A tunnel deformation prediction method based on multi-source data fusion

Abstract

The present invention discloses a tunnel deformation prediction method based on multi-source data fusion, which relates to the technical field of tunnel deformation monitoring. The method comprises collecting tunnel data for preprocessing, fusing the preprocessed tunnel data, extracting tunnel features, constructing a tunnel deformation prediction model to predict tunnel deformation, evaluating and issuing warnings based on real-time prediction results, constructing a visual interface to display tunnel data in real time, and storing, collecting, and analyzing the generated tunnel data. The present invention collects tunnel data for fusion, constructs a tunnel deformation prediction model to predict tunnel deformation, and optimizes the data representation capability. This method not only captures time series features in historical data, but also enables dynamic prediction based on current real-time data, significantly improving the early warning capability of tunnel deformation, providing strong technical support for safe tunnel operation and management, and reducing safety hazards caused by data errors.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel deformation monitoring, and in particular to a tunnel deformation prediction method based on multi-source data fusion. Background Art

[0002] As an important component of modern infrastructure, the stability and safety of tunnel engineering have always been the focus of engineering technology research. Traditional tunnel deformation monitoring methods mostly rely on point-to-point measurement technology, such as using a single sensor for data collection. Although these methods can provide certain monitoring data, they are often limited by the selection of monitoring points and the influence of environmental factors, making it difficult to achieve global dynamic monitoring of the tunnel. These traditional methods often use simple statistical methods for data processing and analysis, such as average calculation or maximum and minimum value analysis. These processing methods cannot fully utilize the complex information contained in the large amount of collected data. Therefore, there is often great uncertainty when predicting tunnel deformation and conducting risk assessments.

[0003] The shortcomings of existing technologies in tunnel deformation prediction and assessment are mainly manifested in the following aspects: data collection in existing technologies mostly relies on a single source, making it difficult to achieve multi-dimensional and comprehensive data integration, which limits the comprehensiveness and representativeness of monitoring data; data processing methods are usually relatively simple and fail to effectively integrate and utilize the inherent connections between various types of data, which means that the useful information in the data is not fully mined. Summary of the Invention

[0004] In view of the above problems existing in the existing tunnel deformation prediction method based on multi-source data fusion, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is that data collection in the existing technology mostly relies on a single source, making it difficult to achieve multi-dimensional and comprehensive data integration, which limits the comprehensiveness and representativeness of the monitoring data. The data processing methods are usually relatively simple and fail to effectively integrate and utilize the inherent connections between various types of data, which means that the useful information in the data is not fully mined.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: a tunnel deformation prediction method based on multi-source data fusion, which includes collecting tunnel data for preprocessing and fusing the preprocessed tunnel data; extracting tunnel features and constructing a tunnel deformation prediction model to predict tunnel deformation; evaluating and issuing early warnings on the results of real-time predictions, constructing a visual interface to display tunnel data in real time, and storing, collecting, and analyzing the generated tunnel data.

[0007] As a preferred embodiment of the tunnel deformation prediction method based on multi-source data fusion described in the present invention, the method of collecting tunnel data for preprocessing includes deploying fiber Bragg grating sensors inside the tunnel, setting the data collection frequency to once per second, collecting sensor data including tunnel displacement and strain, using a geological radar system (GPR) to detect underground structures and obtain tunnel soil compressive strength data, deploying meteorological monitoring stations in and around the tunnel area to collect meteorological data in real time, and deploying a data recording system at the construction site to collect excavation length data.

[0008] The collected tunnel data are cleaned, smoothed using median filtering and moving average methods, missing values in the collected tunnel data are filled using linear interpolation methods, the tunnel data are standardized, the tunnel soil compressive strength and sensor data are spatially aligned using interpolation and supplementation methods, and the collected tunnel data are synchronized according to timestamps using timestamp alignment algorithms.

