Construction management system for core rockfill dam in high-cold and high-altitude area based on digital twinning

Abstract

The application relates to the technical field of water conservancy and hydropower engineering, and discloses a high-cold high-altitude region core wall rockfill dam construction management system based on digital twinning, a construction module is used for constructing a digital twinning model; a preprocessing module is used for preprocessing real-time data; a prediction module is used for inputting the pretreated data into a space-time attention LSTM model to predict a construction risk probability in the next 72 hours; an adjustment module is used for simulating effects of different construction schemes based on the digital twinning model, dynamically adjusting a construction plan and personnel arrangement by adopting a Markov decision model, and generating corresponding regulation and control instructions; a comparison module is used for updating the digital twinning model every 5 minutes by adopting a PSO algorithm, displaying excavation progress, filling height and rolling tracks in real time, comparing deviations of the digital twinning model and an actual project, outputting a deviation analysis report, and iteratively optimizing the space-time attention LSTM model based on the deviation analysis report; and the application improves the efficiency of high-cold high-altitude region core wall rockfill dam construction.

Description

Technical Field

[0001] This invention relates to the field of water conservancy and hydropower engineering technology, specifically to a construction management system for core rockfill dams in high-altitude and cold regions based on digital twins. Background Technology

[0002] Core-wall rockfill dams are key structures in water conservancy and hydropower projects. However, the unique environment of high-altitude and cold regions (altitude ≥3000m, extreme low temperature ≤-15℃) presents numerous technical challenges to the construction of core-wall rockfill dams due to the widespread distribution of permafrost, low oxygen content, high wind speed, and significant diurnal temperature differences. First, the thawing of permafrost leads to large fluctuations in soil moisture content, and traditional empirical compaction parameters are prone to result in substandard compaction, requiring repeated rework. Second, the dynamic changes in the construction environment and the construction process are coupled, making it difficult for traditional 3D simulation technology to adapt to changes in working conditions in real time, resulting in a disconnect between construction progress and quality control. Third, existing digital twin platforms mostly focus on visualization and lack the ability to coordinate and optimize multiple objectives such as construction resources, ecological protection, and project quality. Furthermore, they have not established specific management and control mechanisms for issues such as equipment efficiency degradation and limited working hours in high-altitude and cold regions. Summary of the Invention

[0003] The purpose of this invention is to solve the above-mentioned problems by designing a construction management system for core wall rockfill dams in high-altitude and cold regions based on digital twins.

[0004] This invention provides a construction management system for core-wall rockfill dams in high-altitude and cold regions based on digital twins. The system includes: The module is used to build a digital twin model of a core-wall rockfill dam in a high-altitude and cold region. The digital twin model includes real-time data on dam structure, geological conditions, climate parameters, construction equipment status, and personnel activities. The preprocessing module is used to preprocess real-time data based on the digital twin model using a Kalman filter algorithm with an added temperature compensation factor. The prediction module is used to input preprocessed data into the spatiotemporal attention LSTM model to analyze the impact of changes in environmental parameters on the performance of dam materials and construction progress, and to predict the probability of construction risks in the next 72 hours. The adjustment module is used to simulate the effects of different construction schemes based on the digital twin model, and dynamically adjust the construction plan and personnel arrangement according to the construction risk probability using the Markov decision model to generate corresponding control instructions. The construction decision problem is modeled as a Markov decision process, and the hyperparameters of the Markov decision model are optimized using the adaptive chaotic gray wolf optimization algorithm. The comparison module is used to update the digital twin model every 5 minutes using the PSO algorithm, display the excavation progress, filling height, and compaction trajectory in real time, compare the deviation between the digital twin model and the actual project, output a deviation analysis report, and iteratively optimize the spatiotemporal attention LSTM model based on the deviation analysis report.

[0005] Optionally, in a first implementation of the present invention, the building module includes: The collection submodule is used to collect basic data on core-wall rockfill dams in high-altitude and cold regions and establish a basic database. The embedded submodule is used to build a three-dimensional geometric model based on the basic database, restore the spatial morphology of the dam structure, and embed geological layering information; A submodule is established to acquire real-time data through acquisition devices, associate the real-time data with the three-dimensional geometric model, establish a dynamic mapping relationship, and form a digital twin model containing multi-dimensional information on structure, environment, equipment, and personnel.

[0006] Optionally, in a second implementation of the present invention, the basic data includes at least dam design drawings, geological survey reports, historical climate data, construction equipment parameters, and personnel configuration plans.

