Water-rich tunnel safety early warning method and system based on multi-source heterogeneous data

Abstract

The application discloses a kind of based on multi-source heterogeneous data's rich water tunnel safety early warning method and system, it is related to rich water tunnel early warning technical field, including obtaining historical tunnel rich water detection data, cover rock mass parameter, water body permeation pressure parameter and rock mass displacement parameter, data is analyzed, according to whether rock mass displacement parameter is greater than or equal to preset value, early warning or safety mark is carried out;Mark data is mapped to template, and the parameter state performance model for early warning and safety is formed;Then, compare two models, extract the generalization features and non-generalization features of early warning model;Finally, fuzzy comparison is carried out on real-time data using generalization features, and non-generalization features are accurately compared, to realize rich water tunnel early warning.The method effectively integrates multi-source data, improves the accuracy and real-time performance of early warning, and is suitable for complex geological environment.

Description

Technical Field

[0001] This invention relates to the field of early warning technology for water-rich tunnels, and in particular to a method and system for safety early warning of water-rich tunnels based on multi-source heterogeneous data. Background Technology

[0002] In the field of tunnel engineering, especially during the construction of water-rich tunnels, complex geological conditions, variable hydrological environments, and heterogeneous rock mass structures often present safety hazards such as water inrush and gushing. These hazards can lead to tunnel collapse, equipment damage, or casualties, seriously affecting project progress and safety. While existing technologies employ sensors to monitor parameters such as rock mass displacement and water osmotic pressure, the integration and analysis of multi-source heterogeneous data still have significant shortcomings: on the one hand, historical monitoring data is often processed in isolation, failing to effectively extract generalized and non-generalized features, resulting in insufficient generalization ability of early warning models and an inability to cope with the ambiguity and uncertainty of real-time data; on the other hand, traditional methods often rely on single parameter threshold judgments, ignoring the diversity of rock mass parameter types and the correlation between data, leading to low early warning accuracy, high false alarm rates, or missed alarm risks, especially when rock mass displacement is abnormal, making accurate comparison and timely response difficult. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for early warning of water-rich tunnel safety with high accuracy.

[0004] This invention discloses a safety early warning method for water-rich tunnels based on multi-source heterogeneous data, including:

[0005] Step S100: Obtain several historical tunnel water-rich detection data, including rock mass parameters, water osmotic pressure parameters, and rock mass displacement parameters.

[0006] Step S200: Analyze several historical tunnel water-rich detection data, mark historical tunnel water-rich detection data with rock mass displacement parameters greater than or equal to preset values ​​as early warning data, and mark other historical tunnel water-rich detection data as safe data.

[0007] Step S300: Determine the rock mass parameter type, construct a parameter state representation template for each rock mass parameter type and water permeability parameter, and classify the historical tunnel water-rich detection data with warning and safety markers onto the parameter state representation template for mapping representation, thereby obtaining the parameter state representation model for warning and the parameter state representation model for safety respectively.

[0008] Step S400: Compare the parameter state performance model for early warning and the parameter state performance model for safety to determine the generalization and non-generalization features of the parameter state performance model for early warning. Step S500: Use the generalization features of the parameter state performance model for early warning to perform a fuzzy comparison of the real-time tunnel water-rich detection data, and use the non-generalization features of the parameter state performance model for early warning to perform a precise comparison, thereby achieving early warning for water-rich tunnels.

[0009] In some embodiments disclosed in this invention, the generalization feature is characterized by the fact that the parameter state representation model for early warning has a concentrated representation compared to the state representation model for safety, while the non-generalization feature is characterized by the fact that the parameter state representation model for early warning does not have a concentrated representation compared to the state representation model for safety.

[0010] In some embodiments disclosed in this invention, the method for constructing a parameter state representation template includes:

[0011] Step S301: Construct a parameter vertical axis for each parameter and arrange the parameter vertical axis points in parallel sequentially;

[0012] Step S302: Based on the rock mass parameters and water permeability parameters in the historical tunnel water-rich detection data, configure the corresponding parameter mapping points on the corresponding parameter axis, and connect the parameter mapping points in sequence to form a parameter mapping polyline.

[0013] Step S303: Configure the parameter mapping polylines corresponding to several historical tunnel water-rich detection data on the same parameter number axis system to obtain the parameter state representation template.

[0014] In some embodiments disclosed in this invention, the method for determining the generalization feature representation of the parametric state representation model for early warning includes:

[0015] Step S401: Define a preset parameter range for each inflection point of the parameter mapping polyline in the parameter state representation model for early warning, and count other parameter mapping polylines that fall within the preset parameter range of each inflection point, and record them as the number of similar mapping polylines.

