Underground engineering surrounding rock quality multi-standard intelligent polymerization evaluation method and system

Abstract

The invention discloses an underground engineering surrounding rock quality multi-standard intelligent aggregation evaluation method and system. The method comprises the following steps: acquiring multi-source heterogeneous data related to underground engineering surrounding rock, and performing standardized preprocessing on the multi-source heterogeneous data to obtain standardized surrounding rock data; according to the standardized surrounding rock data, combining the static characteristics and the dynamic evolution characteristics into a surrounding rock characteristic set; constructing and applying an aggregation evaluation model, and dynamically adjusting the weight of each feature in the surrounding rock feature set according to the context information; and according to the weight of each feature in the surrounding rock feature set and the output of fusing the plurality of basic evaluation models, calculating a surrounding rock comprehensive evaluation result and a corresponding confidence coefficient. According to the technical scheme, objectivity, accuracy, timeliness and robustness of surrounding rock quality evaluation are improved, the confidence degree of the evaluation result is systematically quantified, and a basis is provided for reliability of an evaluation conclusion.

Description

Technical Field

[0001] The present invention relates to the technical field of geotechnical engineering, and in particular to a multi-standard intelligent aggregation evaluation method and system for surrounding rock quality of underground engineering. Background Art

[0002] For underground projects such as tunnels, mine workings, hydropower caverns, and underground powerhouses, the quality and stability of the surrounding rock mass (i.e., the rock mass surrounding the excavation face) are key factors in determining project safety, economics, and feasibility during their investigation, design, construction, and long-term operation. Therefore, accurately and reliably evaluating surrounding rock quality and predicting its behavior under engineering disturbances are crucial for optimizing support design, ensuring construction safety, and controlling project risks and costs. Currently, a variety of surrounding rock quality assessment methods have been developed and widely used, including the internationally accepted Rock Mass Quality Index (RMR) method, the Tunnel Quality Index System, and the BQ method, included in the Chinese national standard "Engineering Rock Mass Classification Standard" (GB / T 50218). Based on geological surveys, indoor and outdoor rock mechanics tests, and field engineering experience, these methods score or index key factors such as rock structure, rock strength, groundwater, and stress state. These methods have provided the engineering community with a relatively mature set of surrounding rock classification and engineering guidelines, playing an important role in practice.

[0003] However, as underground projects progress toward deeper, larger, and more complex geological environments (e.g., areas of high stress, strong tectonic influence, and complex lithologic combinations), existing surrounding rock assessment methods have also exposed several limitations. First, traditional methods still suffer from a degree of subjectivity in their application. The scoring process relies on the experience of engineers, and different evaluators may reach different conclusions. Second, existing methods are mostly static assessments, making it difficult to effectively integrate the large amount of real-time, dynamic monitoring data acquired during construction or operation (such as surrounding rock displacement, stress changes, and microseismic activity), and thus fail to timely reflect the dynamic evolution of surrounding rock conditions and potential instability risks. Furthermore, how to scientifically and rationally integrate multi-source heterogeneous information, including geological surveys, geophysical exploration, in-situ testing, laboratory experiments, and dynamic monitoring, while also considering the relative importance (i.e., weighting) of each evaluation indicator in different geological environments, engineering stages, or data quality, is a common challenge faced by existing methods. Empirical fixed weights are often used, lacking adaptability to specific "contextual information." Finally, traditional assessment results often provide a certainty level, rarely considering the inherent uncertainty of geological information. This makes it difficult to provide confidence levels or risk probabilities for the assessment results, limiting their application in refined risk decision-making. How to properly solve the above problems has become a topic that needs to be solved urgently in the industry. Summary of the Invention

[0004] The present invention provides a multi-standard intelligent aggregation evaluation method and system for the quality of surrounding rock in underground projects, which is used to overcome the limitations of existing evaluation methods, intelligently and dynamically evaluate surrounding rock, significantly improve objectivity and accuracy, quantify confidence, achieve adaptive optimization, and provide a reliable decision-making basis for the project.

[0005] According to a first aspect of the present invention, a multi-standard intelligent aggregation evaluation method for underground engineering surrounding rock quality is provided, the multi-standard intelligent aggregation evaluation method for underground engineering surrounding rock quality comprises:

[0006] Acquiring multi-source heterogeneous data related to the surrounding rock of the underground engineering, and performing standardized preprocessing on the multi-source heterogeneous data to obtain standardized surrounding rock data;

[0007] Extracting static features based on the standardized surrounding rock data according to a multi-evaluation standard system, extracting dynamic evolution features based on time series monitoring data in the standardized surrounding rock data, and combining the static features and the dynamic evolution features into a surrounding rock feature set;

[0008] Constructing and applying an aggregated evaluation model, receiving the surrounding rock feature set and preset context information as input through the aggregated evaluation model, and dynamically adjusting the weight of each feature in the surrounding rock feature set according to the context information;

[0009] Based on the weight of each feature in the surrounding rock feature set and the output of the fusion of multiple basic evaluation models, the surrounding rock comprehensive evaluation result and the corresponding confidence level are calculated.

[0010] In one embodiment, dynamically adjusting the weight of each feature in the surrounding rock feature set according to the context information includes:

[0011] Identify the context information, which includes at least one or more of the surrounding rock geological environment, engineering stage and data quality, and the weight of each feature W i The calculation formula is as follows:

[0012] W i =B i *(1+CIF i )

[0013] Among them, CIF i is the contextual impact factor of feature i, W i is the weight of feature i, B i is the basic weight of feature i;

[0014] Contextual Impact Factor (CIF) i The calculation formula is as follows:

[0015] CIF i =α*f(G i)+β*g(P i )+γ*h(Q i )

[0016] G is the geological correlation index of feature i based on the surrounding rock geological environment assessment, P is the stage importance index of feature i based on the engineering stage assessment, Q is the data confidence index of feature i based on the data quality assessment, f, g, and h are preset nonlinear mapping functions used to convert the degree of influence of each geological correlation, stage importance, and data confidence index on the weight, α, β, and γ are the sensitivity coefficients of each contextual information, and their sum is 1 or set according to prior knowledge.

[0017] In one embodiment, the calculation of the surrounding rock comprehensive evaluation result and the corresponding confidence level based on the weight of each feature in the surrounding rock feature set and the integration of the outputs of multiple basic evaluation models includes:

[0018] An integrated learning strategy is used to train and fuse multiple basic evaluation models, wherein the basic evaluation model is based on the surrounding rock feature set and the dynamic weight W i Perform training or weighted fusion;

[0019] generating the surrounding rock comprehensive evaluation result based on the output of the integrated learning strategy;

[0020] The confidence level of the evaluation result is quantified based on the consistency or probability distribution of the output of the basic evaluation model.

[0021] In one embodiment, extracting the dynamic evolution characteristics of the surrounding rock includes:

[0022] Extracting time series monitoring data of at least one of surrounding rock displacement, stress or microseismic events from the standardized surrounding rock data;

[0023] The time series monitoring data are processed using a time series analysis technique, and indicators that can reflect the dynamic evolution of the surrounding rock state are extracted as the dynamic evolution characteristics, and are incorporated into the surrounding rock feature set.

[0024] In one embodiment, it further includes:

[0025] Obtain feedback information on the actual response of surrounding rock in engineering practice;

[0026] Through the feedback information, the constructed intelligent aggregation evaluation model is parameterized or retrained, and the context influence factor CIF is adjusted. i The sensitivity coefficients α, β, and γ in the formula.

[0027] In one embodiment, the calculation of the surrounding rock comprehensive evaluation result and the corresponding confidence level includes:

[0028] Spatially associating the surrounding rock comprehensive evaluation results and corresponding confidence levels with a Geographic Information System (GIS) model or a Building Information Model (BIM) model of the underground project;

[0029] Based on the spatial correlation results, a three-dimensional spatial distribution map of the surrounding rock quality or instability risk probability is generated, and high-risk areas are marked in the map.

