A Smart Decision-Making Method and System for Tunnel Collapse Accident Rescue

Abstract

This invention provides an intelligent decision-making method and system for tunnel collapse accident rescue, which can automatically generate corresponding decision-making plans at the first moment of a tunnel collapse accident. This allows rescuers to promptly set off with the necessary equipment and begin rescue operations, effectively saving rescue preparation time and increasing the survival rate of trapped personnel.

Description

Technical Field

[0001] This invention relates to the field of intelligent decision-making technology, specifically to an intelligent decision-making method and system for tunnel collapse accident rescue plans. Background Technology

[0002] Tunnel collapse accidents refer to engineering disasters that occur during tunnel construction or operation due to sudden changes in geological conditions (such as fault fracture zones, water inrush softening of surrounding rock), support failure, or external disturbances (such as earthquakes, blasting vibrations) that cause the rock and soil to become unstable and collapse, resulting in passage blockage, personnel entrapment, and equipment damage.

[0003] Because the golden rescue window for tunnel collapse accidents is short, and there are often extremely complex factors such as enclosed space, lack of information, and high risk of secondary collapse, traditional manual judgment and decision-making is time-consuming and prone to misjudgment. However, intelligent decision-making can quickly integrate multi-source heterogeneous data such as the length of the collapsed body, gas concentration, and water inflow, thereby reducing rescue time, reducing the risk of secondary disasters, and maximizing the safety of trapped personnel.

[0004] In the prior art, Chinese patent application number CN202210280885.X discloses a rescue decision-making method and system applied to rescue equipment. The method includes: acquiring downhole environmental data, then calculating downhole risk parameters from the downhole environmental data, determining the risk level based on the downhole risk parameters, determining corresponding decision information based on the risk level, and executing corresponding operations based on the decision information. However, it only performs risk prediction based on the acquired parameters and cannot provide a relatively accurate rescue decision. Summary of the Invention

[0005] In order to at least overcome the above-mentioned shortcomings in the prior art, the purpose of this application is to provide an intelligent decision-making method and system for tunnel collapse accident rescue.

[0006] In a first aspect, embodiments of this application provide an intelligent decision-making method for tunnel collapse accident rescue plans, including:

[0007] When a tunnel collapse occurs, the environmental parameters of the current collapse are obtained; the environmental parameters include the length of the collapsed body, the material of the collapsed body, the tunnel diameter, the tunnel burial depth, gas environment data, and water inrush data;

[0008] The environmental parameters are input into a preset decision model, and the decision data output by the decision model is received; the decision data includes the rescue method.

[0009] The decision data is then optimized to form the final rescue plan.

[0010] In one possible implementation, the construction of the decision model includes:

[0011] The sample environmental parameters, sample rescue methods, and rescue duration of historical rescue events are obtained, and the sample environmental parameters are quantified to form multiple sample arrays; each sample array corresponds to one historical rescue event, and the sample array includes the quantified sample environmental parameters;

[0012] A multidimensional clustering space is constructed, and the clustering constraints of the sample array are constructed based on the rescue duration; the dimension of the multidimensional clustering space is equal to the dimension of the sample array.

[0013] In the multidimensional clustering space, the sample array is semi-supervised clustered using the clustering constraints to form multiple final clusters as the decision model; each final cluster corresponds to a sample rescue method.

[0014] In one possible implementation, the clustering constraints for constructing the sample array based on the rescue duration include:

[0015] When the rescue duration exceeds 72 hours, the relationship between the corresponding sample array and the sample rescue method will be marked as negative.

[0016] When the rescue duration is less than or equal to 72 hours, a positive relationship is marked between the corresponding sample array and the sample rescue method.

[0017] In one possible implementation, semi-supervised clustering includes:

[0018] The sample arrays labeled with positive relationships are extracted as positive sample arrays, and the sample arrays labeled with negative relationships are extracted as negative sample arrays. The positive sample arrays of the same sample rescue method are grouped into an initial cluster.

[0019] Calculate the position of the center point of each initial cluster in the multidimensional clustering space and use it as the initial center point;

[0020] The negative sample array is added to the multidimensional clustering space, and the initial clustering calculation is performed with the initial center point as the center of each cluster. The negative sample array will not be clustered into the initial cluster corresponding to the sample rescue method marked as negative.