[0009] As a preferred solution of the tunnel deformation prediction method based on multi-source data fusion of the present invention, the fusion of pre-processed tunnel data refers to the use of the Pearson correlation coefficient to calculate the correlation between historical sensor and meteorological data and tunnel deformation. i , the formula is:

[0010]

[0011] where x i is the observation value of the i-th historical sensor and meteorological data, y i is the observed value of the i-th historical tunnel deformation, and is the average of historical sensor and meteorological data and historical tunnel deformation;

[0012] Calculate the weighted fusion weight as:

[0013]

[0014] where ω i is the weight of sensor and meteorological data, r i is the Pearson correlation coefficient between sensor and meteorological data and tunnel deformation, r j is the sum of the correlation coefficients of the data types, and n is the total number of sensors and meteorological data;

[0015] Perform weighted linear fusion on sensor and meteorological data, the formula is:

[0016]

[0017] Among them D linear is the linear fusion of sensor and meteorological data, Di is the i-th sensor and meteorological data;

[0018] The Gaussian kernel function is used to perform kernel PCA on the tunnel soil compressive strength and excavation length data. The formula is:

[0019]

[0020] Where k(u,v) is the kernel function value between data points u and v, and σ is the broadband parameter of the kernel function;

[0021] Select the best σ value through cross-validation;

[0022] Calculate the kernel function values between all data point pairs to form the kernel matrix K;

[0023] The kernel matrix K is centralized, and the formula is:

[0024] K ′ =K-1 o K-K1 o +1 o K1 o ,

[0025] where K ′ is the centralized kernel matrix, 1 o is an o×o matrix, 1 o K is the column mean matrix, K1 o is the row mean matrix, 1 o K1 o is the adjustment item;

[0026] The centralized kernel matrix K ′ Perform eigendecomposition and select the top three eigenvectors as the main components based on the cumulative contribution rate;

[0027] The original data are mapped into a low-dimensional space using the selected principal components to obtain the tunnel soil compressive strength and excavation length data D after nonlinear fusion. kernel ;

[0028] The sensor and meteorological data after linear fusion are combined with the tunnel soil compressive strength and excavation length data after nonlinear fusion to form comprehensive tunnel data. The formula is:

[0029] D fusion =α·D linear +(1-α)·D kernel ,

[0030] Among them D fusion is the comprehensive tunnel data, α is the adjustment parameter;

[0031] Collect the historical linearly fused sensor and meteorological data and the nonlinearly fused tunnel soil compressive strength and excavation length data, sort them by data size, and assign corresponding ranks. If the data are the same, assign the average rank.

[0032] Calculate linear fusion data D linear and nonlinear fusion data D kernel The Spearman rank correlation coefficient r s , if r s When the value is close to +1, it is a strong correlation, so increase the adjustment parameter α. s When the value is close to 0, it is a weak correlation, so the adjustment parameter α is reduced. s =0.5, the adjustment parameter α remains unchanged;

[0033] Initially, the adjustment parameter α is set to 0.5. The historical linear fusion of sensor and meteorological data and the nonlinear fusion of tunnel soil compressive strength and excavation length data are input into the formula. The adjustment parameter α is adjusted according to the actual application feedback. It is adjusted iteratively until the optimal adjustment parameter α is found. The linear fusion data D linear and nonlinear fusion data D kernel Substitute into the formula to get the comprehensive tunnel data D fusion .

[0034] As a preferred solution of the tunnel deformation prediction method based on multi-source data fusion of the present invention, wherein: extracting tunnel features refers to collecting comprehensive historical tunnel data for preprocessing and standardization to generate an autoencoder training set;

[0035] Construct an autoencoder network, including input layer, hidden layer and output layer;

[0036] Set the input layer to comprehensive tunnel data;

[0037] Use the autoencoder training set to train the autoencoder network model, and use the loss function and Adam optimizer to iteratively optimize the model parameters;

[0038] The integrated tunnel data D fusion Input it into the trained autoencoder network model to obtain the tunnel feature F.

[0039] As a preferred embodiment of the tunnel deformation prediction method based on multi-source data fusion of the present invention, the construction of a tunnel deformation prediction model for tunnel deformation prediction refers to simulating the stress and strain state of the tunnel using the finite element method (FEM), constructing a simulation model based on actual tunnel soil compressive strength and excavation length data, and running the simulation model to obtain deformation data S of the tunnel under different engineering conditions;

[0040] Collect historical tunnel data for preprocessing and standardization, and use the trained autoencoder network model to obtain historical tunnel features F ′ , generate model training set;

[0041] Build a long short-term memory network LSTM model, including input layer, LSTM layer and output layer;

[0042] Set the input layer to the historical tunnel feature F ′ ,Initially set up two layers of LSTM, with 128 units in each layer;

[0043] Use the model training set to train the LSTM model, and use the loss function and Adam optimizer to iteratively optimize the model parameters;

[0044] Input the tunnel feature F into the trained long short-term memory network LSTM model to obtain the tunnel deformation prediction result P ′ ;

[0045] The error E is defined as the difference between the result S of the pseudo model and the prediction result P of the long short-term memory network LSTM model at the same time point in the same tunnel. ′ The difference is:

[0046] E=|SP ′ |;

[0047] Re-adjust the prediction results P of the long short-term memory network LSTM model ′ The adjusted prediction result P is obtained as follows:

[0048]

[0049] Where τ is the standard deviation of the error E, and γ is the weighted adjustment factor;

[0050] Take P as the tunnel deformation prediction result.

