Engineering cost dynamic prediction method and system based on digital twinning

Abstract

The invention provides an engineering cost dynamic prediction method and system based on digital twinning, and the method comprises the steps: generating a construction state map based on a building information model and a construction plan, the map comprises a node set representing a construction entity, an edge set representing a dependency relationship, and an attribute set containing a state field, and each node is associated with a space coordinate; constructing a state time sequence for each node based on construction state data collected in real time, wherein the state time sequence comprises a basic state field and a derived index; predicting the future unit cost based on the state time sequence and the map structure, and using a cost prediction model comprising a state mean item, a risk propagation item based on upstream node propagation and a time compression regular item based on a working condition; and dividing the engineering site into space units, calculating a multi-factor risk score and an operability score of each unit, classifying the space units according to the score matrix, and generating corresponding construction adjustment suggestions.

Description

Technical Field

[0001] This invention belongs to the field of engineering cost estimation, and in particular relates to a method and system for dynamic prediction of engineering costs based on digital twins. Background Technology

[0002] With the continuous expansion and increasing complexity of construction projects, cost control during construction has become one of the key tasks of project management. Project costs are not only affected by planned budgets and market prices, but also closely related to deviations in construction schedules, uneven distribution of resources, overlapping dependencies between work processes, and disruptions from unforeseen events. In practice, managers often rely solely on static budget models or periodic reports, which are insufficient to reflect the dynamic changes in the construction process or reveal the chain effects between different construction units. Traditional cost estimation models are largely based on historical data, lacking a characterization of construction logic, spatial coupling, and real-time status, resulting in often lagging and uninterpretable predictions. Even with the introduction of digital twin platforms, most existing systems remain primarily at the level of visualization, capable of simultaneously displaying individual data such as schedule and cost, but unable to form a complete dynamic prediction mechanism, let alone provide decision-making suggestions for project management. These shortcomings are particularly prominent in resource-intensive, complex, and time-sensitive projects. When delays, material shortages, or construction path conflicts occur, existing methods lack the ability to anticipate and optimize in advance, often leading to localized loss of control escalating into large-scale cost overruns. Summary of the Invention

[0003] The purpose of this invention is to design a method and system for dynamic prediction of engineering costs based on digital twins, which can dynamically predict cost changes, reveal areas of concentrated risk at the spatial level, and output specific intervention measures through data-driven rules.

[0004] To achieve the above objectives, a method for dynamic prediction of engineering costs based on digital twins is provided in a first aspect of the present invention, the method comprising:

[0005] A construction status map is generated based on the building information model and construction plan. The map includes a set of nodes representing construction entities, a set of edges representing dependencies, and a set of attributes containing status fields. Each node is associated with spatial coordinates.

[0006] Based on real-time collected construction status data, a status time series is constructed for each node, including basic status fields and derived indicators. The derived indicators include a status growth rate indicator that considers event disturbance terms and a spatial inconsistency penalty term based on spatial proximity relationships.

[0007] Predicting future unit costs based on state time series and graph structure, using a cost prediction model that includes state mean term, risk propagation term based on upstream node propagation, and time compression regularization term based on rush conditions.

[0008] The project site is divided into spatial units, and the multi-factor risk score and operability score of each unit are calculated. Based on the score matrix, the spatial units are classified and corresponding construction adjustment suggestions are generated.

[0009] Furthermore, the node status fields of the construction status map include progress completion percentage, time offset, and material unit consumption, where the progress completion percentage and time offset are initially set to 0, and the material unit consumption is initially empty.

[0010] Furthermore, the construction status graph is a directed acyclic graph structure, and the dependencies between nodes include the sequence of procedures and resource binding constraints.

[0011] Furthermore, the state growth rate index is obtained by calculating the average progress increment within the sliding window and subtracting the event disturbance term identified based on the construction anomaly log.

[0012] Furthermore, the spatial inconsistency penalty term is obtained by calculating the weighted sum of the differences between the node and its neighboring nodes in terms of schedule and material usage, where the proximity relationship is determined based on the Euclidean distance of the spatial coordinates.

