Information management method and system for project cost

Abstract

The invention relates to the technical field of project cost management, in particular to a project cost-oriented information management method and system, and the method comprises the following steps: carrying out multi-source data collection and standardization processing, carrying out the automatic calculation, logic verification and correction of a project amount based on an AI calculation amount engine and a rule engine, and generating a precise project amount list; calling a dynamic pricing model, carrying out deviation analysis on the actual cost and the plan cost based on a earned value analysis method and a machine learning model, predicting the cost hyper-branched risk, and carrying out graded early warning; constructing a multi-participant collaborative platform, and supporting change application submission, associated cost influence calculation, online examination and approval, problem tracking and progress synchronization; the whole-cycle data is classified and archived, an enterprise-level cost index library is generated based on historical project data, cost reference is provided for a new project, and the method is suitable for cost information efficient management and cost control of the whole life cycle of constructional engineering, municipal engineering, installation engineering and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering cost management, and specifically to an information management method and system for engineering cost. Background Art

[0002] Construction cost management is a core component of project management, encompassing the entire process from investment estimation, design budget, construction drawing budget, and final settlement. It requires the integration of multi-dimensional information, including design drawings, bills of quantities, material prices, progress data, and contract terms. Current construction cost information management faces the following technical bottlenecks: First, regarding data integration, construction cost data is dispersed across multiple parties, including designers, builders, supervisors, and owners, in heterogeneous data formats such as CAD drawings, Excel lists, BIM models, and ERP system data. The lack of unified data standards and integration mechanisms leads to severe information silos, hinders data sharing, delays in information transfer, and high rates of duplicate entry errors. Second, in the calculation and pricing process, traditional methods rely on manual verification of drawings, extraction of construction quantities, such as concrete volume and rebar length, and then calculation based on quotas or market prices. This process is time-consuming and labor-intensive, and prone to errors in quantity calculations due to human oversight, including missed items and double counting, which seriously impact the accuracy of cost results. Furthermore, in terms of cost control, traditional management approaches focus solely on periodic "results control," such as verifying the total price at settlement. They lack real-time monitoring and adjustment mechanisms for dynamic factors during construction, such as material price fluctuations, design changes, and schedule delays. This makes it difficult to promptly identify and address the risk of cost overruns. Finally, in terms of multi-party collaboration, there is a lack of a unified collaboration platform among stakeholders, including owners, designers, contractors, and consultants. Critical information, such as change orders, requires manual processing and signature, resulting in delayed feedback, severely impacting cost management efficiency. Furthermore, traditional methods lack systematic analysis of historical project data and the accumulation of knowledge, making it difficult to provide effective cost references for new projects. To address these issues, existing technologies urgently need improvement. Summary of the Invention

[0003] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide an information management method and system for engineering cost.

[0004] To achieve the above object, the present invention provides the following technical solution: an information management method for construction cost, comprising the following steps:

[0005] S1: Multi-source data collection and standardization processing, integrating all data from the design stage, construction stage, completion stage and external environment, and converting them into structured data that meets preset standards;

[0006] S2: Based on the AI ​​quantity calculation engine and rule engine, it automatically calculates, logically verifies and corrects the engineering quantities to generate an accurate bill of quantities;

[0007] S3: Call the dynamic pricing model, combine the real-time market price, quota database and adjustment parameters, calculate the sub-item project costs, measure project costs and other project costs in real time, and generate a phased cost report;

[0008] S4: Based on earned value analysis (EVM) and machine learning models, we analyze the deviation between actual and planned costs, predict cost overrun risks, and provide graded warnings.

[0009] S5: Build a multi-party collaboration platform that supports change application submission, associated cost impact calculation, online approval, issue tracking, and progress synchronization, making collaborative processes electronic and traceable;

[0010] S6: Classify and archive full-cycle data, generate an enterprise-level cost index library based on historical project data, and provide cost reference for new projects.

[0011] In some embodiments, the multi-source data in step S1 include: BIM models, CAD drawings, and bills of quantities in the design phase; progress records, material delivery ledgers, and change visas in the construction phase; settlement reports and acceptance records in the completion phase; and price information, policy documents, and similar project indicators in the external environment; the standardization processing includes: converting unstructured data (such as PDF drawings, paper change orders) into structured text through OCR recognition, removing redundant information through data cleaning, and classifying materials, processes, and cost types through unified coding rules (such as "Construction Project Bill of Quantities Pricing Specifications" GB50500).

[0012] In some embodiments, the AI ​​quantity calculation engine in step S2 is based on a convolutional neural network (CNN) model, and learns quantity extraction features through a training data set (including annotated BIM models, CAD drawings and corresponding true values ​​of engineering quantities); the rule engine is based on the IFC (Industrial Foundation Class) standard to verify the consistency of the quantity calculation logic (such as checking whether the engineering quantities of the same component in different professional models are repeatedly calculated), and outputs the corrected engineering quantity results.

[0013] In some embodiments, the dynamic pricing model in step S3 includes:

[0014] Dynamic price database: Use crawlers to obtain real-time market prices (such as steel and concrete unit prices) from building materials trading platforms and supplier websites, supporting manual review and automatic updates;

[0015] Quota library: stores the "Construction Project Consumption Quota" and supplementary quotas for each region, and supports user-defined adjustment of consumption coefficients;

[0016] Adjustment parameter module: Dynamically adjust the comprehensive unit price based on material transportation distance, construction difficulty coefficient, and tax rate policy (such as value-added tax rate). The calculation formula is: Comprehensive unit price = (material base price × transportation distance adjustment coefficient + labor / machinery consumption × labor / machinery unit price) × (1 + management fee rate + profit margin) × (1 + tax rate).

[0017] In some embodiments, the dynamic cost monitoring in step S4 includes:

[0018] Cost variance calculation: Using earned value analysis, calculate cost variance (CV = Budgeted Cost of Work Performed (BCWP) - Actual Cost of Work Performed (ACWP)) and schedule variance (SV = BCWP - Budgeted Cost of Work Planned (BCWS)).

[0019] Risk prediction: Based on historical project change data and material price fluctuation data, a prediction model is trained using a random forest algorithm to output the probability and amount of cost overruns within the next 30 days.

[0020] Early warning push: According to the deviation level (yellow warning: CV <-5%; red warning: CV <-10%), early warning information is pushed to the project leader through system messages, SMS or emails, and the specific deviation source is associated (such as a sub-project overspending caused by a material price increase).

[0021] In some embodiments, the multi-party collaboration platform in step S5 includes:

[0022] Permission management module: assign hierarchical permissions to owners, designers, contractors, and consultants (e.g., owners can view all data, while contractors can only submit change requests);

[0023] Change management process: The construction party submits a change application (with change drawings and a certification form attached). The system automatically links the original bill of quantities to calculate the impact of the change (e.g., the cost increase ΔC corresponding to an increase in concrete volume ΔV = ΔV × comprehensive unit price). After online approval by the owner and supervisor (supporting electronic signatures), the cost data is updated and synchronized with all participating parties.

[0024] Progress synchronization module: The construction party uploads progress photos / videos, and the system uses AI image recognition (such as target detection based on YOLOv5) to analyze the construction completion rate (such as the proportion of the main structure topping out) and automatically associates the pricing status of the corresponding sub-project (such as the unfinished part will not be priced for the time being).

[0025] In some embodiments, the data archiving and knowledge accumulation in step S6 includes:

[0026] Full-cycle data storage: Categorized and stored in the database by project stage (design-construction-completion), including structured data (cost ledger, early warning log) and unstructured data (BIM model, change order scans);

[0027] Index library generation: Through data mining technology, statistics are collected on the unit cost indicators of historical projects (such as 3,000 yuan / m2 for residential projects), the proportion indicators of sub-items (such as concrete accounting for 25% of the total cost), and the material consumption indicators (such as steel bars 120kg / m2). It supports filtering by region, structure type, and construction period.

