Method and system for monitoring reliability of tunnel soft rock large deformation support

Abstract

The invention provides a tunnel soft rock large deformation support reliability monitoring method and system, and the method comprises the steps: obtaining and analyzing a plurality of pieces of support structure data of tunnel soft rock, extracting a first deformation feature vector, combining a monitoring intention and a data label, intelligently selecting an appropriate deformation feature conversion strategy, and carrying out the monitoring of the tunnel soft rock large deformation support reliability. And converting the first deformation feature vector into a more representative second deformation feature vector. The method not only realizes accurate capture and conversion of the deformation characteristics of the support structure, but also generates a reliability prediction result for the deformation support monitoring request by performing reliability index decision on the converted deformation characteristics, thereby effectively improving the accuracy and efficiency of tunnel soft rock large deformation support reliability monitoring, and improving the reliability of tunnel soft rock large deformation support monitoring. Therefore, safety construction and operation and maintenance management of tunnel engineering are facilitated.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and more particularly, to a method and system for monitoring the reliability of support for large deformations of soft rock in tunnels. Background Art

[0002] In tunnel engineering, the stability of the support structure in soft rock strata is a key factor to ensure the safe construction and operation of tunnels. Due to the characteristics of soft rock such as easy deformation and low strength, the support structure often faces the risk of large deformations during the excavation process and later operation of the tunnel. Therefore, it is particularly important to monitor the reliability of the support for large deformations of soft rock in tunnels.

[0003] Traditional methods for monitoring the reliability of tunnel support mainly rely on empirical judgment, on-site observation, and simple data analysis. These methods have limitations in dealing with complex and variable soft rock support problems. On the one hand, empirical judgment is greatly affected by subjective factors and is difficult to accurately reflect the actual state of the support structure. On the other hand, on-site observation and simple data analysis can often provide only limited information and are difficult to comprehensively and deeply reveal the deformation characteristics and reliability status of the support structure.

[0004] With the continuous development of information technology, data-driven intelligent monitoring methods have gradually been applied to the monitoring of tunnel support reliability. However, most of the existing intelligent monitoring methods focus on data collection and processing, and lack flexibility and pertinence in the extraction and conversion of deformation characteristics. Different support structures may exhibit different characteristics during the deformation process, and the existing methods often adopt a unified strategy for feature extraction and conversion, making it difficult to adapt to the particularity of different support structures. Summary of the Invention

[0005] In view of the problems mentioned above, in combination with the first aspect of the present invention, embodiments of the present invention provide a method for monitoring the reliability of support for large deformations of soft rock in tunnels, the method comprising: Obtaining a plurality of support structure data of soft rock in a tunnel corresponding to a deformation support monitoring request, and extracting first deformation feature vectors respectively corresponding to the plurality of support structure data, the plurality of support structure data corresponding to a plurality of structural manifestation forms; Obtaining at least two reference deformation feature conversion strategies, each reference deformation feature conversion strategy including at least one step of feature selection and feature focusing on the first deformation feature vector; Determining a target deformation feature conversion strategy from the at least two reference deformation feature conversion strategies according to the monitoring intention corresponding to the deformation support monitoring request and the data label corresponding to the support structure data, the data label representing the reference label knowledge point corresponding to the support structure data under the deformation support monitoring request; Use the target deformation feature conversion strategy to perform deformation feature conversion on the multiple first deformation feature vectors to generate multiple second deformation feature vectors; Perform reliability index decision-making on the multiple second deformation feature vectors to generate reliability index decision-making data corresponding to the deformation support monitoring request, where the reliability index decision-making data represents a reliability prediction result corresponding to the deformation support monitoring request generated by using the multiple support structure data.

[0006] In a possible implementation manner of the first aspect, the determining the target deformation feature conversion strategy from the at least two reference deformation feature conversion strategies according to the monitoring intention corresponding to the deformation support monitoring request and the data label corresponding to the support structure data includes: Loading the monitoring intention corresponding to the deformation support monitoring request, the multiple first deformation feature vectors, and the at least two reference deformation feature conversion strategies into an artificial intelligence network that has completed knowledge learning in advance to obtain the target deformation feature conversion strategy. The artificial intelligence network is used to perform strategy matching on the at least two reference deformation feature conversion strategies according to the monitoring intention corresponding to the deformation support monitoring request and the multiple first deformation feature vectors. The target deformation feature conversion strategy includes at least one of a target feature selection strategy and a target feature focusing strategy.

[0007] In a possible implementation manner of the first aspect, the artificial intelligence network includes a target matching module and a decision execution module; The loading the monitoring intention corresponding to the deformation support monitoring request, the multiple first deformation feature vectors, and the at least two reference deformation feature conversion strategies into an artificial intelligence network that has completed knowledge learning in advance to obtain the target deformation feature conversion strategy includes: Loading the monitoring intention corresponding to the deformation support monitoring request, the multiple first deformation feature vectors, and the at least two reference deformation feature conversion strategies into the target matching module to obtain the matching confidence levels corresponding to the at least two reference deformation feature conversion strategies respectively. The target matching module is used to perform the strategy matching on the at least two reference deformation feature conversion strategies according to the monitoring intention corresponding to the deformation support monitoring request and the multiple first deformation feature vectors; Determining the target feature selection strategy according to the matching confidence levels corresponding to the at least two reference deformation feature conversion strategies respectively; Load the monitoring intention corresponding to the deformed support monitoring request, multiple first deformation feature vectors, and the matching confidence levels corresponding to the at least two reference deformation feature conversion strategies into the decision execution module to obtain the target feature focusing strategy. The decision execution module is used to make a decision based on the multiple first deformation feature vectors and the matching confidence levels corresponding to the at least two reference deformation feature conversion strategies to obtain the target feature focusing strategy that matches the monitoring intention.

[0008] In a possible implementation manner of the first aspect, the matching confidence levels corresponding to the at least two reference deformation feature conversion strategies include the first matching confidence levels corresponding to multiple feature selection strategies; The determining the target feature selection strategy according to the matching confidence levels corresponding to the at least two reference deformation feature conversion strategies includes: Output the feature selection strategy with the largest numerical value of the first matching confidence level from the first matching confidence levels corresponding to the multiple feature selection strategies as the target feature selection strategy.

[0009] In a possible implementation manner of the first aspect, the matching confidence levels corresponding to the at least two reference deformation feature conversion strategies include the second matching confidence levels corresponding to multiple feature focusing strategies; The loading the monitoring intention corresponding to the deformed support monitoring request, multiple first deformation feature vectors, and the matching confidence levels corresponding to the at least two reference deformation feature conversion strategies into the decision execution module to obtain the target feature focusing strategy includes: Load the monitoring intention corresponding to the deformed support monitoring request, multiple first deformation feature vectors, and the second matching confidence levels corresponding to the multiple feature focusing strategies into the decision execution module to obtain the target feature focusing strategy.

[0010] In a possible implementation manner of the first aspect, before loading the monitoring intention corresponding to the deformed support monitoring request, multiple first deformation feature vectors, and the at least two reference deformation feature conversion strategies into the target matching module to obtain the matching confidence levels corresponding to the at least two reference deformation feature conversion strategies, it further includes: Obtain multiple template support structure data corresponding to the template deformed support monitoring request, and extract the first deformation feature vectors corresponding to the multiple template support structure data. The multiple template support structure data correspond to multiple structural manifestation forms; Obtain at least two template deformation feature conversion strategies, and each template deformation feature conversion strategy includes at least one of a template feature selection strategy and a template feature focusing strategy; Load the monitoring intention corresponding to the template deformation support monitoring request, multiple first deformation feature vectors, and the at least two template deformation feature conversion strategies into the template matching module to obtain the matching confidence levels corresponding to the at least two template deformation feature conversion strategies respectively; Load the monitoring intention corresponding to the template deformation support monitoring request, multiple first deformation feature vectors, and the matching confidence levels corresponding to the at least two template deformation feature conversion strategies into the decision execution module to generate reliability index decision data for the feature focusing strategy; Generate a target incentive mechanism network based on the reliability index decision data of the feature focusing strategy; Optimize the template matching module according to the target incentive mechanism network to generate the target matching module.

[0011] In a possible implementation manner of the first aspect, the using the target deformation feature conversion strategy to perform deformation feature conversion on the multiple first deformation feature vectors to generate multiple second deformation feature vectors includes: Use the target feature selection strategy to perform correlation selection on the multiple first deformation feature vectors to generate multiple first sub-deformation feature vectors that meet the correlation requirements, where the correlation requirements represent the correlation between the first deformation feature vector and the monitoring intention; Use the target feature focusing strategy to perform feature focusing on the multiple first sub-deformation feature vectors to generate the multiple second deformation feature vectors.

