An LSTM neuron contains three gating mechanisms: the input gate, the forget gate, and the output gate. These gating functions jointly regulate the transmission process of input, memory, and output values. The specific LSTM computation model is as follows:

[0104] (1) The forget gate is used to control the amount of information that needs to be forgotten in the current state of the neural network. The calculation process of the forget gate is as follows:

[0105]

[0106] In the formula, This represents the output of the forget gate. It was the hidden state from a previous moment. The sigmoid activation function is used. Here is the weight matrix for the forget gate. For the forget gate bias term, This is the input data for the current moment.

[0107] (2) The input gate consists of two parts, namely the input value of the input gate. And new candidate input values:

[0108]

[0109]

[0110] In the formula, the output value of the tanh function ranges from -1 to 1. This represents the candidate memory content for the current moment.

[0111] The input gate is used to filter information from the input layer, and its calculation process is as follows:

[0112]

[0113] In the formula, This represents the updated state of the neural network at the current moment. This refers to the state at the previous moment.

[0114] (3) The calculation process for the output gate and the hidden state is as follows:

[0115]

[0116]

[0117] In the formula, To output the gate value, The final output is controlled by the hidden state at the current moment, which is also the LSTM output at that moment.

[0118] To improve model training efficiency and prediction accuracy, preprocessing of the monitored data is necessary before establishing the database for training and testing. Preprocessing not only accelerates the training speed of neural network models but also effectively prevents gradient explosion during training. The tunnel monitoring data is normalized to ensure its values ​​are distributed within the range [0,1]. The input data normalization uses the min-maxScaler function from the Python deep learning library scikit-learn, and its calculation formula is as follows:

[0119]

[0120] In the formula, x is the original data value. The minimum value in the original data. The maximum value in the original data. These are the normalized data values.

[0121] During normalization, the maximum and minimum values ​​of the training set samples need to be recorded. After the model training is complete, the model's output needs to be denormalized to restore it to the true predicted value.

[0122] Meanwhile, to evaluate the accuracy of the LSTM network model in predicting monitoring data, the root mean square error (RmSE) and the coefficient of determination (R2) were selected as evaluation metrics for the prediction model.

[0123]

[0124]

[0125] In the formula, n is the sample size. For the i-th actual value, For the i-th predicted value, This is the root mean square error. The coefficient of determination.

[0126] In one embodiment of this application, the method further includes:

[0127] Based on the future safety status of the tunnel, a tiered intelligent response will be implemented.

[0128] In one embodiment of this application, targeted, graded intelligent measures are implemented based on the future safety status of the tunnel, as shown in Table 1 below:

[0129] Table 1

[0130]

[0131] In the aforementioned tunnel safety status assessment and prediction method based on lining damage evolution, firstly, tunnel lining crack data is collected using a lining structure-soil composite simulation test system. This data is then input into a TF-Transformer deep learning analysis model to obtain damage characteristics. Based on these characteristics, the damage evolution mechanism and development law of the cracked tunnel are summarized, and a safety evaluation index system for cracked tunnels is established. Next, a PL-VIKOR comprehensive evaluation model is constructed based on the cracked tunnel safety evaluation index system and an improved subjective-objective combination weighting method based on the goodness of fit of a normal distribution. Following this, a multi-dimensional cracked lining monitoring system is established to monitor real-time tunnel cross-sectional data. Finally, a pre-trained LSTM model is used to predict the tunnel cross-sectional development results based on the real-time tunnel cross-sectional data. The future safety status of the tunnel is then assessed based on the tunnel cross-sectional development results combined with the PL-VIKOR comprehensive evaluation model. In other words, by establishing a multi-dimensional cracked lining monitoring system, a cracked tunnel safety evaluation index system, and a cracked tunnel comprehensive evaluation method, and combining it with a prediction and early warning model calculated by LSTM, the evolution trend of the safety status of lining structures with potential risks can be predicted in a forward-looking manner based on monitoring data. This allows for timely and accurate understanding of the tunnel status at the monitoring section, and enables real-time and objective assessment of the safety status of cracked tunnel structures.

