Fabricated building-oriented component-level cost intelligent verification method and system thereof

Abstract

The invention relates to the field of construction engineering cost, in particular to a fabricated building-oriented component-level cost intelligent verification method and system, which perform text analysis through a deep learning model, significantly improve the accuracy of component identification, obtain a cost file and component information of a fabricated building project, and improve the verification efficiency of the fabricated building project. Performing text analysis on the cost file based on a deep learning model, and identifying component description and cost data; multi-dimensional feature vectors of the components are extracted, and a multi-dimensional relation graph reflecting the relation between the components is constructed; on the basis of the multi-dimensional feature vectors and the multi-dimensional relation graph, the actual cost and the budget cost of the component are calculated, and according to the difference between the actual cost and the budget cost, a multi-level deviation calculation system is constructed for difference analysis; establishing a hierarchical early warning mechanism according to the difference analysis result, and generating early warning information when the deviation exceeds a preset threshold value; a deep learning model is introduced, and the accuracy of component recognition is improved from about 85% of traditional manual recognition to 95% or above.

Description

Technical Field

[0001] The present invention relates to the field of construction project cost, and in particular to a component-level intelligent cost verification method and system for prefabricated buildings, which is suitable for cost management and control of prefabricated buildings throughout their entire life cycle. Background Art

[0002] As an important development direction of building industrialization, prefabricated buildings are gradually becoming the mainstream development trend in the construction industry, with advantages such as short construction period, low environmental pollution, and high quality controllability. However, due to the characteristics of prefabricated and standardized components, the cost accounting of prefabricated buildings is significantly different from that of traditional cast-in-place buildings, which is mainly manifested in the following aspects:

[0003] First, the cost structure of prefabricated buildings is more complex. In addition to the material and labor costs of traditional construction, it also includes a variety of unique cost categories, such as component production, transportation, and installation costs. Second, cost accounting for prefabricated buildings requires precise component-level granularity. Components of different types and specifications vary significantly in cost due to factors such as production processes, material requirements, and installation difficulty. Furthermore, the complex spatial, functional, and process relationships between components in prefabricated buildings also have a significant impact on cost accounting.

[0004] Currently, the cost verification of prefabricated buildings relies primarily on manual experience, resulting in low efficiency, poor accuracy, and difficulty in promptly detecting cost deviations. While some existing cost management software can assist with cost calculations, these software primarily relies on simple mathematical formulas and preset rules, unable to effectively handle the complex component relationships and multi-dimensional features, making them difficult to adapt to the unique needs of prefabricated building cost management.

[0005] The development of artificial intelligence (AI), particularly deep learning and natural language processing, has provided a new path for intelligent cost verification of prefabricated buildings. However, the application of existing intelligent technologies in the field of prefabricated building costs is still in its infancy, lacking core functions such as comprehensive component feature extraction, precise modeling of component relationships, and intelligent early warning of cost deviations.

[0006] Therefore, how to combine advanced artificial intelligence technology with professional knowledge of prefabricated building cost management to develop an efficient and accurate component-level cost intelligent verification method and system has become a technical problem that needs to be solved urgently. Summary of the Invention

[0007] In view of the problems existing in the above-mentioned prior art, the purpose of the present invention is to provide a component-level intelligent cost verification method and system for prefabricated buildings, aiming to achieve accurate identification, feature extraction, relationship modeling, cost calculation and deviation warning of prefabricated building components through the integration of deep learning technology and professional cost knowledge, thereby improving the accuracy and efficiency of cost verification.

[0008] The present invention proposes a component-level intelligent cost verification method and system for prefabricated buildings, including:

[0009] Obtain cost documents and component information for prefabricated building projects;

[0010] Performing text analysis on the cost document based on a deep learning model to identify component descriptions and cost data;

[0011] Extracting a multi-dimensional feature vector of the component based on the recognition result, wherein the multi-dimensional feature vector includes a basic feature dimension, a material feature dimension, a process feature dimension, and an installation feature dimension;

[0012] Constructing a multidimensional relationship map reflecting the relationship between components, wherein the multidimensional relationship map includes spatial relationship, functional relationship, process relationship and organizational relationship;

[0013] Calculating the actual cost and budget cost of the component based on the multi-dimensional feature vector and the multi-dimensional relationship map;

[0014] Based on the difference between the actual cost and the budgeted cost, a multi-level deviation calculation system is constructed to perform difference analysis;

[0015] Based on the difference analysis results, a hierarchical early warning mechanism is established to generate early warning information when the deviation exceeds a preset threshold;

[0016] The hierarchical early warning mechanism sets differentiated early warning thresholds according to component importance, project stage and deviation type.

[0017] Preferably, the obtaining of cost documents and component information of the prefabricated building project includes:

[0018] Obtain basic information of prefabricated building projects from the project management system;

[0019] Obtain cost budget files from the cost management system;

[0020] Obtain geometric information and topological relationships of components from the BIM system;

[0021] The basic information, the cost budget file and the geometric information are integrated into structured data for subsequent analysis and processing.

[0022] Preferably, the text analysis of the cost document based on the deep learning model includes:

[0023] Use multi-level text segmentation processing to divide the cost document into project information paragraphs, component description paragraphs and cost data paragraphs;

[0024] Enhance the system's ability to understand prefabricated building terminology using specialized dictionaries, including component type dictionaries, material dictionaries, and process dictionaries;

[0025] Applying a component recognition and positioning system to the segmented text paragraphs to determine the boundaries of component descriptions and extract attribute information;

[0026] The identified components are associated with the corresponding cost information to form a component-cost mapping relationship.

