Step-by-step decision optimization method and system for building design variables

Abstract

The application discloses a kind of building design variable step-by-step decision optimization method and system, belong to building design and performance optimization technical field.The method includes: obtaining design performance target set and design variable set;Establish building performance prediction proxy model;Based on proxy model, sensitivity analysis is carried out, and design variable is divided into decision variable and to-be-decision variable;Random sampling is carried out to to-be-decision variable to generate uncertainty scenario library, and reliability-robustness evaluation model is established based on six sigma theory;Reliability-robustness evaluation result is used as fitness function, and the optimization value range of decision variable is solved using optimization algorithm;Subjective decision is made in combination with building subjective creation demand;To-be-decision variable and undecision variable are combined and iterated until all variable decisions are completed.The application regards to-be-decision variable as uncertainty factor, effectively avoids step-by-step optimization into local optimal solution, and simultaneously considers the flexibility of building creation and the reliability of scheme performance.

Description

Technical Field

[0001] This invention belongs to the field of architectural design and performance optimization technology, and particularly relates to a step-by-step decision optimization method and system for architectural design variables based on the uncertainty of the decision variables. Background Technology

[0002] The potential for performance optimization is enormous in the early stages of architectural design. Achieving the same performance later on would be significantly more costly, making early design decisions crucial for performance-based architectural design.

[0003] Architectural design variables are numerous and diverse, and common architectural optimization processes are mainly divided into two categories: "one-step optimization" and "step-by-step optimization." "One-step optimization" refers to using intelligent algorithms to solve for the optimal values ​​of all design variables at once. It has high computational efficiency, but it has obvious drawbacks: First, when there are categorical, discrete, and continuous variables at the same time, the reliability of intelligent algorithms in solving for the optimal solution set will decrease; second, intelligent algorithms completely dominate the design direction, and architects cannot participate in the decision-making process, resulting in design schemes that cannot take into account both the architect's subjective preferences and the objective requirements of building performance.

[0004] "Step-by-step optimization" refers to dividing design variables into "decision variables" and "variables to be decided," making decisions step by step. Architects can lead the design direction at each step, incorporating their architectural preferences into the design scheme, resulting in high flexibility. However, in existing step-by-step optimization techniques, when making decisions on "decision variables," default values ​​are often assigned to "variables to be decided." This means that each decision is made based on the default values ​​of the variables to be decided, and the optimal solution obtained is only a local optimum, not a global optimum.

[0005] Therefore, how to achieve the complementary advantages of "one-step optimization" and "step-by-step optimization" while taking into account both the flexibility of architectural creation and the overall performance of architectural schemes is a technical problem that urgently needs to be solved in the field of architectural design optimization. Summary of the Invention

[0006] The purpose of this invention is to provide a step-by-step decision optimization method and system for architectural design variables, which aims to solve the problem that existing step-by-step optimization methods cause the design scheme to fall into a local optimum due to setting the decision variables to default values.

[0007] The technical solution adopted by this invention to solve its technical problem is as follows: On the one hand, a step-by-step decision optimization method for architectural design variables is provided, including the following steps: S1: Obtain the design performance target set M and design variable set V of the building scheme, and determine the value range of each design variable; S2: Based on the design variable set V and the design performance target set M, establish a building performance prediction proxy model L; S3: Based on the building performance prediction proxy model L, perform sensitivity analysis on the design variable set V, and in conjunction with architectural design requirements, divide the design variable set V into a decision variable set V. a and the set of variables to be decided V b ; S4: For the set of variables to be decided, V b Random sampling is performed to generate an uncertainty scenario library U. A reliability-robustness assessment model N is established based on Six Sigma theory. The set of decision variables V under the uncertainty scenario library U is then calculated. a The reliability-robustness evaluation results of each candidate scheme; S5: Transfer the decision variable set V a As optimization variables, the uncertainty scenario library U is used as a disturbance factor, and the reliability-robustness assessment result is used as the fitness function. An optimization algorithm is then used to solve the problem to obtain the optimal solution set P, and the decision variable set V is determined. a Optimization range of values ​​V for each variable p ; S6: Based on the optimized value range V p Based on the subjective creative needs of architecture, the decision variable set V is... a Making subjective decisions, denoted as V. p1 The set of undecided variables is denoted as V. p2 ; S7: The set of variables to be decided, V... b With the set of undecided variables V p2 Merge and return to step S3 for iterative optimization until the iteration termination condition is met, completing the decision on all design variables.

