A digital prediction and early warning method for geological disasters of a tunnel in a high-risk karst area

Abstract

The application provides a high-risk karst area tunnel geological disaster digital prediction and early warning method, comprising: obtaining geological data of a tunnel area; determining surrounding rock conditions of each section included in the tunnel area according to the geological data through geological analysis, and performing risk grade division; matching a corresponding prediction mode according to the risk grade division result, and performing geological prediction; obtaining face excavation feedback information of each section of the tunnel area; comparing the geological prediction of the corresponding section with the face excavation feedback information, marking the corresponding section as a prediction error section when it is determined that the two are inconsistent, and issuing an alarm prompt; the application greatly improves the safety of high-risk karst area tunnel geological tunnel construction through high-risk karst area tunnel geological disaster digital prediction and early warning, and ensures that the tunnel construction project proceeds smoothly.

Description

Technical Field

[0001] This invention relates to the field of tunnel construction, and in particular to a digital forecasting and early warning method for geological disasters in tunnels in high-risk karst areas. Background Technology

[0002] Currently, during tunnel construction in my country, various targeted forecasts are made before construction based on the specific conditions of karst areas. Comprehensive advanced geological forecasting is particularly important during construction. Comprehensive advanced geological forecasting is a method based on multiple data sources and technologies to provide early warnings and forecasts of geological disasters, aiming to reduce the harm to human life and property. However, in the current process of comprehensive advanced geological forecasting, when there are errors between the geological forecast information and the feedback information from the tunnel face, timely warnings cannot be generated, which greatly affects the construction process and makes it impossible to guarantee continued comprehensive advanced geological forecasting for later construction. This also prevents the prediction and warning of impending construction risks, significantly increasing the dangers during tunnel construction. Therefore, a digital forecasting and early warning method for geological disasters in tunnels in high-risk karst areas is urgently needed. Summary of the Invention

[0003] This invention aims to at least partially solve one of the technical problems in the aforementioned technologies. Therefore, the purpose of this invention is to propose a digital forecasting and early warning method for geological hazards in tunnels in high-risk karst areas. This method addresses the need for real-time comparison of geological forecast information with tunnel face feedback information during comprehensive advanced geological forecasting. When discrepancies exist between the geological forecast information and the tunnel face feedback information, timely early warnings are generated, ensuring the smooth progress of tunnel construction.

[0004] To achieve the above objectives, this invention proposes a digital forecasting and early warning method for tunnel geological hazards in high-risk karst areas, comprising: Obtain geological data for the tunnel area; Based on geological data and through geological analysis, the surrounding rock conditions of each section within the tunnel area are determined, and the risk levels are classified. Based on the risk level classification results, the corresponding forecasting method is matched, and geological forecasting is performed; Obtain excavation feedback information from the tunnel face of each section; The geological forecasts for the corresponding sections are compared with the feedback information from the excavation of the tunnel face. If the two are found to be inconsistent, the corresponding section is marked as a forecast error section and an alarm is issued.

[0005] Preferably, the geological data includes geological structure data, stratigraphic lithology data, groundwater data, and environmental condition data.

[0006] Preferably, based on geological data and through geological analysis, the surrounding rock conditions of each section within the tunnel area are determined, and a risk level classification is performed, including: Lithological parameters of the determined section When the value is greater than 0.55, the corresponding section is determined to be at the first risk level; The lithological index of the determined section is 0.15 < When the value is less than 0.55, the corresponding section is classified as the second risk level. Lithological parameters of the determined section When the value is less than 0.15, the corresponding section is determined to be at the third risk level.

[0007] Preferably, the forecasting method is matched according to the risk level classification result, including: For sections of the first risk level, a forecasting method using TSP (Transient Spray) for 100-150m is adopted; For sections at the second risk level, a forecast method using TSP (Tracking Point Prediction) with a forecast duration of 100m is adopted; For sections with the third risk level, short-term forecasts are made using ground-penetrating radar (0-20m) or transient electromagnetic forecasts (30-50m), and further confirmation is achieved by combining these forecasts with advance drilling or lateral drilling at the tunnel face.

[0008] Preferably, before determining the surrounding rock conditions of each section within the tunnel area through geological analysis based on geological data, the process further includes: Divide the geological data into several geological text data; Each geological text data is segmented into words to obtain several keywords corresponding to each geological text data. Take any geological text data, calculate the degree of belonging of several keywords in the geological text data to keywords in the preset keyword classification label table, and determine the label in the preset keyword classification table corresponding to the maximum degree of belonging as the pre-selected category label of the geological text data; Obtain the first text vector of the geological text data and the second text vector corresponding to the pre-selected category labels; the second text vector is the average of the text vectors of all keywords in the pre-selected category labels; Calculate the similarity between the first text vector and the second text vector; When the similarity is greater than a preset similarity threshold, the pre-selected category label will be used as the category label corresponding to the text data; Based on the above methods, the category label corresponding to each text data in the geological data is obtained; data with the same category label are merged to obtain geological data of different categories; Select any category of geological data for data cleaning to obtain the first data to be cleaned; obtain a first preset number of data from the first data to be cleaned to obtain the first sub-data to be cleaned; The first sub-data to be cleaned is cleaned using several pre-set data cleaning schemes; the completion rate of each data cleaning scheme for the first sub-data to be cleaned is calculated. The data cleaning scheme with the highest completion rate for cleaning the first sub-data to be cleaned is selected as the first data cleaning scheme, and the cleaning completion rate of the first data cleaning scheme is determined as the first completion rate. Obtain a second preset number of data points from the first set of data to be cleaned to obtain the second set of sub-data points to be cleaned; wherein the second preset number is greater than the first preset number. Based on the first data cleaning scheme, the second sub-data to be cleaned is cleaned, and the cleaning completion rate of the first data cleaning scheme on the second sub-data to be cleaned is calculated to obtain the second completion rate. Calculate the absolute value of the difference between the first completion level and the second completion level. When the absolute value is less than or equal to a preset difference threshold, determine the first data cleaning scheme as the target data cleaning scheme corresponding to the first data to be cleaned. Complete the data cleaning of the first data to be cleaned based on the first data cleaning scheme. Based on the above methods, the geological data cleaning is completed, and the cleaned geological data is obtained.