[0030] According to a second aspect of the present invention, a multi-standard intelligent aggregation evaluation system for surrounding rock quality of underground engineering is provided, comprising:

[0031] an acquisition module, configured to acquire multi-source heterogeneous data related to the surrounding rock of the underground engineering, and perform standardization preprocessing on the multi-source heterogeneous data to obtain standardized surrounding rock data;

[0032] a combination module, configured to extract static features based on the standardized surrounding rock data and a multi-evaluation standard system, and extract dynamic evolution features based on time series monitoring data in the standardized surrounding rock data, and combine the static features and the dynamic evolution features into a surrounding rock feature set;

[0033] an adjustment module, configured to construct and apply an aggregated evaluation model, receive the surrounding rock feature set and preset context information as input through the aggregated evaluation model, and dynamically adjust the weight of each feature in the surrounding rock feature set according to the context information;

[0034] The calculation module is used to calculate the comprehensive evaluation results of the surrounding rock and the corresponding confidence levels according to the weights of the various features in the surrounding rock feature set and the outputs of multiple basic evaluation models.

[0035] In one embodiment, the acquisition module, the combination module, the adjustment module and the calculation module are controlled to execute any one of the above-mentioned multi-standard intelligent aggregation evaluation methods for surrounding rock quality of underground engineering.

[0036] According to a third aspect of the present invention, there is provided an electronic device, the electronic device comprising: a communication interface, a processor, and a memory;

[0037] Wherein, the memory is used to store program instructions, and when the program instructions are executed by the processor that is communicatively connected to the memory through the communication interface, any of the above-mentioned multi-standard intelligent aggregation evaluation methods for surrounding rock quality of underground engineering is implemented.

[0038] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a computer (for example, a processor in a computer), any of the above-mentioned multi-standard intelligent aggregation evaluation methods for the surrounding rock quality of underground engineering is implemented.

[0039] In summary, the present invention provides a multi-standard intelligent aggregation evaluation method and system for the quality of surrounding rock of underground engineering, the method comprising: obtaining multi-source heterogeneous data related to the surrounding rock of the underground engineering, and performing standardized preprocessing on the multi-source heterogeneous data to obtain standardized surrounding rock data; extracting static features based on the standardized surrounding rock data according to a multi-evaluation standard system, and extracting dynamic evolution features based on the time series monitoring data in the standardized surrounding rock data, and combining the static features and the dynamic evolution features into a surrounding rock feature set; constructing and applying an aggregation evaluation model, receiving the surrounding rock feature set and preset context information as input through the aggregation evaluation model, and dynamically adjusting the weight of each feature in the surrounding rock feature set according to the context information; calculating the comprehensive evaluation result of the surrounding rock and the corresponding confidence level according to the weight of each feature in the surrounding rock feature set and the output of the fusion of multiple basic evaluation models. The technical solution of the present application significantly improves the objectivity, accuracy, timeliness and robustness of surrounding rock quality evaluation by introducing a context-aware dynamic weight adjustment mechanism, fusing dynamic monitoring data and adopting a multi-model integrated learning strategy. This method overcomes the limitations of existing surrounding rock assessment methods, such as strong subjectivity, static assessment process, empirical weight allocation, and difficulty in effectively integrating multi-source heterogeneous information. It also systematically quantifies the confidence level of the assessment results, providing a basis for the reliability of the assessment conclusions.

[0040] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0041] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0043] Figure 1 A flowchart of a multi-standard intelligent aggregation evaluation method for surrounding rock quality of underground engineering provided by an embodiment of the present invention;

[0044] Figure 2 A flowchart of step S14 of a multi-standard intelligent aggregation evaluation method for surrounding rock quality of underground engineering provided by an embodiment of the present invention;

[0045] Figure 3A flowchart of step S13 of a multi-standard intelligent aggregation evaluation method for surrounding rock quality of underground engineering provided by an embodiment of the present invention;

[0046] Figure 4 A flowchart of another method for multi-standard intelligent aggregation evaluation of surrounding rock quality in underground engineering provided by an embodiment of the present invention;

[0047] Figure 5 A flowchart of step S14 of another method for multi-standard intelligent aggregation evaluation of surrounding rock quality in underground engineering provided by an embodiment of the present invention;

[0048] Figure 6 A structural diagram of a multi-standard intelligent aggregation evaluation system for surrounding rock quality of underground engineering provided by an embodiment of the present invention;

[0049] Figure 7 A structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The features and exemplary embodiments of various aspects of the present application will be described in detail below. In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is merely to provide a better understanding of the present application by illustrating the examples of the present application.

[0051] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, the elements defined by the phrase "comprising..." do not exclude the presence of other identical elements in the process, method, article, or device comprising the elements.

[0052] like Figure 1 As shown, the present invention provides a multi-standard intelligent aggregation evaluation method for underground engineering surrounding rock quality, which includes:

[0053] In step S11, multi-source heterogeneous data related to the surrounding rock of the underground engineering is acquired, and the multi-source heterogeneous data is subjected to standardization preprocessing to obtain standardized surrounding rock data;

[0054] In step S12, static features are extracted based on the standardized surrounding rock data according to a multi-evaluation standard system, and dynamic evolution features are extracted based on the time series monitoring data in the standardized surrounding rock data, and the static features and the dynamic evolution features are combined into a surrounding rock feature set;

[0055] In step S13, an aggregated evaluation model is constructed and applied, wherein the aggregated evaluation model receives the surrounding rock feature set and preset context information as input, and dynamically adjusts the weight of each feature in the surrounding rock feature set according to the context information;

[0056] In step S14, the surrounding rock comprehensive evaluation result and the corresponding confidence level are calculated based on the weight of each feature in the surrounding rock feature set and the output of the fusion of multiple basic evaluation models.

[0057] In one embodiment, a method for evaluating surrounding rock quality is provided that is objective, dynamic, and intelligently integrates multi-source information, capable of assessing uncertainty. This method includes data acquisition and preprocessing, feature extraction and combination, model application and dynamic weight adjustment, and result calculation and confidence output. This method overcomes the limitations of traditional surrounding rock quality evaluation methods, such as strong subjectivity, static assessment, fixed weights, and a lack of uncertainty quantification. The multi-source information includes geological surveys (e.g., drill core descriptions, geological sketches, and geophysical reports), laboratory tests (e.g., rock physics and mechanical parameter testing), in-situ field tests (e.g., load tests and geostress testing), construction process information (e.g., excavation methods and support parameters), and, crucially, long-term or real-time field monitoring data (e.g., surrounding rock displacement, anchor (cable) axial force, seepage water pressure, and microseismic / acoustic emission signals). Heterogeneous representations of these data vary in form, structure, accuracy, and temporal / spatial scales, including structured data (e.g., experimental values), semi-structured data (e.g., geological report text), and unstructured data (e.g., images and waveform signals), as well as the coexistence of qualitative descriptions and quantitative values. The stability of surrounding rock is a complex and systematic problem affected by multiple factors. Data from a single source or type can often only reflect the characteristics of a certain aspect of the surrounding rock. Acquiring multi-source heterogeneous data can more comprehensively characterize the intrinsic properties (such as rock mechanical properties and structural characteristics) and external responses (such as deformation, stress changes, and fracture activities) of the surrounding rock from different dimensions, scales, and perspectives, thereby minimizing information omissions. For example, geological survey reports alone may not be able to fully foresee the actual dynamic response of high stress or specific structures (such as extremely thin slate or faults in GS projects) under excavation disturbances, and must be combined with on-site monitoring data.