[0008] Preferably, in step S1: the design performance target set M = {M1, M2, ..., M} p}, where p represents the number of performance indicators; the design variable set V={V1, V2, ..., V n} where n represents the number of variables; the architectural design variables include one or more of the following: building layout and orientation, building dimensions, structural selection and material usage, window-to-wall ratio, window selection and window usage, exterior wall construction selection and material usage, roof construction selection and material usage, floor construction selection and material usage, interior wall parameters and material usage, and equipment selection and material usage; the types of design variables include categorical variables and continuous variables; the continuous variables include variables that can continuously take values ​​within a certain interval and variables that take values ​​at a finite number of discrete points (such as standard modular dimensions, standard component models, etc.).

[0009] Preferably, step S2 specifically includes: S21: sampling the design variable set V using a pseudo-random sampling method to generate a design scheme sample library; S22: calculating the performance of each sample scheme using simulation software to obtain a sample dataset containing design variables and performance values; S23: using the sample dataset to fit a quantitative model between the design variable V and the building performance M to obtain the building performance prediction proxy model L. More preferably, the pseudo-random sampling method in step S21 is Latin hypercube sampling; the simulation software in step S22 includes energy consumption simulation software and BIM software, and the performance includes one or more of the following: cooling energy demand, heating energy demand, operational carbon emissions, and implicit carbon emissions during building operation; the building performance prediction proxy model L in step S23 is a regression model, an artificial neural network model, a random forest model, or a decision tree model. The building performance prediction proxy model L is only applicable to the building scheme type determined in step S1.

[0010] Preferably, step S3 specifically includes: S31: performing sensitivity analysis on the design variable set V based on the building performance prediction proxy model L; S32: based on the sensitivity analysis results and the actual building design situation, selecting one or more of the most sensitive variables as decision variables, and the remaining variables as variables to be decided; simultaneously, architects are allowed to include factors that have a significant impact on subjective creation into the decision variables according to their subjective creative needs. More preferably, the sensitivity analysis in step S31 adopts the Sobol global sensitivity analysis method, specifically including: performing Sobol analysis to obtain the main effect index S1 of each variable i on each performance target. i The interaction effect index and total effect index ST between any two or more variables are used. The most sensitive variables are identified based on the cumulative variance contribution rate: the contribution rate of each variable or interaction term to any performance objective is sorted from largest to smallest, and the variables involved in the variables or interaction terms whose cumulative contribution ranks in the top 50% are considered the most sensitive factors. The threshold of the top 50% is determined based on statistical experience that the cumulative contribution rate in engineering can usually explain most of the output variance, and this threshold can be adjusted in subsequent iterations according to actual needs. For reference, Saltelli et al. (2008) pointed out in their monograph on sensitivity analysis (Global Sensitivity Analysis: The Primer) that a cumulative contribution rate of 40% to 60% is a reasonable dividing line; this invention uses the middle value of 50% to balance computational complexity and decision-making efficiency. When multiple performance objectives exist, the sensitivity of each variable to all performance objectives is comprehensively judged, and one or more variables with the highest comprehensive sensitivity are used as decision variables.

