Underground cavern group analysis method combined with disturbance effect, program product and equipment

Abstract

The invention provides an underground cavern group analysis method combined with a disturbance effect, a program product and equipment, and relates to the technical field of engineering data processing. The method comprises the following steps: acquiring monitoring data of a plurality of indexes of an underground cavern group in a construction process; according to the stress parameter of the target cavern disturbed by the adjacent cavern and the working procedure crossing duration of the target cavern and other caverns, the comprehensive disturbance effect intensity is determined; taking the monitoring data and the comprehensive disturbance effect intensity as first time-varying risk factors, screening out second time-varying risk factors from the first time-varying risk factors based on the correlation between the first time-varying risk factors and time-varying progress risks, and forming a time-varying risk factor sequence according to the second time-varying risk factors of different processes; and processing the time-varying risk factor sequence by using a time sequence model with an attention mechanism to obtain an analysis result of the target cavern. The accuracy of the analysis result of the underground cavern group is improved.

Description

Technical Field

[0001] This disclosure relates to the field of engineering data processing technology, and more specifically, to methods, program products, and equipment for analyzing underground cavern groups incorporating disturbance effects. Background Technology

[0002] In large-scale projects such as hydropower stations, underground cavern complexes are typically designed and constructed to serve as core functional spaces for the main powerhouse, main transformer room, and tailrace tunnel. Analyzing the state of these underground cavern complexes, such as the stress and deformation states of each cavern, can guide project management, enabling the optimization of construction plans or the implementation of preventative measures when potential disaster risks are identified.

[0003] In related technologies, most analysis schemes for underground cavern groups focus on a single cavern, resulting in analysis results that cannot accurately depict the actual state of the cavern and are difficult to meet the needs of refined and dynamic engineering management for underground cavern groups. Summary of the Invention

[0004] This disclosure provides a method, program product, and equipment for analyzing underground cavern groups by incorporating disturbance effects, in order to at least partially solve the technical problem of inaccurate results in the analysis of underground cavern groups.

[0005] According to a first aspect of this disclosure, a method for analyzing underground cavern groups incorporating disturbance effects is provided. The method includes: acquiring monitoring data of multiple indicators of the underground cavern group during construction; the construction process includes multiple procedures; the underground cavern group consists of multiple caverns, including a target cavern; the monitoring data of the multiple indicators includes stress parameters; determining the comprehensive disturbance effect intensity based on the stress parameters of the target cavern affected by disturbances from adjacent caverns and the procedure overlap duration between the target cavern and other caverns; the procedure overlap duration refers to the duration during which the target cavern and other caverns are in different procedures; using the monitoring data and the comprehensive disturbance effect intensity as a first time-varying risk factor, selecting a second time-varying risk factor from the first time-varying risk factor based on the correlation between the first time-varying risk factor and time-varying progress risk, and forming a time-varying risk factor sequence based on the second time-varying risk factors for different procedures; processing the time-varying risk factor sequence using a time-series model with an attention mechanism to obtain the analysis results for the target cavern.

[0006] According to a second aspect of this disclosure, an analysis device for underground cavern groups incorporating disturbance effects is provided. The device includes: a monitoring data acquisition module configured to acquire monitoring data of multiple indicators of the underground cavern group during construction; the construction process includes multiple procedures; the underground cavern group consists of multiple caverns, including a target cavern; the monitoring data of the multiple indicators includes stress parameters; and a disturbance effect processing module configured to determine the intensity of a comprehensive disturbance effect based on the stress parameters of the target cavern affected by disturbances from adjacent caverns and the time interval between procedures between the target cavern and other caverns; the procedures... The crossover duration refers to the duration during which the target cavern and other caverns are in different processes; the time-varying risk factor processing module is configured to use the monitoring data and the intensity of the comprehensive disturbance effect as the first time-varying risk factor, and based on the correlation between the first time-varying risk factor and the time-varying progress risk, to select a second time-varying risk factor from the first time-varying risk factor, and to form a time-varying risk factor sequence according to the second time-varying risk factors of different processes; the analysis result determination module is configured to process the time-varying risk factor sequence using a time series model with an attention mechanism to obtain the analysis result of the target cavern.

[0007] According to a third aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the method of the first aspect described above and possible implementations thereof.

[0008] According to a fourth aspect of this disclosure, an electronic device is provided, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform the method of the first aspect and possible implementations thereof by executing the executable instructions.

[0009] The technical solution disclosed herein has the following beneficial effects: On the one hand, by considering and quantifying both geological disturbances and process-related disturbances in the underground cavern group, coupled disturbance factors are incorporated into the cavern analysis process, enabling the analysis results to accurately match the actual state of the underground cavern group and improving the accuracy of the analysis results. On the other hand, by screening time-varying risk factors and combining them with attention mechanisms and time-series models, both global characteristics and temporal patterns are taken into account, ensuring that the analysis results meet engineering requirements and providing a basis for dynamic and refined management of cavern digital modeling, construction control, and other related operations. Attached Figure Description

[0010] Figure 1 A flowchart illustrating an underground cavern group analysis method incorporating disturbance effects is shown in this disclosure embodiment; Figure 2 A flowchart illustrating one embodiment of the present disclosure for determining the intensity of a combined disturbance effect is shown. Figure 3 A flowchart illustrating one method for determining analysis results according to an embodiment of this disclosure is shown; Figure 4 This diagram illustrates a process architecture according to an embodiment of the present disclosure; Figure 5 This diagram illustrates the parameter curves in the embodiments of this disclosure. Figure 6 This diagram illustrates the model loss curve in an embodiment of the present disclosure. Figure 7 A schematic diagram of an underground cavern group analysis device incorporating disturbance effects is shown in an embodiment of this disclosure; Figure 8 A schematic diagram of an electronic device according to an embodiment of the present disclosure is shown. Detailed Implementation

[0011] Exemplary embodiments of this disclosure will be described more fully below with reference to the accompanying drawings.

[0012] The accompanying drawings are schematic illustrations of this disclosure and are not necessarily drawn to scale. Some block diagrams shown in the drawings may be functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in hardware modules or integrated circuits, or in networks, processors, or microcontrollers. Implementations can be carried out in various forms and should not be construed as limited to the examples set forth herein. The features, structures, or characteristics described in this disclosure can be combined in any suitable manner in one or more implementations. In the following description, numerous specific details are provided to give a thorough description of one embodiment of this disclosure. However, those skilled in the art will recognize that one or more specific details may be omitted when implementing the technical solutions provided in an embodiment of this disclosure, or other methods, components, apparatuses, steps, etc., may be used to replace one or more specific details.

[0013] Large-scale underground cavern complexes are characterized by a large number of caverns, dense spatial layout, overlapping construction procedures, and complex and variable geological conditions. During the excavation and construction of underground cavern complexes, in addition to the geological, equipment, and procedural risks of individual caverns, there is also the disturbance effect of the cavern complex coupling. This effect refers to the superposition of construction disturbances in different caverns and the mutual influence of the connection between procedures, resulting in time-varying, correlated, and nonlinear characteristics of schedule risks.

[0014] In related technologies, the approach of analyzing each individual cavern in an underground cavern group neglects the disturbance effect, resulting in analysis results that cannot accurately depict the actual state of the caverns and thus fail to meet the needs of refined and dynamic engineering management for underground cavern groups. Furthermore, these technologies also have the following problems: Poor adaptability of monitoring data processing methods: During the monitoring of underground cavern groups, monitoring data may be missing. Related technologies do not fill in the missing data or only fill in the missing data with low precision. This makes it difficult to provide high-quality dynamic monitoring data and cannot support the analysis and research of underground cavern groups.

[0015] Insufficient model performance: Although some related technologies have introduced machine learning models, most of them use a single model or a simple improvement of the model. They lack the fusion design of global feature extraction and temporal dependency capture, resulting in defects such as slow convergence and low inference accuracy, which makes it difficult to meet the high-precision analysis requirements in engineering.

[0016] The dynamic management and control strategies are not targeted enough: most of the existing management and control strategies are static or simple dynamic management and control, which are not well integrated with the time-varying characteristics of schedule risks and the forecast results. The management and control measures lack targeting and flexibility, making it difficult to achieve accurate and dynamic optimization management and control of schedule risks.

[0017] In view of one or more of the above-mentioned problems, this disclosure provides an analysis method for underground cavern groups that incorporates disturbance effects, which can be used to analyze the state of underground cavern groups in large-scale projects such as hydropower stations. The underground cavern group can be a group of caverns during the excavation and construction process. The analysis results of this disclosure include, but are not limited to: comprehensive mechanical state analysis results of the caverns, cavern stability analysis results, digital model data of the caverns (which can be used for underground cavern group modeling), construction stability analysis results, construction progress risk prediction results, disaster prediction results, etc.