[0058] Standardization preprocessing includes a series of operations, including data cleaning (addressing missing values, outliers, and noise), data conversion (standardizing dimensions, converting formats, and quantifying qualitative information), and data normalization / standardization (mapping data to a specific range, such as [0, 1] or a standard normal distribution, to eliminate the impact of scale differences between different indicators). Data cleaning ensures the quality of model input data; data conversion enables the comparison and integration of different data types; and normalization / standardization ensures the stability and efficiency of model training (especially for models that rely on distance or gradients, such as support vector machines and neural networks), preventing features with large numerical ranges from dominating model results. "Rock integrity descriptions" in geological reports (such as "relatively intact," "relatively fragmented," and "fragmented") can be converted to values in the [0, 1] range using fuzzy membership functions or expert scoring. Ground stress values (MPa) and structural plane spacing (m) with widely varying distributions are normalized. Monitoring data are filtered and denoised, and short-term gaps are filled using interpolation.

[0059] Combining the static features and the dynamic evolution features into a surrounding rock feature set means performing deep processing and information mining to extract features that can effectively characterize the surrounding rock state and potential risks, thereby realizing the transformation from raw data to abstract features.

[0060] Extracting static features based on a multi-evaluation standard system involves calculating parameters reflecting the inherent properties and geological conditions of the surrounding rock from standardized data, based on existing and recognized rock mass evaluation theories and standards. This includes not only the final overall score or grade, but more importantly, the "intermediate" parameters or fundamental indicators that constitute these evaluation systems, such as uniaxial compressive strength (UCS), rock quality index (RQD), structural plane spacing, structural plane conditions (roughness Jr, weathering degree Ja, filling), groundwater conditions (Jw), in-situ stress conditions (SRF or stress reduction factor), lithologic parameters, and structural plane occurrence (dip, inclination). These characteristics can provide richer and more granular information for subsequent intelligent models. For example, two surrounding rock masses with the same RMR level may exhibit distinct instability modes due to differences in structural plane conditions or stress states. Using only the total RMR score would miss this information. Extracting multi-system features helps capture the different aspects emphasized by different standards.

[0061] Extracting dynamic evolution characteristics from time-series monitoring data involves using standardized monitoring data (such as time-varying data sequences of displacement, stress, microseismic event rate, and energy) to extract dynamic indicators that reflect the surrounding rock's response to engineering disturbances, state change trends, and potential precursors to instability through time series analysis and signal processing techniques. Examples include displacement rate, displacement acceleration, relative deformation of characteristic points (such as vaults and sidewalls), stress change rate, stress concentration / relaxation trends, sudden changes in microseismic event frequency / energy, and changes in dominant frequencies. Dynamic characteristics directly reflect the behavior of the surrounding rock under the actual stress path and are irreplaceable for identifying whether deformation is convergent, whether there is a trend of accelerated failure, determining support effectiveness, and providing early warning of rockbursts or large deformations (such as the potential for Grade II to III large deformations in the layered surrounding rock of the GS Hydropower Station).

[0062] Combining the surrounding rock feature set involves organizing all extracted static and dynamic features into a structured dataset, typically a high-dimensional vector or matrix, where each row represents an evaluation unit (e.g., tunnel mileage, specific area) or a time point, and each column represents a feature. This creates a multi-dimensional, static and dynamic "digital portrait" of the surrounding rock state. For example, for a monitoring section in a particular tunnel, its "surrounding rock feature set" might include: UCS = 85 MPa, RQD = 70%, Jn = 9, Jr = 1.5, Ja = 1, Jw = 1, SRF = 1.5 (partial static features from the Q system and RMR), Layer Thickness = 0.2 m, DipAngle = 70° (specific geological static features), MaxDisplacementRate = 0.5 mm / day, DisplacementAcceleration = 0.02 mm / day², MicroseismicEventRate = 5 events / hour (dynamic features).

[0063] Model application and dynamic weight adjustment are based on the surrounding rock feature set and combined with contextual information, and the features are intelligently processed through the aggregate evaluation model. Constructing and applying the aggregate evaluation model means using machine learning, artificial intelligence and other technologies to establish one or more models that can process input features and make evaluation predictions. Model types can be support vector machines (SVM), random forests (Random Forest), neural networks (Neural Networks), Bayesian networks (Bayesian Networks), etc. The construction of the model usually involves a training process based on historical data (including feature sets and corresponding known surrounding rock conditions or stability results). Through the powerful nonlinear mapping capabilities of the AI model and the ability to learn complex patterns from data, the association between feature combinations and surrounding rock conditions that are difficult for human experts to perceive can be discovered, overcoming the limitations of traditional evaluation formulas or simple linear models.

[0064] Contextual information refers to background conditions that can influence the importance of features or the applicability of evaluation criteria. These include the surrounding rock geological environment (e.g., presence in a fault-affected zone, specific lithologic combinations such as layered rock masses or high-stress areas), the project phase (e.g., initial excavation, post-support stabilization, and long-term operation), and data quality (e.g., the reliability of sensor data). Incorporating contextual information enables adaptive evaluation, eliminating the need for a one-size-fits-all model and enabling it to be tailored to local conditions and time. Recognizing that the significance of the same feature varies across different contexts is a significant advancement. For example, the importance of joint density in evaluating the surrounding rock of an intact granite massif should clearly differ from that of a fractured fault zone. The model uses contextual input to automatically and non-fixedly adjust the weight of each feature in the surrounding rock feature set. Features with higher weights have a greater impact on the final results. Weights are assigned differently for different evaluation objects or contextual states. This overcomes the limitations of traditional methods that rely on empirical evidence or fixed weights, making the evaluation more objective, accurate, and practical. For example, in the steeply inclined, low-angle slate section of the GS Hydropower Station spillway tunnel (geological context), the model might automatically increase the weights of bedding-related features (such as bedding dip, rock thickness, and shear strength along the bedding) and dynamic features that predict large deformations. In contrast, in the intact basalt section of the powerhouse area, the model might prioritize rock strength and integrity. When the data quality of a monitoring sensor deteriorates, the weight of the corresponding dynamic feature is automatically lowered to prevent unreliable data from interfering with the results.

[0065] Identify the context information, which includes at least one or more of the surrounding rock geological environment, engineering stage and data quality, and the weight of each feature W i The calculation formula is as follows:

[0066] W i =Bi *(1+CIF i )

[0067] Among them, CIF i is the contextual impact factor of feature i, W i is the weight of feature i, B i is the basic weight of feature i;

[0068] Contextual Impact Factor (CIF) i The calculation formula is as follows:

[0069] CIF i =α*f(G i )+β*g(P i )+γ*h(Q i )

[0070] G is the geological correlation index of feature i based on the surrounding rock geological environment assessment, P is the stage importance index of feature i based on the engineering stage assessment, Q is the data confidence index of feature i based on the data quality assessment, f, g, and h are preset nonlinear mapping functions used to convert the degree of influence of each geological correlation, stage importance, and data confidence index on the weight, α, β, and γ are the sensitivity coefficients of each contextual information, and their sum is 1 or set according to prior knowledge.

[0071] Based on the weights of each feature in the surrounding rock feature set and the outputs of multiple basic evaluation models, a comprehensive evaluation result of the surrounding rock and the corresponding confidence level are calculated. The feature weights are combined with the outputs of multiple basic evaluation models to make a comprehensive judgment and output the comprehensive evaluation result of the surrounding rock and the corresponding confidence level.

[0072] Based on the weights and calculations of the fused multi-model outputs, the final evaluation result is not derived from a single model or simple weighting, but rather integrates the predictions of multiple underlying evaluation models. This takes into account the dynamic weighting of each feature and the potential trust in the underlying models themselves. Common fusion strategies include ensemble learning, such as voting, averaging, and weighted averaging. Ensemble learning generally provides more stable, accurate, and generalizable predictions than a single model. By fusing multiple models, the risk of bias or overfitting from a single model can be effectively reduced, improving the robustness of the evaluation.

[0073] Calculate the comprehensive evaluation results of the surrounding rock, which can be a quantitative index (such as stability index 0-1, safety factor), a qualitative level (such as Class IV surrounding rock, stable / understable / instability), or the probability of occurrence of specific risks (such as the probability of large deformation, rock burst level).

