A generative design and multi-objective optimization method for tropical building envelopes

Abstract

The application discloses a kind of generation design and multi-objective optimization methods for tropical building envelope, data acquisition and environment definition, specification rule executable, candidate scheme generation and durability correction;Digital fingerprint and index table building;Model and drawing isogenous output;Drawing consistency check and quality gating;Performance simulation and robust decision output.The application has beneficial effects: specification articles are parsed into structured rule items and compiled into predicate set, rule source, version information and field relationship are recorded synchronously, reduce the influence caused by artificial interpretation difference, specification reference confusion and rule updating, specification constraint expression is more unified;Using digital fingerprint and index mapping correlation parameter, model component, drawing file and detail drawing index, then execute screening with drawing annotation qualified rate, reduce drawing rework, drawing error and construction deviation, improve the consistency of design results;It is convenient to form implementation document package suitable for engineering application, and enhance the balance of scheme selection.

Description

Technical Field

[0001] This invention relates to the field of green building and building information technology, specifically to a generative design and multi-objective optimization method for tropical building envelopes. Background Technology

[0002] The building envelope is a crucial component of building energy conservation and indoor thermal environment control, directly impacting cooling load, thermal performance, and damp-heat response. In tropical regions characterized by high temperature and humidity, strong radiation, and salt spray corrosion, the building envelope must not only meet requirements for heat transfer, shading, indoor comfort, and regulatory compliance, but also consider effective lifespan, maintenance and replacement, life-cycle carbon emissions, and economic costs. Therefore, its design must comprehensively consider multiple objectives. Existing methods have the following shortcomings: 1. The normative provisions are mostly expressed in natural language and often contain content such as conditional triggers, applicable boundaries, exceptions, partition limits and construction constraints. Traditional practices rely on manual interpretation and clause-by-clause verification, making it difficult to form a unified, stable and easily verifiable rule expression. Rule updates and version tracking are also quite difficult.

[0003] 2. The scheme parameters, 3D model and 2D construction drawings are often updated asynchronously during multiple rounds of modification, which can easily lead to inconsistencies between drawing annotations, detail indexes and database records, increasing the risk of rework in drawing review, construction deviations and quality. Existing methods also rely heavily on manual review and lack a unified standard for determining consistency.

[0004] 3. The degradation of material performance and shortened lifespan in tropical environments have not been fully incorporated into the optimization. The long-term carbon emissions and cost impacts of maintenance and replacement are easily underestimated, and the results are often biased towards the initial stage, making it difficult to truly reflect the comprehensive performance of the building envelope during its service life.

[0005] 4. Multi-objective optimization focuses too much on energy consumption, carbon emissions, and cost, and does not adequately constrain the quality of drawing delivery, the balance of solutions, and the applicability to the project. As a result, the results are difficult to use directly in the construction drawing stage, and there is still a significant disconnect between design optimization and deliverables, which affects the efficiency of subsequent applications and the value of the results.

[0006] Therefore, it is necessary to propose a generative design and multi-objective optimization method for tropical building envelopes to address the above technical problems. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a generative design and multi-objective optimization method for tropical building envelopes; A generative design and multi-objective optimization method for tropical building envelopes includes the following steps: S1. Data Acquisition and Environment Definition: Obtain project boundary conditions, climate boundary files, and corrosion environment levels. Design service life Compared with the operating baseline conditions; S2. Executability of Standard Rules: The large language model is invoked to perform semantic parsing on unstructured standard clauses, outputting structured rule entries according to a preset JSON schema. After schema validation, unit unification, and conflict resolution, the rule entries are compiled into an executable predicate set. And version locking of the rule set; S3. Candidate solution generation and durability correction: Under the predicate constraints, material parameters, structural hierarchy parameters, and geometric parameters are jointly sampled to generate a candidate solution set. Furthermore, a durability degradation factor is introduced to correct the material's effective lifespan and replacement impact, thereby mitigating hidden carbon and replacement costs. S4. Digital fingerprint and index table building: Calculate the constraint state vector for candidate schemes and generate digital fingerprint IDs. Build an index record table with the digital fingerprint IDs as the primary key to achieve consistent association between parameters, models, drawings, verification, and simulation records. S5. The model and drawings are output from the same source. The three-dimensional model and two-dimensional construction drawings of the enclosure structure are generated from the same source based on the same parameter set, and the digital fingerprint ID is written in the model component attributes, drawing file naming and detail index. S6. Drawing consistency check and quality gate control: Analyze key annotation feature points from 2D drawings and make consistency judgments with corresponding parameters in the index record table. Calculate the drawing annotation pass rate and compare it with the quality gate control threshold. Schemes that fail the gate control will not be included in the subsequent output. S7. Performance simulation and robust decision output: The performance of the solution that passes the quality gate is evaluated by performance simulation and the index record table is filled back. Based on the comprehensive decision function, the optimal solution and its implementation file package are output.

