A Parametric Design and Construction Procedure Verification Method and System for New Building Construction Based on Digital Twin

Abstract

This invention belongs to the field of computer-aided process technology and discloses a method and system for parametric design and construction process verification of new building processes based on digital twins. It collects unstructured text data of new building processes and methods, extracts three core process entities (geometry, material properties, and process methods) and three types of relationships (constraints, coupling, and temporal dependencies) based on a domain-fine-tuned pre-trained language model, and constructs a process knowledge graph after entity alignment and disambiguation. This graph is then transformed into a parameter constraint ontology that can be called by a computing engine. The parameter constraint ontology is then connected to a building parametric design engine. A constraint semantic parsing engine transforms natural language process constraints, cost constraints, and performance constraints into multi-dimensional formal constraint expressions, embedding parameter calculation logic to calculate the feasible domain of design parameters, eliminating unqualified parameter combinations, and outputting multiple sets of parametric design models and initial parameter sets with semantic tags.

Description

Technical Field

[0001] This invention belongs to the field of computer-aided process technology, specifically a method and system for parametric design and construction process verification of a new building process based on digital twins. Background Technology

[0002] New building technologies such as prefabricated modular construction, concrete additive manufacturing, and low-carbon precast component production generally feature high parametric levels, strong process coupling, and stringent precision requirements. For real estate developers, the selection of new building technologies and their parametric design directly determine project development costs, construction cycles, and delivery quality. Currently, parametric design technology is widely used in building scheme generation and performance optimization, enabling rapid iteration of geometric models through parameter variables; digital twin technology is also gradually being introduced into construction management to achieve visualized monitoring and progress tracking of the construction process. However, existing technologies still face the following technical challenges regarding parametric design and construction process verification for new building technologies: Existing parametric design methods focus on geometric shape, structural mechanics, and energy performance as core constraints, treating construction feasibility only as a post-construction manual verification step. They fail to embed constraints such as the temporal logic of construction procedures, spatial interference, and resource allocation into the parametric design's parameter calculation engine. For new building technologies, developers cannot simultaneously quantify and assess construction cycle, cost, and quality risks during the design phase. There is a lack of quantitative constraint mapping between design parameters and procedural feasibility, resulting in a disconnect between parameter iteration and procedural verification. This leads to long project development cycles and high design rework costs.

[0003] Existing digital twin systems for building construction mostly focus on 1:1 geometric model replication and progress visualization during the construction phase, without constructing a multi-granularity quantitative mapping model for design parameter variables, twin simulation nodes, and construction process behavior responses. During design parameter tuning, developers cannot synchronously extrapolate the dynamic performance of the entire construction process through the twin; they can only conduct static geometric collision checks. This makes it difficult to cover the dynamic feasibility, quality stability, and cost fluctuation assessments of processes under varying process parameters, and thus cannot support the development team's solution comparison and decision-making.

[0004] Existing construction procedure verification methods are generally used as control tools during the construction phase. They perform static logical checks based on preset rules, only judging the compliance of the sequence and overlapping relationships of procedures, without addressing the transmission chain of construction errors caused by multiple coupled procedures in the design verification stage. Developers cannot identify error-sensitive design parameters and weak points in procedures during the parametric design phase, and cannot predict quality risks and cost deviations in advance. The verification results deviate significantly from the requirements for full-cycle project development control, easily leading to project delivery risks and cost overruns. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for parametric design and construction process verification of a new building process based on digital twins, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for parametric design and construction procedure verification of a new building process based on digital twins, comprising the following specific steps: Preferably, unstructured text data such as instructions for new building processes and methods, construction specifications, historical construction logs, and quality acceptance reports are collected. Based on a general pre-trained language model, domain fine-tuning is performed using labeled building process entities and relation samples to train a process entity recognition and relation extraction model. The process entity recognition and relationship extraction model is invoked to extract three types of core process entities—geometric parameters, material performance parameters, and process method parameters—as well as construction procedure constraint rules, and the constraint relationships, coupling relationships, and temporal dependencies between entities from unstructured text. Entity alignment and disambiguation are then performed to construct a process knowledge graph. Based on ontology modeling, the parameter value range, coupling relationship, and process constraint conditions in the process knowledge graph are transformed into a parameter constraint ontology that can be called by the computing engine.

[0007] Preferably, the parameter constraint ontology is connected to the building parametric design engine, and a constraint semantic parsing engine is built simultaneously to receive process constraints, cost constraints, performance constraints and design boundary conditions in natural language form; The constraint semantic parsing engine processes natural language descriptions of spatial boundary constraints, temporal logic constraints, resource carrying constraints, cost control constraints, and performance index constraints through technologies such as semantic role labeling and constraint predicate recognition, transforming them into multi-dimensional formal constraint expressions that can be recognized by constraint solving algorithms. During the generation of the parametric design model, the formal constraint expression and the constraint rules built into the parameter constraint ontology are synchronously embedded into the parameter solution logic. The feasible domain of the design parameters is solved by the constraint solving algorithm. It supports parallel solution and comparison of multiple sets of design parameter schemes, eliminates parameter combinations that do not meet the process and cost performance requirements, and outputs multiple parametric design models with semantic tags and corresponding initial parameter sets.

[0008] Preferably, the parametric design model with semantic tags is decomposed into multiple granularities, and hierarchical semantic tags are generated for the design parameter variables at different granularity levels: component level, process level, and technology level. Semantic annotation is performed on the geometric nodes, simulation input ports, and process behavior simulation units of the construction digital twin. Through semantic similarity matching algorithms, a one-to-one mapping is completed between design parameter variables, geometric nodes of the construction digital twin, and simulation units of process behavior. Geometrically, component-level and process-level geometric model mapping is performed. In the process dimension, parameter variables are bound to the simulation input ports of the construction digital twin. In the behavior dimension, the association between process actions and the simulation logic of the construction digital twin is performed. When design parameters change, the mapping relationship is updated synchronously with the parameter semantic labels, and a construction digital twin baseline model is built based on the mapping relationship.

[0009] Preferably, the construction process simulation is driven by the construction digital twin baseline model, and the mandatory clauses of the building construction code are accessed simultaneously as a basis for compliance judgment. Perform time sequence logic verification to confirm the compliance of the sequence of procedures and overlap time; perform spatial dynamic verification to simulate the spatial position changes of construction machinery, components, and temporary facilities as the procedures progress, and detect dynamic interference and spatial conflicts; and perform resource matching verification to check the matching degree between the manpower, material, and machinery configuration of each procedure and the procedure rhythm. Each verification result is semantically matched with the text library of mandatory provisions of the building construction code to determine the violation of the mandatory provisions, output a list of process verification issues with natural language attribution descriptions, and associate semantic tags with the corresponding design parameters.