[0051] As a preferred embodiment of the tunnel deformation prediction method based on multi-source data fusion of the present invention, the evaluation and early warning of the real-time prediction results refers to setting the risk level threshold R(t) as low risk, medium risk, high risk, and extremely high risk by statistically analyzing the relationship between tunnel deformation data and known risk events based on historical tunnel data and engineering standards;

[0052] The risk level is assessed using the logistic function, as follows:

[0053]

[0054] Where R(t) is the risk level at time t, β and θ are adjustment parameters;

[0055] If R(t) < 0.25, the risk is low, the tunnel deformation is within the minimum acceptable range, and there is no sign of damage in the near future;

[0056] If 0.25≤R(t)<0.5, it is judged as medium risk and a green warning is issued. The tunnel deformation exceeds the normal operation range, but is not enough to cause immediate structural problems;

[0057] If 0.5≤R(t)<0.75, it is judged as high risk and a yellow warning is issued, indicating that the tunnel is severely deformed and requires immediate inspection and intervention;

[0058] If R(t)≥0.75, it is judged as extremely high risk and a red alert is issued. The tunnel deformation has reached a dangerous level and it is necessary to leave immediately.

[0059] As a preferred solution of the tunnel deformation prediction method based on multi-source data fusion described in the present invention, the construction of a visual interface to display tunnel data in real time refers to using the visualization software Tableau to construct a visual interface, including a real-time data flow chart to display the current deformation status and risk level of the tunnel, making a trend chart of historical data to display the change of tunnel deformation over time and designing an alarm indication interface, setting a prediction threshold, and displaying and warning the user when the prediction result P of the long short-term memory network LSTM model exceeds the prediction threshold.

[0060] As a preferred solution of the tunnel deformation prediction method based on multi-source data fusion described in the present invention, the storage of tunnel data generated by collection and analysis refers to storing the tunnel data generated by tunnel collection and analysis in a database, implementing data security protection measures, and performing cloud backup regularly.

[0061] A computer device comprises: a memory and a processor; the memory stores a computer program, and the processor implements the steps of the above-mentioned tunnel deformation prediction method based on multi-source data fusion when executing the computer program.

[0062] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the tunnel deformation prediction method based on multi-source data fusion.

[0063] The beneficial effects of the present invention are as follows: the present invention collects tunnel data for fusion, extracts tunnel features, and constructs a tunnel deformation prediction model to predict tunnel deformation. The fusion method optimizes the data characterization capability, and can not only capture the time series features in historical data, but also perform dynamic predictions based on current real-time data, significantly improving the early warning capability of tunnel deformation, providing strong technical support for the safe operation and management of tunnels, and reducing safety hazards caused by data errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0065] Figure 1 Schematic diagram of the flow of tunnel deformation prediction method based on multi-source data fusion.

[0066] Figure 2 Schematic diagram of the structure of the tunnel deformation prediction method based on multi-source data fusion. DETAILED DESCRIPTION

[0067] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0068] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0069] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it constitute a separate or selective embodiment that is mutually exclusive with other embodiments.

[0070] Example 1

[0071] Reference Figure 1 and Figure 2 This is the first embodiment of the present invention, which provides a tunnel deformation prediction method based on multi-source data fusion. The tunnel deformation prediction method based on multi-source data fusion includes: S1, collecting tunnel data for preprocessing, and fusing the preprocessed tunnel data;

[0072] Specifically, collecting tunnel data for preprocessing involves deploying fiber grating sensors inside the tunnel, setting the data collection frequency to once per second, collecting sensor data, including tunnel displacement and strain, using a geological radar system (GPR) to detect underground structures and obtain tunnel soil compressive strength data, deploying meteorological monitoring stations in and around the tunnel area to collect meteorological data in real time, and deploying a data recording system at the construction site to collect excavation length data.