[0007] Optionally, in a third implementation of the present invention, the preprocessing module includes: The first initialization submodule is used to initialize the Kalman filter parameters, including the state vector, the initial error covariance matrix, and the observation matrix. The determination submodule is used to statistically analyze the impact of temperature on sensors based on historical data from high-altitude and cold regions, determine the compensation coefficients for different temperature ranges, and obtain the temperature compensation factor. The prediction submodule is used to predict the state vector and error covariance at the current moment based on the optimal estimate and system state equation at the previous moment. A submodule is introduced to correct the observed values ​​by introducing a temperature compensation factor, and then the predicted values ​​are corrected by combining the Kalman gain to obtain the optimal estimate for the current time. The output submodule is used to continuously process real-time data and output preprocessed data after eliminating temperature interference.

[0008] Optionally, in a fourth implementation of the present invention, the prediction module includes: The sorting submodule is used to sort the preprocessed data into spatiotemporal sequence data according to the time and space dimensions. The spatiotemporal sequence data includes environmental parameters, dam parameters and construction progress data. The submodule is set up to build the basic framework of the LSTM network, including the input layer, hidden layer and output layer. The hidden layer is used to capture the temporal correlation of data. The computational submodule is used to introduce a spatial attention mechanism to calculate the weight values ​​of different grid regions and focus on the data characteristics of the core wall and dam abutment. The allocation submodule is used to introduce a time attention mechanism to assign weights to data in different time periods, thereby enhancing the contribution of data during freeze-thaw cycles. The training submodule is used to input spatiotemporally weighted sequence data into the LSTM network for training, adjust the model parameters through historical risk event data, and finally output the construction risk probability of each grid area in the next 72 hours.

[0009] Optionally, in a fifth implementation of the present invention, the adjustment module includes: Define a submodule to model the construction decision problem as a Markov decision process, defining the state space, action space and reward function; The update submodule is used to initialize the hyperparameters of the Markov decision model, optimize the hyperparameters using the Grey Wolf optimization algorithm, and update the Markov decision model. The selection submodule is used to simulate the execution effects of different construction schemes based on the optimized Markov decision model and the digital twin model, select the scheme with the largest reward function value, and generate construction plan adjustment instructions and personnel arrangement instructions.

[0010] Optionally, in the sixth implementation of the present invention, the state space includes the current construction progress, dam status, and environmental conditions; the action space includes adjusting construction intensity, equipment allocation, and personnel scheduling; and the reward function includes progress achievement rate and risk reduction rate.

[0011] Optionally, in a seventh implementation of the present invention, the update submodule includes: Initialize the parameters of the gray wolf optimization algorithm, including wolf pack size, maximum number of iterations, and search space boundary; Initialize the wolf pack positions, with each wolf's position corresponding to a set of hyperparameter combinations, and calculate the fitness of each individual wolf in the pack; Wolf packs are classified into different levels based on their fitness values. The individual with the highest fitness is selected as the alpha wolf, the two individuals with the second highest fitness are selected as the beta wolf and the delta wolf, and the remaining individuals are selected as the omega wolf. A chaotic perturbation mechanism is introduced to perform chaotic mapping on the position of the omega wolf, and the wolf pack search step size is adaptively adjusted. Determine if the maximum number of iterations has been reached. If so, recalculate the fitness of all individuals and update the wolf pack position. If not, stop the iteration and determine the hyperparameter combination corresponding to the current alpha wolf as the optimal hyperparameter. The optimal hyperparameters are substituted into the Markov decision model to replace the original hyperparameters, thus completing the model update.

[0012] Optionally, in the eighth implementation of the present invention, the comparison module includes: The second initialization submodule is used to initialize the particle swarm, setting the position and velocity range of the particles, where each particle represents a set of digital twin model parameters; The smoothing submodule is used to collect actual engineering data every 5 minutes, smooth the data through Kalman filtering, calculate the deviation value between the model output and the actual data for each particle, and use the deviation value as the fitness of the particle. The optimization submodule is used to update the global optimal position and individual optimal position of the particle swarm. Through iterative optimization of particle velocity and position, it finds the parameter combination that minimizes the model deviation and updates the digital twin model. The recording submodule is used to generate a deviation analysis report, which records the differences between the model and the actual project in key indicators and the reasons therefor. The deviation analysis results are fed back to the spatiotemporal attention LSTM model to adjust the input weights of the model and perform iterative optimization of the model.