[0013] Furthermore, the risk propagation term is obtained by aggregating the spatial perturbation scores and state transition amplitudes of upstream nodes, and the set of upstream nodes is determined by the directed edges in the graph.

[0014] Furthermore, the time compression regularization term is triggered when the time offset of a detected node is lower than a set threshold, indicating that the process is entering a rush state.

[0015] Furthermore, the multi-factor risk score is obtained by calculating the average predicted cost of each node within the spatial unit and weighting and integrating the spatial disturbance score, risk accumulation term, and resource tension level.

[0016] Furthermore, the construction adjustment recommendations include one or more measures such as construction delays, material staggering, personnel rotation, and equipment allocation, generating different combinations of recommendations for different types of spatial units.

[0017] A second aspect of the invention provides a dynamic engineering cost prediction system based on digital twins, the system comprising:

[0018] Construction adjustment recommendations include one or more measures such as construction delays, material staggering, personnel rotation, and equipment allocation, generating different combinations of recommendations for different types of spatial units.

[0019] The beneficial technical effects of the present invention are at least as follows:

[0020] To address the aforementioned issues, this invention provides a method and system for dynamic engineering cost prediction based on digital twins. In a digital twin environment, a construction status map containing components and tasks is generated using a BIM model and construction plan, providing a unified semantic structure and spatial interface for subsequent modeling. Subsequently, actual collected status data such as progress ratios, time offsets, and material consumption are embedded into the map nodes to establish a dynamic evolution model capable of expressing state abrupt changes and spatial differences. Based on this, a causal modeling mechanism is constructed, utilizing the upstream path propagation characteristics defined by the map structure and time compression regularization under scheduling rules to generate future cost predictions for each component or task, thereby achieving interpretable inference of the relationship between status and cost changes. Finally, the prediction results are mapped to the construction space, and a multi-factor risk index incorporating spatial disturbances, causal risks, and resource tensions is designed to identify cost risks in different construction areas. Combined with schedule flexibility, construction adjustment suggestions for management are generated. This method not only dynamically predicts cost changes but also reveals concentrated risk areas at the spatial level and outputs specific intervention measures through data-driven rules, achieving integrated design of prediction and management, effectively solving the problems of lack of dynamism and engineering operability in existing methods. Attached Figure Description

[0021] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0022] Figure 1 This is a flowchart of the dynamic prediction method for engineering costs based on digital twins according to the present invention.

[0023] Figure 2 This is a framework diagram of the dynamic engineering cost prediction system based on digital twins of the present invention. Detailed Implementation

[0024] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0025] In one or more embodiments, such as Figure 1 As shown, a method for dynamic prediction of engineering costs based on digital twins is disclosed, the method comprising the following:

[0026] S1: Generate a construction status map based on the building information model and construction plan. The map includes a set of nodes representing construction entities, a set of edges representing dependencies, and a set of attributes containing status fields. Each node is associated with spatial coordinates.

[0027] Specifically, the goal of this step is to construct a construction status graph structure to carry subsequent status information and cost prediction models. This graph is built during the initial stage of the project and does not involve real-time status acquisition or dynamic evolution; it is completed solely based on the engineering structure information and construction plan data already available in the early stages of the project. Therefore, this step is prioritized in terms of time and forms the structural foundation of the entire system. In this graph, each node represents an independent construction entity (including components or construction task units), and edges represent dependencies in the project execution process, such as the sequence of procedures or resource binding constraints. The node structure of the graph must support the mounting of status information, including fields such as progress, time offset, and material usage. The design of the graph structure must ensure: firstly, good consistency in construction logic; secondly, direct interface with the existing BIM and scheduling plans of the digital twin system without secondary development; and thirdly, compatibility with the status input dimensions of the cost prediction task, avoiding a disconnect between the status dimensions and the structural graph.

