Dynamic subway construction progress regulation and control method and system based on digital twinning

Abstract

The invention relates to a subway construction progress dynamic regulation and control method and system based on digital twinning. According to the method, a digital twin model is constructed by collecting multi-source heterogeneous data of a construction site, and a multi-dimensional construction progress characterization model is generated. Based on the model, progress deviation calculation and risk prediction are carried out to obtain a current progress deviation value. And once the value exceeds a preset threshold value, generating a regulation and control scheme set by using a multi-objective optimization algorithm. Then, the schemes are simulated and verified in a digital twinning environment, and a feasible regulation and control scheme set is formed to guide actual construction progress regulation and control. According to the method, by means of the digital twinborn technology and multi-source data fusion, the construction state is presented in real time, progress deviation and risks are rapidly recognized, an effective regulation and control strategy is formulated through an optimization algorithm, the construction progress management efficiency and accuracy are improved, the capability of responding to environment changes and emergencies is enhanced, construction delay and resource waste are reduced, and the construction progress management efficiency and accuracy are improved. And the subway construction management level and quality are optimized.

Description

Technical Field

[0001] The present invention belongs to the field of construction progress control, and in particular relates to a method and system for dynamic control of subway construction progress based on digital twins. Background Art

[0002] With the rapid development of digital twin and IoT technologies, a new approach to construction progress management, leveraging virtual-reality mapping, has emerged in the field of intelligent construction. This digital twin-based technology, by creating a digital mirror of the physical construction scene and combining it with real-time data-driven dynamic simulation, comprehensively reflects changes in construction status. Its significant features include high real-time performance, multi-dimensional integration, and predictive maintenance, providing an innovative solution for complex subway project construction management.

[0003] In actual subway operation and maintenance, construction management faces many challenges. On the one hand, there are few large-scale cross-month construction projects, but the construction tasks in a single night are intensive (such as 4-5 disciplines running in parallel), and they are mainly local maintenance, inspection and inspection. Traditional methods (such as Gantt charts) make it difficult to track the progress details of a single construction in real time, and are even more unable to identify resource conflicts and risks during multi-disciplinary cross-construction (such as the temporal and spatial overlap of power maintenance and track maintenance), resulting in low construction efficiency and frequent safety hazards. On the other hand, for the construction / maintenance plan of a certain discipline (such as the power system) within the annual or multi-year equipment operation cycle, existing static management tools (such as BIM offline simulation) cannot dynamically respond to long-term factors such as equipment aging and environmental changes, resulting in a disconnect between the preview results and actual needs. For example, the periodic maintenance plan of power equipment may fail due to sudden failures or insufficient resources, but there is a lack of a real-time adjustment mechanism.

[0004] The core challenges facing current construction progress management primarily lie in the lag of response mechanisms, insufficient forecasting accuracy, isolated control measures, and difficulty in reusing experience. Traditional management methods struggle to timely capture progress deviations during long-term construction processes, fail to effectively integrate the coupled influence of environmental parameters and resource constraints, and lack synergy in resource allocation, process adjustments, and emergency response plan development. Furthermore, valuable experience from historical construction cases is not systematically applied to current project decision-making processes. These limitations severely restrict the efficiency and accuracy of large-scale subway construction management. Summary of the Invention

[0005] Based on this, it is necessary to provide a method and system for dynamic control of subway construction progress based on digital twins to address the above technical problems.

[0006] In the first aspect, this application provides a method for dynamic control of subway construction progress based on digital twins, including:

[0007] Obtain multi-source heterogeneous data from the construction site collected by the IoT perception layer, perform digital twin modeling based on the multi-source heterogeneous data, and generate a multi-dimensional construction progress representation model. The multi-source heterogeneous data includes historical construction data, real-time monitoring data, and environmental parameters. The historical data includes BIM model data and construction organization design data.

[0008] Calculate progress deviation and predict risks based on a multi-dimensional construction progress representation model to generate the current progress deviation value;

[0009] When the current progress deviation value exceeds the preset threshold, a set of control solutions is generated through a multi-objective optimization algorithm; the control solution set includes resource allocation, process adjustment and emergency measures;

[0010] The set of control schemes is simulated and verified in the digital twin environment to generate a set of verified feasible control schemes; the set of feasible control schemes is used to indicate the progress control operations in the actual construction process.

[0011] In one embodiment, after performing digital twin modeling based on multi-source heterogeneous data from the construction site to generate a multi-dimensional construction progress representation model, the process further includes:

[0012] Conduct construction process network modeling based on BIM model data and construction organization design data to generate a directed acyclic graph of the construction process;

[0013] A multi-dimensional feature vector is defined for each process node of the directed acyclic graph of the construction process to generate a feature vector set; the feature vector set includes time parameters, engineering quantity parameters, resource demand parameters and environmental sensitivity parameters;

[0014] The progress state transfer function is trained based on the feature vector set to generate a prediction model; the prediction model is used to predict the changing trend of the future construction progress; and the changing trend is used to assist in optimizing the set of feasible control schemes.

[0015] In one embodiment, training a progress state transition function based on a feature vector set to generate a prediction model includes:

[0016] Construct training samples based on historical construction data to generate a training data set; the training data set includes feature vectors, environmental parameters, and state changes;

[0017] Perform bidirectional LSTM neural network modeling on the training data set to generate an initial prediction model;

[0018] The parameters of the initial prediction model are optimized to generate a prediction model.