[0012] In a possible implementation manner of the first aspect, before obtaining the at least two reference deformation feature conversion strategies, it further includes: Load the multiple first deformation feature vectors into a feature processing network that has completed prior knowledge learning to generate multiple extended deformation feature vectors, where the feature space fields corresponding to the extended deformation feature vectors are more than the feature space fields corresponding to the first deformation feature vectors, and the feature processing network is used to extract the extended deformation feature vectors corresponding to the first deformation feature vectors.

[0013] In a possible implementation manner of the first aspect, after extracting the first deformation feature vectors corresponding to the multiple support structure data respectively, the method further includes: Obtain the vector attributes corresponding to the multiple first deformation feature vectors respectively, where the vector attributes represent the feature space fields corresponding to the first deformation feature vectors; Fuse the first deformation feature vectors with the same vector attributes among the first deformation feature vectors corresponding to the multiple support structure data respectively to generate a target deformation feature vector.

[0014] In another aspect, an embodiment of the present invention further provides a reliability monitoring system for tunnel soft rock large deformation support, including a processor and a machine-readable storage medium. The machine-readable storage medium is connected to the processor. The machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0015] Based on the above aspects, the embodiment of the present application obtains and analyzes multiple support structure data of tunnel soft rock, extracts its first deformation feature vector, and combines the monitoring intention and data tags to intelligently select an applicable deformation feature conversion strategy to convert the first deformation feature vector into a more representative second deformation feature vector. This method not only realizes the accurate capture and conversion of the deformation characteristics of the support structure, but also generates a reliability prediction result for the deformation support monitoring request by making a reliability index decision on the converted deformation characteristics, effectively improving the accuracy and efficiency of the reliability monitoring of tunnel soft rock large deformation support, thus contributing to the safe construction and operation and maintenance management of tunnel engineering. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a schematic flowchart of the execution of the reliability monitoring method for tunnel soft rock large deformation support provided by an embodiment of the present invention.

[0017] Figure 2 is a schematic diagram of the hardware architecture of the reliability monitoring system for tunnel soft rock large deformation support provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The present invention will be specifically described below with reference to the accompanying drawings of the specification. Figure 1 is a schematic flowchart of the reliability monitoring method for tunnel soft rock large deformation support provided by an embodiment of the present invention. The reliability monitoring method for tunnel soft rock large deformation support will be introduced in detail below.

[0019] Step S110, obtain multiple support structure data of tunnel soft rock corresponding to the deformation support monitoring request, and extract the first deformation feature vectors respectively corresponding to the multiple support structure data. The multiple support structure data correspond to multiple structural manifestation forms.

[0020] In this embodiment, in tunnel engineering, in order to ensure the stability and safety of the soft rock part of the tunnel, it is often necessary to monitor the support structure. For example, in a mountain tunnel project under construction, due to the complex soft rock geological conditions and large differences in soft rock characteristics in different sections, a variety of support structures are set at different positions of the tunnel. These support structures include anchor bolts, shotcrete layers, steel arch frames, etc.

[0021] When a deformed support monitoring request is received, multiple support structure data of the tunnel soft rock corresponding to the deformed support monitoring request can be obtained. For a support structure such as a bolt, data such as its length, diameter, anchoring depth, and spacing can be obtained. Taking the length data as an example, assume that in a certain tunnel section, the designed length of the bolt is 3 meters, but the actually measured length may change due to the deformation of the soft rock, and this actually measured length is part of the support structure data. For the shotcrete layer, data such as its thickness, strength grade, and spraying angle can be obtained. The thickness data may be obtained by a special measuring instrument, such as an ultrasonic thickness gauge, and its measurement result is the support structure data of the shotcrete layer. The data of the steel arch includes the steel type of the arch, the spacing of the arch, and the radian of the arch.

[0022] After obtaining these support structure data, the first deformation feature vectors corresponding to each support structure data can be further extracted. Taking the length data of the bolt as an example, the corresponding first deformation feature vector may include information such as the change rate of the length and the difference in the length change from adjacent bolts. Assume that within a certain period of time, the initial length of the bolt is 3 meters, and after a period of time, the measured length becomes 3.05 meters. Then the length change rate is (3.05 - 3) / 3 = 0.0167, and this 0.0167 is an element of the length change rate feature in the first deformation feature vector. For the thickness data of the shotcrete layer, if the initial designed thickness is 15 cm and it is measured to be 14.5 cm after measurement, data such as the thickness change of -0.5 cm and the change rate of -0.5 / 15 constitute part of the elements of the first deformation feature vector corresponding to the thickness data of the shotcrete layer. The first deformation feature vectors corresponding to these different support structure data each have different features and meanings due to the various structural manifestation forms of the support structure, and they are the basis for subsequent analysis and decision-making.

[0023] Step S120, obtain at least two reference deformation feature conversion strategies, and each reference deformation feature conversion strategy includes at least one of the steps of feature selection and feature focusing on the first deformation feature vector.

[0024] In the process of monitoring and analyzing the support structure of tunnel soft rock, in order to better process and analyze the first deformation feature vector, multiple reference deformation feature conversion strategies need to be obtained. For example, one reference deformation feature conversion strategy may focus on feature selection. Assume that under this strategy, for an element in the first deformation feature vector of the bolt length change rate, if its fluctuation range in the past multiple monitoring data is very small and its correlation with the overall deformation of the tunnel is low, then this strategy may choose to exclude it from the subsequent analysis, and this is the process of feature selection.

[0025] Another reference deformation feature conversion strategy may focus more on feature focusing. For example, for the first deformation feature vector of the shotcrete layer thickness, since the shotcrete layer plays a key role in resisting the lateral pressure of soft rock in tunnel support, this reference deformation feature conversion strategy may focus on the feature of the thickness change rate. For example, through a specific algorithm, such as a weighted algorithm, the weight of the thickness change rate feature can be increased in subsequent analysis, so as to more prominently highlight the importance of this feature in judging the deformation of the support structure.

[0026] Another reference deformation feature conversion strategy may include both feature selection and feature focusing. For example, for the element of the first deformation feature vector of the spacing of the steel arch, first, through the analysis of historical data, it is found that when the arch spacing changes within a certain specific range, the impact on the deformation of the tunnel soft rock is not significant. So, first, feature selection is carried out to screen and exclude part of the data corresponding to this change range. Then, for the remaining data related to the arch spacing, the method of feature focusing is adopted. According to the correlation between the arch spacing and the settlement of the soft rock at the top of the tunnel, different degrees of weights are given, so that in subsequent analysis, the impact of the change of the arch spacing on the overall deformation of the tunnel soft rock support structure can be more accurately reflected. These different reference deformation feature conversion strategies are all constructed based on an in-depth understanding of the characteristics and deformation mechanisms of the tunnel soft rock support structure, providing multiple options for determining the most suitable target deformation feature conversion strategy in the future.

[0027] Step S130, determine the target deformation feature conversion strategy from the at least two reference deformation feature conversion strategies according to the monitoring intention corresponding to the deformation support monitoring request and the data label corresponding to the support structure data, where the data label represents the reference label knowledge point corresponding to the support structure data under the deformation support monitoring request.

[0028] In the actual scenario of tunnel soft rock support structure monitoring, the monitoring intention of the deformation support monitoring request may be to focus on the settlement of the soft rock at the top of the tunnel to prevent collapse accidents. And the data label corresponding to the support structure data contains the reference label knowledge points of each support structure data under this monitoring intention. For example, for the bolt support structure data, its data label may indicate that the bolt mainly plays a role in controlling the settlement of the soft rock at the top of the tunnel through the anchoring force, and the anchoring force is related to data such as the bolt length, diameter, and anchoring depth.

[0029] Suppose there are two reference deformation feature conversion strategies, Strategy A and Strategy B. In terms of feature selection, Strategy A focuses on the two first deformation feature vector elements, namely the anchorage depth and diameter of the rock bolt. In terms of feature focusing, a higher weight is given to the rate of change of the thickness of the shotcrete layer. In feature selection, Strategy B selects several first deformation feature vector elements, including the length and spacing of the rock bolts and the thickness of the shotcrete layer. In terms of feature focusing, a higher weight is given to the change in the curvature of the steel arch.