[0013] Optionally, in the ninth implementation of the present invention, a method for implementing a construction management system for core-wall rockfill dams in high-altitude and cold regions based on digital twins is provided, comprising the following steps: Construct a digital twin model of a core-wall rockfill dam in a high-altitude and cold region. The digital twin model includes real-time data on dam structure, geological conditions, climate parameters, construction equipment status, and personnel activities. Based on the digital twin model, a Kalman filter algorithm with added temperature compensation factor is used to preprocess real-time data; The preprocessed data is input into the spatiotemporal attention LSTM model to analyze the impact of environmental parameter changes on dam material performance and construction progress, and to predict the probability of construction risks in the next 72 hours. The effects of different construction schemes are simulated based on a digital twin model, and the construction plan and personnel arrangement are dynamically adjusted according to the construction risk probability using a Markov decision model to generate corresponding control instructions. The construction decision problem is modeled as a Markov decision process, and the hyperparameters of the Markov decision model are optimized using an adaptive chaotic gray wolf optimization algorithm. The digital twin model is updated every 5 minutes using the PSO algorithm to display the excavation progress, filling height, and compaction trajectory in real time. The deviation between the digital twin model and the actual project is compared, a deviation analysis report is output, and the spatiotemporal attention LSTM model is iteratively optimized based on the deviation analysis report.

[0014] The technical solution provided by this invention involves constructing a digital twin model of a core-wall rockfill dam in high-altitude and cold regions; preprocessing real-time data using a Kalman filter algorithm with added temperature compensation factors based on the digital twin model; inputting the preprocessed data into a spatiotemporal attention LSTM model to analyze the impact of environmental parameter changes on dam material properties and construction progress, and predicting the probability of construction risks in the next 72 hours; simulating the effects of different construction schemes based on the digital twin model, and dynamically adjusting the construction plan and personnel arrangements according to the construction risk probability using a Markov decision model to generate corresponding control instructions; updating the digital twin model every 5 minutes using the PSO algorithm to display excavation progress, filling height, and compaction trajectory in real time, comparing the deviation between the digital twin model and the actual project, outputting a deviation analysis report, and iteratively optimizing the spatiotemporal attention LSTM model based on the deviation analysis report; this invention achieves 5-minute-level updates of the digital twin model, resolving the contradiction between the dynamic changes in working conditions and the static nature of the model in high-altitude and cold regions; and utilizing LSTM... By capturing the temporal correlation of data, the system accurately analyzes the long-term dynamic impact of environmental parameters on dam material performance and construction progress. It also leverages a spatiotemporal attention mechanism to focus on key areas such as the core wall and dam abutments, as well as data characteristics during critical periods like large diurnal temperature variations and strong winds, increasing the weight given to high-risk points. Simultaneously, the comparison module iteratively optimizes the model based on deviation analysis reports, further improving prediction accuracy and providing ample time for proactive risk mitigation. The adjustment module models construction decisions as a Markov decision process, fully considering the dynamic changes in construction conditions in high-altitude and cold regions. Through state transition analysis and reward function evaluation, it quantifies different construction schemes. The study considers both the benefits and risks of construction; simultaneously, it employs an adaptive chaotic gray wolf optimization algorithm to optimize model hyperparameters. This retains the global search capability of the gray wolf optimization algorithm while avoiding local optima through chaotic perturbations. It also adaptively adjusts the search step size to ensure the rapid discovery of optimal hyperparameters suitable for the current construction scenario. This allows the construction plan adjustments and personnel arrangement schemes output by the Markov decision model to better align with actual needs. By simulating the effects of different schemes using a digital twin model, the optimal scheme can be selected without actual trial and error, reducing trial and error costs and improving the flexibility and scientific nature of construction decisions. This continuously improves the quality, safety, and efficiency management level of core wall rockfill dam construction in high-altitude and cold regions. Attached Figure Description

[0015] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention.

[0016] Figure 1 This is a schematic diagram of the structure of the construction management system for core wall rockfill dams in high-altitude and cold regions based on digital twins, provided in an embodiment of the present invention. Figure 2This is a schematic diagram of the structure of the building module provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the preprocessing module provided in an embodiment of the present invention. Detailed Implementation

[0017] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” or “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0018] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 A schematic diagram of the structure of the construction management system for core-wall rockfill dams in high-altitude and cold regions based on digital twins provided in this embodiment of the invention. The system includes: The module is used to build a digital twin model of a core-wall rockfill dam in a high-altitude and cold region. The digital twin model includes real-time data on dam structure, geological conditions, climate parameters, construction equipment status, and personnel activities. The preprocessing module is used to preprocess real-time data based on the digital twin model using a Kalman filter algorithm with an added temperature compensation factor. The prediction module is used to input preprocessed data into the spatiotemporal attention LSTM model to analyze the impact of changes in environmental parameters on the performance of dam materials and construction progress, and to predict the probability of construction risks in the next 72 hours. The adjustment module is used to simulate the effects of different construction schemes based on the digital twin model, and dynamically adjust the construction plan and personnel arrangement according to the construction risk probability using the Markov decision model to generate corresponding control instructions. The construction decision problem is modeled as a Markov decision process, and the hyperparameters of the Markov decision model are optimized using the adaptive chaotic gray wolf optimization algorithm. The comparison module is used to update the digital twin model every 5 minutes using the PSO algorithm, display the excavation progress, filling height, and compaction trajectory in real time, compare the deviation between the digital twin model and the actual project, output a deviation analysis report, and iteratively optimize the spatiotemporal attention LSTM model based on the deviation analysis report.