[0018] Figure 1 An exemplary workflow for analyzing underground cavern groups is shown, including the following steps: S110, acquire monitoring data of multiple indicators during the construction of the underground cavern group; the construction process includes multiple procedures; the underground cavern group consists of multiple caverns, including the target cavern; the monitoring data of multiple indicators includes stress parameters; S120, the intensity of the comprehensive disturbance effect is determined based on the stress parameters of the target cavern being disturbed by adjacent caverns and the process overlap time between the target cavern and other caverns; the process overlap time refers to the duration during which the target cavern and other caverns are in different processes; S130, using monitoring data and the intensity of comprehensive disturbance effects as the first time-varying risk factor, and based on the correlation between the first time-varying risk factor and time-varying schedule risk, a second time-varying risk factor is selected from the first time-varying risk factor, and a time-varying risk factor sequence is formed according to the second time-varying risk factors of different processes. S140 uses a time series model with attention mechanism to process the time-varying risk factor sequence to obtain the analysis results of the target cavern.

[0019] Based on the above methods, on the one hand, by considering and quantifying both geological disturbances and process-related disturbances in the underground cavern group, coupled disturbance factors are incorporated into the cavern analysis process, enabling the analysis results to accurately match the actual state of the underground cavern group and improving the accuracy of the analysis results. On the other hand, by screening time-varying risk factors and combining them with attention mechanisms and time-series models, both global characteristics and temporal patterns are taken into account, ensuring that the analysis results meet engineering requirements and providing a basis for the dynamic and refined management of cavern digital modeling, construction control, and other related operations.

[0020] The following describes, in conjunction with one or more embodiments and related accompanying drawings, Figure 1 Each step in the process will be explained in detail.

[0021] refer to Figure 1 In step S110, monitoring data of multiple indicators of the underground cavern group during the construction process are obtained; the construction process includes multiple procedures; the underground cavern group consists of multiple caverns, including the target cavern; the monitoring data of multiple indicators include stress parameters.

[0022] In this context, the target cavern is the cavern being analyzed, and can be any cavern in an underground cavern group. If it is necessary to control each cavern in the underground cavern group, each cavern can be used as the target cavern, and the method of this disclosed embodiment can be used for analysis to obtain the analysis results for each cavern. For example, in the underground cavern group of a large hydropower station, key caverns such as the main powerhouse, main transformer room, tailrace tunnel, and water diversion tunnel can be used as target caverns for analysis.

[0023] The construction process of underground cavern complexes involves multiple steps, such as excavation, support, lining, and equipment installation. Monitoring data can include relevant status monitoring data of the underground cavern complex, such as stress parameters (rock mass stress, maximum principal stress around the cavern, etc.) and groundwater flow, as well as construction monitoring data during the construction process, such as the time period of each step, the duration of overlapping steps (the length of time the target cavern is in a different step from other caverns), excavation speed, and support strength. Monitoring data can be collected by sensors installed in the underground cavern complex, or obtained from information such as construction records, which include the construction time of each step, thus determining the time period of each step.

[0024] In one implementation, the total time period of the construction process is denoted as T = {t1, t2, …, t}. n}, where t i Indicates the first i Process duration, the unit can be days (d). n This indicates the number of processes. Monitoring data can be collected according to the time period of each process.

[0025] In one implementation, a tiered, lightweight data acquisition mode can be adopted, identifying mandatory measurement indicators, which are all core parameters corresponding to the quantification of disturbance effects, such as stress parameters, process time periods, process overlap durations, groundwater flow, excavation speed, and support strength—all real-time dynamic data. For monitoring data of other indicators, the acquisition intensity can be appropriately reduced, such as by setting longer acquisition cycles.

[0026] In one implementation, after acquiring monitoring data, for missing data, supplementary data can be determined by utilizing similar monitoring data from adjacent processes in the process containing the missing data, as well as different types of monitoring data that have an index coupling relationship with the missing data. During the data acquisition process, data gaps are inevitable. Missing data can refer to blank data in the monitoring data. This implementation determines the corresponding supplementary data through calculation and inference. For example, if the stress parameter of process 3 is missing, the actual stress parameters of processes 2 and 4 can be obtained, along with monitoring data of other indicators that have an index coupling relationship with the stress parameter. That is, monitoring data related to the missing data can be obtained from the longitudinal dimension of time series and the horizontal dimension of different related indicators. By interpolating these monitoring data, supplementary data for filling the gaps can be determined. This approach balances the temporal continuity and index coupling of the monitoring data, improving the accuracy of data filling.

[0027] The coupling relationship between different indicators refers to the correlation between the monitoring data of different indicators, which can be determined through correlation analysis. For example, by determining the coupling coefficient between different indicators through correlation analysis, the magnitude of the coupling coefficient can be used to determine the coupling relationship between rock mass stress and groundwater flow and surrounding rock type, and the coupling relationship between excavation speed and process overlap time, and so on.

[0028] In one implementation, the filling data can be determined using the following formula: (1) in, x i,j Indicates the first i In the process of the first j Fill in the data for each indicator; ω t , ω c These represent time-series correlation weights and indicator coupling weights, respectively. x i-1,j , x i+1,j Indicates the first i -1、 i +1 process jMonitoring data (measured values) for each indicator; t i , t i-1 , t i+1 They represent the first i -1 process, the first i Process, Number i The time period corresponding to +1 process; q Indicates the relationship with the first j The first indicator has an indicator coupling relationship. q One indicator, s Indicates the relationship with the first j The number of indicators that have an indicator coupling relationship; ρ j,q Indicates the first j The first indicator and the first q The coupling coefficient of an indicator represents the degree of correlation between two indicators and can be determined through correlation analysis and other methods. x i,q Indicates the first i In the process of the first q Monitoring data for each indicator (actual measured values, no missing values, with priority given to mandatory indicators).

[0029] Temporal correlation weight ω t Indicator Coupling Weights ω c This refers to the proportion of time-series-related monitoring data and indicator-related monitoring data when calculating supplementary data, with the sum of the two being 1. The specific values ​​of these two proportions can be determined based on experience or specific needs, such as... ω t =0.6~0.7, corresponding ω c =0.3~0.4, if priority needs to be given to the time series continuity of monitoring data, it can be set to 0.3~0.4. ω t The corresponding value is 0.7. ω c The value is 0.3. To strengthen the coupling and adaptation between indicators, it can be set to 0.3. ω t The corresponding value is 0.6. ω c It is 0.4. Compared to the first... j The number of indicators that have an indicator coupling relationship s The value can be determined based on the coupling perturbation quantization requirements. For example, if the perturbation quantization requirements are high, a larger value can be set. s Value. In one implementation, the coupling coefficient can be set based on experience or specific needs. ρ j,qThe threshold is used to filter out those that match the first... j Indicators whose coupling coefficient is not less than the threshold are included in the calculation of formula (1), while for those with coupling coefficients not less than the threshold, the coefficients are not included in the calculation of formula (1). j If the coupling coefficient of an indicator is less than a threshold, it can be considered that there is no indicator coupling relationship.

[0030] By using formula (1), considering both temporal coupling and index coupling factors of the cavern monitoring data, a dual-dimensional data filling process is performed for the missing data, realizing an adaptive interpolation algorithm (the interpolation result is adaptive to the temporal coupling of the missing data process and the index coupling of the missing data index with other indicators). Compared with the traditional linear interpolation method, it considers more comprehensive factors, improves the accuracy of data filling, and retains the data temporal trend caused by construction disturbances, which is conducive to improving the subsequent quantitative accuracy.

[0031] In one implementation, the filled data does not require secondary processing and can be used directly.

[0032] In one implementation, in addition to the data imputation described above, data preprocessing can be performed in other ways, including but not limited to data deduplication and outlier removal. This disclosure does not limit the scope of the application.

[0033] In one implementation, the aforementioned monitoring data is dynamic monitoring data. Dynamic monitoring data refers to monitoring data collected in real time during construction or operation of the underground cavern complex. Furthermore, monitoring locations and configurations can be dynamically adjusted according to changes in the underground cavern complex, resulting in the collection of dynamic monitoring data.

[0034] Continue to refer to Figure 1 In step S120, the intensity of the comprehensive disturbance effect is determined based on the stress parameters of the target cavern being disturbed by adjacent caverns and the process overlap time between the target cavern and other caverns; the process overlap time refers to the time during which the target cavern and other caverns are in different processes.

[0035] For example, if other caverns are in other processes (such as process 2, process 4, etc.) while the target cavern is in process 3, it indicates that there is a process overlap between the target cavern and other caverns, and this part of the time is called the process overlap time. In this embodiment, the disturbance effect is mainly divided into two aspects: geological disturbance effect, which can be understood as the disturbance between adjacent caverns, manifested as changes in stress parameters, etc.; and process overlap disturbance effect, which can be understood as the disturbance caused by the construction overlap between different caverns, which can be quantified by process overlap time, etc. Based on the stress parameters under geological disturbance and the process overlap time, the intensity of the comprehensive disturbance effect is determined, which includes the quantified results of the above two aspects of disturbance effect.

[0036] In one implementation, reference Figure 2 As shown, the determination of the comprehensive disturbance effect intensity based on the stress parameters of the target chamber disturbed by adjacent chambers and the time of process overlap between the target chamber and other chambers includes the following steps S210 to S230: S210. Determine the intensity of the geological disturbance effect of the target cavern based on the stress parameters of the target cavern affected by the disturbance of adjacent caverns and the stress parameters of the target cavern unaffected by the disturbance of adjacent caverns.