[0008] In step S2, the structured rule entries must include at least the fields RuleID, Component, Parameter, Operator, LimitValue, Unit, PreCondition, Severity, Source, and FixAction. After the schema validation is passed, the execution unit is unified and conflicts are resolved. When multiple rules exist for the same Parameter, the one with the higher Severity is selected first. When the Severity and Operator are the same, the stricter limit is selected. When there are mutually exclusive triggering conditions, the activation set is determined by PreCondition. The rule entries are then compiled into binary predicates. If the condition is met, output "1"; otherwise, output "0". Furthermore, the rule version is locked, the constraint state vector, and the digital fingerprint satisfy the following formula constraints: Rule version locked: ; in For hash mapping functions; For normalized serialization functions; For the content of the rule set; To standardize numbering; To standardize version / revision numbers; Generate timestamps for the rules; Constraint state vector: ; in, For the first A vector of parameters for each scheme; For the first A predicate function for each executable rule; This is the total number of rules to activate.

[0009] Digital fingerprint: ; In the formula, For a standardized serialization function that is consistent across platforms; The material characteristic vector (which should include at least thermal conductivity, density, specific heat capacity, carbon emission intensity, and a durability rating label reflecting tropical adaptability). To construct feature vectors (which should at least include the number of layers, the thickness of each layer, the connection method, and the key node implementation); The geometric feature vector (which includes at least thickness, area and orientation, and window-to-wall ratio); This is the constraint state vector; This is a rule version lock identifier.

[0010] Step S3 sets hard constraints that include at least the heat transfer performance of the building envelope, the shading performance of external windows, the occult carbon per unit area, and the total life cycle cost, and incorporates them in the form of predicates. This includes at least: The exterior walls and roof meet the following requirements: ; The exterior windows meet the following requirements: ; And simultaneously satisfy: ; Furthermore, schemes that simultaneously satisfy the aforementioned hard constraints are defined as valid candidate schemes and enter subsequent homogeneous drawing and quality gating.

[0011] In step S3, by material category A durability degradation factor is introduced, and the effective life, replacement cycles, and implicit carbon correction are calculated, where: Effective lifespan: ; Attenuation factor: ; Environmental corrosion rate: ; Number of replacements: ; And based on this, the amount of carbon contained per unit area is obtained: ; in Indicates rounding up; Standard lifespan; The environmental corrosion rate is determined by the corrosion environment level; The baseline corrosion rate; Initially hidden carbon; This represents the incremental increase in implicit carbon during a single replacement.

[0012] The generation of digital fingerprint IDs adopts a cross-platform consistent standardized serialization rule. The serialization includes at least fixed field order, unit normalization, numerical precision rounding, and fixed delimiter rules, and satisfies the following: ; in For a unit normal function, To be accurate Numerical rounding functions, For field order rules, This is a concatenated coding function, ensuring that the same scheme yields consistent results when repeatedly computed on different platforms. .

[0013] In step S4 Create an indexed record table using the primary key, and establish at least the following mapping relationships to achieve full-process traceability: :Will Write the shared parameters of the 3D model component instances and link them to the component instance identifier. One-to-one correspondence; :Will Write the name into the 2D drawing file name or title bar; : Match the detail drawing index number with Related; Record the drawing pass rate, simulation condition identification, simulation version and timestamp.

[0014] In step S6, the key annotation feature point set is used. Calculate the drawing annotation pass rate and implement quality gating, wherein the set of key annotation feature points must include at least the identity and traceability categories ( and Construct hierarchical classes (number of levels) Material labeling With thickness ( ), thermal parameters ( or and / or or ), Window parameters (window-to-wall ratio) Heat transfer coefficient of external window With solar heat gain coefficient ),and =20–50, weight =1–5, the drawing annotation pass rate meets the following requirements: ; The consistency determination function satisfies: ; in, For the first Weights of each feature point; The first one obtained from the analysis of the two-dimensional drawing Each parameter value; This refers to the corresponding parameter value in the index record table; Set the tolerance threshold according to the parameter type and at least meet the following recommended ranges: thickness / size type =1~5mm; Window-to-wall ratio =0.01~0.02; heat transfer coefficient labeling (including , , ) =0.02W / (m 2 K); Solar heat gain coefficient =0.02; And stipulate The consistency is determined to be completely consistent (tolerance is 0), and the following conditions are met:

[0015] The time is used to determine if the gate is passed, among which The value is set between 0.90 and 0.98, and further limited to this value during the construction drawing delivery stage. =0.95~0.98; when At that time, the scheme will not be included in the performance simulation and robust decision output.

[0016] In step S7, the performance simulation should at least calculate the annual energy consumption per unit area. Carbon emissions per unit area Total life cycle cost thermal comfort And calculate as follows: Annual energy consumption per unit area: ; Carbon emissions per unit area: ; Total life cycle cost: ; Thermal comfort index: ; in, For statistical area; For time steps Energy consumption; This represents the total number of time steps within the statistical period. Electricity emission factor; This refers to the initial construction cost; For the first Annual operating, maintenance and replacement costs; The discount rate; This refers to the lifespan in years.