[0016] Step S402: Calculate the ratio of the similar mapping line corresponding to each parameter mapping line to all parameter mapping lines, and record it as the similar mapping volume ratio of the parameter mapping line. If the similar mapping volume ratio is greater than or equal to a preset value, the parameter mapping line is identified as a parameter mapping line that needs attention.

[0017] Step S403: Analyze the proximity features between the mapping lines of the parameters of interest, determine the range mapped by the mapping lines of the parameters of interest that meet the preset proximity criteria, and identify the range mapped by the mapping lines of the parameters of interest that meet the preset proximity criteria as the range of the generalization feature representation mapping.

[0018] Methods for identifying the non-generalization features of the parametric state representation model for early warning include:

[0019] S404, determine the range of the generalized feature representation mapped in the early warning parameter state representation model, determine the range mapped by the parameter mapping polyline in the safety parameter state representation template, calculate the difference range between the range mapped by the generalized feature representation and the range of the generalized feature representation, and identify the difference range as the range of the non-generalized feature representation.

[0020] In some embodiments disclosed in this invention, the method for analyzing the proximity characteristics between the mapping polylines of parameters of interest includes:

[0021] Step S4031: Perform a proximity analysis on adjacent mapping lines of parameters of interest. The proximity analysis method includes determining the mapping area mapped by the corresponding close mapping lines of each and determining the difference in mapping area.

[0022] Step S4032: Determine the distance between each inflection point of adjacent mapping polylines of parameters of interest, and determine the proximity between mapping polylines of parameters of interest by combining the difference in mapping area.

[0023] The expression for calculating the degree of proximity is:

[0024] ;

[0025] in, To approximate the degree of similarity, The function for determining proximity between points is based on the distance interval between points at each turning point. Output the corresponding preset value. This is the preset maximum mapping area difference. This represents the difference in the mapped area between similarly mapped polylines. Adjustment factor for the influence of the difference in mapped area. Adjust the constant to account for the influence of the difference in the mapped area.

[0026] In some embodiments disclosed in this invention, the method for defining a preset parameter range for each inflection point of the parameter mapping polyline in the parameter state representation model for early warning further includes:

[0027] Step S4011: Before defining the preset parameter range for each inflection point in the parameter state representation model for early warning, a portion of the parameter mapping polyline is reserved, which is denoted as the reserved parameter mapping polyline.

[0028] Step S4012: Determine whether the reserved parameter mapping polyline belongs to the range of generalization feature representation mapping. If it does, retain the reserved parameter mapping polyline and remove the reserved parameter mapping polyline that does not belong.

[0029] Step S4013: Compare the reserved parameter mapping polyline with the parameter mapping polyline used to construct the generalization feature representation mapping range. If there is a parameter mapping polyline where the difference between each inflection point is less than the preset value, then mark the reserved parameter mapping polyline.

[0030] Step S4014: Calculate the reserved percentage of the marked reserved parameter mapping line relative to all reserved parameter mapping lines, and determine the preset adjustment coefficient for the preset parameter range based on the preset percentage range to which the marked reserved percentage belongs.

[0031] In some embodiments disclosed in this invention, the method for performing fuzzy comparison of real-time tunnel water-rich detection data using the generalization features of the early warning parametric state representation model includes:

[0032] Step S501: Based on real-time tunnel water-rich detection data, determine real-time rock mass parameters and real-time water osmotic pressure parameters, and configure the real-time rock mass parameters and real-time water osmotic pressure parameters on the parameter state performance model for early warning according to the corresponding relationship. If each real-time parameter mapping point falls within the range corresponding to the generalization feature performance, then an early warning is issued.

[0033] In some embodiments disclosed in this invention, the method for accurate comparison using the non-generalized features of the early warning parametric state representation model includes:

[0034] Step S501: Connect the real-time parameter mapping points sequentially to form a real-time parameter mapping polyline. If part of the real-time parameter mapping polyline falls within the range corresponding to the non-generalized feature expression, find a real-time parameter mapping polyline that matches the real-time parameter mapping polyline within the parameter mapping polyline recorded by the non-generalized feature expression. If a matching real-time parameter mapping polyline does exist, then issue a safety warning for the water-rich tunnel.

[0035] In some embodiments disclosed in this invention, the method for determining whether the parametric mapping polyline recorded by the non-generalized feature representation is a good match with the real-time parametric mapping polyline includes:

[0036] Step S5011: Calculate the matching degree between the parameter mapping polyline recorded by the non-generalized feature performance and the real-time parameter mapping polyline. If the matching degree is greater than or equal to the preset value, then determine that the two are compatible.

[0037] The expression for calculating the degree of matching is:

[0038] ;

[0039] in, For the degree of matching, For the first The function for judging the difference between several inflection points determines the sub-matching degree of the inflection point output based on the preset inflection point difference interval to which the difference belongs. This represents the total number of turning points.

