Tunnel deformation prediction method based on DTW-SSA-DBN and application

Abstract

The invention provides a tunnel deformation prediction method based on DTW-SSA-DBN and an application, and belongs to the technical field of tunnel deformation monitoring, aiming at the problem that the accuracy of a prediction result is not high due to the fact that an existing tunnel deformation prediction method considers few influence factors and too few sample data. The method comprises the following steps: constructing a tunnel deformation prediction index system based on surrounding rock geological parameters, tunnel design parameters and construction factors; based on the tunnel deformation prediction index system, constructing a DTW-SSA-DBN-based tunnel deformation prediction model, and performing model training and verification to obtain a target tunnel deformation prediction model; and inputting tunnel deformation prediction index data of a to-be-predicted target section into the target tunnel deformation prediction model, and outputting a tunnel deformation amount at a specified monitoring time. According to the method, the geological phase change driving force item is added in the energy function of the DBN network, the hidden layer neuron activation mode can be forced to suddenly change, corresponding surrounding rock plastic deformation or damage is achieved, and therefore the accuracy of the prediction model can be improved.

Description

Technical Field

[0001] This application relates to the field of tunnel deformation monitoring technology, and in particular to a tunnel deformation prediction method and its application based on DTW-SSA-DBN. Background Technology

[0002] Tunnel deformation is a crucial indicator of engineering safety, and accurately predicting deformation trends is essential for ensuring construction and operational safety. Tunnel deformation directly reflects the structural stress state; excessive deformation can lead to structural instability, causing serious safety accidents such as collapses and cracking. Tunnel deformation can also cause ground subsidence, threatening the safety of superstructures. In recent years, with the acceleration of urbanization and the increasing depth of underground space development, the importance of tunnel deformation prediction has become increasingly prominent.

[0003] Currently used methods for predicting tunnel deformation all have their own limitations. For example, empirical / semi-empirical formula methods can only be used under specific geological conditions, and their accuracy drops significantly when dealing with complex, heterogeneous rock masses. Numerical simulation methods require a large number of accurate geotechnical parameters; however, reliable input data is often difficult to obtain under complex geological conditions. Machine learning prediction methods, by learning from a large amount of historical monitoring data, can automatically capture the nonlinear relationships between various factors such as different geotechnical conditions, support methods, and construction procedures. This allows them to provide a more unified prediction model across different geological environments, overcoming the limitation of empirical formulas that are "only applicable to specific conditions." In numerical simulations, parameter uncertainty is the main source of error. Machine learning models can directly use monitored quantities (displacement, settlement, stress, etc.) and construction condition parameters as input, eliminating the strict requirements for precise values ​​of elastic modulus, friction coefficient, etc., thus maintaining good prediction accuracy even in complex geological conditions where parameters are difficult to obtain. However, most current methods for tunnel deformation prediction based on neural networks consider relatively few influencing factors. They typically only use monitoring time as the input indicator for the prediction model, neglecting the impact of geological factors (such as uniaxial compressive strength and rock mass integrity index), tunnel design (burial depth), and construction factors (monitoring time) on tunnel deformation. Furthermore, some studies suffer from insufficient sample data, resulting in limited applicability and poor generalization ability of the trained prediction models. For example, the paper "RTCN On-Street Parking Space Prediction Model Considering Regional Occupancy" addresses the problem of ineffective traffic volume caused by repeated parking searches during vehicle parking by proposing an RTCN short-term vacant parking space prediction model considering regional occupancy. The paper points out that the insufficient number of on-street parking space samples is a problem in on-street parking space prediction research. Current convolutional neural networks (CNNs) and recurrent neural networks (RNNs) have shortcomings in short-term vacant parking space prediction research, and this problem further limits the effectiveness of the prediction models. Due to insufficient sample size, the model struggles to fully learn the changing patterns of vacant parking spaces under different times and spaces.

[0004] In conclusion, how to consider the various factors affecting tunnel deformation, establish a more complete and comprehensive tunnel deformation prediction model to improve prediction accuracy, and effectively ensure project safety are urgent problems to be solved. Summary of the Invention

[0005] In view of the above problems, this application takes into account geological factors, tunnel design and construction factors, and proposes a tunnel deformation prediction method and device based on DTW-SSA-DBN, so as to overcome the above problems or at least partially solve them.