[0019] In one embodiment, a progress deviation calculation and risk prediction are performed based on a multi-dimensional construction progress representation model to generate a current progress deviation value, including:

[0020] Based on the planned progress data in the multi-dimensional construction progress representation model and the actual progress data collected in real time, the real-time deviation value of each process is generated through time-weighted deviation calculation. The real-time deviation value is calculated using the following formula:

[0021]

[0022] Among them, Δ i (t) is the real-time deviation value of process i, i is the i-th process node in the construction network, t is the time point for calculating the deviation, T i is the planned completion time of the process, is the planned progress value of process i at time t, is the actual progress value of process i at time t, α is the attenuation coefficient of the control weight;

[0023] The real-time deviation values ​​of each process are aggregated to generate the current progress deviation value that represents the overall construction status. The current progress deviation value is calculated using the following formula:

[0024]

[0025] Among them, Δ(t) is the current progress deviation value, n is the number of process i, ω i is the criticality weight of process i in the construction network.

[0026] In one embodiment, after aggregating the real-time deviation values ​​of each process to generate a current progress deviation value representing the overall construction status, the following steps are further included:

[0027] Based on the current progress deviation value and the logical dependency relationship between processes, risk propagation analysis is performed to generate critical path impact factors;

[0028] Based on the critical path influencing factors and the current progress deviation value, a comprehensive risk assessment is conducted to generate a future risk probability distribution map; based on the future risk probability distribution map, a multi-objective optimization problem model is constructed;

[0029] Based on the multi-source heterogeneous data in the multi-dimensional construction progress representation model, the constraint conditions are defined and processed to generate a constraint condition set for multi-objective optimization;

[0030] Based on the set of constraints, the multi-objective optimization problem model is solved to generate a set of control schemes.

[0031] In one embodiment, based on a set of constraints, a multi-objective optimization problem model is solved to generate a set of control solutions, including:

[0032] Dynamically adjust reference points based on the characteristic data of the current construction stage to generate a reference point distribution plan that adapts to the construction stage;

[0033] Based on the reference point distribution scheme adapted to the construction stage, the optimization solution data matching the current construction stage is selected from the historical case library, and the population is initialized to generate the initial optimization solution set;

[0034] Based on the initial optimization solution set and combined with the iterative process parameters of the optimization algorithm, the crossover and mutation probability is adaptively adjusted to generate a set of dynamically adjusted control schemes.

[0035] In one embodiment, after the control scheme set is simulated and verified in a digital twin environment to generate a verified feasible control scheme set, the following steps are further included:

[0036] Based on the actual construction data after the execution of the feasible control scheme set, the model parameter update amount is calculated to generate the parameter adjustment plan of the digital twin model; the model parameter update amount is calculated using the following formula:

[0037]

[0038] Among them, Δθ is the model parameter update amount, is the composite loss function, η is the learning rate, μ is the momentum coefficient, Δθ p is the historical update amount;

[0039] Based on the parameter adjustment scheme, the case database similarity matrix of the newly generated construction case data is updated to generate an updated case matching feature system;

[0040] Based on the updated case matching feature system and combined with the actual progress deviation observation data, the environmental response coefficient is adjusted to generate an optimized environmental parameter response scheme; the environmental parameter response scheme is used to indicate the adjustment of the constraint condition set of the multi-objective optimization.

[0041] Secondly, this application also provides a subway construction progress dynamic control system based on digital twins, which includes:

[0042] The digital twin modeling module is used to obtain multi-source heterogeneous data of the construction site collected by the IoT perception layer, perform digital twin modeling processing based on the multi-source heterogeneous data, and generate a multi-dimensional construction progress representation model;

[0043] The progress deviation and risk prediction module is used to calculate the progress deviation and predict the risk based on the multi-dimensional construction progress representation model, and generate the current progress deviation value;

[0044] A control scheme generation module is used to generate a set of control schemes through a multi-objective optimization algorithm when the current progress deviation value exceeds a preset threshold;

[0045] The simulation verification module is used to simulate and verify the control scheme set in the digital twin environment and generate a verified feasible control scheme set.

[0046] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements any step of the above embodiment when executing the computer program.

[0047] In a fourth aspect, the present application further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, any step of the above embodiment is implemented.

[0048] The above-mentioned digital twin-based dynamic control method and system for subway construction progress obtains multi-source heterogeneous data from the construction site, constructs a digital twin and generates a multi-dimensional construction progress representation model. Based on this model, the progress deviation and prediction risk are calculated. When the deviation exceeds the threshold, a multi-objective optimization algorithm is used to generate a set of control schemes, and these schemes are verified in the digital twin environment. Finally, a feasible control scheme is generated for actual construction progress control. It can realize dynamic control of the construction progress, improve the ability to respond to changes in the construction site environment and emergencies, reduce construction delays and resource waste, thereby improving the level of subway construction management and construction quality, and effectively solve the problems of delayed response, insufficient prediction accuracy, isolated control measures and difficulty in reusing experience in existing technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0050] Figure 1 This is a flow chart of the method for dynamic control of subway construction progress based on digital twins of the present invention;

[0051] Figure 2 This is another flow chart of the method for dynamic control of subway construction progress based on digital twins of the present invention;

[0052] Figure 3 This is a structural block diagram of the subway construction progress dynamic control system based on digital twins of the present invention. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0054] The embodiments of the present application are applicable to the annual maintenance scenario of the subway. The subway maintenance team collects equipment operation data and construction environment information in real time through the intelligent monitoring equipment deployed at each site. The aforementioned data is transmitted to the cloud server, and a virtual construction model is generated through digital twin technology to accurately present the maintenance progress. Once a delay in progress or potential risks are found, the optimal adjustment plan of automatic planning can be obtained, such as reallocating manpower, adjusting the order of operations, etc., and the plan can be pushed to the mobile terminal of the on-site engineer. Engineers adjust the operations in a timely manner according to the plan to ensure that the maintenance is completed on time, ensure the safety of subway operations, and enhance the travel experience of passengers.

[0055] This is only an example and does not limit the specific application scenario.

[0056] In an exemplary embodiment, Figure 1 As shown, a method for dynamic control of subway construction progress based on digital twins is provided. This method is described using a control terminal as an example. It is understood that the method can also be applied to a server, or to a system including a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps S111 to S114.