[0040] In some embodiments disclosed in this invention, a safety early warning system for water-rich tunnels based on multi-source heterogeneous data is also disclosed, including:

[0041] The first module is used to acquire several historical tunnel water-rich detection data, including rock mass parameters, water osmotic pressure parameters, and rock mass displacement parameters. The water osmotic pressure parameters are detected by osmotic pressure sensors installed on the rock mass, and the rock mass displacement parameters are detected by displacement sensors installed on the rock mass.

[0042] The second module is used to analyze several historical tunnel water-rich detection data, to mark historical tunnel water-rich detection data with rock mass displacement parameters greater than or equal to preset values ​​as early warning data, and to mark other historical tunnel water-rich detection data as safe data.

[0043] The third module is used to determine the rock mass parameter type. For each type of rock mass parameter and water permeability parameter, a parameter status representation template is constructed. Historical tunnel water-rich detection data with warning and safety markers are classified and mapped onto the parameter status representation template to obtain the parameter status representation model for warning and the parameter status representation model for safety, respectively.

[0044] The fourth module is used to compare the parameter state performance model for early warning and the parameter state performance model for safety, and to determine the generalization feature and non-generalization feature of the parameter state performance model for early warning. The generalization feature is the feature that the parameter state performance model for early warning has a concentrated expression relative to the parameter state performance model for safety, and the non-generalization feature is the feature that the parameter state performance model for early warning does not have a concentrated expression relative to the parameter state performance model for safety.

[0045] The fifth module is used to perform fuzzy comparison of real-time tunnel water-rich detection data by utilizing the generalization features of the early warning parameter state representation model, and to perform precise comparison by utilizing the non-generalization features of the early warning parameter state representation model, thereby realizing early warning of water-rich tunnels.

[0046] This invention discloses a method and system for safety early warning of water-rich tunnels based on multi-source heterogeneous data, belonging to the field of water-rich tunnel early warning technology. The method includes acquiring historical water-rich detection data of tunnels, covering rock mass parameters, water osmotic pressure parameters, and rock mass displacement parameters; analyzing the data; issuing early warnings or safety markers based on whether the rock mass displacement parameter is greater than or equal to a preset value; mapping the marked data to templates to form parameter state representation models for early warning and safety purposes; comparing the two models to extract generalized and non-generalized features of the early warning model; finally, using the generalized features to perform fuzzy comparisons of real-time data and precise comparisons of the non-generalized features to achieve early warning of water-rich tunnels. This method effectively integrates multi-source data, improves the accuracy and real-time performance of early warnings, and is suitable for complex geological environments.

[0047] 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

[0048] Figure 1 This is a flowchart illustrating the method steps of the water-rich tunnel safety early warning method based on multi-source heterogeneous data disclosed in this embodiment of the invention. Detailed Implementation

[0049] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0050] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific embodiments. It should be understood that the preferred embodiments described herein are only for illustration and explanation of the present invention and should not be construed as limiting the scope of protection of the present invention. Those skilled in the art can make some non-essential improvements and adjustments based on the following content of the present invention. In the present invention, unless otherwise expressly specified and limited, the technical terms used in the present invention should have the ordinary meaning understood by those skilled in the art.

[0051] Example:

[0052] This invention discloses a safety early warning method for water-rich tunnels based on multi-source heterogeneous data. (See reference...) Figure 1 ,include:

[0053] Step S100: Obtain several historical tunnel water-rich detection data, including rock mass parameters, water osmotic pressure parameters, and rock mass displacement parameters. The water osmotic pressure parameters are detected by osmotic pressure sensors installed on the rock mass, and the rock mass displacement parameters are detected by displacement sensors installed on the rock mass.

[0054] The principle of step S100 is to collect and integrate multi-source heterogeneous historical tunnel water-rich detection data to provide a basic dataset for subsequent analysis and model building. This data includes rock mass parameters (such as rock strength, fracture density, etc.), water osmotic pressure parameters (reflecting the osmotic pressure of water on the rock mass), and rock mass displacement parameters (indicating the degree of rock mass deformation). The water osmotic pressure parameters are acquired in real time by osmotic pressure sensors installed on the rock mass, ensuring that the data reflects the dynamic pressure of the interaction between water and rock mass. The rock mass displacement parameters are captured by displacement sensors to capture the minute displacement changes of the rock mass under water pressure. This data acquisition method realizes a comprehensive quantitative description of the tunnel water-rich environment, avoids the limitations of a single data source, and lays a multi-dimensional historical experience foundation for the training of the early warning system, thereby improving the reliability and comprehensiveness of the early warning.

[0055] Step S200: Analyze several historical tunnel water-rich detection data, mark historical tunnel water-rich detection data with rock mass displacement parameters greater than or equal to preset values ​​as early warning data, and mark other historical tunnel water-rich detection data as safe data.