[0019] In this embodiment, please refer to Figure 2The building blocks include: The collection submodule is used to collect basic data on core-wall rockfill dams in high-altitude and cold regions and establish a basic database. The basic data includes at least dam design drawings, geological survey reports, historical climate data, construction equipment parameters, and personnel configuration plans. The embedded submodule is used to build a three-dimensional geometric model based on the basic database, restore the spatial morphology of the dam structure, and embed geological layering information; A submodule is established to acquire real-time data through acquisition devices, associate the real-time data with the three-dimensional geometric model, establish a dynamic mapping relationship, and form a digital twin model containing multi-dimensional information on structure, environment, equipment, and personnel.

[0020] In this embodiment, when performing data collection and basic database establishment tasks, the dam design drawings are first specifically reviewed. This involves collecting not only macroscopic dimensional data such as the overall dam type, core wall thickness, and slope of the rockfill area, but also extracting key material parameters such as the design compaction degree and permeability coefficient of the core wall fill material, as well as the design scheme for the dam foundation treatment, ensuring that the drawing data covers the structural and material details required for construction. Next, the geological survey report is analyzed in depth, focusing on extracting geological information unique to high-altitude and cold regions, including the distribution range of permafrost, freeze-thaw cycles, degree of rock weathering, groundwater depth, and seasonal variation patterns. The locations of potentially unfavorable geological bodies that may affect the stability of the dam are also marked. In the historical climate data collection phase, the focus is on the extreme low temperature values ​​and daily average temperatures of the construction area over the past 10 years. Data on temperature variation curves, snowfall, wind speed and direction, and permafrost depth are categorized and organized by quarter to reflect seasonal climate characteristics. For construction equipment parameters, priority is given to collecting information on equipment adapted to high-altitude and cold environments, including the operating temperature range and efficiency of cold-resistant compactors, the lifting capacity and climbing speed of low-temperature cranes, and equipment failure rates and maintenance cycles. For personnel configuration plans, information such as the high-altitude operation qualifications of workers, average working hours, and cold-weather protective equipment configuration standards needs to be collected. Finally, all collected data are categorized and stored according to four major categories: structural design, geological conditions, climate environment, and construction resources. A relational database is used to establish data association indexes to ensure that various types of data can be quickly retrieved according to construction stages, dam areas, and other dimensions, forming a basic database to support subsequent modeling. When constructing the 3D geometric model, the dam design drawings in the basic database are used as the basis. 3D modeling software suitable for hydraulic engineering, such as BIM software Revit or Civil 3D, is selected to recreate the overall spatial form of the dam at a 1:1 scale. For key structural components such as the core wall, rockfill zone, transition layer, and filter layer, the dimensional parameters in the design drawings, such as the top and bottom width of the core wall and the layer thickness of the rockfill zone, must be strictly followed for detailed modeling to ensure that the spatial position and connection relationships of each structure are consistent with the design. During the model construction process, the coordinate information of key control points of the dam, such as the dam axis, dam crest elevation, and dam base outline, is simultaneously marked to provide a spatial reference for subsequent association with real-time data. The 3D geometry is then completed. After the model is built, the geological stratification information embedding process is initiated: geological stratification data is extracted from the geological survey report in the basic database, including the thickness, distribution range, and physical and mechanical parameters of each soil and rock layer. Through the attribute editing function of the modeling software, the geological information of each layer is bound to the corresponding spatial area of ​​the three-dimensional geometric model. For example, in the dam foundation area of ​​the model, the spatial range of the frozen soil layer and the non-frozen soil layer is divided according to the frozen soil depth marked in the survey report, and the parameters of the corresponding soil layer are entered into the model attribute library. At the same time, a correlation mechanism between geological information and the spatial location of the model is established, so that when any area of ​​the model is clicked, the geological stratification data of that location can be retrieved in real time, realizing the deep integration of the three-dimensional geometric model and geological information. First, the deployment and debugging of the data acquisition equipment were completed: Based on the dam structure and geological conditions information in the basic database, cold-resistant data acquisition equipment was deployed in key monitoring areas of the dam, such as the core wall axis, dam shoulder slope, and the top of the dam foundation anti-seepage wall. Displacement monitoring used low-temperature GNSS receivers, dam strain monitoring used fiber optic strain sensors, environmental parameter monitoring deployed multi-parameter weather stations capable of collecting air temperature, wind speed, and frost depth, construction equipment status monitoring was achieved by installing IoT sensors such as temperature and speed sensors on cold-resistant compactors and cranes, and personnel activity monitoring was achieved by equipping workers with cold-resistant smart terminals with positioning capabilities. After all data acquisition equipment was debugged, it was connected to the data acquisition system via the LoRa wireless transmission protocol. The platform sets the real-time data acquisition frequency and then establishes a dynamic mapping relationship between real-time data and the 3D geometric model. First, data association rules are defined in the data acquisition platform to bind the installation location coordinates of each acquisition device to the corresponding spatial coordinates in the 3D geometric model, so that the acquired real-time data can be automatically matched to the corresponding position in the model. Next, a data synchronization interface is developed to realize real-time data interaction between the acquisition platform and the 3D model software. When the acquisition device acquires new data, the data is automatically transmitted to the model through the interface to update the status parameters of the corresponding structure, equipment, or personnel in the model. Finally, multi-dimensional information of structure, environment, equipment, and personnel is integrated to form a dynamically updated digital twin model, realizing real-time linkage between the physical engineering and the virtual model.