[0027] Preferably, the multi-dimensional feature vector of the extraction component includes:

[0028] Extract basic feature dimension information, including geometric features, physical features and quantitative features of components;

[0029] Extract material characteristic dimension information, including material type, material grade and material properties;

[0030] Extract process feature dimension information, including production process, connection process and surface treatment;

[0031] Extract installation feature dimension information, including installation method, installation sequence and auxiliary facilities;

[0032] The extracted features are standardized and features of different measurement units are converted into a unified standard.

[0033] Preferably, the constructing of a multidimensional relationship map reflecting the relationship between components includes:

[0034] Initialize component nodes based on component identification results;

[0035] Analyze the relationship descriptions in the cost documents and establish explicit relationships between components;

[0036] Infer the implicit relationship between components based on their positions and functional characteristics;

[0037] Apply relationship transfer rules to expand indirect relationships between components;

[0038] Calculate the relationship strength and divide it into levels according to the importance of the relationship.

[0039] Preferably, the calculating the actual cost and the budgeted cost of a component based on the multi-dimensional feature vector and the multi-dimensional relationship graph includes:

[0040] Establish a multi-level cost decomposition system to decompose the cost into component-level cost, subsystem-level cost and project-level cost;

[0041] Select an appropriate cost calculation model based on component characteristics;

[0042] Consider detailed factors such as component production batch, degree of standardization, technical difficulty, transportation conditions and installation environment;

[0043] Optimize the cost calculation path based on the component relationship diagram to ensure the cost calculation sequence of related components is reasonable;

[0044] According to the correlation and sensitivity analysis results between features, adjust the weight of the impact of features on construction cost.

[0045] Preferably, a multi-level deviation calculation system is constructed based on the difference between the actual cost and the budgeted cost, including:

[0046] Calculate the deviation value of each feature and standardize it into a percentage;

[0047] Calculate component-level comprehensive deviation based on feature weights;

[0048] Based on the component relationship graph, identify functionally related component groups and calculate component group deviations;

[0049] Calculate system-level deviations by functional system, spatial region, and other dimensions;

[0050] Analyze the deviation distribution pattern and identify the deviation concentration areas and propagation paths.

[0051] Preferably, the step of establishing a hierarchical early warning mechanism based on the difference analysis results includes:

[0052] Set differentiated warning thresholds for key components, important components, and general components;

[0053] Dynamically adjust the stringency of warning thresholds based on the project stage;

[0054] Distinguish between cost-increasing deviations and cost-reducing deviations and adopt asymmetric early warning strategies;

[0055] Establish early warning classification standards, and classify early warning into level one, level two, level three, and level four;

[0056] Design differentiated warning push strategies and perform intelligent push based on user roles, warning urgency, and warning relevance.

[0057] Preferably, the establishment of a hierarchical early warning mechanism further comprises:

[0058] Perform intelligent root cause analysis to locate the source and impact of deviations;

[0059] Trace back feature changes and parameter adjustments that caused the deviation;

[0060] Assess the pathways of bias transmission and potential impacts;

[0061] Generate structured root cause analysis reports;

[0062] Provide targeted cost optimization suggestions to reduce or eliminate deviations.

[0063] The component-level intelligent cost verification system for prefabricated buildings includes:

[0064] A text analysis and component identification module is used to obtain cost documents and component information of prefabricated building projects, perform text analysis on the cost documents based on a deep learning model, and identify component descriptions and cost data;

[0065] A feature extraction and matching module is used to extract a multi-dimensional feature vector of a component, wherein the multi-dimensional feature vector includes a basic feature dimension, a material feature dimension, a process feature dimension, and an installation feature dimension;

[0066] A component relationship modeling module is used to construct a multi-dimensional relationship map reflecting the relationship between components, including spatial relationship, functional relationship, process relationship and organizational relationship;

[0067] A cost calculation and analysis module, configured to calculate the actual cost and budgeted cost of a component based on the multi-dimensional feature vector and the multi-dimensional relationship graph;

[0068] A deviation detection and early warning module is used to construct a multi-level deviation calculation system based on the difference between the actual cost and the budgeted cost, perform difference analysis, and establish a hierarchical early warning mechanism based on the difference analysis results to generate early warning information when the deviation exceeds a preset threshold;

[0069] The hierarchical early warning mechanism sets differentiated early warning thresholds according to component importance, project stage and deviation type.

[0070] The beneficial effects of the present invention include:

[0071] 1. By introducing a deep learning model for text analysis, the accuracy of component recognition has been significantly improved, from approximately 85% with traditional manual recognition to over 95%, significantly reducing missed and misidentified instances.

[0072] 2. The use of multi-dimensional feature vector extraction technology comprehensively captures the geometric, material, process and installation characteristics of components, making the consideration of cost influencing factors more comprehensive and improving the cost accounting accuracy by about 20%.

[0073] 3. A multi-dimensional relationship map reflecting the complex relationship between components was constructed. For the first time, the spatial, functional, process and organizational relationships between components were taken into consideration in cost verification, providing key support for accurate cost calculation.

