Intelligent construction site construction progress and material consumption collaborative management platform based on digital twinning

Abstract

The invention relates to the technical field of intelligent construction and digital twinning, and discloses an intelligent construction site construction progress and material consumption collaborative management platform based on digital twinning, comprising a process state evolution graph construction module used for constructing a process state evolution graph containing a probabilistic duration function; the physical digital state synchronization module is used for verifying and updating the current state of the atlas through a trusted event account book; the rolling time domain collaborative controller module is used for generating collaborative construction and material instructions by solving a multi-objective collaborative optimization problem aiming at minimizing a comprehensive cost function based on a current state; and the model self-adaptive correction module is used for carrying out closed-loop self-adaptive correction on the atlas model by comparing actual and prediction results. According to the method, the uncertainty of a construction site can be effectively dealt with, the deep collaboration of the construction progress and the material consumption is realized, and the accuracy of prediction and decision making is continuously improved through a self-learning mechanism.

Description

Technical Field

[0001] This invention relates to the fields of smart construction and digital twin technology, specifically a smart construction site collaborative management platform for construction progress and material consumption based on digital twins. Background Technology

[0002] One of the core tasks of construction project management is to ensure the coordination and unity of construction progress and material supply. Driven by information technology, building information modeling technology has been applied in the engineering design and planning stages, providing a digital foundation for project management.

[0003] However, during the project execution phase, construction schedule management and material consumption management are often conducted as two separate processes. Typically, project managers first develop a construction schedule plan based on a deterministic timeframe, and then material procurement and scheduling plans are passively generated based on this schedule plan. This open-loop, unidirectional management model is inherently fragile in terms of coordination.

[0004] Construction sites are dynamic and unpredictable environments. Factors such as weather changes, equipment failures, fluctuations in personnel efficiency, and the unexpected discovery of concealed works can all cause actual construction progress to deviate from the initial plan. When deviations occur, traditional management methods rely on manual adjustments, which are often delayed. Furthermore, due to information delays and the locality of decision-making, it is difficult to synchronously and optimally correct material supply plans. As a result, materials may arrive prematurely due to project delays, increasing warehousing costs and on-site management difficulties; or, due to expedited project deadlines, supplies may be insufficient, causing work stoppages and directly impacting the overall project duration.

[0005] In recent years, digital twin technology has been introduced into the field of smart construction sites, aiming to enhance on-site visualization and monitoring capabilities by establishing a real-time digital mapping of the physical world. Although existing digital twin applications can reflect the real-time status of construction sites to some extent, most remain at the level of passive information presentation and status tracing. These systems generally lack an inherent mechanism to make forward-looking predictions based on real-time site conditions and inherent uncertainties, and to generate control commands with global coordination. Therefore, how to deeply couple the two links of construction progress and material consumption, and establish a management system that can dynamically adapt to on-site uncertainties and achieve closed-loop collaborative control, is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a smart construction site construction progress and material consumption collaborative management platform based on digital twins. This solves the problem that existing smart construction site management systems typically use deterministic planning or open-loop control when dealing with the collaborative relationship between construction progress and material consumption. This makes it difficult to effectively cope with the inherent uncertainties of the construction site, leading to a disconnect between the plan and the actual situation, and consequently causing delays and material waste.

[0007] The first aspect of this invention provides a smart construction site construction progress and material consumption collaborative management platform based on digital twins, the platform comprising:

[0008] Process state evolution map construction module;

[0009] Physical digital status synchronization module;

[0010] Scrolling time-domain collaborative controller module;

[0011] Model adaptive correction module.

[0012] The process state evolution graph construction module constructs a process state evolution graph based on static design information such as Building Information Modeling (BIM) and Work Breakdown Structure (WBS). This graph is a directed acyclic graph (DAG), where state nodes represent discrete, observable completion states of construction processes. State transitions connecting these nodes represent the specific construction activities required to move from a preceding node to a subsequent node. In a specific implementation, each state transition is associated with a probabilistic duration function and a resource consumption function. The probabilistic duration function is, for example, a normal distribution function. The function is used to quantify the uncertainty of the duration of the construction activity; the resource consumption function defines the materials, labor, and machine shifts required to perform the activity.

[0013] The physical-digital state synchronization module acquires heterogeneous data from multiple sources, such as IoT sensors, mobile terminal reports, and drone inspections, from the physical construction site and transforms this data into reliable state updates for the digital twin model. This module verifies and records the achievement events of state nodes through a trusted event ledger mechanism. In a specific implementation, the module includes a verification rule engine. This engine processes multi-source data pointing to the same state achievement event based on preset cross-validation rules or logical consistency rules. Only verified data can be structured into a trusted event and recorded in the trusted event ledger, thereby triggering the update of the corresponding state node in the process state evolution graph.

[0014] The rolling time-domain collaborative controller module generates a forward-looking collaborative instruction set at each decision cycle based on the current state of the process state evolution graph and model parameters. This is achieved by solving a multi-objective collaborative optimization problem, the objective of which is to minimize the overall cost function. In a specific implementation, this comprehensive cost function consists of the project delay cost. and material supply and demand mismatch costs It consists of two parts, and its mathematical form can be expressed as:

[0015] ;

[0016] in, Represents the statistical expectation under all uncertain future scenarios; cost of material supply and demand mismatch. It can be further broken down into material holding costs caused by supply exceeding consumption, and material shortage costs caused by consumption demand preceding supply.