[0074] Calculate the corresponding confidence level, which can be expressed in various forms, such as: the confidence interval of the evaluation result (such as the stability index is 0.65±0.05), the probability that the result belongs to a certain level (such as the probability of belonging to Grade III surrounding rock is 70%), or a comprehensive confidence score (such as the credibility of the evaluation result is 85%). The method of calculating the confidence level can be based on the consistency of the output of the integrated model (if the predictions of each model are highly consistent, the confidence level is high), using the posterior probability distribution of the probabilistic model (such as the Bayesian network), or by introducing the uncertainty of the input data to perform Monte Carlo simulation. Engineering decisions often need to be made under uncertainty, and it is crucial to understand the credibility of the evaluation results. High-confidence results can provide support for bold or economical plans, while low-confidence results suggest that engineers need to be more cautious and may need to conduct additional surveys, strengthen monitoring, or adopt more conservative designs. For example, if the confidence level is high (such as 90%) for determining that a certain slate at a GS hydropower station may undergo large deformation of level II or III, strong preventive measures must be taken; if the confidence level is low (such as 50%), further strengthened monitoring may be required to confirm the risk.

[0075] The technical solution in this embodiment significantly improves the objectivity, accuracy, timeliness, and robustness of surrounding rock quality assessment by introducing a context-aware dynamic weight adjustment mechanism, integrating dynamic monitoring data, and employing a multi-model ensemble learning strategy. This overcomes the limitations of existing surrounding rock assessment methods, which are characterized by strong subjectivity, static evaluation processes, empirical weight allocation, and difficulty in effectively integrating heterogeneous multi-source information. Furthermore, it systematically quantifies the confidence level of the assessment results, providing a basis for the reliability of the assessment conclusions.

[0076] In one embodiment, Figure 2 As shown, step S14 includes the following steps S21-S23:

[0077] In step S21, an integrated learning strategy is used to train and fuse multiple basic evaluation models, wherein the basic evaluation model is based on the surrounding rock feature set and the dynamic weight W i Perform training or weighted fusion;

[0078] In step S22, based on the output of the integrated learning strategy, the surrounding rock comprehensive evaluation result is generated;

[0079] In step S23 , the confidence of the evaluation result is quantified based on the consistency or probability distribution output by the basic evaluation model.

[0080] In one embodiment, a final evaluation conclusion that is accurate, reliable, and includes a credibility assessment is analyzed by using multi-dimensional feature information with dynamic weights. Model fusion is performed through an ensemble learning strategy, and confidence is quantified based on the model output.

[0081] Ensemble learning fusion based on dynamic weights improves model robustness and accuracy. An ensemble learning strategy is used to train and fuse multiple basic evaluation models, and the basic evaluation model is trained or weighted fused based on the surrounding rock feature set and the dynamic weight Wi. The dynamic weight Wi is integrated into the integration process using the surrounding rock feature set as the data basis. The basic evaluation model refers to a single learner that constitutes the ensemble, which can be multiple models of the same or different types (such as decision trees, SVMs, neural networks, Bayesian classifiers, and may even include some simplified models based on physical mechanisms, etc.). Ensemble learning strategies include Bagging (such as random forests), Boosting (such as AdaBoost, GBDT), Stacking, etc.

[0082] Individual models often have limitations, such as sensitivity to specific data types, prone to overfitting or underfitting, or exhibiting certain predictive biases. Ensemble learning, by constructing and combining multiple learners to accomplish learning tasks, typically achieves significantly superior generalization performance, higher accuracy, and greater robustness than a single learner. This is particularly important for complex problems like surrounding rock stability, characterized by numerous influencing factors, nonlinearity, and high uncertainty, effectively avoiding evaluation errors caused by the inappropriate selection of a single model.

[0083] Weighted fusion assigns different weights to the outputs of each base model when fusing them. For example, if the context gives higher weights to features related to structural planes, then the outputs of the corresponding base models will be assigned higher weights during the final fusion, thus having a greater impact on the final result. Applications in the training process (Training) are used during the training phase of the base models. For example, through weighted loss functions, the model focuses more on information corresponding to features that are more important in the current context. Feature Selection / Subspace: In certain ensemble strategies (such as random forest variants), feature importance (represented by Wi) may influence the probability of a feature being selected for inclusion in the subset used to construct a single base learner (such as a decision tree). This ensures that context-driven feature importance judgments are directly and effectively transmitted to the final evaluation process, making the ensemble model's output context-adaptive. For example, suppose three base models (SVM, Random Forest, GBDT) are all evaluating the stability of a surrounding rock mass segment. The current context is "intense excavation disturbance" (engineering phase), resulting in high Wi values for dynamic features (such as displacement rate). When fusing the results, the system may assign a higher fusion weight to the basic model (such as GBDT) that is most sensitive to changes in displacement rate and has the most accurate predictions, or in the Stacking meta-model training, allow the features related to displacement rate and the output of the corresponding basic model to account for a larger proportion.

[0084] Generate a comprehensive evaluation result based on the output of the ensemble learning strategy. The comprehensive evaluation result (e.g., "Stability Level: Level III," "Probability of Large Deformation in the Next 24 Hours: 12%," "Rockburst Risk: Medium," "Recommended Adjustment Factor for Support Parameters: 1.1") is generated by the ensemble learning strategy and represents the system's assessment of the current surrounding rock state (based on the static and dynamic characteristics of the input and incorporating contextual information).

[0085] The confidence level of the evaluation results is quantified based on the consistency or probability distribution of the outputs of the basic evaluation models. There are two main approaches to quantifying confidence: one is to leverage the differences between the prediction results of the individual basic models in the integrated model, and the other is to utilize the probabilistic information output by the basic models themselves. Underground engineering projects are rife with uncertainty (due to unknown geological conditions, simplified model assumptions, errors in monitoring data, etc.). Traditional deterministic evaluation results (such as "stable" or "unstable") fail to reflect this uncertainty and can mislead decision-making. Providing confidence levels conveys critical information about the reliability of the prediction results to users and represents a responsible and scientific evaluation. For example, the engineering decisions (such as whether to strengthen support) corresponding to a prediction of "stable" (95% confidence) will be very different from those corresponding to a prediction of "stable" (55% confidence). For complex projects like the GS Hydropower Station, including a high-confidence assessment when assessing the risk of "Grade II-III large deformation" will greatly increase decision-makers' resolve to take preventative measures.

[0086] When multiple base models in an ensemble (especially those of different types) reach highly consistent conclusions after their respective judgments (for example, most models predict "stable," or their predicted stability indices are very similar), the system's conclusion about the current input feature data can be considered clear and has low uncertainty, thus giving it a high confidence level. Conversely, if the model outputs differ significantly, this indicates significant uncertainty in the system's judgment of the current situation, and confidence should be lowered accordingly. This consistency can be quantified using metrics such as the variance, entropy, or voting percentage of the model output.

[0087] Many machine learning models can output probabilistic information natively. For example, logistic regression or certain neural networks can output the probability of belonging to a certain class; Bayesian networks can directly generate posterior probability distributions for different states. Using these probabilistic outputs, one can directly obtain a probabilistic description of the outcome, such as P(stable) = 0.85, or calculate a confidence interval (e.g., for a predicted displacement of 5.2 cm, the 90% confidence interval is [4.8 cm, 5.6 cm]). For example, for a given evaluation unit, if 9 out of 10 base models predict "stable" and 1 predicts "understable," the consensus confidence level based on the voting proportion can be set to 90%. Alternatively, if a Bayesian network model outputs P(large deformation | data, context) = 0.15 with a standard deviation of 0.03, one could report "the probability of large deformation is 15% with high confidence (or provide a specific confidence interval)."