[0074] 4. A multi-level deviation calculation system and a hierarchical early warning mechanism have been established, which has upgraded the detection of cost deviations from the traditional project level to the component level and component group level. The timeliness of deviation discovery has been advanced by about 60%, greatly reducing the risk of later rework and cost overruns.

[0075] 5. Overall, the present invention increases the time efficiency of prefabricated building cost verification by 3-5 times, improves cost accuracy by 15%-25%, and achieves an early warning accuracy rate of over 90%, providing strong support for cost control and quality management of prefabricated buildings. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 This is a flow chart of the component-level intelligent cost verification method for prefabricated buildings according to the present invention;

[0077] Figure 2 Schematic diagram of the structure of the text analysis and component recognition module of the present invention;

[0078] Figure 3 A schematic diagram of multi-dimensional feature vector extraction according to the present invention;

[0079] Figure 4 A schematic diagram of the multidimensional relationship map constructed in the present invention;

[0080] Figure 5 A hierarchical structure diagram of the multi-level deviation calculation system of the present invention;

[0081] Figure 6 Schematic diagram of hierarchical warning of the hierarchical warning mechanism of the present invention;

[0082] Figure 7 This is a structural block diagram of the component-level intelligent cost verification system for prefabricated buildings of the present invention. DETAILED DESCRIPTION

[0083] Please refer to the attached Figure 1-7 , the specific implementation of the present invention is further described in detail below with reference to the accompanying drawings.

[0084] Reference Figure 1 The present invention provides a component-level intelligent cost verification method for prefabricated buildings, which includes the following steps:

[0085] Obtaining cost documents and component information of prefabricated building projects: In one embodiment of the present invention, firstly, it is necessary to obtain relevant data of prefabricated building projects as input. Figure 1 As shown, this step mainly involves obtaining data from three different sources and integrating them.

[0086] Specifically, the basic information of the prefabricated building project is first obtained from the project management system, including basic data such as project name, building type, building area, number of floors, and estimated construction period. This data is usually stored in a structured form in the project management database and can be obtained through the API interface or data export function.

[0087] Second, obtain the cost budget file from the cost management system. This file typically contains information such as the total project budget, sub-item costs, and the estimated unit prices of various components. These files may be in Excel spreadsheets, PDF documents, or proprietary formats of professional cost management software, requiring appropriate format conversion and preprocessing.

[0088] Next, the geometric information and topological relationships of the components are obtained from the BIM system. The BIM model contains detailed three-dimensional geometric data, positional information, and the connection relationships between components of the prefabricated building. Preferably, the present invention extracts this information, including component type, size, volume, surface area, and the spatial relationship between components, through the open interface of the BIM system.

[0089] Finally, the three types of data are integrated to form a structured dataset, providing a foundation for subsequent analysis and processing. During this integration process, it is necessary to establish mapping relationships between project information, cost data, and component information to ensure data consistency and integrity. Preferably, the present invention uses a relational database to store this integrated data for efficient subsequent querying and processing.

[0090] Text analysis of cost documents based on deep learning model: Figure 2 The text analysis and component identification module 1 of the present invention is primarily responsible for intelligently parsing the acquired cost documents to identify component descriptions and cost data. This step is the foundation of the entire method, and its core lies in how to accurately extract useful information from unstructured or semi-structured text.

[0091] In a preferred embodiment of the present invention, the text analysis process first utilizes a multi-level text segmentation process to divide the cost document into functionally distinct sections. Specifically, the document is first analyzed holistically to identify its structural features. Next, the document is segmented based on content function into project information sections, component description sections, and cost data sections. Each paragraph is then segmented at the sentence level to identify key information sentences. Finally, key phrases such as component names, parameters, and quantities are extracted from the sentences.

[0092] To improve understanding of specialized terminology, this system has developed a specialized dictionary system for prefabricated buildings, including a component type dictionary, a material dictionary, and a process dictionary. The component type dictionary covers common prefabricated component types such as prefabricated columns, prefabricated beams, prefabricated wall panels, and composite floor slabs; the material dictionary covers common materials such as C30 concrete, HRB400 rebar, and polystyrene boards; and the process dictionary covers common connection methods such as wet joints, bolted connections, and welding. These specialized dictionaries significantly enhance the system's understanding of specialized text.

[0093] To identify and locate components, the present invention first labels potential component names based on a specialized dictionary, then determines the textual boundaries of component descriptions. It then associates attribute information scattered throughout the text with the corresponding components, ultimately creating a unique identifier for each identified component. Preferably, the present invention establishes identification criteria: components must possess basic information such as type and size for valid identification; components with high frequency and direct relevance to construction cost are prioritized; and components containing special markings (such as numbers and drawing references) enhance identification reliability.

[0094] The deep learning model employed in this paper is a pre-trained language model based on BERT (Bidirectional Encoder Representations from Transformers), fine-tuned for the prefabricated building domain. This model's core advantage lies in its ability to understand the meaning of words in different contexts and capture long-range semantic dependencies. The preferred model architecture includes 12 Transformer encoder layers, a hidden layer dimension of 768, 12 attention heads, and a vocabulary of 30,000 words, including 3,000 specialized words for prefabricated buildings.

[0095] The deep learning model was trained using 100,000 real-world prefabricated building cost documents, covering a variety of building types, including residential, commercial, and public. After 200 epochs of training, the model achieved an F1 score of 0.93 on the validation set, significantly higher than the 0.76 achieved by traditional rule-based methods.