[0017] To solve the aforementioned optimization problem, the rolling time-domain cooperative controller module first generates multiple possible future evolution scenarios based on the probabilistic historical functions associated with each state transition by executing stochastic system simulation methods such as Monte Carlo. Subsequently, under the statistical expectations of these multiple possible future evolution scenarios, a heuristic optimization algorithm is used to solve the aforementioned comprehensive cost function, obtaining the optimal control sequence covering the future prediction time domain. Finally, based on the principle of rolling time-domain control, only the first decision step of the sequence is extracted. The system then parses and generates a collaborative instruction set. This collaborative instruction set includes construction task instructions and material scheduling instructions. The material scheduling instructions are derived synchronously from the state transitions and associated resource consumption functions contained in the construction task instructions, thus ensuring the inherent consistency between the two.

[0018] The model adaptive correction module continuously compares the actual execution results and predicted results of state transitions to achieve closed-loop feedback and adaptive correction of the process state evolution graph. This module first calculates the prediction error between the actual duration of a state transition and the expected value of its probabilistic duration function. In a specific implementation, when a systematic bias in the prediction error is diagnosed, the module uses Kalman filtering or other recursive update algorithms to update the expected value and variance parameters of the probabilistic duration function associated with the state transition based on the prediction error. Furthermore, the module monitors whether the prediction error of a specific state transition remains above a preset threshold after parameter updates. If so, it performs correlation analysis of the error patterns to identify potential logical dependency errors or missing resource constraints in the process state evolution graph, and generates interpretable model structure optimization suggestions for managers.

[0019] A second aspect of this invention provides a method for collaborative management of construction progress and material consumption at smart construction sites based on digital twins, the method comprising the following steps:

[0020] Constructing a process state evolution graph: Based on the static design information of the project, construct a process state evolution graph that includes state nodes and state transitions.

[0021] Synchronize physical digital state: Obtain multi-source data from the physical construction site, verify and record the achievement events of state nodes through a trusted event ledger, and update the current state of the process state evolution graph.

[0022] Generate cooperative instructions: Based on the current state of the process state evolution graph, generate a set of cooperative instructions by solving a multi-objective cooperative optimization problem aimed at minimizing the comprehensive cost function.

[0023] Adaptive correction model: By comparing the actual execution results of state transitions with the predicted results, the model parameters or structure of the process state evolution map are adaptively corrected.

[0024] This invention provides a smart construction site collaborative management platform for construction progress and material consumption based on digital twins. It has the following beneficial effects:

[0025] 1. This invention enhances the adaptability of construction management to on-site uncertainties and achieves deep coordination between construction progress and material consumption. Through the rolling time-domain collaborative controller module, the platform utilizes the probabilistic historical function embedded in the process state evolution graph to proactively assess the impact of uncertainties on the future through stochastic system simulation. On this basis, by solving the multi-objective collaborative optimization problem that simultaneously optimizes the cost of construction delays and the cost of material supply and demand mismatch, the platform can generate construction tasks and material scheduling instructions that maintain consistency from the source. This changes the situation in traditional planning where progress and materials are separated and passively adjusted, making management decisions more robust and forward-looking.

[0026] 2. This invention improves the accuracy and reliability of the digital twin model's state, providing a high-quality data foundation for subsequent optimization decisions. Through the physical-digital state synchronization module, the platform introduces a verification rule engine and a trusted event ledger mechanism. Before updating the state of the digital twin model, this mechanism performs necessary cross-validation and logical consistency checks on multi-source heterogeneous data from the physical site, effectively filtering out erroneous or contradictory data. This ensures that only verified state achievement events are used to drive model evolution, thereby significantly improving the fidelity of the digital twin's true mapping to the physical site.

[0027] 3. This invention endows the management platform with self-learning and self-evolution capabilities, enabling its prediction accuracy to continuously improve as the project progresses. Through the model adaptive correction module, the platform establishes a closed-loop feedback path from prediction error to model correction. This model adaptive correction module can not only automatically perform recursive updates on the time function parameters in the process state evolution graph based on the prediction error to adapt to the dynamic changes in on-site operation efficiency, but also further identify potential structural defects such as logical dependency errors in the model through correlation analysis of error patterns. This two-layer correction mechanism enables the model to continuously learn from actual execution data and achieve continuous self-improvement. Attached Figure Description

[0028] Figure 1 This is a system architecture diagram of the present invention;

[0029] Figure 2 This is a flowchart of the method of the present invention;

[0030] Figure 3 This is a schematic diagram of the internal working process of the physical digital state synchronization module of the present invention;

[0031] Figure 4 This is a schematic diagram of the internal working process of the rolling time-domain collaborative controller module of the present invention.

[0032] Among them, 10 is the process state evolution map construction module; 20 is the physical digital state synchronization module; 30 is the rolling time domain collaborative controller module; and 40 is the model adaptive correction module. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] Example:

[0035] Please see the appendix Figure 1 -Appendix Figure 4 This invention provides a smart construction site construction progress and material consumption collaborative management platform based on digital twins, comprising:

[0036] The system includes a process state evolution map construction module 10, a physical digital state synchronization module 20, a rolling time-domain collaborative controller module 30, and a model adaptive correction module 40.