[0037] The geological disturbance effect originates from the stress superposition caused by the excavation of adjacent caverns. The intensity of the geological disturbance effect can be determined based on the difference between the stress parameters of the target cavern affected by disturbances from adjacent caverns and those unaffected by disturbances from adjacent caverns. For example, the intensity of the geological disturbance effect can be determined using the following formula: ; (2) in, D g Indicates the intensity of geological disturbance effects; m The number of adjacent chambers of the target chamber can be determined based on the chamber layout defined by the study area and the excavation status of the chambers under dynamic monitoring. σ k,i Indicates the first i During the process, the target cavern is subjected to the first k Stress parameters after disturbance of adjacent chambers; σ 0,i Indicates the first i During the process, the stress parameters of the target chamber when it is not disturbed by adjacent chambers; ω k Indicates the first k Geological disturbance weighting coefficients of adjacent caverns, L k Indicates the target chamber and the first k The distance between adjacent chambers, such as the center-to-center distance between two chambers.

[0038] In one implementation, the stress parameter in formula (2) can be the maximum principal stress. Specifically, when the target cavity is at the first i During the process, obtain the first k The actual maximum principal stress of the target cavern after the excavation disturbance of adjacent caverns is used as σ k,i The unit can be MPa. Accordingly, σ 0,i This can indicate that the target chamber is located at the first... iDuring the process, the initial maximum principal stress under conditions of no disturbance from adjacent caverns can also be expressed in MPa. The initial maximum principal stress can be calculated by dynamically monitoring geological parameters (such as rock mass integrity coefficient, surrounding rock type, etc.) of the target cavern and combining them with relevant elasticity models or formulas.

[0039] In one implementation, the intensity of the geological disturbance effect D g The value of can be in the range of [0,1]. The larger the value, the more significant the impact of geological disturbance on the construction progress, which in turn leads to a higher risk of time-varying progress.

[0040] Formula (2) is based on the principle of stress superposition in elastic mechanics and combines the dynamic adjustment weight of the cavity spacing. The quantitative results are consistent with the actual stress state of the rock mass, providing reliable geological dimension data for comprehensive disturbance calculation.

[0041] S220, determine the intensity of the process crossover disturbance effect of the target cavern based on the process crossover time of the target cavern and other caverns.

[0042] The process overlap disturbance effect originates from the overlapping interference of construction processes in different caverns, and can be calculated and quantified by the duration of process overlap. For example, the intensity of the process overlap disturbance effect can be determined using the following formula: (3) in, D p Indicates the intensity of the cross-process disturbance effect; T overlap,i Indicates the target chamber is in the first i The duration of overlap between processes in the process and those in other caverns; T total,i Indicates the first i Total duration of the process; p Indicates the first [number] in the target chamber i The number of overlapping processes that intersect with the target cavern in the process; I l,i Indicates the first [number] in the target chamber i In the process, the first l The construction intensity of each overlapping process; I max,i Indicates the first [number] in the target chamber i The maximum construction intensity among all overlapping processes in the process; β l For the first lThe priority coefficient of each cross-process can be determined and adjusted based on the actual construction situation of the underground cavern group. For example, for key caverns, the priority coefficient of core processes (such as main plant excavation and main transformer room support) is 1.0, and the priority coefficient of auxiliary processes (such as drainage ditch construction and temporary support) is 0.3~0.7.

[0043] The specific indicators of construction intensity can be determined according to the process. For example, in the excavation process, the construction intensity can be the excavation volume per unit time, expressed in cubic meters per day (m³ / d). In the support process, the construction intensity can be the support work volume per unit time, expressed in kilonewtons per day (kN / d).

[0044] In one implementation, the intensity of the process cross-disturbance effect D p The value of can be in the range of [0,1]. The larger the value, the more significant the impact of cross-process interference on the construction progress, and the higher the risk of time-varying progress.

[0045] Formula (3) takes into account three factors: the duration of process overlap, construction intensity, and process priority. The quantitative results are consistent with the actual construction organization of underground cavern groups and can accurately reflect the interference of process overlap on construction, providing reliable process dimension data for comprehensive disturbance calculation.

[0046] S230, the comprehensive disturbance effect intensity is determined based on the intensity of geological disturbance effect and the intensity of process cross-disturbance effect.

[0047] Among these, geological disturbance effects and process cross-disruption effects are coupled and synergistically influence time-varying schedule risks. The combined disturbance effect intensity can be obtained by adding, multiplying, or weighting these effects. For example, the combined disturbance effect intensity can be determined using the following formula: (4) in, D total,i Indicates the first i The intensity of the overall disturbance effect of the process can be in the range of [0,1]. D g,i , D p,i They represent the first i The intensity of geological disturbance effect and the intensity of cross-process disturbance effect; α , β These are the weighting coefficients corresponding to the intensity of geological disturbance effect and the intensity of process cross-disturbance effect, respectively, satisfying... α + β =1. In one implementation, the value can be determined based on experience and the construction characteristics of the underground cavern complex. α , βFor example, the weighting coefficient corresponding to the intensity of geological disturbance effects can be set to be slightly higher, taking... α =0.5~0.6, corresponding to the weighting coefficients for the intensity of process cross-disturbance effects. β= The value can be adjusted from 0.4 to 0.5, and can also be finely adjusted according to specific engineering geological conditions, thereby adapting to different engineering scenarios, realizing the unified quantification of the dual coupling disturbance of geology and process, and fully characterizing the overall characteristics of the coupling disturbance of underground cavern groups.

[0048] Based on the above methods, the disturbance effect is decomposed into two independent effects and quantified separately, avoiding the one-sidedness of single disturbance analysis, accurately depicting the dual coupling characteristics of stress superposition and process interference in underground cavern groups, and providing accurate disturbance effect data for further analysis.

[0049] Continue to refer to Figure 1 In step S130, monitoring data and the intensity of comprehensive disturbance effects are used as the first time-varying risk factor. Based on the correlation between the first time-varying risk factor and the time-varying schedule risk, a second time-varying risk factor is selected from the first time-varying risk factor, and a time-varying risk factor sequence is formed according to the second time-varying risk factors of different processes.

[0050] Time-varying schedule risk refers to the risk of abnormal construction progress. Time-varying risk factors are those correlated with time-varying schedule risk. The first time-varying risk factor, initially determined, is composed of one or more indicators from monitoring data combined with the intensity of the overall disturbance effect. Based on the correlation between the first time-varying risk factor and time-varying schedule risk, factors strongly correlated with time-varying schedule risk are selected, while weakly correlated factors are eliminated, resulting in the second time-varying risk factor. The second time-varying risk factors are then arranged according to the temporal sequence of different work processes, forming a time-varying risk factor sequence.

[0051] In one implementation, the first time-varying risk factor can be determined based on the actual construction conditions of the underground cavern group. For example, in a hydropower station scenario, the first time-varying risk factor includes: the intensity of the comprehensive disturbance effect. D total,i Groundwater flow Flow i Rock mass integrity coefficient K v,i Excavation speed v i Support strength S i Equipment availability rate η i The intensity of the comprehensive disturbance effect is calculated using formula (4), while other parameters are derived from dynamically collected monitoring data. The first time-varying risk factor is represented in vector form. Xi =[ X i1 , X i2 , … , X iy ], X i Indicates the first i The first time-varying risk factor vector of the process, X i1、 X i2 The order indicates the first i The values ​​of the first and second time-varying risk factor vectors in the process. y This indicates the number of the first time-varying risk factors (dimensionless).

[0052] In one implementation, the process of selecting a second time-varying risk factor from the first time-varying risk factor based on the correlation between the first time-varying risk factor and the time-varying schedule risk includes the following steps: The correlation coefficient between the first time-varying risk factor and time-varying schedule risk is determined using the following formula: (5) in, r j For the first j The first time-varying risk factor and time-varying schedule risk Y The correlation coefficient; X ij Indicates the first i In the process of the first j The value of the first time-varying risk factor vector; For the first j The mean of the first time-varying risk factor; Y i For the first i Time-varying schedule risk of the process; This represents the average value of time-varying schedule risk. The first time-varying risk factor with a correlation coefficient not less than the correlation threshold is selected as the second time-varying risk factor.

[0053] In one implementation, time-varying schedule risk can be characterized by one or more of the following parameters: delay duration, such as the first... i The delay time of the process is denoted as Δ. T i The unit can be days (d); the probability of delay, such as the first day. i The probability of delay in a process is denoted as P iThese parameters are dimensionless. They can be obtained from dynamic monitoring construction logs. One of these parameters can be directly used as the time-varying schedule risk, or multiple parameters can be combined and calculated (such as weighted average) to obtain the time-varying schedule risk. This disclosure does not limit this.

[0054] In one implementation, a relevance threshold can be set based on experience or specific needs, such as setting it to 0.5 to filter out | r j The first time-varying risk factor with a value ≥ 0.5 is used as the second time-varying risk factor.

[0055] The above method eliminates weakly correlated factors in the first time-varying risk factor through correlation analysis, ensuring that the second time-varying risk factor is strongly correlated with the time-varying progress risk. This helps reduce the redundancy of subsequent model inputs, and all factors come from dynamic monitoring data, avoiding the introduction of static or irrelevant parameters.