[0021] In this embodiment, please refer toFigure 3 The preprocessing module includes: The first initialization submodule is used to initialize the Kalman filter parameters, including the state vector, the initial error covariance matrix, and the observation matrix. The determination submodule is used to statistically analyze the impact of temperature on sensors based on historical data from high-altitude and cold regions, determine the compensation coefficients for different temperature ranges, and obtain the temperature compensation factor. The prediction submodule is used to predict the state vector and error covariance at the current moment based on the optimal estimate and system state equation at the previous moment. A submodule is introduced to correct the observed values ​​by introducing a temperature compensation factor, and then the predicted values ​​are corrected by combining the Kalman gain to obtain the optimal estimate for the current time. The output submodule is used to continuously process real-time data and output preprocessed data after eliminating temperature interference.

[0022] In this embodiment, the prediction module includes: The sorting submodule is used to sort the preprocessed data into spatiotemporal sequence data according to the time and space dimensions. The spatiotemporal sequence data includes environmental parameters, dam parameters and construction progress data. The submodule is set up to build the basic framework of the LSTM network, including the input layer, hidden layer and output layer. The hidden layer is used to capture the temporal correlation of data. The computational submodule is used to introduce a spatial attention mechanism to calculate the weight values ​​of different grid regions and focus on the data characteristics of the core wall and dam abutment. The allocation submodule is used to introduce a time attention mechanism to assign weights to data in different time periods, thereby enhancing the contribution of data during freeze-thaw cycles. The training submodule is used to input spatiotemporally weighted sequence data into the LSTM network for training, adjust the model parameters through historical risk event data, and finally output the construction risk probability of each grid area in the next 72 hours.

[0023] In this embodiment, the adjustment module includes: Define a submodule to model the construction decision problem as a Markov decision process, defining the state space, action space and reward function; The update submodule is used to initialize the hyperparameters of the Markov decision model, optimize the hyperparameters using the Grey Wolf optimization algorithm, and update the Markov decision model. The selection submodule is used to simulate the execution effects of different construction schemes based on the optimized Markov decision model and the digital twin model, select the scheme with the largest reward function value, and generate construction plan adjustment instructions and personnel arrangement instructions.

[0024] In this embodiment, the state space includes the current construction progress, dam status, and environmental conditions; the action space includes adjusting construction intensity, equipment allocation, and personnel scheduling; and the reward function includes progress achievement rate and risk reduction rate.