[0006] In a first aspect, embodiments of this application provide a tunnel deformation prediction method based on DTW-SSA-DBN, including:

[0007] A tunnel deformation prediction index system was constructed based on surrounding rock geological parameters, tunnel design parameters, and construction factors.

[0008] Based on the tunnel deformation prediction index system, a tunnel deformation prediction model based on DTW-SSA-DBN was constructed, and the model was trained and validated to obtain the target tunnel deformation prediction model.

[0009] Input the tunnel deformation prediction index data of the target section to be predicted into the target tunnel deformation prediction model, and output the predicted value of tunnel deformation at the specified monitoring time.

[0010] Optionally, the tunnel deformation prediction index system includes nine primary indicators: rock saturated uniaxial compressive strength, rock integrity index, rock unit weight, elastic resistance coefficient, elastic modulus, Poisson's ratio and internal friction angle, burial depth and monitoring time.

[0011] Optionally, based on the tunnel deformation prediction index system, a tunnel deformation prediction model based on DTW-SSA-DBN is constructed, and the model is trained and validated to obtain the target tunnel deformation prediction model, including:

[0012] Based on the tunnel deformation prediction index system, actual monitoring data of each primary index in tunnel construction are obtained, a tunnel deformation prediction dataset is constructed, and it is divided into a training set and a test set.

[0013] Based on the improved DBN model, a tunnel deformation prediction model based on DTW-SSA-DBN is established;

[0014] The training set data is input into the tunnel deformation prediction model based on DTW-SSA-DBN, the hyperparameters of the DBN network are adjusted, and the DBN network model after the last parameter adjustment is output.

[0015] The DBN network model after the last parameter adjustment is evaluated on the test set, and the predicted value of the DBN network model is output.

[0016] The predicted values ​​for different cross sections were compared and verified with the actual monitoring values, and the verified model was used as the target tunnel deformation prediction model.

[0017] Optionally, establishing the tunnel deformation prediction model based on DTW-SSA-DBN includes:

[0018] The Dynamic Time Warping (DTW) module is used to classify the input tunnel deformation prediction dataset based on the Dynamic Time Warping (DTW) algorithm to obtain a time-consistent standardized dataset.

[0019] The Deep Belief Network module is used to predict tunnel deformation based on time-consistent standardized datasets using an improved DBN network.

[0020] The hyperparameter optimization module is used to optimize the hyperparameters of the improved DBN model based on the SSA algorithm.

[0021] Optionally, the improved DBN network is based on the original DBN network model. A geological phase transition driving force term is introduced into the energy function of its restricted Boltzmann machine (RBM) to model and reconstruct the energy function for abrupt deformation of the surrounding rock.

[0022] Optionally, the reconstructed energy function is:

[0023]

[0024] in, The surrounding rock stress gradient corresponding to the j-th feature is... This represents the critical threshold for rock mass yielding. The phase transition coupling coefficient is... As a softening factor, This represents the bias term for the visible layer. This represents the bias term of the hidden layer. Indicates the state of the visible layer. Indicates the state of the hidden layer. The connection weight between the visible and hidden layers is represented by F, which represents the rock mass unit set.

[0025] Secondly, embodiments of this application provide a tunnel deformation prediction device based on DTW-SSA-DBN, comprising:

[0026] The indicator system construction module is used to construct a tunnel deformation prediction indicator system based on surrounding rock geological parameters, tunnel design parameters, and construction factors.

[0027] The model building module is used to construct a tunnel deformation prediction model based on DTW-SSA-DBN based on the tunnel deformation prediction index system, and to train and validate the model to obtain the target tunnel deformation prediction model.

[0028] The prediction module is used to input the tunnel deformation prediction index data of the target section to be predicted into the target tunnel deformation prediction model, and output the predicted value of tunnel deformation at the specified monitoring time.

[0029] Thirdly, embodiments of this application provide an electronic device, characterized in that it includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the aforementioned tunnel deformation prediction method based on DTW-SSA-DBN.

[0030] Fourthly, embodiments of this application provide a readable storage medium, characterized in that a program or instructions are stored on the readable storage medium, and when the program or instructions are executed by a processor, they implement the above-described tunnel deformation prediction method based on DTW-SSA-DBN.