[0056] In one embodiment, the above-mentioned formation of a time-varying risk factor sequence based on the second time-varying risk factors of different processes includes the following steps: Obtain actual monitoring data of the second time-varying risk factor for the executed process, and obtain predicted data of the second time-varying risk factor for the current process; the executed process includes the current process; By weighting the actual monitoring data of the second time-varying risk factor in the current process with the predicted data of the second time-varying risk factor in the current process, the predicted data of the second time-varying risk factor in the next process is obtained. A time-varying risk factor sequence is formed based on the actual monitoring data of the second time-varying risk factor of the executed process and the predicted data of the second time-varying risk factor of the next process.

[0057] For example, a time-varying risk factor update model can be constructed, and the predicted data for the second time-varying risk factor can be obtained using this model. The time-varying risk factor update model is based on dynamically monitored data, employs the basic concept of exponential smoothing, considers the dynamic evolution trend of risk factors, and predicts and updates risk factors for future periods, ensuring the time-varying nature and accuracy of the data. The time-varying risk factor update model can be referenced by the following formula: (6) in, Indicates the first i +1 process j The predicted data of the second time-varying risk factor can be used as input parameters in the time-varying risk factor update model in the subsequent prediction process. α j For the first jThe smoothing coefficient of the second time-varying risk factor can be determined based on experience or specific needs. The value range can be [0,1]. The larger the value, the higher the weight of the actual monitoring data. Indicates the first i In the process of the first j Actual monitoring data for the second time-varying risk factor; Indicates the first i In the process of the first j The forecast data for the second time-varying risk factor is derived from the previous process (the first... i Output of the time-varying risk factor update model in process -1)

[0058] In one implementation, the initial prediction data of the time-varying risk factor update model, such as... , The monitoring data after preprocessing in the first process (such as the filling data mentioned above) can be used, or the prediction data can be determined by other prediction methods. The prediction data of subsequent processes is determined by iterative updates using the above formula (6). In this way, the entire process relies on dynamic monitoring data to achieve time-varying adaptation and accurate updating of risk factors, laying the foundation for subsequent analysis.

[0059] In one implementation, when constructing a time-varying risk factor sequence, for the second time-varying risk factor of an executed process, its actual monitoring data is used. If there is a lack of actual monitoring data, fill data or predicted data can be used. For the second time-varying risk factor of an unexecuted process (such as the next process), predicted data can be used. These data are arranged in the order of each process to form a time-varying risk factor sequence.

[0060] Continue to refer to Figure 1 In step S140, the time-varying risk factor sequence is processed using a time series model with attention mechanism to obtain the analysis results of the target cavern.

[0061] Among them, the time series model with attention mechanism is a machine learning model based on the time series model, incorporating an attention mechanism. For example, an attention mechanism layer can be integrated into time series models such as LSTM (Long Short Term Memory) and GRU (Gated Recurrent Unit) and trained to obtain a time series model with attention mechanism. The time-varying risk factor sequence is input into the model for processing to obtain the analysis results of the target cavern.

[0062] In one implementation, the temporal model with attention mechanism includes a Transformer, a gated recurrent unit, a fully connected layer, and an output layer. (Reference) Figure 3As shown, the above-mentioned time-varying risk factor sequence is processed using a time-series model with attention mechanism to obtain the analysis results of the target cavern, including the following steps S310 to S340: S310, using Transformer to encode the time-varying risk factor sequence, obtains the first intermediate feature.

[0063] In this embodiment, the time-series model with attention mechanism is a machine learning model (TIGRU) that integrates Transformer and Improved Gated Recurrent Unit (IGRU). In the input layer of TIGRU, the input data can be a normalized, dynamically updated sequence of time-varying risk factors. X =[ X 1, X 2, …, X n ],in X i ∈R 1×y' ( y' (The number of the second time-varying risk factors), the length of the input sequence is n (Total number of processes).

[0064] The Transformer can act as an encoder to encode the input time-varying risk factor sequence, achieving global feature extraction. For example, the Transformer can extract the global correlation between the second time-varying risk factor and the disturbance effect features of underground cavern groups, providing comprehensive feature support for subsequent analysis. The main structure of the Transformer includes attention mechanism layers (such as self-attention mechanism and multi-head attention mechanism layers), and the calculation formula is as follows: (7) in, Q , K , V These are the query matrix, key matrix, and value matrix, each with a dimension of [missing information]. n ×dim k , from the input sequence X Obtained through weight matrix transformation; dim k The attention head dimension can be 16, 32, etc.; QK T This represents the transpose of the query matrix and the key matrix, used to calculate the correlation between features; softmax is the activation function that maps the correlation to weights, and its value can be [0,1], ensuring that the sum of the weights is 1. After processing the input time-varying risk factor sequence, the Transformer outputs the first intermediate feature.

[0065] S320 uses a gated loop unit to serialize the first intermediate feature to obtain the second intermediate feature.

[0066] In one implementation, a basic GRU can be used to process the first intermediate feature, which can further extract serialization information, explore temporal correlations, and output a second intermediate feature.

[0067] In one implementation, the above-described serialization process of the first intermediate feature using a gated loop unit to obtain the second intermediate feature includes the following steps: In the reset gate of the gated loop unit, the reset gate output characteristics are obtained using the following formula: (8) in, res i Indicates the first i The reset gate output feature of the process can take values ​​in the range of [0,1]. W r , b r Represents the weight matrix and bias vector of the reset gate; Z i This represents the first output of the Transformer. i The first intermediate characteristic of the process; h i-1 Indicates the first i -1 hidden layer output features; γ This represents the adaptive adjustment coefficient, which can be determined based on the actual construction conditions of the underground cavern group and the characteristics of dynamic monitoring data; for example, it can be 0.1~0.3; avg( ) represents the mean calculation function; the sigmoid activation function maps the input (linear combination result) of the adaptive reset gate to the interval [0,1], so that the reset gate coefficients can flexibly respond to global coupling perturbation features.

[0068] In the hidden door of the gated loop unit, the candidate hidden state is obtained using the following formula: (9) in, Indicates the first i Candidate hidden state of the process; W h 、b h Represents the weight matrix and bias vector of the hidden gate; This represents the element-wise multiplication operation of matrices, used to achieve feature fusion and updating; tanh( ) represents the hyperbolic tangent function.

[0069] Based on the candidate hidden state and the first i The hidden layer output features of step -1 are obtained to obtain the... i The hidden gate output features of the process; such as updating the gate to match the candidate hidden state with the first... i The hidden layer output features of step -1 are weighted (the weights can be trainable parameters) to obtain the th step. i Hidden door output features of the process.

[0070] The above process of resetting the gate, hiding the gate, and obtaining the output feature of the hidden gate can be implemented based on the IGRU decoder. This decoder is used to decode the first intermediate feature using mechanisms such as adaptive reset gate to obtain the output feature of the hidden gate.

[0071] Next, based on the gated attention mechanism, the hidden gate output features and the second time-varying risk factor are fused using the following formula to obtain the second intermediate feature: ; (10) in, e ij Indicates the first i The hidden door output characteristics of the process and the first j The correlation coefficient of the second time-varying risk factor; W a , b a Represents the weight matrix and bias vector of the gating attention mechanism; h i Indicates the first i Hidden door output features of the process; U a The weight matrix represents the correlation between the second time-varying risk factor and the disturbance effect; d j Indicates the first j The correlation between the second time-varying risk factor and the disturbance effect; α ij Indicates the first i In the process of the first j Attention weights for the second time-varying risk factor; g i Indicates the output characteristics of the gating unit; W g , b g Here are the weight matrix and bias vector of the gated unit; The second intermediate feature is represented by formula (10). Formula (10) can be implemented based on the gated attention mechanism layer. In the gated attention mechanism layer, based on the gated attention mechanism, formula (10) is used to fuse the hidden gate output feature and the second time-varying risk factor to obtain the second intermediate feature.

[0072] In one implementation, the intensity of the combined disturbance effect can be set. D total,i corresponding d j =1, other second time-varying risk factors are based on correlation coefficients. r j It can be determined that, for example, it can be equal to the correlation coefficient and the intensity of the combined disturbance effect corresponding to other second time-varying risk factors. D total,i The ratio of the corresponding correlation coefficients.

[0073] Through the above IGRU and gated attention mechanism layers, two optimization logics were adopted: adaptive reset gate (as shown in the reference formula (8)) and residual connection (as shown in the reference formula (10)). This achieved an adaptive residual enhancement mechanism that integrates global coupling perturbation features, thereby improving the quality of the output second intermediate feature.

[0074] S330 uses a fully connected layer to process the second intermediate feature to obtain the third intermediate feature.

[0075] For example, the second intermediate feature is input into the fully connected layer for feature fusion and dimensionality transformation. Specifically, high-dimensional features can be mapped to low-dimensional features related to time-varying progress risk, as shown in the following formula: (11) in, z i This represents the third intermediate feature, whose dimensions can be set according to the analysis requirements, such as the need to analyze the target cavity. f The analysis results of each dimension, z i The dimension can be f In one implementation, z i The dimensions are 2, corresponding to the time-varying progress risk level and the predicted delay duration, respectively. W z This is the weight matrix of the fully connected layer. b z This is the bias vector for the fully connected layer.

[0076] S340 uses the output layer to process the third intermediate feature to obtain the analysis results.