[0030] Since the monitoring intention is to focus on the settlement of the soft rock at the top of the tunnel, according to the data labels of the support structure data, it can be known that the anchorage depth of the rock bolt is more critical for controlling the settlement of the soft rock at the top, and the change in the thickness of the shotcrete layer also has an important impact on the settlement of the soft rock at the top. Strategy A focuses on the anchorage depth of the rock bolt in feature selection and gives a higher weight to the rate of change of the thickness of the shotcrete layer in feature focusing. In contrast, although Strategy B also involves some relevant elements, its matching degree with the monitoring intention is not as good as that of Strategy A. Therefore, according to the monitoring intention and data labels, Strategy A is finally determined as the target deformation feature conversion strategy. This process is achieved by comprehensively considering the key points concerned by the monitoring intention and the importance of each element in the support structure data under this monitoring intention, ensuring that the selected target deformation feature conversion strategy can most effectively process the first deformation feature vector to meet the requirements of deformation support monitoring.

[0031] Step S140, use the target deformation feature conversion strategy to perform deformation feature conversion on multiple first deformation feature vectors to generate multiple second deformation feature vectors.

[0032] After determining the target deformation feature conversion strategy, deformation feature conversion can be performed on multiple first deformation feature vectors. For example, the target deformation feature conversion strategy is Strategy A mentioned above. For the first deformation feature vector of the rock bolt support structure, since Strategy A focuses on the two elements of the anchorage depth and diameter of the rock bolt in feature selection, then in the process of deformation feature conversion, other first deformation feature vector elements of the rock bolt will be screened first, and only the data related to the anchorage depth and diameter will be retained. Suppose the original first deformation feature vector of the rock bolt anchorage depth contains depth change data at multiple time points, such as [0.02 m (1st week), 0.03 m (2nd week), 0.025 m (3rd week)], and the original first deformation feature vector of the diameter contains [0.002 m (1st week), 0.0015 m (2nd week), 0.0018 m (3rd week)].

[0033] Then, according to the feature focusing part in Strategy A, a higher weight is given to the thickness change rate of the shotcrete layer. Suppose the first deformation feature vector of the original thickness change rate of the shotcrete layer is [0.05 (week 1), 0.06 (week 2), 0.04 (week 3)]. During the feature focusing process, through a specific weight calculation method, such as multiplying by a coefficient greater than 1, it is converted to [0.1 (week 1), 0.12 (week 2), 0.08 (week 3)].

[0034] After such feature selection and feature focusing operations, the processed data of the anchoring depth and diameter of the bolt and the thickness change rate of the shotcrete layer constitute the new second deformation feature vector. For the first deformation feature vectors of other support structures, similar conversion operations are also performed according to the rules in the target deformation feature conversion strategy, so as to obtain multiple second deformation feature vectors. These second deformation feature vectors are optimized and adjusted feature vectors, which can better reflect the key information related to the deformation of tunnel soft rock and provide more effective data support for subsequent reliability index decision-making.

[0035] Step S150, make a reliability index decision on the multiple second deformation feature vectors to generate the reliability index decision data corresponding to the deformation support monitoring request, and the reliability index decision data represents the reliability prediction result corresponding to the deformation support monitoring request generated by using the multiple support structure data.

[0036] After obtaining multiple second deformation feature vectors, the reliability index decision begins. For example, in the tunnel soft rock support structure, for the bolt support part, its second deformation feature vector contains the processed anchoring depth and diameter data. If the change in the anchoring depth is always within a reasonable range and the change in the diameter does not exceed the allowable fluctuation range of the design, this indicates that the reliability of the bolt support is relatively high at this stage. Suppose according to historical experience and engineering standards, the reasonable change range of the anchoring depth is ±0.05 meters, and the reasonable fluctuation range of the diameter is ±0.002 meters. When the change value of the anchoring depth in the second deformation feature vector is within this range and the change value of the diameter is also within the corresponding range, a relatively high reliability score, such as 80 points (out of 100) can be given to the bolt support part.

[0037] For the shotcrete layer, if the data of the thickness change rate in its second deformation feature vector after feature focusing indicates that the thickness change rate is within an acceptable range and there is no trend of sharp change, this also shows that the support reliability of the shotcrete layer is relatively high. Suppose the acceptable range of the thickness change rate is ±0.1. When the actual thickness change rate is within this range, a reliability score of 70 points is given to the support part of the shotcrete layer.

[0038] By performing such reliability analysis on the second deformation feature vectors corresponding to each support structure and comprehensively considering the importance of all support structures in the tunnel soft rock support, such as the importance weight of the bolt being 0.4 and the importance weight of the shotcrete layer being 0.3, and after other support structures such as steel arches also perform reliability analysis according to their respective second deformation feature vectors and corresponding standards and determine the weights, finally, the reliability index decision data of the entire tunnel soft rock support structure under the deformation support monitoring request can be calculated. For example, the calculated reliability index decision data is 75 points, and this score represents the reliability prediction result corresponding to the deformation support monitoring request generated using the data of multiple support structures. This result can provide intuitive information for tunnel engineering construction and maintenance personnel so that they can take timely measures to ensure that the support structures in the soft rock part of the tunnel are always in a reliable working state and guarantee the safety of the tunnel.

[0039] Based on the above steps, the embodiment of the present application obtains and analyzes the data of multiple support structures of tunnel soft rock, extracts their first deformation feature vectors, and combines the monitoring intention and data tags to intelligently select an applicable deformation feature conversion strategy to convert the first deformation feature vectors into more representative second deformation feature vectors. This method not only realizes the accurate capture and conversion of the deformation features of the support structure, but also generates a reliability prediction result for the deformation support monitoring request by making a reliability index decision on the converted deformation features, effectively improving the accuracy and efficiency of the reliability monitoring of the large deformation support of tunnel soft rock, thereby contributing to the safe construction and operation and maintenance management of tunnel engineering.

[0040] In a possible implementation manner, step S130 may include: Loading the monitoring intention corresponding to the deformation support monitoring request, multiple first deformation feature vectors, and the at least two reference deformation feature conversion strategies into an artificial intelligence network that has completed knowledge learning in advance to obtain the target deformation feature conversion strategy. The artificial intelligence network is used to perform strategy matching on the at least two reference deformation feature conversion strategies according to the monitoring intention corresponding to the deformation support monitoring request and the multiple first deformation feature vectors. The target deformation feature conversion strategy includes at least one of a target feature selection strategy and a target feature focusing strategy.

[0041] In a possible implementation manner, the artificial intelligence network includes a target matching module and a decision execution module.

[0042] The loading the monitoring intention corresponding to the deformation support monitoring request, multiple first deformation feature vectors, and the at least two reference deformation feature conversion strategies into an artificial intelligence network that has completed knowledge learning in advance to obtain the target deformation feature conversion strategy includes: Step S131: Load the monitoring intention corresponding to the deformed support monitoring request, multiple first deformation feature vectors, and the at least two reference deformation feature conversion strategies into the target matching module to obtain the matching confidence levels respectively corresponding to the at least two reference deformation feature conversion strategies. The target matching module is used to perform strategy matching on the at least two reference deformation feature conversion strategies according to the monitoring intention corresponding to the deformed support monitoring request and the multiple first deformation feature vectors.

[0043] Step S132: Determine the target feature selection strategy according to the matching confidence levels respectively corresponding to the at least two reference deformation feature conversion strategies.

[0044] Step S133: Load the monitoring intention corresponding to the deformed support monitoring request, multiple first deformation feature vectors, and the matching confidence levels respectively corresponding to the at least two reference deformation feature conversion strategies into the decision execution module to obtain the target feature focusing strategy. The decision execution module is used to make a decision based on the multiple first deformation feature vectors and the matching confidence levels respectively corresponding to the at least two reference deformation feature conversion strategies to obtain the target feature focusing strategy that matches the monitoring intention.

[0045] In this embodiment, taking a specific tunnel project as an example, the monitoring intention of the deformed support monitoring request is to accurately evaluate the deformation of the support structure of the tunnel soft rock under the action of lateral pressure to ensure the safety inside the tunnel and the stability of the structure. At this time, the multiple first deformation feature vectors that have been obtained contain the relevant deformation information of different support structures, such as the axial strain vector of the bolt, the surface crack propagation vector of the shotcrete layer, and the bending degree change vector of the steel arch, etc. At the same time, there are multiple reference deformation feature conversion strategies to choose from, and each reference deformation feature conversion strategy includes different ways of processing the first deformation feature vectors, including operations such as feature selection and feature focusing.