[0031] The specific beneficial effects are as follows:

[0032] First, this invention establishes a tunnel deformation prediction index system that takes into account the geological parameters of the surrounding rock and time factors. This system includes nine easily obtainable indicators: rock saturated uniaxial compressive strength, rock integrity index, rock unit weight, elastic resistance coefficient, elastic modulus, Poisson's ratio, internal friction angle, burial depth, and monitoring time. These evaluation indicators are scientific, reasonable, and easy to obtain. The prediction model established based on this index system has high accuracy and can predict the tunnel deformation situation and trend, provide effective guidance on the timing of secondary lining construction, and effectively ensure the safety and stability of tunnel construction.

[0033] Secondly, this invention constructs a tunnel deformation prediction method based on the DTW-SSA-DBN model. The DTW-SSA-DBN model, through DTW, can perform similarity matching and classification of tunnel deformation monitoring sequences with inconsistent lengths and nonlinear deformations on the time axis (such as monitoring frequency fluctuations and temporal misalignments caused by missing data), eliminating data interference caused by temporal differences and outputting a standardized and highly consistent training dataset. Furthermore, the energy function of the DBN network is improved by strictly correlating the Gibbs phase transition driving force, rock mass stress gradient, and deep learning energy function through Legendre transformation, breaking through the limitations of static data-driven traditional DBN and improving the accuracy of the prediction model. The hyperparameters of the DBN prediction model are optimized using the SSA algorithm, which can utilize the global search capability of a sparrow swarm to quickly traverse the hyperparameter space and find the optimal combination, avoiding the problems of low efficiency and easy getting trapped in local optima in traditional grid search. This also improves the model's generalization ability and prediction accuracy. Finally, experimental results show that the accuracy and precision of the prediction model proposed in this invention meet engineering requirements and have good generalization ability. Attached Figure Description

[0034] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 The basic process of the DTW algorithm;

[0036] Figure 2 This refers to the RBM structure in the DBN model;

[0037] Figure 3 The network structure of the DBN model;

[0038] Figure 4 Here is a flowchart of the SSA algorithm;

[0039] Figure 5 The flowchart shows the tunnel deformation prediction method based on DTW-SSA-DBN proposed in this invention.

[0040] Figure 6 The comparison between the predicted values ​​and the monitored values ​​of the DTW-SSA-DBN model is shown, where: (a) is the crown settlement; (b) is the horizontal convergence. Detailed Implementation

[0041] Exemplary embodiments of this application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of this application are shown in the drawings, it should be understood that this application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of this application and to fully convey the scope of this application to those skilled in the art.

[0042] Example 1:

[0043] This invention provides a tunnel deformation prediction method based on DTW-SSA-DBN, which may include the following steps:

[0044] Step 1: Construct a tunnel deformation prediction index system based on surrounding rock geological parameters, tunnel design parameters, and construction factors;

[0045] In this embodiment, geological parameters of the surrounding rock are obtained based on geological survey data and design manuals for tunnel engineering, and the saturated uniaxial compressive strength of the rock related to the deformation of the surrounding rock is selected from these parameters. Rock mass integrity index , rock weight Elastic resistance coefficient Elastic modulus Poisson's ratio and internal friction angle burial depth and monitoring time Taking into account, a total of 9 parameters constitute the tunnel deformation prediction index system.

[0046] Step 2: Based on the tunnel deformation prediction index system, construct a tunnel deformation prediction model based on DTW-SSA-DBN, and train and validate the model to obtain the target tunnel deformation prediction model;

[0047] Optionally, step 2 may include the following sub-steps:

[0048] Step 2.1: Based on the tunnel deformation prediction index system, obtain the actual monitoring data of each index during tunnel construction and construct a tunnel deformation prediction dataset;

[0049] Specifically, in this embodiment, a tunnel deformation prediction dataset is constructed using seven sets of surrounding rock geological information and actual monitoring data from 21 sections of the Beiziyan Tunnel, Dongfang'ao Tunnel, and Zimeiyan Tunnel in a certain highway project. This dataset includes 522 sets of crown settlement data and 493 sets of horizontal convergence data. Next, the crown settlement data and horizontal convergence data are randomly arranged and divided into a training set (70%) and a test set (30%). The training set is used to train the model and adjust hyperparameters, while the test set is used to evaluate model performance.