[0017] Robust decision-making normalizes the indicators for directional consistency and introduces a dispersion penalty term and a gating penalty term, where: Contrarian indicators where smaller values ​​are better satisfy: ; A positive indicator where a larger value is better satisfy: ; For normalized index vector satisfy: ; And define the durability gating index: ; Set threshold The comprehensive decision function satisfies: ; Among them, weight satisfy =1, and =0.10~0.40 is the dispersion penalty coefficient; , This is the gate control penalty coefficient, thereby reducing the failure to meet the drawing pass rate threshold ( ) or durability threshold ( The schemes are shifted to the end of the sorting process, and the optimal enclosure scheme that is balanced and deliverable is output.

[0018] Compared with the prior art, the present invention has the following beneficial effects: 1. Parse the standard clauses into structured rule entries and compile them into predicate sets, and simultaneously record the rule source, version information and field relationships to reduce the impact of differences in manual interpretation, confusion in clause references and rule updates, and make the expression of standard constraints more consistent; 2. Utilize digital fingerprints and index mapping to associate parameters, model components, drawing files, and detail indexes, and then perform screening based on the drawing annotation pass rate Q to reduce rework during drawing review, drawing errors and omissions, and construction deviations, thereby improving the consistency of design deliverables; 3. Incorporate the durability degradation under salt spray and high humidity conditions into the life cycle calculation, correct the material's effective life, replacement frequency, hidden carbon, and maintenance costs, so that the evaluation results are closer to the long-term use in tropical environments and avoid overestimating short-life solutions. 4. Energy consumption, carbon emissions, cost, and comfort are unified and standardized, and dispersion penalties and threshold penalties are added to ensure that the output solution takes into account multiple indicators, making it easier to form an implementation document package suitable for engineering applications and enhancing the balance of solution selection. Attached Figure Description

[0019] Figure 1 This is a flowchart of the generative design and multi-objective optimization method of the present invention. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments and 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.

[0021] like Figure 1 As shown, the present invention proposes a generative design and multi-objective optimization method for tropical building envelopes, comprising the following steps: S1. Data Acquisition and Environment Definition: Obtain project boundary conditions, climate boundary files, and corrosion environment levels. Design service life Compared with the operating baseline conditions; S2. Executability of Standard Rules: The large language model is invoked to perform semantic parsing on unstructured standard clauses, outputting structured rule entries according to a preset JSON schema. After schema validation, unit unification, and conflict resolution, the rule entries are compiled into an executable predicate set. And version locking of the rule set; S3. Candidate solution generation and durability correction: Under the predicate constraints, material parameters, structural hierarchy parameters, and geometric parameters are jointly sampled to generate a candidate solution set. Furthermore, a durability degradation factor is introduced to correct the material's effective lifespan and replacement impact, thereby mitigating hidden carbon and replacement costs. S4. Digital fingerprint and index table building: Calculate the constraint state vector for candidate schemes and generate digital fingerprint IDs. Build an index record table with the digital fingerprint IDs as the primary key to achieve consistent association between parameters, models, drawings, verification, and simulation records. S5. The model and drawings are output from the same source. The three-dimensional model and two-dimensional construction drawings of the enclosure structure are generated from the same source based on the same parameter set, and the digital fingerprint ID is written in the model component attributes, drawing file naming and detail index. S6. Drawing consistency check and quality gate control: Analyze key annotation feature points from 2D drawings and make consistency judgments with corresponding parameters in the index record table. Calculate the drawing annotation pass rate and compare it with the quality gate control threshold. Schemes that fail the gate control will not be included in the subsequent output. S7. Performance simulation and robust decision output: The performance of the solution that passes the quality gate is evaluated by performance simulation and the index record table is filled back. Based on the comprehensive decision function, the optimal solution and its implementation file package are output.

[0022] In step S2, the structured rule entries must include at least the fields RuleID, Component, Parameter, Operator, LimitValue, Unit, PreCondition, Severity, Source, and FixAction. After the schema validation is passed, the execution unit is unified and conflicts are resolved. When multiple rules exist for the same Parameter, the one with the higher Severity is selected first. When the Severity and Operator are the same, the stricter limit is selected. When there are mutually exclusive triggering conditions, the activation set is determined by PreCondition. The rule entries are then compiled into binary predicates. (output 1 if satisfied, output 0 if not satisfied), and the rule version lock, constraint state vector, and digital fingerprint satisfy the following formula constraints: Rule version locked: ; in For hash mapping functions; For normalized serialization functions; For the content of the rule set; To standardize numbering; To standardize version / revision numbers; Generate timestamps for the rules; Constraint state vector: ; in, For the first A vector of parameters for each scheme; For the first A predicate function for each executable rule; This is the total number of rules to activate.

[0023] Digital fingerprint: ; In the formula, For a standardized serialization function that is consistent across platforms; The material characteristic vector (which should include at least thermal conductivity, density, specific heat capacity, carbon emission intensity, and a durability rating label reflecting tropical adaptability). To construct feature vectors (which should at least include the number of layers, the thickness of each layer, the connection method, and the key node implementation); The geometric feature vector (which includes at least thickness, area and orientation, and window-to-wall ratio); This is the constraint state vector; This is a rule version lock identifier.