[0096] Extracting multi-dimensional feature vectors of components: Figure 3 After component identification, the present invention further extracts the component's multi-dimensional feature vector. The core of this step is to convert the component's various attribute information into a structured feature vector, providing a basis for subsequent cost calculation.

[0097] In a preferred embodiment of the present invention, the feature extraction process establishes a four-dimensional feature system, including a basic feature dimension, a material feature dimension, a process feature dimension, and an installation feature dimension.

[0098] Basic characteristic dimensions primarily include geometric features (such as length, width, height, volume, and surface area), physical features (such as weight, density, and strength), and quantitative features (such as the number of components and units). Ideally, geometric features are extracted directly from the BIM model, while physical features are calculated based on material type and geometric data. For example, the weight of a precast wall panel can be calculated by multiplying its volume by the density of concrete (approximately 2400 kg / m³).

[0099] Material feature dimensions include material type (e.g., concrete, steel, insulation), material grade (e.g., C30, C40 concrete, HRB400, HRB500 rebar), and material properties (e.g., compressive strength, tensile strength, thermal conductivity, etc.). Preferably, the present invention establishes a material property database containing various performance parameters of commonly used building materials for reference and supplementation during feature extraction.

[0100] Process feature dimensions include production processes (such as mold type and vibration method), connection processes (such as welding, bolting, and post-casting), and surface treatments (such as painting, polishing, and roughness). These features are typically extracted from component description text or inferred based on component type and usage. For example, exterior wall panels typically require waterproofing and exterior finishing, while interior partition walls may only require simple plastering.

[0101] Installation feature dimensions include installation method (e.g., hoisting, jacking, sliding), installation sequence (e.g., prioritization, dependencies, etc.), and auxiliary facilities (e.g., temporary supports, positioning tools, etc.). These features are crucial for accurately calculating installation costs. Preferably, the present invention automatically infers the appropriate installation method based on component type, weight, and location. For example, large prefabricated components weighing over 5 tons typically require a large tower crane, while exterior wall panels installed on high floors may require specialized exterior wall lifting equipment.

[0102] The specific steps of feature extraction include: first, based on the component recognition results, locate the component description text; then apply feature vocabulary detection to the text to mark potential feature information; then use the context relationship to confirm the attribution relationship between the feature and the component; then extract the feature value (numerical value, classification value, etc.); finally, standardize the feature value and convert it into a unified measurement unit.

[0103] To ensure the accuracy of feature extraction, this paper introduces a feature matching and alignment mechanism. Specifically, it first performs an exact match to find entries whose feature descriptions are exactly the same as those in the database. Then, it performs a fuzzy match to calculate the semantic similarity between the feature descriptions and the database entries. Finally, it performs a multi-stage verification to ensure that the matching results are internally consistent and have no contradictory features, and to evaluate the match reliability.

[0104] When features are missing or conflicting, the present invention adopts an intelligent completion and conflict resolution strategy: when a feature is missing, it is completed based on inference based on similar components; when features conflict, more specific, recent, and authoritative information is given priority; when the units of feature values ​​are inconsistent, they are automatically converted to standard units.

[0105] Construct a multidimensional relationship map reflecting the relationship between components: Figure 4 After obtaining the component features, the present invention further constructs a multi-dimensional relationship map reflecting the relationship between components. The significance of this step is to capture the complex relationship between components and provide a relationship basis for cost calculation.

[0106] In a preferred embodiment of the present invention, the multidimensional relationship map includes four types of relationships: spatial relationship, functional relationship, process relationship and organizational relationship.

[0107] Spatial relationships describe the relative positions of components in three-dimensional space, including adjacency (components are directly connected in space), containment (components contain or are contained by each other), and distance (the spatial distance between components). These relationships are preferably extracted from the BIM model. For example, the distance between two components can be calculated by calculating the Euclidean distance between their center points:

[0108] ,

[0109] in, Representation component and components The distance between and Respectively represent the center point coordinates of the two components.

[0110] Functional relationships describe the functional connections between components in a building. These include support relationships (one component supports another), connection relationships (how components are connected), and synergy relationships (how components work together to achieve a function). These relationships typically require a combination of BIM models and professional knowledge. For example, when identifying support relationships, one can examine the relative position and type of components. Vertical components (such as columns) often support horizontal components (such as beams).

[0111] Process relationships describe the temporal dependencies of components during construction, including precedence (installation order dependencies), synchronization (components that must be installed simultaneously), and mutual exclusion (components that cannot be installed simultaneously). These relationships are crucial for accurately calculating construction costs and durations. Preferably, the present invention automatically infers process relationships based on component type and building structure knowledge. For example, a support column must be installed before the beam it supports.

[0112] Organizational relationships describe the management and classification of components, including type affiliation (the type system to which the component belongs), system affiliation (the subsystem to which the component belongs), and hierarchy affiliation (the component's position in the building hierarchy). These relationships facilitate system-level cost aggregation and analysis.

[0113] The steps for constructing a relationship graph include: first, initializing component nodes based on component identification and feature extraction results; then analyzing the relationship descriptions in the text to establish explicit relationships; then inferring implicit relationships based on component location, function and other features; then applying transfer rules to expand the relationship network; and finally verifying relationship consistency and resolving conflicting relationships.