[0050] Step 2.2: Establish a tunnel deformation prediction model based on DTW-SSA-DBN;

[0051] The tunnel deformation prediction model of DTW-SSA-DBN includes a dynamic time warping module, a hyperparameter optimization module, and a deep belief network module; wherein:

[0052] (1) Dynamic Time Warping module, which is used to classify the input tunnel deformation prediction dataset based on the Dynamic Time Warping (DTW) algorithm to obtain a time-consistent standardized dataset;

[0053] The DTW algorithm is used to measure the similarity between two time series, and it is particularly suitable for handling sequences with inconsistent lengths or nonlinear deformations of the time axis (such as speed variations). Its core idea is to find the optimal alignment path between two sequences through dynamic programming, minimizing the cumulative distance, thereby more accurately reflecting the shape similarity between the sequences, rather than strict time point alignment.

[0054] The core principle of DTW is to construct a distance matrix, where each element represents the distance between two sequences at different positions (usually Euclidean distance). Then, dynamic programming is used to find the minimum path from the starting point to the ending point, which satisfies constraints such as continuity and monotonicity. Because it allows sequences to be locally stretched or compressed along the time axis, it can handle local deformation or temporal misalignment problems in time series.

[0055] Furthermore, the specific steps of the DTW algorithm include:

[0056] A1: A reference sequence is obtained by standardization based on the historical deformation sequence of monitoring points in the training set;

[0057] Based on the historical deformation sequences of monitoring points in the training set, a standardized baseline sequence is obtained through standardization processing. This standardized baseline sequence is then used as a reference sequence.

[0058] A2: Calculate the Euclidean distance between the historical deformation sequence and the reference sequence at each monitoring point to obtain the Euclidean distance matrix;

[0059] Specifically, based on the historical deformation sequences of each monitoring point extracted from the training set. and reference sequence The Euclidean distance between each monitoring point is calculated according to the following formula, and the Euclidean distance matrix is ​​obtained. :

[0060] (1)

[0061] in, A Euclidean distance matrix representing the historical deformation monitoring point locations and the reference monitoring point locations; This indicates the location of the i-th historical deformation monitoring point. Indicates the location of the j-th reference monitoring point;

[0062] A3: Based on the Euclidean distance matrix between monitoring points, the cumulative distance matrix is ​​obtained recursively using dynamic programming;

[0063] Based on dynamic programming recursion, the recursive formula for the cumulative distance matrix C is obtained as follows:

[0064] (2)

[0065] in, The element in the i-th row and j-th column of the cumulative distance matrix represents the minimum cumulative distance from the starting point (1,1) of the sequence to (i,j).

[0066] A4: Determine the optimal path based on the minimum value of the cumulative distance matrix:

[0067] (3)

[0068] in, For alignment point pairs, T represents the path length.

[0069] (2) Deep belief network module, used to predict tunnel deformation based on time-consistent standardized datasets using an improved DBN network;

[0070] Specifically, the Deep Belief Network (DBN) was first proposed by Hinton, Osindero, and Teh in 2006. It consists of multiple stacked Restricted Boltzmann Machines (RBMs). This multi-layered structure enables DBNs to learn high-order abstract features from data, performing particularly well when dealing with complex data. It can be used for tasks such as feature extraction, data classification, image synthesis, and text generation. Figure 3 As shown, DBN is composed of multiple RBM models connected in series. Its structure includes one visible layer (input layer), multiple hidden layers and one output layer. Every two layers form an RBM, and the hidden layer of each RBM is also the visible layer of the next RBM.

[0071] The RBM is a probabilistic generative model used to learn the probability distribution of input data. An RBM consists of a visible layer and hidden layers. The visible layer receives the input data, while the hidden layers learn data features from the visible layer and extract higher-order features. There are bidirectional connections between the visible and hidden layers, with connection weights representing the connection strength. Neurons within the same layer are not connected to each other. Its structure is as follows: Figure 2 As shown, it uses an energy function to describe the energy of different states in the model. For a given visible and hidden layer state, the energy function of the RBM is... It can be expressed as the following formula:

[0072] (4)

[0073] in, This represents the bias term for the visible layer. This represents the bias term of the hidden layer. Indicates the state of the visible layer. Indicates the state of the hidden layer. The values ​​represent the connection weights between the visible and hidden layers. i represents the index of all neurons in the visible layer, and its range covers the total number of neurons in the visible layer. j represents the index of all neurons in the hidden layer, and its range covers the total number of neurons in the hidden layer.