[0025] In this embodiment, the update submodule includes: When initializing the parameters of the gray wolf optimization algorithm, the parameter values ​​need to be determined in light of the special characteristics of core wall rockfill dam construction in high-altitude and cold regions. The wolf pack size needs to balance computational efficiency and search diversity, usually selecting 15-25 wolves to avoid insufficient search due to an excessively small size, and to prevent excessive computational load due to an excessively large size. The maximum number of iterations is set according to the difficulty of hyperparameter optimization, generally 30-50 times, to ensure that optimization is completed within the time frame required for construction decision-making, such as the daily construction plan adjustment cycle. The search space boundary is strictly defined according to the engineering physical meaning of the hyperparameters of the Markov decision model. For example, the adjustment coefficient of the state transition probability matrix is ​​limited to 0-1 to conform to the probability value range, the discount factor is limited to 0.6-0.9 to adapt to the medium- and long-term decision-making needs of construction in high-altitude and cold regions, and the sum of the reward function weight parameters is set to 1 to ensure reasonable allocation of target weights. When initializing the wolf pack positions, within the preset search space boundary, specific values ​​for each set of hyperparameters are generated through uniform random sampling. Each wolf's position corresponds to a complete set of hyperparameters, including state transition adjustment coefficients, discount factors, and three types of reward weight parameters, ensuring that the initial population covers the entire feasible domain of the hyperparameters. When calculating individual fitness, this set of hyperparameters is substituted into a Markov decision model to simulate typical construction scenarios in cold environments, such as compaction operations during low-temperature freeze-thaw periods and personnel scheduling in strong winds. By comparing the decision results output by the model with historical best construction cases, such as successful progress control schemes under the same environmental conditions, the fitness value is quantified based on the degree of matching in key indicators such as progress consistency, risk reduction rate, and cost control level. The higher the matching degree, the greater the fitness value. The wolf pack is sorted in descending order based on its fitness value, and then categorized into different levels. The individual ranked first is designated as the alpha wolf, whose hyperparameter combination represents the currently found optimal solution and serves as the core guide for the global search. The second and third ranked individuals are designated as beta and delta wolves, respectively, assisting the alpha wolf in guiding the search direction and providing reference information for suboptimal solutions. The remaining individuals are classified as omega wolves, primarily following the guidance of the alpha, beta, and delta wolves to expand the exploration range of the high-quality solution region. The core purpose of this level division is to establish a hierarchical search mechanism, allowing high-quality individuals to dominate the search direction while ensuring the breadth of the search through a large number of omega wolves, thus avoiding the omission of potential optimal solutions. When introducing a chaotic perturbation mechanism, individuals with low fitness among the omega wolves are typically the bottom 50%. A chaotic sequence is generated using a Logistic chaotic mapping. This sequence is then combined with the current position parameters of the omega wolves, and the values ​​of their hyperparameter combinations are adjusted through nonlinear transformations. This allows the positions of omega wolves, which were originally concentrated in local areas, to spread to new potential high-quality areas, increasing population diversity and preventing the algorithm from getting stuck in local optima. This is especially suitable for scenarios in high-altitude and cold regions where construction parameters fluctuate greatly and the optimal solution may change abruptly. When adaptively adjusting the search step size, the step size coefficient is dynamically adjusted according to the current iteration number: the step size is larger in the early stages of iteration, such as 80%-100% of the initial step size, to enhance the global search capability and cover a wider hyperparameter space; in the later stages of iteration, the step size is gradually reduced, such as to 20%-30% of the initial step size, to focus on the area near the alpha wolves for fine-grained searching and improve optimization accuracy. The algorithm determines whether the current iteration count has reached the preset maximum iteration count. If not, it recalculates the fitness of all individuals, including the omega wolves after chaotic perturbation and step size adjustment, as well as the alpha, beta, and delta wolves. The wolf pack level is updated by comparing the new fitness values. If the new fitness of an omega wolf exceeds that of the original alpha wolf, the alpha wolf is updated to that individual, and the affiliation of the beta and delta wolves is adjusted accordingly. Subsequently, the distance parameters between wolves are recalculated based on the updated level to guide the omega wolves to move towards new high-quality individuals, completing the iterative update of the wolf pack position. If the maximum iteration count has been reached, the search process stops, and the hyperparameter combination corresponding to the alpha wolf in the current level is determined as the optimal hyperparameter. This combination has been optimized through multiple rounds of iterations and can best adapt to the construction decision-making scenario of high-altitude and cold-weather core rockfill dams. The determined optimal hyperparameters are substituted into the Markov decision model, and the original hyperparameters in the model are replaced one by one: the new state transition probability matrix is ​​used to adjust the coefficients to correct the transition probabilities between different construction states, such as normal operation, freeze-thaw effects, and equipment failure, so that it is more in line with the state evolution law in high-altitude and cold environments; the optimized discount factor is used to balance the weight of immediate benefits, such as daily progress, and long-term risks, such as dam stability, in construction decisions; the new reward function weight parameters are used to adjust the proportion of progress, quality, and safety objectives in decision evaluation, highlighting the priority of safety indicators such as frost heave prevention in high-altitude and cold regions; after the hyperparameter replacement is completed, the Markov decision model can more accurately simulate the construction decision process, and the output control instructions, such as construction plan adjustments and personnel scheduling, will be more adapted to the special construction needs of high-altitude and cold regions.