[0077] For example, the third intermediate feature can be directly output as the final analysis result, or the third intermediate feature can be further processed (such as activation) to obtain the analysis result.

[0078] In one implementation, the analysis results include comprehensive mechanical state analysis results of the target cavern. Based on the mechanical properties of the surrounding rock and the stress characteristics of the support structure, it integrates the combined effects of stress superposition disturbance between caverns in the underground cavern group and the disturbance caused by cross-process construction, comprehensively reflecting the true and dynamic overall mechanical performance of the target cavern throughout the entire construction cycle or operation period. For example, the comprehensive mechanical state analysis results can be dynamic, multi-dimensional, and practical engineering mechanical information, including but not limited to the following data: Real-time stress state data of the cavern, such as the real-time values ​​of the maximum and minimum principal stresses around the target cavern, the location, range and degree of stress concentration areas, the stress superposition increase caused by geological disturbances in adjacent caverns, and the comparison results between the cavern stress values ​​and the allowable stress of the surrounding rock, intuitively reflect the current stress state of the cavern.

[0079] Information on the deformation state of the surrounding rock, such as the real-time values ​​and rates of change of cavern convergence and surrounding rock settlement, the spatial distribution of the deformation area, and the amplitude of instantaneous deformation fluctuations caused by cross-process disturbances, can fully depict the dynamic development characteristics of the surrounding rock deformation.

[0080] Information on the dynamic effects of coupled disturbances includes the degree of influence of geological disturbances on the stress state, the interference amplitude of cross-process disturbances on the deformation state, the degree of attenuation or amplification of the overall mechanical properties of the cavern by the comprehensive disturbance effect, the disturbance stress characteristics of key load-bearing parts such as inter-cave walls, and the specific degree of influence of group effects on the mechanical state.

[0081] Information on the overall mechanical stability level of the surrounding rock, such as the comprehensive judgment based on stress state, deformation state, and intensity of comprehensive disturbance effect, divides mechanical stability into four levels: stable, basically stable, understability, and unstable. This clarifies the current overall mechanical safety status of the target cavern and provides an intuitive basis for construction control.

[0082] Information on the time-varying evolution trend of mechanical state can predict the stress change trend, deformation development trend, and dynamic influence trend of coupled disturbance in the target cavern under multiple construction procedures in the future, and mark the key turning points and time nodes of mechanical state, so as to realize the early prediction of mechanical state.

[0083] Information on the mechanical adaptation status of the support structure, such as the matching degree between the current support strength and the surrounding rock stress, the uniformity of the force and stress distribution of the support structure, and the force increase of the support structure under coupled disturbance, ensures that the support structure is highly adapted to the mechanical state of the cavern.

[0084] Information on mechanical anomaly risk points, such as identifying areas of high stress concentration, large deformation, and disturbance-sensitive areas, can be used to mark the anomaly risk level and cause, and provide targeted mechanical prevention and control tips to achieve accurate positioning of mechanical risks.

[0085] The above comprehensive mechanical state analysis results cover multiple dimensions such as stress, deformation, disturbance effect, stability, and evolution trend. One or more dimensions can be determined according to specific needs, and TIGRU can be built and trained based on these dimensions so that the analysis results containing the corresponding dimension data can be output through the output layer of TIGRU in practical applications.

[0086] In one implementation, the analysis results include the delay probability and delay duration of the target chamber in the current process. The third intermediate feature can be processed using the output layer to obtain output data; the output data includes two dimensions; the first dimension of the output data is activated to obtain the delay probability; the second dimension of the output data is used as the delay duration.

[0087] For example, it is necessary to analyze and predict the delay probability and delay duration of the target chamber. The analysis results include two dimensions, corresponding to the delay probability and delay duration, respectively. The third intermediate feature is input into the output layer, and the time-varying progress risk level is mapped to the [0,1] interval using the sigmoid function, as shown in the following formula: (12) in, P i Indicates the predicted first i The probability of delay in a process can range from [0,1]. z i1 Indicates the first i The first dimension of the third intermediate feature corresponding to the process. Furthermore, based on actual project management needs, the probability of delay can be classified into time-varying risk levels, for example: P i <0.3 (low risk, no special control required), 0.3≤ P i <0.7 (medium risk, monitoring needs to be strengthened) P i ≥0.7 (High risk, control measures need to be taken immediately).

[0088] The predicted delay time can be directly output based on the third intermediate feature. Alternatively, a linear activation function can be used to output the specific numerical value, as shown below: (13) Where, Δ T i Indicates the predicted first i Delay duration of the process (unit can be days); z i2 Indicates the first i The second dimension of the third intermediate feature corresponding to the process. This provides a quantitative reference for construction management.

[0089] In one implementation, corresponding control measures can be determined based on the analysis results of the target chamber. For example, the above-mentioned TIGRU method yields a two-dimensional analysis and prediction result (delay probability). P i+1 Delay duration Δ T i+1 Based on the construction control threshold of the underground cavern group, a four-level risk early warning mechanism and dynamic control strategy are constructed to achieve early warning and graded control of risks, which is in line with the actual operation and maintenance needs of the project. The specific division is shown in Table 1: Table 1

[0090] This system allows for the collection of actual monitoring data (including construction data) for each completed process, which is then fed back to TIGRU to fine-tune the model's parameters and hyperparameters. Based on the updated TIGRU, the system predicts risks for the next period and updates control strategies accordingly, achieving a dynamic iteration of "prediction-control-verification-optimization".

[0091] Figure 4 A schematic diagram of the process architecture of an embodiment of this disclosure is shown. (Reference) Figure 4 As shown, firstly, dynamic monitoring data is collected, and adaptive interpolation is performed based on temporal coupling (i.e., coupling between different processes) and index coupling, followed by data preprocessing. Then, the geological disturbance effect and the process cross-disturbance effect are quantified to obtain the quantified result of the comprehensive disturbance effect (i.e., the comprehensive disturbance effect intensity). The time-varying risk factor update model is used to update the second time-varying risk factor, which includes the comprehensive disturbance effect intensity, and a time-varying risk factor sequence is constructed. Next, the time-varying risk factor sequence is input into the TIGRU model, passing sequentially through the input layer, Transformer encoder, IGU decoder, gated attention mechanism layer, fully connected layer, and output layer, outputting the underground cavern group analysis results. Subsequent optimization of the TIGRU model can be performed based on the underground cavern group analysis results or other relevant information.

[0092] The embodiments of this disclosure will be further illustrated and verified through examples below.

[0093] Taking the underground cavern complex of a large hydropower station as an example, the study area is defined as the main powerhouse (120m long, 25m wide, and 30m high), the main transformer room (80m long, 18m wide, and 22m high), and the tailrace tunnels (3 parallel tunnels, each 150m long and 6m in diameter). Twenty process time periods (t1-t) are divided. 20, 7 days per period). Dynamic monitoring data for 20 periods were collected using sensors (stress, strain sensors, etc.) set up on site, and missing data were filled in using formula (1) or related methods. Two coupling factors strongly correlated with rock mass stress (i.e., other indicators that have an index coupling relationship with rock mass stress) were selected: groundwater flow rate Flow (m) 3 / d), excavation speed v (m / d) was used as an auxiliary interpolation basis. The results of the filling are shown in Table 2. A total of 3 missing values ​​were filled, all of which were t8 and t. 13 Rock mass stress data for a given period of time.

[0094] Table 2

[0095] Based on the monitoring data after filling in the gaps, the intensity of geological disturbance effect, the intensity of process cross-effect, the intensity of comprehensive disturbance effect, and the correlation coefficient (i.e., the correlation coefficient between the intensity of comprehensive disturbance effect and time-varying progress risk) were calculated. The calculation results are as follows: Figure 5 As shown, the correlation coefficients for different time periods are all greater than 0.6, meeting the correlation threshold condition. This confirms a good correlation between the intensity of the comprehensive disturbance effect obtained from the collected monitoring data and the time-varying progress risk, and can be used for subsequent research. Besides... Figure 5 In addition to the correlation coefficients shown, correlation coefficients between other first time-varying risk factors and time-varying schedule risk can also be calculated. Ultimately, six first time-varying risk factors significantly correlated with time-varying schedule risk were selected as second time-varying risk factors, all with correlation coefficients greater than 0.55. The second time-varying risk factors specifically include: the intensity of the comprehensive disturbance effect. D total,i Groundwater flow Flow i Rock mass integrity coefficient K v,i Excavation speed v i Support strength S i Equipment availability rate η i By combining monitoring data and the results of coupled perturbation quantification, dynamic updates of each second time-varying risk factor were completed for the 20 process periods.

[0096] The updated second time-varying risk factor is formed into a time-varying risk factor sequence, and divided into training sets according to a 7:2:1 ratio (e.g., t1-t). 14 ), validation set (such as t) 15 -t 17 ), test set (such as t) 18 -t 20A TIGRU model was constructed, using MSE (Mean Squared Error) as the loss function and AdaBelief as the optimizer. An early stopping strategy was employed to train the model, and hyperparameters were fine-tuned to ensure convergence. After 50 iterations, the model converged. At convergence, the training set loss value stabilized at 0.012, and the validation set loss value was 0.015, with a difference of only 0.003, indicating no overfitting or underfitting. The loss curves during model training and validation are shown below. Figure 6 As shown, the accuracy of the risk level prediction results reached 93.1%, proving that the model can effectively extract the correlation between time-varying factors and coupled perturbation features, and the training effect is good.