[0132] The following specific embodiment illustrates the detailed implementation steps of the tunnel safety status assessment and prediction method based on lining damage evolution of this application. Figure 5As shown, the tunnel lining damage evolution testing system first consists of an analysis model based on TF-Transformer deep learning and a lining structure-soil composite simulation test system. Through this damage evolution testing system, the damage evolution mechanism and development law of cracked tunnels are summarized, providing theoretical judgment for the establishment of a safety evaluation index system for cracked tunnels. For example, theoretically, cracking occurs at 300kN, so 0-300kN is set as Level 1 structural safety, and complete failure occurs after 900kN, correspondingly 900-1200kN is set as Level 4 structural unsafety. A multi-dimensional cracked tunnel lining monitoring system is established. By monitoring the lining in the field, data support can be provided for the establishment of a safety evaluation index system for cracked tunnels. For example, if the maximum monitored force is 1000kN, the upper limit of the evaluation index benchmark can be set to 1200kN (including some redundancy). An improved AHP-EWM combined weighting method based on the goodness of fit of a normal distribution is constructed, which can assign appropriate weights to the safety evaluation index system of cracked tunnels, thereby enabling timely evaluation of the safety status of cracked tunnels through the PL-VIKOR comprehensive evaluation method. The monitoring data provided by the multi-dimensional cracked tunnel lining monitoring system is preprocessed using the min-maxScaler function in the Python deep learning library scikit-learn, and an LSTM model is established to predict the development results of the monitoring data. The predicted data is then fed into the PL-VIKOR comprehensive evaluation model to predict the future safety status of cracked tunnels and propose targeted measures based on the safety status prediction.

[0133] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0134] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0135] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0136] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0137] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for assessing and predicting the safety status of tunnels based on the evolution of lining damage, characterized in that, The method includes: Based on the simulation test system of lining structure-stratum complex, tunnel lining crack damage data is collected, and the data is input into the TF-Transformer deep learning analysis model to obtain damage characteristics. Based on the damage characteristics, the damage evolution mechanism and development law of cracked tunnels are summarized, and a safety evaluation index system for cracked tunnels is established. A PL-VIKOR comprehensive evaluation model is constructed based on the safety evaluation index system for cracked tunnels and an improved subjective and objective combined weighting method based on the goodness of fit of normal distribution. Establish a multi-dimensional cracked lining monitoring system to monitor real-time tunnel cross-section data; Based on the real-time tunnel cross-section data, a pre-trained LSTM model is used to predict the tunnel cross-section development results. Based on the tunnel cross-section development results, the PL-VIKOR comprehensive evaluation model is used to assess the future safety status of the tunnel. The improved subjective-objective combination weighting method based on the goodness of fit of the normal distribution includes: The subjective weights are calculated using an improved analytic hierarchy process (AHP), and the objective weights are calculated using an improved entropy weighting method. A combined weighting method is used based on the goodness of fit of the normal distribution of the sample evaluation result sequence; The PL-VIKOR comprehensive evaluation model includes: Calculate the group utility value and individual regret value based on safety evaluation index data, index weights, positive ideal solutions, and negative ideal solutions; Calculate the cosine angle value based on safety evaluation index data; The benefit ratio is calculated based on the group utility value, individual regret value, and cosine angle value. The safety level of the monitoring section is determined based on the aforementioned benefit ratio.

2. The tunnel safety status assessment and prediction method based on lining damage evolution according to claim 1, characterized in that, The safety evaluation index system for cracked tunnels includes structural mechanical characteristic indexes, tunnel geological condition indexes, surrounding rock engineering characteristic indexes, and lining cracking characteristic indexes.

3. The tunnel safety status assessment and prediction method based on lining damage evolution according to claim 1, characterized in that, The calculation of subjective weights using the improved analytic hierarchy process and the calculation of objective weights using the improved entropy weight method include: An improved analytic hierarchy process (AHP) is based on a quasi-optimal transfer matrix, and an improved entropy weighting method is based on relative weights.

4. The tunnel safety status assessment and prediction method based on lining damage evolution according to claim 1, characterized in that, The method for combining weights based on the goodness of fit of the normal distribution of the sample evaluation result sequence includes: Calculate the goodness of fit of the normal distribution under different combination coefficients of each method, and determine the final combination weighting result based on the goodness of fit of the normal distribution.

5. The tunnel safety status assessment and prediction method based on lining damage evolution according to claim 1, characterized in that, The method further includes: Based on the future safety status of the tunnel, a tiered intelligent response will be implemented.

Images