[0026] In this embodiment, the comparison module includes: The second initialization submodule is used to initialize the particle swarm, setting the position and velocity range of the particles, where each particle represents a set of digital twin model parameters; The smoothing submodule is used to collect actual engineering data every 5 minutes, smooth the data through Kalman filtering, calculate the deviation value between the model output and the actual data for each particle, and use the deviation value as the fitness of the particle. The optimization submodule is used to update the global optimal position and individual optimal position of the particle swarm. Through iterative optimization of particle velocity and position, it finds the parameter combination that minimizes the model deviation and updates the digital twin model. The recording submodule is used to generate a deviation analysis report, which records the differences between the model and the actual project in key indicators and the reasons therefor. The deviation analysis results are fed back to the spatiotemporal attention LSTM model to adjust the input weights of the model and perform iterative optimization of the model.

[0027] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A construction management system for core-wall rockfill dams in high-altitude and cold regions based on digital twins, characterized in that: The system includes: The module is used to build a digital twin model of a core-wall rockfill dam in a high-altitude and cold region. The digital twin model includes real-time data on dam structure, geological conditions, climate parameters, construction equipment status, and personnel activities. The preprocessing module is used to preprocess real-time data based on the digital twin model using a Kalman filter algorithm with an added temperature compensation factor. The prediction module is used to input preprocessed data into the spatiotemporal attention LSTM model to analyze the impact of changes in environmental parameters on the performance of dam materials and construction progress, and to predict the probability of construction risks in the next 72 hours. The adjustment module is used to simulate the effects of different construction schemes based on the digital twin model, and dynamically adjust the construction plan and personnel arrangement according to the construction risk probability using the Markov decision model to generate corresponding control instructions. The construction decision problem is modeled as a Markov decision process, and the hyperparameters of the Markov decision model are optimized using the adaptive chaotic gray wolf optimization algorithm. The comparison module is used to update the digital twin model every 5 minutes using the PSO algorithm, display the excavation progress, filling height, and compaction trajectory in real time, compare the deviation between the digital twin model and the actual project, output a deviation analysis report, and iteratively optimize the spatiotemporal attention LSTM model based on the deviation analysis report.

2. The construction management system for core-wall rockfill dams in high-altitude and cold regions based on digital twins as described in claim 1, characterized in that, The building module includes: The collection submodule is used to collect basic data on core-wall rockfill dams in high-altitude and cold regions and establish a basic database. The embedded submodule is used to build a three-dimensional geometric model based on the basic database, restore the spatial morphology of the dam structure, and embed geological layering information; A submodule is established to acquire real-time data through acquisition devices, associate the real-time data with the three-dimensional geometric model, establish a dynamic mapping relationship, and form a digital twin model containing multi-dimensional information on structure, environment, equipment, and personnel.

3. The construction management system for core-wall rockfill dams in high-altitude and cold regions based on digital twins as described in claim 2, characterized in that, The basic data includes at least dam design drawings, geological survey reports, historical climate data, construction equipment parameters, and personnel configuration plans.

4. The construction management system for core-wall rockfill dams in high-altitude and cold regions based on digital twins as described in claim 1, characterized in that, The preprocessing module includes: The first initialization submodule is used to initialize the Kalman filter parameters, including the state vector, the initial error covariance matrix, and the observation matrix. The determination submodule is used to statistically analyze the impact of temperature on sensors based on historical data from high-altitude and cold regions, determine the compensation coefficients for different temperature ranges, and obtain the temperature compensation factor. The prediction submodule is used to predict the state vector and error covariance at the current moment based on the optimal estimate and system state equation at the previous moment. A submodule is introduced to correct the observed values ​​by introducing a temperature compensation factor, and then the predicted values ​​are corrected by combining the Kalman gain to obtain the optimal estimate for the current time. The output submodule is used to continuously process real-time data and output preprocessed data after eliminating temperature interference.

5. The construction management system for core-wall rockfill dams in high-altitude and cold regions based on digital twins as described in claim 1, characterized in that, The prediction module includes: The sorting submodule is used to sort the preprocessed data into spatiotemporal sequence data according to the time and space dimensions. The spatiotemporal sequence data includes environmental parameters, dam parameters and construction progress data. The submodule is set up to build the basic framework of the LSTM network, including the input layer, hidden layer and output layer. The hidden layer is used to capture the temporal correlation of data. The computational submodule is used to introduce a spatial attention mechanism to calculate the weight values ​​of different grid regions and focus on the data characteristics of the core wall and dam abutment. The allocation submodule is used to introduce a time attention mechanism to assign weights to data in different time periods, thereby enhancing the contribution of data during freeze-thaw cycles. The training submodule is used to input spatiotemporally weighted sequence data into the LSTM network for training, adjust the model parameters through historical risk event data, and finally output the construction risk probability of each grid area in the next 72 hours.

