Supervision method for construction passage of underground powerhouse of hydropower station based on continuous TBM excavation

Abstract

This invention discloses a method for supervising the construction passage of a hydropower station's underground powerhouse based on continuous TBM excavation. The method first uses the ROBOTEC measurement system to collect three-dimensional spatial geometric data of the construction passage, and the Buildertrend monitoring system to obtain behavioral data such as personnel operations. A multidimensional perception data screening model for the construction passage is then used to perform spatiotemporal fusion of the two types of data, extracting abnormal data features. A coupled relationship model is then established based on TBM excavation and surrounding rock geological parameters to deduce the theoretical excavation state, and actual deviation data is determined through multidimensional comparison. Finally, a construction intervention decision set is generated based on the deviations through the Buildertrend monitoring system and fed back to the construction terminal. This method enables precise supervision and intelligent decision-making throughout the entire construction process, effectively improving construction safety and project quality.

Description

Technical Field

[0001] The present invention relates to the field of construction supervision of hydropower stations, and in particular to a method for supervising construction passages of underground powerhouses of hydropower stations based on TBM continuous excavation. Background Art

[0002] Under the strategic development of hydropower, the construction of underground powerhouses for hydropower stations is progressing towards deeper and more complex geological environments. TBM (Transport Block Machine) continuous tunneling technology, with its efficient and stable construction advantages, has become a core method for constructing underground powerhouse construction passages. However, the construction process faces challenges such as complex and changing geological conditions and the difficulty in collaboratively processing multi-source heterogeneous data. Traditional construction supervision methods are no longer able to meet the urgent need for high-precision, real-time monitoring.

[0003] There are two major key flaws in existing technologies. First, the data collection and processing system is imperfect. Traditional supervision methods rely on decentralized data collection methods, which cannot achieve the simultaneous collection and deep integration of multiple types of data such as the three-dimensional spatial form of the construction channel and the dynamic information of personnel and equipment. As a result, abnormal features in the construction process are difficult to detect in a timely manner, and there is a significant lag in information feedback, which makes it difficult to meet the supervision requirements in a dynamic construction environment. Second, the construction status analysis and decision-making mechanism lack scientificity. Previous technologies have found it difficult to accurately analyze the intrinsic relationship between TBM excavation parameters and surrounding rock response, and cannot reliably predict the construction status; and in the deviation analysis and intervention decision-making links, they rely more on experience judgment and lack data-driven intelligent decision-making capabilities, making it difficult to effectively ensure construction safety and project quality. Summary of the Invention

[0004] In order to overcome the shortcomings and deficiencies of the prior art, the present invention provides a method for supervising the construction passage of an underground powerhouse of a hydropower station based on continuous tunneling by a TBM.

[0005] The technical solution adopted by the present invention is a method for supervising the construction passage of a hydropower station underground powerhouse based on continuous TBM excavation, comprising the following steps:

[0006] Step S1: Using the ROBOTEC measurement system, three-dimensional geometric data of the TBM continuous excavation operation area is collected according to a preset spatial coordinate network within the construction channel of the hydropower station's underground powerhouse. This initial measurement data set, including construction channel contour point cloud data and real-time TBM equipment posture data, is obtained.

[0007] Step S2: Connect the Buildertrend monitoring system to the construction management network to simultaneously collect personnel operation information, equipment operation status information, and material flow data within the construction channel to build a construction process behavior data set;

[0008] Step S3: Construct a multi-dimensional perception data screening model for the construction channel, fuse the initial measurement data set with the construction process behavior data set in the spatiotemporal dimension, and extract abnormal data features from the fused data by establishing an association rule base based on the channel design parameter threshold;

[0009] Step S4: Based on the dynamic mechanical parameters of the TBM's continuous excavation and the geological parameters of the surrounding rock of the construction channel, a tunneling parameter-surrounding rock response coupling relationship model is established. The current TBM tunneling parameters are deduced in real time using this model to obtain theoretical tunneling status data;

[0010] Step S5: performing a multi-dimensional feature comparison between the theoretical excavation state data and the abnormal data features, and using a deviation analysis algorithm based on the convergence threshold of the construction channel section to determine the actual deviation data during the TBM excavation process;

[0011] Step S6: Based on the actual deviation data, the Buildertrend monitoring system generates a construction intervention decision set, including excavation parameter adjustment instructions and personnel and equipment scheduling plans. The construction intervention decision set is then fed back to the TBM control system and the on-site construction management terminal.

[0012] Furthermore, in step S3, the construction channel multi-dimensional perception data screening model adopts the following data fusion formula:

[0013] ,in, is the fused data vector; The first A geometric data vector; For the The spatial weight coefficient corresponding to each geometric data vector is determined by the calibrated structural coordinates in the construction channel design drawings; The construction process behavior data set behavioral data vectors; For the The time weight coefficient corresponding to each behavior data vector is calculated based on the construction process time constraint matrix; are weighted factors that characterize the importance of geometric data and behavioral data, respectively, and are determined by the construction channel safety level parameters.

[0014] Furthermore, in step S4, the tunneling parameter-surrounding rock response coupling relationship model adopts the following modeling formula:

[0015] ,in, is the theoretical excavation state data vector; is the tunneling parameter matrix including TBM thrust, torque, and cutterhead speed parameters; The surrounding rock mechanical property matrix is ​​constructed based on the geological exploration data of the construction channel area, including rock hardness coefficient and joint development degree parameters; is the construction channel cross-sectional dimensions and buried depth structural parameter matrix; It is a coupling relationship function based on nonlinear mapping and is determined by a deep learning network trained with historical construction data.

[0016] Furthermore, in step S5, the deviation analysis algorithm based on the construction channel section shrinkage threshold adopts the following deviation calculation model:

[0017] ,in, is the comprehensive deviation measure; The actual measured construction channel characteristic parameter values; is the first Theoretical values ​​of characteristic parameters; The first The convergence threshold of the characteristic parameters; The number of feature parameters to be calculated for the participation bias.

[0018] Furthermore, in step S3, the association rule base is constructed using the following rule generation formula:

[0019] ,in, is a set of association rules; Conditional items formed by the combination of data features; is the corresponding abnormal behavior action item; is the probability that the condition item and the action item appear at the same time; is the probability of the conditional item occurring; is the minimum confidence threshold set according to the risk level of the construction channel.

[0020] Furthermore, in step S6, the construction intervention decision set is generated using the following decision model:

[0021] ,in, To ultimately generate construction intervention decisions; is the set of all decision options; For decision-making The implementation cost is composed of equipment scheduling cost and material consumption cost; For decision-making The estimated execution time; For decision-making Residual risk value after implementation; The weight coefficients corresponding to cost, time and risk are determined according to the construction schedule and safety goals.

[0022] Furthermore, the step S3 includes the following sub-steps:

[0023] Step S3-1: Establish a spatial topological structure model of the construction channel, divide the construction channel into multiple three-dimensional grid sub-regions, and determine the spatiotemporal reference for data collection in each sub-region. By constructing a three-dimensional grid system, independent spatiotemporal coordinates are assigned to data in different regions to form a data spatial positioning framework.