[0009] Preferably, obtaining the excavation feedback information of each section of the tunnel face includes: Obtain the tunnel face images of each section of the tunnel area; Image enhancement was performed on the tunnel face images of each section of the tunnel area; The enhanced tunnel face images of each section are input into a pre-trained neural network model for rock strata feature extraction and identification. The identification results are analyzed to obtain the excavation feedback information of each section of the tunnel face.

[0010] Preferably, the image enhancement of the tunnel face images of each section of the tunnel area includes: The tunnel face images of each section are segmented into edge images and non-edge images; The edge image is converted to grayscale to obtain the first grayscale image; Obtain the grayscale value of each pixel in the first grayscale image; randomly select a pixel in the first grayscale image to obtain the first pixel; determine an n×n sliding window centered on the first pixel; calculate the absolute value of the difference between the grayscale value of the first pixel and the grayscale values ​​of each pixel in the first preset direction in the sliding window, and determine the largest absolute value as the first absolute value; Calculate the absolute value of the difference in grayscale value between the first pixel in the sliding window and each pixel in the second preset direction, and determine the largest absolute value as the second absolute value; wherein the first preset direction and the second preset direction are opposite directions; Calculate the absolute value of the difference in grayscale value between the first pixel in the sliding window and each pixel in the third preset direction, and determine the largest absolute value as the third absolute value; wherein, the third preset direction is perpendicular to the first preset direction and the second preset direction; Calculate the absolute value of the difference in grayscale value between the first pixel in the sliding window and each pixel in the fourth preset direction, and determine the largest absolute value as the fourth absolute value; wherein the third preset direction and the fourth preset direction are opposite directions; Compare the first absolute value with the second absolute value, determine the gray value of each pixel and the total number of pixels in the preset direction corresponding to the largest absolute value, and obtain the first set of pixels. Compare the third absolute value with the fourth absolute value, determine the gray value of each pixel and the total number of pixels in the preset direction corresponding to the largest absolute value, and obtain the second set of pixels. Calculate the weighted average of the gray values ​​of all pixels in the first pixel set and the second pixel set, and replace the gray value of the first pixel with the weighted average. Iterate through all pixels in the first grayscale image to obtain the edge-enhanced image; The non-edge image is processed to obtain a second grayscale image; Obtain the grayscale value of each pixel in the second grayscale image, and enhance the brightness of the second grayscale image to obtain the grayscale enhanced image; The edge enhancement image and the grayscale enhancement image are overlaid and merged to enhance the face image of each section of the tunnel region.

[0011] Preferably, after marking the corresponding segment as a prediction error segment and issuing an alarm, the method further includes: Based on the feedback information from the excavation of the tunnel face corresponding to the incorrect forecast section, a new forecasting method is selected to conduct a new geological forecast. Then, the construction plan is adjusted according to the new geological forecast to carry out tunnel construction.

[0012] Preferably, after determining the cleaned geological data for the final tunnel area, the method further includes: Record the cleaning steps, processing methods, and data changes of geological data in the tunnel area during the data cleaning process to obtain a data cleaning document.

[0013] Preferably, comparing the geological forecast of the corresponding section with the excavation feedback information of the tunnel face includes: Based on the geological forecast of the corresponding section, predict the lithology, rock mass structure and geological structure of the working face; Based on the excavation feedback information of the corresponding section of the working face, we can obtain the lithology, rock mass structure and geological structure of the corresponding section of the working face; The lithology, rock mass structure, and geological structure of the corresponding section's face predicted by geological forecasts are compared with the lithology, rock mass structure, and geological structure of the corresponding section's face obtained from the excavation feedback information.

[0014] The beneficial effects of this invention are as follows: By acquiring geological data of the tunnel area and conducting geological analysis, a deeper understanding of the surrounding rock conditions in each section can be obtained, including lithology, deformation, and groundwater level, which helps predict potential geological problems encountered during tunnel face excavation; based on the geological data and analysis results, the surrounding rock conditions in each section of the tunnel area can be classified into risk levels, identifying high, medium, and low-risk sections, allowing for targeted development of construction plans and safety measures; and based on the risk level classification results, suitable forecasting methods can be selected for geological forecasting, such as ground-penetrating radar and artificial observation. Through methods such as surveying and measurement, potential geological problems can be identified and warned of in a timely manner during construction. By installing monitoring equipment and sensors, feedback information on the excavation face of each section of the tunnel area can be obtained, including indicators such as deformation, displacement, and stress, to monitor and grasp geological changes in real time during construction. By comparing the geological forecasts of corresponding sections with the excavation face feedback information, inconsistencies between the forecasts and the actual situation can be identified in a timely manner, marked as incorrect forecast sections, and alarm prompts can be issued. Construction plans and safety measures can be adjusted in a timely manner to avoid accidents. Through the above processes and methods, geological risks in tunnel construction can be effectively managed and addressed, improving the safety and efficiency of the project.