[0073] The collected tunnel data are cleaned, smoothed using median filtering and moving average methods, missing values in the collected tunnel data are filled using linear interpolation methods, the tunnel data are standardized, the tunnel soil compressive strength and sensor data are spatially aligned using interpolation and supplementation methods, and the collected tunnel data are synchronized according to timestamps using timestamp alignment algorithms.

[0074] By setting the data collection frequency to once per second, it is possible to ensure that minute changes in the tunnel structure are captured in real time, improving the sensitivity and timeliness of monitoring, which is of great significance for the rapid detection and response to potential tunnel deformation risks. By using the median filtering and moving average methods, the noise and outliers in the data can be effectively removed, the data curve can be smoothed, and the authenticity and accuracy of the data can be ensured. The linear interpolation method is used to fill the missing values in the collected data to ensure the integrity of the data, which is crucial for subsequent data analysis and model training. Standardization of tunnel data can eliminate the influence of different dimensions and enhance the comparability of data. The timestamp alignment algorithm is used to synchronize data from different data sources to ensure the consistency of the data in time and improve the effect of multi-source data fusion.

[0075] Furthermore, the pre-processed tunnel data is fused by using the Pearson correlation coefficient to calculate the correlation between the historical sensor and meteorological data and the tunnel deformation. i , the formula is:

[0076]

[0077] where x i is the observation value of the i-th historical sensor and meteorological data, y i is the observed value of the i-th historical tunnel deformation, and is the average of historical sensor and meteorological data and historical tunnel deformation;

[0078] Calculate the weighted fusion weight as:

[0079]

[0080] where ω i is the weight of sensor and meteorological data, r i is the Pearson correlation coefficient between sensor and meteorological data and tunnel deformation, r j is the sum of the correlation coefficients of the data types, and n is the total number of sensors and meteorological data;

[0081] Perform weighted linear fusion on sensor and meteorological data, the formula is:

[0082]

[0083] Among them D linear is the linear fusion of sensor and meteorological data, D i is the i-th sensor and meteorological data;

[0084] The Gaussian kernel function is used to perform kernel PCA on the tunnel soil compressive strength and excavation length data. The formula is:

[0085]

[0086] Where k(u,v) is the kernel function value between data points u and v, and σ is the broadband parameter of the kernel function;

[0087] Select the best σ value through cross-validation;

[0088] Calculate the kernel function values between all data point pairs to form the kernel matrix K;

[0089] The kernel matrix K is centralized, and the formula is:

[0090] K ′ =K-1 o K-K1 o +1 o K1 o ,

[0091] where K ′ is the centralized kernel matrix, 1 o is an o×o matrix, 1 o K is the column mean matrix, K1 o is the row mean matrix, 1 o K1 o is the adjustment item;

[0092] The centralized kernel matrix K ′ Perform eigendecomposition and select the top three eigenvectors as the main components based on the cumulative contribution rate;

[0093] The original data are mapped into a low-dimensional space using the selected principal components to obtain the tunnel soil compressive strength and excavation length data D after nonlinear fusion. kernel ;

[0094] The sensor and meteorological data after linear fusion are combined with the tunnel soil compressive strength and excavation length data after nonlinear fusion to form comprehensive tunnel data. The formula is:

[0095] D fusion =α·D linear +(1-α)·D kernel ,

[0096] Among them D fusionis the comprehensive tunnel data, α is the adjustment parameter;

[0097] Collect the historical linearly fused sensor and meteorological data and the nonlinearly fused tunnel soil compressive strength and excavation length data, sort them by data size, and assign corresponding ranks. If the data are the same, assign the average rank.

[0098] Calculate linear fusion data D linear and nonlinear fusion data D kernel The Spearman rank correlation coefficient r s , if r s When the value is close to +1, it is a strong correlation, so increase the adjustment parameter α. s When the value is close to 0, it is a weak correlation, so the adjustment parameter α is reduced. s =0.5, the adjustment parameter α remains unchanged;

[0099] Initially, the adjustment parameter α is set to 0.5. The historical linear fusion of sensor and meteorological data and the nonlinear fusion of tunnel soil compressive strength and excavation length data are input into the formula. The adjustment parameter α is adjusted according to the actual application feedback. It is adjusted iteratively until the optimal adjustment parameter α is found. The linear fusion data D linear and nonlinear fusion data D kernel Substitute into the formula to get the comprehensive tunnel data D fusion .