[0024] Step S3-2: Perform a rough match between the initial measurement data set and the construction process behavior data set based on spatiotemporal tags to eliminate data records with spatiotemporal conflicts; perform a preliminary alignment of multi-source data based on spatiotemporal tags to identify and eliminate data with conflicting timestamps and spatial locations;

[0025] Step S3-3: Based on the construction channel design parameters, a multi-dimensional data screening template is constructed, including a cross-sectional dimension threshold and an equipment operation parameter threshold. A multi-dimensional threshold system is set based on the design parameters to form a standard template for data screening.

[0026] Step S3-4: Filter the roughly matched data through a multidimensional data screening template to extract abnormal data features; use the screening template to filter the data, identify data features that exceed the threshold range, and form an abnormal data feature set.

[0027] Furthermore, the step S4 includes the following sub-steps:

[0028] Step S4-1: Collect geological drilling data in the construction channel area and construct a spatial distribution model of surrounding rock mechanical parameters; integrate geological exploration data, establish a spatial distribution model of surrounding rock mechanical parameters, and digitally characterize the geological characteristics of the construction area;

[0029] Step S4-2: Acquire real-time measurement values ​​of the TBM's current excavation parameters and, in combination with the construction channel structural parameters, construct a current excavation condition parameter matrix; combine the real-time excavation parameters with the construction channel structural parameters to form a parameter matrix reflecting the current construction status;

[0030] Step S4-3: Input the spatial distribution model of surrounding rock mechanical parameters and the current tunneling condition parameter matrix into the tunneling parameter-surrounding rock response coupling relationship model to calculate theoretical tunneling state data; simulate the theoretical tunneling state of the TBM under the current working conditions through the coupling relationship model to generate a theoretical state data vector;

[0031] Step S4-4: Performing spatiotemporal interpolation processing on the theoretical excavation status data to make it consistent with the spatiotemporal scale of the actual measurement data; performing scale conversion on the theoretical data using a spatiotemporal interpolation algorithm.

[0032] Furthermore, the step S5 includes the following sub-steps:

[0033] Step S5-1: Align the theoretical excavation status data with the abnormal data features based on the feature dimension to establish a data mapping relationship; through the alignment of the feature dimension, a mapping bridge is built between the theoretical data and the actual data;

[0034] Step S5-2: Determine the deviation calculation weights of different characteristic parameters based on the cross-section convergence threshold of the construction channel; assign corresponding weights to different characteristic parameters based on the cross-section convergence threshold system;

[0035] Step S5-3: using a weighted Euclidean distance algorithm to calculate the deviation vector between the actual measured data and the theoretical data; using the weighted Euclidean distance algorithm to quantify the difference between the actual and theoretical data to form a multi-dimensional deviation vector;

[0036] Step S5-4: Perform dimensionality reduction processing on the deviation vector, extract the calibrated deviation features, and generate actual deviation data; compress the deviation vector dimension through the dimensionality reduction algorithm, extract the deviation features of different construction nodes, and form a construction node deviation data set.

[0037] Beneficial Effects: This invention proposes a method for supervising the construction passage of a hydropower station's underground powerhouse based on continuous TBM excavation. In terms of data acquisition and processing, this method uses the ROBOTEC measurement system and the Buildertrend monitoring system to simultaneously collect multi-source heterogeneous data, including the three-dimensional geometry of the construction passage and the dynamics of personnel and equipment. It then utilizes a multidimensional perception data screening model for spatiotemporal fusion, addressing the data fragmentation and difficulty in collaborative analysis inherent in traditional methods. This method not only rapidly captures abnormal data characteristics but also significantly shortens information feedback time. In the construction status analysis and decision-making phase, a coupling relationship model is established based on TBM excavation parameters and surrounding rock geological parameters to accurately deduce the theoretical excavation status. This, combined with a deviation analysis algorithm based on cross-section convergence thresholds, accurately determines actual deviation data, thus changing the previous model that relied on empirical judgment. Furthermore, through a scientific construction intervention decision-making model, the optimal decision set is generated by integrating factors such as cost, time, and risk, achieving closed-loop management of the entire process from data acquisition and analysis to decision-making. This significantly improves the accuracy and scientific nature of construction supervision, provides strong technical support for ensuring the construction safety and quality of the hydropower station's underground powerhouse construction passage, and effectively addresses the shortcomings of existing technologies in meeting the requirements of high-precision, real-time supervision. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is a flow chart of the method steps of the present invention;

[0039] Figure 2This is a diagram of the unit composition of the method implementation of the present invention. DETAILED DESCRIPTION

[0040] It should be noted that, unless there is a conflict, the embodiments in this application and the features described in the embodiments can be combined with each other. The application is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0041] like Figure 1 As shown in FIG, the method for supervising the construction passage of the underground powerhouse of a hydropower station based on continuous TBM excavation includes the following steps:

[0042] Step S1: Using the ROBOTEC measurement system, three-dimensional geometric data of the TBM continuous excavation operation area is collected according to a preset spatial coordinate network within the construction channel of the hydropower station's underground powerhouse. This initial measurement data set, including construction channel contour point cloud data and real-time TBM equipment posture data, is obtained.

[0043] Specifically, step S1 is the fundamental data collection phase of the entire supervision method. It utilizes the ROBOTEC measurement system to collect three-dimensional spatial geometric data of the TBM's continuous excavation area. Within the construction tunnel, measurements are conducted according to a pre-defined spatial coordinate network. This coordinate network is divided according to the tunnel's design dimensions and precision requirements. For example, in critical areas requiring high precision, the grid spacing may be set to 0.1 meters, while in conventional areas, the grid spacing can be set to 0.5-1 meters. Laser scanning and other technologies are used to obtain point cloud data of the tunnel's contours. This point cloud data typically has a density of 1,000-5,000 points per square meter, accurately representing the tunnel's geometry. Simultaneously, the system collects real-time TBM posture data at a rate of 1-10 times per second, covering information such as the machine's position and posture. This data together constitutes the initial measurement dataset. This step is crucial for providing accurate spatial geometric foundational data for subsequent construction status analysis. Inaccurate or missing data can render subsequent analysis and decision-making unreliable.

[0044] The ROBOTEC measurement system, short for the Robotic Total Station Electronic Control measurement system, is an advanced device widely used in the field of engineering surveying. In the scenario of supervising the construction passage of the underground powerhouse of a hydropower station, it works based on a preset spatial coordinate network. Through laser scanning technology, the system can quickly obtain point cloud data of the construction passage contour, with a point cloud density of typically 1,000 to 5,000 points per square meter, which enables it to accurately outline the geometry of the passage. At the same time, the ROBOTEC measurement system collects real-time posture data of the TBM equipment at a frequency of 1 to 10 times per second, covering key information such as the spatial position and posture of the equipment. These data together constitute the initial measurement data set, providing crucial spatial geometric basic data for subsequent construction status analysis.