[0011] Preferably, step S4 specifically includes: S41: for the set of variables to be decided V bPseudo-random sampling is performed on the values ​​to generate an uncertainty scenario library U={U1, U2, ..., U...} k}, where k represents the number of uncertain scenarios; S42: Based on the 6σ theory, establish a reliability-robustness evaluation model N for each performance of the building scheme, and the reliability evaluation result of any scheme i and any target performance m is N. mi Under p performance objectives, the reliability-robustness evaluation result N of scheme i is... i ={N 1i , N 2i , ..., N pi More preferably, the pseudo-random sampling in step S41 adopts the Latin hypercube sampling method, and the sampling size k is more than 20 times the number of the variables to be decided; when the variables to be decided are categorical variables, all their possible values ​​are included in the uncertainty scenario library U. The mathematical expression of the reliability-robustness assessment model N in step S42 is: N mi = μ mi + 6σ mi , Where, μ mi and σ mi Let M represent the performance of any architectural design scheme i under k uncertain scenarios. m The mean and standard deviation; , , M represents the performance of any architectural design scheme i under any uncertainty scenario j. m ; M represents the performance of scheme i. m The reliability and robustness assessment results are presented in this invention. The 6σ coefficient is chosen based on the Six Sigma design philosophy. This coefficient balances performance mean and volatility and is widely used for robustness optimization in engineering practice. Its selection criteria can refer to relevant Six Sigma design standards (such as ISO 13053). In practical applications, if the design has low reliability requirements (such as ordinary residential buildings), 3σ (theoretical coverage 99.73%) can be used; for important public buildings, 4σ or 5σ can be used; for critical infrastructure, 6σ (coverage 99.99966%) is selected. The specific coefficient is pre-set by the designer based on the project's importance, risk appetite, and cost constraints.

[0012] Preferably, step S5 specifically includes: S51: Integrating the decision variable set V a The uncertainty scenario library U and the reliability-robustness assessment model N are combined to form a complete optimization model; S52: The decision variable set V is...a Using the uncertainty scenario library U as the solution variable, the reliability-robustness assessment result as the fitness function, and intelligent algorithms to find the optimal solution set P, the decision variable set V is determined. a The optimal value range V of each design variable in the process p .

[0013] More preferably, the fitness function in step S52 is: Fitness = [f N1 (X), f N2 (X), ..., f Np (X) ], Among them, f Nm (X) represents the reliability-robustness assessment result N. m The functional relationship between the design variable X and p is the number of performance objectives; when the performance objective is maximized, the corresponding f Nm Add a negative sign before (X).

[0014] The intelligent algorithm mentioned in step S52 is a multi-objective intelligent optimization algorithm, including but not limited to NSGA-II, MOEA / D, and SPEA2; when the number of candidate solutions is small, manual optimization is used instead of the intelligent algorithm optimization.

[0015] Preferably, in step S6: in the optimized value range V p Within the scope, the architect assigns a value to each design variable based on subjective judgment. This value is either a fixed value or a range of values. If the value is fixed, it indicates that the decision for that variable has been completed, and the variable is recorded in the set of decided variables V. p1 If the value is within a range, it indicates that the variable has a relatively small impact on the subjective creation of the architecture, and multiple values ​​are acceptable. This variable should be recorded in the undecided variable set V. p2 A decision will be made after the potential for performance improvement of the architectural design has approached zero. The architect's subjective judgment is based on one or more of the following factors: architectural spatial form requirements, aesthetic requirements, owner needs, and cost factors.

[0016] Preferably, step S7 specifically includes: S71: taking the set of variables to be decided V from step S3... b and the set of undecided variables V p2 Merge and redivide into a new set of decision variables V c and the new set of decision variables V dS72: Repeat steps S3 to S5 until all variables that significantly impact the design decision have been decided, and the potential for performance improvement of the architectural scheme approaches zero, at which point the iteration terminates. S73: Perform a trade-off decision on the variables that remain undecided at the end of the last optimization. More preferably, the quantitative criterion for terminating the iteration in step S72 is: reaching the maximum number of iterations, or convergence when the optimal solution set distribution stabilizes. The maximum number of iterations is 50 iterations. The trade-off decision in step S73 includes: when multiple design variables can be selected within their value range, the architect independently selects values ​​for each variable within that range; variables with less impact on the architect's subjective creative process undergo a comprehensive trade-off decision after the last optimization to improve the diversity of design schemes.