[0037] Among them, the process state evolution map construction module 10 is used to establish a basic mathematical model describing the inherent logic and uncertainty of the construction process, and the physical digital state synchronization module 20 serves as a bridge connecting the digital model and the physical construction site, responsible for providing highly reliable real-time on-site state data.

[0038] The rolling time-domain collaborative controller module 30 is the decision-making core of the system. It receives the current status information provided by the physical-digital state synchronization module 20, performs forward-looking optimization calculations based on the model established by the process state evolution map construction module 10, and outputs collaborative construction and material instructions.

[0039] The model adaptive correction module 40 forms a feedback correction closed loop. It continuously corrects the model parameters in the process state evolution map construction module 10 by comparing the prediction results of the rolling time domain collaborative controller module 30 with the actual results collected by the physical digital state synchronization module 20, so that the decision basis of the entire system can dynamically adapt to the actual changes on site.

[0040] See attached document Figure 2 A smart construction site collaborative management method for construction progress and material consumption based on digital twins includes the following steps:

[0041] S101. Constructing a process state evolution map: Based on the project's building information model, construction organization design, and historical project data, the construction process is decomposed into a set of discrete and physically verifiable state nodes, and the transition relationships between state nodes are established; for each state transition relationship, its resource consumption function is set, and a probabilistic duration function containing the expected duration and duration variance is initialized, thereby forming an initial process state evolution map.

[0042] S102. Synchronize physical and digital states: By deploying data acquisition equipment at the construction site, real-time on-site data related to construction progress and material consumption is obtained; after verifying the collected data, it is recorded as a credible event in the event ledger, and the current state node in the process state evolution graph is updated according to the event, so as to realize the synchronization of the state of the digital model and the physical entity.

[0043] S103. Generate collaborative control instructions: Starting from the updated current state node, perform random path simulation based on the probabilistic duration function of each state transition within a preset future prediction time domain; determine an optimal sequence of collaborative control instructions by solving a multi-objective optimization problem consisting of project delay cost and material mismatch cost.

[0044] S104. Issue and execute the first step instruction: Extract the instruction set of the first time step in the aforementioned optimal collaborative control instruction sequence, and issue the construction task instructions and material scheduling instructions contained therein to the corresponding field execution systems through the interface.

[0045] S105. Correct model parameters: Continuously compare the actual duration of state transitions recorded in the event ledger with the predicted duration generated in step S103, and calculate the prediction error; when the statistical result of the prediction error meets the preset systematic deviation condition, update the parameters of the probabilistic duration function of the corresponding state transition in the process state evolution diagram according to the error. Steps S102 to S105 are executed cyclically to form a continuously operating closed-loop control system.

[0046] The specific implementation of the process state evolution graph construction module 10 lies in its transformation of the project's static design information into a dynamic graph model that can be used for subsequent optimization calculations. The core of this process lies in the digital definition and instantiation of state nodes, specifically including the following steps:

[0047] The module for parsing design information and identifying construction components receives and processes the project's building information model, computer-aided design drawings, and work breakdown structure files. For methods of parsing the geometric and non-geometric information of components from the building information model, those skilled in the art can use existing standard interfaces or software development kits. The specific parsing process is well-known in the field and will not be elaborated here. The purpose of this step is to atomize the engineering entity and identify each independent construction component.

[0048] This invention decomposes the construction components into process states. Traditional construction plans typically use activities or tasks as basic units, while this invention further decomposes tasks into a series of physically observable and verifiable states. For example, for a specific reinforced concrete column component, its construction process can be decomposed into multiple successive state nodes, including: completion of column bottom reinforcement pre-embedding, completion of column reinforcement binding, completion of column formwork installation, completion of concrete pouring, completion of concrete curing to the specified strength, and completion of column formwork removal. This decomposition method transforms the management object from process-oriented activities to result-oriented states, providing a foundation for accurate state synchronization through subsequent sensor or manual verification.

[0049] Define the data structure for each state node, obtained through the above steps. All data objects are instantiated as structured data objects in the system. The specific composition of these data objects, i.e., their data structure, is defined as follows:

[0050] ;

[0051] The meanings of each field are as follows:

[0052] : A globally unique identifier for a state node.

[0053] The identifier of the construction component to which this status node belongs, used to associate and identify the component.

[0054] The spatial location information of this state node is usually inherited from the three-dimensional coordinates of its component.

[0055] : A textual description of the state.

[0056] : Quality acceptance flag. This is an enumeration type whose values ​​can be not accepted, under acceptance, accepted, unaccepted, etc. It is used to record quality control information related to this status.

[0057] The set of preceding state nodes, recording all nodes that must precede the current state. Other state nodes that have been achieved.

[0058] The set of subsequent state nodes, recording all nodes in the current state. A state node can only begin after it has been achieved.

[0059] The process state evolution graph construction module 10 batch instantiates and generates node sets, traverses all identified construction components, decomposes the process state rules for each component, and, based on the data structure, batch creates and populates data objects for each state node in the database or memory. By performing the above state decomposition and instantiation process on all components in the project, the process state evolution graph is finally generated. Complete set of state nodes .

[0060] Generate a set of state nodes Subsequently, the process state evolution graph construction module 10 further establishes the connection relationships between nodes and models these connection relationships, i.e. state transitions, in a probabilistic and resource-related manner. This modeling process is the key difference from the traditional deterministic planning model, as it intrinsically integrates uncertainty and resource constraints into the process model.