[0028] This step does not rely on the input of other steps. All the information required for its construction comes from the structural and planning data that was determined before the formal construction of the project. Specifically, this includes: the project's BIM (Building Information Modeling) model, read from the twin system via the IFC format interface, including component ID, component category, floor location, spatial coordinates, etc.; the project's construction plan information, usually expressed in Gantt charts or CPM networks to represent task relationships, which can be imported through scheduling management systems (such as Project, Primavera) via XML or API interfaces; and the initial cost number (non-numerical) of components or tasks in the project control budget system, used to uniquely identify the cost allocation unit. All this information is existing data from the project initiation and technical briefing stages and does not involve real-time collection later.

[0029] First, component IDs are extracted from the component list recorded in the BIM model, and key components for prediction are selected, such as load-bearing structures (columns, beams, floor slabs), enclosure structures (exterior walls, curtain walls), and large nodes (foundation raft slabs, large-span floor slabs, etc.). Each component ID is mapped to a node in the graph. In cases involving non-component elements, such as "slab pouring task in a certain area" or "specific process such as concrete curing," nodes can be defined by the task name in the construction plan.

[0030] After a node is created, its state field needs to be initialized. ,include:

[0031] : Progress completion percentage, initially set to 0, and subsequently updated periodically by the twin system's progress sensor or manual input;

[0032] Time offset, initially set to 0, and subsequently calculated by the scheduling system after comparing actual and planned times;

[0033] : Unit usage of materials, such as concrete , steel bars Initially empty, it will be recorded and uploaded by the construction site material system or sensors.

[0034] For example, the node ID of the raft foundation component on the basement floor of an office building is F-B1-001, the component type is "raft foundation," and the corresponding task is "concrete pouring." This component is represented as a node in the diagram. Its status field is initialized as follows:

[0035] ;

[0036] ;

[0037] ;

[0038] Next, based on the task dependencies imported in the scheduling plan, the edge set of the graph is constructed. For each pair of components or tasks and If the plan clearly states You can only begin after it's finished. Then, a directed edge is created in the graph. Such dependencies may include structural construction sequence (e.g., building the ground beams before the floor slabs) and resource conflict constraints (e.g., the same template system cannot be used for multiple tasks simultaneously).

[0039] The final structure of the map is defined as follows:

[0040] ;

[0041] in, It is a collection of components and task nodes. It is a set of dependencies. This is a collection of state fields attached to each node. It's important to emphasize that the graph structure is a directed acyclic graph (DAG), ensuring no circular dependencies exist. In a project, this structure can be automatically generated using a CPM graph. For example, exporting an XML file from Primavera P6 allows direct construction of the graph structure, with each node having WBS encoding, and the system automatically generating a set of dependency edges based on its preceding task table.

[0042] In addition, to enable subsequent spatial visualization mapping operations, it is necessary to record the spatial center point of each component or task. These spatial coordinates are read from the BIM system, typically calculated through the geometric center or obtained via the "ReferencePoint" attribute field. For example, a node... The spatial location is:

[0043] ;

[0044] This indicates that the center of the component is located 4.2 meters underground in the BIM coordinate system, with an X-axis distance of 35.6 meters and a Y-axis distance of 12.4 meters. If a node is not a component but an area task (such as "second floor slab pouring"), its center coordinates can be estimated by the average spatial position of multiple components in its affected area.

[0045] S2: Construct a state time series for each node based on real-time collected construction status data, including basic state fields and derived indicators. The derived indicators include a state growth rate indicator that considers event disturbance terms and a spatial inconsistency penalty term based on spatial proximity relationships.

[0046] Specifically, this step is based on the construction status map constructed in step one. and spatial location set For each node in the graph The goal of constructing a state evolution model is to establish a causal relationship between dynamic engineering states and future cost changes. This modeling process involves not only filling in state data but also transforming a static construction structure map into a dynamic semantic graph that tracks the evolution path of construction behavior. Compared to traditional models based on single-point estimation of time series data, this step proposes a three-dimensional state modeling mechanism oriented towards nodes, structural dependencies, and spatial locations. State modeling needs to embed unique construction dynamic characteristics such as "state transition behavior," "spatial volatility," and "resource lag" to support subsequent causal modeling and prediction accuracy.