[0015] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0016] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a digital forecasting and early warning method for tunnel geological hazards in high-risk karst areas according to an embodiment of the present invention; Figure 2This is a flowchart of obtaining excavation feedback information of each section of the tunnel face according to an embodiment of the present invention; Figure 3 This is a flowchart illustrating the process after marking the corresponding segment as a predicted error segment and issuing an alarm notification according to an embodiment of the present invention. Detailed Implementation

[0018] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0019] Example 1 like Figure 1 As shown, this invention provides a digital forecasting and early warning method for tunnel geological hazards in high-risk karst areas, comprising: Obtain geological data for the tunnel area; Based on geological data and through geological analysis, the surrounding rock conditions of each section within the tunnel area are determined, and the risk levels are classified. Based on the risk level classification results, the corresponding forecasting method is matched, and geological forecasting is performed; Obtain excavation feedback information from the tunnel face of each section; The geological forecasts for the corresponding sections are compared with the feedback information from the excavation of the tunnel face. If the two are found to be inconsistent, the corresponding section is marked as a forecast error section and an alarm is issued.

[0020] In this embodiment, the specific implementation method of geological analysis based on geological data is as follows: by using rock and mineral sample data in the geological data, their composition and characteristics can be determined, thereby inferring geological processes and resource potential; by using data such as seismic exploration, gravity measurement, electromagnetic measurement, and magnetic measurement, information on underground structure and properties can be obtained, thereby inferring crustal structure and resource geological characteristics; by using chemical composition data of samples such as surface, groundwater, and soil, the element cycle and geological processes in the Earth system can be understood; by using observation and measurement data of surface landforms and stratigraphic characteristics, geological history and geological processes, such as faults, folds, and sedimentary records, can be inferred.

[0021] In this embodiment, the risk levels include, but are not limited to, the following classifications. The classification of risk levels should be comprehensively considered based on specific circumstances and adjusted and improved in conjunction with actual construction conditions. Examples are provided in this embodiment: Level A: Areas with significant geological hazard risks, such as large underground river systems, contact zones between soluble and insoluble rocks, weak and fractured zones, etc., which may result in large or extra-large water and mud inrushes, areas prone to major environmental geological disasters, and areas with high ground stress, gas, natural gas, and serious radioactive problems. Level B: Medium and small water and mud inrush areas, areas with significant geophysical anomalies, fault zones, etc. Level C: Carbonate and clastic rock areas with relatively good hydrogeological conditions, small fault fracture zones, etc., with a low probability of water and mud inrushes. Level D: Insoluble rock areas, with an extremely low probability of water and mud inrushes.

[0022] In this embodiment, geological forecasting includes one or more combinations of long-range forecasting, short-range forecasting, advanced drilling, and geological feedback analysis.

[0023] In this embodiment, the specific implementation methods for obtaining the excavation feedback information of each section of the tunnel face include: **Spotting Method:** During the tunnel face excavation process, a series of measuring points and spotting instruments are installed to monitor the deformation and movement of the tunnel. This includes measuring surface settlement, internal tunnel stress, and deformation of the support structure; **Borehole Measurement Method:** Using drilling and measurement equipment, boreholes are drilled on the ground or in the strata penetrating the tunnel. By measuring the displacement and stress changes at the borehole opening, the impact range and deformation of the tunnel excavation are determined; **Groundwater Level Monitoring:** Changes in the groundwater level during the tunnel excavation process are monitored, including water level height and water pressure, and corresponding drainage measures are taken in a timely manner to ensure the safe progress of construction; **Tunnel Construction Monitoring System:** Utilizing modern monitoring and Internet of Things (IoT) technologies, sensors and monitoring equipment installed at the tunnel construction site monitor parameters such as deformation, displacement, and stress of the tunnel face excavation in real time, and perform data collection and analysis.

[0024] The above implementation method has the following advantages: By acquiring geological data of the tunnel area and conducting geological analysis, a deeper understanding of the surrounding rock conditions in each section can be obtained, including lithology, deformation, and groundwater level, which helps predict potential geological problems encountered during tunnel face excavation. Based on the geological data and analysis results, the surrounding rock conditions in each section of the tunnel area can be classified into risk levels, identifying high, medium, and low-risk sections, allowing for targeted development of construction plans and safety measures. Based on the risk level classification, suitable forecasting methods can be selected for geological forecasting, such as ground-penetrating radar and artificial... Observation and measurement methods are used to promptly identify and warn of potential geological problems during construction. By installing monitoring equipment and sensors, feedback information from the tunnel face excavation in each section of the tunnel area is obtained, including indicators such as deformation, displacement, and stress, allowing for real-time monitoring and understanding of geological changes during construction. By comparing the geological forecasts for corresponding sections with the excavation feedback information, discrepancies between the forecasts and actual conditions are promptly identified, marked as erroneous forecast sections, and alarm prompts are issued. This allows for timely adjustments to construction plans and safety measures to prevent accidents. Through the above processes and methods, geological risks in tunnel construction can be effectively managed and addressed, improving the safety and efficiency of the project.

[0025] Example 2 The geological data includes geological structure data, stratigraphic lithology data, groundwater data, and environmental condition data.

[0026] The beneficial effects of the above scheme are as follows: This data provides a comprehensive understanding of geological structures, including the distribution, thickness, and lithology of strata, which helps in better understanding geological tectonic movements and geological history; it provides in-depth knowledge of groundwater conditions, including groundwater type, water level, and flow rate, which helps in better predicting and planning the management and utilization of groundwater resources; and it provides access to environmental condition data, including climate, soil, and vegetation characteristics, which helps in better assessing and planning land use, ecological protection, and other related work.

[0027] Example 3 Based on geological data and through geological analysis, the surrounding rock conditions of each section within the tunnel area were determined, and risk levels were classified, including: Lithological parameters of the determined section When the value is greater than 0.55, the corresponding section is determined to be at the first risk level; The lithological index of the determined section is 0.15 < When the value is less than 0.55, the corresponding section is classified as the second risk level. Lithological parameters of the determined section When the value is less than 0.15, the corresponding section is determined to be at the third risk level.

[0028] In this embodiment, The indicators reflect a comprehensive set of factors that reflect the mechanical properties of rocks and the degree of fragmentation of rock masses; ; in, The mechanical properties of the rocks and the degree of fragmentation of the rock mass are shown in Table 1. The elastic wave propagation velocity of the rock mass, This represents the elastic wave propagation speed of the rock.