[0017] Preferably, the step-by-step decision optimization method for architectural design variables further includes the following steps: based on the values ​​of the design variables after the final decision is completed, a comprehensive evaluation of the performance of the architectural scheme is conducted, and the final optimized design scheme and its performance evaluation results are output.

[0018] The method is applied to a Building Information Modeling (BIM) environment. Some or all variables in the design variable set V originate from component parameters in the BIM model. Performance indicators in the performance target set M are calculated through the interface between the BIM model and simulation software. Specifically, the BIM model provides geometric and material information, which is transmitted to energy consumption simulation software (such as EnergyPlus) and carbon emission calculation tools via open data formats (such as IFC) or API interfaces to achieve automated performance evaluation, thus seamlessly integrating with the method steps of this invention. (Modified: The interaction method within the BIM environment has been refined.) The above steps can be executed by a general-purpose computer, workstation, or cloud server. The program modules are stored in memory, executed by the processor, and interact with Building Information Modeling (BIM) software and energy simulation software through input / output interfaces. The computer program product is provided as Software as a Service (SaaS) or deployed locally for use by the design team.

[0019] On the other hand, a step-by-step decision optimization system for architectural design variables is provided, including: The data acquisition module is used to acquire the design performance target set M and design variable set V of the building scheme, and to determine the value range of each design variable; The proxy model construction module is used to establish a building performance prediction proxy model L based on the design variable set V and the design performance target set M; The sensitivity analysis and variable partitioning module is used to perform sensitivity analysis on the design variable set V based on the building performance prediction proxy model L, and, in conjunction with architectural design requirements, partition the design variable set V into a decision variable set V. a and the set of variables to be decided Vb ; The uncertainty scenario generation and evaluation module is used to evaluate the set of variables to be decided, V. b Random sampling is performed to generate an uncertainty scenario library U, and a reliability-robustness assessment model N is established based on Six Sigma theory. The set of decision variables V under the uncertainty scenario library U is then calculated. a The reliability-robustness evaluation results of each candidate scheme; The optimization solution module is used to optimize the decision variable set V. a As optimization variables, the uncertainty scenario library U is used as a disturbance factor, and the reliability-robustness assessment result is used as the fitness function. An optimization algorithm is then used to solve the problem to obtain the optimal solution set P, and the decision variable set V is determined. a Optimization range of values ​​V for each variable p ; The subjective decision-making module is used to base decisions on the optimized value range V. p Based on the subjective creative needs of architecture, the decision variable set V is... a Making subjective decisions, denoted as V. p1 The set of undecided variables is denoted as V. p2 ; Iterative control module: used to process the set of variables to be decided, V b With the set of undecided variables V p2 Merge and return to the sensitivity analysis and variable partitioning module for iterative optimization until the iteration termination condition is met, thus completing the decision-making for all design variables.

[0020] Furthermore, a computer-readable storage medium is provided, on which a computer program is stored. When executed by a processor, the computer program implements the steps of the step-by-step decision optimization method for building design variables as described above. The computer program runs on a general-purpose computer, workstation, or cloud server. The program modules are stored in memory, executed by the processor, and interact with Building Information Modeling (BIM) software and energy consumption simulation software through input / output interfaces.

[0021] One of the above technical solutions has the following advantages or beneficial effects: This invention treats the decision-making variables as uncertainties rather than assigning them fixed default values, ensuring that each decision is made based on a full consideration of these uncertainties. This effectively avoids the problem of step-by-step optimization getting trapped in local optima. Simultaneously, this invention preserves the architect's subjective decision-making power at each step, achieving an organic combination of building performance optimization and subjective architectural creation.

[0022] Specifically, the present invention has the following beneficial effects: (1) Through sensitivity analysis, architects can clearly understand the degree of influence of each design variable on building performance and make trade-off decisions accordingly, making the building design process more scientific and transparent.

[0023] (2) The reliability-robustness assessment model based on the Six Sigma theory is adopted, which fully considers the impact of the uncertainty of the decision variables on the building performance. The optimal solution obtained has better robustness and reliability.