[0074] To overcome the shortcomings of traditional DBN models where the RBM energy function relies solely on statistical data characteristics and is disconnected from the physical mechanism of tunnel surrounding rock deformation, and to address its inability to respond to the critical abrupt change in surrounding rock deformation from elastic to plastic, this application improves the RBM energy function in the original DBN model. Specific improvements include:

[0075] First, according to the second law of thermodynamics, under isothermal and isobaric conditions, the direction of the phase transition process of "elastic deformation → plastic deformation / failure" in the surrounding rock due to stress is determined by the decrease in Gibbs free energy (G).

[0076] When the parent phase (α) of the rock mass transforms into the new phase (β), the driving force is the Gibbs free energy difference:

[0077] (5)

[0078] in, For Gibbs free energy difference, For the Gibbs free energy of the new phase β, The Gibbs free energy of the parent phase α;

[0079] When the supercooling is relatively small It can be represented as:

[0080] (6)

[0081] Where T represents the actual temperature, and L represents the latent heat of phase change per unit volume. m This indicates the equilibrium temperature between the two phases.

[0082] Treating temperature and stress gradients as conjugate driving sources, the phase transformation of rock mass is essentially caused by external stress performing work leading to mineral lattice reorganization. According to the principle of energy equivalence, the stress field performs work... With thermal driving energy W T Equivalence during phase transition, i.e.:

[0083] (7)

[0084] in, This is plastic strain (macroscopic manifestation of phase transition). This represents the actual stress in the surrounding rock of the tunnel. Entropy;

[0085] Secondly, the temperature term in equation (6) is replaced with a stress term, specifically including:

[0086] 1) Define the critical threshold for rock mass yielding. :

[0087] (8)

[0088] in, , These are geological calibration coefficients, determined by triaxial tests. It represents the compressive strength of the rock.

[0089] 2) Establish analogy:

[0090] (9)

[0091] in, For the stress gradient of the surrounding rock, Indicates proportional to; Represents the elastic state (analogous to T>T) m ), Indicates plastic yielding / phase transformation (analogous to T) <T m );

[0092] 3) Based on the analogy, replace the temperature term with a stress term:

[0093] Substituting equation (9) into equation (6), we get:

[0094] (10);

[0095] Furthermore, in order to accurately capture yield mutations ( (the nonlinear response at the given point) is constructed as follows:

[0096] (11)

[0097] in, This is the softening factor, used to control the steepness of the function; the default value is [value to be filled in]. =0.05;

[0098] From equations (10) and (11), the driving force term for geological phase transition is defined as follows:

[0099] (12)

[0100] in, This is the driving force term for geological phase transition. The phase transition coupling coefficient is calibrated through geostress field simulation, converting the geological driving force (MPa) into an energy function scalar (J).

[0101] Finally, a geological phase transition driving force term is introduced into the original RBM energy function to model the abrupt deformation of the surrounding rock and reconstruct the energy function:

[0102] (13)

[0103] in, The stress gradient of the surrounding rock corresponding to the j-th feature (from geological survey). is the critical threshold for rock mass yielding, and F represents the rock mass unit set.

[0104] The improved RBM energy function directly correlates physical parameters such as surrounding rock stress gradient and yield threshold with the energy state through the addition of a geological phase transition driving force term. Therefore, this application, by improving the energy function, directly correlates local stress gradient... hour, Adding a positive penalty term (geological phase transition driving force term) to the energy function can force the activation mode of hidden layer neurons to mutate, corresponding to plastic deformation or failure of the surrounding rock, thereby improving the accuracy of the prediction model.

[0105] The joint probability distribution of RBMs, used to provide probabilistic support for feature learning in DBN models, is defined by the Boltzmann distribution. Based on a new energy function, the new probability mass function is shown in the following equation:

[0106] (14)

[0107] Where Z represents the partition function, used for normalization;

[0108] (3) Hyperparameter optimization module, used to optimize the hyperparameters of the improved DBN model based on the SSA algorithm.

[0109] To further improve the performance and accuracy of the model, this invention employs the Sparrow Search Algorithm (SSA) to optimize the hyperparameters of the DBN network. The Sparrow Search Algorithm (SSA), proposed in 2020 by Xue and Shen, is a novel bee colony optimization algorithm inspired by the swarm intelligence inherent in the foraging and cooperative behaviors of sparrows. It boasts advantages such as simple execution, few parameters, and high flexibility, and possesses a strong ability to explore globally potential optimal regions, thus avoiding local optima problems. Its flowchart is shown below. Figure 4 As shown.

[0110] Combination Figure 4 The basic steps of the SSA algorithm include:

[0111] B1: Group initialization;

[0112] A group of sparrows is randomly generated, and each sparrow represents a potential solution in the problem space.