[0056] The principle of step S200 is to perform preliminary classification and labeling of the collected historical data. By setting a preset threshold for rock mass displacement as a judgment standard, when the rock mass displacement parameter is greater than or equal to the threshold, it indicates that the rock mass has undergone significant deformation, which may indicate a critical state of water-rich risk. Therefore, it is marked as a warning to highlight potentially dangerous samples. Conversely, if the displacement is less than the threshold, it is considered a safe state and is marked as safe. This threshold-based labeling mechanism realizes the binary partitioning of data, effectively separating high-risk and low-risk datasets, providing a supervised learning foundation for the construction of subsequent models, while reducing noise interference, improving the targeting and efficiency of data processing, and ensuring that the early warning model can focus on learning the real risk characteristics.

[0057] Step S300: Determine the rock mass parameter type, construct a parameter state representation template for each rock mass parameter type and water permeability parameter, and map the historical tunnel water-rich detection data with early warning and safety markers onto the parameter state representation template to obtain the parameter state representation model for early warning and the parameter state representation model for safety.

[0058] The principle of step S300 is to first identify the diverse types of rock mass parameters (such as rock type, porosity, etc.), and then construct a unified parameter state representation template for each type of rock mass parameter and water permeability parameter. This template is essentially a multi-dimensional coordinate system or line graph framework used to visualize and quantify the relationship between parameters. Subsequently, historical data with warning markers are mapped onto the template to form a parameter state representation model for warning purposes. Similarly, safety marker data is mapped to form a safety model. This mapping process captures the pattern and distribution differences between parameters by connecting data points into lines or range representations, thereby transforming abstract heterogeneous data into a comparable graphical model. This facilitates subsequent feature extraction and the abstract representation of complex geological conditions, improving the interpretability and generalization ability of the model.

[0059] Step S400: Compare the parameter state performance model for early warning and the parameter state performance model for safety, and determine the generalization feature performance and non-generalization feature performance of the parameter state performance model for early warning. The generalization feature performance is the feature performance of the parameter state performance model for early warning that has a concentrated performance relative to the parameter state performance model for safety, and the non-generalization feature performance is the feature performance of the parameter state performance model for early warning that does not have a concentrated performance relative to the parameter state performance model for safety.

[0060] The principle of step S400 is to identify the unique feature distribution in the early warning model relative to the safety model by comparing the parameter state performance model used for early warning with the safety model used for safety. The generalized feature performance refers to the concentrated and repetitive patterns (such as dense broken lines within a specific parameter range) in the early warning model. These patterns represent general indicators of water-rich risk that are applicable to a variety of scenarios. The non-generalized feature performance refers to those patterns that are not concentrated, are more dispersed, or are unique in the early warning model, indicating risk variations under specific conditions. This comparison mechanism uses statistical or cluster analysis methods to quantify the differences, ensuring that the extracted features can cover both generalized risks and special cases (non-generalized), thereby constructing a robust early warning framework, avoiding model overfitting or underfitting, and improving adaptability to unknown data.

[0061] Step S500: Use the generalization feature of the parameter state representation model for early warning to perform fuzzy comparison on the real-time tunnel water-rich detection data, and use the non-generalization feature of the parameter state representation model for early warning to perform precise comparison, so as to realize early warning of water-rich tunnels.

[0062] The principle of step S500 is to use the extracted generalized features to perform fuzzy comparison on the real-time tunnel water-rich detection data. That is, by checking whether the real-time parameters fall within the concentrated range of the generalized features (such as the similarity of broken lines or the overlap of areas), a tolerant matching of uncertain or noisy data is achieved to capture the overall trend of potential risks. At the same time, non-generalized features are used for precise comparison, that is, the real-time data is strictly matched with the non-generalized pattern (such as the distance between points or the calculation of differences) to verify the precise existence of specific risks. This dual comparison strategy combines the fault tolerance of fuzzy logic and the accuracy of precise algorithms. Finally, the early warning mechanism of water-rich tunnels is triggered based on the comparison results to ensure timely response to potential hazards in complex environments, reduce false alarms, and improve the level of safety assurance.

[0063] In some embodiments disclosed in this invention, the method for constructing a parameter state representation template includes:

[0064] Step S301: Construct a parameter vertical axis for each parameter and arrange the parameter vertical axis points in parallel sequentially.

[0065] Step S302: Based on the rock mass parameters and water permeability parameters in the historical tunnel water-rich detection data, configure the corresponding parameter mapping points on the corresponding parameter axes, and connect the parameter mapping points in sequence to form a parameter mapping polyline.