[0046] First, load the monitoring intention corresponding to the deformed support monitoring request, multiple first deformation feature vectors, and at least two reference deformation feature conversion strategies into the target matching module. The target matching module performs strategy matching on the at least two reference deformation feature conversion strategies according to the monitoring intention corresponding to the deformed support monitoring request and the multiple first deformation feature vectors, so as to obtain the matching confidence levels respectively corresponding to the at least two reference deformation feature conversion strategies.

[0047] Suppose there are two reference deformation feature conversion strategies in this tunnel project, namely Strategy 1 and Strategy 2. For Strategy 1, in terms of feature selection, it tends to select two first deformation feature vector elements, namely the axial strain of the bolt and the surface crack propagation of the shotcrete layer. In terms of feature focusing, it focuses on the change trend of the axial strain. For Strategy 2, it pays more attention to the change in the curvature of the steel arch and the surface crack propagation of the shotcrete layer in feature selection, and focuses on the propagation speed of the surface cracks in feature focusing.

[0048] When the target matching module performs strategy matching, it comprehensively considers the monitoring intention and the actual situation of each first deformation feature vector. Since the monitoring intention is to evaluate the deformation of the support structure of the tunnel soft rock under lateral pressure, from the perspective of multiple first deformation feature vectors, the axial strain of the bolt is of great significance for reflecting the deformation of the support structure under lateral pressure, and the surface crack propagation of the shotcrete layer is also a key indication under lateral pressure. The target matching module evaluates Strategy 1 and Strategy 2 respectively through internal algorithms and learned knowledge, and obtains their respective matching confidence levels. Suppose the matching confidence level of Strategy 1 is 0.7 and that of Strategy 2 is 0.5.

[0049] Determine the target feature selection strategy based on the matching confidence levels corresponding to at least two reference deformation feature conversion strategies. In the above example, since the matching confidence level of Strategy 1, 0.7, is higher than that of Strategy 2, 0.5, the feature selection part in Strategy 1 is determined as the target feature selection strategy, that is, select the two first deformation feature vector elements, namely the axial strain of the bolt and the surface crack propagation of the shotcrete layer, as the content of the target feature selection strategy.

[0050] Next, load the monitoring intention corresponding to the deformed support monitoring request, multiple first deformation feature vectors, and the matching confidence levels corresponding to at least two reference deformation feature conversion strategies into the decision execution module. The decision execution module makes a decision based on multiple first deformation feature vectors and the matching confidence levels corresponding to at least two reference deformation feature conversion strategies to obtain the target feature focusing strategy that matches the monitoring intention.

[0051] Continuing with this example, after the decision execution module receives relevant information, since the first deformation feature vector elements of interest in the target feature selection strategy have been determined, it will further determine the target feature focusing strategy based on these elements and the matching confidence. In this process, the decision execution module will deeply analyze the relationship between the monitoring intention and these elements. For example, because the monitoring intention is to evaluate the deformation under lateral pressure, and the axial strain change trend of the anchor bolt occupies an important position in reflecting this deformation, the surface crack propagation speed of the shotcrete layer is also crucial for judging the deformation under lateral pressure. Based on the matching confidence of Strategy 1 and the importance of these elements, the decision execution module determines that in the target feature focusing strategy, specific weights and focusing methods are given to the axial strain change trend of the anchor bolt and the surface crack propagation speed of the shotcrete layer to better match the monitoring intention.

[0052] Thus, by using the target matching module and the decision execution module in the artificial intelligence network, based on the monitoring intention corresponding to the deformation support monitoring request and the data label corresponding to the support structure data, a target deformation feature conversion strategy including at least one of the target feature selection strategy and the target feature focusing strategy is determined from at least two reference deformation feature conversion strategies. This target deformation feature conversion strategy can process the first deformation feature vector more accurately, thereby providing an effective basis for subsequent tunnel soft rock support structure deformation analysis and decision-making.

[0053] In this process, the knowledge learning of the artificial intelligence network is crucial, and it is learned based on a large amount of tunnel soft rock support structure data, different monitoring intentions, and corresponding deformation feature conversion strategies. For example, in the learning process of the network, historical data of many tunnel projects may be used. These data include various first deformation feature vectors and corresponding effective deformation feature conversion strategies under different geological conditions and different support structure designs. By learning these massive data, the artificial intelligence network can accurately match the reference deformation feature conversion strategies in different monitoring intention and first deformation feature vector situations, obtain reasonable matching confidences, and determine the most suitable target deformation feature conversion strategy.

[0054] Meanwhile, the internal algorithms and logics of the target matching module and the decision execution module are also carefully designed. When calculating the matching confidence, the target matching module will consider various factors, such as the correlation strength between each element in the first deformation feature vector and the monitoring intention, and the adaptability of the feature selection and feature focusing operations in different reference deformation feature conversion strategies to the monitoring intention. When determining the target feature focusing strategy, the decision execution module not only needs to consider the matching confidence, but also combine the actual physical meaning of the first deformation feature vector and the core requirements of the monitoring intention to ensure that the determined target feature focusing strategy can best meet the accurate assessment requirements of the deformation situation of the tunnel soft rock support structure.

[0055] In a possible implementation manner, the matching confidences respectively corresponding to the at least two reference deformation feature conversion strategies include first matching confidences respectively corresponding to multiple feature selection strategies.

[0056] Step S132 may include: outputting, as the target feature selection strategy, the feature selection strategy with the largest numerical value of the first matching confidence from the first matching confidences respectively corresponding to the multiple feature selection strategies.

[0057] In a possible implementation manner, the matching confidences respectively corresponding to the at least two reference deformation feature conversion strategies include second matching confidences respectively corresponding to multiple feature focusing strategies.

[0058] Step S133 may include: loading the monitoring intention corresponding to the deformed support monitoring request, multiple first deformation feature vectors, and the second matching confidences respectively corresponding to the multiple feature focusing strategies into the decision execution module to obtain the target feature focusing strategy.

[0059] In this embodiment, taking a certain large tunnel project as an example, the tunnel passes through a complex soft rock formation. To ensure the safety and stability of the tunnel, the deformation monitoring of the support structure is crucial. During this process, there are multiple reference deformation feature conversion strategies, each strategy includes different feature selection strategies and feature focusing strategies, and the matching confidence corresponding to each reference deformation feature conversion strategy will be calculated, including first matching confidences respectively corresponding to multiple feature selection strategies and second matching confidences respectively corresponding to multiple feature focusing strategies.

[0060] For the first matching confidences respectively corresponding to the multiple feature selection strategies in the matching confidences respectively corresponding to the at least two reference deformation feature conversion strategies, when determining the target feature selection strategy based on these matching confidences, it is necessary to output, as the target feature selection strategy, the feature selection strategy with the largest numerical value of the first matching confidence from the first matching confidences respectively corresponding to the multiple feature selection strategies.

[0061] Suppose there are three reference deformation feature conversion strategies in this tunnel project, namely Strategy A, Strategy B, and Strategy C. Each strategy has different focuses in the feature selection strategy part. The feature selection strategy of Strategy A focuses on three first deformation feature vector elements: the anchorage depth of the bolt, the thickness of the shotcrete layer, and the spacing of the steel arch. The feature selection strategy of Strategy B focuses on several first deformation feature vector elements such as the axial strain of the bolt, the compressive strength of the shotcrete layer, and the steel type of the steel arch. The feature selection strategy of Strategy C mainly focuses on several first deformation feature vector elements such as the anchorage angle of the bolt, the elastic modulus of the shotcrete layer, and the arch radian of the steel arch.

[0062] When calculating the first batch of configuration confidence levels corresponding to these feature selection strategies, the target matching module will conduct a comprehensive evaluation based on the monitoring intention corresponding to the deformation support monitoring request and multiple first deformation feature vectors. Suppose the monitoring intention is to accurately judge the stability of the support structure of the tunnel soft rock under vertical pressure. For Strategy A, since the anchorage depth of the bolt is directly related to resisting vertical pressure, the thickness of the shotcrete layer also reflects the bearing capacity for vertical pressure to a certain extent, and the spacing of the steel arch is related to the overall stability of the support structure. After complex calculations and evaluations, the first batch of configuration confidence level corresponding to the feature selection strategy of Strategy A is 0.7. For Strategy B, although the axial strain of the bolt and the compressive strength of the shotcrete layer are related to vertical pressure to some extent, compared with Strategy A, their direct relevance to the monitoring intention is slightly lower. After evaluation, its first batch of configuration confidence level is 0.5. For Strategy C, the anchorage angle of the bolt, the elastic modulus of the shotcrete layer, and the arch radian of the steel arch have relatively weak relevance in reflecting the stability of the support structure under vertical pressure, and its first batch of configuration confidence level is 0.4.