[0024] Step S3 sets hard constraints that include at least the heat transfer performance of the building envelope, the shading performance of external windows, the occult carbon per unit area, and the total life cycle cost, and incorporates them in the form of predicates. This includes at least: The exterior walls and roof meet the following requirements: ; The exterior windows meet the following requirements: ; And simultaneously satisfy: ; Furthermore, schemes that simultaneously satisfy the aforementioned hard constraints are defined as valid candidate schemes and enter subsequent homogeneous drawing and quality gating.

[0025] In step S3, by material category A durability degradation factor is introduced, and the effective life, replacement cycles, and implicit carbon correction are calculated, where: Effective lifespan: ; Attenuation factor: ; Environmental corrosion rate: ; Number of replacements: ; And based on this, the amount of carbon contained per unit area is obtained: ; in Indicates rounding up; Standard lifespan; The environmental corrosion rate is determined by the corrosion environment level; The baseline corrosion rate; Initially hidden carbon; This represents the incremental increase in implicit carbon during a single replacement.

[0026] The generation of digital fingerprint IDs adopts a cross-platform consistent standardized serialization rule. The serialization includes at least fixed field order, unit normalization, numerical precision rounding, and fixed delimiter rules, and satisfies the following: ; in For a unit normal function, To be accurate Numerical rounding functions, For field order rules, This is a concatenated coding function, ensuring that the same scheme yields consistent results when repeatedly computed on different platforms. .

[0027] In step S4 Create an indexed record table using the primary key, and establish at least the following mapping relationships to achieve full-process traceability: :Will Write the shared parameters of the 3D model component instances and link them to the component instance identifier. One-to-one correspondence; :Will Write the name into the 2D drawing file name or title bar; : Match the detail drawing index number with Related; Record the drawing pass rate, simulation condition identification, simulation version and timestamp.

[0028] In step S6, the key annotation feature point set is used. Calculate the drawing annotation pass rate and implement quality gating, wherein the set of key annotation feature points must include at least the identity and traceability categories ( and Construct hierarchical classes (number of levels) Material labeling With thickness ( ), thermal parameters ( or and / or or ), Window parameters (window-to-wall ratio) Heat transfer coefficient of external window With solar heat gain coefficient ),and =20–50, weight =1–5, the drawing annotation pass rate meets the following requirements: ; The consistency determination function satisfies: ; in, For the first Weights of each feature point; The first one obtained from the analysis of the two-dimensional drawing Each parameter value; This refers to the corresponding parameter value in the index record table; Set the tolerance threshold according to the parameter type and at least meet the following recommended ranges: thickness / size type =1~5mm; Window-to-wall ratio =0.01~0.02; heat transfer coefficient labeling (including , , ) =0.02W / (m 2 K); Solar heat gain coefficient =0.02; And stipulate The consistency is determined to be completely consistent (tolerance is 0), and the following conditions are met: ; The time is used to determine if the gate is passed, among which The value is set between 0.90 and 0.98, and further limited to this value during the construction drawing delivery stage. =0.95~0.98; when At that time, the scheme will not be included in the performance simulation and robust decision output.

[0029] In step S7, the performance simulation should at least calculate the annual energy consumption per unit area. Carbon emissions per unit area Total life cycle cost thermal comfort And calculate as follows: Annual energy consumption per unit area: ; Carbon emissions per unit area: ; Total life cycle cost: ; Thermal comfort index: ; in, For statistical area; For time steps Energy consumption; This represents the total number of time steps within the statistical period. Electricity emission factor; This refers to the initial construction cost; For the first Annual operating, maintenance and replacement costs; The discount rate; This refers to the lifespan in years.

[0030] Robust decision-making normalizes the indicators for directional consistency and introduces a dispersion penalty term and a gating penalty term, where: Contrarian indicators where smaller values ​​are better satisfy: ; A positive indicator where a larger value is better satisfy: ; For normalized index vector satisfy: ; And define the durability gating index: ; Set threshold The comprehensive decision function satisfies: ; Among them, weight satisfy =1, and =0.10~0.40 is the dispersion penalty coefficient; , This is the gate control penalty coefficient, thereby reducing the failure to meet the drawing pass rate threshold ( ) or durability threshold ( The schemes are shifted to the end of the sorting process, and the optimal enclosure scheme that is balanced and deliverable is output.

[0031] Compared with the prior art, the present invention has the following beneficial effects: 1. Parse the standard clauses into structured rule entries and compile them into predicate sets, and simultaneously record the rule source, version information and field relationships to reduce the impact of differences in manual interpretation, confusion in clause references and rule updates, and make the expression of standard constraints more consistent; 2. Utilize digital fingerprints and index mapping to associate parameters, model components, drawing files, and detail indexes, and then perform screening based on the drawing annotation pass rate Q to reduce rework during drawing review, drawing errors and omissions, and construction deviations, thereby improving the consistency of design deliverables; 3. Incorporate the durability degradation under salt spray and high humidity conditions into the life cycle calculation, correct the material's effective life, replacement frequency, hidden carbon, and maintenance costs, so that the evaluation results are closer to the long-term use in tropical environments and avoid overestimating short-life solutions. 4. Energy consumption, carbon emissions, cost, and comfort are unified and standardized, and dispersion penalties and threshold penalties are added to ensure that the output solution takes into account multiple indicators, making it easier to form an implementation document package suitable for engineering applications and enhancing the balance of solution selection.