[0010] Preferably, unstructured text data such as historical construction logs and quality deviation reports of new building technologies are collected, and combined with the problem node information output by process verification, the construction error range, error propagation coefficient and error accumulation rules of the process are extracted through information extraction and processing to construct the construction process error propagation matrix; Based on the aforementioned construction digital twin baseline model, the transmission and accumulation process of construction errors along the process chain is simulated, the actual deviation value after each process is completed is calculated, and the process quality margin corresponding to the design parameters is evaluated. Identify error-sensitive design parameters and weak points in processes, generate error risk assessment reports in natural language, and output the probability of quality compliance of the process chain.

[0011] Preferably, the system receives the process verification problem list, process weak points and error-sensitive design parameters, and receives optimization requirements and priority requirements in natural language form. Through a requirement semantic parsing model, the natural language optimization requirements are transformed into a computable optimization objective function and constraint boundary. Using the feasible region of design parameters as the boundary, and with process compliance, quality margin, construction period, and project cost as multi-dimensional optimization objectives, a constraint solving algorithm is used for collaborative optimization, adjusting the design parameter values ​​within the feasible region of parameters. The construction sequence arrangement scheme is adjusted synchronously, including adjusting the sequence of procedures, optimizing the overlapping relationship, and adjusting the resource allocation. The matching design parameter set and construction sequence scheme are output through iterative calculation.

[0012] Preferably, the text data such as design parameters and process combination schemes, verification reports, and optimization records after collaborative optimization verification are used as incremental inputs. Through incremental knowledge extraction algorithms, the supplementary parameter correlations, process constraints, and error propagation experience data are extracted. Entity completion and relationship update operations are performed. The extracted parameter relationships, process constraints and error propagation experience data are structured and stored in the process knowledge graph. The parameter constraint ontology is iteratively optimized through ontology update algorithm to supplement the feasible parameter range and error propagation coefficient of the new process. The updated parameter constraint ontology supports the constraint solution of subsequent parametric design.

[0013] This invention also provides a new parametric design and construction procedure verification system for building processes based on digital twins, used to implement the above-mentioned method, comprising the following modules: The knowledge ontology construction module collects unstructured text data of new building technologies, trains a technology entity recognition and relationship extraction model based on a general pre-trained language model, extracts core technology entities and their relationships, constructs a technology knowledge graph through entity alignment and disambiguation, and transforms it into a callable parameter constraint ontology. The parametric design solution module connects to the parameter constraint ontology, builds a constraint semantic parsing engine, transforms the input natural language form of process constraints, cost constraints, and performance constraints into multi-dimensional formal constraint expressions, embeds the constraints into the parameter solution logic, solves the feasible domain of design parameters and supports parallel calculation and comparison of multiple schemes, and outputs multiple parametric design models with semantic tags and corresponding initial parameter sets; The twin baseline construction module decomposes the parametric design model into multiple granularities and generates hierarchical semantic labels. It performs semantic annotation on each unit of the construction digital twin and completes multi-dimensional mapping through semantic similarity matching to construct the construction digital twin baseline model. The process coupling verification module drives the simulation of construction processes, accesses the text library of mandatory clauses of building construction specifications, and verifies the processes from the dimensions of temporal logic, spatial dynamics, and resource matching. Combined with the semantic verification of specifications, it outputs a list of process verification issues related to design parameters. The error simulation and evaluation module extracts construction error parameters to construct an error propagation matrix, simulates the propagation and accumulation process of errors along the process chain based on the twin baseline model, calculates process deviation values ​​and evaluates process quality margins, and identifies error-sensitive design parameters and weak points in the process. The collaborative optimization module parses natural language optimization requirements, determines optimization objectives and constraint boundaries, uses the feasible domain of design parameters as boundaries, employs constraint solving algorithms for collaborative optimization, synchronously adjusts design parameters and process arrangement schemes, and outputs a matching parameter set and process scheme. The knowledge iteration and update module uses optimized and verified text data as incremental input to extract new constraint rules and experience data, and updates the process knowledge graph and parameter constraint ontology.

[0014] The beneficial effects of this invention are as follows: 1. This invention collects unstructured text data such as construction process manuals, construction specifications, historical construction logs, and quality acceptance reports for new building technologies. Based on a domain-fine-tuned pre-trained language model, it extracts three core technological entities: geometry, material properties, and construction methods, as well as three types of relationships: constraints, coupling, and temporal dependencies. After entity alignment and disambiguation, a technological knowledge graph is constructed, which is then transformed into a parameter constraint ontology that can be called by the computing engine. The parameter constraint ontology is then integrated into the building parametric design engine. Through a constraint semantic parsing engine, natural language process constraints, cost constraints, and performance constraints are transformed into multi-dimensional formal constraint expressions, embedded with parameter calculation logic, to calculate the feasible domain of design parameters. This supports parallel calculation and comparison of multiple design schemes, eliminates unqualified parameter combinations, and outputs multiple sets of parametric design models and initial parameter sets with semantic tags. This technology embeds construction processes, costs, and quality constraints into the front end of parametric design, supporting development units to complete scheme selection and parameter optimization during the design phase, reducing design rework costs, shortening project development cycles, and significantly improving the construction adaptability and scheme decision-making efficiency of parametric design for new building technologies.

[0015] 2. This invention decomposes the parametric design model with semantic tags into three levels: component level, process level, and technology level, generating hierarchical semantic tags for design parameter variables at different levels. Simultaneously, it performs full-element semantic annotation on the geometric nodes, simulation input ports, and process behavior simulation units of the construction digital twin. Through a semantic similarity matching algorithm, it completes multi-dimensional correspondences of geometric mapping, parameter port binding, and process behavior association, constructing a construction digital twin baseline model. This model drives process simulation and connects to a mandatory code library, performing process coupling verification and compliance judgment from three dimensions: temporal, spatial, and resource. This solution establishes a multi-granular quantitative mapping between design parameters and twin simulation, enabling dynamic process verification under parameter changes, comprehensively covering process compliance risks, and improving the accuracy and response efficiency of construction verification for new building technologies.