[0063] According to the rules, since the first batch of configuration confidence level of Strategy A, which is 0.7, is the largest among these three strategies, the feature selection strategy in Strategy A is determined as the target feature selection strategy, that is, the three first deformation feature vector elements: the anchorage depth of the bolt, the thickness of the shotcrete layer, and the spacing of the steel arch are selected as the key focus objects for subsequent operations.

[0064] In the case where the configuration confidence level also includes the second batch of configuration confidence levels corresponding to multiple feature focusing strategies, the process of loading the monitoring intention corresponding to the deformation support monitoring request, multiple first deformation feature vectors, and the configuration confidence levels corresponding to at least two reference deformation feature conversion strategies into the decision execution module to obtain the target feature focusing strategy is as follows.

[0065] Continuing with the above tunnel project as an example, assume that the feature focusing strategies in each reference deformation feature conversion strategy also vary. The feature focusing strategy of Strategy A assigns a higher weight to the rate of change of the thickness of the shotcrete layer, aiming to highlight the importance of thickness changes to the stability of the support structure under vertical pressure; the feature focusing strategy of Strategy B focuses on the strength change coefficient corresponding to the steel type of the steel arch, because the strength change of the steel type has a greater impact on the overall performance of the support structure; the feature focusing strategy of Strategy C focuses on the change trend of the anchoring angle of the anchor bolts, believing that the change trend of the anchoring angle has special significance for the stability of the support structure under vertical pressure.

[0066] When calculating the second batch of configuration confidence levels corresponding to these feature focusing strategies, it is also based on multiple factors such as the monitoring intention corresponding to the deformed support monitoring request and multiple first deformation feature vectors. For Strategy A, since the rate of change of the thickness of the shotcrete layer has a high correlation in reflecting the stability of the support structure under vertical pressure, after detailed analysis and calculation, the second batch of configuration confidence level corresponding to its feature focusing strategy is 0.6. For Strategy B, although the strength change coefficient corresponding to the steel type of the steel arch has an impact on the performance of the support structure, its importance in reflecting the stability under vertical pressure is slightly lower than that of the rate of change of the thickness of the shotcrete layer, and its second batch of configuration confidence level is 0.5. For Strategy C, the importance of the change trend of the anchoring angle of the anchor bolts in reflecting the stability of the support structure under vertical pressure is relatively lower, and its second batch of configuration confidence level is 0.4.

[0067] Then, load the monitoring intention corresponding to the deformed support monitoring request (i.e., accurately judge the stability of the support structure of the tunnel soft rock under vertical pressure), multiple first deformation feature vectors (including relevant vector elements such as the anchoring depth of the anchor bolts, the thickness of the shotcrete layer, and the spacing of the steel arches), and the second batch of configuration confidence levels corresponding to multiple feature focusing strategies (Strategy A is 0.6, Strategy B is 0.5, and Strategy C is 0.4) into the decision execution module. Based on this input information, the decision execution module deeply analyzes the relationship between the monitoring intention and each feature focusing strategy. Since the second batch of configuration confidence level of Strategy A, 0.6, is the highest, and the rate of change of the thickness of the shotcrete layer is of great significance in monitoring the stability of the support structure under vertical pressure, the decision execution module determines that the target feature focusing strategy is to assign a higher weight to the rate of change of the thickness of the shotcrete layer to better match the monitoring intention.

[0068] Throughout the process, the work of the target matching module and the decision execution module highly depends on a large amount of engineering data and precise algorithms. When calculating the first configuration confidence and the second configuration confidence, the target matching module needs to consider numerous factors. For the calculation of the first configuration confidence, not only the direct correlation between the selected first deformation feature vector elements in the feature selection strategy and the monitoring intention needs to be considered, but also the interaction relationship among these elements in the entire support structure system. For example, although the anchoring depth of the bolt is directly related to the resistance to vertical pressure, there is a synergistic relationship between it and the thickness of the shotcrete layer and the spacing of the steel arch. This synergistic relationship also needs to be accurately considered when calculating the first configuration confidence.

[0069] For the calculation of the second configuration confidence, the decision execution module needs to consider the sensitivity and importance degree of the elements concerned in the feature focusing strategy in reflecting the monitoring intention. For example, the change rate of the thickness of the shotcrete layer has different sensitivities to reflecting the stability of the support structure under vertical pressure in different engineering environments and support structure states. This requires precise analysis and evaluation based on a large amount of historical data and engineering experience to obtain the accurate second configuration confidence.

[0070] When determining the target feature focusing strategy, the decision execution module, in addition to relying on the level of the second configuration confidence, also needs to comprehensively consider the overall situation of the first deformation feature vector and the core requirements of the monitoring intention. For example, although the second configuration confidence of strategy A is the highest, the decision execution module also needs to consider whether weighting the change rate of the thickness of the shotcrete layer can most optimally reflect the stability of the support structure under vertical pressure on the basis that the target feature selection strategy has already determined to focus on elements such as the anchoring depth of the bolt, the thickness of the shotcrete layer, and the spacing of the steel arch. This requires in-depth understanding and precise grasp of the mechanical properties, deformation mechanism of the entire support structure, and the change law of the monitoring data.

[0071] Thus, the most suitable strategy content can be accurately selected from numerous reference deformation feature conversion strategies according to the specific monitoring intention, thereby improving the accuracy and effectiveness of the deformation monitoring of the support structure and providing reliable technical support for ensuring the safety and stability of the tunnel project.

[0072] In a possible implementation manner, before step S131, the method further includes: Step A110, obtaining multiple template support structure data corresponding to the template deformation support monitoring request, and extracting the first deformation feature vectors respectively corresponding to the multiple template support structure data, where the multiple template support structure data correspond to multiple structural manifestation forms.

[0073] Step A120: Obtain at least two template deformation feature conversion strategies, where each template deformation feature conversion strategy includes at least one of a template feature selection strategy and a template feature focusing strategy.

[0074] Step A130: Load the monitoring intention corresponding to the template deformation support monitoring request, multiple first deformation feature vectors, and the at least two template deformation feature conversion strategies into the template matching module to obtain the matching confidence levels corresponding to the at least two template deformation feature conversion strategies respectively.

[0075] Step A140: Load the monitoring intention corresponding to the template deformation support monitoring request, multiple first deformation feature vectors, and the matching confidence levels corresponding to the at least two template deformation feature conversion strategies respectively into the decision execution module to generate reliability index decision data for the feature focusing strategy.

[0076] Step A150: Generate a target incentive mechanism network based on the reliability index decision data of the feature focusing strategy.

[0077] Step A160: Optimize the template matching module according to the target incentive mechanism network to generate the target matching module.

[0078] In this embodiment, taking a typical tunnel project as an example, the template deformation support monitoring request is set to evaluate the stability of the support structure of soft rock in a specific area of the tunnel under various complex working conditions. For this request, the corresponding template support structure data covers various types. For example, the relevant data of the bolt includes the depth values of its different anchorage sections, the axial stress values at different construction stages, etc.; the data of the shotcrete layer contains the thickness measurement values at different positions, the compressive strength values under different curing times, etc.; the data of the steel arch includes the arch radius at different parts, the elastic modulus of the steel, and the node force values at the arch connections. These data reflect various structural performance forms of different support structures in the tunnel soft rock environment. From these template support structure data, according to the established feature extraction algorithm, the corresponding first deformation feature vectors are extracted respectively. For the anchorage depth data of the bolt, its first deformation feature vector may include the change rate of the depth, the change trend of the depth difference from the adjacent bolt, etc.; the first deformation feature vector corresponding to the shotcrete layer thickness data may include the reduction rate of the thickness, the amplitude of the thickness fluctuation within a specific time period, etc.

[0079] Next, the template feature selection strategy of the first template deformation feature conversion strategy focuses on the two first deformation feature vector elements, namely the axial stress of the bolt and the compressive strength of the shotcrete layer. In terms of the template feature focusing strategy, a relatively high weight is given to the change rate of the compressive strength of the shotcrete layer because in soft rock tunnels, the change in the compressive strength of the shotcrete layer has a greater impact on the stability of the overall support structure. For the second template deformation feature conversion strategy, its template feature selection strategy focuses on the force on the nodes of the steel arch and the anchoring depth of the bolts, and the template feature focusing strategy focuses on the change trend of the force on the nodes of the steel arch because the force on the nodes of the steel arch can intuitively reflect the local stability of the support structure.