[0114] To reflect the differences in the importance of relationships, this paper introduces a relationship strength calculation and hierarchical division mechanism. Relationship strength is determined based on three factors: the strength of textual evidence (directly stated relationships receive high strength, while inferred relationships receive lower strength), relationship importance (safety-critical relationships receive high strength, while auxiliary relationships receive lower strength), and relationship reliability (relationships supported by multiple textual evidence receive increased strength, while relationships with contradictory evidence receive reduced strength). Based on relationship strength, this paper divides relationships into three levels: first-level relationships (core relationships with direct connection, force, or functional dependence), second-level relationships (indirectly related but significantly influential relationships), and third-level relationships (auxiliary, alternative, or distant relationships).

[0115] In addition, the present invention constructs a component family tree and dependency chain to better understand the hierarchical structure and dependency relationships between components. The component family tree uses the main structural component as the root node, builds a trunk based on support and inclusion relationships, and adds branches based on connection and collaboration relationships to form a complete component hierarchy. The dependency chain identifies each component's immediate predecessor and subsequent dependent components, constructs a complete dependency graph, marks critical paths, detects circular dependencies, and optimizes the dependency chain to reduce unnecessary dependencies.

[0116] Calculate the actual cost and budget cost of components based on multi-dimensional feature vectors and multi-dimensional relationship graphs:

[0117] After obtaining the component features and relationships, the present invention further calculates the actual cost and budgeted cost of the component. This step is the core of the entire method, and its goal is to accurately calculate the cost of each component based on comprehensive feature and relationship information.

[0118] In a preferred embodiment of the present invention, the cost calculation adopts a multi-level cost decomposition system to decompose the cost into component-level cost, subsystem-level cost and project-level cost.

[0119] Component-level costs are further broken down into material costs, processing costs, transportation costs, and installation costs. Material costs are calculated based on material type, grade, and usage; processing costs are based on the processing technology and complexity; transportation costs are based on distance, weight, and special requirements; and installation costs are based on the installation method and difficulty.

[0120] Taking prefabricated wall panels as an example, the material cost can be calculated using the following formula:

[0121] ,

[0122] in, represents the material cost, represents the volume of concrete, represents the density of concrete, Indicates the unit price of concrete, represents the cross-sectional area of ​​steel bars, Indicates the unit price of steel bars, Indicates the density of steel bars. The processing cost can be calculated by the following formula:

[0123] ,

[0124] in, represents the processing cost, represents the basic processing cost, and is the weight coefficient, represents the complexity coefficient, represents the special process coefficient. Preferably, The value is 0.3, The value is 0.5, which is an empirical value obtained based on the analysis of a large amount of actual engineering data. The transportation cost can be calculated using the following formula:

[0125] ,

[0126] in, represents the transportation cost, Indicates the weight of the component, Indicates the transport distance, Indicates the unit transportation price, is the weight coefficient, represents the special transport coefficient. Preferably, The value is 0.4, which indicates the degree of impact of special transportation requirements (such as extra length, extra width, and extra height) on transportation costs. The installation cost can be calculated using the following formula:

[0127] ,

[0128] in, represents the installation cost, Represents labor cost, represents the equipment cost, Represents the cost of auxiliary materials.

[0129] The total cost of a component is the sum of all costs:

[0130] ,

[0131] Subsystem-level costs are calculated by aggregating the costs of related components. These can be grouped by functional system (e.g., structural system, enclosure system), spatial system (e.g., floor, area), or process system (e.g., foundation engineering, main structure engineering). Project-level costs further include direct engineering costs, indirect costs, and taxes.

[0132] During the cost calculation process, this invention considers a series of refined factors, including component production batch (scale effect), component standardization (customization cost), component technical difficulty (special process cost), component transportation conditions (special transportation requirements), and component installation environment (special conditions such as high altitude and confined spaces). These factors are incorporated into the cost calculation formula through adjustment coefficients, significantly improving calculation accuracy.

[0133] To further optimize cost calculations, this invention introduces a correlation and sensitivity analysis system. Correlation analysis includes intra-component correlation (the cost impact relationship between different features of the same component) and inter-component correlation (the coordinated pattern of cost changes among different components). Sensitivity analysis varies key cost-influencing factors, observes the magnitude and trend of cost changes, calculates sensitivity indicators, and identifies the most sensitive factors. These analysis results are used to optimize the cost calculation model, improve calculation accuracy, and provide a basis for subsequent deviation analysis.

[0134] According to the difference between actual cost and budget cost, a multi-level deviation calculation system is established: Figure 5 After calculating the actual and budgeted component costs, the present invention constructs a multi-level deviation calculation system to systematically analyze cost discrepancies. This step is crucial for pinpointing the source and impact of cost deviations, providing a foundation for subsequent early warning.

[0135] In a preferred embodiment of the present invention, the deviation calculation system includes four levels: single feature deviation, component level deviation, component group deviation and system level deviation.

[0136] The most basic form of deviation is the single feature deviation, which calculates the difference between the actual value of a single feature and the budgeted value and normalizes it into a percentage. For numerical features, the deviation calculation formula is:

[0137]

[0138] in, represents the characteristic deviation percentage, represents the actual value of the feature, Represents the feature budget value.

[0139] For categorized features, this method determines a deviation based on the cost impact of each classification. For example, changing the concrete strength grade from C30 to C35 may result in an approximately 8%-12% increase in material costs. For missing features, a missing penalty is assigned based on their importance; missing important features will result in a higher deviation.