[0097] The effectiveness of the model was verified by comparing it with traditional models using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), accuracy, and the number of convergence iterations. MAE and RMSE measure the deviation between the predicted and actual values ​​of time-varying schedule risk, with the deviation measured in days (d) of construction delay. The model comparison results are shown in Table 3. The TIGRU model constructed in this embodiment outperforms traditional LSTM and GRU models in terms of MAE, RMSE, and accuracy (higher is better), verifying that TIGRU has higher quality and performance. This demonstrates that the design integrating Transformer global feature extraction and IGRU temporal dependency capture is effective and can solve the shortcomings of traditional models such as insufficient global feature extraction and inaccurate temporal correlation characterization.

[0098] Table 3

[0099] Five key time periods were selected: t1 (initial excavation), t5 (mid-stage of main plant excavation), t 10 (Peak excavation of main transformer room), t 15 (Peak of tailrace tunnel excavation), t 20 Table 4 shows the comparison of risk prediction, control, and actual results for five key time periods (at the end of excavation). P Let Δ be the probability of risk. T For the delay duration (in days), the prediction error = |predictionΔ T -actual Δ T |

[0100] Table 4

[0101] Based on the comparison of predicted data, actual data, and control effects for key time periods in Table 4, it can be concluded that: the prediction errors for all five key time periods are ≤0.3 days, the prediction deviation is extremely small, the constructed risk prediction model has high prediction accuracy and controllable error, and can effectively support time-varying progress risk prediction; the construction progress risk level is highly correlated with the construction period (i.e., the process period), and the peak excavation period (t 10 t 15 The highest risk level and longest delay period clearly define the key time period for risk management. A comparison of all 20 time periods shows that without any control measures, the simulated total delay time was 38 days; after implementing the dynamic management strategy of this implementation method, the actual total delay time was only 12 days, a reduction of 26 days, or 68.4%. This demonstrates that the dynamic management strategy of this implementation method can effectively avoid schedule risks caused by coupled disturbances and construction procedures, exhibiting outstanding management results and can be applied to similar project management practices.

[0102] In summary, the embodiments disclosed herein achieve the following technical effects: By decomposing the disturbance effects of underground cavern groups into two main categories—geological disturbance effects and process cross-effects—a two-dimensional quantitative index system was constructed. Combining coupled geomechanics theory and construction organization, the correlation function between the intensity of various disturbance effects and schedule risks was derived. This successfully solved the problems of difficulty in quantifying the disturbance effects of underground cavern groups and their inability to be integrated into relevant risk prediction models, achieving accurate quantification of disturbance effects. At the same time, focusing on core disturbance effects improved the efficiency and accuracy of quantification.

[0103] A three-dimensional time-varying framework of "construction period - cavern group status - risk factors" is introduced. Relying solely on dynamic monitoring data, this framework considers the dynamic evolution of risk factors such as disturbance intensity, geological parameters, and construction parameters as construction progresses, and constructs a time-varying risk factor update model. This replaces the traditional static risk prediction approach, achieving dynamic adaptation and lightweight data in risk prediction. It can capture the time-varying characteristics of risk factors in real time, ensuring that risk prediction is synchronized with the actual construction progress.

[0104] Abandoning traditional LSTM and GRU models, a TIGRU model was designed, introducing a gated attention mechanism to strengthen the weights of features related to perturbation effects. This effectively solves the problems of insufficient global feature extraction, inaccurate characterization of temporal correlations, model redundancy, and weak generalization ability of traditional algorithms. It significantly improves the accuracy and efficiency of cavern analysis and construction progress risk prediction, and achieves a more lightweight model. The TIGRU model takes into account both global and temporal features, resulting in higher prediction accuracy. At the same time, its low complexity reduces the model training and running costs, making it adaptable to underground cavern clusters of different geological conditions and scales, and demonstrating high practicality and applicability.

[0105] This disclosure also provides an analysis device for underground cavern groups incorporating disturbance effects. (Reference) Figure 7 As shown, the underground cavern group analysis device 700 includes: The monitoring data acquisition module 710 is configured to acquire monitoring data of multiple indicators during the construction process of the underground cavern group; the construction process includes multiple procedures; the underground cavern group consists of multiple caverns, including a target cavern; the monitoring data of the multiple indicators includes stress parameters. The disturbance effect processing module 720 is configured to determine the comprehensive disturbance effect intensity based on the stress parameters of the target cavern being disturbed by adjacent caverns and the process overlap duration between the target cavern and other caverns; the process overlap duration refers to the duration during which the target cavern and other caverns are in different processes. The time-varying risk factor processing module 730 is configured to use the monitoring data and the intensity of the comprehensive disturbance effect as the first time-varying risk factor, and based on the correlation between the first time-varying risk factor and the time-varying schedule risk, to select a second time-varying risk factor from the first time-varying risk factor, and to form a time-varying risk factor sequence according to the second time-varying risk factors of different processes. The analysis result determination module 740 is configured to process the time-varying risk factor sequence using a time series model with an attention mechanism to obtain the analysis results of the target cavern.

[0106] In one embodiment, after acquiring the monitoring data, the device is further configured to: for missing data in the monitoring data, utilize similar monitoring data from adjacent processes in the process where the missing data is located, as well as different types of monitoring data that have an index coupling relationship with the missing data, to determine filler data for filling the missing data.

[0107] In one implementation, determining the filler data for filling the missing data using similar monitoring data from adjacent processes in the process containing the missing data, as well as different types of monitoring data that have an index coupling relationship with the missing data, includes: determining the filler data using the following formula: ; in, x i,j Indicates the first i In the process of the first j Fill in the data for each indicator; ω t , ω c These represent time-series correlation weights and indicator coupling weights, respectively. x i-1,j , x i+1,j Indicates the first i-1、 i +1 process j Monitoring data for each indicator; t i , t i-1 , t i+1 They represent the first i -1 process, the first i Process, Number i The time period corresponding to +1 process; q Indicates the relationship with the first j The first indicator has an indicator coupling relationship. q One indicator, s Indicates the relationship with the first j The number of indicators that have an indicator coupling relationship; ρ j,q Indicates the first j The first indicator and the first q The coupling coefficient of each indicator; x i,q Indicates the first i In the process of the first q Monitoring data for each indicator.

[0108] In one embodiment, determining the intensity of the comprehensive disturbance effect based on the stress parameters of the target cavern affected by disturbances from adjacent caverns and the time intervals of process overlap between the target cavern and other caverns includes: determining the intensity of the geological disturbance effect of the target cavern based on the stress parameters of the target cavern affected by disturbances from adjacent caverns and the stress parameters not affected by disturbances from adjacent caverns; determining the intensity of the process overlap disturbance effect of the target cavern based on the time intervals of process overlap between the target cavern and other caverns; and determining the intensity of the comprehensive disturbance effect based on the intensity of the geological disturbance effect and the intensity of the process overlap disturbance effect.

[0109] In one embodiment, determining the intensity of the geological disturbance effect of the target cavern based on the stress parameters of the target cavern affected by disturbances from adjacent caverns and the stress parameters unaffected by disturbances from adjacent caverns includes: determining the intensity of the geological disturbance effect using the following formula: ; ; in, D g This indicates the intensity of the geological disturbance effect; m This indicates the number of adjacent chambers of the target chamber; σ k,i Indicates the first i During the process, the target cavity is subjected to the first k Stress parameters after disturbance of adjacent chambers; σ0,i Indicates the first i During the process, the stress parameters of the target cavity when it is not disturbed by adjacent cavities; ω k Indicates the first k Geological disturbance weighting coefficients of adjacent caverns, L k Indicates that the target chamber is related to the first k The distance between adjacent chambers.

[0110] In one embodiment, determining the intensity of the process cross-disturbance effect of the target cavern based on the process cross-disturbance duration between the target cavern and other caverns includes: determining the intensity of the process cross-disturbance effect using the following formula: ; in, D p This indicates the intensity of the cross-disturbance effect of the process; T overlap,i Indicates that the target chamber is in the first i The duration of overlap between processes in the process and those in other caverns; T total,i Indicates the first i Total duration of the process; p Indicated in the target chamber the first i The number of overlapping processes that intersect with the target cavity in the process; I l,i Indicated in the target chamber the first i In the process, the first l The construction intensity of each overlapping process; I max,i Indicated in the target chamber the first i The maximum construction intensity among all overlapping processes in the process; β l For the first l Priority coefficients for cross-processes.

[0111] In one implementation, determining the comprehensive disturbance effect intensity based on the intensity of the geological disturbance effect and the intensity of the process cross-disturbance effect includes: determining the comprehensive disturbance effect intensity using the following formula: ; in, D total,i Indicates the first i The intensity of the overall disturbance effect of the process; D g,i , D p,i They represent the first iThe intensity of geological disturbance effect and the intensity of cross-process disturbance effect; α , β The weighting coefficients corresponding to the intensity of the geological disturbance effect and the intensity of the process cross-disturbance effect are respectively, satisfying the following conditions. α + β =1.