[0029] Table 1 The above embodiments offer the following benefits: Geological analysis and risk level classification accurately reveal the surrounding rock quality of each tunnel section, allowing for the implementation of corresponding support measures based on the risk level. This helps improve tunnel construction safety, reducing construction accidents and delays caused by surrounding rock issues; it provides a reference for tunnel operation; after determining the surrounding rock conditions and risk level, targeted inspections and maintenance can be conducted to promptly identify and address potential surrounding rock problems, ensuring the tunnel's operational safety and continuous stability; it allows for advance understanding of the surrounding rock conditions in each tunnel section, thereby optimizing engineering design and construction plans; and it enables the rational selection of support measures and materials, avoiding unnecessary over-support and thus saving construction costs.

[0030] Example 4 Matching the corresponding forecasting method based on the risk level classification results includes: For sections of the first risk level, a forecasting method using TSP (Transient Spray) for 100-150m is adopted; For sections at the second risk level, a forecast method using TSP (Tracking Point Prediction) with a forecast duration of 100m is adopted; For sections with the third risk level, short-term forecasts are made using ground-penetrating radar (0-20m) or transient electromagnetic forecasts (30-50m), and further confirmation is achieved by combining these forecasts with advance drilling or lateral drilling at the tunnel face.

[0031] In this embodiment, TSP is a method for assessing and predicting tunnel rock mass stability based on on-site geological conditions and parameters such as groundwater.

[0032] In this embodiment, ground-penetrating radar is a tool that uses radar technology to explore and detect underground geological structures.

[0033] In this embodiment, the transient electromagnetic method is a geophysical exploration method that is widely used in mineral resource exploration, groundwater detection, geological engineering and other fields.

[0034] The above technical solutions offer the following benefits: Sections with different risk levels possess varying geological conditions and risk degrees. Selecting a forecasting method based on the risk level classification allows for a more accurate understanding of the geological conditions and potential risks of that section. Choosing a suitable forecasting method improves the accuracy and reliability of the forecast. Different forecasting methods have different principles and applicable scopes; selecting one based on geological conditions and risk level allows for better capture and assessment of potential geological risks. By employing appropriate forecasting methods, potential risks and problems can be warned in advance, enabling the implementation of corresponding preventative and control measures to mitigate potential risks and losses. Through the rational selection of forecasting methods, the construction risks in sections with different risk levels can be reduced to the greatest extent possible.

[0035] Example 5 Before determining the surrounding rock conditions of each section within the tunnel area through geological analysis based on geological data, the following steps are also included: Divide the geological data into several geological text data; Each geological text data is segmented into words to obtain several keywords corresponding to each geological text data. Take any geological text data, calculate the degree of belonging of several keywords in the geological text data to keywords in the preset keyword classification label table, and determine the label in the preset keyword classification table corresponding to the maximum degree of belonging as the pre-selected category label of the geological text data; Obtain the first text vector of the geological text data and the second text vector corresponding to the pre-selected category labels; the second text vector is the average of the text vectors of all keywords in the pre-selected category labels; Calculate the similarity between the first text vector and the second text vector; When the similarity is greater than a preset similarity threshold, the pre-selected category label will be used as the category label corresponding to the text data; Based on the above methods, the category label corresponding to each text data in the geological data is obtained; data with the same category label are merged to obtain geological data of different categories; Select any category of geological data for data cleaning to obtain the first data to be cleaned; obtain a first preset number of data from the first data to be cleaned to obtain the first sub-data to be cleaned; The first sub-data to be cleaned is cleaned using several pre-set data cleaning schemes; the completion rate of each data cleaning scheme for the first sub-data to be cleaned is calculated. The data cleaning scheme with the highest completion rate for cleaning the first sub-data to be cleaned is selected as the first data cleaning scheme, and the cleaning completion rate of the first data cleaning scheme is determined as the first completion rate. Obtain a second preset number of data points from the first set of data to be cleaned to obtain the second set of sub-data points to be cleaned; wherein the second preset number is greater than the first preset number. Based on the first data cleaning scheme, the second sub-data to be cleaned is cleaned, and the cleaning completion rate of the first data cleaning scheme on the second sub-data to be cleaned is calculated to obtain the second completion rate. Calculate the absolute value of the difference between the first completion level and the second completion level. When the absolute value is less than or equal to a preset difference threshold, determine the first data cleaning scheme as the target data cleaning scheme corresponding to the first data to be cleaned. Complete the data cleaning of the first data to be cleaned based on the first data cleaning scheme. Based on the above methods, the geological data cleaning is completed, and the cleaned geological data is obtained.

[0036] In this embodiment, the specific implementation method for dividing geological data into several geological text data is to divide the geological data in a certain way according to actual needs. It can be divided according to time, location, region, etc., and the large segment of geological data is divided into multiple smaller text data.

[0037] In this embodiment, the specific implementation method for segmenting each geological text data can be to use existing word segmentation libraries or develop a word segmentation tool.

[0038] In this embodiment, the specific calculation method for the degree of belonging is as follows: ; in, For the first The text data is compared to the first text in the preset keyword category label table. Class affiliation degree For the first The first text data contained in the text The number of keyword categories; n is the total number of categories in the preset keyword category label table.

[0039] In this embodiment, the specific implementation of obtaining the first text vector of geological text data and the second text vector corresponding to the pre-selected category label uses word vector models such as Word2Vec and GloVe to perform vector representation learning on the word segmentation results, mapping each word in the vocabulary list to a vector; these vectors can capture the semantic relationships and contextual information between words; the vectors of each word in the text are averaged or summed to obtain the vector representation of the text; this representation method transforms the entire text into a high-dimensional vector, which facilitates subsequent text mining and analysis.