[0140] Component-level deviation comprehensively considers the deviations of all component features and performs weighted calculation based on feature weights. The calculation formula is:

[0141] ,

[0142] in, represents the component-level deviation, represents the weight of the i-th feature, represents the deviation of the i-th feature, and n represents the total number of features. Feature weights are preferably determined based on their impact on cost. For example, for prefabricated wall panels, the weights of material type and strength grade are typically between 0.3 and 0.4, while the weight of surface treatment may be between 0.1 and 0.2.

[0143] Component group deviation is based on the component relationship diagram, identifying functionally related component groups and calculating the overall deviation of the component group. Component groups can be structurally related components (such as a column and its connected beams), functionally related components (such as the components of an exterior wall insulation system), or constructionally related components (such as all components in the same construction section). The formula for calculating component group deviation is:

[0144] ,

[0145] in, represents the component group deviation, Indicates the The cost of each component, Indicates the Deviation of components, This weighted average approach ensures that components with higher costs have a greater impact on the deviation of the component group.

[0146] System-level deviation calculates the overall deviation of the system by functional system, spatial area, and other dimensions, and analyzes the distribution pattern of deviation within the system. A system can include the building's main structure, enclosure system, equipment system, etc., or it can be a specific floor, area, or functional space. The formula for calculating system-level deviation is similar to that for component group deviation, but it considers a wider range.

[0147] In addition to calculating deviation values, the present invention also analyzes deviation distribution patterns to identify concentrated areas and propagation paths. For example, if deviations are found to be generally high across the exterior wall system, primarily concentrated in material costs, this could be due to price fluctuations or changes in exterior wall material specifications. Alternatively, if deviations in one component are found to cause deviations in multiple related components, the propagation path of the deviation can be identified, allowing timely measures to control the spread.

[0148] Based on the results of the difference analysis, a hierarchical early warning mechanism is established: Figure 6 After completing the deviation calculation, the present invention further establishes a hierarchical early warning mechanism to generate early warning information when the deviation exceeds a preset threshold. This step is the end point of the entire method and is also the key to achieving intelligent cost monitoring.

[0149] In a preferred embodiment of the present invention, the early warning mechanism sets differentiated early warning thresholds based on three factors: component importance, project phase, and deviation type.

[0150] Based on the importance of components, the present invention classifies components into key components, important components, and general components, and sets different warning thresholds. Key components are those that have a significant impact on the safety and functionality of the building, such as major load-bearing components and key nodes. Their warning threshold is relatively low, typically 3%. Important components have a certain impact on building performance but do not affect basic safety, such as ordinary beams and columns, exterior wall panels, etc., and their warning threshold is 5%. General components have less impact on the overall building, such as decorative components and non-load-bearing partitions, and their warning threshold is higher, at 8%.

[0151] The present invention dynamically adjusts the strictness of the warning threshold based on the project stage. During the preliminary design phase, due to the incompleteness of the plan and frequent changes, the warning threshold is relatively relaxed and can be increased by 30%. During the construction drawing design phase, the plan is basically finalized, and the warning threshold remains at the standard level. During the construction implementation phase, changes will have a significant impact on costs and schedules, so the warning threshold is relatively strict and can be lowered by 20%.

[0152] Based on the deviation type, this invention distinguishes between cost-increasing and cost-decreasing deviations and adopts an asymmetric early warning strategy. For cost-increasing deviations, as they may lead to budget overruns, the early warning threshold is stricter. For cost-decreasing deviations, while not leading to overruns, they may indicate quality issues or specification degradation, so the early warning threshold is more relaxed, but monitoring is still required.

[0153] Warning classification is an important part of the warning mechanism. The present invention divides warnings into four levels:

[0154] Level 1 Warning (Minor): A single non-critical component exceeds the threshold, and the impact range is small. Usually, only attention is required and no immediate processing is required.

[0155] Level 2 warning (moderate): Multiple related components exceed the threshold or a single key component exceeds the threshold. The impact is limited and needs to be resolved in the near future.

[0156] Level 3 warning (serious): The system-level deviation exceeds the threshold or the key component group exceeds the threshold across the board, with a wide impact and requiring immediate attention.

[0157] Level 4 Warning (Critical): The total project cost deviation exceeds the threshold and multiple systems have serious deviations, affecting the entire project and requiring immediate suspension of work.

[0158] In order to effectively convey warning information to relevant personnel, the present invention designs an intelligent warning push strategy, including role-based differentiated push, urgency-based push and intelligent aggregate push.

[0159] Role-based, differentiated push notifications address the information needs of different users: project managers require a global deviation overview and critical alerts; technicians require detailed information on specific component and feature deviations; and procurement personnel require information on material and equipment deviations. Urgency-based push notifications determine the timing and method of notifications based on the alert level: critical alerts are instantly pushed through multiple channels; severe alerts are pushed within the business day; and moderate / minor alerts are aggregated and pushed regularly. Intelligent aggregate push notifications make alert information more systematic and easier to understand by aggregating related alerts, categorizing them by root cause, and providing trend analysis and forecasts.

[0160] Furthermore, in a preferred embodiment of the present invention, the early warning mechanism also includes intelligent root cause analysis. This function locates the component or system with the largest deviation, traces the characteristics or parameters that cause the deviation, analyzes the causes of characteristic value changes (such as design changes and market fluctuations), assesses the deviation's propagation path and impact, and ultimately generates a structured root cause report. Based on the root cause analysis results, the system also provides targeted cost optimization recommendations to reduce or eliminate deviations.