[0112] In one implementation, the step of selecting a second time-varying risk factor from the first time-varying risk factors based on the correlation between the first time-varying risk factor and the time-varying schedule risk includes: determining the correlation coefficient between the first time-varying risk factor and the time-varying schedule risk using the following formula: ; in, r j For the first j The first time-varying risk factor and time-varying schedule risk Y The correlation coefficient; X ij Indicates the first i In the process of the first j The value of the first time-varying risk factor vector; For the first j The mean of the first time-varying risk factor; Y i For the first i Time-varying schedule risk of the process; The mean of time-varying schedule risk is used; the first time-varying risk factor with a correlation coefficient not less than the correlation threshold is selected as the second time-varying risk factor.

[0113] In one embodiment, forming a time-varying risk factor sequence based on the second time-varying risk factors of different processes includes: acquiring actual monitoring data of the second time-varying risk factors of the executed processes and acquiring predicted data of the second time-varying risk factors of the current process; the executed processes include the current process; obtaining predicted data of the second time-varying risk factors of the next process by weighting the actual monitoring data of the second time-varying risk factors of the current process with the predicted data of the second time-varying risk factors of the current process; and forming the time-varying risk factor sequence based on the actual monitoring data of the second time-varying risk factors of the executed processes and the predicted data of the second time-varying risk factors of the next process.

[0114] In one embodiment, the attention-based time series model includes a Transformer, a gated recurrent unit, a fully connected layer, and an output layer. The process of using the attention-based time series model to process the time-varying risk factor sequence to obtain the analysis result of the target cavern includes: encoding the time-varying risk factor sequence using a Transformer to obtain a first intermediate feature; serializing the first intermediate feature using a gated recurrent unit to obtain a second intermediate feature; performing a fully connected process on the second intermediate feature using a fully connected layer to obtain a third intermediate feature; and processing the third intermediate feature using the output layer to obtain the analysis result.

[0115] In one embodiment, the step of serializing the first intermediate feature using a gated loop unit to obtain the second intermediate feature includes: in the reset gate of the gated loop unit, obtaining the reset gate output feature using the following formula: ; in, res i Indicates the first i Process reset gate output feature; W r , b r Represents the weight matrix and bias vector of the reset gate; Z i This represents the first output of the Transformer. i The first intermediate characteristic of the process; h i-1 Indicates the first i -1 hidden layer output features; γ This represents the adaptive adjustment coefficient; in the hidden gate of the gated loop unit, the candidate hidden state is obtained using the following formula: ; in, Indicates the first i Candidate hidden state of the process; W h 、b h Represents the weight matrix and bias vector of the hidden gate; This represents the operation of multiplying corresponding elements of a matrix; based on the candidate hidden state and the... i The hidden layer output features of step -1 are obtained to obtain the... i The hidden door output feature of the process; based on the gated attention mechanism, the hidden door output feature and the second time-varying risk factor are fused using the following formula to obtain the second intermediate feature: ; ; in, e ij Indicates the first i The hidden door output characteristics of the process and the first j The correlation coefficient of the second time-varying risk factor; W a , b a Represents the weight matrix and bias vector of the gating attention mechanism; h i Indicates the first i Hidden door output features of the process; U a The weight matrix represents the correlation between the second time-varying risk factor and the disturbance effect; d j Indicates the first j The correlation between the second time-varying risk factor and the disturbance effect; α ij Indicates the first i In the process of the first j Attention weights for the second time-varying risk factor; g i Indicates the output characteristics of the gating unit; W g , b g Here are the weight matrix and bias vector of the gated unit; This represents the second intermediate feature.

[0116] The specific details of each part of the above-mentioned device have been described in detail in the method section of the implementation plan. For any undisclosed details, please refer to the implementation plan of the method section, and therefore will not be repeated here.

[0117] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to exemplary embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0118] This disclosure also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the method steps of various exemplary embodiments of this disclosure.

[0119] In one implementation, the computer program product can be a tangible product, such as a computer-readable storage medium storing a computer program. The readable storage medium can be based on electrical, magnetic, optical, electromagnetic, infrared, or other signals, and includes, but is not limited to: Random Access Memory (RAM), Read-Only Memory (ROM), magnetic tape, floppy disk, flash memory, Hard Disk Drive (HDD), Solid State Disk (SSD), etc. For example, the computer program product can be a non-volatile storage medium storing a computer program, such as read-only memory, NAND flash memory, etc.

[0120] In one implementation, the computer program product can be an intangible product. For example, the computer program product can be a virtual digital product, such as an executable file or installation package containing a computer program.

[0121] Computer program code can be written in one or more programming languages. Examples of programming languages ​​include C, Java, and C++. Program code can execute entirely on the user's computing device, partially on the user's computing device, or as a standalone software package. It can also execute partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, such as a Local Area Network (LAN) or a Wide Area Network (WAN), or it can be connected to an external computing device (e.g., via an internet connection provided by a mobile network operator).

[0122] Computer programs can be carried or transmitted via signals such as electrical, magnetic, optical, electromagnetic, and infrared rays. Electronic devices can convert the signals carrying computer programs into digital signals, thereby running the computer programs. When a computer program runs on an electronic device, its code is used to cause the electronic device to execute (more specifically, to be executed by the processor of the electronic device) the method steps of various embodiments of this disclosure, such as... Figure 1 The method and steps.

[0123] Implementing the above methods and steps through computer programs yields the following technical benefits: Firstly, by considering and quantifying both geological disturbances and cross-process disturbances within the underground cavern group, coupled disturbance factors are incorporated into the cavern analysis process, ensuring that the analysis results accurately match the actual state of the underground cavern group and improving the accuracy of the analysis results. Secondly, by screening time-varying risk factors and combining them with attention mechanisms and time-series models, both global characteristics and temporal patterns are taken into account, ensuring that the analysis results meet engineering requirements and providing a basis for dynamic and refined management of cavern digital modeling, construction control, and other related operations.

[0124] This disclosure also provides an electronic device. The electronic device includes a processor and a memory. The memory stores executable instructions for the processor, such as computer programs. The processor executes the executable instructions to perform the method steps of various exemplary embodiments of this disclosure.

[0125] The following is for reference. Figure 8 The electronic device is illustrated by way of a general-purpose computing device. It should be understood that... Figure 8 The electronic device 800 shown is merely an example and should not be construed as limiting the functionality or scope of this disclosure.

[0126] like Figure 8 As shown, the electronic device 800 may include: a processor 810, a memory 820, a bus 830, an I / O (input / output) interface 840, and a network adapter 850.

[0127] The memory 820 may include volatile memory, such as RAM 821 and cache unit 822, and may also include non-volatile memory, such as ROM 823. The memory 820 may also include one or more program modules 824, including but not limited to: an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. For example, program module 824 may include the modules described above.

[0128] The processor 810 may include one or more processing units, such as an AP (Application Processor), a modem processor, a GPU (Graphics Processing Unit), an ISP (Image Signal Processor), a controller, an encoder, a decoder, a DSP (Digital Signal Processor), a baseband processor, and / or an NPU (Neural-Network Processing Unit).

[0129] The processor 810 can be used to execute executable instructions stored in the memory 820 to perform method steps of various embodiments of this disclosure, such as... Figure 1 The method and steps.

[0130] By executing the above-mentioned method steps through processor 810, the following technical effects are achieved: Firstly, by considering and quantifying both geological disturbances and process cross-disturbances of the underground cavern group, coupled disturbance factors are incorporated into the cavern analysis process, enabling the analysis results to accurately match the actual state of the underground cavern group and improving the accuracy of the analysis results. Secondly, by screening time-varying risk factors and combining them with attention mechanisms and time-series models, both global characteristics and temporal patterns are taken into account, ensuring that the analysis results meet engineering requirements and providing a basis for the dynamic and refined management of cavern digital modeling, construction control, and other related businesses.

[0131] Bus 830 is used to connect different components of electronic device 800 and may include data bus, address bus and control bus.

[0132] Electronic device 800 can communicate with one or more external devices 900 (such as keyboard, mouse, external controller, etc.) through I / O interface 840.

[0133] Electronic device 800 can communicate with one or more networks via network adapter 850. For example, network adapter 850 can provide mobile communication solutions such as 3G / 4G / 5G, or wireless communication solutions such as wireless LAN, Bluetooth, and near-field communication. Network adapter 850 can communicate with other modules of electronic device 800 via bus 830.

[0134] In one embodiment, the electronic device 800 further includes a display for displaying a graphical user interface.

[0135] although Figure 8As not shown in the diagram, other hardware and / or software modules may also be configured in the electronic device 800, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (Redundant Arrays of Independent Disks) systems, tape drives, and data backup storage systems.

[0136] As can be seen from the above, the technical solutions disclosed herein can be implemented as methods, apparatus, systems, computer program products, storage media, electronic devices, etc. Those skilled in the art will understand that various aspects of this disclosure can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or an implementation combining hardware and software aspects. Exemplarily, these three forms can be referred to as "circuit," "module," and "system," respectively.