[0113] B2: Objective function evaluation;

[0114] For each individual, calculate its objective function value in the optimization problem, i.e., its fitness; the objective function is the evaluation function of the problem to be optimized, which can be an objective that needs to be minimized or maximized.

[0115] B3: Group Behavior Simulation;

[0116] SSA divides sparrows into Producers and Scroungers. Individual sparrows move according to the following rules: ① An individual uses its current position as the center and generates a new position through a certain random perturbation; ② If the fitness of the new position is better than the current position, the individual moves towards the new position; ③ By considering information from other sparrows in the group, an individual can adjust its movement direction.

[0117] B4: Local search and global search;

[0118] SSA balances exploration and utilization through two levels: local search and global search. Local search seeks a better solution within the neighborhood by moving the individual itself, while global search explores a wider range by considering information from the entire population.

[0119] B5: Iterative update;

[0120] By repeatedly simulating individual behavior, evaluating the objective function, and updating the position, the position of individuals in the group is gradually optimized, so that their fitness gradually approaches the optimization objective.

[0121] B6: Stopping Criteria;

[0122] The iteration stops when the predetermined number of iterations is reached or a specific convergence criterion is met.

[0123] B7: Output of the optimal solution;

[0124] After the algorithm finishes running, it outputs the individual with the best fitness value, which is the optimal solution to the optimization problem.

[0125] This invention improves the efficiency of hyperparameter optimization in the DBN model by introducing SSA into it. The hyperparameters of SSA are set as follows: number of sparrows pop=6, maximum number of iterations Max_iteration=30, and the remaining hyperparameters are consistent with those of the DBN model, thus forming an optimized tunnel deformation prediction SSA-DBN model.

[0126] Step 2.3: Input the training set data into the tunnel deformation prediction model based on DTW-SSA-DBN, adjust the hyperparameters of the DBN network, and output the DBN network model after the last parameter adjustment.

[0127] The training process of DBN is divided into two stages: pre-training and fine-tuning. In the pre-training stage, each RBM is greedily trained sequentially, learning different levels of abstract feature representations of the input data layer by layer through unsupervised learning. In the fine-tuning stage, supervised learning methods (such as backpropagation BP networks) are used to fine-tune the entire network structure in reverse, further improving performance.

[0128] The hyperparameters of DBN were set as follows: number of hidden layer nodes dbn.sizes=20, number of training iterations numepochs=500, learning rate alpha=0.01, number of training samples per batch size=5, and number of back-tuning iterations numepochs=500. Finally, a DBN model for tunnel deformation prediction was constructed.

[0129] Step 2.4: Evaluate the DBN network model after the last parameter adjustment on the test set and output the predicted value of the DBN network model;

[0130] Step 2.5: Compare and verify the predicted values ​​of different cross sections with the actual monitoring values, and use the verified model as the target tunnel deformation prediction model;

[0131] In the embodiments of this application, the accuracy and feasibility of the model are verified by comparing the actual monitored values ​​with the predicted values ​​of the DTW-SSA-DBN model.

[0132] Step 3: Input the tunnel deformation prediction index data of the target section into the target tunnel deformation prediction model, and output the tunnel deformation at the specified monitoring time. The tunnel deformation is the predicted value of crown settlement and horizontal convergence.

[0133] Simulation Case

[0134] This simulation case verifies the effectiveness of the proposed method in the following aspects through simulation experiments. Three typical tunnels (Tunnel 1, Tunnel 2, and Tunnel 3) from a highway project in a certain area are selected as the data source, as shown in Table 1. Furthermore, to adapt to the training and prediction requirements of the DTW-SSA-DBN model, the original data undergoes the following preprocessing:

[0135] 1. Data cleaning: Remove outliers (such as sudden data changes caused by instrument malfunctions) and missing values, and use linear interpolation to complete a small amount of discontinuous monitoring data to ensure the continuity of time series data.

[0136] 2. Indicator Standardization: The nine primary indicators are normalized (the values ​​are mapped to the [0,1] interval) to eliminate the interference of dimensional differences (such as intensity unit MPa and burial depth unit m) on model training.

[0137] 3. Time series normalization: The deformation time series data of each section are classified and processed by the DTW algorithm to eliminate the time series misalignment caused by slight differences in the monitoring frequency of different sections, and output a standardized time series dataset.