[0047] For the status acquisition section, this step uses data from three sources to dynamically populate the data:

[0048] The progress is obtained by comparing the planned progress with the actual construction logs on site through the digital twin platform. The planned progress comes from the task time interval of the scheduling system, while the actual progress comes from the on-site check-in records, the updated construction status node annotations in the BIM model, or the video recognition interface.

[0049] It is the time difference between the actual start of the on-site task and the planned start time, generated by the task status log of the scheduling system;

[0050] It is the amount of material used per unit component, which comes from the construction material system. It is accumulated through methods such as batch material entry into the warehouse and on-site scanning and binding to components, and is allocated to components or task units from the total usage by the component binding table.

[0051] The local state growth rate index is used to measure the average rate of change of a state per unit time. Taking construction progress as an example, its growth rate index is:

[0052] ;

[0053] in, The width of the sliding window (e.g., 3 days) is used to calculate the average increment; This is a state disturbance item, used to indicate whether the node has experienced a sudden event (such as staff shortage or equipment failure) in the past three days. The value is 0 or 1, and it is automatically identified by the construction anomaly log in the digital twin system. It is the disturbance weighting coefficient, used to express the inhibitory effect of sudden events on the construction rhythm.

[0054] This formula introduces an event disturbance term as a "negative adjustment factor" to reflect the state stagnation or repetition caused by nonlinear sudden behaviors during construction. Such disturbances are extremely common in actual engineering, while traditional state modeling methods often ignore this behavior. The state evolution modeled in this way has a more realistic ability to express the semantics of construction.

[0055] Simultaneously, this step introduces a spatial inconsistency penalty term to reflect the degree of inconsistency between the state changes of a component or task and the states of its neighboring components. This term is not only used in subsequent prediction models but also helps in real-time identification of regional state anomalies. This spatial fluctuation term... The definition is as follows:

[0056] ;

[0057] in, This represents a set of spatially adjacent components, with adjacency determined by spatial location. and The Euclidean distance is less than the defined threshold. Position weights, typically ; This is a material difference penalty factor, used to amplify anomaly scores when materials are used unevenly.

[0058] This spatial penalty mechanism reflects a core management logic in construction scenarios: spatial continuity is a crucial foundation for project quality and efficiency management. When the progress or material usage of a component deviates significantly from that of its spatial neighbors, potential problems such as planning chaos, construction sequence skipping, or resource lag are highly likely. Traditional state modeling methods often focus on single-node behavior, neglecting this spatial coordination relationship. This solution introduces... This allows state evolution to not only express changes in the time dimension, but also reflect local spatial coordination.

[0059] Finally, the state sequence of each graph node is expanded into an enhanced state vector sequence, including... and derived indicators These values ​​will be embedded into the original map structure. The fields do not change the graph structure, but enhance the semantic information of the graph. The entire state sequence is organized with a uniform period (such as daily) through a timestamp alignment mechanism, forming a standardized time-series input structure that supports sequence analysis tasks in subsequent causal modeling.

[0060] S3: Predict future unit costs based on state time series and graph structure, using a cost prediction model that includes state mean, risk propagation based on upstream node propagation, and time compression regularization based on rush conditions.

[0061] Specifically, this step uses the state graph constructed in the second step. Based on this, and oriented towards each node in the graph This involves predicting engineering costs at future points in time. The goal is to construct a cost prediction model with engineering semantic constraints, state logic consistency, and spatial propagation interpretability to achieve the "interpretable dynamic prediction of engineering costs based on digital twins" required by this solution. This step is central to the entire system; its input comes from the output of the state evolution model, and its result will serve as the direct input for the next step of spatial cost mapping. Therefore, the quality of its modeling directly determines the system's prediction accuracy and management usability.

[0062] First, define the target prediction value as a node. In the future The unit cost of the Queen The overall cost prediction model is as follows:

[0063] ;

[0064] in, and They represent the past. The average progress and average material usage within a day are directly derived from the window aggregation results of the state sequence; This indicates propagation from the upstream node to... The aggregate value of cost risk This is the time compression adjustment item for the current node.