[0161] For example, if the cost of a cladding panel exceeds the budget by 15%, root cause analysis may reveal that a design change increased the panel thickness from 120mm to 150mm, increasing material usage by 25%. This increase in thickness also increases transportation and installation costs. The system then provides optimization suggestions, such as using stronger concrete to reduce panel thickness or adjusting the construction method to reduce weight, thereby reducing costs.

[0162] Component-level cost intelligent verification system for prefabricated buildings: Reference Figure 7 The present invention also provides a component-level intelligent cost verification system for prefabricated buildings, which includes five main modules: a text analysis and component recognition module 1, a feature extraction and matching module 2, a component relationship modeling module 3, a cost calculation and analysis module 4, and a deviation detection and early warning module 5.

[0163] The text analysis and component identification module 1 is responsible for acquiring cost documents and component information for prefabricated building projects. It then performs text analysis on the cost documents based on a deep learning model to identify component descriptions and cost data. This module includes a data acquisition unit 11, a text segmentation unit 12, a professional dictionary unit 13, and a component identification unit 14.

[0164] The feature extraction and matching module 2 is responsible for extracting the multi-dimensional feature vector of the component, including the basic feature dimension, material feature dimension, process feature dimension and installation feature dimension. This module includes a feature extraction unit 21, a feature matching unit 22 and a feature correlation analysis unit 23.

[0165] The component relationship modeling module 3 is responsible for constructing a multidimensional relationship map that reflects the relationship between components, including spatial relationships, functional relationships, process relationships, and organizational relationships. This module includes a relationship identification unit 31, a relationship map construction unit 32, and a relationship strength calculation unit 33.

[0166] The cost calculation and analysis module 4 is responsible for calculating the actual cost and budget cost of the component based on the multi-dimensional feature vector and multi-dimensional relationship map. This module includes a cost decomposition unit, a cost calculation unit, and a cost analysis unit.

[0167] Deviation Detection and Warning Module 5 is responsible for building a multi-level deviation calculation system based on the difference between the actual cost and the budgeted cost, performing variance analysis, and establishing a hierarchical warning mechanism based on the variance analysis results. It generates warning information when the deviation exceeds the preset threshold. This module includes a deviation calculation unit, a warning threshold management unit, a warning generation unit, and a root cause analysis unit.

[0168] Each module interacts with each other through standardized data interfaces to ensure the consistency and integrity of data flow. The system also provides external interfaces with BIM systems, cost management systems, and project management systems to achieve two-way data synchronization.

[0169] Preferably, the system adopts a distributed architecture, where each module can be deployed independently or in an integrated manner. The core algorithms and models of the system are stored in the central server, while data processing and calculations can be performed at the edge nodes, reducing the amount of data transmission and improving the system response speed.

[0170] The system provides a user-friendly interface, including a cost deviation visualization dashboard, interactive component relationship diagram browsing, an early warning message center, a root cause analysis report generator, and a cost optimization suggestion system. These interfaces allow users to intuitively understand project cost status and promptly identify and address cost deviations.

[0171] The component-level intelligent cost verification method for prefabricated buildings and the system thereof of the present invention have high industrial applicability and broad application prospects.

[0172] First, this invention addresses key technical challenges in prefabricated building cost verification, including low component recognition accuracy, incomplete feature extraction, insufficient component relationship modeling, poor cost calculation accuracy, and untimely deviation warnings. By integrating deep learning technology with professional cost knowledge, this invention significantly improves the accuracy and efficiency of cost verification.

[0173] Secondly, the present invention is applicable to all stages of the prefabricated building lifecycle, including the design, procurement, construction, and delivery and acceptance stages. During the design phase, the present invention can provide cost feedback to guide the optimization of design solutions; during the procurement phase, it can assist in material and component procurement decisions and optimize procurement plans; during the construction phase, it can monitor cost changes in real time and adjust construction plans in a timely manner; and during the delivery and acceptance phase, it can verify the final cost's compliance with the budget, ensuring project quality.

[0174] Finally, this invention contributes to the standardization, digitization, and intelligent development of the prefabricated building industry. Standardized component descriptions and cost calculations promote the formation of industry standards; digital cost management tools accelerate the industry's digital transformation; and intelligent analysis and early warning systems enhance management capabilities and lower industry barriers to entry.

[0175] In summary, the component-level intelligent cost verification method and system for prefabricated buildings of the present invention achieve efficient and accurate cost verification by deeply integrating artificial intelligence technology with professional knowledge of prefabricated buildings, providing strong support for cost control and quality management of prefabricated buildings, and has important theoretical and practical value.

[0176] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. The intelligent component-level cost verification method for prefabricated buildings is characterized by: include: Obtain cost documents and component information for prefabricated building projects; Performing text analysis on the cost document based on a deep learning model to identify component descriptions and cost data; Extracting a multi-dimensional feature vector of the component based on the recognition result, wherein the multi-dimensional feature vector includes a basic feature dimension, a material feature dimension, a process feature dimension, and an installation feature dimension; Constructing a multidimensional relationship map reflecting the relationship between components, wherein the multidimensional relationship map includes spatial relationship, functional relationship, process relationship and organizational relationship; Calculating the actual cost and budget cost of the component based on the multi-dimensional feature vector and the multi-dimensional relationship map; Based on the difference between the actual cost and the budgeted cost, a multi-level deviation calculation system is constructed to perform difference analysis; Based on the difference analysis results, a hierarchical early warning mechanism is established to generate early warning information when the deviation exceeds a preset threshold; The hierarchical early warning mechanism sets differentiated early warning thresholds according to component importance, project stage and deviation type.