[0080] Then, the template matching module calculates according to the preset algorithms and models. For the first template deformation feature conversion strategy, since the monitoring intention is to evaluate the stability of the support structure under various complex working conditions, the axial stress of the bolt and the compressive strength of the shotcrete layer selected in the template feature selection strategy are strongly related to this intention. When calculating the matching confidence, the module will consider factors such as the influence weight of these elements on the stability of the support structure under different working conditions and the correlation in historical data. Suppose after detailed calculation, the matching confidence of the first template deformation feature conversion strategy is 0.7. For the second template deformation feature conversion strategy, although the force on the nodes of the steel arch and the anchoring depth of the bolts are also important factors, their overall matching degree with the monitoring intention is slightly lower, and after evaluation, its matching confidence is 0.5.

[0081] After that, the decision execution module conducts in-depth analysis based on the input information. For the first template deformation feature conversion strategy, due to its relatively high matching confidence and the relatively high weight given to the change rate of the compressive strength of the shotcrete layer in the template feature focusing strategy, the decision execution module will evaluate the reliability of this feature focusing strategy in accurately reflecting the stability of the support structure under different working conditions according to a large amount of engineering data and empirical models. For example, by analyzing the relationship between the change rate of the compressive strength of the shotcrete layer and the actual stability of the support structure under different geological conditions and different construction progress, a specific reliability index value is obtained. Suppose after calculation, the reliability index decision data of the feature focusing strategy corresponding to the first template deformation feature conversion strategy is 0.8. For the second template deformation feature conversion strategy, analysis and calculation are carried out according to the same principle, and the reliability index decision data of its feature focusing strategy is obtained as 0.6.

[0082] Furthermore, the construction of the target incentive mechanism network is based on the reliability evaluation of the feature focusing strategy for different template deformation feature transformation strategies. Taking the previous results as an example, the decision data of the reliability index of template deformation feature transformation strategy 1 is 0.8, and that of strategy 2 is 0.6. The target incentive mechanism network will set different incentive weights according to these data. For the template deformation feature transformation strategy 1 with higher reliability, a larger incentive weight will be given in the target incentive mechanism network, which means that in the subsequent optimization process, the feature selection and feature focusing operations related to strategy 1 will be more emphasized.

[0083] Finally, in the optimization process, the incentive weights of the target incentive mechanism network will affect the algorithm parameters and matching logic inside the template matching module. For the matching factors related to the template deformation feature transformation strategy 1, since it has a larger incentive weight in the target incentive mechanism network, the template matching module will adjust the internal algorithm, so that in the subsequent actual deformation support monitoring requests, it is more inclined to identify a reference deformation feature transformation strategy similar to the template deformation feature transformation strategy 1 and give a higher matching confidence. For example, if there is a reference deformation feature transformation strategy in the actual deformation support monitoring request that is similar to the template deformation feature transformation strategy 1 in terms of feature selection and feature focusing, then the optimized template matching module will give it a relatively higher matching confidence, thereby improving the accuracy and rationality of the entire monitoring system when selecting the target deformation feature transformation strategy.

[0084] By processing the template-related data and strategies, an optimized target matching module is finally generated, which lays a solid foundation for accurately obtaining the matching confidence of the reference deformation feature transformation strategy, thereby ensuring that in the monitoring of the tunnel soft rock support structure, the appropriate deformation feature transformation strategy can be more accurately selected according to the monitoring intention, and improving the effectiveness and reliability of the deformation evaluation and monitoring of the support structure.

[0085] In a possible implementation manner, step S140 includes: Step S141, performing correlation selection on the multiple first deformation feature vectors by using the target feature selection strategy to generate multiple first sub-deformation feature vectors that meet the correlation requirements, where the correlation requirements represent the correlation between the first deformation feature vector and the monitoring intention.

[0086] Step S142, performing feature focusing on the multiple first sub-deformation feature vectors by using the target feature focusing strategy to generate the multiple second deformation feature vectors.

[0087] In this embodiment, taking the previously mentioned tunnel project as an example, the monitoring intention of the deformation support monitoring request is to accurately evaluate the deformation of the support structure of the tunnel soft rock under the action of lateral pressure to ensure the safety inside the tunnel and the stability of the structure. Through the previous steps, the target deformation feature conversion strategy is determined, which includes the target feature selection strategy and the target feature focusing strategy.

[0088] First, use the target feature selection strategy to perform correlation selection on multiple first deformation feature vectors to generate multiple first sub-deformation feature vectors that meet the correlation requirements. Here, the correlation requirements characterize the correlation between the first deformation feature vector and the monitoring intention. For the support structure in this tunnel project, the first deformation feature vector contains information in multiple aspects, such as the axial strain vector of the bolt, the surface crack propagation vector of the shotcrete layer, and the bending degree change vector of the steel arch.

[0089] The target feature selection strategy focuses on the elements of the first deformation feature vector that are closely related to the monitoring intention. Under the monitoring intention of judging the influence of lateral pressure on the deformation of the support structure, when the bolt resists lateral pressure, its axial strain is a key factor. The change in axial strain can directly reflect the magnitude of the lateral force borne by the bolt and the deformation of the bolt itself. Therefore, the axial strain vector has a high correlation with the monitoring intention. For the shotcrete layer, the propagation direction and speed of its surface cracks are greatly affected by lateral pressure, and the crack propagation vector can reflect the stress state and structural integrity of the shotcrete layer under lateral pressure. So the surface crack propagation vector is also closely related to the monitoring intention. For the steel arch, its bending degree change directly reflects the structural deformation under the action of lateral pressure, and the bending degree change vector is also closely related to the monitoring intention.

[0090] Through the target feature selection strategy, screen the original multiple first deformation feature vectors, remove the elements with low correlation with the monitoring intention, and retain the elements that meet the correlation requirements, such as the axial strain vector of the bolt, the surface crack propagation vector of the shotcrete layer, and the bending degree change vector of the steel arch, so as to generate multiple first sub-deformation feature vectors. This process is based on an in-depth understanding of the mechanical behavior of the tunnel soft rock support structure under lateral pressure and an accurate grasp of the internal connection between each first deformation feature vector and the monitoring intention.

[0091] Next, use the target feature focusing strategy to focus on the features of multiple first sub-deformation feature vectors to generate multiple second deformation feature vectors. The target feature focusing strategy aims to further highlight the importance of certain key information in the first sub-deformation feature vector to better reflect the monitoring intention.

[0092] For the axial strain vector of the bolt, it is assumed that in the target feature focusing strategy, based on engineering experience and data analysis, it is found that the change trend of the axial strain within a specific time period is more critical for judging the stability of the support structure. For example, during the process of the lateral pressure of the soft rock in the tunnel gradually increasing, if the axial strain of the bolt shows a sharp change within a short period of time, this may indicate that the support structure is about to face the risk of instability. Therefore, the target feature focusing strategy will focus on this change trend in the axial strain vector, and through specific algorithms, such as weighted algorithms or data transformation, enhance the influence of this change trend in the entire axial strain vector.

[0093] For the surface crack propagation vector of the shotcrete layer, the target feature focusing strategy may focus on the acceleration of crack propagation. Because under the action of lateral pressure, the acceleration of crack propagation can more sensitively reflect the structural change rate of the shotcrete layer. If the acceleration of crack propagation suddenly increases, it means that the structural integrity of the shotcrete layer is more seriously threatened. Therefore, the target feature focusing strategy will focus on the acceleration element in the crack propagation vector, and by adjusting the data processing method, the importance of this element in the subsequent analysis is enhanced.

[0094] For the bending change vector of the steel arch, the target feature focusing strategy may focus on the bending change in specific key parts. For example, at the connection part between the steel arch and the tunnel wall or the crown part, the bending change in these parts has a greater impact on the stability of the entire support structure. The target feature focusing strategy will use specific data processing means, such as amplifying or weighting the bending change data of these key parts, so that the bending change in these parts has an increased weight in the entire bending change vector.