[0045] During implementation, the ROBOTEC measurement system was installed in a fixed and stable position within the construction tunnel, ensuring that the measurement process was unaffected by factors such as construction vibration. The system automatically scanned and measured the work area according to a preset coordinate network. After each scan, the collected raw data was initially filtered and denoised to remove invalid data points caused by environmental interference. The processed data was labeled and organized according to spatial coordinates and acquisition time, and stored in standardized data formats, such as common point cloud data formats and equipment posture data formats, to facilitate subsequent integration and analysis with other data. The entire acquisition process was continuous, with data constantly updated as the TBM progressed, ensuring real-time and continuous data.

[0046] Step S2: Connect the Buildertrend monitoring system to the construction management network to simultaneously collect personnel operation information, equipment operation status information, and material flow data within the construction channel to build a construction process behavior data set;

[0047] Specifically, step S2 aims to collect behavioral data during the construction process through the Buildertrend monitoring system to construct a data set. Once connected to the construction management network, the system establishes data exchange with various sensors and information systems deployed at the construction site. For collecting personnel work information, positioning sensors and identity recognition devices are used to obtain real-time location information of personnel within the construction corridor. The collection frequency is set based on the type of work and safety management requirements. For example, the frequency of personnel location collection in high-risk work areas can be up to 5 times per minute, while in regular work areas, it is 1-2 times per minute. Information such as personnel work time and operation content is also recorded. Regarding equipment operating condition information collection, sensors monitor dynamic parameters such as thrust, torque, and cutterhead speed of key equipment such as TBMs in real time through real-time monitoring. The collection frequency is typically 1-10 times per second. For other auxiliary equipment, corresponding operating status parameters are also collected. Material flow data collection covers information such as the time, quantity, and storage location of materials entering and leaving the warehouse. The collection frequency is determined by the material management process, generally 1-4 times per hour. These multiple data types together constitute the construction process behavior data set, providing a rich source of data for a comprehensive understanding of construction dynamics.

[0048] The Buildertrend Supervision System, or Buildertrend Construction Management, is primarily used to collect and integrate construction process behavior data. Once connected to the construction management network, it works closely with various sensors and information systems deployed at the construction site. Using positioning sensors and identity recognition devices, the system can obtain real-time information about personnel operations, including their location within the construction corridor, work duration, and operation content. By connecting to the equipment's built-in sensors, it can monitor the operating conditions of equipment like TBMs, such as dynamic parameters like thrust, torque, and cutterhead speed. It also collects material flow data, such as the time, quantity, and storage location of materials entering and leaving the warehouse. This data forms a comprehensive data set for understanding construction dynamics. In tandem, the initial measurement dataset collected by the ROBOTEC measurement system and the construction process behavior data collected by the Buildertrend Supervision System serve as inputs to a multi-dimensional perception data screening model for the construction corridor. The two are integrated in time and space, laying the foundation for subsequent extraction of abnormal data features based on an association rule library. Together, they contribute to the entire construction corridor supervision process, ensuring construction safety and quality.

[0049] During implementation, the Buildertrend monitoring system utilizes a distributed data collection architecture, establishing data collection nodes in various areas of the construction site. Data collection devices are deployed at key equipment locations, personnel entrances and exits, and material storage points, enabling local data collection and reducing data transmission delays. Each collection node transmits data to the system server via a wireless network or dedicated communication lines. The server then performs preliminary collation and classification of the collected data, formatting it according to unified data standards. For example, personnel information, equipment parameters, and material data are categorized and converted into standard data formats. The system updates data in real time, ensuring that construction management personnel and subsequent data analysis modules have access to the latest construction process information, providing strong support for construction supervision and decision-making.

[0050] Step S3: Construct a multi-dimensional perception data screening model for the construction channel, fuse the initial measurement data set with the construction process behavior data set in the spatiotemporal dimension, and extract abnormal data features from the fused data by establishing an association rule base based on the channel design parameter threshold;

[0051] Specifically, step S3 constructs a multidimensional perception data screening model for the construction channel, deeply processing the initial measurement datasets and construction process behavior data sets acquired in the first two steps. First, in the spatiotemporal fusion phase, a unified spatiotemporal reference system is established to map data from different sources and at different acquisition times into the same spatiotemporal coordinate system. Specifically, the two types of data are synchronized and spatially aligned based on the acquisition time and spatial coordinate information to ensure temporal and spatial consistency. Then, an association rule base is established based on channel design parameter thresholds. This rule base contains various rules, such as thresholds for construction channel cross-sectional dimensions, thresholds for equipment operating parameters, and personnel operating specifications. For example, the allowable deviation range for construction channel cross-sectional dimensions, the normal range for various TBM equipment operating parameters, and safe operating procedures for personnel. By comparing the fused data against the association rule base one by one, the system can automatically identify data features that exceed the threshold range and extract abnormal data features. The significance of this step lies in integrating multi-source data, uncovering hidden anomalies, and promptly identifying potential risks in the construction process.

[0052] During the implementation process, a modular model architecture is used to realize the data screening function. The data fusion module adopts different fusion algorithms according to the type and characteristics of the data. For example, for geometric data and behavioral data, appropriate spatial fusion algorithms and time series fusion algorithms are used respectively to give full play to the advantages of each type of data. The construction of the association rule base is based on the design drawings of the construction channel, equipment operation manuals, safety construction standards and other materials, and is optimized and adjusted in combination with historical construction data to ensure the accuracy and applicability of the rules. In the process of extracting abnormal data features, the system uses sliding window technology and statistical analysis methods to monitor and analyze the data in real time. Once it is found that the data exceeds the threshold range set by the rule base, the relevant data features are immediately marked and recorded to generate an abnormal data feature set to provide key information for subsequent construction status analysis.

[0053] Step S4: Based on the dynamic mechanical parameters of the TBM's continuous excavation and the geological parameters of the surrounding rock of the construction channel, a tunneling parameter-surrounding rock response coupling relationship model is established. The current TBM tunneling parameters are deduced in real time using this model to obtain theoretical tunneling status data;

[0054] Specifically, step S4 establishes a coupled relationship model between excavation parameters and surrounding rock response based on the dynamic mechanical parameters of the TBM's continuous excavation and the geological parameters of the surrounding rock in the construction channel, thereby deducing the TBM's theoretical excavation status data. To construct the model, dynamic mechanical parameters such as the TBM's thrust, torque, and cutterhead speed are first collected. These parameters are measured in real time by sensors built into the TBM at a high frequency, typically 1-10 times per second. Simultaneously, a model of the surrounding rock's mechanical properties is constructed, incorporating parameters such as rock hardness coefficient, joint development, and elastic modulus, in combination with detailed data obtained from previous geological surveys of the construction channel area. By studying and analyzing a large amount of historical construction data and utilizing techniques such as deep learning, the complex nonlinear mapping relationship between excavation parameters and surrounding rock response is determined, thereby establishing a coupled relationship model between excavation parameters and surrounding rock response. Based on this model, the TBM's current excavation parameters are input to derive theoretical excavation status data in real time, including information such as the projected excavation speed and surrounding rock deformation. This step is crucial for predicting construction status and identifying potential problems in advance, helping construction personnel prepare countermeasures in advance.