[0088] The technical solution in this embodiment clarifies the application of ensemble learning strategies, explains the role of dynamic weights in fusion, and specifies methods for quantifying confidence based on model consistency or probability distribution. This not only ensures that the final evaluation results are the product of integrating the wisdom of multiple models and fully considering contextual influences, thereby improving accuracy and robustness, but also, by introducing confidence quantification, enables the evaluation results to reflect inherent uncertainty, providing an unprecedentedly reliable basis for underground engineering risk management and scientific decision-making.

[0089] In one embodiment, Figure 3 As shown, step S13 includes the following steps S31-S32:

[0090] In step S31, time series monitoring data of at least one of surrounding rock displacement, stress or microseismic events is extracted from the standardized surrounding rock data;

[0091] In step S32, the time series analysis technology is applied to process the time series monitoring data, and the indicators that can reflect the dynamic evolution of the surrounding rock state are extracted as the dynamic evolution characteristics, and are included in the surrounding rock feature set.

[0092] In one embodiment, time-series monitoring data related to displacement, stress, or microseismic events is selected from standardized surrounding rock data. Multiple time series analysis techniques are applied to this data to extract key indicators reflecting the evolutionary trends of the surrounding rock state. These indicators are then incorporated into the surrounding rock feature set as dynamic evolution characteristics. While ensuring data comparability and consistency, this approach allows for timely capture of surrounding rock state changes over time, providing a scientific basis for subsequent comprehensive evaluation and construction decisions.

[0093] First, the multi-source heterogeneous data needs to be standardized. Standardized surrounding rock data refers to a data set that meets unified data structure and dimensionality requirements after data cleaning, dimensional unification, and outlier removal. For example, if the stress units of different monitoring devices are MPa and kPa, respectively, they need to be unified into the same dimension (such as MPa); if the sampling frequencies of different sensors are different, interpolation or resampling strategies can be used during the standardization process to ensure that the final data series has the same or sufficiently close sampling points on the time axis. After completing the above standardization, at least one time-series monitoring data of surrounding rock displacement data, surrounding rock stress data, or microseismic event data can be selected for subsequent analysis. The above data are usually collected in real time or periodically by monitoring equipment (such as multi-point displacement meters, stress meters, or microseismic monitoring systems) to reflect the changing patterns of the surrounding rock over a period of time.

[0094] Screen and preprocess the selected surrounding rock displacement, stress, or microseismic event data. If there are missing values in the monitoring data, interpolation algorithms (such as linear interpolation and spline interpolation) can be used to estimate and fill them. If extreme outliers are generated during certain periods due to sensor failure or external interference, they need to be eliminated or corrected according to statistical rules (such as box plot method and triple standard deviation method). For high-frequency noise in time series monitoring data, filtering algorithms (such as wavelet noise reduction and Kalman filtering) can be used for smoothing to highlight the main trends in the evolution of the surrounding rock state.

[0095] After preprocessing, various time series analysis methods can be applied to extract important indicators that reflect the dynamic evolution of the surrounding rock, based on the properties of the selected data and the analysis objectives. For the overall trend of surrounding rock displacement or stress data, methods such as sliding average and exponential smoothing can be used to determine whether the surrounding rock has sustained settlement, deformation, or stress accumulation on a long-term scale. If the monitoring data exhibits periodic variation (such as stress changes caused by periodic fluctuations in groundwater levels), a periodic decomposition model can be used to decompose the data and extract the amplitude and frequency of changes in the surrounding rock under different periodic components. In some cases, to better capture the possible future evolution trends of the surrounding rock, methods such as classic ARIMA, LSTM, or prediction models based on gray system theory can be used to make short-term or medium- to long-term predictions of the time series monitoring data, thereby obtaining possible scenarios for the future evolution of the surrounding rock. For microseismic data, characteristics such as the frequency of events and energy release rate can be focused on, and combined with geological conditions and construction progress, to determine whether the surrounding rock structure has potential instability or rupture risks.

[0096] Based on the above time series analysis results, the extracted features need to be quantified in the form of indicators or feature vectors so that they can be included in the surrounding rock feature set. If the analysis results show that the surrounding rock displacement is showing a cumulative increase trend, indicators such as the displacement increment rate and acceleration in different time periods can be used as dynamic evolution characteristics. For surrounding rock stress monitoring data, key values such as the stress change rate per unit time and the time evolution of the difference between maximum and minimum stresses can be extracted to characterize the stress concentration or relaxation of the surrounding rock. For microseismic data, indicators such as the number of events within a certain time window and the cumulative energy of events can be included in the feature set to assess whether there are risks such as crack expansion and rockburst in the surrounding rock in a local area. If a neural network or prediction model is used to make a short-term prediction of the surrounding rock state, the deviation between the actual measured value and the predicted value can be used as an indicator to measure the abnormal surrounding rock state. When the deviation exceeds a predetermined threshold, it indicates that the surrounding rock is experiencing accelerated changes or abnormal state.

[0097] After the dynamic evolution features are extracted, these features need to be integrated with previously obtained static features (e.g., rock strength, rock formation occurrence, geological structure, etc.) to form a complete set of surrounding rock features. The extracted dynamic evolution features are uniformly numbered or labeled, such as "surrounding rock deformation rate" and "stress concentration coefficient", and stored in a database or feature management module; static features and dynamic evolution features are merged using data structures (such as vectors or tables), and it is ensured that the two have the same or comparable identification information (such as the same monitoring points, the same timestamp range, etc.), thereby forming a comprehensive feature set that can be directly called by subsequent algorithms; key dynamic evolution features are visualized in the time dimension, such as drawing trend charts or heat maps, and compared with the actual situation on site or expert judgment to verify the correctness of the extracted features. If significant deviations or errors are found, the data processing process should be traced back in time and corrected again.

[0098] Imagine that during tunnel construction, multiple displacement and stress sensors are deployed, and a microseismic monitoring system is used to record potential rock failure events. After data collection and standardization, a set of surrounding rock deformation data, stress data, and microseismic event logs covering three consecutive months are obtained. Trend analysis of multi-point displacement meter data revealed a significant increase in the displacement growth rate of the tunnel vault and sidewalls between days 30 and 45, inferring significant deformation of the surrounding rock during this period. Periodic decomposition of the stress data revealed that stress fluctuated approximately every 10 days due to cyclical changes in the groundwater level. However, the peak stress fluctuation, starting on the 35th day, was significantly higher than in previous cycles, suggesting a high risk of stress concentration in the surrounding rock during this period. Microseismic data from the same period revealed a significant increase in the frequency of multiple high-energy microseismic events between days 38 and 42. The results of this time series analysis were quantified into indicators such as deformation rate, stress concentration factor, and microseismic event energy release rate. These indicators were incorporated into a dynamic evolution feature set and integrated with static features (such as stratum lithology and fault structure). This resulted in a complete surrounding rock feature set encompassing several static and dynamic features. This was then used in a subsequent aggregated evaluation model to assess the comprehensive stability of the surrounding rock. The results were consistent with field observations, demonstrating the practical value of this dynamic evolution feature extraction process.

[0099] The technical solution in this embodiment enables in-depth analysis and indexing of time-series monitoring data such as surrounding rock displacement, stress, and microseismic events, yielding dynamic evolutionary characteristics that reflect how surrounding rock conditions change over time. Combining dynamic and static characteristics not only comprehensively characterizes the complex evolution of surrounding rock during construction and operation, but also helps engineers make more scientific and accurate construction decisions and implement risk prevention and control measures.

[0100] In one embodiment, Figure 4 As shown, the following steps S41-S42 are also included:

[0101] In step S41, time series monitoring data of at least one of surrounding rock displacement, stress or microseismic events is extracted from the standardized surrounding rock data;

[0102] In step S42, the time series monitoring data is processed using a time series analysis technique, and indicators that can reflect the dynamic evolution of the surrounding rock state are extracted as the dynamic evolution characteristics, and are incorporated into the surrounding rock feature set.