[0016] 3. This invention collects unstructured text data such as historical construction logs and quality deviation reports of new building technologies. Combined with problem node information from process verification output, it extracts the construction error range, propagation coefficient, and accumulation rules to construct a construction process error propagation matrix. Based on a digital twin baseline model, it simulates the propagation and accumulation process of errors along the process chain, calculates the actual deviation value of each process, assesses the quality margin, and identifies error-sensitive design parameters and weak points in the process. Combining the verification problems and risk nodes with the analysis of optimization needs, it uses a constraint-solving algorithm to collaboratively optimize design parameters and process schemes, and incrementally updates the knowledge graph and parameter constraint ontology with the optimization results. This solution advances error assessment and quality control to the parametric design stage, helping development units accurately predict project quality risks and cost deviations, quantify the development value of each design parameter, achieve collaborative iterative optimization of design and construction, and continuously improve the full-cycle management level and scheme adaptability of new building technologies. Attached Figure Description

[0017] Figure 1 This is a complete flowchart of the overall method of the present invention; Figure 2 This is a flowchart of the process coupling and hole alignment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] like Figures 1 to 2 As shown, this embodiment of the invention provides a parametric design and construction procedure verification method for a new building process based on digital twins. This embodiment takes the new precast concrete composite floor slab support-free construction process as a fixed application scenario, and the method includes the following specific steps: This embodiment uses the new construction process of precast concrete composite floor slabs without support as a fixed scenario. Unstructured text data such as the "Code for Construction of Concrete Structures" GB50666-2011, the "Technical Standard for Prefabricated Concrete Buildings" GB / T51231-2016, construction logs of composite floor slab projects in the past two years, and quality acceptance reports of sub-projects are collected. Based on a general pre-trained language model with a Transformer encoder architecture, the composite floor slab process entity recognition and relation extraction model is trained by fine-tuning the domain using labeled composite floor slab process entity and relation samples.

[0020] The composite slab process entity recognition and relation extraction model is set up with 12 bidirectional attention coding layers, with one entity classification head and one relation classification head. The labeled samples are made according to the labeling specifications of the construction process field. The entity types are divided into three core entities: geometric parameters including composite slab thickness, truss reinforcement height, support lap length, etc.; material performance parameters including concrete strength grade, truss reinforcement yield strength, etc.; process and method parameters including hoisting and placement time, temporary support spacing, pouring slump, etc. The relation types cover three types of association relationships: lap length constraint, parameter coupling effect, and process sequence dependency, and adopt the BIO sequence labeling system.

[0021] The training is divided into two phases, based on 12,000 composite floor slab process annotation samples. In the first phase, the parameters of the bottom 10 coding layers are frozen, and only the parameters of the top classification head and the 11th and 12th coding layers are fine-tuned. In the second phase, all coding layer parameters are unfrozen and full parameter fine-tuning is performed.

[0022] The initial learning rate for the first stage is set to 2e-5, and the initial learning rate for the second stage of full parameter fine-tuning is set to 5e-6. The batch size is uniformly set to 16, and the maximum number of training rounds per stage is set to 20. The F1 scores for entity recognition and relation extraction in the verification set of composite slab process are used as the early termination criteria for each stage. If the criteria do not improve for 3 consecutive rounds, the training is terminated. The trained model takes a single unstructured text statement of composite slab process as input and outputs a sequence of entity category labels and a list of relation triples, which can be directly used for the construction of a knowledge graph of composite slab process.

[0023] The above model is used to extract three types of core process entities from the unstructured text of composite floor slabs: geometric parameters, material performance parameters, and process and method parameters. Construction process constraints such as hoisting operations must be carried out after the component strength meets the standard and insufficient lap length will reduce the bearing capacity of the support are extracted, as well as the constraints, coupling, and temporal dependencies between entities. After entity alignment and disambiguation processing, a composite floor slab process knowledge graph is constructed.

[0024] In the entity alignment stage, candidate entity names are first matched with the standard terminology dictionary of the construction engineering industry for literal matching and semantic similarity calculation. Different entities with similarity scores higher than the 0.85 threshold are uniformly mapped to standard term nodes. For example, different expressions such as slab length and support extension length are uniformly mapped to the standard term support lap length. Synonymous entities with large literal differences are normalized and merged after a second judgment based on the context of the composite slab hoisting process in the sentence in which they are located and the associated parameter features.

[0025] Disambiguation processing targets entries with the same name by matching three types of contextual features: the stage of the process, the type of component, and the category of the construction method applied. By combining these features, the specific scope of the entity is clarified. For example, when the entry for "thickness" is matched to a precast composite slab component, it refers to the thickness of the precast layer; when matched to a post-cast layer process, it refers to the thickness of the post-cast concrete, thus avoiding confusion between entities with the same name under different processes.

[0026] After alignment and disambiguation are completed, a unique identifier is assigned to all entity nodes such as composite slab thickness and support lap length. The relationships such as lap length constraining support bearing capacity and hoisting sequence depending on component strength are bound to the corresponding entity nodes, which together constitute the basic topology of the composite floor slab process knowledge graph.

[0027] Based on ontology modeling, the parameter value range, parameter coupling relationship, and process constraint conditions in the composite floor slab process knowledge graph are transformed into a composite slab parameter constraint ontology that can be called by the parametric design engine.

[0028] The composite slab parameter constraint body is set with three top-level categories: parameter entity class, constraint relationship class, and construction method / process class. Each category is configured with corresponding subclasses and data attributes: The parameter entity class stores the value range and dimension information of design parameters such as composite slab thickness 120-180mm and truss reinforcement height 80-120mm; The constraint relationship class stores parameter coupling rules such as "lap length ≥10mm" and "truss reinforcement height to composite slab thickness ratio ≤0.7", process constraint axioms, unit cost thresholds, component performance indicators, and other cost and performance constraint rules; The construction method / process class stores the timing logic of each process such as hoisting, grouting, and pouring, as well as the resource allocation and corresponding time cost calculation rules for hoisting 8 slabs per day by a single shift.

[0029] The ontology connects to the parametric design engine of the composite slab through a standardized REST data interface. The interface layer maps the built-in constraint axioms and attribute data into constraint entries and parameter baseline values ​​that the solution engine can directly read, supporting the automatic calculation of composite slab parameters.

[0030] In this embodiment, the composite slab parameter constraint ontology is connected to the building parametric design engine, and a constraint semantic parsing engine is built simultaneously. The engine receives process constraint requirements and design boundary conditions in natural language form, such as "the overlap length of the composite slab support shall not be less than 15mm, and the interval between adjacent hoisting processes shall not be less than 2 hours". The constraint semantic parsing engine is set up with four processing units in sequence: text preprocessing, semantic role labeling, constraint predicate recognition, and expression generation. It transforms the spatial boundary constraints, temporal logic constraints, resource carrying capacity constraints, cost control constraints, and performance index constraints described in natural language into multi-dimensional formal constraint expressions that can be recognized by the constraint solving algorithm.