[0057] in:

[0058] S111, obtain multi-source heterogeneous data of the construction site collected by the IoT perception layer, perform digital twin modeling processing based on the multi-source heterogeneous data, and generate a multi-dimensional construction progress representation model; the multi-source heterogeneous data includes historical construction data, real-time monitoring data and environmental parameters; the historical data includes BIM model data and construction organization design data.

[0059] Optionally, IoT devices, such as sensors and surveillance cameras, can continuously collect real-time data from the construction site. This data can also be accessed from stored historical construction data (e.g., BIM models and construction plans), along with environmental monitoring data, such as temperature and humidity. This heterogeneous data can be standardized to eliminate format differences. A virtual construction environment can then be constructed based on digital twin technology, mapping the physical state of the construction site to a digital model. Ultimately, a multi-dimensional representation of the construction rush, encompassing time, space, and resources, can be generated for subsequent analysis.

[0060] S112: Perform progress deviation calculation and risk prediction based on the multi-dimensional construction progress representation model to generate a current progress deviation value.

[0061] Specifically, schedule deviation calculation refers to comparing the difference between the actual construction progress and the planned progress, and usually uses methods such as the critical path method or earned value management to quantify the deviation; risk prediction is based on historical data and current status to evaluate the probability and impact of possible risk events (such as process delays and resource conflicts) in the future.

[0062] Compare real-time construction data with the planned progress in the digital twin model to calculate the deviation between the actual completion status of key nodes and the plan. Combined with a risk case library from historical data, statistical analysis or machine learning methods are used to predict the risks that may arise from current progress deviations (such as cross-construction conflicts and equipment failures). Risk warning indicators are generated, and the current progress deviation value and potential risk level are ultimately output for management's decision-making reference.

[0063] S113, when the current progress deviation value exceeds a preset threshold, a control plan set is generated through a multi-objective optimization algorithm; the control plan set includes resource allocation, process adjustment and emergency measures.

[0064] For example, a multi-objective optimization algorithm is a computational method that seeks the optimal solution among multiple conflicting objectives (such as shortest construction period, lowest cost, and highest resource utilization), and is often used to solve complex resource scheduling problems; a set of control schemes is a variety of adjustment strategies proposed for progress deviations, including reallocation of manpower / equipment, adjustment of process sequence, or initiation of emergency response.

[0065] When the schedule deviation exceeds a preset threshold, an optimization algorithm is automatically triggered to generate multiple feasible control plans with the goal of minimizing construction delays, resource waste, and cost overruns. Specifically, the algorithm comprehensively considers the current resource status (such as personnel and equipment availability), process dependencies (such as whether predecessor tasks have been completed), and environmental constraints (such as weather and nighttime construction restrictions). It calculates the effectiveness of different adjustment strategies and outputs a set of control plans for managers to choose from, including resource reallocation (such as adding workers and replacing equipment), process adjustments (such as parallel construction and delaying non-critical tasks), and contingency measures (such as suspending high-risk operations and initiating backup plans).

[0066] S114, simulate and verify the control scheme set in the digital twin environment to generate a verified feasible control scheme set; the feasible control scheme set is used to indicate the progress control operation in the actual construction process.

[0067] Optionally, a digital twin environment refers to a digital platform that performs real-time simulation of real scenes based on virtual models, which can simulate various dynamic changes in the construction process; simulation verification is to run the control plan and observe its impact on progress, resources, and risks to screen out the optimal solution.

[0068] Each generated control plan is simulated in a virtual environment using a digital twin model to evaluate its execution. For example, adjusting the process sequence is used to observe whether it will lead to new resource conflicts or extended construction time. Furthermore, increasing equipment investment is used to assess its combined impact on cost and schedule. Through multiple simulations, feasible solutions that can correct deviations without introducing new risks are identified. These solutions are then labeled with their expected effects (e.g., shortening construction time by X days, reducing risk by Y%). Ultimately, a set of optimized control plans is formed to guide on-site construction adjustments.

[0069] In the above-mentioned dynamic control method of subway construction progress based on digital twins, by acquiring multi-source heterogeneous data for modeling, calculating progress deviations, generating control plans and conducting simulation verification, the optimal control of construction progress is finally achieved. This can solve the problems of extensive progress tracking, delayed risk identification, lack of coordination of control measures, and difficulty in reusing experience in traditional management, and improve the real-time, accuracy and foresight of construction management.

[0070] In one embodiment, after performing digital twin modeling based on multi-source heterogeneous data from the construction site to generate a multi-dimensional construction progress representation model, the process further includes:

[0071] S211: Perform construction process network modeling based on BIM model data and construction organization design data to generate a directed acyclic graph of the construction process.

[0072] Preferably, the directed acyclic graph of construction processes is a graph structure that describes task dependencies, in which nodes represent construction processes and edges represent sequence constraints, ensuring that tasks are executed in a logical order without circular dependencies. The spatial relationship and logical dependencies of each construction unit are extracted from the BIM model (such as pipeline installation must be carried out after civil engineering is completed), and the initial task list is constructed in combination with the process division and time arrangement in the construction organization design. The sequence constraints between each process are analyzed (such as "equipment debugging must be completed after pipeline laying"), directed edges are established, and it is ensured that there are no circular dependencies in the graph (such as avoiding the situation of A→B→C→A). Finally, a directed acyclic graph of construction processes is generated, which intuitively displays the execution order and dependency of each task.

[0073] S212, defining a multi-dimensional feature vector for each process node of the directed acyclic graph of the construction process to generate a feature vector set; the feature vector set includes time parameters, engineering quantity parameters, resource demand parameters and environmental sensitivity parameters.

[0074] For example, a feature vector is a multidimensional data structure used to describe process attributes, where each dimension corresponds to a feature (such as time and resource requirements). Quantifying the above features can support intelligent analysis and decision-making.