[0065] The construction employs a structurally constrained adjacency propagation mechanism to simulate the risk accumulation process of "upstream state chaos leading to increased costs for this node." It is defined as follows:

[0066] ;

[0067] in, Indicates the direction in the spectrum The set of all upstream nodes, These are the edge weights of the structural path, which can be set according to the scheduling order weights as either task urgency or structural relevance. It is a spatial disturbance score. This indicates the magnitude of the state transition of the upstream node. This is the perturbation amplification factor, representing the sensitivity of state transitions to cost impact. This propagation term incorporates the graph structure (through...). and The quantitative scoring of node state jumps and spatial chaos truly reflects the phenomenon in the project that "the instability of a certain construction unit will spread its influence downstream".

[0068] Subtractions in the cost model This is a time compression regularization term used to express the strong upward trend in unit cost when a task is being rushed. It is defined as:

[0069] ;

[0070] in, This is an indicator function that indicates whether the rush work condition (i.e., the planned delay exceeds a threshold) is triggered. If satisfied, then the penalty term is taken. Otherwise, it is 0. This simulation describes a typical phenomenon in engineering scheduling where additional resources are forcibly invested due to schedule delays. This usually manifests as overtime work, additional materials, and the rental of emergency equipment, all of which significantly increase unit construction costs. This aspect is difficult to reflect in traditional machine learning prediction models, but it is embedded in this solution in a rule-based manner, taking into account both interpretability and deployability.

[0071] The entire model structure uses only state means, local propagation terms, and regularization terms driven by engineering rules. With a very small number of variables, it effectively captures key cost change drivers and preserves crucial causal path information within the graph structure. (Each coefficient...) and , All parameters can be set by fitting historical data or expert experience, and interfaces can be provided during the system deployment phase for project managers to fine-tune them to adapt to the cost control styles of different construction projects.

[0072] S4: Divide the engineering site into spatial units, calculate the multi-factor risk score and operability score of each unit, classify the spatial units according to the score matrix, and generate corresponding construction adjustment suggestions.

[0073] Specifically, this step serves as the final stage of the entire system, building upon the unit cost prediction results generated in the previous phase. Spatial coordinates of the map nodes This completes a closed-loop process from numerical prediction to spatial risk identification and management recommendations. Compared to conventional engineering data visualization methods, this step not only represents the prediction results continuously in the spatial dimension, but also incorporates typical high-risk factors in engineering construction scenarios such as "resource stacking," "work process overlap," and "plan compression." By designing a multi-dimensional weight aggregation mechanism, it achieves dynamic risk identification and generates actionable construction adjustment recommendations based on schedule flexibility and resource allocation rules, thus truly embodying a closed-loop system encompassing the entire process of "prediction-identification-control."

[0074] To achieve spatialized risk aggregation, this step divides the engineering site into a set of discrete spatial units. each It contains several graph nodes. The division method is based on the actual construction management logic (such as floor, area, construction section), and each unit forms a local risk analysis area. For each spatial unit, a multi-factor risk scoring function combining cost prediction results, state disturbance intensity, and resource saturation is designed. This is used to assess the potential risk of "cost runaway" in the region during the forecast period. The specific construction is as follows:

[0075] ;

[0076] in, This is the unit cost predicted in the previous step; The spatial perturbation score is derived from the state modeling in the second step. This represents the risk accumulation term in the causal propagation path, derived from the third step of modeling. Resource tension level scoring is provided by the construction resource scheduling system. It assesses the quantity of resources (manpower, machinery, equipment) required for the construction of a single component, as well as the intensity of concurrent demand. The levels are typically 1 (low), 2 (medium), and 3 (high). (Coefficient) , and These represent the sensitivity to spatial conflicts, path risks, and resource conflicts, respectively, and can be set through historical project optimization or specified by the user.

[0077] The formula is designed not only to consider cost prediction itself, but also to incorporate three factors that have a strong risk amplification effect in actual construction: the inconsistency of disturbances between construction blocks ( ), chain reaction risks along the construction path ( ), the execution pressure brought about by resource concentration This combination achieves a three-dimensional improvement from "numerical cost" to "space-organization-planning linkage risk," possessing extremely strong engineering scenario matching capabilities, representing a significant technological leap for this step compared to general heatmap rendering.