[0100] By using the Pearson correlation coefficient to calculate the correlation between sensors, meteorological data and tunnel deformation, the data source with the greatest impact on tunnel deformation can be accurately identified, ensuring the scientific nature and effectiveness of data fusion, making subsequent prediction models more accurate and reliable. The calculated weighted fusion weights are assigned different weights according to the correlation of each data. Such a dynamic weight adjustment mechanism enables the model to be flexibly adjusted when dealing with tunnel deformation under different environments and conditions, thereby improving the adaptability and accuracy of the prediction. Through linear fusion and nonlinear fusion of kernel PCA, linear and nonlinear data features can be comprehensively utilized to optimize the expression of data and enhance the model's ability to capture complex data structures. This fusion technology is particularly suitable for geological and construction data, which often contain complex Spatial and temporal relationships can be better understood and predicted through advanced fusion methods. In kernel PCA processing, the main components selected can represent the core features of the original data. The data mapped to low-dimensional space not only reduces the computational complexity, but also retains the most critical information, providing the possibility for high-accuracy prediction. Combining linear and nonlinear fusion results, the comprehensive tunnel data formed can fully reflect the actual deformation of the tunnel. By sorting and rank correlation analysis of these data, the present invention further optimizes the data processing flow, making the final tunnel deformation prediction result more accurate, calculating the Spearman rank correlation coefficient, and adjusting the model parameters according to the strength of the data correlation, so that the prediction model can be dynamically adjusted according to the actual performance of the data, thereby improving the flexibility and response speed of the model.

[0101] S2. Extract tunnel features and construct a tunnel deformation prediction model to predict tunnel deformation;

[0102] Specifically, extracting tunnel features refers to collecting comprehensive historical tunnel data, preprocessing and standardizing it, and generating an autoencoder training set;

[0103] Construct an autoencoder network, including input layer, hidden layer and output layer;

[0104] Set the input layer to the integrated tunnel data, and the nodes of the output layer to the integrated tunnel data D fusion The number of features is equal;

[0105] Use the autoencoder training set to train the autoencoder network model, and use the loss function and Adam optimizer to iteratively optimize the model parameters;

[0106] The integrated tunnel data D fusion Input it into the trained autoencoder network model to obtain the tunnel feature F.

[0107] By preprocessing and standardizing the collected comprehensive historical tunnel data, the data is ensured to have a consistent format and scale before being input into the autoencoder. Standardization ensures that data of different magnitudes or units can be analyzed at the same scale, enhancing the reliability of data processing and the effectiveness of subsequent learning models. The construction of the training set is based on actual historical data, ensuring the practical application value of the learning process and the representativeness of the data. The equal number of nodes in the input and output layers ensures that the model can attempt to restore the input data during the learning process, allowing the model to learn the most critical features in the data, which is the key to extracting the inherent structure of the data. Deep feature extraction methods are superior to traditional statistical methods because they can reveal nonlinearities and complex patterns in the data, providing more in-depth and comprehensive insights for tunnel deformation prediction.

[0108] Furthermore, constructing a tunnel deformation prediction model to predict tunnel deformation refers to using the finite element method (FEM) to simulate the stress and strain state of the tunnel, building a simulation model based on the actual tunnel soil compressive strength and excavation length data, running the simulation model, and obtaining the deformation data S of the tunnel under different engineering conditions;

[0109] Collect historical tunnel data for preprocessing and standardization, and use the trained autoencoder network model to obtain historical tunnel features F ′ , generate model training set;

[0110] Build a long short-term memory network LSTM model, including input layer, LSTM layer and output layer;

[0111] Set the input layer to the historical tunnel feature F ′ ,Initially set up two layers of LSTM, with 128 units in each layer;

[0112] Use the model training set to train the LSTM model, and use the loss function and Adam optimizer to iteratively optimize the model parameters;

[0113] Input the tunnel feature F into the trained long short-term memory network LSTM model to obtain the tunnel deformation prediction result P ′ ;

[0114] The error E is defined as the difference between the result S of the pseudo model and the prediction result P of the long short-term memory network LSTM model at the same time point in the same tunnel. ′ The difference is:

[0115] E=|SP ′ |;

[0116] Re-adjust the prediction results P of the long short-term memory network LSTM model ′ The adjusted prediction result P is obtained as follows:

[0117]

[0118] Where τ is the standard deviation of the error E, and γ is the weighted adjustment factor;

[0119] Take P as the tunnel deformation prediction result.