[0040] In this embodiment, the specific implementation method for calculating the similarity between the first text vector and the second text vector is to determine it using methods such as Euclidean distance, Manhattan distance, Minkowski distance, and cosine similarity.

[0041] Here is an example of a specific implementation method for calculating similarity using Euclidean distance: ; in, The similarity between the first text vector and the second text vector; and Let i and n represent the i-th elements of the first and second text vectors, respectively; n is the dimension of the vector.

[0042] In this embodiment, the preset similarity threshold is a similarity judgment standard value preset in advance according to the actual application needs.

[0043] In this embodiment, the purpose of data cleaning is to eliminate erroneous, abnormal, and incomplete data, thereby improving data quality. By removing duplicate data, correcting erroneous and missing values ​​in geological data, interference with geological data analysis and model building can be reduced, ensuring the accuracy and consistency of geological data.

[0044] In this embodiment, several data cleaning schemes are targeted data cleaning schemes that are pre-set based on the characteristics of geological data before data cleaning.

[0045] In this embodiment, the specific implementation method for achieving cleaning completion is as follows: Calculate the percentage of valid data in the dataset. Valid data refers to observations that do not contain missing values, outliers, or other invalid data. Completeness = (Number of valid data / Total number of observations) * 100%. The higher the completeness of valid data in the dataset, the higher the quality of the data after cleaning.

[0046] In this embodiment, when the absolute value is less than or equal to a preset difference threshold, it indicates that the first data cleaning scheme has a good data cleaning effect on the first data to be cleaned; when the absolute value is greater than the preset difference threshold, it indicates that the first data cleaning scheme has a very unstable data cleaning effect on the first data to be cleaned, and it is necessary to carefully analyze the quality differences of different numbers of the first sub-data to be cleaned, including missing values, outliers, duplicate values, etc., to determine the main reasons for the quality differences between different numbers of data; based on the analysis results of the quality problems of the first data to be cleaned, the parameters and strategies of the first data cleaning scheme are appropriately adjusted. For example, for missing value processing, the filling strategy can be adjusted according to the proportion of missing values ​​in different numbers of geological data; the sample size of different numbers of the first sub-data to be cleaned is increased to increase the accuracy and representativeness of the cleaning; the final target data cleaning scheme is determined through these methods.The beneficial effects of the above technical solution are as follows: First, the geological data is divided into several geological text data segments; second, each geological text data segment is processed to obtain several keywords corresponding to it; third, pre-selected category labels for the geological text data are selected by calculating the affiliation degree between several keywords in the geological text data and keywords in a preset keyword classification label table; fourth, the pre-selected category labels for the geological text data are selected by calculating the similarity between the first text vector of the geological text data and the second text vector corresponding to the pre-selected category labels. Label verification improves the accuracy of geological data classification; data cleaning of classified data eliminates errors, anomalies, and incomplete data, thereby improving data quality; removing duplicate data and repairing erroneous and missing values ​​reduces interference with geological data analysis and model building, ensuring the accuracy and consistency of geological data; specifically, several pre-set data cleaning schemes are used to clean the first sub-data set to be cleaned, and the completion rate of each data cleaning scheme is calculated; the data cleaning scheme with the highest completion rate is selected as the first... A data cleaning scheme is proposed, and the cleaning completion rate of the first data cleaning scheme is determined as the first completion rate to complete the first cleaning. A second preset number of data points are obtained from the first data to be cleaned to obtain the second sub-data to be cleaned. The second sub-data to be cleaned is then cleaned based on the first data cleaning scheme, and the cleaning completion rate of the first data cleaning scheme on the second sub-data to be cleaned is calculated to obtain the second completion rate, which is a verification of the first data cleaning scheme. When the absolute value of the difference between the first completion rate and the second completion rate is less than or equal to a preset difference threshold, the first data cleaning scheme is determined as the target data cleaning scheme corresponding to the first data to be cleaned. The solution completes the data cleaning of the first set of data to be cleaned. Through data classification and cleaning, geological data can be organized and structured, making it more regular and easier to understand. This helps improve the readability and usability of geological data, preparing it for subsequent analysis and application. It can identify and handle quality issues such as outliers, missing values, and duplicate values ​​in geological data, thereby improving the quality of the geological data. High-quality geological data helps improve the accuracy and reliability of decision-making, avoiding errors and biases caused by geological data quality issues during analysis and application. It can filter out invalid or redundant data and extract key features and variables. This helps reduce the complexity of geological data analysis, reduce the workload, and focus efforts on the analysis and mining of core geological data.

[0047] Example 6 like Figure 2 As shown, obtaining the excavation feedback information of each section of the tunnel face includes: Obtain the tunnel face images of each section of the tunnel area; Image enhancement was performed on the tunnel face images of each section of the tunnel area; The enhanced tunnel face images of each section are input into a pre-trained neural network model for rock strata feature extraction and identification. The identification results are analyzed to obtain the excavation feedback information of each section of the tunnel face.

[0048] In this embodiment, the specific method for acquiring the working face image is to use a camera device, such as a webcam or camcorder, directly mounted on the working face for real-time shooting.

[0049] In this embodiment, image noise reduction may be included before image enhancement is performed on the face images of each section of the tunnel region.

[0050] In this embodiment, the pre-trained neural network model is a neural network model trained in advance.

[0051] The above technical solution offers the following benefits: Through a pre-trained neural network model, rock strata features in various sections of the tunnel area can be automatically extracted and identified. Compared to manual identification, the neural network model can quickly and accurately identify the types and properties of different rock strata, improving work efficiency. Real-time analysis and identification of tunnel face images allows for timely acquisition of rock strata features and excavation status information for each section of the tunnel area. This helps engineers and operators understand the tunnel excavation process and make corresponding adjustments and decisions based on feedback information. Accurate rock strata feature identification can help promptly identify potential rock strata problems such as rock bursts and local collapses, enabling the implementation of necessary safety measures. Furthermore, real-time analysis of excavation status and rock strata properties can optimize excavation strategies, improving construction efficiency and quality. Using deep learning models for rock strata feature extraction and identification allows for the accumulation of a large amount of image data, from which the relationship between rock strata features and excavation feedback information can be learned and summarized. This helps form a knowledge base, providing experience and reference for similar engineering projects.