[0028] To achieve the above-mentioned object, the present invention further provides the following technical solution: an information management system for project cost, used to implement the information management method for project cost, comprising:

[0029] Data layer: Relational databases (such as MySQL) and non-relational databases (such as MongoDB) that store multi-source structured / unstructured data, with RESTful API interfaces configured for connecting to design software (Revit), project management software (Primavera), and building materials trading platforms;

[0030] Processing layer: Integrates AI quantity calculation engine (including CNN model and IFC rule engine), dynamic pricing model (including price library, quota library, calculation engine), cost warning model (including EVM analysis module and random forest prediction module), and collaboration engine (including workflow engine and log module);

[0031] Application layer: Deployment of project management module (creating / importing projects), intelligent quantity calculation module (automatically generating lists), dynamic pricing module (generating reports in real time), cost monitoring module (deviation analysis and early warning push), and collaborative workbench (change management, issue tracking, and progress synchronization);

[0032] User layer: supports multi-terminal access from PC and mobile terminals (iOS / Android), and configures role permission management modules (such as differentiated interfaces for the owner and construction terminals).

[0033] In some embodiments, the CNN model of the AI ​​quantity calculation engine uses ResNet-50 as the backbone network, and is fine-tuned on a small sample engineering quantity annotation dataset (≥100 projects) through transfer learning technology, with the model output accuracy ≥98%; the IFC rule engine has a built-in quantity calculation logic rule library of the "Uniform Standard for the Application of Building Information Modeling" GB / T 51212-2016, and supports custom rule extensions.

[0034] In some embodiments, the price library module of the dynamic pricing model is configured with a web crawler tool (such as the Scrapy framework) to periodically (daily / weekly) crawl price data from building materials trading platforms (such as China Building Materials Network), and the crawling frequency is configurable; the workflow module of the collaborative engine supports drag-and-drop process design (such as the change approval process can customize the node order and approval roles).

[0035] Compared with the existing technology, the beneficial effects of the present invention are: data integration and sharing: through multi-source data collection and standardized processing, it breaks the information island, realizes the unified management of design, construction, and completion cycle data and real-time sharing among multiple parties, and reduces repeated entry and communication costs (tested, collaborative efficiency is improved by more than 50%).

[0036] Intelligent quantity calculation and pricing: AI image recognition is combined with a rule engine to achieve automatic calculation of engineering quantities (accuracy ≥ 98%), avoiding manual omissions / miscalculations; the dynamic pricing model links to market prices in real time, improving cost accounting efficiency (the traditional manual list compilation takes 3 days, but this system only takes 4 hours).

[0037] Dynamic cost control: A cost early warning mechanism based on EVM and machine learning can predict cost overrun risks (such as hidden costs caused by rising material prices) 7-15 days in advance, helping project teams to adjust procurement strategies or optimize construction plans in a timely manner to reduce overrun losses (case verification shows that the cost deviation rate can be controlled within ±3%).

[0038] Improved collaborative efficiency: The online collaboration module enables online change approval and issue tracking, shortening the process from the traditional 3-5 days to within 1 day. All operations are traceable, reducing the risk of disputes.

[0039] Details of one or more embodiments of the present application are presented in the following drawings and descriptions to make other features, purposes and advantages of the present application more concise and easy to understand, and the present application is fully described and understood through the embodiments of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a flow chart of the method of the present invention;

[0041] Figure 2 This is a system architecture diagram of the present invention;

[0042] Figure 3 This is a schematic diagram of the technical principle of the intelligent calculation engine of the present invention;

[0043] Figure 4 This is a schematic diagram of the dynamic cost warning logic of the present invention. DETAILED DESCRIPTION

[0044] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0045] In traditional construction cost information management, multi-source, heterogeneous data is stored in independent systems across the design, construction, and supervision departments, leading to format compatibility and semantic consistency issues during data integration. For example, BIM models in the design phase are encoded using the IFC standard, progress records in the construction phase are stored in Excel spreadsheets, and acceptance data at the completion stage is archived as PDF documents. Differences in field definitions between different data sources require extensive manual mapping and format conversion when interoperating across systems. Manual quantity calculations rely on cost engineers individually verifying component dimensions and material specifications in CAD drawings. This can easily lead to errors in rebar length calculations exceeding the allowable tolerance due to visual fatigue or experience differences, resulting in cumulative discrepancies between concrete usage and total rebar in the bill of quantities. Dynamic cost control lacks a real-time response mechanism to market price fluctuations. For example, if steel prices rise by more than the benchmark price in a single day due to market supply and demand fluctuations, the traditional pricing model cannot automatically update the comprehensive unit price of the associated sub-projects. Corrections must wait for the manual price adjustment approval process to complete, causing cost forecasts to lag behind actual expenditures. Collaboration among multiple parties relies on the offline delivery of paper change orders. The average cycle for change applications from submission by the construction party to approval by the owner is as long as 72 hours. If design parameters are modified during this period, the relevant bill of quantities and pricing data cannot be updated synchronously, resulting in version conflicts and duplicate calculations.

[0046] For example, in a high-rise commercial complex project, the designer's Revit model contained over 100,000 building components. The bill of quantities (BOQ) generated by the contractor using Glodon software did not fully match the BIM model's calculation rules for embedded components and pipeline intersections, resulting in the omission of the casing area volume in the concrete calculation. During the main structure construction phase, unexpected policy adjustments caused price fluctuations in commercial concrete supply. However, the cost management department failed to obtain the latest price data in a timely manner and continued to calculate progress payments based on the previous week's benchmark price, causing the actual cost of completed work that month to exceed the budget by 8%. Furthermore, the contractor's request for a curtain wall design change was delayed due to the supervisor's business trip. As a result, the revised keel specifications and the aluminum plate thickness parameters in the original BIM list were not synchronized with the cost system in a timely manner. This led to a discrepancy between the pricing basis and the construction records during the final settlement.

[0047] If the above problems are not resolved, the inefficiency of multi-source data integration will directly lead to a decrease in the confidence of key indicators in the cost management process. For example, the cumulative errors in the bill of quantities may lead to disputes over project payment and even trigger contract breach clauses. The lag of dynamic cost control makes the project lose its risk buffer ability when facing drastic fluctuations in market prices. The risk of overspending may spread from local material price differences to overall budget out of control. The asynchronous nature of the collaborative process not only increases communication costs, but also causes the chain reaction of engineering changes to not be accurately captured due to inconsistent information versions. For example, the design modification of a beam-column node may not be transmitted to the associated steel bar quantity calculation module in a timely manner, thereby affecting the structural safety verification results. The long-term existence of data silos will hinder the construction of an enterprise-level cost knowledge base. Key parameters in historical projects cannot be effectively extracted and reused, resulting in a lack of reliable benchmarks for cost forecasts of similar projects.

[0048] When faced with the above problems, this application first considers how to break the multi-source data silos and improve integration efficiency. The traditional method relies on manual mapping and format conversion, which is time-consuming and error-prone. To this end, this application proposes to establish a unified data collection and standardization processing mechanism, and realize the structured conversion of multi-format data through automation technology. In response to the problem of manual calculation errors, the traditional method relies on the experience of engineers and is prone to visual fatigue. This application explores the introduction of intelligent algorithms to replace manual verification, and improves the calculation accuracy through model training and rule verification. Faced with the lag in dynamic cost control, the traditional model cannot respond to price fluctuations in real time. This application studies the construction of a dynamic pricing model to automatically associate market data to update unit prices. In response to low collaborative efficiency, this application designs a multi-party collaborative platform to eliminate offline process delays, and realizes real-time linkage between change approval and data synchronization through electronic processes. In addition, the traditional method lacks a historical data precipitation mechanism. This application plans to establish a full-cycle data archive and indicator library, and extract reusable cost reference indicators through data mining technology.