[0031] The text preprocessing unit performs word segmentation, stop word filtering, and domain terminology normalization on the input text, such as normalizing "overlap length" to "support overlap length". The semantic role labeling unit identifies three types of semantic roles in the constraint statement: constraint subject, constraint object, and constraint condition. For example, the constraint subject is the composite slab hoisting process, the constraint object is the support overlap length, and the constraint condition is not less than 15mm. The constraint predicate recognition unit identifies constraint predicates such as greater than, less than, earlier than, later than, and containing based on a pre-set predicate dictionary in the construction field. The expression generation unit generates constraint expressions in standard first-order logic form according to the mapping relationship between semantic roles and constraint predicates, such as support overlap length ≥ 15mm and hoisting process interval ≥ 2h.

[0032] The engine takes as input the natural language process constraint text of composite floor slabs and outputs as structured constraint expressions that can be directly connected to constraint solving algorithms. It has a built-in terminology mapping table for composite floor slab process, which can uniformly map common terms and non-standard expressions such as "lap length" and "extend length" into the standard entity name "support lap length" in the parameter constraint ontology.

[0033] During the generation of the parametric design model for composite slabs, multi-dimensional formal constraint expressions and the constraint rules built into the parameter constraint ontology are synchronously embedded into the parameter solution logic. The constraint solution for multiple schemes is completed through feasible domain traversal: the parameter value ranges of composite slab thickness (120-180mm) and overlap length (10-25mm) built into the ontology, along with coupling rules, are used as hard constraint boundaries. The formal constraint expressions of external input, such as support overlap length ≥15mm and unit cost control within a preset range, are used as additional constraint conditions. All parameter value ranges that satisfy the constraints are searched in the 12-dimensional parameter space. Three sets of differentiated parameter schemes are generated in parallel, and the corresponding construction cycle and cost calculation results are matched synchronously. Parameter combinations that do not meet the hoisting procedures and cost requirements are eliminated. Multiple sets of parametric design models of composite floor slabs with semantic tags and corresponding initial parameter sets are output to support the development unit in carrying out scheme comparison and decision-making.

[0034] In this embodiment, the parametric design model of the composite slab with semantic tags is decomposed into three levels: component level, process level, and technology level. Hierarchical semantic tags are generated for the design parameter variables at different levels. At the same time, semantic annotation is completed for the geometric nodes, simulation input ports, and process behavior simulation units of the composite slab construction digital twin.

[0035] The hierarchical semantic tags adopt a hierarchical coding structure. The component-level tag format is DHB-001-thickness, containing three fields: component type code, component unique identifier, and parameter name, corresponding to independent design parameters such as the thickness of a single precast composite slab and the height of the truss reinforcement. The process-level tag format is GX002-hoisting and positioning-DHB001-lap length, containing four fields: process number, process name, associated component identifier, and parameter name, corresponding to individual parameters such as lap length and hoisting speed covered by a single hoisting process. The process-level parameter tag format is "GY01-supportless composite slab-hoisting efficiency," containing three fields: construction method category code, process name, and global parameter name, corresponding to individual global constraint parameters such as hoisting efficiency and curing cycle of the entire new process. A corresponding process-level master tag, GY01-supportless composite slab-GX001~GX006, is provided to associate the entire process set.

[0036] The three granularity levels are hierarchically related from top to bottom: the overall process-level label corresponds to all process-level labels, and the process-level labels correspond to all component-level labels under their jurisdiction; the process-level parameter label is a dedicated label for global constraint parameters and has a constraint relationship with the process-level labels; different level labels establish a hierarchical mapping through the identifier field, and the unique identifier field of the label remains fixed when the parameter value changes.

[0037] The semantic similarity matching algorithm completes the one-to-one mapping of composite slab design parameter variables, construction digital twin geometric nodes, and process behavior simulation units. The semantic similarity matching algorithm adopts a calculation logic that combines pre-trained semantic encoding vectors with domain rules weighting: first, the two sets of semantic label texts to be matched are input into the semantic encoding model to generate 768-dimensional semantic vectors, and cosine similarity is used as the basic similarity value. Furthermore, a weighting term is introduced from the rules of the construction process field. The three rules of component type matching, process level correspondence, and parameter category consistency are assigned weight coefficients of 0.3, 0.3, and 0.4 respectively to adjust the basic similarity value.

[0038] Weighted semantic similarity calculation formula: This represents the final weighted semantic similarity value between two sets of semantic tags, used to determine the matching relationship between entities, ports, and processes. This represents the cosine similarity of the 768-dimensional semantic vectors corresponding to the two sets of semantic labels, and is the basic similarity calculation value. , , The weight coefficients of the three domain matching rules are 0.3, 0.3, and 0.4 respectively, corresponding to component type matching, process level matching, and parameter category consistency. , , This represents the matching coefficient of the three rules. The value is 1 when the rule is satisfied and 0 when the rule is not satisfied.

[0039] For the composite slab twin mapping scenario, the algorithm sets differentiated matching thresholds of 0.8, 0.85, and 0.75 for component geometric labels, parameter port labels, and process behavior labels, respectively, and sets corresponding rule weights. If the corrected score meets the threshold of the corresponding dimension, the matching is considered successful, adapting to the accuracy requirements of the three dimensions. The algorithm takes two sets of semantic label texts with hierarchical attributes as input and outputs a list of matching pairs and corresponding similarity scores.

[0040] Based on the matching results, the geometric dimension completes the mapping of the three-dimensional geometric model at the component level and hoisting process level of a single composite slab, the process dimension completes the binding of the support overlap length parameter to the simulation input port of the hoisting process, and the behavior dimension completes the association between the hoisting and positioning action and the hoisting simulation logic of the twin. When the design parameters change, the mapping relationship is updated synchronously with the parameter semantic tags, and finally a digital twin baseline model for composite slab construction is constructed.

[0041] The digital twin baseline model for composite slab construction consists of four parts: geometric twin units, parameter port units, process behavior units, and state output units. The geometric twin units store the three-dimensional geometric models and corresponding spatial coordinate data of 20 composite slabs (DHB001 to DHB020), supporting construction machinery, and temporary facilities. The parameter port units provide input interfaces and initial values ​​for design parameters such as composite slab thickness and support overlap length. The process behavior units contain the simulation execution logic and state transition rules for six processes: precast component arrival and acceptance, hoisting and positioning, grouting and leveling, rebar binding, concrete pouring, and curing and acceptance. The state output units output state data such as component spatial position, construction progress, and tower crane resource consumption during the process simulation.