[0095] After the target feature focusing strategy performs feature focusing processing on multiple first sub-deformation feature vectors, the obtained result is multiple second deformation feature vectors. These second deformation feature vectors have undergone double optimization of correlation selection and feature focusing, and more accurately reflect the deformation situation of the support structure of the tunnel soft rock under the action of lateral pressure, providing a more targeted and effective data basis for subsequent reliability index decision-making.

[0096] This conversion process from the first deformation feature vector to the second deformation feature vector is an important link in the monitoring and analysis of the tunnel soft rock support structure, which can effectively extract the key information in the original data, provide more accurate and valuable data support for the entire monitoring system, thereby improving the accuracy and reliability of the deformation monitoring of the tunnel soft rock support structure and ensuring the safety and stability of the tunnel project.

[0097] In a possible implementation manner, before obtaining at least two reference deformation feature conversion strategies, it further includes: Load the multiple first deformation feature vectors into a feature processing network that has completed knowledge learning beforehand to generate multiple extended deformation feature vectors. The feature space fields corresponding to the extended deformation feature vectors are more than those corresponding to the first deformation feature vectors. The feature processing network is used to extract the extended deformation feature vectors corresponding to the first deformation feature vectors.

[0098] In this embodiment, in the monitoring and analysis process of the tunnel soft rock support structure, there are some important pre - operations before obtaining at least two reference deformation feature conversion strategies. These operations help to enrich data features and optimize data structures so as to more effectively determine the deformation feature conversion strategy subsequently. Also, after extracting the first deformation feature vectors corresponding to multiple support structure data respectively, there are corresponding operations to further process these vectors to improve the usability and representativeness of the data.

[0099] First, before obtaining at least two reference deformation feature conversion strategies, load the multiple first deformation feature vectors into a feature processing network that has completed knowledge learning beforehand to generate multiple extended deformation feature vectors. Taking the previously mentioned tunnel project as an example, in this project, for the monitoring of the tunnel soft rock support structure, multiple first deformation feature vectors have been obtained, and these vectors contain information such as the axial strain vector of the anchor bolt, the surface crack propagation vector of the shotcrete layer, and the bending degree change vector of the steel arch.

[0100] The feature processing network is a network that has undergone a large amount of data learning and algorithm optimization. It is designed to extract the extended deformation feature vectors corresponding to the first deformation feature vectors. When these first deformation feature vectors are loaded into the feature processing network, the network processes them based on the knowledge and algorithms it has learned beforehand. For example, for the axial strain vector of the anchor bolt, the feature processing network may consider factors such as the stress - strain relationship between the axial strain and the surrounding soft rock, the influence of the anchor bolt installation angle on the axial strain, and the variation law of the axial strain at different time scales. Thus, on the basis of the original axial strain vector, more feature information related to the axial strain is added to generate the extended deformation feature vector. These newly added feature information may include the interaction term between the elastic modulus of the soft rock around the anchor bolt and the axial strain, the correction value of the axial strain in different seasons (considering the influence of factors such as temperature), etc.

[0101] The eigen - space field corresponding to the extended deformation eigen - vector has more fields than the eigen - space field corresponding to the first deformation eigen - vector. This means that the extended deformation eigen - vector contains richer information. For the surface crack - propagation vector of the shotcrete layer, the feature - processing network may introduce feature fields related to the influence of factors such as the concrete mix ratio, curing conditions, and environmental humidity on crack propagation. For example, under different concrete mix ratios, the rate and pattern of crack propagation will be different. By incorporating these factors into the extended deformation eigen - vector, the surface crack - propagation situation of the shotcrete layer can be described more comprehensively. Similarly, for the bending - degree change vector of the steel arch, the feature - processing network may add feature fields related to factors such as the corrosion degree of the steel arch, the fatigue characteristics of the steel, and the uneven distribution of the lateral pressure on the arch, thus obtaining a more complex and information - rich extended deformation eigen - vector.

[0102] In a possible implementation manner, after extracting the first deformation eigen - vectors corresponding to the multiple support - structure data respectively, the method further includes: Obtaining the vector attributes corresponding to the multiple first deformation eigen - vectors respectively, where the vector attributes characterize the eigen - space fields corresponding to the first deformation eigen - vectors.

[0103] Fusing the first deformation eigen - vectors with the same vector attributes among the first deformation eigen - vectors corresponding to the multiple support - structure data respectively to generate a target deformation eigen - vector.

[0104] In this embodiment, after extracting the first deformation eigen - vectors corresponding to the multiple support - structure data respectively, some subsequent operations are also required. First, obtain the vector attributes corresponding to the multiple first deformation eigen - vectors respectively. The vector attributes characterize the eigen - space fields corresponding to the first deformation eigen - vectors. Continuing with the tunnel - engineering example, for the axial - strain vector of the bolt, its vector attributes may include the measurement direction of the strain (such as the longitudinal direction of the axis), the accuracy range of strain measurement (for example, accurate to 0.001 mm / m), the time resolution of the strain data (for example, measured once per hour), etc. These vector attributes clarify the specific description of the axial - strain vector in the feature space and determine its corresponding eigen - space field.

[0105] For the surface crack - propagation vector of the shotcrete layer, its vector attributes may include the plane direction of crack propagation (such as the vertical direction or the horizontal direction), the accuracy of crack - width measurement (for example, accurate to 0.1 mm), the spatial resolution of crack - propagation data (for example, the statistics of the number of cracks per square meter), etc. The vector attributes of the bending - degree change vector of the steel arch may include the reference point for bending - degree measurement (such as the crown or the springing), the error range of bending - degree measurement (for example, the error does not exceed 0.5 degrees), the update frequency of bending - degree data (for example, updated once a day), etc.

[0106] After obtaining these vector attributes, the first deformation feature vectors with the same vector attributes in the first deformation feature vectors corresponding to multiple support structure data are fused to generate a target deformation feature vector. For example, in tunnel engineering, there may be axial strain vectors of multiple anchor bolts within the same measurement accuracy range and in the same measurement direction. By fusing these axial strain vectors with the same vector attributes, a more representative target deformation feature vector can be obtained. The fusion process may involve operations such as weighted averaging of data, data screening, or data merging.

[0107] For the surface crack propagation vectors of the shotcrete layer, if there are multiple crack propagation vectors in the same crack plane direction and with the same crack width measurement accuracy, by fusing these vectors, the crack propagation conditions at different positions or at different time periods can be comprehensively considered, and a target deformation feature vector that more comprehensively reflects the surface crack propagation of the shotcrete layer can be generated. Similarly, for the bending degree change vectors of the steel arch, by fusing the bending degree change vectors with the same bending degree measurement reference point and the same measurement error range, a target deformation feature vector that can more accurately reflect the overall bending degree change of the steel arch can be obtained.

[0108] This fusion operation helps to reduce data redundancy and enhance data representativeness. In the monitoring of tunnel soft rock support structures, by generating a target deformation feature vector, the subsequent analysis process can be simplified, and the efficiency and accuracy of data analysis can be improved. For example, in the subsequent determination process of the deformation feature conversion strategy, directly using the target deformation feature vector can avoid repeated processing of a large number of similar first deformation feature vectors. At the same time, since the target deformation feature vector integrates the information of multiple vectors with the same vector attributes, it can more accurately reflect the overall deformation of the support structure, thus providing a more reliable basis for selecting an appropriate deformation feature conversion strategy.

[0109] Figure 2 FIG. shows the hardware structure diagram of a tunnel soft rock large deformation support reliability monitoring system 100 provided by an embodiment of the present invention for implementing the above-mentioned tunnel soft rock large deformation support reliability monitoring method, as Figure 2 shown, the tunnel soft rock large deformation support reliability monitoring system 100 may include a processor 110, a machine-readable storage medium 120, a bus 130, and a communication unit 140.

[0110] The machine-readable storage medium 120 may store data and / or instructions. In some embodiments, the machine-readable storage medium 120 may store data obtained from an external terminal. In some embodiments, the machine-readable storage medium 120 may store data and / or instructions that the tunnel soft rock large deformation support reliability monitoring system 100 uses to execute or complete the exemplary methods described in the present invention.

[0111] In a specific implementation process, one or more processors 110 execute the computer-executable instructions stored in the machine-readable storage medium 120, so that the processors 110 can execute the tunnel soft rock large deformation support reliability monitoring method in the above method embodiments. The processors 110, the machine-readable storage medium 120, and the communication unit 140 are connected through a bus 130. The processors 110 can be used to control the transceiver actions of the communication unit 140.

[0112] For the specific implementation process of the processors 110, reference may be made to the various method embodiments executed by the above tunnel soft rock large deformation support reliability monitoring system 100. Their implementation principles and technical effects are similar, and will not be elaborated here in this embodiment.