[0049] In this regard, Figures 1 to 4 As shown, this application proposes an information management method and system for engineering cost.

[0050] It includes the following steps: multi-source data collection and standardized processing, integrating all data from the design stage, construction stage, completion stage and external environment, and converting them into structured data that meets preset standards; based on the AI ​​quantity calculation engine and rule engine, automatically calculate, logically verify and correct the engineering quantities to generate an accurate bill of quantities; call the dynamic pricing model, combine real-time market prices, quota database and adjustment parameters, calculate the sub-item engineering costs, measure project costs and other project costs in real time, and generate phased cost reports; based on earned value analysis and machine learning models, analyze the deviation between actual costs and planned costs, predict cost overrun risks and issue graded warnings; build a multi-party collaborative platform to support change application submission, associated cost impact calculation, online approval, problem tracking and progress synchronization, and realize the electronic and traceable collaborative process; classify and archive full-cycle data, generate an enterprise-level cost index library based on historical project data, and provide cost reference for new projects.

[0051] Among them, multi-source data collection and standardized processing refers to obtaining data from different sources from the design stage, construction stage, completion stage and external environment, and converting it into structured data in a unified format. Specifically, OCR recognition technology can be used to convert unstructured data into text, redundant information can be removed through data cleaning, and unified coding rules can be used for classification to solve the problems of data fragmentation and islands, and realize full-process data integration. Among them, AI quantity calculation engine and rule engine refer to the use of artificial intelligence models to automatically calculate engineering quantities and perform logical verification. Specifically, convolutional neural network models can be used to extract engineering quantity features, and the logical consistency of quantity calculation can be verified in combination with IFC standards to improve the efficiency and accuracy of quantity calculation and avoid errors caused by human omissions. Among them, dynamic pricing model refers to the calculation of engineering costs based on real-time market prices, quota database and adjustment parameters. Specifically, the real-time price of building materials can be obtained through web crawlers, and the comprehensive unit price can be dynamically adjusted in combination with the transportation distance adjustment coefficient and tax rate policy to cope with material price fluctuations and realize dynamic cost control. Among them, earned value analysis and machine learning models refer to predicting risks by comparing the deviation between actual costs and planned costs. Specifically, earned value analysis can be used to calculate cost and schedule deviations, and the random forest algorithm can be used to train historical data to predict the probability of overspending, so as to monitor cost risks in real time and issue graded warnings. Among them, the multi-party collaboration platform refers to a system that supports multi-party online collaboration in processing change applications, approvals, and progress synchronization. Specifically, role permissions can be assigned through the permission management module, and the workflow engine can be used to achieve electronic approval and data synchronization to solve the problem of low collaboration efficiency and improve the timeliness of information transmission. Among them, full-cycle data classification and archiving and cost index library generation refer to the structured storage of data from each stage of the project and the extraction of reference indicators. Specifically, data mining technology can be used to calculate indicators such as unit cost and proportion of sub-projects to accumulate historical project experience and provide cost reference basis for new projects.

[0052] The core innovation of this application lies in integrating multi-source data from the entire process and standardizing the processing, combining the AI ​​quantity calculation engine with the dynamic pricing model to realize automatic calculation of engineering quantities and real-time cost updates, while introducing machine learning models to predict cost deviation risks, building a multi-party collaborative platform to improve information sharing efficiency, and forming an intelligent management closed loop covering the entire life cycle of engineering costs.

[0053] The working process and principle of this application are as follows: multi-source data collection and standardized processing steps integrate the full amount of data on design, construction, completion and external environment, and convert unstructured data into structured data through data cleaning and unified coding rules. The AI ​​quantity calculation engine and rule engine automatically calculate, logically verify and correct the engineering quantities to generate an accurate bill of quantities. The dynamic pricing model combines real-time market prices, quota databases and adjustment parameters to calculate the costs of various projects in real time and generate phased cost reports. Earned value analysis and machine learning models analyze the deviation between actual costs and planned costs, predict the risk of cost overruns and issue graded warnings. The multi-party collaborative platform supports change application submission, associated cost impact calculation, online approval, problem tracking and progress synchronization, realizing the electronic and traceable collaborative process. Full-cycle data classification and archiving, based on historical project data, generate an enterprise-level cost index database to provide cost reference for new projects. Each step works together to achieve comprehensive management of engineering cost information.

[0054] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0055] The multi-source data collection and standardization process involves design phase data, including BIM models, CAD drawings, and bills of quantities; construction phase data, including progress records, material delivery records, and change certificates; completion phase data, including settlement reports and acceptance records; and external environmental data, including pricing information, policy documents, and indicators of similar projects. Unstructured data is converted into structured text using optical character recognition (OCR) technology, cleansed to remove redundant information, and categorized and coded using the "Construction Project Bill of Quantities Pricing Specification" (GB50500).

[0056] The AI ​​quantity calculation engine, based on a convolutional neural network model, learns quantity extraction features from a training dataset. The rules engine, based on the IFC standard, verifies the consistency of quantity calculation logic and outputs revised quantity results.

[0057] The dynamic pricing model includes a dynamic price library, a quota library, and an adjustment parameter module. The dynamic price library uses a crawler to obtain real-time market prices, the quota library stores consumption quotas, and the adjustment parameter module dynamically adjusts the comprehensive unit price based on material transportation distance, construction difficulty coefficient, and tax rate policies.

[0058] Earned value analysis calculates cost and schedule variances. A random forest algorithm trains a predictive model to output the probability and amount of future cost overruns. Early warning messages are issued based on the variance level, linking the specific source of the variance.

[0059] The multi-party collaboration platform includes a permissions management module, a change management process, and a progress synchronization module. Permission management assigns hierarchical permissions to each party. The change management process supports change application submission, impact calculation, and online approval. The progress synchronization module uses AI image recognition to analyze construction completion.

[0060] Full-cycle data is stored by project phase, including both structured and unstructured data. Data mining techniques are used to generate indicators such as unit cost, proportion of sub-projects, and material consumption, with support for filtering by region, structure type, and duration.

[0061] Through the above scheme, this application realizes the efficient integration of multi-source heterogeneous data, improves the accuracy of quantity calculation and pricing, realizes dynamic cost control, improves the efficiency of multi-party collaboration, and builds an enterprise-level cost knowledge base. Specifically, data standardization solves the problems of format compatibility and semantic consistency, the AI ​​quantity calculation engine reduces human errors, the dynamic pricing model realizes real-time response to market price fluctuations, the collaborative platform shortens the change approval cycle, and the data archiving and indicator library provide reliable cost references for new projects. As a result, this application effectively solves the problems of data fragmentation, low quantity calculation efficiency, lagging cost control, and low collaboration efficiency in traditional engineering cost information management, and improves the accuracy, real-time and collaboration of cost management.

[0062] In some of the above-mentioned solutions in this application, multi-source data collection and standardized processing require the integration of full data from the design stage, construction stage, completion stage and external environment. However, in traditional methods, the data formats of different stages vary greatly, and unstructured data is difficult to process directly, resulting in low data integration efficiency and redundant information interference.

[0063] This application further proposes that multi-source data include BIM models, CAD drawings, and bills of quantities in the design phase; progress records, material delivery ledgers, and change visas in the construction phase; settlement reports and acceptance records in the completion phase; and price information, policy documents, and similar project indicators in the external environment; standardized processing includes converting unstructured data into structured text through OCR recognition, removing redundant information through data cleaning, and classifying materials, processes, and cost types through unified coding rules.

[0064] Among them, design phase data covers BIM models and CAD drawings, construction phase data includes progress records and change certificates, completion phase data includes settlement reports and acceptance records, and external environment data involves price information and policy documents. Unstructured data is converted into structured text through OCR recognition. For example, PDF drawings and paper change orders are processed by OCR to generate editable text data. During the data cleaning process, redundant information is filtered out. For example, duplicate ledger entries are automatically removed. The unified coding rules adopt the "Construction Project Bill of Quantities Pricing Specification" GB50500 standard to classify and code material types, construction processes, and cost categories to ensure data classification consistency.