[0066] Step S303: Configure the parameter mapping polylines corresponding to several historical tunnel water-rich detection data on the same parameter number axis system to obtain the parameter state representation template.

[0067] In some embodiments disclosed in this invention, the method for determining the generalization feature representation of the parametric state representation model for early warning includes:

[0068] Step S401: Define a preset parameter range for each inflection point of the parameter mapping polyline in the parameter state representation model for early warning, and count other parameter mapping polylines that fall within the preset parameter range of each inflection point, and record them as the number of similar mapping polylines.

[0069] The principle of step S401 is to define the local range of the inflection point of each parameter mapping polyline in the parameter state representation model for early warning. That is, to set a preset parameter range (such as a small interval centered on the inflection point) for each inflection point, and then to count the number of other parameter mapping polylines falling within these ranges and calculate the number of similar mapping polylines. This statistical mechanism is essentially a local similarity assessment, which can quantify the degree of clustering of each polyline at key points, avoid the coarseness of global analysis, and thus identify potential pattern-dense areas. This provides a data foundation for subsequent screening of high-influence polylines, ensures that the extracted generalized features are based on real distribution concentration rather than random noise, and improves the accuracy and robustness of feature extraction.

[0070] Step S402: Calculate the ratio of the similar mapping line to all parameter mapping lines corresponding to each parameter mapping line, and record it as the similar mapping volume ratio of the parameter mapping line. If the similar mapping volume ratio is greater than or equal to a preset value, the parameter mapping line is identified as a parameter mapping line that needs attention.

[0071] The principle of step S402 is to calculate the ratio of the number of similar mapping lines obtained in step S401 to the total number of all parameter mapping lines in the model, and obtain the proportion of similar mapping volume of each line. This proportion reflects the representativeness and repetition of the line in the overall dataset. When the proportion is greater than or equal to a preset threshold, it is identified as a parameter mapping line that needs attention. This threshold filtering mechanism is similar to the core point identification in density clustering. It can effectively remove isolated or low-frequency abnormal lines and focus on those statistically significant patterns, thereby laying a reliable sample set for the construction of generalization features, avoiding overfitting to noisy data, and improving the model's ability to capture general risk features.

[0072] Step S403: Analyze the proximity features between the mapping lines of the parameters of interest, determine the range mapped by the mapping lines of the parameters of interest that meet the preset proximity criteria, and identify the range mapped by the mapping lines of the parameters of interest that meet the preset proximity criteria as the range of the generalization feature representation mapping.

[0073] In some embodiments disclosed in this invention, the method for determining the non-generalization feature representation of the parametric state representation model for early warning includes:

[0074] S404, determine the range of the generalized feature representation mapped in the early warning parameter state representation model, determine the range mapped by the parameter mapping polyline in the safety parameter state representation template, calculate the difference range between the range mapped by the generalized feature representation and the range of the generalized feature representation, and identify the difference range as the range of the non-generalized feature representation.

[0075] In some embodiments disclosed in this invention, the method for analyzing the proximity characteristics between the mapping polylines of parameters of interest includes:

[0076] Step S4031: Perform a proximity analysis on adjacent mapping lines of the parameters of interest. The proximity analysis method includes determining the mapping area mapped by the corresponding close mapping lines of each and determining the difference in mapping area.

[0077] Step S4032: Determine the distance between each inflection point of adjacent mapping lines of parameters of interest, and determine the proximity between mapping lines of parameters of interest by combining the difference in mapping area.

[0078] The expression for calculating the degree of proximity is:

[0079] .

[0080] in, To approximate the degree of similarity, The function for determining proximity between points is based on the distance interval between points at each turning point. Output the corresponding preset value. This is the preset maximum mapping area difference. This represents the difference in the mapped area between similarly mapped polylines. Adjustment factor for the influence of the difference in mapped area. Adjust the constant to account for the influence of the difference in the mapped area.

[0081] In some embodiments disclosed in this invention, the method for defining a preset parameter range for each inflection point of the parameter mapping polyline in the parameter state representation model for early warning further includes:

[0082] Step S4011: Before defining the preset parameter range for each inflection point in the parameter state representation model for early warning, a portion of the parameter mapping polyline is reserved, which is denoted as the reserved parameter mapping polyline.

[0083] Step S4012: Determine whether the reserved parameter mapping line belongs to the range of generalization feature representation mapping. If it does, retain the reserved parameter mapping line and remove the reserved parameter mapping line that does not belong.

[0084] Step S4013: Compare the reserved parameter mapping polyline with the parameter mapping polyline used to construct the generalization feature representation mapping range. If there is a parameter mapping polyline where the difference between each inflection point is less than the preset value, then mark the reserved parameter mapping polyline.

[0085] Step S4014: Calculate the reserved percentage of the marked reserved parameter mapping line relative to all reserved parameter mapping lines, and determine the preset adjustment coefficient for the preset parameter range based on the preset percentage range to which the marked reserved percentage belongs.