[0052] Example 7 The image enhancement of the tunnel face images of each section of the tunnel area includes: The tunnel face images of each section are segmented into edge images and non-edge images; The edge image is converted to grayscale to obtain the first grayscale image; Obtain the grayscale value of each pixel in the first grayscale image; randomly select a pixel in the first grayscale image to obtain the first pixel; determine an n×n sliding window centered on the first pixel; calculate the absolute value of the difference between the grayscale value of the first pixel and the grayscale values ​​of each pixel in the first preset direction in the sliding window, and determine the largest absolute value as the first absolute value; Calculate the absolute value of the difference in grayscale value between the first pixel in the sliding window and each pixel in the second preset direction, and determine the largest absolute value as the second absolute value; wherein the first preset direction and the second preset direction are opposite directions; Calculate the absolute value of the difference in grayscale value between the first pixel in the sliding window and each pixel in the third preset direction, and determine the largest absolute value as the third absolute value; wherein, the third preset direction is perpendicular to the first preset direction and the second preset direction; Calculate the absolute value of the difference in grayscale value between the first pixel in the sliding window and each pixel in the fourth preset direction, and determine the largest absolute value as the fourth absolute value; wherein the third preset direction and the fourth preset direction are opposite directions; Compare the first absolute value with the second absolute value, determine the gray value of each pixel and the total number of pixels in the preset direction corresponding to the largest absolute value, and obtain the first set of pixels. Compare the third absolute value with the fourth absolute value, determine the gray value of each pixel and the total number of pixels in the preset direction corresponding to the largest absolute value, and obtain the second set of pixels. Calculate the weighted average of the gray values ​​of all pixels in the first pixel set and the second pixel set, and replace the gray value of the first pixel with the weighted average. Iterate through all pixels in the first grayscale image to obtain the edge-enhanced image; The non-edge image is processed to obtain a second grayscale image; Obtain the grayscale value of each pixel in the second grayscale image, and enhance the brightness of the second grayscale image to obtain the grayscale enhanced image; The edge enhancement image and the grayscale enhancement image are overlaid and merged to enhance the face image of each section of the tunnel region.

[0053] In this embodiment, edge image grayscale processing is performed. Color images contain rich color information, requiring the processing of a large amount of color data, while grayscale images only contain brightness information, requiring only the processing of brightness data, thus reducing processing complexity. Since grayscale images have fewer pixels, processing is simpler and more convenient, while avoiding interference and errors caused by color in image recognition, thereby improving image processing speed and accuracy. The grayscale image can highlight the target area, increase contrast, and make the image clearer and easier to see.

[0054] In this embodiment, the technical solution is illustrated with specific examples; Assuming the first pixel is k, a 5×5 sliding window is defined with the first pixel k as the center, as shown in the table: The specific grayscale values ​​in the table are all assumed settings to facilitate a better understanding of the technical solution; Assuming the first preset direction is horizontal to the right of pixel k; the second preset direction is horizontal to the left of pixel k; the third preset direction is vertical upward above pixel k; and the fourth preset direction is vertical downward below pixel k. Assuming the grayscale value of pixel k is 40, then the absolute values ​​of the differences between the grayscale value of pixel k and the grayscale values ​​of all pixels in the first preset direction are 20 and 10, respectively. Therefore, the first absolute value is 20; similarly, the second absolute value is 10, the third absolute value is 28, and the fourth absolute value is 9. Since the first absolute value is greater than the second absolute value, the first set of pixels is {60, 50}; since the third absolute value is greater than the fourth absolute value, the second set of pixels is {25, 12}. The weighted average is then rounded to 45. This scheme enhances edge pixels by taking into full account multiple preset directions. Reference pixels are selected from multiple preset directions for enhancement. Since a single preset direction cannot accurately determine the edge gradient, it is easily affected by noise interference or the presence of edge details, which can cause errors. This scheme can effectively avoid such errors. The enhanced grayscale value is determined by weighted average of the pixels, which fully considers the importance of each pixel, solves the problem of sample imbalance, considers the reliability of the data, and reduces the impact of outliers on the results.

[0055] In this embodiment, the specific implementation method for enhancing the brightness of the second grayscale image can be linear transformation, histogram equalization, contrast enhancement, nonlinear transformation, etc., and no specific limitation is made here.

[0056] The beneficial effects of the above technical solution are as follows: The tunnel face image of each section is segmented into edge images and non-edge images. Image enhancement is performed on both edge and non-edge images separately. Enhancing the edge image makes the edges more prominent, thus improving the accuracy of the edge detection algorithm; it makes the target boundaries clearer, helping target detection and image segmentation algorithms to better identify and locate target objects; enhancing the non-edge image increases the image's contrast, brightness, and color vibrancy; it makes details in the image more clearly visible, thus increasing the image quality and detail expression ability, and enhancing the details of the tunnel face image; the sliding window image enhancement of the first grayscale image, instead of using overall enhancement, effectively… This method increases the accuracy of edge enhancement; it selects reference pixels from multiple preset directions for enhancement, as a single preset direction cannot accurately determine the edge gradient and is easily affected by noise interference or the presence of edge details, thus avoiding such errors; it determines the enhanced grayscale value by weighted averaging of pixels, fully considering the importance of each pixel, solving the sample imbalance problem, considering data reliability, and reducing the impact of outliers on the results; enhancing non-edge images makes the details in the tunnel face image clearer, thereby increasing the quality and detail expression of the tunnel face image; and it combines the edge enhancement image with the grayscale enhancement image to complete the enhancement of the tunnel face image of each segment of the tunnel region.