[0075] A set of characteristic parameters is defined for each process node in the directed acyclic graph. The used parameters are combined into a characteristic vector and normalized to ensure comparability. Finally, a characteristic vector set is formed for subsequent progress status prediction and optimization calculation.

[0076] S213, training the progress state transfer function based on the feature vector set to generate a prediction model; the prediction model is used to predict the changing trend of the future construction progress; the changing trend is used to assist in optimizing the set of feasible control schemes.

[0077] Specifically, historical data from feature vector sets (such as progress records from similar projects) is used to train machine learning models (such as LSTM neural networks) to learn the patterns of progress state transitions. This model takes as input the characteristics of the current process (such as remaining work volume and resource availability) and outputs a progress forecast for a future point in time (such as the probability distribution of completion). Through continuous iterative optimization, a high-precision prediction model is generated, and combined with risk warning rules, potential delay points (such as process delays on the critical path) are identified. The prediction results are fed back into the control plan set, assisting managers in dynamically adjusting strategies to ensure construction is completed on schedule.

[0078] In the above-mentioned method of dynamic control of subway construction progress based on digital twins, the method of construction process network modeling and feature vector definition is carried out through BIM model and construction organization design data, which can accurately analyze the logical dependencies of multi-professional construction processes and quantify the key attributes of each process, providing a scientific decision-making basis for dynamic control of construction progress.

[0079] In an exemplary embodiment, training a progress state transition function based on a feature vector set to generate a prediction model includes:

[0080] S311, constructing training samples based on historical construction data to generate a training data set; the training data set includes feature vectors, environmental parameters and state changes.

[0081] For example, multi-source heterogeneous data is extracted from historical databases, including process parameters from BIM models, environmental monitoring data collected by IoT devices, and manually submitted progress reports. This data is cleaned and feature-engineered, converting each construction task into a fixed-dimensional feature vector and associating it with the corresponding timestamp and environmental parameters. Finally, this information is integrated in a time series to form a structured dataset containing feature vectors, environmental parameters, and state changes, ensuring temporal continuity and causal relationships.

[0082] S312, performing bidirectional LSTM neural network modeling processing on the training data set to generate an initial prediction model.

[0083] Optionally, time series data from the training dataset (e.g., daily records of project completion) is fed into a bidirectional LSTM network, with the feature vector serving as the input layer and the state change serving as the target value at the output layer. Hyperparameters such as the number of network layers and neurons are automatically set, and weights are adjusted using a backpropagation algorithm, enabling the model to learn how construction progress changes over time. Because the construction process is context-dependent (e.g., delays in previous steps can impact subsequent tasks), the bidirectional LSTM can simultaneously reference historical data and future trends (e.g., resource allocation plans) to generate an initial prediction model, enabling a preliminary dynamic simulation of construction progress.

[0084] S313, performing parameter optimization processing on the initial prediction model to generate a prediction model.

[0085] Based on the initial prediction model, cross-validation was used to evaluate its performance on a test dataset, and grid search was used to adjust hyperparameters (such as the number of LSTM layers and time step). Regularization techniques were also introduced to prevent overfitting and ensure model generalization. To address noise issues in construction data, a dropout layer or data smoothing can be automatically added. Through multiple iterative training cycles, an optimized prediction model was generated that more accurately reflects construction progress trends and provides early warning of delay risks caused by resource conflicts or environmental changes.

[0086] In the above-mentioned digital twin-based dynamic control method for subway construction progress, a training dataset is constructed through historical construction data and a bidirectional LSTM neural network is used for modeling and optimization. This enables accurate prediction of construction progress trends and early identification of cross-construction risks, thereby effectively improving the controllability and management efficiency of construction plans.

[0087] In one embodiment, a progress deviation calculation and risk prediction are performed based on a multi-dimensional construction progress representation model to generate a current progress deviation value, including:

[0088] S411, based on the planned progress data in the multi-dimensional construction progress representation model and the actual progress data collected in real time, a time-weighted deviation calculation is performed to generate a real-time deviation value for each process. The real-time deviation value is calculated using the following formula:

[0089]

[0090] Among them, Δ i (t) is the real-time deviation value of process i, i is the i-th process node in the construction network, t is the time point for calculating the deviation, T i is the planned completion time of the process, is the planned progress value of process i at time t, is the actual progress value of process i at time t, and α is the attenuation coefficient of the control weight.

[0091] Specifically, time-weighted deviation calculation is a method for dynamically evaluating the degree of deviation of construction progress, which uses an exponential decay function to calculate the time-weighted deviation. Different weights are assigned to deviations at different time points, so that the deviation of processes close to the planned completion time has a greater impact on the overall situation, thereby more accurately reflecting the deviation between the current construction status and the plan. Real-time acquisition of planned progress data in the multi-dimensional construction progress representation model (such as process time nodes in the BIM model) and actual progress data collected by IoT devices (such as the completion status of the project monitored by sensors). For each process node i, when calculating the time point t, first compare its planned progress value and actual progress value The absolute difference is then weighted by an exponential decay function.

[0092] In one embodiment, the power maintenance process A is planned to be completed on the 10th day (T i =10), current time t=8, planned progress Actual progress The real-time deviation value is Δ i (8) = e -0.1×(10-8) |80% - 60%| = 0.8187 × 20% = 16.37%. The larger the value, the more serious the progress lag.

[0093] S412: The real-time deviation values ​​of each process are aggregated to generate a current progress deviation value representing the overall construction status. The current progress deviation value is calculated using the following formula:

[0094]

[0095] Among them, Δ(t) is the current progress deviation value, n is the number of process i, ω i is the criticality weight of process i in the construction network.

[0096] Specifically, progress status aggregation is a comprehensive quantitative evaluation method for the real-time deviation value of each process, through the criticality weight ω i The deviations of different processes are weighted and summed up to make the process deviations on the critical path have a more significant impact on the overall progress evaluation, thereby accurately reflecting the overall health status of the construction network.