[0032] Example 1: This embodiment takes office buildings in tropical marine climates as the object and executes steps S1–S7 to realize the entire process of building envelope generation, model / drawing co-source output, drawing consistency gating, performance simulation and robust decision output.

[0033] Step S1: Data Acquisition and Environment Definition (1) Project boundary conditions and operating baseline conditions: The project is located in the near-shore area of ​​Sanya, Hainan. The project is about 500m from the coastline, the relative humidity of the air is greater than 80% all year round, and salt spray corrosion is significant; the total building area is about 12,000m². 2 Five stories, with a shape factor of approximately 0.28, facing approximately 15° south; the design service life is set at... ; (2) Climate boundary file and key meteorological statistics: The typical meteorological year file CHN_HI_Sanya.599480_TMYx.2004-2018.epw was used, and the statistical results are shown in Table 1.

[0034] Table 1. Statistical data on climate information in typical meteorological years of Sanya City

[0035] In Table 1, Q1 is the first quartile, which is the 25th percentile of the sample after it is arranged from smallest to largest; Q2 is the second quartile, which is the 50th percentile of the sample after it is arranged from smallest to largest; Q3 is the third quartile, which is the 75th percentile of the sample after it is arranged from smallest to largest; the difference between Q3 and Q1 is called the interquartile range (IQR).

[0036] Statistics show that the average annual outdoor temperature is approximately 26.8℃, and the highest annual temperature period is ( >30°C) Approximately 2160 hours, annual total solar radiation approximately 5300 MJ / (m²) 2 ·a).

[0037] (3) Corrosion environment level: Based on the high salt spray conditions in the nearshore area, the corrosion environment level is determined as follows: .

[0038] (4) Operating baseline conditions: Air conditioning set temperature 26℃, operating time 7:00–18:00; lighting power density 8W / m² 2 The equipment power density is 12W / m². 2 Personnel density: 0.1 people / m² 2 This benchmark will serve as the Schedule input source for the simulation model, and its source and version will be recorded in the structured constraint entries.

[0039] Step S2: Make the rules executable (1) Structured rule entries: The large language model is called to perform semantic parsing on the unstructured specification text, and the structured rule entries are output according to the preset JSON Schema. The fields must include at least RuleID, Component, Parameter, Operator, LimitValue, Unit, PreCondition, Severity, Source, FixAction, and after Schema validation, unit unification and conflict resolution, they are compiled into a set of binary predicates. (Output 1 if satisfied, output 0 if not satisfied).

[0040] (2) Rule version locking: To ensure the traceability and verifiability of rules, the rule set is version locked, satisfying the following: ; in For hash mapping functions; For normalized serialization functions; For the content of the rule set; To standardize numbering; To standardize version / revision numbers; Generate timestamps for the rules.

[0041] (3) Constraint state vector: for each candidate solution Calculate the constraint state vector of its activation rule: ; in, For the first A vector of parameters for each scheme; For the first A predicate function for each executable rule; This is the total number of rules to activate.

[0042] (4) Digital fingerprint: To achieve the identity association of "parameters-model-drawings-verification-simulation records", a digital fingerprint is generated for each scheme: ; in, For a standardized serialization function that is consistent across platforms; For material feature vectors; To construct feature vectors; These are geometric eigenvectors; This is the constraint state vector; This is a rule version lock identifier.

[0043] Step S3: Candidate solution generation and durability correction (1) Hard constraints and candidate validity: Under predicate constraints, material parameters, structural level parameters and geometric parameters are jointly sampled to generate a set of candidate schemes. And set hard constraints that include at least the heat transfer performance of the building envelope, the shading performance of external windows, the occult carbon per unit area, and the total life cycle cost. These constraints should include at least: The exterior walls and roof meet the following requirements: ; The exterior windows meet the following requirements: ; And simultaneously satisfy: ; Those that meet the hard constraints are allowed to proceed with subsequent same-source drawing and quality gating.

[0044] (2) Durability degradation correction: for material categories Introducing a durability degradation factor This is used to correct for the impact of material lifespan and replacement, and accordingly adjust for hidden carbon and replacement costs. Durability correction meets the following requirements: ; in For material standard life, For effective lifespan.

[0045] In this embodiment, for C4 level salt spray environment, an environmental durability degradation factor is introduced ( ), setting the building's design service life =50 years, based on the corrosion rate of ISO 9223 standard, defining the modified life of the main materials: Ordinary exterior wall coatings: standard life 10 years, degradation factor =0.5, actual effective lifespan 5 years. PVDF fluorocarbon coated aluminum sheet (three coats and three bakes): standard lifespan 25 years, degradation factor =0.8, with an actual effective lifespan of 20 years. Compared to the benchmark solution of ordinary exterior wall coatings requiring frequent recoating, this significantly reduces the total lifecycle carbon emissions and costs associated with maintenance and replacement.

[0046] Step S4: Digital fingerprint and index table creation (1) Index record table: based on digital fingerprints A primary key is used to create an indexed record table to record and associate parameter vectors. ), constraint state vector Model component instance identifier Drawing file name / detail index number, drawing pass rate Q Simulation output and version timestamps, etc., enable end-to-end consistency traceability.