6. The construction management system for core-wall rockfill dams in high-altitude and cold regions based on digital twins as described in claim 1, characterized in that, The adjustment module includes: Define a submodule to model the construction decision problem as a Markov decision process, defining the state space, action space and reward function; The update submodule is used to initialize the hyperparameters of the Markov decision model, optimize the hyperparameters using the Grey Wolf optimization algorithm, and update the Markov decision model. The selection submodule is used to simulate the execution effects of different construction schemes based on the optimized Markov decision model and the digital twin model, select the scheme with the largest reward function value, and generate construction plan adjustment instructions and personnel arrangement instructions.

7. The construction management system for core-wall rockfill dams in high-altitude and cold regions based on digital twins as described in claim 6, characterized in that, The state space includes the current construction progress, dam status, and environmental conditions; the action space includes adjusting construction intensity, equipment allocation, and personnel scheduling; and the reward function includes progress achievement rate and risk reduction rate.

8. The construction management system for core-wall rockfill dams in high-altitude and cold regions based on digital twins as described in claim 6, characterized in that, The update submodule includes: Initialize the parameters of the gray wolf optimization algorithm, including wolf pack size, maximum number of iterations, and search space boundary; Initialize the wolf pack positions, with each wolf's position corresponding to a set of hyperparameter combinations, and calculate the fitness of each individual in the wolf pack; Wolf packs are classified into different levels based on their fitness values. The individual with the highest fitness is selected as the alpha wolf, the two individuals with the second highest fitness are selected as the beta wolf and the delta wolf, and the remaining individuals are selected as the omega wolf. A chaotic perturbation mechanism is introduced to perform chaotic mapping on the position of the omega wolf, and the wolf pack search step size is adaptively adjusted. Determine if the maximum number of iterations has been reached. If so, recalculate the fitness of all individuals and update the wolf pack position. If not, stop the iteration and determine the hyperparameter combination corresponding to the current alpha wolf as the optimal hyperparameter. The optimal hyperparameters are substituted into the Markov decision model to replace the original hyperparameters, thus completing the model update.

9. The construction management system for core-wall rockfill dams in high-altitude and cold regions based on digital twins as described in claim 1, characterized in that, The comparison module includes: The second initialization submodule is used to initialize the particle swarm, setting the position and velocity range of the particles, where each particle represents a set of digital twin model parameters; The smoothing submodule is used to collect actual engineering data every 5 minutes, smooth the data through Kalman filtering, calculate the deviation value between the model output and the actual data for each particle, and use the deviation value as the fitness of the particle. The optimization submodule is used to update the global optimal position and individual optimal position of the particle swarm. Through iterative optimization of particle velocity and position, it finds the parameter combination that minimizes the model deviation and updates the digital twin model. The recording submodule is used to generate a deviation analysis report, which records the differences between the model and the actual project in key indicators and the reasons therefor. The deviation analysis results are fed back to the spatiotemporal attention LSTM model to adjust the input weights of the model and perform iterative optimization of the model.

10. A method for implementing the construction management system for core-wall rockfill dams in high-altitude and cold regions based on digital twins as described in claim 1, characterized in that, The method includes the following steps: Construct a digital twin model of a core-wall rockfill dam in a high-altitude and cold region. The digital twin model includes real-time data on dam structure, geological conditions, climate parameters, construction equipment status, and personnel activities. Based on the digital twin model, a Kalman filter algorithm with added temperature compensation factor is used to preprocess real-time data; The preprocessed data is input into the spatiotemporal attention LSTM model to analyze the impact of environmental parameter changes on dam material performance and construction progress, and to predict the probability of construction risks in the next 72 hours. The effects of different construction schemes are simulated based on a digital twin model, and the construction plan and personnel arrangement are dynamically adjusted according to the construction risk probability using a Markov decision model to generate corresponding control instructions. The construction decision problem is modeled as a Markov decision process, and the hyperparameters of the Markov decision model are optimized using an adaptive chaotic gray wolf optimization algorithm. The digital twin model is updated every 5 minutes using the PSO algorithm to display the excavation progress, filling height, and compaction trajectory in real time. The deviation between the digital twin model and the actual project is compared, a deviation analysis report is output, and the spatiotemporal attention LSTM model is iteratively optimized based on the deviation analysis report.

Images