[0120] Detailed simulation models based on actual geological and construction data allow engineers to predict the structural stress response of tunnels before actual construction, thereby optimizing the design and preventing potential structural problems. By running these simulation models, deformation data of the tunnel under different engineering conditions can be obtained, providing scientific decision-making support for actual construction. The LSTM model can fully utilize the deep information contained in historical data. The autoencoder plays a role in data noise reduction and feature compression, making the data input to the LSTM network more refined and having a higher information density. By defining the error E to evaluate the difference between the finite element simulation result S and the LSTM prediction result P, this method not only verifies the accuracy of the prediction model, but also uses real-time data to correct the model and adjust the LSTM model using weighted adjustment factors, further enhancing the model's adaptability to new data. This dynamic adjustment mechanism ensures that the model can provide reliable prediction results throughout the life cycle of the tunnel project, significantly improving the safety and efficiency of tunnel projects.

[0121] S3. Evaluate and issue early warnings based on the real-time prediction results, build a visual interface to display tunnel data in real time, and store, collect, and analyze the generated tunnel data;

[0122] Specifically, the evaluation and early warning of the real-time prediction results refers to setting the risk level threshold R(t) as low risk, medium risk, high risk and extremely high risk based on historical tunnel data and engineering standards through statistical analysis of the relationship between tunnel deformation data and known risk events;

[0123] The risk level is assessed using the logistic function, as follows:

[0124]

[0125] Where R(t) is the risk level at time t, β and θ are adjustment parameters, which are obtained based on historical data statistics and control the slope and midpoint of the function respectively;

[0126] If R(t) < 0.25, the risk is low, the tunnel deformation is within the minimum acceptable range, and there is no sign of damage in the near future;

[0127] If 0.25≤R(t)<0.5, it is judged as medium risk and a green warning is issued. The tunnel deformation exceeds the normal operation range, but is not enough to cause immediate structural problems;

[0128] If 0.5≤R(t)<0.75, it is judged as high risk and a yellow warning is issued, indicating that the tunnel is severely deformed and requires immediate inspection and intervention;

[0129] If R(t)≥0.75, it is judged as extremely high risk and a red alert is issued. The tunnel deformation has reached a dangerous level and it is necessary to leave immediately.

[0130] By setting a risk level threshold R(t) based on historical tunnel data and engineering standards, the present invention can identify potential risk levels in advance. This step provides a scientific assessment basis for tunnel safety management, making risk management more systematic and quantitative. By statistically analyzing the relationship between tunnel deformation data and known risk events, this method not only enhances the practical application value of the prediction model but also improves the accuracy of its early warning. The use of a logical function to assess risk levels is an innovation of the present invention. This method can smoothly handle risk level transitions and avoid misjudgments caused by sudden changes. The logical function can reasonably classify risk levels into low, medium, high, and very high levels based on the actual distribution of risk data, making risk assessment more consistent with the risk characteristics of actual deformation conditions. For different risk levels, the present invention defines clear early warning signals, such as green, yellow, and red warnings. These warnings can promptly notify relevant personnel and systems to take appropriate measures. Low risk levels maintain normal monitoring, medium risk levels require enhanced monitoring, and high and very high risk levels require emergency measures, such as reinforcement or evacuation, to ensure the safety of personnel and equipment.

[0131] Furthermore, building a visualization interface to display tunnel data in real time refers to using the visualization software Tableau to build a visualization interface, including a real-time data flow chart to display the current deformation status and risk level of the tunnel, making a trend chart of historical data to display the change of tunnel deformation over time and designing an alarm indication interface, setting a prediction threshold, and displaying and warning the user when the prediction result P of the long short-term memory network LSTM model exceeds the prediction threshold.

[0132] By building a real-time visualization interface, the safety status of the tunnel can be fully monitored, including displacement, strain, and environmental parameters. Through real-time data flow charts and historical trend charts, managers can understand the current status and historical changes of the tunnel at any time, providing a reliable basis for tunnel safety management. During the tunnel construction process, the real-time display of construction data and tunnel deformation through the visualization interface can help construction personnel to adjust the construction plan in time, reduce the impact on the tunnel structure, and improve construction safety and efficiency. During the tunnel construction process, the real-time display of construction data and tunnel deformation through the visualization interface can help construction personnel to adjust the construction plan in time, reduce the impact on the tunnel structure, and improve construction safety and efficiency. The visualization interface can not only display monitoring data in real time, but also perform deformation prediction and risk assessment through the LSTM model. Managers can make scientific decisions and emergency plans based on the prediction results and risk levels to ensure the long-term safe operation of the tunnel.