[0103] In one embodiment, the actual response feedback information of the surrounding rock in engineering practice is obtained, and the feedback information is used to correct or retrain the parameters of the intelligent aggregation evaluation model and adjust the context influence factor CIF iThe sensitivity coefficients α, β, and γ in the formula are as follows. In engineering practice, the response of surrounding rock is often affected by a combination of factors, such as geological conditions, construction methods, and support measures. In order for the intelligent aggregation evaluation model to more accurately reflect the actual surrounding rock state, it is necessary to continuously collect actual response data of the surrounding rock during the use of the model and introduce this feedback information into the model update process. By correcting or retraining the model parameters and adjusting the sensitivity coefficients α, β, and γ in the contextual influence factor formula, the model's prediction and evaluation results can be better aligned with the actual situation.

[0104] Feedback on the actual response of the surrounding rock is collected through on-site monitoring, construction logs, and acceptance assessments to obtain data on actual changes in the surrounding rock during construction or operation. The feedback information is processed and screened, and the collected data is cleaned, standardized, and structured to remove noise and irrelevant information, resulting in high-quality samples that can be directly used for model updates. The intelligent aggregation evaluation model is parameter-corrected or retrained, using newly collected samples for incremental learning or secondary training of the model to correct parameters in the model that deviate significantly from the actual response of the surrounding rock. The values of α, β, and γ are dynamically adjusted for different periods, working conditions, and geological backgrounds, allowing the evaluation model to more flexibly respond to on-site changes.

[0105] Multi-point displacement meters, stress sensors, and microseismic monitoring systems are used to obtain real-time or periodic information on surrounding rock deformation, stress, and fracture events. Detailed information on geological advance forecasts, abnormal events such as rock fall or collapse, and surrounding rock quality assessments and safety ratings performed by third-party expert groups or supervisory units during project node acceptance or phased evaluations is also recorded. After obtaining the raw data, various data types need to be preprocessed to ensure compatibility with previously standardized surrounding rock data. Data derived from different sensors or monitoring systems should be unified into the same file format or database table structure. If different units exist for stress, displacement, and other data, they should be converted or normalized. Clearly erroneous or invalid data points (such as extreme values caused by sensor failure) can be eliminated or corrected based on statistical or expert judgment. By comparing the feedback information with the existing construction timeline, corresponding sample data can be provided for subsequent model updates. This technical solution yields a set of actual surrounding rock response feedback data that matches the input format of the intelligent aggregation evaluation model.

[0106] In the intelligent aggregation evaluation model described in this application, the context influence factor CIF iThe sensitivity coefficients α, β, and γ in the formula are used to characterize the degree of influence of different contextual variables on the comprehensive surrounding rock assessment results. Contextual variables include, but are not limited to, geological background information (GEO), construction condition information (CON), and environmental and operational information (ENV). Geological background information (GEO) refers to factors such as fracture development, groundwater conditions, and rock formation structure; construction condition information (CON) refers to factors such as blasting or mechanical excavation methods, support type, and grouting measures; and environmental and operational information (ENV) refers to factors such as temperature, humidity, surface load, and vehicle load during the operational phase. Certain contextual factors may have a more pronounced impact on surrounding rock conditions at different times and under different geological environments. If newly collected feedback data reveals that a certain variable (such as groundwater seepage pressure) has a stronger triggering effect on surrounding rock deformation or fracture risk, the corresponding sensitivity coefficient can be appropriately increased. Based on professional engineering experience and field monitoring results, if the surrounding rock is found to be extremely sensitive to groundwater changes or excavation methods over a certain period of time, the corresponding coefficient can be directly increased. The sensitivity coefficient is considered a parameter to be optimized, linked to the model's overall evaluation accuracy, stability, or other performance indicators. Evolutionary algorithms (such as genetic algorithms) or gradient algorithms are used for automatic search and adjustment. An adaptive update rule is designed so that the model analyzes the contribution of contextual variables after each new feedback information arrives. If the contribution of a variable significantly exceeds that of other factors, the corresponding sensitivity coefficient is automatically adjusted upward, and vice versa.

[0107] For example, in a high-ground stress tunnel project, geological surveys showed that the fault zone was severely water-rich. In the early stages of construction, the dynamic changes in groundwater pressure were not fully taken into account, and the sensitivity coefficient β (used to characterize the influence of groundwater-related contextual factors) was set relatively low. After a period of monitoring, it was found that surrounding rock collapse events occurred mainly during periods of intensive rainfall, and there was a clear correlation between the peak groundwater pressure and surrounding rock deformation. Combined with on-site feedback information, the project team correspondingly raised the sensitivity coefficient β value related to groundwater pressure, so that the model can more sensitively capture the impact of groundwater changes on surrounding rock stability in subsequent evaluations. After re-evaluation, the model's prediction accuracy for high-risk periods of surrounding rock deformation has significantly improved, proving the effectiveness of the adjustment strategy.

[0108] The technical solution in this embodiment corrects the internal parameters of the model through incremental learning or comprehensive retraining, and flexibly adjusts the sensitivity coefficients in the contextual influencing factors, thereby realizing the continuous adaptation and accurate prediction of the model under changing construction conditions and geological environments, and fully considering the complex coupling relationship between the geological environment and the construction method, thereby effectively improving the reliability and accuracy of the comprehensive evaluation results of the surrounding rock.

[0109] In one embodiment, Figure 5 As shown, step S14 further includes the following steps S51-S52:

[0110] In step S51, feedback information on actual response of surrounding rock in engineering practice is obtained;

[0111] In step S52, the constructed intelligent aggregation evaluation model is parameterized or retrained based on the feedback information, and the context influence factor CIF is adjusted. i The sensitivity coefficients α, β, and γ in the formula.

[0112] In one embodiment, the comprehensive surrounding rock assessment results and their confidence levels are mapped to a GIS or BIM model of the underground project. Based on this, a three-dimensional spatial distribution map of surrounding rock quality or instability risk probability is generated. By spatially correlating and visualizing the assessment results, high-risk areas of the surrounding rock can be more intuitively identified, providing strong support for engineering decision-making.

[0113] After completing the comprehensive evaluation of the surrounding rock mass, the evaluation value and confidence value corresponding to each monitoring point or grid unit can usually be obtained. For example, the underground tunnel or cavern area is divided into several grid units, and each grid unit corresponds to a set of evaluation values and confidence information. GIS models are usually based on geographic coordinates or engineering coordinates, and contain spatial information such as regional topography, landforms, pipelines, and construction area boundaries; BIM models are based on building information modeling technology and contain more detailed three-dimensional physical components, construction stage information, and related structural properties. The comprehensive evaluation results of the surrounding rock mass can be mapped into a GIS or BIM environment to achieve more three-dimensional and comprehensive visualization.

[0114] Align or convert the coordinate system of the evaluation results with the coordinate system in the GIS / BIM model. If the data sources originally belong to different coordinate systems (such as WGS84, UTM or local engineering coordinate system), geographic information transformation or projection algorithm needs to be used to ensure the consistency of spatial position. At the data level, each grid unit or monitoring point needs to be assigned a unique identifier corresponding to the GIS / BIM element (such as feature ID, component ID, etc.). In this way, when a specific unit is retrieved in the GIS or BIM software, the comprehensive evaluation results of the surrounding rock and its confidence level can be quickly located. In order to achieve subsequent visualization, the surrounding rock evaluation results and confidence data need to be mounted on the corresponding spatial objects in the GIS or BIM in the form of attributes. For example, add fields such as "surrounding rock quality score" and "instability risk probability" in the GIS database, and add "surrounding rock comprehensive evaluation value" and "confidence index" in the component properties of the BIM model.