[0057] Example 8 like Figure 3 As shown, after marking the corresponding segment as a prediction error segment and issuing an alarm, the method further includes: Based on the feedback information from the excavation of the tunnel face corresponding to the incorrect forecast section, a new forecasting method is selected to conduct a new geological forecast. Then, the construction plan is adjusted according to the new geological forecast to carry out tunnel construction.

[0058] The above technical solution offers the following benefits: By re-forecasting and adjusting construction plans for sections with erroneous forecasts, safety risks arising from discrepancies between geological conditions and expectations can be avoided. This helps reduce construction accidents and ensures the safety of construction personnel and equipment. By selecting a new forecasting method for geological prediction and adjusting the construction plan based on the new forecast results, geological conditions can be predicted more accurately, avoiding unnecessary construction adjustments and work stoppages, thereby improving construction efficiency. By accurately forecasting geological conditions and adjusting the construction plan accordingly, construction adjustments and repetitive work caused by geological problems can be reduced, lowering construction costs. By re-selecting a forecasting method for geological prediction and adjusting the construction plan based on the new forecast results, geological risks can be better controlled, thereby improving the quality and stability of tunnel construction.

[0059] Example 9 After determining the final geological data for the tunnel area, the cleaned data also includes: Record the cleaning steps, processing methods, and data changes of geological data in the tunnel area during the data cleaning process to obtain a data cleaning document.

[0060] The above technical solution offers the following benefits: By recording the cleaning steps, processing methods, and data changes, the data cleaning process can be easily traced, allowing for understanding of each step of the operation and data processing, ensuring data reliability and credibility. During the data cleaning process, if data anomalies or errors are discovered, reviewing the cleaning steps and processing methods allows for identifying the root cause of the data problem and implementing appropriate corrections and repairs, improving data accuracy and quality. The data cleaning documentation records the cleaning steps and processing methods, serving as an accumulation of experience and knowledge for reference in subsequent similar projects, improving the efficiency and accuracy of data cleaning and promoting knowledge sharing and collaboration within the team. Recording the data cleaning process and generating cleaning documentation reduces repetitive work, eliminates redundant data cleaning steps, and improves overall data cleaning efficiency and work efficiency.

[0061] Example 10 The comparison of the geological forecast for the corresponding section with the excavation feedback information from the tunnel face includes: Based on the geological forecast of the corresponding section, predict the lithology, rock mass structure and geological structure of the working face; Based on the excavation feedback information of the corresponding section of the working face, we can obtain the lithology, rock mass structure and geological structure of the corresponding section of the working face; The lithology, rock mass structure, and geological structure of the corresponding section's face predicted by geological forecasts are compared with the lithology, rock mass structure, and geological structure of the corresponding section's face obtained from the excavation feedback information.

[0062] The beneficial effects of the above technical solution are as follows: By comparing the geological forecasts for the corresponding sections with the feedback information from the tunnel face excavation, the accuracy and reliability of the geological forecasts can be assessed. If the geological forecast matches the actual tunnel face conditions, it indicates that the geological forecast has high accuracy; if there are discrepancies, the accuracy of the geological forecast needs to be reassessed to improve prediction precision. Differences between the geological forecasts and the actual excavation conditions for the corresponding sections can be identified, i.e., geological changes can be detected. This helps to identify geological risks and trends, take timely countermeasures and adjust construction plans, and reduce unexpected and unpredictable geological problems. Differences between predictions and actual data can be analyzed, and the model can be optimized. By continuously adjusting and updating the geological forecast model, the accuracy and reliability of predictions can be improved, providing a more reliable geological forecast basis for tunnel construction. Inappropriate prediction results can be identified in advance, and construction plans can be adjusted in a timely manner to improve construction efficiency and ensure construction safety. Accurate geological forecasts can avoid geological problems during construction, reduce construction risks, and improve construction quality and efficiency.

[0063] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A digital forecasting and early warning method for tunnel geological hazards in high-risk karst areas, characterized in that, include: Obtain geological data for the tunnel area; The geological data includes geological structure data, stratigraphic lithology data, groundwater data, and environmental condition data; Based on geological data and through geological analysis, the surrounding rock conditions of each section within the tunnel area are determined, and a risk level classification is established. This includes: lithological indicators of the determined section When the value is greater than 0.55, the corresponding section is determined to be at the first risk level; The lithological index of the determined section is 0.15 < When the value is less than 0.55, the corresponding section is classified as the second risk level. lithological indicators of the determined section When the value is less than 0.15, the corresponding section is classified as the third risk level. Based on the risk level classification results, the corresponding forecasting method is matched, and geological forecasting is performed; Matching the corresponding forecasting method based on the risk level classification results includes: For sections of the first risk level, a forecasting method using TSP (Transient Spray) for 100-150m is adopted; For sections of the second risk level, a forecast method using TSP forecasts at 100m is adopted; For sections with the third risk level, short-term forecasts are made using ground-penetrating radar (0-20m) or transient electromagnetic forecasts (30-50m), and further confirmed by advance drilling or by lateral drilling at the tunnel face in the tunnel area. Obtain excavation feedback information from the tunnel face of each section; The acquisition of excavation feedback information from each section of the tunnel face includes: Obtain the tunnel face images of each section of the tunnel area; Image enhancement was performed on the tunnel face images of each section of the tunnel area; The enhanced tunnel face images of each section are input into a pre-trained neural network model for rock strata feature extraction and identification. The identification results are analyzed to obtain the excavation feedback information of each section of the tunnel face. The geological forecasts for the corresponding sections are compared with the feedback information from the excavation at the tunnel face. If the two are found to be inconsistent, the corresponding section is marked as a forecast error section and an alarm is issued. After marking the corresponding segment as a prediction error segment and issuing an alarm, the system further includes: Based on the feedback information from the excavation of the tunnel face corresponding to the incorrect forecast section, a new forecasting method is selected to conduct a new geological forecast. Then, the construction plan is adjusted according to the new geological forecast to carry out tunnel construction.