[0024] (3) By combining step-by-step iterative optimization with subjective decision-making, the efficiency of the optimization algorithm is guaranteed, while the architect's creative initiative is fully utilized, thereby enhancing the diversity and flexibility of the design scheme.

[0025] (4) This invention integrates the advantages of "one-step optimization" and "step-by-step optimization", so that the optimization process has both high computational efficiency and can take into account the global optimality of the design scheme and the flexibility of architectural creation.

[0026] (5) This invention employs multiple technical means—including Sobol sensitivity analysis, Latin hypercube sampling, 6σ reliability-robustness assessment, intelligent algorithm optimization, and subjective decision-making iteration—which, through organic integration, generate a synergistic effect: sensitivity analysis guides variable stratification, prioritizing the most critical variables in optimization and reducing the search space; the combination of Latin hypercube sampling and 6σ assessment significantly reduces computational costs while ensuring assessment accuracy (computational load is reduced by approximately 80% compared to Monte Carlo full sampling), enabling efficient acquisition of globally robust solutions within a step-by-step optimization framework; intelligent algorithms provide global search capabilities, avoiding local optima; subjective decision-making incorporates design wisdom, ensuring the humanistic feasibility of the solution; and the iterative mechanism ensures the completeness of the final solution. These means work together to achieve a balance between "global optimization + subjective creation," achieving unexpected technical effects that cannot be achieved by a single technical means. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating a step-by-step decision optimization method for architectural design variables, according to an exemplary embodiment. Figure 2 This is a schematic diagram of a step-by-step decision optimization system for architectural design variables, according to an exemplary embodiment. Figure 3 This is a schematic diagram illustrating a sensitivity analysis and iterative decision result according to an exemplary embodiment. Detailed Implementation

[0028] To more clearly illustrate the technical features of the present invention, the invention will be described in detail below through specific embodiments and in conjunction with the accompanying drawings. The following disclosure provides many different embodiments or examples for implementing different structures of the present invention. To simplify the disclosure of the present invention, the components and arrangements of specific examples are described below. Of course, these are merely examples and are not intended to limit the invention.

[0029] Example 1 like Figure 1 As shown in the figure, an embodiment of the present invention provides a step-by-step decision optimization method for architectural design variables, which includes the following steps: S1: Obtain the design performance target set M and design variable set V of the building scheme, and determine the value range of each design variable.

[0030] The design performance target set M = {M1, M2, ..., M} p}, where p represents the number of performance indicators; the design variable set V={V1, V2, ..., V n} where n represents the number of variables; the architectural design variables include one or more of the following: building layout and orientation, building dimensions, structural selection and material usage, window-to-wall ratio, window selection and window usage, exterior wall construction selection and material usage, roof construction selection and material usage, floor construction selection and material usage, interior wall parameters and material usage, and equipment selection and material usage; the types of design variables include categorical variables and continuous variables; the continuous variables include variables that can take values ​​continuously within a certain interval and variables that take values ​​at a finite number of discrete points.

[0031] S2: Based on the design variable set V and the design performance target set M, establish a building performance prediction proxy model L.

[0032] Step S2 specifically includes: S21: Use a pseudo-random sampling method to sample the design variable set V to generate a design scheme sample library; the pseudo-random sampling method is Latin hypercube sampling; S22: Use simulation software to calculate the performance of each sample scheme and obtain a sample dataset containing design variables and performance values; the simulation software includes energy consumption simulation software and BIM software; the performance includes one or more of the following: cooling energy demand, heating energy demand, operational carbon emissions, and implicit carbon emissions during building operation. S23: Using the sample dataset, fit a quantitative model between the design variable V and the building performance M to obtain the building performance prediction proxy model L; the building performance prediction proxy model L is a regression model, an artificial neural network model, a random forest model, or a decision tree model. The building performance prediction proxy model L is only applicable to the building scheme type determined in step S1.

[0033] S3: Based on the building performance prediction proxy model L, perform sensitivity analysis on the design variable set V, and in conjunction with architectural design requirements, divide the design variable set V into a decision variable set V. a and the set of variables to be decided V b .