[0086] The principle of step S4014 is to calculate the ratio of the number of reserved parameter mapping lines of the markers to all initial reserved parameter mapping lines, obtain the reserved ratio of the markers, and then determine a corresponding preset adjustment coefficient based on the preset interval (such as low, medium and high ratio intervals) into which this ratio falls, so as to dynamically correct the size or boundary of the preset parameter range. This ratio-driven adjustment mechanism is essentially an adaptive feedback loop, which can optimize the range delineation parameters based on the success rate of the verification results (high ratio indicates that the generalization range is reliable, and low ratio indicates that the range needs to be expanded), so as to ensure that the preset parameter range is neither too loose to cause false noise, nor too strict to cause omission of features, thereby improving the adaptability and accuracy of the overall early warning model.

[0087] In some embodiments disclosed in this invention, the method for performing fuzzy comparison of real-time tunnel water-rich detection data using the generalization features of the early warning parametric state representation model includes:

[0088] Step S501: Based on real-time tunnel water-rich detection data, determine real-time rock mass parameters and real-time water osmotic pressure parameters, and configure the real-time rock mass parameters and real-time water osmotic pressure parameters on the parameter state performance model for early warning according to the corresponding relationship. If each real-time parameter mapping point falls within the range corresponding to the generalization feature performance, then an early warning is issued.

[0089] In some embodiments disclosed in this invention, the method for accurate comparison using the non-generalized features of the early warning parametric state representation model includes:

[0090] Step S501: Connect the real-time parameter mapping points sequentially to form a real-time parameter mapping polyline. If part of the real-time parameter mapping polyline falls within the range corresponding to the non-generalized feature expression, find a real-time parameter mapping polyline that matches the real-time parameter mapping polyline within the parameter mapping polyline recorded by the non-generalized feature expression. If a matching real-time parameter mapping polyline does exist, then issue a safety warning for the water-rich tunnel.

[0091] In some embodiments disclosed in this invention, the method for determining whether the parametric mapping polyline recorded by the non-generalized feature representation is a good match with the real-time parametric mapping polyline includes:

[0092] Step S5011: Calculate the matching degree between the parameter mapping polyline recorded by the non-generalized feature performance and the real-time parameter mapping polyline. If the matching degree is greater than or equal to the preset value, then determine that the two are compatible.

[0093] The expression for calculating the degree of matching is:

[0094] .

[0095] in, For the degree of matching, For the first The function for judging the difference between several inflection points determines the sub-matching degree of the inflection point output based on the preset inflection point difference interval to which the difference belongs. This represents the total number of turning points.

[0096] In some embodiments disclosed in this invention, a safety early warning system for water-rich tunnels based on multi-source heterogeneous data is also disclosed, including:

[0097] The first module is used to acquire several historical tunnel water-rich detection data, including rock mass parameters, water osmotic pressure parameters, and rock mass displacement parameters. The water osmotic pressure parameters are detected by osmotic pressure sensors installed on the rock mass, and the rock mass displacement parameters are detected by displacement sensors installed on the rock mass.

[0098] The second module is used to analyze several historical tunnel water-rich detection data, to issue early warning marks for historical tunnel water-rich detection data with rock mass displacement parameters greater than or equal to preset values, and to issue safety marks for other historical tunnel water-rich detection data.

[0099] The third module is used to determine the rock mass parameter type. For each type of rock mass parameter and water permeability parameter, a parameter state representation template is constructed. Historical tunnel water-rich detection data with warning and safety markers are classified and mapped onto the parameter state representation template to obtain the parameter state representation model for warning and the parameter state representation model for safety, respectively.

[0100] The fourth module is used to compare the parameter state performance model for early warning and the parameter state performance model for safety, and to determine the generalization and non-generalization features of the parameter state performance model for early warning. The generalization features are those features that the parameter state performance model for early warning has a concentrated expression compared to the parameter state performance model for safety, while the non-generalization features are those features that the parameter state performance model for early warning does not have a concentrated expression compared to the parameter state performance model for safety.

[0101] The fifth module is used to perform fuzzy comparison of real-time tunnel water-rich detection data by utilizing the generalization features of the early warning parameter state representation model, and to perform precise comparison by utilizing the non-generalization features of the early warning parameter state representation model, thereby realizing early warning of water-rich tunnels.