[0042] When the thickness of the composite slab is adjusted from 150mm to 160mm or the range of the overlap length is changed, only the parameter values ​​and range attribute fields in the hierarchical semantic label are modified accordingly; when adding a post-cast layer slump parameter, the corresponding label entries are added or removed, and the label's hierarchical code and unique identifier field (such as DHB-001) remain fixed.

[0043] The mapping relationship relies on the unique identifier of the tag to quickly locate the corresponding unit of the twin: value-type changes such as the adjustment of the thickness of the composite plate are synchronized with the update of the input values ​​and boundary configurations of the parameter port unit; structural changes such as the addition of temporary support parameters are synchronized with the adjustment of the port binding relationship and the process behavior association logic; local rematching is only triggered when the overall tag hierarchy structure is adjusted.

[0044] When the support overlap length parameter changes, the input value of the hoisting process parameter port is updated synchronously through the mapping relationship, triggering the hoisting process behavior unit to re-execute the overall process simulation at a preset step size of 1 hour. For spatial interference verification, a minute-level sub-step deduction is adopted, and the verification results are updated synchronously.

[0045] In this embodiment, the construction process simulation is driven by a digital twin baseline model of the composite slab construction. Simultaneously, a pre-annotated text library of mandatory clauses of the "Code for Acceptance of Construction Quality of Concrete Structures" GB50204-2015 is accessed as the basis for compliance judgment. Each clause in this text library stores four pieces of structured information: clause number, standard text, applicable composite slab process scope, and violation judgment criteria. A semantic index structure is also established, with each index item corresponding to the core semantic vector and keyword set of each clause, supporting rapid retrieval and hit determination based on semantic similarity matching.

[0046] The process simulation is executed sequentially according to the process chain sequence of precast component arrival and acceptance, hoisting and positioning, grouting and leveling, rebar binding, concrete pouring, and curing and acceptance. Before starting a single hoisting process, the current design parameters such as the composite slab thickness of 150mm and the support overlap length of 15mm are read from the corresponding parameter port. The component qualification status data and initial spatial pose data of the previous arrival and acceptance process are read. The tower crane hoisting action simulation logic built into the process behavior unit is called, and the overall process flow is advanced according to the preset step size of 1 hour. For dynamic processes such as hoisting operations, the spatial pose deduction is performed with a sub-step size of minutes. The single-step simulation step size can be adjusted from minutes to man-days as needed.

[0047] After a single hoisting and positioning process is completed, the spatial position, installation status, and overlap length of the composite plate of the geometric twin unit are updated synchronously. This generates the output parameter set for this process and transmits it to the corresponding input port of the downstream grouting and leveling process. This process continues until all processes are completed.

[0048] The process coupling verification is carried out simultaneously from three dimensions: the temporal logic verification is based on the temporal dependency rules stored in the process class of the parameter constraint ontology, to verify the compliance of the process sequence of "the grouting process must be completed within 30 minutes after hoisting and positioning", and to compare the overlap time of grouting and hoisting with the requirement of a minimum process interval of 15 minutes to determine the compliance of the overlap. The spatial dynamic verification extracts the bounding box data of the tower crane, composite slab components, and surrounding scaffolding according to the simulation step size, and detects the overlap of bounding boxes between elements during the hoisting operation frame by frame to identify dynamic interference and spatial conflicts; the resource matching verification is based on the resource consumption simulation logic built into the process behavior unit, and calculates the total demand of 2 hoisting teams and 1 tower crane during the daily construction period, compares it with the preset total resource configuration, and determines the degree of matching between resource supply and demand and the daily hoisting process rhythm of 8 slabs.

[0049] Each verification result is semantically matched with the mandatory clause text library of prefabricated construction specifications to determine the violation of mandatory clauses. For specification matching scenarios, a special semantic encoding for specification clauses is added to the original semantic similarity matching algorithm to fine-tune the weight and violation judgment rules: the verification problem description of "no temporary fixation after hoisting and positioning" and the specification clause are respectively encoded into semantic vectors to calculate the basic similarity. Then, a keyword weighting rule for mandatory clauses is introduced to give higher weight to core keywords related to safety and quality, such as "temporary fixation" and "hoisting safety," to correct the similarity score. If the score is greater than the specification matching threshold of 0.82, it is determined that the clause is matched and the correlation between the matched clause number GB50204-2015 Article 9.2.3 and the violation is output simultaneously.

[0050] Upon encountering clause 9.2.3 of GB50204-2015, the process enters the secondary judgment process, verifying each of the clause's preset criteria for violations, such as failure to install temporary fixing components, component position deviation exceeding 5mm, and spatial position violations. If all criteria are triggered simultaneously, it is formally judged as a mandatory clause violation; if only some criteria are triggered and core safety and quality requirements are not met, it is marked as a general risk warning item and not included in the formal violation list. After the judgment is completed, all violations and risk items are associated with the corresponding clause number and requirement trigger details. The final output includes a list of process verification issues with natural language attribution descriptions such as failure to configure temporary supports after the composite slab DHB001 is hoisted and positioned, and violation of the temporary fixing requirements for prefabricated components, and is associated with the corresponding process-level label "GX002-Hoisting and Positioning-DHB001-Temporary Support Spacing".

[0051] In this embodiment, unstructured text data such as historical construction logs and dimensional deviation quality reports of composite floor slab projects in the past two years are collected. Combined with the hoisting and overlapping deviation problem node information output by process verification, the construction error range, error transmission coefficient and error accumulation rules of each process are extracted through information extraction and processing to construct the composite slab construction process error transmission matrix.

[0052] The error propagation matrix for composite slab construction is a 12-order square matrix, with 12 core input and output parameters involved in the entire composite slab construction process as its dimensions. Rows correspond to upstream process output parameters, and columns correspond to downstream process input parameters. Matrix elements are error propagation coefficients between upstream and downstream parameters. For example, the coefficient for the propagation of hoisting positioning deviation to support overlap length deviation is 0.8, used to characterize the influence of upstream parameter deviation on downstream parameters. The diagonal elements of the matrix correspond to the inherent error coefficients of each process output parameter. For example, the construction error coefficient of the hoisting positioning parameter itself is ±3mm, used to characterize the contribution of the construction deviation of the corresponding output parameter in a single process. All coefficients were obtained through statistical fitting of 320 sets of historical construction deviation data.