[0021] After the initial clustering calculation is completed, the centroid of each cluster is recalculated, and the positive or negative relationships of the sample arrays that are farthest from the recalculated centroid in the same cluster are removed before clustering is performed again.

[0022] Repeatedly calculate the center point of each cluster and remove the positive or negative relationship of the farthest sample array, then perform clustering until there is no positive or negative relationship between the sample array farthest from the center point in each cluster;

[0023] All clusters at this point are considered as the final cluster.

[0024] In one possible implementation, the generation of the decision data includes:

[0025] The actual environmental parameters of the tunnel collapse accident are quantified to form actual quantified parameters, and the actual Euclidean distance between the actual quantified parameters and the center point of each final cluster is calculated.

[0026] All final clusters are sorted from smallest to largest based on the actual Euclidean distance, and the rescue method corresponding to the final cluster with the highest sorting result and that meets the rigid requirements is selected as the decision data.

[0027] In one possible implementation, the rigidity requirement includes:

[0028] The vertical shaft method is not used when the tunnel burial depth is greater than 50m.

[0029] Large-diameter horizontal drilling method is not used when the length of the collapsed body is greater than 50m and / or the longitudinal slope ratio of the tunnel is greater than 3%.

[0030] When the water inflow data is greater than 0, the small pilot tunnel method is not selected and is replaced by the pipe jacking method.

[0031] In one possible implementation, quantizing the sample environment parameters and the environment parameters includes:

[0032] The length of the collapsed body is normalized within the range of [10m, 100m], with a value of 0 for lengths less than 10m and a value of 1 for lengths greater than 100m.

[0033] The average compressive strength of the collapsed body is calculated based on the proportions of loose soil, hard materials, and metallic materials in the collapsed body. The average compressive strength is then normalized within the range of [10 MPa, 90 MPa], with a value of 0 for values ​​less than 10 MPa and 1 for values ​​greater than 90 MPa. If the specific distribution of materials in the collapsed body is unknown, the strength of the surrounding rock is used as the average compressive strength. The average compressive strength is then used as the material composition of the collapsed body.

[0034] The tunnel diameter is normalized within the range of [6m, 14m], with a value of 0 for diameters less than 6m and a value of 1 for diameters greater than 14m.

[0035] The tunnel burial depth is normalized within the range of [8m, 80m], with a value of 0 for depths less than 8m and a value of 1 for depths greater than 80m.

[0036] The methane content of the gas environment data is normalized within the range of [0.3%, 1%], with 0 being less than 0.3% and 1 being greater than 1%.

[0037] The water inrush data is set in [100m] 3 / d,1000m 3 Normalization calculations are performed within the interval [ / d], and the result is less than 100m. 3 / d is taken as 0, and is greater than 1000m 3 / d takes the value 1.

[0038] In one possible implementation, optimizing the decision data to form a final rescue plan includes:

[0039] Based on the site conditions of the collapse accident, the rescue methods based on the decision data are planned and a final rescue plan is formed.

[0040] Secondly, embodiments of this application also provide an intelligent decision-making system for tunnel collapse accident rescue plans, including:

[0041] The acquisition unit is configured to acquire environmental parameters of the current tunnel collapse accident when such an accident occurs; the environmental parameters include the length of the collapsed body, the material of the collapsed body, the tunnel diameter, the tunnel burial depth, gas environment data, and water inrush data.

[0042] The decision-making unit is configured to input the environmental parameters into a preset decision-making model and receive decision data output by the decision-making model; the decision data includes rescue methods.

[0043] The optimization unit is configured to optimize the decision data to form a final rescue plan.

[0044] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0045] This invention provides an intelligent decision-making method and system for tunnel collapse accident rescue, which can automatically generate corresponding decision-making plans at the first moment of a tunnel collapse accident. This allows rescuers to promptly set off with the necessary equipment and begin rescue operations, effectively saving rescue preparation time and increasing the survival rate of trapped personnel. Attached Figure Description

[0046] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:

[0047] Figure 1 This is a schematic diagram of the method steps in an embodiment of this application;

[0048] Figure 2 This is a schematic diagram illustrating the decision model construction steps in an embodiment of this application;

[0049] Figure 3This is a schematic diagram illustrating the semi-supervised clustering construction steps in an embodiment of this application;

[0050] Figure 4 This is a schematic diagram illustrating the overall steps of an embodiment of this application. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the accompanying drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0052] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0053] Please refer to the following: Figure 1 and Figure 4 This is a flowchart illustrating an intelligent decision-making method for a tunnel collapse accident rescue plan provided in an embodiment of the present invention. Further, the intelligent decision-making method for a tunnel collapse accident rescue plan may specifically include the contents described in steps S1-S3.