[0065] Specifically, BIM models and CAD drawings from the design phase are parsed to extract engineering quantity information. Paper change certificates from the construction phase are converted into structured text through OCR recognition and matched with electronic ledgers. The data cleaning module removes duplicate or invalid entries based on preset rules. For example, progress records with missing key fields are marked as abnormal data. Unified coding rules map material names from different sources to standard codes. For example, "rebar" and "rebar" are uniformly coded as "A001" to eliminate semantic ambiguity. Through the above processing, multi-source heterogeneous data is converted into structured data in a unified format, providing standardized input for subsequent engineering quantity calculations and cost analysis.

[0066] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0067] The multi-source data collection and standardization process involves integrating all data from the design, construction, and completion phases, as well as the external environment. Design phase data includes BIM models, CAD drawings, and bills of quantities. Construction phase data includes progress records, material delivery records, and change approvals. Completion phase data includes settlement reports and acceptance records. External environment data includes pricing information, policy documents, and similar project indicators.

[0068] Standardization begins by using OCR technology to identify unstructured data, such as PDF drawings and paper change orders, and converting them into structured text. Data cleaning is then performed to remove redundant information. Finally, materials, processes, and expense types are uniformly coded and categorized according to GB50500, the "Construction Project Bill of Quantities Pricing Specification."

[0069] For example, for construction drawings in PDF format, a deep learning-based OCR model is used for text recognition and graphic element extraction. The recognition results are cleaned to remove non-critical information such as frames and title bars. The identified component information (such as beams, columns, walls, etc.) is then coded according to the classification system in the GB50500 standard. For paper change orders, OCR recognition is also used to extract key fields such as change content and approval information, and then link them to the corresponding bill of quantities items.

[0070] Through the above technical solution, this application realizes the automated collection and standardized processing of multi-source heterogeneous data. It avoids errors and omissions caused by manual data entry and improves the integrity and accuracy of the data. The standardized data format facilitates subsequent automated analysis and processing, laying the foundation for accurate engineering quantity calculation and cost management. At the same time, through unified coding rules, the association and integration of data from different stages and different participants are achieved, solving the problems of data fragmentation and information islands in traditional methods.

[0071] In some of the above-mentioned solutions of this application, when automatically calculating engineering quantities based on the AI ​​quantity calculation engine and the rule engine, the traditional method relies on manual verification of drawings to extract engineering quantities, which leads to errors in engineering quantity calculation due to insufficient feature extraction accuracy, as well as repeated calculation of components between different professional models.

[0072] This application further proposes an AI quantity calculation engine based on a convolutional neural network model, which learns engineering quantity extraction features through a training data set; the rule engine is based on the IFC standard, verifies the consistency of the quantity calculation logic, and outputs the corrected engineering quantity results.

[0073] The convolutional neural network model uses ResNet-50 as its backbone network and is fine-tuned using transfer learning techniques on a small sample quantity annotation dataset. The training dataset includes annotated BIM models, CAD drawings, and the corresponding true quantity values. The model output accuracy exceeds 98%. The rule engine has a built-in quantity calculation logic rule library that complies with the "Unified Standard for the Application of Building Information Modeling" (GB / T 51212-2016). It uses the IFC standard to analyze component properties in different professional models. During logic verification, it automatically detects duplicate calculations of the same component in different models. For example, a duplicate calculation alarm is triggered when the spatial position of a beam component in a steel structure model overlaps with an embedded part in the electromechanical model.

[0074] Specifically, when processing the BIM model in the design phase, the convolutional neural network model first extracts the geometric features and material properties of the components and outputs the initial engineering quantity calculation results. The rule engine then verifies the calculation results in multiple dimensions: it parses the unique coding and spatial coordinate data of the components through the IFC standard, compares the spatial position relationship of the components in different professional models, and automatically eliminates redundant calculation items if it detects that the same component is repeatedly referenced by multiple models. For example, when a concrete column has the same IFC code in the architectural model and the structural model but belongs to different professional views, the system only retains one calculation record. The verified engineering quantity results are corrected through log records, and finally a bill of quantities that fully matches the design drawings and construction specifications is generated. This process controls the error of engineering quantity calculation within 2% through a dual verification mechanism of feature extraction and rule verification, effectively avoiding the common problems of omissions and repeated calculations in traditional manual quantity calculations.

[0075] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0076] The AI ​​quantity calculation engine, based on a convolutional neural network model, learns quantity extraction features from a training dataset consisting of annotated BIM models, CAD drawings, and the corresponding ground-truth quantities. The convolutional neural network model uses a ResNet-50 network as its backbone and is fine-tuned using transfer learning techniques on a small sample quantity annotation dataset, achieving output accuracy exceeding 98%.

[0077] The rules engine verifies the consistency of quantity calculation logic based on the IFC standard. It includes a built-in quantity calculation logic rule library based on the "Uniform Standard for the Application of Building Information Modeling," GB / T 51212-2016, and supports custom rule extensions. For example, it can check whether quantities for the same component in different professional models are double-counted and output the corrected quantity results.

[0078] During implementation, the BIM model or CAD drawing is first input into the AI ​​quantity calculation engine. Using a convolutional neural network, the AI ​​quantity calculation engine extracts component features from the drawing, identifies structural elements such as walls, beams, and columns, and calculates the corresponding engineering quantity data. The rules engine then performs logical verification on the AI ​​quantity calculation results. The rules engine checks for duplicate calculations, missing items, and other issues, automatically correcting them according to pre-set rules. Finally, a verified and corrected bill of quantities is output.

[0079] Through the above technical solution, this application achieves automated and accurate calculation of engineering quantities. The AI ​​quantity calculation engine improves the efficiency and accuracy of engineering quantity extraction and reduces manual operation errors. The rule engine further ensures the logical consistency of the quantity calculation results, avoiding common problems such as repeated calculations. As a result, this solution significantly improves the efficiency and accuracy of engineering quantity calculations in engineering cost management and provides a reliable data foundation for subsequent cost estimation.

[0080] In some of the above-mentioned solutions in this application, traditional cost management methods lack a real-time response mechanism to material price fluctuations and tax policy changes during the construction process, resulting in the calculation of comprehensive unit prices relying on static benchmark prices and fixed parameters, and being unable to dynamically reflect market changes and differences in construction conditions, causing the cost calculation results to deviate from the actual cost.

[0081] This application further proposes a dynamic pricing model including a dynamic price library, a quota library and an adjustment parameter module.

[0082] Among them, the dynamic price library configures web crawler tools to regularly capture steel and concrete unit price data from building materials trading platforms and suppliers' official websites, and supports automatic updates to the database after manual review; the quota library stores construction project consumption quotas and regional supplementary quotas, allowing users to adjust labor and machinery consumption coefficients according to project requirements; the adjustment parameter module calculates the transportation distance adjustment coefficient based on the material transportation distance, and dynamically calculates the comprehensive unit price through a formula based on the construction difficulty coefficient and the current value-added tax rate.

[0083] Specifically, the dynamic price database's crawler tool performs a daily data crawling task, comparing the acquired building materials market prices with historical database records. If price fluctuations exceed a preset threshold, a manual review process is triggered. Upon approval, the price benchmark data is updated. The quota database draws on machine-hour consumption data from regional supplementary quotas. Users adjust the coefficients through the user interface, and the system automatically updates the quota consumption for the corresponding sub-project. The adjustment parameter module receives the transport distance input and generates a transport distance adjustment coefficient based on a preset transport distance-coefficient mapping table. It then substitutes the construction difficulty coefficient and the current tax rate into the formula to calculate the comprehensive unit price in real time. For example, if the concrete transport distance increases to 10 kilometers, the transport distance adjustment coefficient increases from 1.0 to 1.2. The system automatically multiplies the adjusted price benchmark by the updated labor consumption and tax rate to generate a new comprehensive unit price, which is then synchronized to the cost report. This enables real-time response to material price fluctuations, policy changes, and variations in construction conditions, ensuring that cost calculations dynamically align with actual costs.