[0047] (2) Vectorization and Digital Fingerprinting: For Candidate Solutions The exterior wall layers consist of PVDF fluorocarbon aluminum panels, an air gap, rock wool, autoclaved aerated concrete blocks, and cement mortar; the windows are Low-E double-glazed, with a south-facing window-to-wall ratio of 0.35. K =2.2, SHGC =0.28.

[0048] Exterior wall construction ( ): Layer 1: 3mm PVDF fluorocarbon aluminum veneer (corrosion resistance grade A); Layer 2: 30mm airflow gap (auxiliary insulation); Layer 3: 40mm rock wool board ( =0.045W / (m K), =140kg / m 3 ); Layer 4: 200mm autoclaved aerated concrete blocks B06 ( =0.019W / (m K); Layer 5: 20mm cement mortar.

[0049] Thermal verification: Calculated thermal resistance ≈0.76m 2 K / W, heat storage coefficient ≈4.08, thermal inertia index = =3.1.

[0050] Compliance assessment: =3.1>2.5, trigger rule R_Wall_Heavy_001 The limit is 1.50. This scheme calculates the heat transfer coefficient. =0.72 < 1.50, which meets the requirements.

[0051] External window parameters ( ): The ratio of the south-facing window to the wall is 0.35.

[0052] Low-E insulated glass is selected. =2.2, =0.28.

[0053] Compliance determination: triggering rules R_Window_Medium_001 ,Require ≤2.50, ≤0.30. This solution meets the requirements.

[0054] Based on this, normalized serialization and hash generation are performed, for Implement tamper-proof encoding: (Materials): [PVDF_Alu, Life20y, RockWool_40, Concrete_200]; (Construction): [5 Layers, 273mm, AirGap]; (Geometry): [South, WWR_0.35]; (State): [1, 1, 1, 1].

[0055] Hash operation: ; Result generated: The first 16 digits are truncated to obtain a unique identifier: ID:7F3A-9C2E-B1D4-88F0. Note: If the designer manually changes the rock wool thickness to 39mm, the hash value will change completely, and the system will immediately trigger a "data inconsistency" alarm.

[0056] Step S5: Output the model and drawings from the same source. (1) Same source mold and drawing: based on the same parameter set (same The 3D BIM model and 2D construction drawings of the building envelope are generated from the same source, and digital fingerprints are written into the model component attributes, drawing file naming, and detail index. This avoids inconsistencies between "model parameters - drawing annotations - simulation input".

[0057] (2) Example of a delivery package: Output a 3D BIM model file Model_7F3A-9C2E.rvt, and write shared parameters into the enclosure structural members within the model. Output a 2D compliant drawing (Draw_7F3A-9C2E.pdf), with the title block containing a digital fingerprint. Simultaneously, it outputs reports and log files for record keeping and review.

[0058] Step S6: Drawing consistency verification and quality gate control (1) Set of key annotation feature points: Analyzing the set of key annotation feature points from two-dimensional drawings It must include, and must be included in, the following mandatory options for calculating the pass rate: Identity and Traceability: Digital Fingerprint Solution Model component instance identifier ; Constructing a hierarchical class: number of levels Material labeling Thickness of each layer ; Thermal parameters: or and / or or At least one; External window parameters: Window-to-wall ratio Heat transfer coefficient of external window With solar heat gain coefficient .

[0059] (2) Quantity, weight and classification: Number of key annotation feature points Take a value between 20 and 50, with a weight of 10. Choose from 1 to 5 and configure them according to importance, with the weight of identity and traceability categories not less than 4, and the weight of construction level / thermal parameters / external window parameters categories not less than 3.

[0060] (3) Consistency judgment and pass rate calculation: Define the consistency judgment function ; in The first one obtained from the analysis of the two-dimensional drawing Each parameter value; This refers to the corresponding parameter value in the index record table; Tolerance for corresponding feature points. Drawing annotation compliance rate. Calculate using the following formula: ; (4) Recommended tolerance range and gate threshold: tolerance Set according to parameter type and at least meet the following requirements: thickness / size type. =1~5mm; Window-to-wall ratio =0.01~0.02; heat transfer coefficient labeling (including , , ) =0.02W / (m 2 K); Solar heat gain coefficient =0.02.

[0061] When the following conditions are met: The time is used to determine if the gate is passed, among which The value is set between 0.90 and 0.98, and further limited to this value during the construction drawing delivery stage. =0.95~0.98; when At that time, the scheme will not be included in the performance simulation and robust decision output.

[0062] (5) Gating example: for exist Key annotation feature points were selected within the scope for full verification; the verification results were obtained. Q =1.0, meeting the gate control requirements of the construction drawing stage, and is fixed in the drawing delivery documents.

[0063] Step S7: Performance Simulation and Robust Decision Output (1) Performance evaluation indicators: Perform performance simulation on the scheme that passes the quality gate and backfill the index record table, including at least the annual energy consumption per unit area. Carbon emissions per unit area Total life cycle cost thermal comfort The optimal solution and implementation package are output using a comprehensive decision function with discreteness penalty.