[0115] The spatially associated data is loaded through a professional GIS 3D module or a 3D visualization engine built into the BIM software. The engine can draw the three-dimensional structure of the tunnel, cavern or surrounding rock mass based on the coordinate information in the 3D scene, and display the corresponding attribute information for each grid unit or monitoring point. In the 3D view, in order to better distinguish the quality of the surrounding rock or the level of instability risk, a color gradient map (Color Map) can be introduced. For example: high-risk areas use more eye-catching colors (such as red), indicating that the surrounding rock evaluation value is low or the probability of instability risk is high; medium-risk areas use transition colors (such as yellow); low-risk areas use cool colors (such as blue). The transparency or brightness of the color is corrected based on the confidence information. The higher the confidence level, the more vivid or focused the results will be in the figure. Data with lower confidence levels can be displayed in translucent or light colors. By selecting a grid unit or monitoring point in the three-dimensional scene, the corresponding detailed data such as the surrounding rock evaluation value, instability risk probability and confidence level will pop up; the three-dimensional distribution map can be displayed in layers according to the evaluation threshold or confidence threshold, hiding low-risk areas or highlighting high-risk areas, helping decision makers quickly identify potential risk points; if a new round of real-time monitoring data or model iteration output generates an updated comprehensive evaluation result, the system can automatically refresh the corresponding three-dimensional visualization effect, allowing technical personnel to grasp the changing situation of the surrounding rock at any time.

[0116] In engineering practice, the project team or expert group can formulate a threshold for the probability of instability risk or the surrounding rock quality score based on the surrounding rock stability assessment standards and relevant experience. Once the threshold is exceeded, it will be regarded as a high-risk or concern area. When the system detects that the evaluation result of a certain area exceeds the set threshold, it can add a striking mark (such as a flashing icon, special texture or text prompt) to the area in the three-dimensional distribution map, and trigger an early warning message at the same time (such as displaying a high-risk alarm in the model or sending an alarm to the operator). Engineering management personnel can formulate corresponding inspection and reinforcement measures based on the marked areas in the visual interface and the actual situation of the construction site; if necessary, dense monitoring points can be added in high-risk areas or reinforced support can be carried out to ensure the safety of the project.

[0117] In a large-section highway tunnel project, the comprehensive evaluation results of the surrounding rock obtained by multi-point displacement meters, stress sensors and geological advance forecasts were imported into the GIS platform for standardization, and the risk probability and confidence level of each section within the evaluation period were transferred to the attribute database. Subsequently, the integrated evaluation results were superimposed on the BIM model of the tunnel structure through the three-dimensional visualization function of the BIM software. In the three-dimensional interface, the view can be rotated and roamed, and the quality of the surrounding rock and the risk of instability in different sections of the tunnel can be clearly understood at a glance through the color gradient. For key sections with high confidence and risk scores exceeding the set warning line, an automatic pop-up window will issue an early warning, prompting the project team to install arch supports, strengthen grouting and other support measures in this section.

[0118] In one embodiment, Figure 6 This is a block diagram of a multi-standard intelligent aggregation evaluation system for surrounding rock quality of underground engineering projects according to an exemplary embodiment. Figure 6 As shown, the multi-standard intelligent aggregation evaluation system for surrounding rock quality of underground engineering includes an acquisition module 61, a combination module 62, an adjustment module 63 and a calculation module 64.

[0119] The acquisition module 61 is used to acquire multi-source heterogeneous data related to the surrounding rock of the underground engineering, and perform standardization preprocessing on the multi-source heterogeneous data to obtain standardized surrounding rock data;

[0120] The combination module 62 is used to extract static features based on the standardized surrounding rock data according to a multi-evaluation standard system, and extract dynamic evolution features based on the time series monitoring data in the standardized surrounding rock data, and combine the static features and the dynamic evolution features into a surrounding rock feature set;

[0121] The adjustment module 63 is configured to construct and apply an aggregated evaluation model, receive the surrounding rock feature set and preset context information as input through the aggregated evaluation model, and dynamically adjust the weight of each feature in the surrounding rock feature set according to the context information;

[0122] The calculation module 64 is used to calculate the surrounding rock comprehensive evaluation result and the corresponding confidence level according to the weight of each feature in the surrounding rock feature set and the output of the fusion of multiple basic evaluation models.

[0123] The acquisition module 61, the combination module 62, the adjustment module 63 and the calculation module 64 included in the block diagram of the multi-standard intelligent aggregation evaluation system for the surrounding rock quality of underground engineering are controlled to execute the multi-standard intelligent aggregation evaluation method for the surrounding rock quality of underground engineering described in any of the above embodiments.

[0124] like Figure 7 As shown, the present invention provides an electronic device 700, which includes: a communication interface, a processor 701, and a memory 702;

[0125] Among them, the memory 702 is used to store program instructions. When the program instructions are executed by the processor 701 that is communicatively connected to the memory 702 through the communication interface, the multi-source heterogeneous data related to the surrounding rock of the underground engineering is obtained, and the multi-source heterogeneous data is standardized and preprocessed to obtain standardized surrounding rock data; based on the standardized surrounding rock data, static features are extracted according to a multi-evaluation standard system, and dynamic evolution features are extracted according to the time series monitoring data in the standardized surrounding rock data, and the static features and the dynamic evolution features are combined into a surrounding rock feature set; an aggregated evaluation model is constructed and applied, and the surrounding rock feature set and preset context information are received as input through the aggregated evaluation model, and the weight of each feature in the surrounding rock feature set is dynamically adjusted according to the context information; based on the weight of each feature in the surrounding rock feature set and the output of the fusion of multiple basic evaluation models, the comprehensive evaluation results of the surrounding rock and the corresponding confidence levels are calculated.

[0126] The present invention provides a computer-readable storage medium having computer program instructions stored thereon. When the computer program instructions are executed by a processor, multi-source heterogeneous data related to the surrounding rock of the underground engineering is obtained, and the multi-source heterogeneous data is standardized and preprocessed to obtain standardized surrounding rock data; static features are extracted based on the standardized surrounding rock data according to a multi-evaluation standard system, and dynamic evolution features are extracted based on time series monitoring data in the standardized surrounding rock data, and the static features and the dynamic evolution features are combined into a surrounding rock feature set; an aggregate evaluation model is constructed and applied, and the surrounding rock feature set and preset context information are received as input through the aggregate evaluation model, and the weight of each feature in the surrounding rock feature set is dynamically adjusted according to the context information; and a comprehensive evaluation result of the surrounding rock and the corresponding confidence level are calculated based on the weight of each feature in the surrounding rock feature set and the output of the fusion of multiple basic evaluation models.

[0127] It should be understood that the specific features, operations and details described herein above with respect to the method of the present invention may also be similarly applied to the apparatus and system of the present invention, or vice versa. In addition, each step of the method of the present invention described above may be performed by the corresponding components or units of the apparatus or system of the present invention.

[0128] It should be understood that the various modules / units of the apparatus of the present invention may be implemented in whole or in part by software, hardware, firmware, or a combination thereof. Each module / unit may be embedded in a processor of a computer device in the form of hardware or firmware or may be independent of the processor, or may be stored in a memory of a computer device in the form of software for the processor to call to execute the operations of each module / unit. Each module / unit may be implemented as an independent component or module, or two or more modules / units may be implemented as a single component or module.

[0129] In one embodiment, a computer device is provided, comprising a memory and a processor. The memory stores computer instructions executable by the processor, which, when executed by the processor, instruct the processor to perform the steps of the method according to an embodiment of the present invention. The computer device can be broadly defined as a server, a terminal, or any other electronic device with the necessary computing and / or processing capabilities. In one embodiment, the computer device can include a processor, memory, a network interface, a communication interface, etc. connected via a system bus. The processor of the computer device can be used to provide the necessary computing, processing, and / or control capabilities. The memory of the computer device can include a non-volatile storage medium and an internal memory. An operating system, a computer program, etc. can be stored in or on the non-volatile storage medium. The internal memory can provide an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface and communication interface of the computer device can be used to connect to and communicate with external devices via a network. When the computer program is executed by the processor, the steps of the method according to the present invention are performed.