[0084] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0085] The dynamic pricing model includes a dynamic price library, a quota library and an adjustment parameter module.

[0086] The dynamic price database uses a crawler to obtain real-time market prices from building materials trading platforms and supplier websites. For example, it can retrieve steel prices daily from "My Steel Network" and concrete unit prices weekly from "China Concrete Network." The obtained price data is manually reviewed and automatically updated to the system.

[0087] The quota library stores the "Construction Project Consumption Quotas" and supplementary quotas for each region. Users can adjust consumption coefficients based on project specifics. For example, for a high-rise building project, the labor consumption coefficient for concrete work could be increased by 10%.

[0088] The parameter adjustment module dynamically adjusts the comprehensive unit price based on material transportation distance, construction difficulty coefficient, and tax rate policies. The calculation formula is: Comprehensive Unit Price = (Material Base Price × Transportation Adjustment Coefficient + Labor / Machinery Consumption × Labor / Machinery Unit Price) × (1 + Management Fee Rate + Profit Margin) × (1 + Tax Rate). The transportation adjustment coefficient can be set based on the distance between the project location and the material production site, such as 1.0 for within 100 kilometers and 1.05 for 100-500 kilometers. The construction difficulty coefficient can be set based on the project type, such as 1.0 for ordinary residential buildings and 1.2 for super-high-rise buildings. The tax rate is set according to the latest VAT policy, such as the 9% VAT rate for the construction industry.

[0089] Through the above technical solution, this application achieves dynamic and accurate calculation of project costs. The dynamic price database ensures the real-time and accuracy of material prices, avoiding the deviations caused by traditional fixed prices. The adjustability of the quota database makes pricing more tailored to the actual project situation. The parameter adjustment module takes into account multiple influencing factors, making the calculation of the comprehensive unit price more comprehensive and accurate. This dynamic pricing model significantly improves the accuracy and flexibility of project cost management and provides a reliable basis for project cost control.

[0090] In some of the above-mentioned solutions in this application, traditional methods lack a real-time monitoring and adjustment mechanism for dynamic factors such as material price fluctuations, design changes, and schedule delays during the construction process, resulting in the inability to timely predict and warn of cost overrun risks, affecting the effectiveness of cost control.

[0091] This application further proposes dynamic cost monitoring including cost deviation calculation, risk prediction and early warning push. The cost deviation calculation adopts the earned value analysis method, which quantifies the cost deviation by the difference between the budget cost of completed work and the actual cost of completed work, and quantifies the progress deviation by the difference between the budget cost of completed work and the budget cost of planned work. Risk prediction is based on the change data of historical projects and material price fluctuation data. The prediction model is constructed through the random forest algorithm to output the probability and amount of cost overruns in the next 30 days. The early warning push sets yellow and red warning thresholds according to the deviation level, pushes warning information to the project leader through system messages, SMS or emails, and associates the specific deviation source.

[0092] The cost deviation calculation module establishes a real-time data interface between BCWP and ACWP, automatically capturing the difference between actual and planned project cost data, and storing the deviation results numerically in a database. The risk prediction module utilizes a historical project database, storing change records and material price fluctuation data for at least 100 projects. A random forest model is trained using ten-fold cross-validation, outputting overrun probabilities with an accuracy of two decimal places. The early warning push module employs a two-level trigger mechanism: a yellow alert is triggered when the cost deviation rate exceeds 5%, and a red alert is triggered when it exceeds 10%. The alert information is automatically linked to the deviation source data table, extracting the specific sub-project code and material number that caused the overrun.

[0093] Specifically, during the construction phase, the system collects progress records and material delivery data in real time, generating a cost estimate for work completed through the BCWP calculation module. The ACWP calculation module simultaneously obtains actual payment data from the financial system, and the difference between the two is updated in real time to the monitoring interface using the CV formula. The random forest model uses the latest price fluctuation data and change records daily to predict the probability of overspending over the next 30 days. When the probability exceeds a preset threshold, an early warning process is triggered. Early warning information is pushed to the approval system via a message queue, allowing project managers to view details of associated deviation sources, such as increased costs for a sub-project due to rising steel unit prices. Early warning processing logs record the person responsible for the response time and handling measures, creating a closed-loop management system.

[0094] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0095] Dynamic cost monitoring includes three modules: cost deviation calculation, risk prediction and early warning push.

[0096] The Cost Variance Calculation module uses earned value analysis to calculate cost and schedule variances. Cost Variance (CV) is calculated by subtracting Actual Cost of Work Performed (ACWP) from Budgeted Cost of Work Performed (BCWP). Schedule Variance (SV) is calculated by subtracting Budgeted Cost of Work Planned (BCWS) from Budgeted Cost of Work Performed (BCWP).

[0097] The risk prediction module uses a random forest algorithm to train a predictive model based on historical project change data and material price fluctuations. This module outputs the probability and amount of cost overruns within the next 30 days. The random forest model consists of 100 decision trees, each with a maximum depth of 10 and a minimum leaf node size of 5. The model's input features include current cost deviation, schedule deviation, material price index, and change frequency.

[0098] The early warning push module sends warning messages to project managers based on the deviation level. A yellow warning is triggered when the cost deviation CV is less than -5%, and a red warning is triggered when the CV is less than -10%. Warning messages are sent via system message, SMS, or email, and are linked to the specific source of the deviation, such as a subproject overrun caused by a material price increase. Warning messages include the project name, deviation type, deviation amount, cause of the deviation, and recommended actions.

[0099] Through the above technical solutions, this application realizes dynamic monitoring of project costs and risk warning. The cost deviation calculation module provides real-time and accurate cost status assessment. The risk prediction module uses historical data and machine learning algorithms to quantitatively predict future cost risks. The early warning push module promptly transmits risk information to relevant personnel and provides specific deviation source analysis. This dynamic and intelligent cost monitoring method significantly improves the timeliness and accuracy of project cost management, helping project managers to quickly identify and respond to cost risks, thereby better controlling the total project cost.

[0100] In some of the above-mentioned solutions in this application, multiple parties lack a unified collaborative platform in the traditional engineering cost management process. Change applications need to be signed through manual circulation, and information feedback is delayed, resulting in low efficiency in change impact calculation and approval. At the same time, the construction progress and pricing status cannot be linked in real time, affecting the timely update of cost data and the efficiency of multi-party collaboration.

[0101] This application further proposes to build a multi-party collaborative platform, including a permission management module, a change management process, and a progress synchronization module.

[0102] Among them, the authority management module assigns hierarchical permissions to the owner, designer, contractor, and consultant. For example, the owner can view all data, while the contractor can only submit change applications. The change management process allows the contractor to submit change applications and attach change drawings and visa forms. The system automatically associates the original bill of quantities to calculate the impact of the change. For example, the cost increase corresponding to the increase in concrete usage is equal to the change in usage multiplied by the comprehensive unit price. After online approval by the owner and supervisor, the cost data is updated and synchronized to all participants. The progress synchronization module allows the contractor to upload progress photos or videos. The system uses YOLOv5-based target detection technology to analyze the completion of construction, such as the proportion of the main structure topping out, and automatically associates the pricing status of the corresponding sub-items. The unfinished parts are not priced for the time being.