[0097] Collect the real-time deviation value Δ of all process nodes i (t), and analyze the criticality weight of each process ω according to the structure of the construction network i (For example, the importance of the process in the overall progress is determined by the critical path method, and the weight of the critical process ω i =1.0, non-critical process weight ω i=0.5). Multiply the real-time deviation value of each process by its corresponding weight, and add up the weighted deviation values ​​of all processes.

[0098] In one embodiment, the power maintenance project B includes three processes:

[0099] Process 1 (critical process, ω1 = 1.0, Δ1(t) = 16.37%),

[0100] Process 2 (non-critical process, ω2 = 0.5, Δ2(t) = 10%),

[0101] Process 3 (non-critical process, ω3=0.5, Δ3(t)=5%).

[0102] The current progress deviation is Δ(t) = 1.0 × 16.37% + 0.5 × 10% + 0.5 × 5% = 23.87%. A larger value indicates a higher risk of overall progress delay, providing managers with a basis for dynamic adjustment decisions.

[0103] In the above-mentioned digital twin-based dynamic control method for subway construction progress, time-weighted deviation calculation and dynamic aggregation analysis are used to accurately identify the risk of progress delays in large-scale construction, and real-time feedback on the overall construction status is provided through weighted evaluation of key processes.

[0104] In one embodiment, after aggregating the real-time deviation values ​​of each process to generate a current progress deviation value representing the overall construction status, the following steps are further included:

[0105] S511, based on the current progress deviation value and combined with the logical dependency relationship between processes, conduct risk propagation analysis and generate critical path impact factors.

[0106] Optionally, risk propagation analysis is to quantify the cascading impact of a single process schedule deviation on the critical path by analyzing the logical dependencies between processes; the critical path impact factor is a quantitative indicator that reflects the degree of impact of each process deviation on the overall construction period. The larger the value, the greater the threat to the critical path.

[0107] Based on the process dependency network of multi-disciplinary construction (such as the delay in the maintenance process of a certain equipment in the power industry), the propagation path of the deviation in the logic chain is automatically tracked, and its time compression effect on subsequent professional processes such as signals and tracks is calculated to generate an impact factor that characterizes the vulnerability of the critical path.

[0108] S512: Based on the critical path influencing factors and the current progress deviation value, a comprehensive risk assessment process is performed to generate a future risk probability distribution map; and a multi-objective optimization problem model is constructed based on the future risk probability distribution map.

[0109] Specifically, comprehensive risk assessment combines schedule deviation values ​​with critical path influencing factors to predict the spatial distribution and occurrence probability of future construction risks through a probabilistic statistical model; the multi-objective optimization problem model aims to minimize construction delays, minimize resource conflicts, and maximize equipment utilization, and construct a mathematical model that includes multiple optimization objectives.

[0110] The real-time deviation data of each discipline (such as a 30-minute delay in power maintenance and a 15-minute advance in signal debugging) and the critical path influencing factors are input into the risk assessment algorithm to generate a risk probability distribution map within the next 12 hours (such as the probability of cross-construction risk in a certain work area reaches 70%). Based on this map, a multi-objective optimization model is constructed with the goal of cross-disciplinary collaborative regulation, defining multi-dimensional optimization goals such as construction period, resources, and safety.

[0111] S513, based on the multi-source heterogeneous data in the multi-dimensional construction progress representation model, constraint definition processing is performed to generate a constraint condition set for multi-objective optimization.

[0112] For example, constraints are defined by extracting limiting factors influencing the feasibility of control plans from a multidimensional construction progress representation model, including resource capacity, process logic, and safety regulations. The constraint set for multi-objective optimization is a set of rules encompassing time constraints (e.g., the latest completion time for a single specialized process), resource constraints (e.g., the upper limit on the number of specialized workers), and spatial constraints (e.g., the minimum safe distance between multi-disciplinary work areas). By traversing the multi-source heterogeneous data in the multidimensional model (e.g., power equipment maintenance requires two high-voltage electricians, signal debugging requires three specialized test instruments), resource bottlenecks for each specialized construction project (e.g., only four high-voltage electricians are required for the entire station), process time windows (e.g., track operations must be completed between 2:00 AM and 4:00 AM), and safety regulations (e.g., the power and signal operation areas must be at least 5 meters apart), generating a set of 23 constraints.

[0113] S514, based on the constraint condition set, solving the multi-objective optimization problem model to generate a set of control schemes.

[0114] The multi-objective optimization solution uses intelligent algorithms (such as an improved particle swarm optimization algorithm) to search for the optimal solution within a set of constraints, balancing multiple mutually exclusive objectives (such as shortening construction period and reducing costs). Based on the set of constraints (e.g., only two high-voltage electricians are available and the track operation time window is fixed), an improved non-dominated sorting genetic algorithm (NSGA-II) is used to iteratively solve the multi-objective model, generating three feasible solutions: Solution 1 involves deploying a high-voltage electrician from a neighboring station to support power maintenance; Solution 2 adjusts signal debugging to the early morning of the next day; and Solution 3 consolidates some non-critical processes. Each solution is evaluated with indicators such as the reduction in construction period and the increase in resource consumption.

[0115] In the above-mentioned dynamic control method of subway construction progress based on digital twins, a multi-objective optimization model is constructed through risk propagation analysis and comprehensive risk assessment, and a control plan is generated in combination with constraint conditions. This realizes quantitative analysis of the progress deviation transmission path of multi-disciplinary construction and cross-disciplinary collaborative control, thereby improving the risk prediction accuracy and resource allocation efficiency in complex construction scenarios during the operation period.

[0116] In an exemplary embodiment, based on a set of constraints, a multi-objective optimization problem model is solved to generate a set of control solutions, including:

[0117] S611: Dynamically adjust the reference points based on the characteristic data of the current construction stage to generate a reference point distribution plan that adapts to the construction stage.