[0055] During the implementation process, geological drilling data and other relevant information for the construction channel area were first retrieved from the geological exploration database. Using professional geological data analysis software, a spatial distribution model of the surrounding rock mechanical parameters was constructed, visually demonstrating the differences in mechanical properties of the surrounding rock at different locations. Next, the measured values ​​of the TBM's current excavation parameters were acquired in real time. Combined with structural parameters such as the construction channel's cross-sectional dimensions and burial depth, a matrix of current excavation condition parameters was constructed. This spatial distribution model of surrounding rock mechanical parameters, along with the matrix of current excavation condition parameters, was input into the established excavation parameter-surrounding rock response coupling relationship model. Using the model's internal algorithms and computational logic, theoretical excavation status data was calculated. Because the calculated theoretical data may differ from the actual measured data in terms of time and space, the theoretical excavation status data required temporal and spatial interpolation. Appropriate interpolation algorithms, such as linear interpolation and spline interpolation, were employed to ensure consistency in temporal and spatial scale with the actual measured data, facilitating subsequent comparative analysis.

[0056] Step S5: performing a multi-dimensional feature comparison between the theoretical excavation state data and the abnormal data features, and using a deviation analysis algorithm based on the convergence threshold of the construction channel section to determine the actual deviation data during the TBM excavation process;

[0057] Specifically, step S5 performs a multi-dimensional feature comparison between the theoretical excavation status data and the abnormal data features to determine the actual deviation data during the TBM excavation process. Before the comparison, the theoretical excavation status data and the abnormal data features are first aligned based on their feature dimensions. Specifically, a data mapping relationship is established between the two based on the physical meaning and feature type represented by the data, such as excavation speed, equipment posture, and surrounding rock deformation, to ensure accurate correspondence between data with the same feature dimensions. Next, deviation calculation weights are determined for each feature parameter based on design specifications such as the convergence threshold of the construction channel section. Different feature parameters have varying degrees of impact on construction safety and quality. For example, surrounding rock deformation may have a relatively high weight, while some minor equipment parameters may have lower weights. A weighted Euclidean distance algorithm is then used to quantify the differences between the actual measured data and the theoretical data in each feature dimension, resulting in a deviation vector between the actual measured data and the theoretical data. Finally, to highlight key deviation information, the deviation vector is subjected to dimensionality reduction. Using dimensionality reduction algorithms such as principal component analysis, key deviation features are extracted from it, thereby generating actual deviation data. This step can accurately quantify the actual deviations during the construction process and provide an accurate basis for subsequent decision-making.

[0058] During the implementation process, a specialized data processing program was developed to implement the above operations. The program first reads the theoretical excavation status data and abnormal data features, parses and organizes the data according to the preset feature dimension division rules, and completes the feature dimension-based alignment operation. By querying the construction channel design documents and relevant specifications and standards, information such as the cross-section convergence threshold of each feature parameter is obtained, and the deviation calculation weight is calculated to determine the deviation calculation weight. Using the pre-written weighted Euclidean distance algorithm program module, the actual measurement data and theoretical data are calculated to obtain the deviation vector. Finally, the dimensionality reduction algorithm program is used to process the deviation vector, extract key deviation features, and organize them into the format of actual deviation data for storage and output. The entire process runs automatically on the data processing server, ensuring that the deviation analysis work can be completed quickly and accurately, saving time for construction decision-making.

[0059] Step S6: Based on the actual deviation data, the Buildertrend monitoring system generates a construction intervention decision set, including excavation parameter adjustment instructions and personnel and equipment scheduling plans. The construction intervention decision set is then fed back to the TBM control system and the on-site construction management terminal.

[0060] Specifically, in step S6, based on the actual deviation data determined in step S5, the Buildertrend monitoring system generates a construction intervention decision set and feeds it back to the relevant terminals. This decision set takes into account a variety of factors, including instructions for adjusting tunneling parameters and scheduling personnel and equipment. Based on the type and extent of the actual deviation, combined with the construction schedule and safety objectives, the system analyzes the implementation costs of different decision options, such as equipment scheduling costs and material consumption costs; estimates the execution time of each option; and assesses the remaining risk after implementation. By comprehensively evaluating and comparing all possible decision options, the optimal option in terms of cost, time, and risk is selected to form a construction intervention decision set. This decision set includes specific operational instructions, such as adjusting the TBM's thrust and torque parameters and scheduling specific personnel and equipment. The significance of this step lies in translating the data analysis results into practical and feasible construction operational instructions, enabling timely correction of construction deviations and ensuring construction safety and project quality.

[0061] During the implementation process, the Buildertrend monitoring system incorporated a dedicated decision-making module. After receiving actual deviation data, this module first generated a series of possible decision scenarios based on pre-set decision-making rules and algorithms. It then invoked submodules such as cost calculation, time estimation, and risk assessment to calculate and evaluate the implementation cost, execution time, and residual risk of each scenario. Using a multi-objective optimization algorithm, all scenarios were sorted and screened to determine the optimal construction intervention decision, forming a construction intervention decision set. Once the decision set was generated, instructions were distributed to the TBM control system via the industrial control network, enabling automatic adjustment of excavation parameters. Simultaneously, the personnel and equipment scheduling plan was transmitted to the on-site construction management terminal, where on-site managers made personnel and equipment arrangements based on the instructions. Furthermore, the system incorporated a feedback mechanism to receive real-time feedback from the TBM control system and on-site construction management terminal regarding decision execution, enabling evaluation and subsequent adjustments.

[0062] Preferably, in step S3, the construction channel multi-dimensional perception data screening model adopts the following data fusion formula:

[0063] ,in, is the fused data vector; The first A geometric data vector; For the The spatial weight coefficient corresponding to each geometric data vector is determined by the calibrated structural coordinates in the construction channel design drawings; The construction process behavior data set behavioral data vectors; For the The time weight coefficient corresponding to each behavior data vector is calculated based on the construction process time constraint matrix; are weighted factors that characterize the importance of geometric data and behavioral data, respectively, and are determined by the construction channel safety level parameters.