[0054] S1: When a tunnel collapse occurs, obtain the environmental parameters of the current collapse; the environmental parameters include the length of the collapsed body, the material of the collapsed body, the tunnel diameter, the tunnel burial depth, gas environment data, and water inrush data;

[0055] S2: Input the environmental parameters into a preset decision model and receive the decision data output by the decision model; the decision data includes the rescue method;

[0056] S3: Optimize the decision data to form the final rescue plan.

[0057] In the implementation of this application embodiment, when a tunnel collapse occurs, environmental parameters can be obtained through on-site detection combined with construction operation reports. The length and material of the collapsed body can be obtained using advanced ground-penetrating radar or similar detection methods, while tunnel diameter, tunnel depth, gas environment data, and water inflow data can be obtained from data acquired during on-site construction. Generally, gas environment data refers to methane content data.

[0058] In this embodiment, a decision-making model needs to be pre-constructed. This model needs to be able to select a specific rescue method under known environmental parameters, and it also needs to consider the rigid requirements of different rescue methods. Generally, the rescue methods described in this embodiment include vertical shaft method, pipe jacking method, small shield tunneling method, large-diameter horizontal drilling method, small pilot tunnel method, and dredging of central drainage ditch method. After the decision-making model selects a rescue method, rescuers can begin preparing the corresponding materials, equipment, personnel, and vehicles. At the same time, the rescue method can be refined and optimized to generate a final rescue plan. This embodiment can automatically generate the corresponding decision plan at the first moment of a tunnel collapse accident, allowing rescuers to promptly set off with the appropriate equipment and begin rescue, thereby effectively saving rescue preparation time and increasing the survival rate of trapped personnel.

[0059] In one possible implementation, the construction of the decision model includes:

[0060] The sample environmental parameters, sample rescue methods, and rescue duration of historical rescue events are obtained, and the sample environmental parameters are quantified to form multiple sample arrays; each sample array corresponds to one historical rescue event, and the sample array includes the quantified sample environmental parameters;

[0061] A multidimensional clustering space is constructed, and the clustering constraints of the sample array are constructed based on the rescue duration; the dimension of the multidimensional clustering space is equal to the dimension of the sample array.

[0062] In the multidimensional clustering space, the sample array is semi-supervised clustered using the clustering constraints to form multiple final clusters as the decision model; each final cluster corresponds to a sample rescue method.

[0063] When implementing the embodiments of this application, please refer to Figure 2Historical rescue events can be collected by combining domestic and international data. For example, relevant international data can be retrieved from the MIT Tunnel Accident Database and the eMARS database, while domestic data can be obtained by requesting access to the Ministry of Emergency Management's Production Safety Accident Statistical Information Direct Reporting System. Since the number of historical rescue events related to tunnel collapses obtained through the above methods is relatively small, only around 600, it is difficult to use algorithms such as neural network models for analysis. Therefore, in this embodiment, a semi-supervised clustering method is used to classify the environmental parameters of accident-related samples.

[0064] In this embodiment, since a clustering analysis algorithm is used, the environmental parameters of the samples need to be quantified and normalized; the normalized data constitutes the sample array. During clustering, the dimension of the clustering space needs to be the same as the dimension of the sample array. For example, if the sample array contains normalized data on the length of the collapsed body, the material of the collapsed body, the tunnel diameter, the tunnel depth, gas environment data, and water inrush data—a total of six dimensions—then a six-dimensional clustering space is required. In this embodiment, the clustering constraint refers to the constraint on the rescue duration. Generally, the golden rescue time is 72 hours. In this example, rescues completed within 72 hours are considered positive samples, while those exceeding 72 hours are considered negative samples, thus constructing the clustering constraint. After performing semi-supervised clustering with constraints, multiple classifications, or final clusters, can be formed. These final clusters have a one-to-one correspondence with the sample rescue methods. Specifically, when one of the following rescue methods is used: vertical shaft method, pipe jacking method, small shield tunneling method, large-diameter horizontal borehole method, small pilot tunnel method, or dredging the central drainage ditch method, both the number of final clusters and the number of sample rescue methods are six. The classified final clusters can be used to make decisions regarding the selection of rescue methods.