[0118] Optionally, by collecting stage characteristic data of multi-professional construction in real time (such as 60% completion of insulation testing of power professional equipment and entry of signal and track professional into the cross-operation stage), the reference points of multi-objective optimization are dynamically adjusted based on the fuzzy logic algorithm (such as increasing the resource balance target weight from 30% to 40%), and generating a reference point distribution plan that is suitable for the mid-term of nighttime maintenance, guiding the algorithm to prioritize solving cross-professional resource conflicts.

[0119] S612, based on the reference point distribution scheme adapted to the construction stage, select the optimization solution data matching the current construction stage from the historical case library, perform population initialization processing, and generate an initial optimization solution set.

[0120] Preferably, population initialization is performed at the initial stage of the optimization algorithm startup, selecting optimized solutions for similar working conditions from historical cases as the initial solution set to shorten the algorithm convergence time. By performing cosine similarity matching between the current construction phase characteristics (such as the 30th day of the annual power maintenance and humidity exceeding 75%) and 32 nighttime multi-disciplinary maintenance cases in the historical case library, 8 cases with a similarity greater than 85% are selected (such as the power and signal cross-maintenance case at a station in September 2024). Their resource allocation strategies (such as the cross-professional support ratio and the adjustment range of the process time window) are extracted as the initial optimization solution set to initialize the particle position parameters of the particle swarm algorithm.

[0121] S613 , based on the initial optimization solution set and in combination with the iterative process parameters of the optimization algorithm, an adaptive adjustment process of the crossover and mutation probability is performed to generate a dynamically adjusted control scheme set.

[0122] For example, the crossover probability is a key parameter in the optimization algorithm that controls the diversity of solutions and the speed of convergence. The crossover probability determines the frequency of solution combinations, while the mutation probability determines the degree of random perturbation of the solution. During the particle swarm algorithm iteration process, the fitness value of the initial set of optimized solutions (such as the reduction in construction delays and the number of resource conflicts) is monitored in real time. When no better solution appears in five consecutive generations, the mutation probability is automatically increased from 10% to 15%, and the crossover probability is reduced to 60%. This prompts the algorithm to explore new control paths such as the deployment of specialized power backup equipment and the segmented execution of signal processes, ultimately generating a set of solutions that include dynamic adjustment strategies (such as time-slot staggered operations and dynamic scheduling of cross-site resources).

[0123] In the above-mentioned digital twin-based dynamic control method for subway construction progress, by dynamically adjusting reference points, matching historical cases to initialize the optimization solution and adaptively adjusting algorithm parameters, precise adaptation of multi-professional characteristics of large-scale construction stages and intelligent generation of control plans are achieved, thereby improving the real-time performance and global optimization capabilities of progress control under complex working conditions.

[0124] In one embodiment, Figure 2 As shown, after the control scheme set is simulated and verified in the digital twin environment and a set of verified feasible control schemes is generated, the following is also included:

[0125] S711: Based on the actual construction data after the execution of the feasible control scheme set, the model parameter update amount is calculated to generate a parameter adjustment scheme for the digital twin model; the model parameter update amount is calculated using the following formula:

[0126]

[0127] Among them, Δθ is the model parameter update amount, is the composite loss function, η is the learning rate, μ is the momentum coefficient, Δθ p This is the historical update amount.

[0128] Specifically, actual data after the implementation of feasible control plans is collected (such as the actual time consumed in power maintenance is 20 minutes shorter than predicted, and the number of conflicts after the signal professional process is adjusted is 0), and the data is substituted into the formula to calculate the model parameter update amount. Weighted adjustments are made to the cross-influence factors of multi-professional construction, the time parameters of equipment periodic maintenance, etc., to generate a model parameter adjustment plan that is suitable for the current working conditions.

[0129] S712: Based on the parameter adjustment plan, the case library similarity matrix of the newly generated construction case data is updated to generate an updated case matching feature system.

[0130] Optionally, the case library similarity matrix quantifies the degree of match between historical construction cases and current working conditions. A high-dimensional vector space is constructed based on dimensions such as process logic, resource allocation, and environmental parameters, with the matrix values ​​reflecting the cosine similarity between cases. The case matching feature system is a set of key features used to retrieve historical cases, including professional type, construction phase, and environmental indicators. Based on the parameter-adjusted digital twin model, key features (single-professional progress deviation rate, cross-professional conflict type) are extracted for newly generated construction cases (such as the nighttime maintenance of four professional units at a station in May 2025). The case library similarity matrix is ​​updated (for example, the weight of the "ageing degree of power equipment" feature is increased from 15% to 20%), generating a case matching feature system that is more suitable for multi-professional collaboration scenarios during the operation period.

[0131] S713, based on the updated case matching feature system and combined with the actual progress deviation observation data, the environmental response coefficient is adjusted to generate an optimized environmental parameter response plan; the environmental parameter response plan is used to indicate the adjustment of the constraint condition set of the multi-objective optimization.

[0132] For example, the environmental response coefficient measures the impact of environmental parameters on construction progress. A larger value indicates a greater contribution of environmental changes to process delays. Constraint set adjustment dynamically revises the constraints of multi-objective optimization based on the environmental response coefficient (e.g., when humidity is >80%, the power industry is required to increase the frequency of insulation testing). By combining actual progress deviation observation data (e.g., humidity reaching 85% one night, causing a one-hour delay in power cable laying), the environmental response coefficient is adjusted (increasing the humidity response coefficient from 0.3 to 0.5), and the constraint set of multi-objective optimization is optimized accordingly (e.g., when humidity exceeds the standard, "power workers must wear moisture-proof equipment" is added to the resource constraint), generating a response plan that adapts to real-time environmental changes.