[0064] Specifically, the core of the data fusion link in step S3 is to achieve deep fusion of multi-source data through data fusion formulas. During the implementation process, first, for the geometric data vectors in the initial measurement data set, the spatial weight coefficient is determined based on the key structural coordinates in the construction channel design drawings. These key structural coordinates, such as the coordinates of the key points of the arch and side walls of the channel, directly affect the safety and quality of construction. The corresponding data are given a higher weight to highlight their importance. The behavioral data vectors in the construction process behavior data set are calculated based on the time constraint matrix of the construction process. The construction process has strict sequence and time connection requirements. The time weight coefficient can reflect the degree of correlation between each behavioral data in the time dimension. The weighting factors α and β are determined by the construction channel safety level parameters. For channels with high safety levels, the weighting factor α of the geometric data is relatively large, ensuring that the focus is on the changes in the construction channel morphology. Through this formula, geometric data and behavioral data are weightedly fused to generate a fused data vector, which lays the foundation for the subsequent abnormal data feature extraction based on the association rule library, realizes the transformation of multi-source data from dispersion to organic integration, makes the construction status information contained in the data richer and more accurate, and effectively improves the comprehensiveness and accuracy of abnormal data identification.

[0065] Preferably, in step S4, the tunneling parameter-surrounding rock response coupling relationship model adopts the following modeling formula:

[0066] ,in, is the theoretical excavation state data vector; is the tunneling parameter matrix including TBM thrust, torque, and cutterhead speed parameters; The surrounding rock mechanical property matrix is ​​constructed based on the geological exploration data of the construction channel area, including rock hardness coefficient and joint development degree parameters; is the construction channel cross-sectional dimensions and buried depth structural parameter matrix; It is a coupling relationship function based on nonlinear mapping and is determined by a deep learning network trained with historical construction data.

[0067] Specifically, the excavation parameter-surrounding rock response coupling relationship model of step S4 is constructed. When implementing it, we first start with the geological exploration data of the construction channel area and construct a surrounding rock mechanical property matrix including parameters such as rock hardness coefficient and joint development degree. The rock hardness coefficient directly affects the TBM tool wear and excavation efficiency, and the joint development degree is related to the stability of the surrounding rock. These parameters are obtained through geological drilling, on-site investigation and other means, and the matrix is ​​constructed based on the spatial position distribution. At the same time, the excavation parameters such as thrust, torque, and cutter head speed of the TBM are collected, and combined with the structural parameters such as the cross-sectional size and burial depth of the construction channel, the excavation parameter matrix and the structural parameter matrix are constructed respectively. These three matrices are used as input, and the nonlinear mapping function determined by the deep learning network trained with historical construction data is used. , enabling modeling of the complex relationship between tunneling parameters and surrounding rock response. Based on current tunneling parameters, this model can deduce theoretical tunneling status data vectors in real time, including key information such as projected tunneling speed and surrounding rock deformation. Its significance lies in breaking the limitations of traditional empirical judgment of construction status. Through a data-driven approach, it accurately predicts the surrounding rock response to TBM tunneling during construction, providing a scientific basis for construction parameter adjustment and risk prediction, thereby ensuring construction safety and efficiency.

[0068] Preferably, in step S5, the deviation analysis algorithm based on the construction channel section shrinkage threshold adopts the following deviation calculation model:

[0069] ,in, is the comprehensive deviation measure; The actual measured construction channel characteristic parameter values; is the first Theoretical values ​​of characteristic parameters; The first The convergence threshold of the characteristic parameters; The number of feature parameters to be calculated for the participation bias.

[0070] Specifically, in practical applications, the deviation calculation model in step S5 first defines the convergence thresholds for each characteristic parameter specified in the construction channel design specifications. These thresholds are key indicators for measuring construction compliance. For example, the allowable range of channel cross-sectional convergence is directly related to channel stability. The measured construction channel characteristic parameter values ​​are then compared with the theoretical values ​​calculated using a coupled model of tunneling parameters and surrounding rock response. For each characteristic parameter, the difference between the actual and theoretical values ​​is calculated and normalized using the corresponding convergence threshold to eliminate the influence of different parameter dimensions. The square root of the sum of the normalized differences for all characteristic parameters involved in the deviation calculation is taken to obtain a comprehensive deviation metric. This process accurately quantifies the difference between the actual measured data and the theoretical data, visually reflecting the degree of deviation during TBM excavation in numerical form. During implementation, specialized data processing software or programming algorithms rapidly calculate large amounts of real-time construction data. Once the comprehensive deviation metric exceeds a preset safety threshold, the system immediately issues an alert, prompting construction personnel to take appropriate measures to ensure the construction process remains under control.

[0071] Preferably, in step S3, the association rule base is constructed using the following rule generation formula:

[0072] ,in, is a set of association rules; Conditional items formed by the combination of data features; is the corresponding abnormal behavior action item; is the probability that the condition item and the action item appear at the same time; is the probability of the conditional item appearing; It is the minimum confidence threshold set according to the risk level of the construction channel.

[0073] Specifically, during the construction of the association rule base in step S3, the implementation process first involves an in-depth analysis of the fused multi-source data, combining various features within the data to form conditional items. These features include construction channel geometry parameters, equipment operating parameters, and personnel operation behavior parameters, such as variations in channel cross-sectional dimensions, TBM torque fluctuations, and abnormal personnel operation time. Next, for each conditional item, its corresponding abnormal behavior action item is analyzed. Specifically, when a specific combination of conditions occurs, the construction anomaly that may be triggered, such as the risk of collapse caused by channel cross-sectional dimensions exceeding tolerance, is considered, and the corresponding action item for strengthening support is taken. Through statistical analysis of historical construction data, the probability of the conditional item and action item occurring simultaneously, as well as the probability of the conditional item occurring alone, is calculated. A minimum confidence threshold, determined by the risk level of the construction channel, is set. For channels with higher risk levels, the minimum confidence threshold is increased accordingly to ensure the reliability of the rules. When the ratio of the probability of the conditional item and action item occurring simultaneously to the probability of the conditional item occurring is greater than the minimum confidence threshold, the association between the conditional item and the action item is incorporated into the association rule base. After the association rule library is built, it can be used to screen the fused data in real time, quickly identify abnormal data features, provide rule support for construction safety supervision, and realize active monitoring and early warning of abnormal situations in the construction process.

[0074] Preferably, in step S6, the construction intervention decision set is generated using the following decision model:

[0075] ,in, To ultimately generate construction intervention decisions; is the set of all decision options; For decision-making The implementation cost is composed of equipment scheduling cost and material consumption cost; For decision-making The estimated execution time; For decision-making Residual risk value after implementation; The weight coefficients corresponding to cost, time and risk are determined according to the construction schedule and safety goals.

[0076] Specifically, during the generation of the construction intervention decision set in step S6, the actual deviation data is first used, combined with the construction process and resource allocation, to generate a set of all possible decision options. These options include different combinations of adjusting TBM excavation parameters (such as thrust and torque adjustment ranges) and personnel and equipment scheduling (number of personnel deployed, equipment movement routes). For each decision option, the implementation cost is calculated, including equipment scheduling costs (equipment movement and maintenance costs) and material consumption costs (support materials and tool wear costs). The expected execution time is estimated, taking into account factors such as the complexity of the implementation process and the time it takes for resources to be available. The residual risk value after implementation is assessed, combining construction safety regulations and historical accident data to analyze the degree of safety hazards remaining after the implementation of the option. Based on the construction schedule and safety objectives, weighting coefficients for cost, time, and risk are determined. For example, the time weighting coefficient increases when the construction schedule is tight, while the risk weighting coefficient increases when safety requirements are high. All options are evaluated using a decision model, and the option with the optimal overall cost, time, and risk is selected as the final construction intervention decision, forming the construction intervention decision set. After the decision set is generated, it is promptly distributed to the construction management terminal through the Buildertrend supervision system to guide construction personnel to adjust construction operations and achieve dynamic optimization and precise control of the construction process.