[0137] It should be understood that this disclosure is not limited to the specific methods, steps, or structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. Those skilled in the art will readily conceive of other embodiments based on the specific implementations provided in this disclosure. Therefore, the specific implementations provided in this disclosure are merely exemplary, and the scope and spirit of this disclosure are indicated by the claims, and should cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary technical means in the art not disclosed in this disclosure.

Claims

1. A method for analyzing underground cavern groups incorporating disturbance effects, characterized in that, The method includes: The monitoring data of multiple indicators during the construction of the underground cavern complex are obtained; the construction process includes multiple procedures; the underground cavern complex consists of multiple caverns, including the target cavern; the monitoring data of the multiple indicators include stress parameters. The intensity of the comprehensive disturbance effect is determined based on the stress parameters of the target cavern affected by disturbances from adjacent caverns and the duration of process overlap between the target cavern and other caverns; the duration of process overlap refers to the duration during which the target cavern and other caverns are in different processes. Using the monitoring data and the intensity of the comprehensive disturbance effect as the first time-varying risk factor, and based on the correlation between the first time-varying risk factor and the time-varying schedule risk, a second time-varying risk factor is selected from the first time-varying risk factor, and a time-varying risk factor sequence is formed according to the second time-varying risk factors of different processes. The time-varying risk factor sequence is processed using a time-series model with an attention mechanism to obtain the analysis results of the target cavern.

2. The method according to claim 1, characterized in that, After acquiring the monitoring data, the method further includes: For missing data in the monitoring data, filler data is determined by using similar monitoring data from adjacent processes in the process where the missing data is located, as well as different types of monitoring data that have an index coupling relationship with the missing data.

3. The method according to claim 2, characterized in that, The step of determining filler data for filling the missing data by utilizing similar monitoring data from adjacent processes in the process containing the missing data, as well as different types of monitoring data that have an indicator coupling relationship with the missing data, includes: The filling data is determined using the following formula: ； in, x i,j Indicates the first i In the process of the first j Fill in the data for each indicator; ω t , ω c These represent time-series correlation weights and indicator coupling weights, respectively. x i-1,j , x i+1,j Indicates the first i -1、 i +1 process j Monitoring data for each indicator; t i , t i-1 , t i+1 They represent the first i -1 process, the first i Process, Number i The time period corresponding to +1 process; q Indicates the relationship with the first j The first indicator has an indicator coupling relationship. q One indicator, s Indicates the relationship with the first j The number of indicators that have an indicator coupling relationship; ρ j,q Indicates the first j The first indicator and the first q The coupling coefficient of each indicator; x i,q Indicates the first i In the process of the first q Monitoring data for each indicator.

4. The method according to claim 1, characterized in that, The determination of the comprehensive disturbance effect intensity based on the stress parameters of the target cavity affected by disturbances from adjacent cavity and the time intervals of process overlap between the target cavity and other cavity includes: The intensity of the geological disturbance effect of the target cavern is determined based on the stress parameters of the target cavern when disturbed by adjacent caverns and the stress parameters when not disturbed by adjacent caverns. The intensity of the process crossover disturbance effect of the target cavern is determined based on the process crossover time between the target cavern and other caverns. The overall disturbance effect intensity is determined based on the intensity of the geological disturbance effect and the intensity of the process cross-disturbance effect.

5. The method according to claim 4, characterized in that, The step of determining the intensity of the geological disturbance effect of the target cavern based on the stress parameters of the target cavern affected by disturbances from adjacent caverns and the stress parameters unaffected by disturbances from adjacent caverns includes: The intensity of the geological disturbance effect is determined using the following formula: ； ； in, D g This indicates the intensity of the geological disturbance effect; m This indicates the number of adjacent chambers of the target chamber; σ k,i Indicates the first i During the process, the target cavity is subjected to the first k Stress parameters of adjacent chambers after disturbance; σ 0,i Indicates the first i During the process, the stress parameters of the target cavity when it is not disturbed by adjacent cavities; ω k Indicates the first k Geological disturbance weighting coefficients for adjacent caverns, L k Indicates that the target chamber is related to the first k The distance between adjacent chambers.

6. The method according to claim 4, characterized in that, The step of determining the intensity of the process cross-disturbance effect of the target cavern based on the process cross-disturbance duration between the target cavern and other caverns includes: The intensity of the cross-disturbance effect of the process is determined using the following formula: ； in, D p This indicates the intensity of the cross-disturbance effect of the process; T overlap,i Indicates that the target chamber is in the first i The duration of overlap between processes in the process and those in other caverns; T total,i Indicates the first i Total duration of the process; p Indicates the first in the target chamber i The number of overlapping processes that intersect with the target cavity in the process; I l,i Indicates the first in the target chamber i In the process, the first l The construction intensity of each overlapping process; I max,i Indicates the first in the target chamber i The maximum construction intensity among all overlapping processes in the process; β l For the first l Priority coefficients for cross-processes.

7. The method according to claim 4, characterized in that, The determination of the comprehensive disturbance effect intensity based on the intensity of the geological disturbance effect and the intensity of the process cross-disturbance effect includes: The intensity of the combined disturbance effect is determined using the following formula: ； in, D total,i Indicates the first i The intensity of the overall disturbance effect of the process; D g,i , D p,i They represent the first i The intensity of geological disturbance effect of the process and the intensity of cross-process disturbance effect; α , β The weighting coefficients corresponding to the intensity of the geological disturbance effect and the intensity of the process cross-disturbance effect are respectively, satisfying the following conditions. α + β =1.

8. The method according to claim 1, characterized in that, The step of selecting a second time-varying risk factor from the first time-varying risk factor based on the correlation between the first time-varying risk factor and the time-varying schedule risk includes: The correlation coefficient between the first time-varying risk factor and the time-varying schedule risk is determined using the following formula: ； in, r j For the first j The first time-varying risk factor and time-varying schedule risk Y The correlation coefficient; X ij Indicates the first i In the process of the first j The value of the first time-varying risk factor vector; For the first j The mean of the first time-varying risk factor; Y i For the first i Time-varying schedule risk of the process; This represents the average value of time-varying schedule risk. The first time-varying risk factor with a correlation coefficient not less than the correlation threshold is selected as the second time-varying risk factor.

9. The method according to claim 1, characterized in that, The step of forming a time-varying risk factor sequence based on the second time-varying risk factors of different processes includes: Obtain actual monitoring data of the second time-varying risk factor of the executed process, and obtain predicted data of the second time-varying risk factor of the current process; the executed process includes the current process; By weighting the actual monitoring data of the second time-varying risk factor of the current process with the predicted data of the second time-varying risk factor of the current process, the predicted data of the second time-varying risk factor of the next process is obtained. The time-varying risk factor sequence is formed based on the actual monitoring data of the second time-varying risk factor of the executed process and the predicted data of the second time-varying risk factor of the next process.

10. The method according to any one of claims 1 to 9, characterized in that, The attention-based time series model includes a Transformer, a gated recurrent unit, a fully connected layer, and an output layer; the process of using the attention-based time series model to process the time-varying risk factor sequence to obtain the analysis results of the target cavern includes: The time-varying risk factor sequence is encoded using a Transformer to obtain the first intermediate feature; The first intermediate feature is serialized using a gated loop unit to obtain the second intermediate feature; The second intermediate feature is processed by a fully connected layer to obtain the third intermediate feature; The third intermediate feature is processed using the output layer to obtain the analysis result.

11. The method according to claim 10, characterized in that, The step of using a gated loop unit to serialize the first intermediate feature to obtain the second intermediate feature includes: In the reset gate of the gated loop unit, the reset gate output characteristics are obtained using the following formula: ； in, res i Indicates the first i Process reset gate output feature; W r , b r Represents the weight matrix and bias vector of the reset gate; Z i This represents the first output of the Transformer. i The first intermediate characteristic of the process; h i-1 Indicates the first i -1 hidden layer output features; γ This represents the adaptive adjustment coefficient; In the hidden door of the gated loop unit, the candidate hidden state is obtained using the following formula: ； in, Indicates the first i Candidate hidden state of the process; W h 、b h Represents the weight matrix and bias vector of the hidden gate; This represents the operation of multiplying corresponding elements of a matrix; Based on the candidate hidden state and the first i The hidden layer output features of step -1 are obtained to obtain the... i Hidden door output features of the process; Based on the gated attention mechanism, the second intermediate feature is obtained by fusing the hidden gate output feature and the second time-varying risk factor using the following formula: ； ； in, e ij Indicates the first i The hidden door output characteristics of the process and the first j The correlation coefficient of the second time-varying risk factor; W a , b a Represents the weight matrix and bias vector of the gating attention mechanism; h i Indicates the first i Hidden door output features of the process; U a The weight matrix represents the correlation between the second time-varying risk factor and the disturbance effect; d j Indicates the first j The correlation between the second time-varying risk factor and the disturbance effect; α ij Indicates the first i In the process of the first j Attention weights for the second time-varying risk factor; g i Indicates the output characteristics of the gating unit; W g , b g Here are the weight matrix and bias vector of the gated unit; This represents the second intermediate feature.

12. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method according to any one of claims 1 to 11.

13. An electronic device, characterized in that, include: processor; as well as Memory for storing the executable instructions of the processor; The processor is configured to implement the method of any one of claims 1 to 11 by executing the executable instructions.

Images