[0133] Furthermore, storing the tunnel data generated by the collection and analysis means storing the tunnel data generated by the tunnel collection and analysis in a database, implementing data security protection measures, and performing cloud backup on a regular basis.

[0134] Storing the collected tunnel data and analysis results in the database can provide high reliability and durability. The database system ensures the security and integrity of data during storage through data redundancy and logging mechanisms, and can perform data recovery even in the event of a system failure. Storing tunnel data in a structured form in the database can facilitate data classification and organization. By implementing data security protection measures, data leakage and illegal access can be effectively prevented. Regular cloud backup can ensure the high availability and elasticity of tunnel data.

[0135] Example 2

[0136] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0137] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0138] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0139] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

Claims

1. A tunnel deformation prediction method based on multi-source data fusion, characterized by: include, Collect tunnel data for preprocessing and fuse the preprocessed tunnel data; Extract tunnel features and build a tunnel deformation prediction model to predict tunnel deformation; Evaluate and issue early warnings based on real-time prediction results, build a visual interface to display tunnel data in real time, and store, collect, and analyze generated tunnel data; The collection of tunnel data for preprocessing includes deploying fiber grating sensors inside the tunnel, setting the data collection frequency to once per second, collecting sensor data, including tunnel displacement and strain, using a geological radar system (GPR) to detect underground structures and obtain tunnel soil compressive strength data, deploying meteorological monitoring stations in and around the tunnel area to collect meteorological data in real time, and deploying a data recording system at the construction site to collect excavation length data; The collected tunnel data was cleaned and smoothed using median filtering and moving average methods. Missing values in the collected tunnel data were filled using linear interpolation. The tunnel data was standardized. Interpolation and infill methods were used to spatially align the tunnel soil compressive strength and sensor data. The collected tunnel data was synchronized according to timestamps using a timestamp alignment algorithm. Calculate the correlation between historical sensor and meteorological data and tunnel deformation using the Pearson correlation coefficient r i , the formula is: where x i is the observation value of the i-th historical sensor and meteorological data, y i is the observed value of the i-th historical tunnel deformation, and is the average of historical sensor and meteorological data and historical tunnel deformation; Calculate the weighted fusion weight as: where ω i is the weight of sensor and meteorological data, r i is the Pearson correlation coefficient between sensor and meteorological data and tunnel deformation, r j is the sum of the correlation coefficients of the data types, and n is the total number of sensors and meteorological data; Perform weighted linear fusion on sensor and meteorological data, the formula is: Among them D linear is the linear fusion of sensor and meteorological data, D i is the i-th sensor and meteorological data; The Gaussian kernel function is used to perform kernel PCA on the tunnel soil compressive strength and excavation length data. The formula is: Where k(u,v) is the kernel function value between data points u and v, and σ is the broadband parameter of the kernel function; Select the best σ value through cross-validation; Calculate the kernel function values between all data point pairs to form the kernel matrix K; The kernel matrix K is centralized, and the formula is: K ′ =K-1 o K-K1 o +1 o K1 o , where K ′ is the centralized kernel matrix, 1 o is an o×o matrix, 1 o K is the column mean matrix, K1 o is the row mean matrix, 1 o K1 o is the adjustment item; The centralized kernel matrix K ′ Perform eigendecomposition and select the top three eigenvectors as the main components based on the cumulative contribution rate; The original data are mapped into a low-dimensional space using the selected principal components to obtain the tunnel soil compressive strength and excavation length data D after nonlinear fusion. kernel ; The sensor and meteorological data after linear fusion are combined with the tunnel soil compressive strength and excavation length data after nonlinear fusion to form comprehensive tunnel data. The formula is: D fusion =α·D linear +(1-a)·D kernel , Among them D fusion is the comprehensive tunnel data, α is the adjustment parameter; Extracting tunnel features refers to collecting comprehensive historical tunnel data, performing preprocessing and standardization, and generating an autoencoder training set; Construct an autoencoder network, including input layer, hidden layer and output layer; Set the input layer to comprehensive tunnel data; Use the autoencoder training set to train the autoencoder network model, and use the loss function and Adam optimizer to iteratively optimize the model parameters; The integrated tunnel data D fusion Input into the trained autoencoder network model to obtain the tunnel feature F; Constructing a tunnel deformation prediction model to predict tunnel deformation refers to using a finite element method (FEM) to simulate the stress and strain state of the tunnel, constructing a simulation model based on actual tunnel soil compressive strength and excavation length data, and running the simulation model to obtain deformation data S of the tunnel under different engineering conditions; Collect historical tunnel data for preprocessing and standardization, and use the trained autoencoder network model to obtain historical tunnel features F ′ , generate model training set; Build a long short-term memory network LSTM model, including input layer, LSTM layer and output layer; Set the input layer to the historical tunnel feature F ′ ,Initially set up two layers of LSTM, with 128 units in each layer; Use the model training set to train the LSTM model, and use the loss function and Adam optimizer to iteratively optimize the model parameters; Input the tunnel feature F into the trained long short-term memory network LSTM model to obtain the tunnel deformation prediction result P ′ ; The error E is defined as the difference between the result S of the pseudo model and the prediction result P of the long short-term memory network LSTM model at the same time point in the same tunnel. ′ The difference is: E=|S-P ′ |; Re-adjust the prediction results P of the long short-term memory network LSTM model ′ The adjusted prediction result P is obtained as follows: Where τ is the standard deviation of the error E, and γ is the weighted adjustment factor; Take P as the tunnel deformation prediction result.