[0077] Preferably, the step S3 includes the following sub-steps:

[0078] Step S3-1: Establish a spatial topological structure model of the construction channel, divide the construction channel into multiple three-dimensional grid sub-areas, and determine the spatiotemporal benchmark for data collection in each sub-area; by constructing a three-dimensional grid system, assign independent spatiotemporal coordinates to data in different areas, forming a data spatial positioning framework to ensure the spatial consistency of subsequent data fusion.

[0079] Step S3-2: Perform a rough match between the initial measurement data set and the construction process behavior data set based on spatiotemporal tags to eliminate data records with spatiotemporal conflicts; perform a preliminary alignment of multi-source data based on spatiotemporal tags to identify and eliminate data with timestamp conflicts and spatial location conflicts, thereby improving the reliability of data fusion.

[0080] Step S3-3: Based on the design parameters of the construction channel, a multidimensional data screening template is constructed, including cross-sectional dimension thresholds and equipment operation parameter thresholds. A multidimensional threshold system is set based on the design parameters to form a standard template for data screening, providing a judgment basis for the accurate identification of abnormal data.

[0081] Step S3-4: Filter the roughly matched data through a multidimensional data screening template to extract abnormal data features; use the screening template to filter the data, identify data features that exceed the threshold range, and form an abnormal data feature set.

[0082] Specifically, in step S3-1, a spatial topological structure model of the construction channel is established. Based on the construction channel design drawings, the channel is divided into multiple three-dimensional grid sub-areas. The size of each sub-area is determined according to the channel size and accuracy requirements, so as to determine an accurate spatiotemporal benchmark for data acquisition and ensure accurate data spatial positioning. In step S3-2, the initial measurement data set and the construction process behavior data set are roughly matched according to the timestamp and spatial coordinates of the data acquisition, and data records with spatiotemporal contradictions are identified and eliminated, such as data of equipment in two different locations at the same time, to improve data quality. In step S3-3, a multidimensional data screening template containing cross-sectional dimension thresholds, equipment operating parameter thresholds, etc. is constructed based on the construction channel design parameters. This template acts as a "ruler" for data screening and provides a standard for abnormal data identification. In step S3-4, the roughly matched data is input into the multidimensional data screening template. The data is filtered one by one according to the threshold conditions set in the template, and abnormal data features that exceed the threshold range are extracted, thus realizing a complete processing flow of multi-source data from original acquisition to abnormal feature extraction, providing key data for subsequent construction status analysis.

[0083] Preferably, the step S4 includes the following sub-steps:

[0084] Step S4-1: Collect geological drilling data in the construction channel area and construct a spatial distribution model of surrounding rock mechanical parameters; integrate geological exploration data, establish a spatial distribution model of surrounding rock mechanical parameters, and digitally characterize the geological characteristics of the construction area.

[0085] Step S4-2: Obtain real-time measurement values ​​of the TBM's current excavation parameters, combine them with the construction channel structural parameters, and construct a parameter matrix for the current excavation condition. Combine the real-time excavation parameters with the construction channel structural parameters to form a parameter matrix reflecting the current construction status, providing a data basis for theoretical state deduction.

[0086] Step S4-3: Input the spatial distribution model of surrounding rock mechanical parameters and the current tunneling condition parameter matrix into the tunneling parameter-surrounding rock response coupling relationship model to calculate the theoretical tunneling state data; through the coupling relationship model, simulate the theoretical tunneling state of the TBM under the current working conditions and generate a theoretical state data vector.

[0087] Step S4-4: Perform temporal and spatial interpolation processing on the theoretical excavation status data to make it consistent with the temporal and spatial scales of the actual measurement data; perform scale conversion on the theoretical data through the temporal and spatial interpolation algorithm to ensure comparability with the actual measurement data in temporal and spatial dimensions.

[0088] Specifically, in step S4-1, geological drilling data of the construction channel area is collected, including core sample analysis, geological radar detection results, etc., and a spatial distribution model of the surrounding rock mechanical parameters is constructed using professional geological analysis software to intuitively present the differences in the mechanical properties of the surrounding rock at different locations. In step S4-2, the current tunneling parameter measurement values ​​of the TBM are obtained in real time. Combined with structural parameters such as the cross-sectional dimensions and burial depth of the construction channel, a current tunneling condition parameter matrix is ​​constructed in a specific data format to comprehensively reflect the current construction conditions. In step S4-3, the spatial distribution model of the surrounding rock mechanical parameters and the current tunneling condition parameter matrix are input into the tunneling parameter-surrounding rock response coupling relationship model, and the theoretical tunneling state data is calculated through the model's internal algorithm to achieve theoretical deduction of the construction state. In step S4-4, since the theoretical data and the actual measurement data may differ in time and space scales, the theoretical tunneling state data is processed in the time and space dimensions using an interpolation algorithm to match the actual measurement data in time frequency and spatial position, facilitating subsequent comparative analysis and ensuring that the theoretical data can accurately guide construction practice.

[0089] Preferably, the step S5 includes the following sub-steps:

[0090] Step S5-1: Align the theoretical excavation status data with the abnormal data features based on the feature dimension to establish a data mapping relationship; through the alignment of the feature dimension, a mapping bridge between the theoretical data and the actual data is constructed, laying the foundation for deviation calculation.

[0091] Step S5-2: Determine the deviation calculation weights for different characteristic parameters based on the cross-section convergence threshold of the construction channel; assign corresponding weights to different characteristic parameters based on the cross-section convergence threshold system, highlighting the role of key parameters in deviation calculation.

[0092] Step S5-3: Using a weighted Euclidean distance algorithm, calculate the deviation vector between the actual measurement data and the theoretical data; using the weighted Euclidean distance algorithm, quantify the difference between the actual and theoretical data to form a multi-dimensional deviation vector.

[0093] Step S5-4: Perform dimensionality reduction processing on the deviation vector, extract the calibrated deviation features, and generate actual deviation data; compress the deviation vector dimension through the dimensionality reduction algorithm, extract the deviation features of different construction nodes, and form a streamlined construction node deviation data set.