[0065] In one possible implementation, the clustering constraints for constructing the sample array based on the rescue duration include:

[0066] When the rescue duration exceeds 72 hours, the relationship between the corresponding sample array and the sample rescue method will be marked as negative.

[0067] When the rescue duration is less than or equal to 72 hours, a positive relationship is marked between the corresponding sample array and the sample rescue method.

[0068] In the implementation of this application embodiment, when the rescue duration is less than or equal to 72 hours, the historical rescue event can be considered a successful rescue, and the sample array and the corresponding sample rescue method can be marked as a positive relationship and classified into the positive feedback set M; when the rescue duration is greater than 72 hours, the historical rescue event can be considered a failed rescue, and the sample array and the corresponding sample rescue method can be marked as a negative relationship and classified into the negative feedback set C.

[0069] In one possible implementation, semi-supervised clustering includes:

[0070] The sample arrays labeled with positive relationships are extracted as positive sample arrays, and the sample arrays labeled with negative relationships are extracted as negative sample arrays. The positive sample arrays of the same sample rescue method are grouped into an initial cluster.

[0071] Calculate the position of the center point of each initial cluster in the multidimensional clustering space and use it as the initial center point;

[0072] The negative sample array is added to the multidimensional clustering space, and the initial clustering calculation is performed with the initial center point as the center of each cluster. The negative sample array will not be clustered into the initial cluster corresponding to the sample rescue method marked as negative.

[0073] After the initial clustering calculation is completed, the centroid of each cluster is recalculated, and the positive or negative relationships of the sample arrays that are farthest from the recalculated centroid in the same cluster are removed before clustering is performed again.

[0074] Repeatedly calculate the center point of each cluster and remove the positive or negative relationship of the farthest sample array, then perform clustering until there is no positive or negative relationship between the sample array farthest from the center point in each cluster;

[0075] All clusters at this point are considered as the final cluster.

[0076] When implementing the embodiments of this application, please refer to Figure 3 Positive sample arrays are placed into the positive feedback set M, and negative sample arrays are placed into the negative feedback set C, at which point clustering initialization is performed. Since each positive sample array actually corresponds to a sample rescue method, positive sample arrays with the same sample rescue method can be grouped into the same class, i.e., the initial cluster. Unlike ordinary clustering analysis, the clustering initialization stage in this embodiment already has fixed categories, and some data groups have already been classified.

[0077] In this embodiment, to classify the subsequently added negative sample array, it is necessary to first determine the initial cluster center point as the initial center point, which can generally be the centroid. Then, the negative sample array is added to the multidimensional clustering space, and clustering is performed. The objective function for clustering is as follows:

[0078]

[0079] In the formula, It is the classification matrix of the i-th array in the k-th cluster, where n is the number of arrays and K is the number of clusters. It is the i-th array To the center point of the kth cluster The distance between them is generally expressed using Euclidean distance; M is the set of positive feedback, and C is the set of negative feedback. It is the positive feedback penalty coefficient. It is the negative feedback penalty coefficient, which is usually a large number, such as 100; To constrain the weights, in this embodiment of the application, since hard constraints are used, the weight is set to 1; function This is an indicator function that takes the value 1 when the condition in the function is true and 0 when the condition in the function is false. This is the classification matrix of the cluster to which the rescue method of the sample corresponding to the i-th array belongs. When performing clustering classification using the above objective function, the objective requirement is to minimize the objective function value. In this case, if the arrays in M ​​or C do not meet the corresponding requirements, a very large penalty value will be obtained, thus avoiding this situation.

[0080] In this embodiment, after the initial clustering calculation is completed, the positive sample array is assigned to the cluster corresponding to the rescue method, while the negative sample array is assigned to the cluster of non-corresponding rescue methods. Due to the specific nature of on-site rescue, the rescue method corresponding to the positive sample array may not be the optimal solution, and the negative relationship in the negative sample array may also be due to other factors on-site. Therefore, in this embodiment, extracting the sample array furthest from the recalculated center point in each cluster, removing the positive or negative relationship, and then re-clustering can assign the corresponding array to a more reasonable cluster. After iterative calculation, the furthest sample array will no longer have a positive or negative relationship. This indicates that the furthest sample array itself has already been classified as a free array without constraints, and therefore can serve as a reasonable classification boundary. At this point, all clusters can be used as the final clusters for rescue method screening.