[0113] In addition, an embodiment of the present invention also provides a readable storage medium, in which computer-executable instructions are preset. When a processor executes the computer-executable instructions, the above tunnel soft rock large deformation support reliability monitoring method is implemented.

[0114] It should be noted that, in order to simplify the description of the present invention disclosure and thus help the understanding of one or more embodiments of the invention, in the previous description of the embodiments of the present invention, sometimes multiple features are merged into one embodiment, drawing, or description thereof.

Claims

1. A reliability monitoring method for supporting large deformation of soft rock in tunnels, characterized in that, The method includes: Obtaining a plurality of support structure data of the tunnel soft rock corresponding to the deformed support monitoring request, and extracting first deformation feature vectors respectively corresponding to the plurality of support structure data, where the plurality of support structure data correspond to a plurality of structural manifestation forms; Obtaining at least two reference deformation feature conversion strategies, each of which includes at least one step of feature selection and feature focusing on the first deformation feature vector; Determining a target deformation feature conversion strategy from the at least two reference deformation feature conversion strategies according to the monitoring intention corresponding to the deformed support monitoring request and the data label corresponding to the support structure data, where the data label represents the reference label knowledge point corresponding to the support structure data under the deformed support monitoring request; Using the target deformation feature conversion strategy to perform deformation feature conversion on the plurality of first deformation feature vectors to generate a plurality of second deformation feature vectors; Performing a reliability index decision on the plurality of second deformation feature vectors to generate reliability index decision data corresponding to the deformed support monitoring request, where the reliability index decision data represents a reliability prediction result corresponding to the deformed support monitoring request generated using the plurality of support structure data.

2. The reliability monitoring method for supporting large deformation of soft rock in tunnels according to Claim 1, wherein The determining a target deformation feature conversion strategy from the at least two reference deformation feature conversion strategies according to the monitoring intention corresponding to the deformed support monitoring request and the data label corresponding to the support structure data includes: Loading the monitoring intention corresponding to the deformed support monitoring request, the plurality of first deformation feature vectors, and the at least two reference deformation feature conversion strategies into an artificial intelligence network that has completed prior knowledge learning to obtain the target deformation feature conversion strategy. The artificial intelligence network is used to perform strategy matching on the at least two reference deformation feature conversion strategies according to the monitoring intention corresponding to the deformed support monitoring request and the plurality of first deformation feature vectors. The target deformation feature conversion strategy includes at least one of a target feature selection strategy and a target feature focusing strategy.

3. The reliability monitoring method for supporting large deformation of soft rock in tunnels according to claim 2, characterized in that The artificial intelligence network includes a target matching module and a decision execution module; The loading the monitoring intention corresponding to the deformed support monitoring request, the plurality of first deformation feature vectors, and the at least two reference deformation feature conversion strategies into an artificial intelligence network that has completed prior knowledge learning to obtain the target deformation feature conversion strategy includes: Loading the monitoring intention corresponding to the deformed support monitoring request, the plurality of first deformation feature vectors, and the at least two reference deformation feature conversion strategies into the target matching module to obtain the matching confidence levels respectively corresponding to the at least two reference deformation feature conversion strategies. The target matching module is used to perform the strategy matching on the at least two reference deformation feature conversion strategies according to the monitoring intention corresponding to the deformed support monitoring request and the plurality of first deformation feature vectors; Determining the target feature selection strategy according to the matching confidence levels respectively corresponding to the at least two reference deformation feature conversion strategies; Load the monitoring intention corresponding to the deformed support monitoring request, multiple first deformation feature vectors, and the matching confidence levels corresponding to the at least two reference deformation feature conversion strategies into the decision execution module to obtain the target feature focusing strategy. The decision execution module is used to make a decision based on the multiple first deformation feature vectors and the matching confidence levels corresponding to the at least two reference deformation feature conversion strategies to obtain the target feature focusing strategy that matches the monitoring intention.

4. The reliability monitoring method for supporting large deformation of soft rock in tunnels according to claim 3, characterized in that The matching confidence levels corresponding to the at least two reference deformation feature conversion strategies include the first matching confidence levels corresponding to multiple feature selection strategies; Determining the target feature selection strategy according to the matching confidence levels corresponding to the at least two reference deformation feature conversion strategies includes: Output the feature selection strategy with the largest numerical value of the first matching confidence level from the first matching confidence levels corresponding to the multiple feature selection strategies as the target feature selection strategy.

5. The reliability monitoring method for large deformation support of soft rock in tunnels according to claim 3, characterized in that The matching confidence levels corresponding to the at least two reference deformation feature conversion strategies include the second matching confidence levels corresponding to multiple feature focusing strategies; Loading the monitoring intention corresponding to the deformed support monitoring request, multiple first deformation feature vectors, and the matching confidence levels corresponding to the at least two reference deformation feature conversion strategies into the decision execution module to obtain the target feature focusing strategy includes: Load the monitoring intention corresponding to the deformed support monitoring request, multiple first deformation feature vectors, and the second matching confidence levels corresponding to the multiple feature focusing strategies into the decision execution module to obtain the target feature focusing strategy.

6. The reliability monitoring method for supporting large deformation of soft rock in tunnels according to claim 3, characterized in that, Before loading the monitoring intention corresponding to the deformed support monitoring request, multiple first deformation feature vectors, and the at least two reference deformation feature conversion strategies into the target matching module to obtain the matching confidence levels corresponding to the at least two reference deformation feature conversion strategies, it further includes: Obtain multiple template support structure data corresponding to the template deformed support monitoring request, and extract the first deformation feature vectors corresponding to the multiple template support structure data. The multiple template support structure data correspond to multiple structural manifestation forms; Obtain at least two template deformation feature conversion strategies, and each template deformation feature conversion strategy includes at least one of a template feature selection strategy and a template feature focusing strategy; Load the monitoring intention corresponding to the template deformed support monitoring request, multiple first deformation feature vectors, and the at least two template deformation feature conversion strategies into the template matching module to obtain the matching confidence levels corresponding to the at least two template deformation feature conversion strategies; Load the monitoring intention corresponding to the template deformed support monitoring request, multiple first deformation feature vectors, and the matching confidence levels corresponding to the at least two template deformation feature conversion strategies into the decision execution module to generate reliability index decision data for the feature focusing strategy; Generate a target incentive mechanism network according to the reliability index decision data of the feature focusing strategy; Optimize the template matching module according to the target incentive mechanism network to generate the target matching module.

7. The reliability monitoring method for large deformation support of soft rock in tunnels according to any one of claims 2-6, characterized in that The step of using the target deformation feature conversion strategy to perform deformation feature conversion on the multiple first deformation feature vectors to generate multiple second deformation feature vectors includes: Performing correlation selection on the multiple first deformation feature vectors by using the target feature selection strategy to generate multiple first sub-deformation feature vectors that meet the correlation requirements, where the correlation requirements characterize the correlation between the first deformation feature vector and the monitoring intention; Performing feature focusing on the multiple first sub-deformation feature vectors by using the target feature focusing strategy to generate the multiple second deformation feature vectors.

8. The reliability monitoring method for supporting large deformation of soft rock in tunnels according to any one of claims 2-6, characterized in that, Before obtaining at least two reference deformation feature conversion strategies, it further includes: Loading the multiple first deformation feature vectors into a feature processing network that has completed knowledge learning in advance to generate multiple extended deformation feature vectors, where the feature space fields corresponding to the extended deformation feature vectors are more than those corresponding to the first deformation feature vectors, and the feature processing network is used to extract the extended deformation feature vectors corresponding to the first deformation feature vectors.

9. The reliability monitoring method for supporting large deformation of soft rock in tunnels according to any one of claims 2-6, characterized in that, After extracting the first deformation feature vectors respectively corresponding to the multiple support structure data, the method further includes: Obtaining the vector attributes respectively corresponding to the multiple first deformation feature vectors, where the vector attributes characterize the feature space fields corresponding to the first deformation feature vectors; Fusing the first deformation feature vectors with the same vector attributes among the first deformation feature vectors respectively corresponding to the multiple support structure data to generate a target deformation feature vector.

10. A reliability monitoring system for supporting large deformation of soft rock in tunnels, characterized in that, The tunnel soft rock large deformation support reliability monitoring system includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the tunnel soft rock large deformation support reliability monitoring method according to any one of claims 1-9 above.

Images