[0064] (2) Durability gating quantification: Introducing a minimum effective life index And set a threshold ,satisfy: ; when At that time, the scheme is significantly downgraded in the overall ranking by using a gating penalty.

[0065] In this embodiment, The effective lifespan of PVDF fluorocarbon aluminum panels for exterior wall cladding is approximately 20 years. Therefore, this material corresponds to... =0.4, which is The calculation provides a lower bound to reflect the constraint of durability differences on robust ordering.

[0066] (3) Recommendation of comprehensive decision function and parameters: The comprehensive decision function adopted is: ; in, This is the vector of the direction-consistent normalized result; The degree of dispersion; The weights of each indicator and satisfying =1; =0.1 0.3 is the dispersion penalty coefficient; , These are the penalty weights for quality gating and durability gating, respectively.

[0067] This example provides sample weights and penalty coefficients: =0.35、 =0.25、 =0.20、 =0.20.

[0068] (4) Simulation comparison and robustness interpretation with the baseline: The baseline (SetBaseline) uses ordinary exterior wall paint + 30mm XPS insulation board; the optimized scheme... PVDF fluorocarbon aluminum plate + 40mm rock wool is used. The comparison results are: annual power consumption. E Baseline 72.0 kWh / m 2 , 65.5 kWh / m 2 Total carbon emissions over 50 years C Baseline: 2243 kg CO2e / m³ 2 , 1963 kg CO2e / m 2 ; Full life cycle LCC Base price: 1920 yuan / m 2 Set088 is priced at 1580 yuan / m 2 Drawing pass rate Q Baseline 0.85 It is 1.0.

[0069] The key mechanism for improving carbon emissions and costs lies in: The number of maintenance and replacement cycles has been reduced from the baseline of 9 to 2, thereby reducing carbon emissions and scaffolding / labor costs throughout the maintenance cycle. This difference is then stabilized through durability correction and gating mechanisms.

[0070] (5) Output the implementation package: output the parameter vector and rule version corresponding to the optimal solution. Digital fingerprint Gating results Comparison with simulation results Write back the index record table and output an implementation document package that can be directly used for design delivery and review, including at least: a 3D BIM model, 2D compliant drawings, verification reports, and performance comparison reports; digital fingerprints are permanently written into the naming and title blocks of the above documents. To facilitate auditing and tracing.

[0071] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application also intends to include these modifications and variations.

Claims

1. A generative design and multi-objective optimization method for tropical building envelopes, characterized in that: Includes the following steps: S1. Data Acquisition and Environment Definition: Obtain project boundary conditions, climate boundary files, and corrosion environment levels. Design service life Compared with the operating baseline conditions; S2. Executability of Standard Rules: The large language model is invoked to perform semantic parsing on unstructured standard clauses, outputting structured rule entries according to a preset JSON Schema. After schema validation, unit unification, and conflict resolution, the rule entries are compiled into an executable predicate set. And version locking of the rule set; S3. Candidate solution generation and durability correction: Under the predicate constraints, material parameters, structural hierarchy parameters, and geometric parameters are jointly sampled to generate a candidate solution set. Furthermore, a durability degradation factor is introduced to correct the material's effective lifespan and replacement impact, thereby mitigating hidden carbon and replacement costs. S4. Digital fingerprint and index table building: Calculate the constraint state vector for candidate schemes and generate digital fingerprint IDs. Build an index record table with the digital fingerprint IDs as the primary key to achieve consistent association between parameters, models, drawings, verification, and simulation records. S5. The model and drawings are output from the same source. The three-dimensional model and two-dimensional construction drawings of the enclosure structure are generated from the same source based on the same parameter set, and the digital fingerprint ID is written in the model component attributes, drawing file naming and detail index. S6. Drawing consistency check and quality gate control: Analyze key annotation feature points from 2D drawings and make consistency judgments with corresponding parameters in the index record table. Calculate the drawing annotation pass rate and compare it with the quality gate control threshold. Schemes that fail the gate control will not be included in the subsequent output. S7. Performance simulation and robust decision output: The performance of the scheme that passes the quality gate is evaluated by performance simulation and the index record table is filled back. Based on the comprehensive decision function, the optimal scheme and its implementation file package are output.

2. The generative design and multi-objective optimization method for tropical building envelopes according to claim 1, characterized in that: In step S2, the structured rule entries must include at least the fields RuleID, Component, Parameter, Operator, LimitValue, Unit, PreCondition, Severity, Source, and FixAction. After the schema validation is passed, the execution unit is unified and conflicts are resolved. When multiple rules exist for the same Parameter, the one with the higher Severity is selected first. When the Severity and Operator are the same, the stricter limit is selected. When mutually exclusive triggering conditions exist, the activation set is determined by PreCondition. The rule entries are then compiled into binary predicates. (Satisfaction results in output "1", non-satisfaction results in output "0"), and the rule version is locked, the constraint state vector and the digital fingerprint satisfy the following formula constraints: Rule version locked: ； in For hash mapping functions; For normalized serialization functions; For the content of the rule set; To standardize numbering; To standardize version / revision numbers; Generate timestamps for the rules; Constraint state vector: ； in, For the first A vector of parameters for each scheme; For the first A predicate function for each executable rule; To the total number of activated rules; Digital fingerprint: ； In the formula, For a standardized serialization function that is consistent across platforms; For material feature vectors; To construct feature vectors; These are geometric eigenvectors; This is the constraint state vector; This is a rule version lock identifier.