[0061] Establish state transition connections, and the process state evolution graph construction module 10 constructs state nodes based on the state nodes established in step S213. The defined set of preceding state nodes and the set of subsequent state nodes are connected by directed connections between corresponding nodes to form a state transition. This directed connection clarifies the technological logic and sequential constraints of the construction process.

[0062] Associating state transitions with resource consumption functions, for each state transition... This module establishes structured resource consumption functions for it. This function is a quantitative description of all the resources required to complete this state transition, and its specific form is:

[0063] ;

[0064] The bill of materials details the specific types, specifications, and quantities of materials required to complete the state transition. The data in this bill of materials is derived from the project's overall bill of materials and corresponds precisely to the components and construction techniques associated with the transition.

[0065] For the human resource requirements, the required job types, skill levels, and estimated working hours were defined.

[0066] For the mechanical equipment requirements, the specific models and number of shifts of the required equipment were defined.

[0067] To establish probabilistic time-series functions for state transitions, and to quantify and handle inherent uncertainties in the construction process, such as time deviations caused by weather changes, fluctuations in personnel efficiency, or delays in material supply, this invention provides a method for each state transition. Establish a probabilistic time function Instead of a fixed, deterministic duration, in specific embodiments, this probabilistic duration function can be modeled using a normal distribution:

[0068] ;

[0069] in: To complete this state transition The expected average duration is an estimate based on standard operating conditions.

[0070] The variance of the state transition time is used to quantify the degree of time uncertainty in this process. A larger variance value indicates that the actual time of the process is more likely to deviate from its expected value.

[0071] It should be understood that the normal distribution is only an exemplary implementation. In other embodiments, the probabilistic duration function can also select other probabilistic distribution models according to the duration distribution characteristics of different types of tasks, such as the beta distribution or triangular distribution that can better describe the project activity time. The probabilistic duration function here is a generalization of all mathematical models that can describe time uncertainty.

[0072] Initialize model parameters, the above resource consumption function With probabilistic time function The initial parameters can be obtained from the project's historical data warehouse through statistical analysis, or set according to industry standard construction quotas, or obtained through expert evaluation methods when data is lacking. These initial parameters will be dynamically updated by the model adaptive correction module 40 during subsequent system operation.

[0073] By generating a set of state transitions and performing the above steps on the logical relationships between all state nodes, the process state evolution graph construction module 10 finally generates the graph. Complete set of state transitions Thus, the complete process state evolution map, which includes state, logical relationships, resource constraints, and uncertainty quantification, has been constructed.

[0074] The primary responsibility of the physical digital status synchronization module 20 is to acquire various data representing the current status of the physical construction site, and to process and integrate these data from different sources and in different formats, providing standardized data input for subsequent status verification and updates. This process specifically includes the following steps:

[0075] The physical digital status synchronization module 20 configures and manages various data acquisition interfaces to receive data from various data acquisition devices deployed at the construction site. These devices may include, but are not limited to:

[0076] Radio frequency identification (RFID) readers used to identify materials and components;

[0077] Fixed or mobile video surveillance equipment used to monitor construction areas and progress;

[0078] Drones equipped with high-precision cameras;

[0079] Global positioning system modules and sensors installed on large machinery to monitor its operating conditions and location;

[0080] And mobile smart terminals with specific applications installed for use by on-site managers, workers and quality inspectors. The specific implementation of data transmission through standard communication protocols for the above devices is well known in the field and will not be described in detail here.

[0081] The module preprocesses and standardizes the raw collected data. Since data from different sources differ in format, timestamps, units, etc., this module first preprocesses the received raw data. This preprocessing includes:

[0082] Data cleaning is used to filter out noise or obviously abnormal data points generated during transmission.

[0083] Timestamp synchronization calibrates the time information of all data to Coordinated Universal Time (UTC) to establish a unified time base.

[0084] In addition, it standardizes the format, parses and converts data of different structures into a predefined standard data structure within the system. This standard data structure may include: data source ID, collection timestamp, data type, and data payload.

[0085] Performing the spatiotemporal association of data with the process model is the core of data fusion. It aims to establish a connection between standardized data and specific elements in the process state evolution map. This association process includes:

[0086] Spatial Association: For data containing geographic location information, the system associates its spatial coordinates with the state nodes in the process state evolution map. Defined spatial location attributes By comparing the data and using a spatial nearest neighbor algorithm, the data is associated with the construction components or areas corresponding to its location, and then with the relevant status nodes.

[0087] Semantic Association: For data uploaded via mobile terminals that contains explicit semantic information, the system uses natural language processing or preset business logic rules to directly associate it with the corresponding state descriptions in the data and process state evolution graph. The state node.

[0088] After generating a fused data packet and performing spatiotemporal correlation, the original data payload is appended with its contextual information in the data and process state evolution graph model. The physical-digital state synchronization module 20 encapsulates the preprocessed and spatiotemporally correlated data into a fused data packet. This data packet not only contains the standardized original data but also explicitly includes the identifiers of one or more associated state nodes. This fused data packet will serve as input for status event verification in subsequent steps.

[0089] Upon receiving the fusion data packet generated in the aforementioned steps, the physical-digital state synchronization module 20 performs state verification and update. This process aims to ensure that only verified and traceable data can change the state of the digital twin model, thereby providing a highly reliable data foundation for subsequent decision-making and learning.