[0103] Specifically, the authority management module ensures data security and operational compliance through role-based authority division, avoiding information leakage or misoperation caused by unauthorized access. In the change management process, the system automatically calculates the impact of the change, eliminates manual calculation errors, and implements online approval through electronic signatures, shortening the process cycle. For example, after a certain concrete usage change application is submitted, the system automatically extracts the comprehensive unit price in the original list, generates a cost increment based on the change amount, and updates the cost report in real time after approval. The progress synchronization module uses AI image recognition technology to analyze the uploaded construction images and identify the completion status of key nodes. For example, it determines the completion degree of the concrete structure by detecting the status of formwork removal, and then triggers the pricing status update of the corresponding sub-project. As a result, the construction progress and cost data are linked in real time to avoid pricing deviations caused by progress delays, while reducing the workload of manual on-site verification.

[0104] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0105] The multi-party collaboration platform includes a rights management module, a change management process, and a progress synchronization module.

[0106] The permissions management module assigns hierarchical permissions to owners, designers, contractors, and consultants. For example, owners can view all data, while contractors can only submit change requests.

[0107] The change management process includes the following steps: the construction party submits a change request, attaching the change drawings and a certification form. The system automatically links the change to the original bill of quantities and calculates the impact of the change. For example, an increase in concrete volume ΔV corresponds to a cost increase ΔC = ΔV × comprehensive unit price. After online approval by the owner and supervisor, the cost data is updated and synchronized with all parties involved. Online approval supports electronic signatures.

[0108] The progress synchronization module works as follows: the construction party uploads progress photos or videos. The system uses AI image recognition to analyze construction completion, such as the percentage of the main structure topping out. The YOLOv5-based object detection algorithm is used for image recognition. The system automatically associates the pricing status of the corresponding sub-projects, temporarily excluding any unfinished parts from being priced.

[0109] Through the above technical solutions, this application achieves efficient collaboration among multiple parties involved in project cost management. Permission tiering ensures data security, the change management process enables automatic calculation and online approval of change impacts, and the progress synchronization module uses AI technology to automatically link construction progress and pricing status. These features collectively improve information transmission efficiency, reduce manual errors, and enhance the real-time and accuracy of cost management.

[0110] In some of the above-mentioned solutions of this application, historical engineering cost data are stored in different stages and systems in a scattered manner, and there is a lack of a unified classification and archiving mechanism, resulting in low data utilization, inability to effectively extract reusable cost indicators, and lack of reliable reference basis for estimating the cost of new projects.

[0111] This application further proposes to classify and archive full-cycle data, generate an enterprise-level cost index library based on historical project data, and provide cost reference for new projects.

[0112] Full-cycle data storage is achieved by categorizing and storing it in a database by project phase, encompassing both structured and unstructured data. Structured data includes cost records and early warning logs, while unstructured data includes BIM models and scanned copies of change orders. The indicator library utilizes data mining techniques to compile historical project metrics, including unit cost, percentage of sub-projects, and material consumption, with support for filtering by region, structure type, and construction period.

[0113] Specifically, throughout the project lifecycle, structured data from the design phase is automatically synchronized to a relational database via a pre-set interface, while unstructured data generated during the construction phase is scanned and uploaded to a non-relational database. During the completion phase, a data cleansing module verifies the correlation of data across phases to form a complete project dataset. The indicator generation module uses a clustering algorithm to extract features from historical projects. For example, the unit cost indicator for residential projects is calculated by dividing the total cost by the built area, and the proportion indicator for sub-items is calculated by dividing the cost of each sub-item by the total cost. Material consumption indicators are calculated by calculating the amount of steel used per unit area. When users create a new project and enter project attribute parameters, the system automatically matches cost indicators from similar historical projects. For example, if a user enters "frame structure + East China region + 12-month construction period," the system will return a reference range of 25% ± 3% for the proportion of concrete work in this category, assisting in the preparation of budget estimates.

[0114] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0115] Data archiving and knowledge accumulation include two parts: full-cycle data storage and indicator library generation.

[0116] Data storage throughout the project lifecycle is categorized and stored in a database by project phase. Design phase data includes BIM models, CAD drawings, and bills of quantities; construction phase data includes progress records, material delivery records, and change certificates; and completion phase data includes settlement reports and acceptance records. This data is divided into structured and unstructured data. Structured data, such as cost records and early warning logs, is stored in a MySQL relational database, while unstructured data, such as BIM models and scanned copies of change orders, is stored in a MongoDB non-relational database.

[0117] The indicator library is generated through data mining techniques, which statistically analyze historical project data. First, the unit cost is calculated, such as 3,000 yuan / ㎡ for a residential project. Second, the proportion of sub-projects to the total cost is analyzed, such as concrete accounting for 25%. Third, material consumption is calculated, such as 120 kg / ㎡ for steel bars. Finally, a multi-dimensional filtering mechanism is established to support the selection of indicators by region, structural type, construction period, and other criteria.

[0118] Through the above technical solutions, this application achieves the systematic archiving and knowledge accumulation of data throughout the entire construction cost cycle. Full-cycle data storage improves the integrity and traceability of data, facilitating subsequent analysis and auditing. The generation of an indicator library provides an accurate cost reference for new projects, helping to improve the accuracy and efficiency of cost forecasting. The multi-dimensional screening mechanism enhances the applicability of indicators, enabling cost estimators to quickly obtain historical data that matches the current project. This data-driven approach significantly improves the scientific nature and accuracy of cost management.

[0119] In some of the aforementioned solutions, the construction cost information management system suffers from low data integration efficiency, insufficient quantity and price calculation accuracy, weak dynamic cost control capabilities, and difficulty in multi-party collaboration. Traditional methods rely on manual operations and decentralized tools, resulting in data silos, information lags, high duplication error rates, difficulty in real-time monitoring of cost fluctuations and the impact of changes, and a lack of a unified collaborative platform to support efficient collaboration among multiple stakeholders.

[0120] This application further proposes an information management system for engineering cost, including a data layer, a processing layer, an application layer, and a user layer. The data layer stores multi-source structured and unstructured data through relational and non-relational databases, and configures a RESTful API interface to connect to design software, project management software, and building materials trading platforms. The processing layer integrates an AI quantity calculation engine, a dynamic pricing model, a cost warning model, and a collaborative engine. The application layer deploys a project management module, an intelligent quantity calculation module, a dynamic pricing module, a cost monitoring module, and a collaborative workbench. The user layer supports multi-terminal access from PC and mobile terminals, and is configured with a role permission management module.

[0121] The data layer's relational database uses MySQL to store structured data, while the non-relational database uses MongoDB to store unstructured data such as BIM models and images. A RESTful API interface uses standardized data formats to enable real-time data interaction with systems such as Revit and Primavera. The processing layer's AI quantity calculation engine incorporates a ResNet-50-based CNN model. Through transfer learning, it fine-tunes the model on a labeled dataset of over 100 projects, achieving over 98% accuracy in output quantity calculations. The price library module of the dynamic pricing model uses the Scrapy framework as a web crawler to scrape steel and concrete price data from platforms such as China Building Materials Network daily. The scraping frequency can be adjusted to weekly or custom intervals as needed. The collaborative engine's workflow module supports drag-and-drop process design, allowing users to customize the node sequence and approval roles of change approval processes. For example, change requests submitted by the construction party can be automatically linked to the original bill of quantities to calculate cost increments, and cost data can be updated after electronic signature approval by the supervisor.