[0094] Specifically, in sub-step S5-1, the theoretical excavation status data and the abnormal data features are aligned according to the physical feature dimensions represented by the data, such as excavation speed, equipment posture, surrounding rock deformation, etc., and a data mapping relationship is established so that the two can be directly compared in the same dimension. In sub-step S5-2, based on the design specifications such as the convergence threshold of the construction channel section and the degree of influence of each characteristic parameter on construction safety and quality, the deviation calculation weight is determined to highlight the importance of key parameters in deviation analysis. In sub-step S5-3, a weighted Euclidean distance algorithm is used to quantify the differences between the actual measurement data and the theoretical data in each characteristic dimension, and a deviation vector reflecting the overall deviation is obtained to achieve a numerical expression of the deviation. In sub-step S5-4, a dimensionality reduction algorithm is used to process the deviation vector, remove redundant information, extract key deviation features, and generate concise and representative actual deviation data, providing accurate deviation information for construction decision-making, helping construction personnel to quickly locate problems and take effective measures.

[0095] like Figure 2 As shown in FIG, a method for supervising the construction passage of a hydropower station underground powerhouse based on continuous TBM excavation is implemented through different units, including:

[0096] The spatiotemporal synchronization acquisition unit for heterogeneous data sources is used to connect the ROBOTEC measurement system and the Buildertrend supervision system, respectively. By establishing a unified spatiotemporal coordinate reference, it synchronizes the acquisition and transmission of the initial measurement data set and the construction process behavior data set. Through a data interface adapter, this unit converts data from different systems into a unified format and synchronizes the acquisition at preset time intervals.

[0097] The multimodal data spatiotemporal fusion processing unit has its input connected to the spatiotemporal synchronization acquisition unit of heterogeneous data sources. Through the data fusion formula, it performs weighted fusion of geometric data and behavioral data in the spatiotemporal dimension. The unit adopts the spatiotemporal sliding window mechanism to dynamically fuse real-time data streams to ensure the timeliness of the data.

[0098] The tunneling-surrounding rock coupling behavior deduction and calculation unit has its input connected to the multimodal data spatiotemporal fusion processing unit. Based on the tunneling parameter-surrounding rock response coupling relationship model, it deduces the theoretical tunneling state of the TBM by inputting the current tunneling parameters and geological parameters. The unit has a built-in deep learning model that continuously optimizes the coupling relationship function through historical data.

[0099] The multi-dimensional deviation feature extraction and judgment unit has its input ends connected to the tunneling-surrounding rock coupling behavior deduction and calculation unit and the multimodal data spatiotemporal fusion processing unit respectively. Through the deviation analysis algorithm, the theoretical data and actual data are compared and analyzed; this unit adopts a parallel computing architecture to simultaneously process deviation calculation tasks in multiple dimensions.

[0100] The construction intervention decision intelligent generation unit has its input connected to the multi-dimensional deviation feature extraction and judgment unit. It uses the decision model to generate the optimal intervention plan based on the deviation data. The unit adopts a multi-objective optimization algorithm to seek the optimal balance between cost, time and risk.

[0101] The closed-loop control instruction execution feedback unit has its input connected to the construction intervention decision intelligent generation unit, and its output connected to the TBM control system and the on-site construction management terminal respectively. It issues and provides execution status feedback on the construction intervention decision set implemented through the industrial control network. This unit has fault diagnosis and redundant control functions to ensure the reliability of instruction execution.

[0102] The method for supervising the construction passages of underground powerhouses at hydropower stations, based on continuous TBM excavation, has undergone systematic technological innovation to address the two major shortcomings of existing technologies: the lagging data collection and processing capabilities, and the lack of accurate construction status deduction and intelligent decision-making mechanisms. At the data collection level, this method uses the ROBOTEC measurement system and the Buildertrend supervision system to build a comprehensive data collection network, enabling the simultaneous collection of multi-source heterogeneous data such as the three-dimensional spatial geometry data of the construction passage, personnel work information, and equipment operating condition data. This completely changes the situation of traditional methods where data is dispersed and difficult to coordinate. At the same time, by constructing a multi-dimensional perception data screening model, fusing the collected data in the spatial and temporal dimensions, and using an association rule library to extract abnormal data features, the efficiency and accuracy of data processing are significantly improved, solving the problem of delayed detection of abnormal situations.

[0103] In terms of construction status analysis and decision-making, this method establishes a coupled relationship model between tunneling parameters and surrounding rock responses based on the dynamic mechanical parameters of continuous TBM excavation and the geological parameters of the surrounding rock of the construction channel. This model can deeply analyze the inherent relationship between tunneling parameters and surrounding rock responses and accurately deduce the theoretical tunneling status of the TBM. Combined with a deviation analysis algorithm based on the convergence threshold of the construction channel section, a multi-dimensional feature comparison is performed between theoretical and actual data to accurately determine the actual deviation data during the construction process, changing the previous extensive model that relied on empirical judgment. In addition, by comprehensively considering factors such as cost, time, and risk, a scientific and reasonable set of construction intervention decisions can be generated and promptly fed back to the construction management terminal, achieving closed-loop management of the entire process from data collection and analysis to decision-making, greatly enhancing the scientific nature and reliability of construction supervision.

[0104] The application of this supervision method not only effectively addresses the shortcomings of existing technologies but also demonstrates significant advantages in multiple areas. It enables comprehensive, real-time, and precise supervision of the construction process, significantly improving construction safety. Through scientific data analysis and decision-making, it optimizes construction resource allocation and improves construction efficiency. It also provides a standardized, intelligent supervision solution for the construction of underground powerhouse construction passages at hydropower stations, significantly contributing to technological advancement in the hydropower engineering construction industry.

[0105] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "disposed," "installed," "connected," "connected," and "fixed" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; they may refer to mechanical connections or electrical connections; they may refer to direct connections or indirect connections through an intermediate medium; and they may refer to internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0106] While embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for supervising the construction passage of a hydropower station underground powerhouse based on continuous TBM excavation, characterized in that: The steps include: Step S1: Using the ROBOTEC measurement system, three-dimensional geometric data of the TBM continuous excavation operation area is collected according to a preset spatial coordinate network within the construction channel of the hydropower station's underground powerhouse. This initial measurement data set, including construction channel contour point cloud data and real-time TBM equipment posture data, is obtained. Step S2: Connect the Buildertrend monitoring system to the construction management network to simultaneously collect personnel operation information, equipment operation status information, and material flow data within the construction channel to build a construction process behavior data set; Step S3: Construct a multi-dimensional perception data screening model for the construction channel, fuse the initial measurement data set with the construction process behavior data set in the spatiotemporal dimension, and extract abnormal data features from the fused data by establishing an association rule base based on the channel design parameter threshold; Step S4: Based on the dynamic mechanical parameters of the TBM's continuous excavation and the geological parameters of the surrounding rock of the construction channel, a tunneling parameter-surrounding rock response coupling relationship model is established. The current TBM tunneling parameters are deduced in real time using this model to obtain theoretical tunneling status data; Step S5: performing a multi-dimensional feature comparison between the theoretical excavation state data and the abnormal data features, and using a deviation analysis algorithm based on the convergence threshold of the construction channel section to determine the actual deviation data during the TBM excavation process; Step S6: Based on the actual deviation data, the Buildertrend monitoring system generates a construction intervention decision set, including excavation parameter adjustment instructions and personnel and equipment scheduling plans. The construction intervention decision set is then fed back to the TBM control system and the on-site construction management terminal. In step S3, the construction channel multi-dimensional perception data screening model adopts the following data fusion formula: Among them, F fusion is the fused data vector; P geometry，i is the i-th geometric data vector in the initial measurement data set; W geometry，i is the spatial weight coefficient corresponding to the i-th geometric data vector, which is determined by the calibrated structural coordinates in the construction channel design drawings; P behavior，j is the jth behavior data vector in the construction process behavior data set; W behavior，j is the time weight coefficient corresponding to the j-th behavior data vector, which is calculated according to the construction process time constraint matrix; α and β are weighting factors that represent the importance of geometric data and behavior data, respectively, and are determined by the construction channel safety level parameters.