3. The generative design and multi-objective optimization method for tropical building envelopes according to claim 1, characterized in that: Step S3 sets hard constraints, including at least the heat transfer performance of the building envelope, the shading performance of the external windows, the occult carbon per unit area, and the total life cycle cost, and incorporates them in the form of predicates. At least including: The exterior walls and roof meet the following requirements: ； The exterior windows meet the following requirements: ； And simultaneously satisfy: ； Furthermore, schemes that simultaneously satisfy the aforementioned hard constraints are defined as valid candidate schemes and enter subsequent homogeneous drawing and quality gating.

4. The generative design and multi-objective optimization method for tropical building envelopes according to claim 1, characterized in that: In step S3, the materials are categorized. A durability degradation factor is introduced, and the effective life, replacement cycles, and implicit carbon correction are calculated, where: Effective lifespan: ； Attenuation factor: ； Environmental corrosion rate: ； Number of replacements: ； And based on this, the amount of carbon contained per unit area is obtained: ； Where, represents rounding up; Standard lifespan; The environmental corrosion rate is determined by the corrosion environment level; The baseline corrosion rate; Initially hidden carbon; This represents the incremental increase in implicit carbon during a single replacement.

5. The generative design and multi-objective optimization method for tropical building envelopes according to claim 1, characterized in that: The digital fingerprint ID is generated using a cross-platform consistent standardized serialization rule. The serialization includes at least fixed field order, unit normalization, numerical precision rounding, and fixed separator rules, and satisfies the following: ； in For a unit normal function, To be accurate Numerical rounding functions, For field order rules, This is a concatenated coding function, ensuring that the same scheme yields consistent results when repeatedly computed on different platforms. .

6. The generative design and multi-objective optimization method for tropical building envelopes according to claim 1, characterized in that: In step S4 Create an indexed record table using the primary key, and establish at least the following mapping relationships to achieve full-process traceability: :Will Write the shared parameters of the 3D model component instances and link them to the component instance identifier. One-to-one correspondence; :Will Write the name into the 2D drawing file name or title bar; : Match the detail drawing index number with Related; Record the drawing pass rate, simulation condition identification, simulation version and timestamp.

7. The generative design and multi-objective optimization method for tropical building envelopes according to claim 1, characterized in that: In step S6, the key annotation feature point set is used. Calculate the compliance rate of drawing annotations and implement quality gating. The set of key annotation feature points must include at least the following categories: identity and traceability, structural hierarchy, thermal parameters, and external window parameters. =20–50, weight =1–5, the drawing annotation pass rate meets the following requirements: ； The consistency determination function satisfies: ； in, For the first Weights of each feature point; The first one obtained from the analysis of the two-dimensional drawing Each parameter value; This refers to the corresponding parameter value in the index record table; Set the tolerance threshold according to the parameter type and at least meet the following recommended ranges: thickness / size class =1~5mm; Window-to-wall ratio =0.01~0.02; Heat transfer coefficient labeling =0.02W / (m 2 K); Solar heat gain coefficient =0.02; And stipulate The consistency determination is complete consistency, and the following conditions are met: ； The time is used to determine if the gate is passed, among which The value is set between 0.90 and 0.98, and further limited to this value during the construction drawing delivery stage. =0.95~0.98; when At that time, the scheme will not be included in the performance simulation and robust decision output.

8. The generative design and multi-objective optimization method for tropical building envelopes according to claim 1, characterized in that: In step S7, the performance simulation should at least calculate the annual energy consumption per unit area. Carbon emissions per unit area Total life cycle cost thermal comfort And calculate as follows: Annual energy consumption per unit area: ； Carbon emissions per unit area: ； Total life cycle cost: ； Thermal comfort index: ； in, For statistical area; For time step Energy consumption; This represents the total number of time steps within the statistical period. Electricity emission factor; This refers to the initial construction cost; For the first Annual operating, maintenance and replacement costs; The discount rate; This refers to the lifespan in years.

9. The generative design and multi-objective optimization method for tropical building envelopes according to claim 1, characterized in that: The robust decision-making process normalizes the indicators for directional consistency and introduces a dispersion penalty term and a gating penalty term, wherein: Contrarian indicators where smaller values ​​are better satisfy: ； Positive indicators where larger values ​​are better satisfy: ； For normalized index vector satisfy: ； And define the durability gating index: ； Set threshold The comprehensive decision function satisfies: ； Among them, weight satisfy =1, and =0.10~0.40 is the dispersion penalty coefficient; , This is the gate control penalty coefficient, thereby reducing the failure to meet the drawing pass rate threshold ( ) or durability threshold ( The schemes are shifted to the end of the sorting process, and the optimal enclosure scheme that is balanced and deliverable is output.

Images