[0090] The physical-digital state synchronization module 20 verifies the execution status achievement event. It contains a configurable verification rule engine to determine the authenticity and validity of events represented by the fused data packets. This verification is not based on a simple trigger from a single data source, but rather on a preset, composable set of rules. In this embodiment, these rules may include:

[0091] Multi-source cross-validation rule: The achievement of a state node must be jointly confirmed by at least two different types of data sources. For example, the status of template installation completion requires the simultaneous receipt of a completion report submitted by the construction worker's mobile terminal and on-site images taken by a fixed camera or drone and identified by computer vision algorithms.

[0092] Authorized signature verification rules: For critical quality control nodes, such as the acceptance of concealed works, the event data must contain the digital signature of the quality inspection engineer or supervisor with specific permissions.

[0093] Logical consistency rule: Verify whether the current state event to be confirmed is consistent with the current state of its predecessor state node defined in the process state evolution diagram. For example, the event of concrete pouring completion can only be verified if its predecessor state of formwork reinforcement acceptance has been confirmed as achieved.

[0094] Structured trusted events are generated. Only data that passes the verification rules is considered a confirmed state achievement event. The physical digital state synchronization module 20 then generates trusted event objects according to the standardized data structure. The structure of the object is defined as follows:

[0095] ;

[0096] in:

[0097] : A globally unique identifier generated for this trusted event.

[0098] The precise timestamp at which the event was verified.

[0099] The identifier of the state node that has been achieved, i.e. ID.

[0100] The set of evidence used to support this verification includes an index or summary of all the original data packets used for verification.

[0101] : The rule ID or responsible person ID that triggered this successful verification.

[0102] Write trusted events to the event ledger and generate trusted event objects. The data is submitted and recorded in a trusted event ledger with tamper-proof features. Here, the trusted event ledger is a summary of all data that can guarantee data integrity, immutability, and traceability. In specific embodiments, this ledger can be implemented based on blockchain or distributed ledger technology, utilizing its decentralized and chained hash characteristics to ensure that each event record cannot be deleted or modified. In another embodiment, the ledger can also be a centralized database table with access control and operation log recording, allowing only append writes, and ensuring its integrity by hash signing each record.

[0103] The system triggers a state update in the process model. A listening process within the system continuously monitors the trusted event ledger. Once a new trusted event is successfully written, the process reads the state node identifier from the new event. Subsequently, the process accesses the process state evolution graph model and assigns the state node corresponding to the identifier. The current state is marked as achieved, and this update operation is the sole basis for subsequent decisions made by the rolling time-domain collaborative controller module 30, ensuring that all decisions are based on verified and undeniable field facts.

[0104] After acquiring the current system state, the core function of the rolling time-domain collaborative controller module 30 is to make forward-looking predictions of future system evolution. This prediction is not a simple deterministic deduction, but rather generates a large number of possible future scenarios by executing stochastic system simulations, providing a data foundation for subsequent optimization decisions. The simulation process specifically includes the following steps:

[0105] To set simulation parameters and initialize the simulation environment, the rolling time-domain co-controller module 30 first obtains the verified set of current state nodes from the physical digital state synchronization module 20, and sets a finite future prediction time domain length. At the same time, the total number of simulation iterations is set. This value determines the breadth of exploration into future uncertainties.

[0106] A single random evolution path is generated. In each iteration, the system starts from the current state and performs a complete random deduction. The specific process of this deduction is as follows:

[0107] First, identify the set of executable state transitions that have been achieved by all their preceding state nodes. Then, for each state transition in this set... Call the random number generator, based on its associated probabilistic time function. The duration of generating random samples The duration of this sample is the simulation time for that transition in this iteration. The system records the simulation completion time of all executable transitions and marks their subsequent state nodes as completed in the simulation environment according to their chronological order. This process occurs in the prediction time domain. Repeat this process until the simulation time exceeds the limit. .

[0108] The rolling temporal collaborative controller module constructs a set of future scenarios, and repeatedly executes 30 steps to generate a single random evolution path. Since the duration of the samples drawn from the probability distribution in each iteration is random, each iteration generates a unique future evolutionary path. These paths collectively constitute a set of future scenarios, which statistically reflects the system's performance in the prediction time domain. The Monte Carlo method can be used to simulate all possible state evolution trajectories within the system. The specific algorithm implementation is a well-known technology in this field and will not be elaborated here.

[0109] The output of the stochastic system simulation, namely the set of future scenarios, serves as the input for solving the subsequent multi-objective collaborative optimization problem. It presents uncertainty in the form of a series of specific scenarios, enabling the optimization process to be carried out on the basis of considering multiple possibilities.

[0110] After obtaining a set of future scenarios through random system simulation, the core task of the rolling time-domain collaborative controller module 30 is to evaluate different control strategies and find the optimal strategy. This process is achieved by constructing and solving a multi-objective collaborative optimization problem that unifies the two objectives of construction progress and material supply within the same digital framework.

[0111] The rolling time-domain collaborative controller module 30 constructs a multi-objective collaborative optimization function and establishes a comprehensive cost function. Used for quantization in the prediction time domain Internal, specific control sequences Overall performance of this control sequence Defined from the current moment To the future The daily construction task scheduling at any given time aims to find the optimal control sequence. This minimizes the expected value of the comprehensive cost function, which is defined as follows:

[0112] ;

[0113] in:

[0114] The expected value operator is implemented by using the control sequence... Applied to all generated in the preceding steps For each of the three random simulated paths, calculate its cost and then take the arithmetic mean.