[0133] In the above-mentioned digital twin-based dynamic control method for subway construction progress, through iterative updating of model parameters, optimization of case library features and adjustment of environmental response coefficients, the digital twin model can dynamically adapt to multi-professional construction scenarios during the operation period and continuously reuse historical experience, thereby improving the accuracy of the progress control strategy and the system's self-evolution capability in complex environments.

[0134] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0135] Based on the same inventive concept, the embodiments of the present application also provide a system for implementing the aforementioned method for dynamically controlling subway construction progress based on digital twins. The implementation solution provided by this system is similar to the implementation solution described in the aforementioned method. Therefore, the specific limitations of one or more embodiments of the system for dynamically controlling subway construction progress based on digital twins provided below can be found in the limitations of the method for dynamically controlling subway construction progress based on digital twins above, and will not be repeated here.

[0136] In an exemplary embodiment, Figure 3 As shown, a subway construction progress dynamic control system 10 based on digital twin is provided, including:

[0137] The digital twin modeling module 11 is used to obtain multi-source heterogeneous data of the construction site collected by the IoT perception layer, perform digital twin modeling processing based on the multi-source heterogeneous data, and generate a multi-dimensional construction progress representation model;

[0138] The progress deviation and risk prediction module 12 is used to calculate the progress deviation and predict the risk based on the multi-dimensional construction progress representation model, and generate the current progress deviation value;

[0139] The control scheme generating module 13 is used to generate a control scheme set through a multi-objective optimization algorithm when the current progress deviation value exceeds a preset threshold;

[0140] The simulation verification module 14 is used to simulate and verify the control scheme set in the digital twin environment to generate a verified feasible control scheme set.

[0141] In one embodiment, the digital twin modeling module is further configured to:

[0142] Conduct construction process network modeling based on BIM model data and construction organization design data to generate a directed acyclic graph of the construction process;

[0143] A multi-dimensional feature vector is defined for each process node of the directed acyclic graph of the construction process to generate a feature vector set; the feature vector set includes time parameters, engineering quantity parameters, resource demand parameters and environmental sensitivity parameters;

[0144] The progress state transfer function is trained based on the feature vector set to generate a prediction model; the prediction model is used to predict the changing trend of the future construction progress; and the changing trend is used to assist in optimizing the set of feasible control schemes.

[0145] In one embodiment, the digital twin modeling module is further configured to:

[0146] Construct training samples based on historical construction data to generate a training data set; the training data set includes feature vectors, environmental parameters, and state changes;

[0147] Perform bidirectional LSTM neural network modeling on the training data set to generate an initial prediction model;

[0148] The parameters of the initial prediction model are optimized to generate a prediction model.

[0149] In one embodiment, the schedule deviation and risk prediction module is further configured to:

[0150] Based on the planned progress data in the multi-dimensional construction progress representation model and the actual progress data collected in real time, the real-time deviation value of each process is generated through time-weighted deviation calculation. The real-time deviation value is calculated using the following formula:

[0151]

[0152] Among them, Δ i (t) is the real-time deviation value of process i, i is the i-th process node in the construction network, t is the time point for calculating the deviation, T i is the planned completion time of the process, is the planned progress value of process i at time t, is the actual progress value of process i at time t, α is the attenuation coefficient of the control weight;

[0153] The real-time deviation values ​​of each process are aggregated to generate the current progress deviation value that represents the overall construction status. The current progress deviation value is calculated using the following formula:

[0154]

[0155] Among them, Δ(t) is the current progress deviation value, n is the number of process i, ω i is the criticality weight of process i in the construction network.

[0156] In one embodiment, the schedule deviation and risk prediction module is further configured to:

[0157] Based on the current progress deviation value and the logical dependency relationship between processes, risk propagation analysis is performed to generate critical path impact factors;

[0158] Based on the critical path influencing factors and the current progress deviation value, a comprehensive risk assessment is conducted to generate a future risk probability distribution map; based on the future risk probability distribution map, a multi-objective optimization problem model is constructed;

[0159] Based on the multi-source heterogeneous data in the multi-dimensional construction progress representation model, the constraint conditions are defined and processed to generate a constraint condition set for multi-objective optimization;

[0160] Based on the set of constraints, the multi-objective optimization problem model is solved to generate a set of control schemes.

[0161] In one embodiment, the schedule deviation and risk prediction module is further configured to:

[0162] Dynamically adjust reference points based on the characteristic data of the current construction stage to generate a reference point distribution plan that adapts to the construction stage;

[0163] Based on the reference point distribution scheme adapted to the construction stage, the optimization solution data matching the current construction stage is selected from the historical case library, and the population is initialized to generate the initial optimization solution set;

[0164] Based on the initial optimization solution set and combined with the iterative process parameters of the optimization algorithm, the crossover and mutation probability is adaptively adjusted to generate a set of dynamically adjusted control schemes.

[0165] In one embodiment, the simulation verification module is further configured to:

[0166] Based on the actual construction data after the execution of the feasible control scheme set, the model parameter update amount is calculated to generate the parameter adjustment plan of the digital twin model; the model parameter update amount is calculated using the following formula:

[0167]

[0168] Among them, Δθ is the model parameter update amount, is the composite loss function, η is the learning rate, μ is the momentum coefficient, Δθ p is the historical update amount;

[0169] Based on the parameter adjustment scheme, the case database similarity matrix of the newly generated construction case data is updated to generate an updated case matching feature system;

[0170] Based on the updated case matching feature system and combined with the actual progress deviation observation data, the environmental response coefficient is adjusted to generate an optimized environmental parameter response scheme; the environmental parameter response scheme is used to indicate the adjustment of the constraint condition set of the multi-objective optimization.

[0171] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method for dynamic control of subway construction progress based on digital twins as described above are implemented.

[0172] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0173] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0174] The above-described embodiments merely represent several implementation methods of the embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the concept of the embodiments of the present application, and these modifications and improvements fall within the scope of protection of the embodiments of the present application.