[0138] 4. Dataset partitioning: The complete cross-sectional data is divided into a training set (for model training and hyperparameter optimization) and a test set (for model validation) in a 7:3 ratio. The data from the three cross-sections in Table 1 are all from the test set and are used for final effect validation.

[0139] Specifically, three typical cross-sections with different burial depths and surrounding rock parameters were selected from the test set; the data of nine indicators for each cross-section were organized according to the monitoring time series to form an input data matrix; the input data matrix was then substituted into the trained DTW-SSA-DBN target model to output the arch crown settlement prediction value (GD01) and horizontal convergence prediction value (SL01) for each monitoring time point (0 days, 1 day, 2 days...21 days); finally, the model prediction values ​​were compared with the deformation data measured in the field during the same period, sorted by "monitoring time", and the final Table 1 format was obtained.

[0140] The results of the example are shown in Table 1. As can be seen from Table 1, the settlement of the crown and the deformation rate of horizontal convergence of the three sections on the 21st day are all less than 0.2 mm / d, which can be judged as basically stable and secondary lining construction can be carried out.

[0141] Table 1

[0142]

[0143]

[0144] Figure 6 The prediction results are based on monitoring data from three typical tunnels in a highway project in the aforementioned area. As can be seen from the figure, regardless of the crown settlement ( Figure 6 (a) or horizontal convergence (a) Figure 6 In section (b), the predicted and measured values ​​showed a high degree of consistency in their trends, exhibiting the typical tunnel deformation pattern of "rapid growth in the early stage and gradual flattening in the later stage," with no deviation from the trend. The deviation between the predicted and measured values ​​for all sections was controlled within 0.5 mm, which is lower than the allowable error threshold of ±1 mm for engineering.

[0145] Example 2:

[0146] This invention provides a tunnel deformation prediction device based on DTW-SSA-DBN, which is implemented according to a tunnel deformation prediction method based on DTW-SSA-DBN, and includes the following modules:

[0147] The indicator system construction module is used to construct a tunnel deformation prediction indicator system based on surrounding rock geological parameters, tunnel design parameters, and construction factors.

[0148] The model building module is used to construct a tunnel deformation prediction model based on DTW-SSA-DBN based on the tunnel deformation prediction index system, and to train and validate the model to obtain the target tunnel deformation prediction model.

[0149] The prediction module is used to input the tunnel deformation prediction index data of the target section to be predicted into the target tunnel deformation prediction model and output the tunnel deformation amount at the specified monitoring time.

[0150] Optionally, the model building module includes:

[0151] The dataset construction unit is used to obtain actual monitoring data of various indicators in tunnel construction based on the tunnel deformation prediction index system, construct the tunnel deformation prediction dataset, and divide it into training set and test set.

[0152] Building blocks are used to establish tunnel deformation prediction models based on DTW-SSA-DBN;

[0153] The training unit is used to input the training set data into the tunnel deformation prediction model based on DTW-SSA-DBN, adjust the hyperparameters of the DBN network, and output the DBN network model after the last parameter adjustment.

[0154] The evaluation unit is used to evaluate the DBN network model after the last parameter adjustment on the test set and output the predicted value of the DBN network model.

[0155] The verification unit is used to compare and verify the predicted values ​​with the actual monitoring values ​​of different cross sections, and the verified model is used as the target tunnel deformation prediction model.

[0156] This application provides an electronic device, which includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the method of the aforementioned embodiments.

[0157] This application also provides a computer-readable storage medium storing a computer program / instructions thereon, which, when executed by a processor, implements the steps in the tunnel deformation prediction method based on DTW-SSA-DBN disclosed in this application.

[0158] This application also provides a computer program product that, when run on an electronic device, enables a processor to execute the steps in the tunnel deformation prediction method based on DTW-SSA-DBN disclosed in this application.

[0159] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0160] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, apparatuses, electronic devices, and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0161] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0162] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0163] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0164] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0165] The tunnel deformation prediction method and apparatus based on DTW-SSA-DBN provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and its core ideas. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A tunnel deformation prediction method based on DTW-SSA-DBN, characterized in that, The method includes: A tunnel deformation prediction index system was constructed based on surrounding rock geological parameters, tunnel design parameters, and construction factors. Based on the tunnel deformation prediction index system, a tunnel deformation prediction model based on DTW-SSA-DBN was constructed, and the model was trained and validated to obtain the target tunnel deformation prediction model. Input the tunnel deformation prediction index data of the target section to be predicted into the target tunnel deformation prediction model, and output the predicted value of tunnel deformation at the specified monitoring time.