[0115] For at any time The project delay cost is a quantified penalty for delays in key project milestones or the overall project duration. In this embodiment, it can be defined as the penalty resulting from the difference between the predicted completion time and the planned completion time of a key milestone in the simulated path.

[0116] For at any time The cost of material supply and demand mismatch is a key quantitative indicator for achieving schedule material coordination in this invention. It penalizes both advance supply and supply shortage of materials, and its specific composition is as follows:

[0117] ;

[0118] For at any time The amount of stockpiled material in the inventory resulting from supply planning exceeding actual consumption.

[0119] For at any time The inventory contains the amount of materials in short supply due to construction demand preceding the supply plan.

[0120] and These are the unit time holding cost coefficient and the shortage cost coefficient for each unit of material, respectively. These two coefficients are determined by project financial data or industry standards.

[0121] and These are two configurable weighting factors used to adjust the relative importance of the schedule target and the material cost target. Project managers can flexibly adjust these two weights according to different stages of the project.

[0122] To minimize the comprehensive cost function described above, a series of realistic physical and resource constraints must be satisfied, including:

[0123] Process logic constraint: The start of any state transition must be after all its preceding state nodes have been achieved. This constraint is directly defined by the topology of the process state evolution graph.

[0124] Resource availability constraints: at any given time The total amount of all resources required for all construction tasks assigned to be carried out shall not exceed the total amount of corresponding resources available for the project at that moment.

[0125] Solving this multi-objective collaborative optimization problem is a typical problem due to its inclusion of randomness and complex constraints. Therefore, the rolling time-domain collaborative controller module 30 employs a heuristic optimization algorithm to find an approximate optimal solution. Here, the heuristic optimization algorithm is a generalization of a class of algorithms that can find a relatively optimal solution within a reasonable time. In specific embodiments, genetic algorithms or particle swarm optimization algorithms can be used. The specific implementation of heuristic algorithms such as genetic algorithms is well-known in the art and will not be elaborated here. This algorithm iteratively optimizes and ultimately outputs a value that statistically satisfies the overall cost function. The optimal control sequence with the minimum expected value .

[0126] To obtain the optimal control sequence Subsequently, the rolling time-domain collaborative controller module 30 is responsible for converting this optimization result into specific instructions that can be executed on-site. This process reflects the rolling time-domain control concept and collaborative decision-making characteristics of the present invention.

[0127] Extracting the initial optimal control decision, the rolling time-domain cooperative controller module 30 extracts the complete predicted time-domain optimal control sequence. In this process, only the control decision at its first time step is extracted and adopted. This decision It contains a set of all state transitions that should be initiated within the current decision-making cycle. By adopting this rolling time-domain strategy, the system avoids fixing plans for a long period of time at once. Instead, it re-optimizes and calculates each decision-making cycle as time goes by and new information is acquired, so that the decision can continuously adapt to the actual changes and uncertainties on site.

[0128] The rolling time-domain cooperative controller module 30 generates a cooperative instruction set and makes the first-step optimal control decision. The code performs parsing and simultaneously generates two types of interrelated instructions, which together form a cooperative instruction set. The specific process for generating this cooperative instruction set is as follows:

[0129] Generate construction task instructions: Module traversal decision Includes all state transitions to be started The list of these state transitions, along with contextual information such as their respective components and spatial locations, is combined into a structured construction task instruction. This instruction clarifies the specific tasks that the construction team needs to perform in the next work cycle.

[0130] Generate material scheduling instructions: While generating construction task instructions, the module transitions based on these selected states. Query the resource consumption function that is pre-associated in the process state evolution graph. The module extracts the bill of materials from this function. It also summarizes and schedules all the materials required for the tasks, thereby generating precise material dispatch instructions. These instructions not only include the exact type and quantity of the required materials, but also the latest delivery time calculated based on the task start time.

[0131] The synergy here is reflected in the fact that material scheduling instructions are not based on independent inventory levels or consumption forecasts, but are directly derived from the optimal construction task instructions. This ensures that the material supply plan and the optimized construction schedule are consistent and coordinated from the source.

[0132] By issuing instruction sets through the interface, the rolling time-domain collaborative controller module 30 encapsulates the construction task instructions and material scheduling instructions generated in the previous step into standardized structured data formats, and issues them to different external execution systems through a preset application programming interface. In a specific embodiment, the construction task instructions can be pushed to the project's construction execution system or project management platform through the interface and assigned to the on-site construction teams in the form of electronic work packages. At the same time, the material scheduling instructions are sent to the enterprise's resource planning system or supply chain management system through another interface to trigger the corresponding material procurement, warehousing and picking, or logistics and distribution processes.

[0133] The model adaptive correction module 40 diagnoses potential systematic biases in the model by quantitatively comparing the predicted results with the actual execution results, providing a basis for subsequent model parameter correction. This process is the foundation for realizing the system's self-learning capability.

[0134] To obtain the actual duration of a state transition, for any completed state transition... The model's adaptive correction module 40 queries the trusted event ledger and obtains the results. and Confirmed timestamp and The actual duration of this state transition Calculated as:

[0135] ;

[0136] After obtaining the corresponding prediction duration, the model adaptive correction module 40 queries the same state transition from the process state evolution map. The associated probabilistic time function The prediction duration used for comparison is not the sample value in a random simulation, but rather the probabilistic duration function of the state transition. The expected value, i.e. This value represents the model's estimate of the central tendency over the duration of the transition when making a decision.