[0053] Based on the digital twin baseline model of composite slab construction, the transmission and accumulation process of construction errors along the six-step process chain is simulated, the actual deviation value after the completion of each step is calculated, and the process quality margin of corresponding design parameters such as support lap length is evaluated.

[0054] The quality margin assessment is based on the allowable deviation of ±5mm for the support lap joint as specified in the construction acceptance specification. Based on the simulation results of error propagation in the whole process chain, the final cumulative absolute value of the support lap joint length is obtained as 3.2mm. The cumulative absolute value of the deviation is compared with the allowable deviation threshold, and the actual deviation ratio is calculated to be 64%. The remaining allowable deviation ratio of 36% is the final quality margin of the support lap joint length parameter. At the same time, the process-level deviation ratio of the hoisting process itself is output separately for single-process accuracy analysis.

[0055] Formula for calculating process quality margin: This represents the process quality margin of the corresponding design parameter, with a value range of 0 to 1. The larger the value, the more sufficient the remaining fault tolerance space of the parameter. This represents the final cumulative absolute value of the deviation of this parameter after the entire process chain is completed; This indicates the allowable deviation threshold for this parameter as specified in the building construction acceptance standard, and serves as the criterion for determining quality margin.

[0056] The error propagation simulation starts from the first incoming inspection process. The initial deviation value is taken from the inherent error coefficient of the corresponding process output parameter in the error propagation matrix, ±1.5mm. Combined with the nominal value of 150mm for the composite slab thickness, the initial deviation range is generated. After each process is completed, the propagation coefficient of the corresponding parameter of the current process and the downstream process in the error propagation matrix is ​​called. For example, the deviation of the hoisting process is converted into the input deviation of the grouting process with a coefficient of 0.8. The inherent error of the grouting process itself, ±2mm, is added to obtain the actual total deviation of the process.

[0057] Cumulative recursive formula for process error: This represents the total cumulative deviation of the construction parameters after the i-th process is completed; This represents the cumulative deviation value of the output parameters of the (i-1)th upstream process, which is the input quantity for error propagation; In the error propagation matrix of construction process, the error propagation coefficient from the output parameter of the (i-1)th process to the input parameter of the ith process represents the degree of influence of the upstream parameter deviation on the downstream parameter. It represents the inherent construction error coefficient of the i-th process, which is the contribution of the corresponding parameter of a single process to its own construction deviation.

[0058] Deviation calculation covers three types of construction deviations: composite slab dimensions, installation space location, and bearing capacity performance of supports. It is deduced step by step according to the six-process chain to complete the accumulation and transmission of deviations throughout the entire process.

[0059] Based on the deviation extrapolation results, error-sensitive design parameters and weak points in the process are identified, and a composite slab error risk assessment report is generated, outputting the probability of the process chain quality meeting the standard. Among them, the support overlap length is an error-sensitive design parameter. Based on the composite slab support bearing capacity calculation model and the mapping relationship between size deviation, a unit 1mm fluctuation corresponds to a final bearing capacity deviation change rate of 12%, which is higher than the preset threshold of 10%. The hoisting and positioning process is a weak point, with its own quality margin of 52%, which is lower than the single process margin threshold of 60%, and its cumulative contribution to the error of downstream processes accounts for 42%.

[0060] The overall quality compliance probability of the composite slab process chain is obtained through Monte Carlo simulation statistics based on the quality margin and error probability distribution of each process. Under the current scheme, the value is 89.2%, which represents the overall level of construction quality that meets acceptance requirements.

[0061] In this embodiment, the system receives a list of issues related to the verification of procedures such as temporary fixing violations during hoisting, as well as sensitive parameters such as weak nodes in hoisting and the lap length of supports. It also receives natural language optimization requirements and priority requirements such as "prioritize improving quality margin, then compress the construction cycle, and the lap length of supports shall not be less than 12mm". These requirements are then transformed into a computable optimization objective function and constraint boundary through a requirement semantic parsing model.

[0062] The requirement semantic parsing model has two built-in units: optimization target identification and constraint boundary extraction. The optimization target identification unit identifies three core optimization directions from the natural language requirements: quality margin improvement, construction period compression, and project cost control. These are mapped to multi-dimensional standard objective functions according to priority: maximizing the quality margin coefficient, minimizing the total process time, and controlling the total project cost. The constraint boundary extraction unit extracts constraints such as support overlap length ≥12mm, composite slab thickness 120-180mm, and tower crane configuration cannot be adjusted from the requirement text, and transforms them into parameter feasible domain boundaries and hard constraint rules.

[0063] The model takes as input the natural language optimization requirements text and priority annotations for composite slabs, and outputs a structured objective function expression and a set of constraint boundaries, which can be directly connected to constraint solving algorithms; it has built-in objective weight mapping rules for composite slab process optimization scenarios, which can automatically adjust the weight coefficients of multi-objective optimization according to the input priority.

[0064] Using the feasible domain of composite slab design parameters as the search boundary, and taking process compliance, quality margin, construction period, and project cost as multi-dimensional core optimization objectives, a constraint-solving algorithm with a multi-objective iterative optimization mode is adopted to perform collaborative optimization of design parameters and process schemes.

[0065] The optimization adopts an iterative logic of alternating parameter adjustment and process adjustment to adapt to the multi-objective optimization needs of the development unit: after each round of design parameter value adjustment, such as adjusting the support overlap length from 15mm to 18mm, the parameter port input data of the construction digital twin baseline model is updated simultaneously, the full process simulation and error propagation calculation are re-executed, the construction period and project cost of the corresponding scheme are calculated simultaneously, and the updated process compliance results, quality margin values ​​and cost and schedule indicators are obtained. If the quality margin still does not meet the target requirement of 50% after parameter adjustment, the construction process precision control plan will be further adjusted. For example, a secondary positioning calibration step will be added in the hoisting process, and the leveling precision control threshold of the grouting process will be tightened from ±2mm to ±1.5mm. After the process control plan is adjusted, the adaptability of the composite slab design parameters will be checked again. If it is necessary to compress the construction cycle at the same time, the overlap time between the grouting process and the rebar binding process will be adjusted accordingly, from 2 hours to 1.5 hours.

[0066] After three rounds of alternating iterations of parameters and procedures, the number of procedure violations was cleared, the quality margin reached 58% to meet the optimization target, and the iteration was terminated. The final design parameter set (composite slab thickness 150mm, support overlap length 17mm, truss reinforcement height 100mm) and the optimized construction procedure scheme were calculated and output. With the baseline total construction period of the construction section corresponding to the initial scheme as a reference, the total construction period was shortened by 8% after optimization.