Claims

1. A method for dynamic control of subway construction progress based on digital twins, characterized in that: The method comprises: Obtain multi-source heterogeneous data from the construction site collected by the IoT perception layer, perform digital twin modeling processing based on the multi-source heterogeneous data, and generate a multi-dimensional construction progress representation model; the multi-source heterogeneous data includes historical construction data, real-time monitoring data, and environmental parameters; the historical data includes BIM model data and construction organization design data; Perform progress deviation calculation and risk prediction based on the multi-dimensional construction progress representation model to generate a current progress deviation value; When the current progress deviation value exceeds a preset threshold, a control scheme set is generated through a multi-objective optimization algorithm; the control scheme set includes resource allocation, process adjustment and emergency measures; The control scheme set is simulated and verified in a digital twin environment to generate a verified feasible control scheme set; the feasible control scheme set is used to indicate the progress control operation in the actual construction process.

2. The method according to claim 1, characterized in that After performing digital twin modeling processing based on the multi-source heterogeneous data of the construction site to generate a multi-dimensional construction progress representation model, the method further includes: Performing construction process network modeling based on the BIM model data and the construction organization design data to generate a directed acyclic graph of the construction process; Defining a multi-dimensional feature vector for each process node of the directed acyclic graph of the construction process to generate a feature vector set; the feature vector set includes a time parameter, a project quantity parameter, a resource demand parameter, and an environmental sensitivity parameter; The progress state transfer function is trained based on the feature vector set to generate a prediction model; the prediction model is used to predict the changing trend of the future construction progress; the changing trend is used to assist in optimizing the set of feasible control schemes.

3. The method according to claim 2, characterized in that The training of the progress state transfer function based on the feature vector set to generate a prediction model includes: Constructing training samples based on historical construction data to generate a training data set; the training data set includes feature vectors, environmental parameters, and state changes; Performing bidirectional LSTM neural network modeling on the training data set to generate an initial prediction model; Parameter optimization processing is performed on the initial prediction model to generate a prediction model.

4. The method according to claim 1, wherein The performing of progress deviation calculation and risk prediction based on the multi-dimensional construction progress representation model to generate a current progress deviation value includes: Based on the planned progress data in the multi-dimensional construction progress representation model and the actual progress data collected in real time, a real-time deviation value of each process is generated through time-weighted deviation calculation; the real-time deviation value is calculated using the following formula: Among them, Δ i (t) is the real-time deviation value of process i, i is the i-th process node in the construction network, t is the time point for calculating the deviation, T i is the planned completion time of the process, is the planned progress value of process i at time t, is the actual progress value of process i at time t, α is the attenuation coefficient of the control weight; The real-time deviation values ​​of each process are aggregated to generate a current progress deviation value representing the overall construction status; the current progress deviation value is calculated using the following formula: Among them, Δ(t) is the current progress deviation value, n is the number of process i, ω i is the criticality weight of process i in the construction network.

5. The method according to claim 4, characterized in that After aggregating the real-time deviation values ​​of each process to generate a current progress deviation value representing the overall construction status, the method further includes: Based on the current progress deviation value and the logical dependency relationship between processes, a risk propagation analysis is performed to generate a critical path impact factor; Based on the critical path influencing factors and the current progress deviation value, a comprehensive risk assessment process is performed to generate a future risk probability distribution map; and a multi-objective optimization problem model is constructed based on the future risk probability distribution map; Based on the multi-source heterogeneous data in the multi-dimensional construction progress representation model, a constraint condition definition process is performed to generate a constraint condition set for multi-objective optimization; Based on the set of constraints, the multi-objective optimization problem model is solved to generate a set of control schemes.

6. The method according to claim 5, characterized in that Solving the multi-objective optimization problem model based on the constraint condition set to generate a set of control schemes includes: Dynamically adjust reference points based on the characteristic data of the current construction stage to generate a reference point distribution plan that adapts to the construction stage; Based on the reference point distribution scheme adapted to the construction stage, the optimization solution data matching the current construction stage is selected from the historical case library, and a population initialization process is performed to generate an initial optimization solution set; Based on the initial optimization solution set and in combination with the iterative process parameters of the optimization algorithm, a crossover and mutation probability adaptive adjustment process is performed to generate a dynamically adjusted control scheme set.

7. The method according to claim 1, characterized in that After simulating and verifying the control scheme set in the digital twin environment to generate a verified feasible control scheme set, the method further includes: Based on the actual construction data after the execution of the set of feasible control schemes, the model parameter update amount is calculated to generate a parameter adjustment scheme for the digital twin model; the model parameter update amount is calculated using the following formula: Among them, Δθ is the model parameter update amount, is the composite loss function, η is the learning rate, μ is the momentum coefficient, Δθ p is the historical update amount; Based on the parameter adjustment scheme, the case library similarity matrix of the newly generated construction case data is updated to generate an updated case matching feature system; Based on the updated case matching feature system, the environmental response coefficient is adjusted in combination with the actual progress deviation observation data to generate an optimized environmental parameter response scheme; the environmental parameter response scheme is used to indicate the adjustment of the constraint condition set of the multi-objective optimization.

8. A subway construction progress dynamic control system based on digital twin, characterized by: The system comprises: A digital twin modeling module is used to obtain multi-source heterogeneous data of the construction site collected by the IoT perception layer, perform digital twin modeling processing based on the multi-source heterogeneous data, and generate a multi-dimensional construction progress representation model; A progress deviation and risk prediction module is used to calculate progress deviation and predict risks based on the multi-dimensional construction progress representation model to generate a current progress deviation value; A control scheme generating module is used to generate a control scheme set through a multi-objective optimization algorithm when the current progress deviation value exceeds a preset threshold; The simulation verification module is used to simulate and verify the control scheme set in the digital twin environment to generate a verified feasible control scheme set.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images