2. The method according to claim 1, characterized in that The acquisition of cost documents and component information of the prefabricated building project includes: Obtain basic information of prefabricated building projects from the project management system; Obtain cost budget files from the cost management system; Obtain geometric information and topological relationships of components from the BIM system; The basic information, the cost budget file and the geometric information are integrated into structured data for subsequent analysis and processing.

3. The method according to claim 1, characterized in that The text analysis of the cost document based on the deep learning model includes: Use multi-level text segmentation processing to divide the cost document into project information paragraphs, component description paragraphs and cost data paragraphs; Enhance the system's ability to understand prefabricated building terminology using specialized dictionaries, including component type dictionaries, material dictionaries, and process dictionaries; Applying a component recognition and positioning system to the segmented text paragraphs to determine the boundaries of component descriptions and extract attribute information; The identified components are associated with the corresponding cost information to form a component-cost mapping relationship.

4. The method according to claim 1, wherein The multi-dimensional feature vector of the extraction component includes: Extract basic feature dimension information, including geometric features, physical features and quantitative features of components; Extract material characteristic dimension information, including material type, material grade and material properties; Extract process feature dimension information, including production process, connection process and surface treatment; Extract installation feature dimension information, including installation method, installation sequence and auxiliary facilities; The extracted features are standardized and features of different measurement units are converted into a unified standard.

5. The method according to claim 1, wherein The construction of a multi-dimensional relationship map reflecting the relationship between components includes: Initialize component nodes based on component identification results; Analyze the relationship descriptions in the cost documents and establish explicit relationships between components; Infer the implicit relationship between components based on their positions and functional characteristics; Apply relationship transfer rules to expand indirect relationships between components; Calculate the relationship strength and divide it into levels according to the importance of the relationship.

6. The method according to claim 1, characterized in that The calculating the actual cost and the budgeted cost of the component based on the multi-dimensional feature vector and the multi-dimensional relationship graph includes: Establish a multi-level cost decomposition system to decompose the cost into component-level cost, subsystem-level cost and project-level cost; Select an appropriate cost calculation model based on component characteristics; Consider detailed factors such as component production batch, degree of standardization, technical difficulty, transportation conditions and installation environment; Optimize the cost calculation path based on the component relationship diagram to ensure the cost calculation sequence of related components is reasonable; According to the correlation and sensitivity analysis results between features, adjust the weight of the impact of features on construction cost.

7. The method according to claim 1, characterized in that According to the difference between the actual cost and the budgeted cost, a multi-level deviation calculation system is constructed, including: Calculate the deviation value of each feature and standardize it into a percentage; Calculate component-level comprehensive deviation based on feature weights; Based on the component relationship graph, identify functionally related component groups and calculate component group deviations; Calculate system-level deviations by functional system, spatial region, and other dimensions; Analyze the deviation distribution pattern and identify the deviation concentration areas and propagation paths.

8. The method according to claim 1, characterized in that The step of establishing a hierarchical early warning mechanism based on the difference analysis results includes: Set differentiated warning thresholds for key components, important components, and general components; Dynamically adjust the stringency of warning thresholds based on the project stage; Distinguish between cost-increasing deviations and cost-reducing deviations and adopt asymmetric early warning strategies; Establish early warning classification standards, and classify early warning into level one, level two, level three, and level four; Design differentiated warning push strategies and perform intelligent push based on user roles, warning urgency, and warning relevance.

9. The method according to claim 8, characterized in that The establishment of a hierarchical early warning mechanism also includes: Perform intelligent root cause analysis to locate the source and impact of deviations; Trace back feature changes and parameter adjustments that caused the deviation; Assess the pathways of bias transmission and potential impacts; Generate structured root cause analysis reports; Provide targeted cost optimization suggestions to reduce or eliminate deviations.

10. The intelligent component-level cost verification system for prefabricated buildings is characterized by: include: A text analysis and component identification module is used to obtain cost documents and component information of prefabricated building projects, perform text analysis on the cost documents based on a deep learning model, and identify component descriptions and cost data; A feature extraction and matching module is used to extract a multi-dimensional feature vector of a component, wherein the multi-dimensional feature vector includes a basic feature dimension, a material feature dimension, a process feature dimension, and an installation feature dimension; A component relationship modeling module is used to construct a multi-dimensional relationship map reflecting the relationship between components, including spatial relationship, functional relationship, process relationship and organizational relationship; A cost calculation and analysis module, configured to calculate the actual cost and budgeted cost of a component based on the multi-dimensional feature vector and the multi-dimensional relationship graph; A deviation detection and early warning module is used to construct a multi-level deviation calculation system based on the difference between the actual cost and the budgeted cost, perform difference analysis, and establish a hierarchical early warning mechanism based on the difference analysis results to generate early warning information when the deviation exceeds a preset threshold; The hierarchical early warning mechanism sets differentiated early warning thresholds according to component importance, project stage and deviation type.

Images