2. The digital forecasting and early warning method for tunnel geological hazards in high-risk karst areas according to claim 1, characterized in that, Before determining the surrounding rock conditions of each section within the tunnel area through geological analysis based on geological data, the following steps are also included: Divide the geological data into several geological text data; Each geological text data is segmented into words to obtain several keywords corresponding to each geological text data. Take any geological text data, calculate the degree of belonging of several keywords in the geological text data to keywords in the preset keyword classification label table, and determine the label in the preset keyword classification table corresponding to the maximum degree of belonging as the pre-selected category label of the geological text data; Obtain the first text vector of the geological text data and the second text vector corresponding to the pre-selected category labels; the second text vector is the average of the text vectors of all keywords in the pre-selected category labels; Calculate the similarity between the first text vector and the second text vector; When the similarity is determined to be greater than the preset similarity threshold, the pre-selected category label will be used as the category label corresponding to the text data; Based on the above methods, the category label corresponding to each text data in the geological data is obtained; data with the same category label are merged to obtain geological data of different categories; Select any category of geological data for data cleaning to obtain the first data to be cleaned; obtain a first preset number of data from the first data to be cleaned to obtain the first sub-data to be cleaned; The first sub-data to be cleaned is cleaned using several pre-set data cleaning schemes; the completion rate of each data cleaning scheme for the first sub-data to be cleaned is calculated. The data cleaning scheme with the highest completion rate for cleaning the first sub-data to be cleaned is selected as the first data cleaning scheme, and the cleaning completion rate of the first data cleaning scheme is determined as the first completion rate. Obtain a second preset number of data points from the first set of data to be cleaned to obtain the second set of sub-data points to be cleaned; wherein the second preset number is greater than the first preset number. Based on the first data cleaning scheme, the second sub-data to be cleaned is cleaned, and the cleaning completion rate of the first data cleaning scheme on the second sub-data to be cleaned is calculated to obtain the second completion rate. Calculate the absolute value of the difference between the first completion level and the second completion level. When the absolute value is less than or equal to a preset difference threshold, determine the first data cleaning scheme as the target data cleaning scheme corresponding to the first data to be cleaned. Complete the data cleaning of the first data to be cleaned based on the first data cleaning scheme. Based on the above methods, the geological data cleaning is completed, and the cleaned geological data is obtained.

3. The digital forecasting and early warning method for tunnel geological hazards in high-risk karst areas according to claim 1, characterized in that, The image enhancement of the tunnel face images of each section of the tunnel area includes: The tunnel face images of each section are segmented into edge images and non-edge images; The edge image is converted to grayscale to obtain the first grayscale image; Obtain the grayscale value of each pixel in the first grayscale image; randomly select a pixel in the first grayscale image to obtain the first pixel; determine an n×n sliding window centered on the first pixel; calculate the absolute value of the difference between the grayscale value of the first pixel and the grayscale values ​​of each pixel in the first preset direction in the sliding window, and determine the largest absolute value as the first absolute value; Calculate the absolute value of the difference in grayscale value between the first pixel in the sliding window and each pixel in the second preset direction, and determine the largest absolute value as the second absolute value; wherein the first preset direction and the second preset direction are opposite directions; Calculate the absolute value of the difference in grayscale value between the first pixel in the sliding window and each pixel in the third preset direction, and determine the largest absolute value as the third absolute value; wherein, the third preset direction is perpendicular to the first preset direction and the second preset direction; Calculate the absolute value of the difference in grayscale value between the first pixel in the sliding window and each pixel in the fourth preset direction, and determine the largest absolute value as the fourth absolute value; wherein the third preset direction and the fourth preset direction are opposite directions; Compare the first absolute value with the second absolute value, determine the gray value of each pixel and the total number of pixels in the preset direction corresponding to the largest absolute value, and obtain the first set of pixels. Compare the third absolute value with the fourth absolute value, determine the gray value of each pixel and the total number of pixels in the preset direction corresponding to the largest absolute value, and obtain the second set of pixels. Calculate the weighted average of the gray values ​​of all pixels in the first pixel set and the second pixel set, and replace the gray value of the first pixel with the weighted average. Iterate through all pixels in the first grayscale image to obtain the edge-enhanced image; The non-edge image is processed to obtain a second grayscale image; Obtain the grayscale value of each pixel in the second grayscale image, and enhance the brightness of the second grayscale image to obtain the grayscale enhanced image; The edge enhancement image and the grayscale enhancement image are overlaid and merged to enhance the face image of each section of the tunnel region.

4. The digital forecasting and early warning method for tunnel geological hazards in high-risk karst areas according to claim 2, characterized in that, After determining the final geological data for the tunnel area, the cleaned data also includes: Record the cleaning steps, processing methods, and data changes of geological data in the tunnel area during the data cleaning process to obtain a data cleaning document.

5. The digital forecasting and early warning method for tunnel geological hazards in high-risk karst areas according to claim 1, characterized in that, The comparison of the geological forecast for the corresponding section with the excavation feedback information from the tunnel face includes: Based on the geological forecast of the corresponding section, predict the lithology, rock mass structure and geological structure of the working face; Based on the excavation feedback information of the corresponding section of the working face, we can obtain the lithology, rock mass structure and geological structure of the corresponding section of the working face; The lithology, rock mass structure, and geological structure of the corresponding section's face predicted by geological forecasts are compared with the lithology, rock mass structure, and geological structure of the corresponding section's face obtained from the excavation feedback information.

Images