[0122] Specifically, the data layer utilizes a hybrid storage architecture combining structured and unstructured databases to address the issue of inconsistent data formats in traditional construction cost data. A MySQL database stores structured data such as cost ledgers and early warning logs, while a MongoDB database stores unstructured data such as BIM models and scanned copies of change orders. A RESTful API enables seamless integration with external systems, eliminating data silos. The processing layer's AI quantity calculation engine automatically extracts quantity features using a high-precision CNN model. This, combined with the IFC standard rule base, verifies the logical consistency of quantity calculations. This, for example, detects duplicate calculations of the same component across different specialized models, improving calculation efficiency and accuracy. The dynamic pricing model dynamically calculates comprehensive unit prices through real-time price capture and parameter adjustments. For example, it updates unit prices based on material transportation distance adjustment factors and tax policies, generating phased cost reports. The cost early warning model uses earned value analysis to calculate cost deviations and incorporates a random forest algorithm to predict overrun risks. A yellow alert is triggered when a cost deviation exceeding -5% is detected, and a system message is sent to the project manager. The collaborative workbench at the application layer integrates the change management process. After the construction party submits an application with change drawings, the system automatically calculates the cost changes corresponding to the increase in concrete usage, and updates it to all participants after online approval, reducing manual flow delays. The role permission management module at the user layer assigns differentiated permissions to different participants. For example, the owner-side interface displays the full amount of data, the construction-side interface only supports change application submission and progress upload, and the mobile terminal uses iOS / Android applications to achieve real-time upload and AI recognition of on-site progress photos. For example, the topping-out ratio of the main structure is analyzed based on the YOLOv5 model, and the pricing status of unfinished sub-projects is automatically associated. Therefore, the system realizes the full-process integration of engineering cost data, dynamic cost control and efficient collaboration among multiple parties through the collaborative operation of a multi-layer architecture.

[0123] As a preferred embodiment, the solution of this application is implemented as follows: the data layer uses a MySQL relational database to store structured data, including cost ledgers, quota databases, and early warning logs; and a MongoDB non-relational database to store BIM models, CAD drawings, and scanned copies of change orders. A RESTful API interface is configured to connect to Revit design software, enabling automatic parsing of BIM models and synchronization of engineering quantity data. The processing layer is deployed in a Kubernetes container cluster. The AI ​​quantity calculation engine loads a pre-trained ResNet-50 convolutional neural network model and fine-tunes it using transfer learning techniques on a dataset of 100 annotated BIM models from historical projects to output engineering quantity calculation results for concrete volume and rebar length. The dynamic pricing model integrates a web crawler tool developed with the Scrapy framework to crawl steel and concrete price data from the China Building Materials Network daily, which is then manually reviewed and updated to the price library module. The application layer deploys a project management module that supports importing Primavera schedule files when creating new projects. The intelligent quantity calculation module uses the AI ​​engine to automatically generate detailed itemized bills of quantities. The dynamic pricing module uses real-time price data to generate phased cost reports. The user layer develops a PC interface based on the Vue framework and a mobile application based on the React Native framework, configuring full data viewing permissions for the owner side, and the construction side interface only displays the change application submission and progress photo upload functions.

[0124] Through the above technical solutions, this application realizes the unified storage and cross-platform interaction of engineering cost data, solving the data island problem existing in traditional management; through the containerized deployment of AI quantity calculation engine and dynamic pricing model, it improves the efficiency of engineering quantity calculation and the real-time performance of price updates, avoiding errors and delays caused by manual operations; the multi-terminal collaborative platform supports automatic association of electronic approval and progress data, reducing the time cost of multi-party communication and the risk of information transmission errors.

[0125] In some of the solutions mentioned above, traditional engineering cost information management relies on manual operations and decentralized tools, resulting in data fragmentation and siloed data, which leads to low efficiency and poor accuracy in quantity calculation and pricing. While AI quantity calculation engines can extract engineering quantity characteristics, model training is difficult in small sample scenarios, and the calculation logic lacks standardization and scalability, affecting the accuracy and applicability of engineering quantity calculations.

[0126] This application further proposes that the CNN model of the AI ​​quantity calculation engine adopts ResNet-50 as the backbone network, and fine-tunes it on a small sample engineering quantity annotation dataset through transfer learning technology, with the model output accuracy greater than or equal to 98%; the IFC rule engine has a built-in quantity calculation logic rule library with unified standards for building information modeling applications, and supports custom rule extensions.

[0127] The ResNet-50 backbone network uses a residual structure to prevent gradient vanishing, making it suitable for processing deep features in BIM models and CAD drawings. Transfer learning technology leverages the general feature extraction capabilities of pre-trained models to fine-tune the model when the amount of engineering quantity annotation data is greater than or equal to 100 projects, improving model generalization in small sample scenarios. Model output accuracy is evaluated through cross-validation to ensure that the error in engineering quantity calculation results is controlled within 2%. The IFC rule engine establishes a logical rule library by parsing the component classification and engineering quantity calculation rules in the GB / T51212-2016 standard. Custom rule extensions add conditional statements through a graphical interface, such as setting deduction rules at beam-slab-column junctions, enabling flexible adaptation of quantity calculation logic.

[0128] Specifically, the residual connection structure of ResNet-50 can effectively extract multi-scale features when processing the three-dimensional geometric data of BIM models, such as identifying the density of steel bar layouts at beam-column joints. Transfer learning freezes the parameters of the underlying convolutional layer and only fine-tunes the fully connected layer. After training on a dataset of 100 annotated projects, the model's relative error in concrete volume calculations was reduced to 1.5%. The IFC rule engine automatically matches component types with standard rules when verifying project quantities, for example, checking whether door and window openings are deducted from wall quantities. When a project uses special construction techniques, users can add custom rules, such as defining the template area calculation coefficient for curved components, to ensure that the quantity calculation logic is consistent with actual on-site conditions. This solution solves the problem of quantity calculation errors caused by small sample training by combining a high-precision model with an extensible rule base, while adapting to differences in quantity calculation standards in different regions.

[0129] As a preferred embodiment, the solution of this application is implemented as follows: When building an AI quantity calculation engine, ResNet-50 is used as the backbone network. This network incorporates a residual structure to address the vanishing gradient problem in deep network training. Using transfer learning techniques, the model first loads the weights of a ResNet-50 model pre-trained on the ImageNet dataset, retaining all convolutional layer parameters except for the fully connected layers. Subsequently, fine-tuning training is performed on a small sample quantity annotation dataset of at least 100 historical projects. Each project dataset includes BIM models, CAD drawings, and corresponding ground-truth quantity labels. Training utilizes a cross-entropy loss function and the Adam optimizer, with a learning rate set to 0.001 and 200 iterations. The fine-tuned model outputs quantity calculation results with an accuracy exceeding 98%. During the construction of the IFC rule engine, the quantity calculation logic rules are encoded as executable verification scripts based on the "Uniform Standard for Building Information Modeling Applications" (GB / T 51212-2016). For example, verification of whether the rebar lap length at beam-column joints meets regulatory requirements is performed. The engine supports adding custom rules through a visual interface, such as adding special verification conditions for the volume calculation of special-shaped components. The expanded rule library can be automatically loaded into the quantity calculation process to perform logical verification.

[0130] Through the above technical solution, this application solves the problems of low efficiency and error accumulation caused by reliance on manual experience in the traditional quantity calculation process. It realizes high-precision engineering quantity extraction in small sample scenarios through transfer learning, reduces dependence on large-scale labeled data, and at the same time, based on a standardized rule base and scalable mechanism, ensures the rigor and adaptability of the quantity calculation logic, effectively avoiding repeated calculations and omission errors.

[0131] In some of the above-mentioned solutions of this application, the price library module of the dynamic pricing model needs to obtain price data of the building materials trading platform in real time, but the capture frequency in the existing technology is fixed and cannot be adjusted, and the data update cycle cannot be flexibly configured according to the material price fluctuation characteristics or project requirements, resulting in untimely data updates or waste of resources.

[0132] This application further proposes that the price library module of the dynamic pricing model is configured with a web crawler tool to periodically capture price data from the building materials trading platform, and the capture frequency is configurable; the workflow module of the collaborative engine supports drag-and-drop process design, and the change approval process can customize the node order and approval role.

[0133] The web crawler tool is implemented using the Scrapy framework, which supports the structured extraction of dynamic price data from building materials trading platforms by defining crawler rules and data parsing pipelines. The scheduled crawling function uses the task scheduler to set daily or weekly crawling cycles. The crawling frequency is adjusted in the system configuration interface based on material type or project stage requirements, such as daily crawling of steel prices and weekly crawling of concrete prices. The drag-and-drop process design presents approval nodes as graphical components through a visual interface. Users define the node sequence of the approval process by dragging and dropping components. Each node is associated with the approval role permissions. For example, a change application must be submitted by the construction party, reviewed by the supervisor, and approved by the owner.