[0081] In one possible implementation, the generation of the decision data includes:

[0082] The actual environmental parameters of the tunnel collapse accident are quantified to form actual quantified parameters, and the actual Euclidean distance between the actual quantified parameters and the center point of each final cluster is calculated.

[0083] All final clusters are sorted from smallest to largest based on the actual Euclidean distance, and the rescue method corresponding to the final cluster with the highest sorting result and that meets the rigid requirements is selected as the decision data.

[0084] In implementing this application's embodiments, the actual environmental parameters also need to be quantified, and the quantification process should be consistent with the sample environmental parameters. Furthermore, the final selected rescue methods also need to consider rigid requirements; that is, each rescue method must meet certain conditions to be selected. By sorting by Euclidean distance and screening based on rigid requirements, the optimal rescue method can be chosen.

[0085] In one possible implementation, the rigidity requirement includes:

[0086] The vertical shaft method is not used when the tunnel burial depth is greater than 50m.

[0087] Large-diameter horizontal drilling method is not used when the length of the collapsed body is greater than 50m and / or the longitudinal slope ratio of the tunnel is greater than 3%.

[0088] When the water inflow data is greater than 0, the small pilot tunnel method is not selected and is replaced by the pipe jacking method.

[0089] In one possible implementation, quantizing the sample environment parameters and the environment parameters includes:

[0090] The length of the collapsed body is normalized within the range of [10m, 100m], with a value of 0 for lengths less than 10m and a value of 1 for lengths greater than 100m.

[0091] The average compressive strength of the collapsed body is calculated based on the proportions of loose soil, hard materials, and metallic materials in the collapsed body. The average compressive strength is then normalized within the range of [10 MPa, 90 MPa], with a value of 0 for values ​​less than 10 MPa and 1 for values ​​greater than 90 MPa. If the specific distribution of materials in the collapsed body is unknown, the strength of the surrounding rock is used as the average compressive strength. The average compressive strength is then used as the material composition of the collapsed body.

[0092] The tunnel diameter is normalized within the range of [6m, 14m], with a value of 0 for diameters less than 6m and a value of 1 for diameters greater than 14m.

[0093] The tunnel burial depth is normalized within the range of [8m, 80m], with a value of 0 for depths less than 8m and a value of 1 for depths greater than 80m.

[0094] The methane content of the gas environment data is normalized within the range of [0.3%, 1%], with 0 being less than 0.3% and 1 being greater than 1%.

[0095] The water inrush data is set in [100m] 3 / d,1000m 3 Normalization calculations are performed within the interval [ / d], and the result is less than 100m. 3 / d is taken as 0, and is greater than 1000m 3 / d takes the value 1.

[0096] In implementing this application's embodiments, the sample environmental parameters and environmental parameters need to be quantified, including normalization. The principle of normalization is that if the minimum value in the interval before normalization is lower than this value, it will not have a significant impact on the selection of the rescue method; similarly, if the maximum value is higher than this value, it will not have a significant impact on the selection of the rescue method. At the same time, the completeness of the sample environmental parameters and environmental parameters also needs to be considered.

[0097] For the length of the collapsed body, the excavation and tunneling difficulty of collapsed bodies less than 10m can be kept at the same level, while collapsed bodies exceeding 100m are all at the level of extremely difficult to handle, with similar processing difficulty.

[0098] For the materials of the collapsed body, the concept of average compressive strength is used. Measurements within the collapsed body itself rely on on-site advanced exploration, so this data is often the most severely lacking. When this data is known, the average compressive strength is calculated based on the proportion of different materials and then normalized. A lower limit of 10 MPa is used because soil below 10 MPa tends to be loose soil, which can be excavated using relatively simple methods, while soil above 90 MPa contains a large amount of metallic material, requiring the same cutting and other methods of treatment. For the missing materials of the collapsed body, the strength of the surrounding rock, which is always known, is directly used.

[0099] Tunnel depth generally affects shaft excavation and tunnel ground stress. Other technologies besides shaft operation are not very sensitive to tunnel depth. Strong support is required when the depth is greater than 80m, while simpler support is often sufficient when the depth is less than 8m.