2. The method for supervising the construction passage of a hydropower station underground powerhouse based on TBM continuous excavation according to claim 1 is characterized in that: In step S4, the tunneling parameter-surrounding rock response coupling relationship model adopts the following modeling formula: R response =f(P TBM ·M geology ·S channel ) Among them, R response is the theoretical excavation state data vector; P TBM M is the tunneling parameter matrix including TBM thrust, torque, and cutterhead speed parameters; geology The surrounding rock mechanical property matrix is ​​constructed based on the geological exploration data of the construction channel area, including rock hardness coefficient and joint development degree parameters; S channel is the structural parameter matrix of the construction channel cross-sectional dimensions and burial depth; f is the coupling relationship function based on nonlinear mapping, which is determined by the deep learning network trained with historical construction data.

3. The method for supervising the construction passage of a hydropower station underground powerhouse based on TBM continuous excavation according to claim 1 is characterized in that: In step S5, the deviation analysis algorithm based on the construction channel section shrinkage threshold adopts the following deviation calculation model: Among them, D deviation is the comprehensive deviation measurement value; V actual，k is the characteristic parameter value of the kth construction channel obtained by actual measurement; V theory，k is the theoretical value of the kth characteristic parameter calculated by the excavation parameter-surrounding rock response coupling relationship model; T threshold，k is the convergence threshold of the kth characteristic parameter determined according to the construction channel design specification; l is the number of characteristic parameters involved in the deviation calculation.

4. The method for supervising the construction passage of a hydropower station underground powerhouse based on TBM continuous excavation according to claim 1 is characterized in that: In step S3, the association rule base is constructed using the following rule generation formula: Among them, R rule is the association rule set; C condition A is a conditional item formed by combining data features; action is the corresponding abnormal behavior action item; P(C condition ∩A action ) is the probability of condition item and action item appearing at the same time; P(C condition ) is the probability of the conditional item occurring; θ is the minimum confidence threshold set according to the risk level of the construction channel.

5. The method for supervising the construction passage of a hydropower station underground powerhouse based on TBM continuous excavation according to claim 1 is characterized in that: In step S6, the construction intervention decision set is generated using the following decision model: Among them, D decision is the final construction intervention decision; D is the set of all decision options; C cost (d) is the implementation cost of decision plan d, which is composed of equipment scheduling cost and material consumption cost; T time (d) is the expected execution time of decision plan d; R risk (d) is the residual risk value after the implementation of decision plan d; λ1, λ2, and λ3 are the weight coefficients corresponding to cost, time, and risk, respectively, which are determined according to the construction schedule and safety goals.

6. The method for supervising the construction passage of a hydropower station underground powerhouse based on TBM continuous excavation according to claim 1 is characterized in that: The step S3 includes the following sub-steps: Step S3-1: Establish a spatial topological structure model of the construction channel, divide the construction channel into multiple three-dimensional grid sub-regions, and determine the spatiotemporal reference for data collection in each sub-region. By constructing a three-dimensional grid system, independent spatiotemporal coordinates are assigned to data in different regions to form a data spatial positioning framework. Step S3-2: Perform a rough match between the initial measurement data set and the construction process behavior data set based on spatiotemporal tags to eliminate data records with spatiotemporal conflicts; perform a preliminary alignment of multi-source data based on spatiotemporal tags to identify and eliminate data with conflicting timestamps and spatial locations; Step S3-3: Construct a multi-dimensional data screening template including section size thresholds and equipment operation parameter thresholds based on the construction channel design parameters; Set a multi-dimensional threshold system based on design parameters to form a standard template for data screening; Step S3-4: Filter the roughly matched data through a multi-dimensional data screening template to extract abnormal data features; The data is filtered using the screening template to identify data features that exceed the threshold range and form a set of abnormal data features.

7. The method for supervising the construction passage of a hydropower station underground powerhouse based on TBM continuous excavation according to claim 1 is characterized in that: The step S4 includes the following sub-steps: Step S4-1: Collect geological drilling data in the construction channel area and construct a spatial distribution model of surrounding rock mechanical parameters; integrate geological exploration data, establish a spatial distribution model of surrounding rock mechanical parameters, and digitally characterize the geological characteristics of the construction area; Step S4-2: Acquire real-time measurement values ​​of the TBM's current excavation parameters and, in combination with the construction channel structural parameters, construct a current excavation condition parameter matrix; combine the real-time excavation parameters with the construction channel structural parameters to form a parameter matrix reflecting the current construction status; Step S4-3: Input the spatial distribution model of surrounding rock mechanical parameters and the current tunneling condition parameter matrix into the tunneling parameter-surrounding rock response coupling relationship model to calculate theoretical tunneling state data; Through the coupling relationship model, the theoretical excavation state of the TBM under the current working conditions is simulated to generate the theoretical state data vector; Step S4-4: Performing spatiotemporal interpolation processing on the theoretical excavation status data to make it consistent with the spatiotemporal scale of the actual measurement data; performing scale conversion on the theoretical data using a spatiotemporal interpolation algorithm.

8. The method for supervising the construction passage of a hydropower station underground powerhouse based on TBM continuous excavation according to claim 1 is characterized in that: The step S5 includes the following sub-steps: Step S5-1: Align the theoretical excavation status data with the abnormal data features based on the feature dimension to establish a data mapping relationship; through the alignment of the feature dimension, a mapping bridge is built between the theoretical data and the actual data; Step S5-2: Determine the deviation calculation weights of different characteristic parameters based on the cross-section convergence threshold of the construction channel; assign corresponding weights to different characteristic parameters based on the cross-section convergence threshold system; Step S5-3: using a weighted Euclidean distance algorithm to calculate the deviation vector between the actual measured data and the theoretical data; using the weighted Euclidean distance algorithm to quantify the difference between the actual and theoretical data to form a multi-dimensional deviation vector; Step S5-4: Perform dimensionality reduction processing on the deviation vector, extract the calibrated deviation features, and generate actual deviation data; compress the deviation vector dimension through the dimensionality reduction algorithm, extract the deviation features of different construction nodes, and form a construction node deviation data set.

Images