[0137] The model's adaptive correction module 40 quantifies and records the prediction error for each completed state transition. Calculate its single prediction error The calculation method is as follows:

[0138] ;

[0139] in: This represents the error of a single prediction. A positive error indicates that the actual time taken is longer than the model expects, while a negative error indicates that it is shorter than expected. This refers to the actual duration of the state transition; The error value is the expected value of the time taken for the state transition. This error value and the identifier of the corresponding state transition are stored together in the historical error database for subsequent statistical analysis.

[0140] To avoid overreacting to random fluctuations, this module does not perform model corrections every time an error occurs. Instead, it employs a statistical process control method to diagnose systematic biases. This systematic bias diagnosis is a generalization of all statistical test methods that can identify non-random patterns from a series of error data. In a specific embodiment, the model adaptive correction module 40 can maintain a cumulative sum control chart for each type of state transition. This method continuously accumulates positive and negative prediction errors. When the cumulative sum in either direction exceeds a preset control limit, the system determines that the historical prediction of that state transition has a significant and persistent systematic bias, rather than an accidental fluctuation. This diagnostic result will serve as a condition for triggering subsequent adaptive updates of model parameters.

[0141] When the diagnostic results of a systematic deviation indicate a state transition When there is a systematic bias in the prediction, the model adaptive correction module 40 is triggered to execute the probabilistic time function for that state transition. The parameters are adaptively updated. The prediction error-driven adaptive parameter update mechanism here is a generalization of all algorithms that can adjust model parameters to reduce future prediction errors by using recursive or batch updates based on historical error sequences. In a specific embodiment, this update mechanism can be implemented using a recursive update algorithm, which specifically includes the following steps:

[0142] The expected duration parameters are updated, and the model adaptive correction module 40 adjusts the state transition based on the most recent prediction error. For its probabilistic time function Expected value parameter The update process can follow the recursive formula below:

[0143] ;

[0144] in: The updated expected duration; The expected duration before the update; The most recent calculated single prediction error, i.e. .

[0145] The learning rate or gain factor is a configurable parameter ranging from 0 to 1, used to control the weight of the impact of the current error on the model parameter correction. A larger value indicates a higher learning rate. A higher value means the model is more sensitive to new data, while a smaller value makes the changes in model parameters smoother.

[0146] The model adaptive correction module 40 updates the variance parameter of the duration while simultaneously updating the expected duration and the variance parameter of the quantified uncertainty. This update allows the model to learn not only the average time of the process but also changes in its stability. This update can be achieved using the following recursive formula:

[0147] ;

[0148] in: This is the updated historical variance; The variance over time before the update; This is the square of the most recent prediction error, which reflects the extent to which the current observation deviates from the expectation; The learning rate for variance updates is also a configurable parameter that takes a value between 0 and 1.

[0149] This update mechanism ensures that when there is a persistent deviation between the actual duration and the expected duration, the model's expected value... It will be gradually corrected to tend towards the actual average time. At the same time, if the volatility of the process increases, the variance of the model will also increase. This will also increase accordingly, which will manifest as greater uncertainty in subsequent stochastic simulations, thus making decision-making more robust.

[0150] Persist the updated parameters, and then calculate the new parameters. and Subsequently, the model adaptive correction module 40 writes these two new values ​​back into the process state evolution graph, overwriting the corresponding state transitions. The old parameters are now complete, marking the end of a full parameter adaptive update closed loop. All subsequent collaborative control decisions based on this graph will use these revised model parameters, which better reflect the actual situation on site.

[0151] In addition to parameter-level corrections, the model adaptive correction module 40 also includes a higher-order correction mechanism to identify potential defects in the topology of the process state evolution map and generate interpretable optimization suggestions.

[0152] The model's adaptive correction module 40 continuously monitors the historical error database to monitor and identify persistently high prediction errors, and adjusts the module accordingly when specific state transitions occur. After a preset number of adaptive parameter updates, the absolute value of its prediction error The error remains consistently higher than the preset threshold. or its variance parameter If the growth is divergent, the system will determine that parameter adjustment alone is no longer sufficient to describe the actual behavior of the process. This situation indicates that the root cause of the problem may not be inaccurate parameters, but rather a defect in the model structure itself.

[0153] After triggering a condition that monitors and identifies persistently high prediction errors, the model adaptive correction module 40 extracts longer-term time series error data from the historical error database and performs correlation analysis to identify potential model structural problems. This correlation analysis of error patterns is a generalization of all data mining methods capable of discovering potential causal or correlational relationships from time series data. In specific embodiments, this analysis may include:

[0154] Error correlation calculation: The model adaptive correction module 40 calculates the cross-correlation coefficient between the prediction error sequences of different state transitions. If two state transitions that do not have a direct logical connection in the model show a significant positive or negative correlation in their prediction errors, it indicates that there may be a hidden dependency between them that has not been captured by the model.

[0155] Waiting Time Analysis: Model Adaptive Correction Module 40 analyzes the transitions between two consecutive states in the model. and The model adaptive correction module 40 calculates the actual completion time from the previous transfer. The actual start time of subsequent transfers The time interval between them, i.e., the waiting time If this waiting time is consistently significantly greater than zero and exhibits systematic characteristics, this indicates that at the state node... and One or more necessary state nodes may have been missed.