[0100] The tunnel diameter generally affects the stability of the collapsed body. Therefore, the upper limit is selected as 14m, which is larger for a two-lane highway tunnel, while the lower limit is selected as 6m, which is common for shield tunnels. When the diameter is less than 6m, the stability of the collapsed body will tend to be the same.

[0101] Gas environment data generally uses methane content data, which examines the gas situation in the tunnel. According to the specifications, if it is below 0.3%, no corresponding treatment is required, only monitoring is needed; however, if it is above 1%, ventilation and other methods are required to reduce the methane content in advance. Therefore, if it is above 1%, it needs to be reduced to below 1% after corresponding treatment.

[0102] Water inflow data above 1000m 3 When the collapse reaches / d, a large number of pumps need to be deployed for treatment. At this point, the collapsed body has already exhibited strong fluid plasticity, so the treatment methods are relatively similar; while for collapses below 100 m 3 When the time is / d, a single pump can be used for disposal, which has little impact on the rescue method.

[0103] In one possible implementation, optimizing the decision data to form a final rescue plan includes:

[0104] Based on the site conditions of the collapse accident, the rescue methods based on the decision data are planned and a final rescue plan is formed.

[0105] When implementing the embodiments of this application, after the rescue method is determined, the rescuers can set off with the preliminary equipment and personnel for that rescue method. At this time, it is necessary to specify the rescue method, such as determining the excavation location, ventilation and drainage arrangements, etc., which are relatively mature existing technologies, and the embodiments of this application will not impose any limitations.

[0106] Based on the same inventive concept, embodiments of this application also provide an intelligent decision-making system for tunnel collapse accident rescue plans, including:

[0107] The acquisition unit is configured to acquire environmental parameters of the current tunnel collapse accident when such an accident occurs; the environmental parameters include the length of the collapsed body, the material of the collapsed body, the tunnel diameter, the tunnel burial depth, gas environment data, and water inrush data.

[0108] The decision-making unit is configured to input the environmental parameters into a preset decision-making model and receive decision data output by the decision-making model; the decision data includes rescue methods.

[0109] The optimization unit is configured to optimize the decision data to form a final rescue plan.

[0110] In one possible implementation, the construction of the decision model includes:

[0111] The sample environmental parameters, sample rescue methods, and rescue duration of historical rescue events are obtained, and the sample environmental parameters are quantified to form multiple sample arrays; each sample array corresponds to one historical rescue event, and the sample array includes the quantified sample environmental parameters;

[0112] A multidimensional clustering space is constructed, and the clustering constraints of the sample array are constructed based on the rescue duration; the dimension of the multidimensional clustering space is equal to the dimension of the sample array.

[0113] In the multidimensional clustering space, the sample array is semi-supervised clustered using the clustering constraints to form multiple final clusters as the decision model; each final cluster corresponds to a sample rescue method.

[0114] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0115] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices or units, or may be electrical, mechanical or other forms of connection.

[0116] The units described as separate components may or may not be physically separate. As will be apparent to those skilled in the art, the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0117] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0118] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or grid device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0119] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An intelligent decision-making method for tunnel collapse accident rescue plans, characterized in that, include: When a tunnel collapse occurs, obtain the environmental parameters of the current collapse. The environmental parameters include the length of the collapsed body, the material of the collapsed body, the tunnel diameter, the tunnel burial depth, gas environment data, and water inrush data; The environmental parameters are input into a preset decision model, and the decision data output by the decision model is received; the decision data includes the rescue method. The construction of the decision-making model includes: The sample environmental parameters, sample rescue methods, and rescue duration of historical rescue events are obtained, and the sample environmental parameters are quantified to form multiple sample arrays; each sample array corresponds to one historical rescue event, and the sample array includes the quantified sample environmental parameters; A multidimensional clustering space is constructed, and the clustering constraints of the sample array are constructed based on the rescue duration; the dimension of the multidimensional clustering space is equal to the dimension of the sample array. In the multidimensional clustering space, the sample array is semi-supervised clustered using the clustering constraints to form multiple final clusters as the decision model; each final cluster corresponds to a sample rescue method. Semi-supervised clustering includes: The sample arrays labeled with positive relationships are extracted as positive sample arrays, and the sample arrays labeled with negative relationships are extracted as negative sample arrays. The positive sample arrays of the same sample rescue method are grouped into an initial cluster. Calculate the position of the center point of each initial cluster in the multidimensional clustering space and use it as the initial center point; The negative sample array is added to the multidimensional clustering space, and the initial clustering calculation is performed with the initial center point as the center of each cluster. The negative sample array will not be clustered into the initial cluster corresponding to the sample rescue method marked as negative. After the initial clustering calculation is completed, the centroid of each cluster is recalculated, and the positive or negative relationships of the sample arrays that are farthest from the recalculated centroid in the same cluster are removed before clustering is performed again. Repeatedly calculate the center point of each cluster and remove the positive or negative relationship of the farthest sample array, then perform clustering until there is no positive or negative relationship between the sample array farthest from the center point in each cluster; All clusters at this point are taken as the final cluster; The decision data is then optimized to form the final rescue plan.