2. The tunnel deformation prediction method based on multi-source data fusion according to claim 1, characterized in that: The fusion of the pre-processed tunnel data refers to collecting the sensor and meteorological data after historical linear fusion and the tunnel soil compressive strength and excavation length data after nonlinear fusion, sorting them according to the size of the data, and assigning corresponding ranks. If the data are the same, the average rank is assigned; Calculate linear fusion data D linear and nonlinear fusion data D kernel The Spearman rank correlation coefficient r s , if r s When the value is close to +1, it is a strong correlation, so increase the adjustment parameter α. s When the value is close to 0, it is a weak correlation, so the adjustment parameter α is reduced. s =0.5, the adjustment parameter α remains unchanged; Initially, the adjustment parameter α is set to 0.

5. The historical linear fusion of sensor and meteorological data and the nonlinear fusion of tunnel soil compressive strength and excavation length data are input into the formula. The adjustment parameter α is adjusted according to the actual application feedback. It is adjusted iteratively until the optimal adjustment parameter α is found. The linear fusion data D linear and nonlinear fusion data D kernel Substitute into the formula to get the comprehensive tunnel data D fusion .

3. The tunnel deformation prediction method based on multi-source data fusion according to claim 2, characterized in that: The evaluation and early warning of the real-time prediction results refers to setting the risk level threshold R(t) as low risk, medium risk, high risk and very high risk by statistically analyzing the relationship between tunnel deformation data and known risk events based on historical tunnel data and engineering standards; The risk level is assessed using the logistic function, as follows: Where R(t) is the risk level at time t, β and θ are adjustment parameters; If R(t) < 0.25, the risk is low, the tunnel deformation is within the minimum acceptable range, and there is no sign of damage in the near future; If 0.25≤R(t)<0.5, it is judged as medium risk and a green warning is issued. The tunnel deformation exceeds the normal operation range, but is not enough to cause immediate structural problems; If 0.5≤R(t)<0.75, it is judged as high risk and a yellow warning is issued, indicating that the tunnel is severely deformed and requires immediate inspection and intervention; If R(t)≥0.75, it is judged as extremely high risk and a red alert is issued. The tunnel deformation has reached a dangerous level and it is necessary to leave immediately.

4. The tunnel deformation prediction method based on multi-source data fusion according to claim 3, characterized in that: The said construction of a visualization interface to display tunnel data in real time refers to using the visualization software Tableau to construct a visualization interface, including a real-time data flow chart to display the current deformation status and risk level of the tunnel, making a trend chart of historical data to display the change of tunnel deformation over time and designing an alarm indication interface, setting a prediction threshold, and displaying and warning the user when the prediction result P of the long short-term memory network LSTM model exceeds the prediction threshold.

5. The tunnel deformation prediction method based on multi-source data fusion according to claim 4, characterized in that: The storing of the tunnel data generated by the collection and analysis refers to storing the tunnel data generated by the collection and analysis of the tunnel in a database, implementing data security protection measures, and performing cloud backup on a regular basis.

6. A computer device comprising: memory and processor; The memory stores a computer program, characterized in that when the processor executes the computer program, the steps of the tunnel deformation prediction method based on multi-source data fusion according to any one of claims 1 to 5 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the tunnel deformation prediction method based on multi-source data fusion according to any one of claims 1 to 5 are implemented.

Images