[0156] The model adaptive correction module 40 generates an interpretable model structure optimization suggestion based on the analysis results of the correlation analysis of the execution error pattern. This suggestion does not directly modify the model, but reports the problem to the user in text form and proposes possible correction directions.

[0157] If an error correlation is identified, the generated suggestion might be: A strong correlation has been detected between the time-based errors of the state transition task A description and the task B description. It is recommended to check whether these two tasks share constraint resources that are not explicitly defined in the model, or whether there are hidden process dependencies.

[0158] If a systematic wait time is identified, the generated suggestion may be: If a persistent delay is detected between the completion of the state J description and the start of the state K description, it is recommended to check and confirm whether any necessary intermediate steps or prerequisites are missing between the two.

[0159] These suggestions are presented to project managers or BIM engineers through the system's user interface, providing data-driven and specific guidance for their continuous optimization and iterative process state evolution map, thereby achieving a higher level of system learning and evolution.

Claims

1. A smart construction site progress and material consumption collaborative management platform based on digital twins, characterized in that, include: The process state evolution graph construction module is used to construct a process state evolution graph based on the static design information of the project. The process state evolution graph includes a series of state nodes and state transitions connecting the state nodes. The physical digital state synchronization module is used to acquire multi-source data from the physical construction site and verify and record the achievement events of the state nodes through a trusted event ledger in order to update the current state of the process state evolution graph. The rolling time-domain collaborative controller module is configured to proactively predict future evolution based on the current state of the process state evolution map, and generate a collaborative instruction set by solving a multi-objective collaborative optimization problem aimed at minimizing the comprehensive cost function. The model adaptive correction module is used to adaptively correct the model parameters or structure of the process state evolution map by comparing the actual execution results with the predicted results of the state transition.

2. The smart construction site progress and material consumption collaborative management platform based on digital twins as described in claim 1, characterized in that, The process state evolution map construction module is specifically used for: Associate each of the aforementioned state transitions with a probabilistic duration function to quantify its duration uncertainty; Associate a resource consumption function with each of the state transitions to define the required materials, labor, and machinery.

3. The smart construction site progress and material consumption collaborative management platform based on digital twins as described in claim 1, characterized in that, The physical digital state synchronization module is specifically used for: A verification rule engine is provided to perform cross-validation or logical consistency verification on the multi-source data according to preset rules before recording the state achievement event into the trusted event ledger.

4. The smart construction site progress and material consumption collaborative management platform based on digital twins as described in claim 1, characterized in that, The optimization objective of the comprehensive cost function solved by the rolling time-domain cooperative controller module includes: Minimize the cost of a project delay; Minimize the cost of a material supply-demand mismatch.

5. The smart construction site progress and material consumption collaborative management platform based on digital twins according to claim 4, characterized in that, Before solving the multi-objective cooperative optimization problem, the rolling time-domain cooperative controller module is also used for: Based on the probabilistic time function, multiple possible future evolution scenarios are generated by performing stochastic system simulations, and the comprehensive cost function is minimized under the statistical expectation of the multiple possible future evolution scenarios.

6. The smart construction site progress and material consumption collaborative management platform based on digital twins according to claim 1, characterized in that, The cooperative instruction set generated by the rolling time-domain cooperative controller module includes: Construction task instructions; Material dispatch instructions; The material scheduling instruction is derived synchronously from the state transitions and associated resource consumption functions contained in the construction task instruction.

7. The smart construction site progress and material consumption collaborative management platform based on digital twins as described in claim 1, characterized in that, The model adaptive correction module is specifically used for: Calculate the prediction error between the actual duration of the state transition and the expected value of the probabilistic duration function; When a systematic bias is diagnosed in the prediction error, the expected value and variance parameter of the probabilistic time function are updated using a recursive update algorithm based on the prediction error.

8. The smart construction site progress and material consumption collaborative management platform based on digital twins according to claim 1, characterized in that, The model adaptive correction module is also used for: When the prediction error of a specific state transition is found to remain higher than a preset threshold even after parameter updates, an error pattern correlation analysis is performed to identify potential structural defects in the process state evolution map and generate interpretable model structure optimization suggestions.

9. The smart construction site progress and material consumption collaborative management platform based on digital twins according to claim 4, characterized in that, The cost of matching material supply and demand includes: Material holding costs arising from supply exceeding consumption; and Costs arising from material shortages due to demand preceding supply.

10. A method for collaborative management of construction progress and material consumption at smart construction sites based on digital twins, wherein the platform for collaborative management of construction progress and material consumption at smart construction sites based on digital twins as described in any one of claims 1-9 is characterized in that, Includes the following steps: Constructing a process state evolution graph: Based on the static design information of the project, a process state evolution graph is constructed, which includes a series of state nodes and state transitions connecting the state nodes; Synchronize physical digital state: Obtain multi-source data from the physical construction site, and verify and record the achievement events of the state nodes through a trusted event ledger to update the current state of the process state evolution map; Generate cooperative instructions: Based on the current state of the process state evolution map, predict future evolution in a forward-looking manner, and generate a set of cooperative instructions by solving a multi-objective cooperative optimization problem aimed at minimizing the comprehensive cost function; Adaptive correction model: By comparing the actual execution results with the predicted results of the state transition, the model parameters or structure of the process state evolution map are adaptively corrected.

Images