[0078] Risk index for completing all spatial units After calculation, the system compares it with the dynamically generated construction schedule and calculates the operability score for each risk unit. This is used to determine whether there is a realistic possibility of adjusting the construction strategy in the area. It is defined as follows:

[0079] ;

[0080] in, Representing spatial units The remaining adjustable time flexibility (in days) in the current plan is the construction window from the current time to the latest time when it must not be delayed. It is a tiny constant to prevent the denominator from being zero. This indicator is used to quantify the "risk / elasticity" ratio. The larger the value, the higher the risk but the more room for adjustment. The smaller the value, the more difficult it is to adjust even if the risk is high, and early warning is required.

[0081] The system according to and The resulting two-dimensional analysis matrix divides all spatial units into four states:

[0082] High risk - high flexibility: It is recommended to immediately adjust the work process, such as delaying the task or redistributing materials;

[0083] High risk - low flexibility: It is recommended to trigger a cross-disciplinary coordination mechanism, and if necessary, seize resources or restructure the construction plan;

[0084] Low risk - low flexibility: It is recommended to proceed conservatively, maintain the current pace, and increase the frequency of monitoring;

[0085] Low risk - high elasticity: No processing is required; it can be considered a scheduling buffer.

[0086] Based on the above classification, the system automatically generates corresponding management suggestion texts and suggested parameters for each type of spatial unit, including: suggested construction delay days, number of material batches, personnel rotation rhythm, equipment allocation sequence, etc., and outputs them in the form of structured data, which can be connected to the construction scheduling platform for secondary verification and manual intervention.

[0087] The final output includes two objects: one is a spatial heatmap, which displays... The system provides two main features: first, it displays the distribution of risk information and allows for color overlay in the digital twin platform; second, it includes a management suggestion parameter package for each high-risk area, containing fields such as suggested actions, risk level, intervention priority, and plan adjustment range, which project managers can access and execute.

[0088] In one or more embodiments, such as Figure 2 As shown, a dynamic engineering cost prediction system based on digital twins is disclosed, the system comprising:

[0089] The graph construction module is used to generate a construction status graph based on the building information model and construction plan. The graph includes a set of nodes representing construction entities, a set of edges representing dependencies, and a set of attributes containing status fields. Each node is associated with spatial coordinates.

[0090] The state evolution module is used to construct a state time series for each node based on real-time collected construction state data, including basic state fields and derived indicators. The derived indicators include a state growth rate indicator that considers event disturbance terms and a spatial inconsistency penalty term based on spatial proximity relationships.

[0091] The causal prediction module is used to predict future unit costs based on state time series and graph structure. It uses a cost prediction model that includes a state mean term, a risk propagation term based on upstream node propagation, and a time compression regularization term based on rush conditions.

[0092] The spatial analysis module is used to divide the engineering site into spatial units, calculate the multi-factor risk score and operability score of each unit, classify the spatial units according to the score matrix, and generate corresponding construction adjustment suggestions.

[0093] It is worth noting that the specific workflow of the engineering cost dynamic prediction system based on digital twins provided in this embodiment of the invention is the same as that of the engineering cost dynamic prediction method based on digital twins described in the above embodiments, and will not be repeated here.

[0094] This invention also provides a digital twin-based dynamic engineering cost prediction device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the steps described in the above embodiments of the digital twin-based dynamic engineering cost prediction method, for example... Figure 1 The steps S1 to S4 described above; or, when the processor executes the computer program, it implements the functions of each module in the above system embodiments.

[0095] For example, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the digital twin-based dynamic engineering cost prediction device.

[0096] The digital twin-based dynamic cost prediction device can be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. This device may include, but is not limited to, processors and memory. Those skilled in the art will understand that the digital twin-based dynamic cost prediction device may also include input / output devices, network access devices, buses, etc.

[0097] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASACs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. This processor is the control center of the digital twin-based dynamic cost prediction device, connecting all parts of the device via various interfaces and lines.