[0102] This invention discloses a method and system for safety early warning of water-rich tunnels based on multi-source heterogeneous data, belonging to the field of water-rich tunnel early warning technology. The method includes acquiring historical water-rich detection data of tunnels, covering rock mass parameters, water osmotic pressure parameters, and rock mass displacement parameters; analyzing the data; issuing early warnings or safety markers based on whether the rock mass displacement parameter is greater than or equal to a preset value; mapping the marked data to templates to form parameter state representation models for early warning and safety purposes; comparing the two models to extract generalized and non-generalized features of the early warning model; finally, using the generalized features to perform fuzzy comparisons of real-time data and precise comparisons of the non-generalized features to achieve early warning of water-rich tunnels. This method effectively integrates multi-source data, improves the accuracy and real-time performance of early warnings, and is suitable for complex geological environments.

[0103] Through the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented in hardware or by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0104] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A safety early warning method for water-rich tunnels based on multi-source heterogeneous data, characterized in that, include: Step S100: Obtain several historical tunnel water-rich detection data, including rock mass parameters, water osmotic pressure parameters, and rock mass displacement parameters. Step S200: Analyze several historical tunnel water-rich detection data, mark historical tunnel water-rich detection data with rock mass displacement parameters greater than or equal to preset values ​​as early warning data, and mark other historical tunnel water-rich detection data as safe data. Step S300: Determine the rock mass parameter type, construct a parameter state representation template for each rock mass parameter type and water permeability parameter, and classify the historical tunnel water-rich detection data with warning and safety markers onto the parameter state representation template for mapping representation, thereby obtaining the parameter state representation model for warning and the parameter state representation model for safety respectively. Step S400: Compare the parameter state performance model for early warning and the parameter state performance model for safety to determine the generalization and non-generalization features of the parameter state performance model for early warning. Step S500: Use the generalization feature of the parameter state representation model for early warning to perform fuzzy comparison on real-time tunnel water-rich detection data, and use the non-generalization feature of the parameter state representation model for early warning to perform precise comparison, so as to realize early warning of water-rich tunnels. The generalization feature is characterized by the fact that the parametric state representation model for early warning has a concentrated representation compared to the state representation model for safety, while the non-generalization feature is characterized by the fact that the parametric state representation model for early warning does not have a concentrated representation compared to the state representation model for safety. Methods for fuzzy comparison of real-time tunnel water-rich detection data using the generalization features of the early warning parametric state representation model include: Step S501: Based on real-time tunnel water-rich detection data, determine real-time rock mass parameters and real-time water osmotic pressure parameters, and configure the real-time rock mass parameters and real-time water osmotic pressure parameters on the parameter state performance model for early warning according to the corresponding relationship. If each real-time parameter mapping point falls within the range corresponding to the generalization feature performance, then an early warning is issued. Methods for accurate comparison using the non-generalization features of the early warning parametric state representation model include: Step S501: Connect the real-time parameter mapping points sequentially to form a real-time parameter mapping polyline. If part of the real-time parameter mapping polyline falls within the range corresponding to the non-generalized feature expression, find a real-time parameter mapping polyline that matches the real-time parameter mapping polyline within the parameter mapping polyline recorded by the non-generalized feature expression. If a matching real-time parameter mapping polyline does exist, then issue a safety warning for the water-rich tunnel.

2. The method for safety early warning of water-rich tunnels based on multi-source heterogeneous data according to claim 1, characterized in that, Methods for constructing parameter state representation templates include: Step S301: Construct a parameter vertical axis for each parameter and arrange the parameter vertical axis points in parallel sequentially; Step S302: Based on the rock mass parameters and water permeability parameters in the historical tunnel water-rich detection data, configure the corresponding parameter mapping points on the corresponding parameter axis, and connect the parameter mapping points in sequence to form a parameter mapping polyline. Step S303: Configure the parameter mapping polylines corresponding to several historical tunnel water-rich detection data on the same parameter number axis system to obtain the parameter state representation template.

3. The method for safety early warning of water-rich tunnels based on multi-source heterogeneous data according to claim 2, characterized in that, Methods for determining the generalization features of the parametric state representation model for early warning include: Step S401: Define a preset parameter range for each inflection point of the parameter mapping polyline in the parameter state representation model for early warning, and count other parameter mapping polylines that fall within the preset parameter range of each inflection point, and record them as the number of similar mapping polylines. Step S402: Calculate the ratio of the similar mapping line corresponding to each parameter mapping line to all parameter mapping lines, and record it as the similar mapping volume ratio of the parameter mapping line. If the similar mapping volume ratio is greater than or equal to a preset value, the parameter mapping line is identified as a parameter mapping line that needs attention. Step S403: Analyze the proximity features between the mapping lines of the parameters of interest, determine the range mapped by the mapping lines of the parameters of interest that meet the preset proximity criteria, and identify the range mapped by the mapping lines of the parameters of interest that meet the preset proximity criteria as the range of the generalization feature representation mapping. Methods for identifying the non-generalization features of the parametric state representation model for early warning include: S404, determine the range of the generalized feature representation mapped in the early warning parameter state representation model, determine the range mapped by the parameter mapping polyline in the safety parameter state representation template, calculate the difference range between the range mapped by the generalized feature representation and the range of the generalized feature representation, and identify the difference range as the range of the non-generalized feature representation.