[0130] The present invention can be implemented as a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, causes the steps of the method of an embodiment of the present invention to be performed. In one embodiment, the computer program is distributed on a plurality of computer devices or processors coupled to a network so that the computer program is stored, accessed, and executed in a distributed manner by one or more computer devices or processors. A single method step / operation, or two or more method steps / operations, can be performed by a single computer device or processor or by two or more computer devices or processors. One or more method steps / operations can be performed by one or more computer devices or processors, and one or more other method steps / operations can be performed by one or more other computer devices or processors. One or more computer devices or processors can perform a single method step / operation, or perform two or more method steps / operations.

[0131] It will be understood by those skilled in the art that the method steps of the present invention can be performed by instructing related hardware such as a computer device or a processor through a computer program, and the computer program can be stored in a non-transitory computer-readable storage medium, which causes the steps of the present invention to be performed when the computer program is executed. Depending on the circumstances, any reference to memory, storage, database or other media in this document may include non-volatile and / or volatile memory. Examples of non-volatile memory include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (PROM), flash memory, magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid-state disk, etc. By introducing a context-aware dynamic weight adjustment mechanism, integrating dynamic monitoring data and adopting a multi-model integrated learning strategy, the objectivity, accuracy, timeliness and robustness of surrounding rock quality evaluation are significantly improved. The limitations of existing surrounding rock evaluation methods, such as strong subjectivity, static evaluation process, empirical weight allocation and difficulty in effectively integrating multi-source heterogeneous information, are overcome. The confidence of the evaluation results is systematically quantified, providing a basis for the reliability of the evaluation conclusion. Examples of volatile memory include random access memory (RAM), external cache memory, etc.

[0132] The various technical features described above can be combined arbitrarily. Although not all possible combinations of these technical features are described, any combination of these technical features should be considered to be covered by this specification as long as such combination does not conflict.

[0133] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-standard intelligent aggregation evaluation method for underground engineering surrounding rock quality, characterized by: include: Acquiring multi-source heterogeneous data related to the surrounding rock of the underground engineering, and performing standardized preprocessing on the multi-source heterogeneous data to obtain standardized surrounding rock data; Extracting static features based on the standardized surrounding rock data according to a multi-evaluation standard system, extracting dynamic evolution features based on time series monitoring data in the standardized surrounding rock data, and combining the static features and the dynamic evolution features into a surrounding rock feature set; Constructing and applying an aggregated evaluation model, receiving the surrounding rock feature set and preset context information as input through the aggregated evaluation model, and dynamically adjusting the weight of each feature in the surrounding rock feature set according to the context information; Based on the weight of each feature in the surrounding rock feature set and the output of the fusion of multiple basic evaluation models, the surrounding rock comprehensive evaluation result and the corresponding confidence level are calculated.

2. The multi-standard intelligent aggregation evaluation method for surrounding rock quality of underground engineering according to claim 1, characterized in that: The dynamically adjusting the weight of each feature in the surrounding rock feature set according to the context information includes: Identify the context information, which includes at least one or more of the surrounding rock geological environment, engineering stage and data quality, and the weight of each feature W i The calculation formula is as follows: W i =B i *(1+CIF i ) Among them, CIF i is the contextual impact factor of feature i, W i is the weight of feature i, B i is the basic weight of feature i; Contextual Impact Factor (CIF) i The calculation formula is as follows: CIF i =α*f(G i )+β*g(P i )+γ*h(Q i ) G i is the geological correlation index of feature i according to the surrounding rock geological environment assessment, P i is the stage importance index of feature i evaluated according to the engineering stage, Q i is the data confidence index of feature i based on the data quality assessment, f, g, and h are preset nonlinear mapping functions used to convert the influence of each geological correlation, stage importance, and data confidence index on the weight, α, β, and γ are the sensitivity coefficients of each context information, and their sum is 1 or set according to prior knowledge.

3. The multi-standard intelligent aggregation evaluation method for surrounding rock quality of underground engineering according to claim 2, characterized in that: The calculation of the surrounding rock comprehensive evaluation result and the corresponding confidence level based on the weight of each feature in the surrounding rock feature set and the output of the fusion of multiple basic evaluation models includes: An integrated learning strategy is used to train and fuse multiple basic evaluation models, wherein the basic evaluation model is based on the surrounding rock feature set and the dynamic weight W i Perform training or weighted fusion; generating the surrounding rock comprehensive evaluation result based on the output of the integrated learning strategy; The confidence level of the evaluation result is quantified based on the consistency or probability distribution of the output of the basic evaluation model.

4. The multi-standard intelligent aggregation evaluation method for surrounding rock quality of underground engineering according to claim 1, characterized in that: The extracting of the dynamic evolution characteristics of the surrounding rock comprises: Extracting time series monitoring data of at least one of surrounding rock displacement, stress or microseismic events from the standardized surrounding rock data; The time series monitoring data are processed using a time series analysis technique, and indicators that can reflect the dynamic evolution of the surrounding rock state are extracted as the dynamic evolution characteristics, and are incorporated into the surrounding rock feature set.

5. The multi-standard intelligent aggregation evaluation method for underground engineering surrounding rock quality according to claim 1 is characterized in that: Also includes: Obtain feedback information on the actual response of surrounding rock in engineering practice; Through the feedback information, the parameters of the constructed aggregation evaluation model are corrected or retrained, and the context influence factor CIF is adjusted. i The sensitivity coefficients α, β, and γ in the formula.

6. The multi-standard intelligent aggregation evaluation method for surrounding rock quality of underground engineering according to claim 1, characterized in that: The calculated surrounding rock comprehensive evaluation results and corresponding confidence levels include: Spatially associating the surrounding rock comprehensive evaluation results and corresponding confidence levels with a Geographic Information System (GIS) model or a Building Information Model (BIM) model of the underground project; Based on the spatial correlation results, a three-dimensional spatial distribution map of the surrounding rock quality or instability risk probability is generated, and high-risk areas are marked in the map.

7. A multi-standard intelligent aggregation evaluation system for surrounding rock quality of underground engineering, characterized by: include: an acquisition module, configured to acquire multi-source heterogeneous data related to the surrounding rock of the underground engineering, and perform standardization preprocessing on the multi-source heterogeneous data to obtain standardized surrounding rock data; a combination module, configured to extract static features based on the standardized surrounding rock data and a multi-evaluation standard system, and extract dynamic evolution features based on time series monitoring data in the standardized surrounding rock data, and combine the static features and the dynamic evolution features into a surrounding rock feature set; an adjustment module, configured to construct and apply an aggregated evaluation model, receive the surrounding rock feature set and preset context information as input through the aggregated evaluation model, and dynamically adjust the weight of each feature in the surrounding rock feature set according to the context information; The calculation module is used to calculate the comprehensive evaluation results of the surrounding rock and the corresponding confidence levels according to the weights of the various features in the surrounding rock feature set and the outputs of multiple basic evaluation models.

8. The multi-standard intelligent aggregation evaluation system for surrounding rock quality of underground engineering according to claim 7, characterized in that: The acquisition module, the combination module, the adjustment module and the calculation module are controlled to execute the multi-standard intelligent aggregation evaluation method for underground engineering surrounding rock quality according to any one of claims 1 to 6.

9. An electronic device, characterized in that: include: Communication interface, processor, memory; In which, the memory is used to store program instructions, and when the program instructions are executed by the processor that is communicatively connected to the memory through the communication interface, the electronic device implements the multi-standard intelligent aggregation evaluation method for underground engineering surrounding rock quality as described in any one of claims 1 to 6.

10. A computer-readable storage medium having program instructions stored thereon, characterized in that: When the program instructions are executed by a computer, the computer implements the multi-standard intelligent aggregation evaluation method for underground engineering surrounding rock quality as described in any one of claims 1 to 6.

Images