2. The intelligent decision-making method for tunnel collapse accident rescue plan according to claim 1, characterized in that, The clustering constraints for constructing the sample array based on the rescue duration include: When the rescue duration exceeds 72 hours, the relationship between the corresponding sample array and the sample rescue method will be marked as negative. When the rescue duration is less than or equal to 72 hours, a positive relationship is marked between the corresponding sample array and the sample rescue method.

3. The intelligent decision-making method for tunnel collapse accident rescue plan according to claim 1, characterized in that, The generation of the decision data includes: The actual environmental parameters of the tunnel collapse accident are quantified to form actual quantified parameters, and the actual Euclidean distance between the actual quantified parameters and the center point of each final cluster is calculated. All final clusters are sorted from smallest to largest based on the actual Euclidean distance, and the rescue method corresponding to the final cluster with the highest sorting result and that meets the rigid requirements is selected as the decision data.

4. The intelligent decision-making method for tunnel collapse accident rescue plan according to claim 3, characterized in that, The rigid requirements include: The vertical shaft method is not used when the tunnel burial depth is greater than 50m. Large-diameter horizontal drilling method is not used when the length of the collapsed body is greater than 50m and / or the longitudinal slope ratio of the tunnel is greater than 3%. When the water inflow data is greater than 0, the small pilot tunnel method is not selected and is replaced by the pipe jacking method.

5. The intelligent decision-making method for tunnel collapse accident rescue plan according to claim 3, characterized in that, The quantification of the sample environmental parameters and the environmental parameters includes: The length of the collapsed body is normalized within the range of [10m, 100m], with a value of 0 for lengths less than 10m and a value of 1 for lengths greater than 100m. The average compressive strength of the collapsed body is calculated based on the proportions of loose soil, hard materials, and metallic materials in the collapsed body. The average compressive strength is then normalized within the range of [10 MPa, 90 MPa], with a value of 0 for values ​​less than 10 MPa and 1 for values ​​greater than 90 MPa. If the specific distribution of materials in the collapsed body is unknown, the strength of the surrounding rock is used as the average compressive strength. The average compressive strength is then used as the material composition of the collapsed body. The tunnel diameter is normalized within the range of [6m, 14m], with a value of 0 for diameters less than 6m and a value of 1 for diameters greater than 14m. The tunnel burial depth is normalized within the range of [8m, 80m], with a value of 0 for depths less than 8m and a value of 1 for depths greater than 80m. The methane content of the gas environment data is normalized within the range of [0.3%, 1%], with 0 being less than 0.3% and 1 being greater than 1%. The water inrush data is set in [100m] 3 / d,1000m 3 Normalization calculations are performed within the interval [ / d], and the result is less than 100m. 3 / d is taken as 0, and is greater than 1000m 3 / d takes the value 1.

6. The intelligent decision-making method for tunnel collapse accident rescue plan according to claim 1, characterized in that, Optimizing the decision data to form the final rescue plan includes: Based on the site conditions of the collapse accident, the rescue methods based on the decision data are planned and a final rescue plan is formed.

7. An intelligent decision-making system for tunnel collapse accident rescue, used to execute the method described in any one of claims 1 to 6, characterized in that, include: The acquisition unit is configured to acquire environmental parameters of the current tunnel collapse when a tunnel collapse occurs. The environmental parameters include the length of the collapsed body, the material of the collapsed body, the tunnel diameter, the tunnel burial depth, gas environment data, and water inrush data; The decision-making unit is configured to input the environmental parameters into a preset decision-making model and receive decision data output by the decision-making model; the decision data includes rescue methods. The optimization unit is configured to optimize the decision data to form a final rescue plan.

Images