4. The method for safety early warning of water-rich tunnels based on multi-source heterogeneous data according to claim 3, characterized in that, Methods for analyzing the proximity characteristics between the mapping lines of parameters of interest include: Step S4031: Perform a proximity analysis on adjacent mapping lines of parameters of interest. The proximity analysis method includes determining the mapping area mapped by the corresponding close mapping lines of each and determining the difference in mapping area. Step S4032: Determine the distance between each inflection point of adjacent mapping polylines of parameters of interest, and determine the proximity between mapping polylines of parameters of interest by combining the difference in mapping area. The expression for calculating the degree of proximity is: ； Where J represents the degree of proximity. The function for determining proximity between points is based on the distance interval between points at each turning point. Output the corresponding preset value. This is the preset maximum mapping area difference. K represents the difference in the mapped area between similar mapped lines, K is the adjustment coefficient for the influence of the difference in the mapped area, and b is the adjustment constant for the influence of the difference in the mapped area.

5. The method for safety early warning of water-rich tunnels based on multi-source heterogeneous data according to claim 3, characterized in that, The method for defining a preset parameter range for each inflection point of the parametric mapping polyline in the parametric state representation model for early warning also includes: Step S4011: Before defining the preset parameter range for each inflection point in the parameter state representation model for early warning, a portion of the parameter mapping polyline is reserved, which is denoted as the reserved parameter mapping polyline. Step S4012: Determine whether the reserved parameter mapping polyline belongs to the range of generalization feature representation mapping. If it does, retain the reserved parameter mapping polyline and remove the reserved parameter mapping polyline that does not belong. Step S4013: Compare the reserved parameter mapping polyline with the parameter mapping polyline used to construct the generalization feature representation mapping range. If there is a parameter mapping polyline where the difference between each inflection point is less than the preset value, then mark the reserved parameter mapping polyline. Step S4014: Calculate the reserved percentage of the marked reserved parameter mapping line relative to all reserved parameter mapping lines, and determine the preset adjustment coefficient for the preset parameter range based on the preset percentage range to which the marked reserved percentage belongs.

6. The method for safety early warning of water-rich tunnels based on multi-source heterogeneous data according to claim 1, characterized in that, Methods for determining whether the parametric mapping polylines recorded by the non-generalized feature representation are a good fit with the real-time parametric mapping polylines include: Step S5011: Calculate the matching degree between the parameter mapping polyline recorded by the non-generalized feature performance and the real-time parameter mapping polyline. If the matching degree is greater than or equal to the preset value, then determine that the two are compatible. The expression for calculating the degree of matching is: ； Where P represents the degree of matching. The function is used to determine the difference between the i-th turning points. Based on the preset difference interval of the turning point, the sub-matching degree of the turning point output is determined, where n is the total number of turning points.

7. A safety early warning system for water-rich tunnels based on multi-source heterogeneous data, characterized in that, The method for safety early warning of water-rich tunnels according to any one of claims 1-6 includes: The first module is used to acquire several historical tunnel water-rich detection data, including rock mass parameters, water osmotic pressure parameters, and rock mass displacement parameters. The water osmotic pressure parameters are detected by osmotic pressure sensors installed on the rock mass, and the rock mass displacement parameters are detected by displacement sensors installed on the rock mass. The second module is used to analyze several historical tunnel water-rich detection data, to mark historical tunnel water-rich detection data with rock mass displacement parameters greater than or equal to preset values ​​as early warning data, and to mark other historical tunnel water-rich detection data as safe data. The third module is used to determine the rock mass parameter type. For each type of rock mass parameter and water permeability parameter, a parameter status representation template is constructed. Historical tunnel water-rich detection data with warning and safety markers are classified and mapped onto the parameter status representation template to obtain the parameter status representation model for warning and the parameter status representation model for safety, respectively. The fourth module is used to compare the parameter state performance model for early warning and the parameter state performance model for safety, and to determine the generalization feature and non-generalization feature of the parameter state performance model for early warning. The generalization feature is the feature that the parameter state performance model for early warning has a concentrated expression relative to the parameter state performance model for safety, and the non-generalization feature is the feature that the parameter state performance model for early warning does not have a concentrated expression relative to the parameter state performance model for safety. The fifth module is used to perform fuzzy comparison of real-time tunnel water-rich detection data by utilizing the generalization features of the early warning parameter state representation model, and to perform precise comparison by utilizing the non-generalization features of the early warning parameter state representation model, thereby realizing early warning of water-rich tunnels.

Images