[0134] Specifically, after the web crawler tool of the price library module is started, it will regularly access the target building materials trading platform according to the preset crawling frequency, extract fields such as material name, specification model, price and update time by parsing the HTML structure of the web page, and store them in the dynamic price library after data cleaning. Users can modify the crawling frequency of specific materials in the system management interface. For example, during the peak construction period, the frequency of steel bar price crawling can be adjusted to twice a day. In the workflow module of the collaborative engine, the administrator builds a process template by dragging and dropping the approval node components. For example, the design change approval process is defined as "Construction party submission → Supervision review → Owner approval → Cost update", and each node is configured with the corresponding approval personnel authority. When the construction party submits a change application, the system automatically pushes the application to the designated role according to the process template, and triggers the synchronous update of the cost data after approval. This solution ensures the timeliness of price data through configurable crawling frequency, and improves the efficiency of multi-party collaboration through customized approval processes.

[0135] As a preferred embodiment, the solution of the present application is specifically implemented as follows: the price library module of the dynamic pricing model integrates a web crawler component developed based on the Scrapy framework, which configures a timed task trigger and a proxy IP pool. The crawler task automatically accesses building materials trading platforms such as China Building Materials Network and My Steel Network according to the daily or weekly crawling cycle preset by the user, and extracts the latest market quotes of key materials such as rebar and commercial concrete by parsing page elements through XPath. The crawling frequency is set through a visual configuration interface, and supports setting differentiated update strategies by material category, such as daily crawling of steel and weekly crawling of decorative materials. The workflow module of the collaborative engine integrates a process designer, providing a graphical drag-and-drop interface that allows users to arrange and combine change approval process nodes (such as "construction party submits application", "supervision review", and "owner final review") as needed. Approval role permissions are defined through a role-node mapping table. For example, a "technical person in charge review" node is inserted into the "major design change" process, and the approval authority of this node is bound to the chief engineer role.

[0136] Through the above technical solution, this application solves the problem of pricing deviations caused by delayed building material price updates in traditional cost management. A configurable timed capture mechanism ensures the real-time nature of price database data, avoiding cost accounting errors caused by manual data collection delays. Furthermore, the drag-and-drop process design eliminates the collaborative inefficiencies associated with the rigidity of traditional offline approval processes, allowing different projects to flexibly configure approval paths based on actual management needs, reducing information flow blockages caused by process mismatches.

[0137] The above embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

[0138] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for information management of construction cost, characterized by: The following steps are involved: S1: Multi-source data collection and standardization processing, integrating all data from the design stage, construction stage, completion stage and external environment, and converting them into structured data that meets preset standards; S2: Based on the AI ​​quantity calculation engine and rule engine, it automatically calculates, logically verifies and corrects the engineering quantities to generate an accurate bill of quantities; S3: Call the dynamic pricing model, combine the real-time market price, quota database and adjustment parameters, calculate the sub-item project costs, measure project costs and other project costs in real time, and generate a phased cost report; S4: Based on earned value analysis and machine learning models, we analyze the deviation between actual and planned costs, predict cost overrun risks, and issue graded warnings. S5: Build a multi-party collaboration platform that supports change application submission, associated cost impact calculation, online approval, issue tracking, and progress synchronization, making collaborative processes electronic and traceable; S6: Classify and archive full-cycle data, generate an enterprise-level cost index library based on historical project data, and provide cost reference for new projects.

2. The information management method for construction cost according to claim 1, characterized in that: The multi-source data in step S1 includes BIM models, CAD drawings, and bills of quantities in the design phase; progress records, material delivery records, and change certificates in the construction phase; Settlement report and acceptance record at the completion stage; As well as price information, policy documents, and indicators of similar projects in the external environment; The standardization process includes: converting unstructured data into structured text through OCR recognition, removing redundant information through data cleaning, and classifying materials, processes, and cost types through unified coding rules.

3. The information management method for construction cost according to claim 1, characterized in that: The AI ​​quantity calculation engine in step S2 is based on a convolutional neural network model and learns engineering quantity extraction features through a training data set; the rule engine verifies the consistency of the quantity calculation logic based on the IFC standard and outputs the corrected engineering quantity results.

4. The information management method for construction cost according to claim 1, characterized in that: The dynamic pricing model in step S3 includes: Dynamic price database: Use crawlers to obtain market prices from building materials trading platforms and supplier websites in real time, supporting manual review and automatic updates; Quota library: stores construction project consumption quotas and regional supplementary quotas, and supports user-defined adjustment of consumption coefficients; Adjustment parameter module: Dynamically adjust the comprehensive unit price according to the material transportation distance, construction difficulty coefficient, and tax rate policy. The calculation formula is: Comprehensive unit price = (material base price × transportation distance adjustment coefficient + labor / machinery consumption × labor / machinery unit price) × (1 + management fee rate + profit margin) × (1 + tax rate).

5. The information management method for construction cost according to claim 1, characterized in that: The dynamic cost monitoring in step S4 includes: Cost deviation calculation: Use earned value analysis to calculate cost deviation and schedule deviation; Risk prediction: Based on historical project change data and material price fluctuation data, a prediction model is trained using a random forest algorithm to output the probability and amount of cost overruns within the next 30 days. Early warning push: According to the deviation level, early warning information is pushed to the project leader through system messages, SMS or email, and the specific deviation source is associated.

6. The information management method for construction cost according to claim 1, characterized in that: The multi-party collaboration platform in step S5 includes: Permission management module: assign hierarchical permissions to owners, designers, contractors, and consultants; Change management process: When the construction party submits a change application, the system automatically links the original bill of quantities to calculate the impact of the change. After online approval by the owner and supervisor, the cost data is updated and synchronized to all participating parties. Progress synchronization module: The construction party uploads progress photos / videos, and the system analyzes the construction completion degree through AI image recognition and automatically associates the pricing status of the corresponding sub-projects.

7. The information management method for construction cost according to claim 1, characterized in that: The data archiving and knowledge accumulation in step S6 include: Full-cycle data storage: stored in the database according to project stages, including structured data (and unstructured data; Index library generation: Through data mining technology, statistics are collected on the unit cost indicators, proportion indicators of sub-items and parts of historical projects, and material consumption indicators, and support is provided for filtering by region, structure type and construction period.

8. An information management system for construction cost, characterized by: The method for implementing any one of claims 1 to 7 comprises: Data layer: Relational and non-relational databases that store multi-source structured / unstructured data, with RESTful API interfaces configured for connecting to design software, project management software, and building materials trading platforms; Processing layer: Integrates AI calculation engine, dynamic pricing model, cost warning model, and collaboration engine; Application layer: deploy project management module, intelligent quantity calculation module, dynamic pricing module, cost monitoring module, and collaborative workbench; User layer: supports multi-terminal access on PC and mobile terminals, and configures role permission management module.

9. The information management system for construction cost according to claim 8, characterized in that: The CNN model of the AI ​​quantity calculation engine uses ResNet-50 as the backbone network and is fine-tuned on a small sample engineering quantity annotation dataset through transfer learning technology. The model output accuracy is ≥98%; the IFC rule engine has a built-in quantity calculation logic rule library of the "Uniform Standard for the Application of Building Information Modeling" GB / T51212-2016, and supports custom rule extensions.

10. The information management system for construction cost according to claim 8, characterized in that: The price library module of the dynamic pricing model is configured with a web crawler tool to periodically (daily / weekly) capture price data from the building materials trading platform, with a configurable capture frequency; the workflow module of the collaborative engine supports drag-and-drop process design.

Images