[0098] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the digital twin-based dynamic engineering cost prediction device by running or executing the computer programs and / or modules stored in the memory, and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function, etc.; the data storage area may store data created based on the operation of the air conditioner controller, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0099] The modules integrated into the digital twin-based dynamic cost prediction device for engineering projects, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0100] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0101] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for dynamic prediction of engineering costs based on digital twins, characterized in that, The method includes: A construction status map is generated based on the building information model and construction plan. The map includes a set of nodes representing construction entities, a set of edges representing dependencies, and a set of attributes containing status fields. Each node is associated with spatial coordinates. Based on real-time collected construction status data, a status time series is constructed for each node, including basic status fields and derived indicators. The derived indicators include a status growth rate indicator that considers event disturbance terms and a spatial inconsistency penalty term based on spatial proximity relationships. Predicting future unit costs based on state time series and graph structure, using a cost prediction model that includes state mean term, risk propagation term based on upstream node propagation, and time compression regularization term based on rush conditions. The project site is divided into spatial units, and the multi-factor risk score and operability score of each unit are calculated. Based on the score matrix, the spatial units are classified and corresponding construction adjustment suggestions are generated.

2. The method for dynamic prediction of engineering costs based on digital twins according to claim 1, characterized in that, The node status fields of the construction status map include progress completion percentage, time offset, and material unit usage, wherein the progress completion percentage and time offset are initially set to 0, and the material unit usage is initially empty.

3. The method for dynamic prediction of engineering costs based on digital twins according to claim 1, characterized in that, The construction status graph is a directed acyclic graph structure, and the dependencies between nodes include the sequence of procedures and resource binding constraints.

4. The method for dynamic prediction of engineering costs based on digital twins according to claim 1, characterized in that, The state growth rate index is obtained by calculating the average progress increment within the sliding window and subtracting the event disturbance term identified based on the construction anomaly log.

5. The method for dynamic prediction of engineering costs based on digital twins according to claim 1, characterized in that, The spatial inconsistency penalty term is obtained by calculating the weighted sum of the differences between the node and its neighboring nodes in terms of progress and material usage, wherein the proximity relationship is determined based on the Euclidean distance of the spatial coordinates.

6. The method for dynamic prediction of engineering costs based on digital twins according to claim 1, characterized in that, The risk propagation term is obtained by aggregating the spatial perturbation scores and state transition amplitudes of upstream nodes, and the set of upstream nodes is determined by the directed edges in the graph.

7. The method for dynamic prediction of engineering costs based on digital twins according to claim 1, characterized in that, The time compression regularization term is triggered when the time offset of a node is detected to be lower than a set threshold, indicating that the process has entered a rush state.

8. The method for dynamic prediction of engineering costs based on digital twins according to claim 1, characterized in that, The multi-factor risk score is obtained by calculating the average predicted cost of each node within a spatial unit and then weighting and integrating the spatial disturbance score, risk accumulation term, and resource tension level.

9. The method for dynamic prediction of engineering costs based on digital twins according to claim 1, characterized in that, The proposed construction adjustments include one or more measures such as construction delays, material staggering, personnel rotation, and equipment allocation, generating different combinations of recommendations for different types of spatial units.

10. A dynamic engineering cost prediction system based on digital twins, characterized in that, The system includes: The graph construction module is used to generate a construction status graph based on the building information model and construction plan. The graph includes a set of nodes representing construction entities, a set of edges representing dependencies, and a set of attributes containing status fields. Each node is associated with spatial coordinates. The state evolution module is used to construct a state time series for each node based on real-time collected construction state data, including basic state fields and derived indicators. The derived indicators include a state growth rate indicator that considers event disturbance terms and a spatial inconsistency penalty term based on spatial proximity relationships. The causal prediction module is used to predict future unit costs based on state time series and graph structure. It uses a cost prediction model that includes a state mean term, a risk propagation term based on upstream node propagation, and a time compression regularization term based on rush conditions. The spatial analysis module is used to divide the engineering site into spatial units, calculate the multi-factor risk score and operability score of each unit, classify the spatial units according to the score matrix, and generate corresponding construction adjustment suggestions.

Images