[0067] In this embodiment, the composite plate design parameters and process combination schemes, process verification reports, parameter optimization records and other text data verified through collaborative optimization are used as incremental inputs. Supplementary parameter relationships, process constraints and error propagation experience data are extracted through incremental knowledge extraction algorithms.

[0068] The incremental knowledge extraction algorithm adopts a processing flow that combines entity linking with relation confidence determination: First, candidate entities in the incremental text are linked with existing composite plate process knowledge graph entities, and "support overlap length" is identified as an existing entity; for existing entities, the new association relationship and value of "optimal quality margin when overlap length is 17mm" are extracted, and the relation confidence is calculated to be 0.92, which is greater than the set threshold of 0.8, and is determined to be a valid new relation; after performing entity type classification and attribute completion on the new entity, it is stored in the knowledge graph.

[0069] The algorithm takes incremental unstructured text data of the composite plate as input and outputs a set of newly added entities and a set of newly added relation triples. Based on the extraction results, it performs entity completion and relation update operations, stores the structured data in the composite plate process knowledge graph, and then iteratively optimizes the composite plate parameter constraint ontology through the ontology update algorithm.

[0070] The ontology update algorithm performs incremental ontology evolution based on newly added knowledge graph data: First, it compares the concept system of the new triples with that of the existing ontology, adds a "performance impact relationship" subclass under the constraint relationship class, and transforms the new performance impact relationship of "lap length-mass margin" into ontology axioms under the corresponding subclass; then, it updates the optimal value range of the support lap length to 16-18mm based on the new data, and simultaneously updates the attribute data of the corresponding parameters in the parameter entity class, and updates the error propagation coefficient of the hoisting-grouting process to 0.75; after the update is completed, it performs ontology consistency verification, and completes ontology iteration after confirming that there are no logical conflicts.

[0071] The updated composite slab parameter constraint body supplements the optimal parameter range and the corrected error propagation coefficient of the new support-free process, which can support the constraint solution of parametric design for subsequent similar composite floor projects.

[0072] This invention also provides a parametric design and construction procedure verification system for a new building process based on digital twins, used to realize the parametric design and procedure verification method for the aforementioned precast concrete composite floor slab support-free new process, including the following modules: The knowledge ontology construction module collects unstructured text data such as construction method manuals, construction specifications, historical construction logs, and quality acceptance reports for the new technology of precast concrete composite floor slabs without support. Based on a general pre-trained language model, it trains a composite slab process entity recognition and relationship extraction model, extracts three types of core process entities—geometric parameters, material performance parameters, and process method parameters—as well as their constraints, coupling, and temporal dependencies. After entity alignment and disambiguation, it constructs a composite floor slab process knowledge graph and transforms it into a composite slab parameter constraint ontology that can be called by the parametric design engine. The parametric design solution module connects to the composite slab parameter constraint ontology, builds a constraint semantic parsing engine, transforms the input natural language form of process constraints, cost constraints, and performance constraints into multi-dimensional formal constraint expressions, embeds the constraints into the parameter solution logic, solves the feasible domain of composite slab design parameters and supports parallel calculation and comparison of multiple schemes, and outputs multiple sets of parametric design models of composite floor slabs with semantic tags and corresponding initial parameter sets; The twin baseline construction module decomposes the parametric design model of the composite slab at multiple granularities (component level, process level, and technology level) and generates hierarchical semantic tags. It performs semantic annotation on the geometric nodes, simulation input ports, and process behavior simulation units of the digital twin of the composite slab construction. Through semantic similarity matching, it completes multi-dimensional mapping of geometric mapping, parameter port binding, and process behavior association, and constructs the digital twin baseline model of the composite slab construction. The process coupling verification module drives the construction process simulation through the digital twin baseline model of the composite slab construction, accesses the mandatory clause text library of the prefabricated building construction code, performs process coupling verification from the dimensions of temporal logic, spatial dynamics, and resource matching, and outputs a list of process verification issues with corresponding design parameter semantic tags in combination with the semantic verification of the code. The error simulation and evaluation module combines historical construction deviation data with process verification problem nodes to extract construction error characteristic parameters and construct the composite slab construction process error transmission matrix. Based on the twin baseline model, it simulates the transmission and accumulation process of error along the process chain, calculates the actual deviation value of the process and evaluates the process quality margin, and identifies error-sensitive design parameters and weak nodes in the process. The collaborative optimization module analyzes the optimization requirements and priority requirements in natural language, determines the optimization objective function and constraint boundary, uses the feasible region of composite slab design parameters as the boundary, and adopts constraint solving algorithm to perform collaborative optimization of design parameters and process schemes, synchronously adjusts the design parameter values ​​and process arrangement schemes, and outputs a set of design parameters and construction process schemes that match the optimization objective. The knowledge iteration and update module takes optimized and verified design parameter schemes, verification reports, optimization records and other text data as incremental inputs, extracts the new parameter associations, process constraint rules and error propagation experience data, and incrementally updates the composite floor slab process knowledge graph and parameter constraint ontology.

[0073] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0074] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for parametric design and construction procedure verification of a new building process based on digital twins, characterized in that, The specific steps include the following: Collect unstructured text data of new building technologies, extract technology entities and their relationships, construct a technology knowledge graph, and generate a parameter constraint ontology; The parameter constraint ontology is integrated into the building parametric design engine. The input natural language form process constraints, cost constraints, and performance constraints are semantically parsed to generate multi-dimensional formal constraint expressions and embed parameter calculation logic. It supports parallel calculation and comparison of multiple sets of design parameter schemes and outputs multiple parametric design models with semantic tags and corresponding initial parameter sets. The parametric design model with semantic tags is decomposed into multiple granularities and hierarchical semantic tags are generated. Semantic annotation is performed on each unit of the construction digital twin. The mapping is completed through semantic similarity matching, and the construction digital twin baseline model is constructed. Based on the aforementioned construction digital twin baseline model, the simulation of construction procedures is driven, and the mandatory clauses of building construction specifications are accessed to perform multi-dimensional process coupling verification and specification matching, and output a list of process verification issues. By combining historical unstructured text data of construction and process verification problem nodes, error feature parameters are extracted and a construction process error transmission matrix is ​​constructed. The transmission and accumulation process of error along the process chain is simulated, the process quality margin is evaluated, and the weak nodes of the process and error-sensitive design parameters are output. Receive verification and evaluation results, parse natural language optimization requirements to determine optimization objectives and constraint boundaries, complete collaborative optimization through constraint solving, and output design parameter set and construction procedure scheme; The optimized and verified text data is used as incremental input to update the process knowledge graph and the parameter constraint ontology.