2. The method according to claim 1, characterized in that, The tunnel deformation prediction index system includes nine primary indicators: rock saturated uniaxial compressive strength, rock integrity index, rock unit weight, elastic resistance coefficient, elastic modulus, Poisson's ratio, internal friction angle, burial depth, and monitoring time.

3. The method according to claim 2, characterized in that, The tunnel deformation prediction index system is used to construct a tunnel deformation prediction model based on DTW-SSA-DBN, and the model is trained and validated to obtain the target tunnel deformation prediction model, including: Based on the tunnel deformation prediction index system, actual monitoring data of each primary index in tunnel construction are obtained, a tunnel deformation prediction dataset is constructed, and it is divided into a training set and a test set. Based on the improved DBN model, a tunnel deformation prediction model based on DTW-SSA-DBN is established; The training set data is input into the tunnel deformation prediction model based on DTW-SSA-DBN, the hyperparameters of the DBN network are adjusted, and the DBN network model after the last parameter adjustment is output. The DBN network model after the last parameter adjustment is evaluated on the test set, and the predicted value of the DBN network model is output. The predicted values ​​for different cross sections were compared and verified with the actual monitoring values, and the verified model was used as the target tunnel deformation prediction model.

4. The method according to claim 3, characterized in that, The establishment of the tunnel deformation prediction model based on DTW-SSA-DBN includes: The Dynamic Time Warping (DTW) module is used to classify the input tunnel deformation prediction dataset based on the Dynamic Time Warping (DTW) algorithm to obtain a time-consistent standardized dataset. The Deep Belief Network module is used to predict tunnel deformation based on time-consistent standardized datasets using an improved DBN network. The hyperparameter optimization module is used to optimize the hyperparameters of the improved DBN model based on the SSA algorithm.

5. The method according to claim 4, characterized in that, The improved DBN network is based on the original DBN network model. A geological phase transition driving force term is introduced into the energy function of its restricted Boltzmann machine (RBM) to model and reconstruct the energy function for abrupt deformation of the surrounding rock.

6. The method according to claim 5, characterized in that, The energy function for reconstruction is: ； in, The surrounding rock stress gradient corresponding to the j-th feature is... This represents the critical threshold for rock mass yielding. The phase transition coupling coefficient is... As a softening factor, This represents the bias term for the visible layer. This represents the bias term of the hidden layer. Indicates the state of the visible layer. Indicates the state of the hidden layer. The connection weight between the visible and hidden layers is represented by F, which represents the rock mass unit set.

7. A tunnel deformation prediction device based on DTW-SSA-DBN, characterized in that, Implemented based on any one of the methods described in claims 1-6, comprising: The indicator system construction module is used to construct a tunnel deformation prediction indicator system based on surrounding rock geological parameters, tunnel design parameters, and construction factors. The model building module is used to construct a tunnel deformation prediction model based on DTW-SSA-DBN based on the tunnel deformation prediction index system, and to train and validate the model to obtain the target tunnel deformation prediction model. The prediction module is used to input the tunnel deformation prediction index data of the target section to be predicted into the target tunnel deformation prediction model, and output the predicted value of tunnel deformation at the specified monitoring time.

8. The apparatus as claimed in claim 7, characterized in that, The model building module includes: The dataset construction unit is used to obtain actual monitoring data of each primary indicator in tunnel construction, construct a tunnel deformation prediction dataset, and divide it into training set and test set; Building units are used to establish a tunnel deformation prediction model based on DTW-SSA-DBN based on the improved DBN model; The training unit is used to input the training set data into the tunnel deformation prediction model based on DTW-SSA-DBN, adjust the hyperparameters of the DBN network, and output the DBN network model after the last parameter adjustment. The evaluation unit is used to evaluate the DBN network model after the last parameter adjustment on the test set and output the predicted value of the DBN network model. The verification unit is used to compare and verify the predicted values ​​of different cross sections with the actual monitoring values, and to use the verified model as the target tunnel deformation prediction model.

9. An electronic device, characterized in that, The method includes a memory, a processor, and a computer program stored on the memory, wherein the processor executes the computer program to implement the method as claimed in any one of claims 1-6.

10. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the method as described in any one of claims 1-6.

Images