2. The method for parametric design and construction procedure verification of a new building process based on digital twins according to claim 1, characterized in that, The unstructured text data of the new building technology includes construction method manuals, construction specifications, construction logs, and quality acceptance reports; an entity recognition and relationship extraction model finely tuned in the field of building technology is used to extract various technology entities and their interrelationships. The process entity includes three core entities: geometric parameters, material performance parameters, and process method parameters. The relationships include process constraint rules and the constraints, coupling, and temporal dependencies between entities. After entity alignment and disambiguation, a process knowledge graph is constructed, which is then converted into a callable parameter constraint ontology through ontology modeling.

3. The method for parametric design and construction procedure verification of a new building process based on digital twins according to claim 2, characterized in that, While accessing the parameter constraint ontology, a constraint semantic parsing engine is built. Through semantic role labeling and constraint predicate recognition, the spatial boundary constraints, temporal logic constraints, resource carrying capacity constraints, cost control constraints, and performance index constraints described in natural language are transformed into the multi-dimensional formal constraint expressions. During the parameter solution process, the multi-dimensional formal constraint expression and the constraint rules built into the parameter constraint ontology are synchronously embedded into the parameter solution logic. The feasible domain of the design parameters is solved through the constraint solving algorithm. The construction period and engineering cost parameters are synchronously associated to carry out parallel calculation of multiple schemes. After eliminating unqualified parameter combinations, multiple sets of comparable design parameter schemes are output.

4. The method for parametric design and construction procedure verification of a new building process based on digital twins according to claim 3, characterized in that, The parametric design model with semantic tags is decomposed into three levels of granularity: component level, process level, and technology level, and the hierarchical semantic tags are generated for the design parameter variables at each level. The geometric nodes, simulation input ports, and process behavior simulation units of the construction digital twin are respectively semantically annotated; Through semantic similarity matching, geometric dimension model mapping, process dimension parameter port binding, and behavioral dimension process logic association are completed. When design parameters change, the mapping relationship is updated synchronously with the hierarchical semantic tags.

5. The method for parametric design and construction procedure verification of a new building process based on digital twins according to claim 4, characterized in that, The process coupling verification is carried out from three dimensions: temporal logic, spatial dynamics, and resource matching. The temporal logic check is used to verify the compliance of the sequence of operations and the overlap time; the spatial dynamic check is used to simulate the spatial position changes of construction elements and detect dynamic interference; and the resource matching check is used to verify the matching degree between the configuration of manpower, materials, and machinery and the rhythm of the operation. The verification results are matched with the mandatory provisions of the building construction code for semantic similarity. After determining the violations, a list of verification issues for the process with attribution descriptions is output and associated with the corresponding hierarchical semantic tags.

6. The method for parametric design and construction procedure verification of a new building process based on digital twins according to claim 5, characterized in that, The historical construction unstructured text data includes construction logs and quality deviation reports; Based on the problem nodes of the aforementioned process verification, the construction error range, error propagation coefficient, and error accumulation rules are extracted to construct the error propagation matrix of the aforementioned construction process; Based on the simulation error propagation and accumulation of the construction digital twin baseline model, the actual deviation value of each process is calculated and the quality margin of the corresponding design parameters is evaluated. The error-sensitive design parameters and the weak nodes of the process are identified, and the probability of quality compliance is output.

7. The method for parametric design and construction procedure verification of a new building process based on digital twins as described in claim 6, characterized in that, The natural language optimization requirements and priority requirements are received and transformed into a computable optimization objective function and the constraint boundary through a requirement semantic parsing model; Using the feasible domain of the design parameters as the boundary, and taking process compliance, quality margin, construction period, and project cost as multi-dimensional optimization objectives, the constraint solving algorithm is used to adjust the values ​​of the design parameters, and simultaneously optimize the process sequence, overlap relationship, and resource allocation. Through iterative calculation, the set of design parameters and the construction process scheme that match the optimization objectives are output.

8. The method for parametric design and construction procedure verification of a new building process based on digital twins according to claim 7, characterized in that, The incremental input text data includes optimized and verified design parameters and process combination schemes, verification reports, and optimization records; The incremental knowledge extraction algorithm extracts supplementary parameter relationships, process constraints, and error propagation data to complete entity completion and relationship updates. The extracted structured data is stored in the process knowledge graph, and the parameter constraint ontology is iteratively optimized through an ontology update algorithm.

9. A parameterized design and construction procedure verification system for a new building process based on digital twins, used to implement the method described in any one of claims 1-8, characterized in that, Includes the following modules: The knowledge ontology construction module is used to collect unstructured text data of new building technologies, extract technology entities and their relationships, construct a technology knowledge graph, and generate a parameter constraint ontology. The parametric design solution module is used to access the parameter constraint ontology, parse the process constraints, cost constraints, and performance constraints in natural language form, and generate multi-dimensional formal constraint expressions. After embedding the parameter solution logic, it supports parallel solution and comparison of multiple schemes and outputs multiple parametric design models with semantic tags and corresponding initial parameter sets. The twin baseline construction module is used to decompose the parametric design model with semantic labels at multiple granularities and generate hierarchical semantic labels, perform semantic annotation on each unit of the construction digital twin, and construct the construction digital twin baseline model through matching and mapping. The process coupling verification module is used to drive the simulation of construction processes, access the text library of mandatory clauses of building construction specifications to perform multi-dimensional process coupling verification and specification matching, and output a list of process verification issues; The error simulation and evaluation module is used to extract error characteristic parameters to construct the error transmission matrix of construction procedures, simulate error transmission accumulation and evaluate quality margin, and output weak nodes of procedures and error-sensitive design parameters. The collaborative optimization module is used to parse natural language optimization requirements, determine optimization objectives and constraint boundaries, and complete the collaborative optimization of design parameters and process schemes. The knowledge iteration and update module is used to update the process knowledge graph and parameter constraint ontology using optimized and verified text data as increments.

10. A parameterized design and construction procedure verification system for a new building process based on digital twins as described in claim 9, characterized in that, The process coupling verification module has three types of verification units built in: temporal logic, spatial dynamics, and resource matching. These correspond to process overlap verification, dynamic interference detection, and resource configuration matching verification, respectively. Each unit is associated with the semantic matching unit of the mandatory clauses of the building construction code, and the output results are automatically associated with the corresponding hierarchical semantic tags.

Images