[0034] Step S3 specifically includes: S31: Based on the building performance prediction proxy model L, perform sensitivity analysis on the design variable set V; S32: Based on the sensitivity analysis results and the actual situation of the architectural design, one or more of the most sensitive variables are selected as decision variables, and the remaining variables are selected as variables to be decided. At the same time, architects are allowed to include factors that have a greater impact on subjective creation into decision variables according to their subjective creative needs.

[0035] The sensitivity analysis described in step S31 uses the Sobol global sensitivity analysis method, specifically including: performing Sobol analysis to obtain the main effect index S1 of each variable i on each performance objective. i The interaction effect index and total effect index ST between any two or more variables; the most sensitive variables are identified based on the cumulative variance contribution rate: the contribution rate of each variable or interaction term to any performance target is sorted from largest to smallest, and the variables involved in the variables or interaction terms whose cumulative contribution enters the top 50% are regarded as the most sensitive factors; the threshold of the top 50% is determined based on the statistical experience that the cumulative contribution rate can usually explain most of the output variance in engineering, and this threshold can be adjusted in subsequent iterations according to actual needs; for reference, Saltelli et al. (2008) pointed out in their monograph on sensitivity analysis (Global Sensitivity Analysis: The Primer) that a cumulative contribution rate of 40% to 60% is a reasonable dividing range, and this invention takes the middle value of 50% to balance computational complexity and decision-making efficiency.

[0036] When there are multiple performance objectives, the sensitivity of each variable to all performance objectives is comprehensively assessed, and one or more variables with the highest overall sensitivity are used as decision variables.

[0037] S4: For the set of variables to be decided, V bRandom sampling is performed to generate an uncertainty scenario library U. A reliability-robustness assessment model N is established based on Six Sigma theory. The set of decision variables V under the uncertainty scenario library U is then calculated. a The reliability-robustness evaluation results of each candidate scheme.

[0038] Step S4 specifically includes: S41: For the set of variables to be decided V b Pseudo-random sampling is performed on the values ​​to generate an uncertainty scenario library U={U1, U2, ..., U...} k}, where k represents the number of uncertainty scenarios; the pseudo-random sampling adopts the Latin hypercube sampling method, and the sampling size k is more than 20 times the number of the variables to be decided; when the variables to be decided are categorical variables, all their possible values ​​are included in the uncertainty scenario library U; S42: Based on the 6σ theory, establish a reliability-robustness evaluation model N for each performance of the building scheme. The reliability evaluation result of any scheme i and any target performance m is N. mi Under p performance objectives, the reliability-robustness evaluation result N of scheme i is... i ={N 1i , N 2i , ..., N pi The mathematical expression for the reliability-robustness assessment model N is: N mi = μ mi + 6σ mi , Where, μ mi and σ mi Let M represent the performance of any architectural design scheme i under k uncertain scenarios. m The mean and standard deviation; , , M represents the performance of any architectural design scheme i under any uncertainty scenario j. m ; M represents the performance of scheme i. m The results of the reliability and robustness assessment.

[0039] The 6σ coefficient in the reliability-robustness assessment model N can be adjusted to 3σ, 4σ, or 5σ according to the reliability requirements of the design objectives. Among them, 3σ corresponds to a theoretical coverage rate of 99.73%, 4σ corresponds to 99.9937%, 5σ corresponds to 99.99994%, and 6σ corresponds to 99.99966%. The specific coefficients are determined by the designer based on the importance of the building, risk preference, and cost constraints, and are preset before optimization.

[0040] S5: Transfer the decision variable set V a As optimization variables, the uncertainty scenario library U is used as a disturbance factor, and the reliability-robustness assessment result is used as the fitness function. An optimization algorithm is then used to solve the problem to obtain the optimal solution set P, and the decision variable set V is determined. a Optimization range of values ​​V for each variable p .

[0041] Step S5 specifically includes: S51: Integrate the decision variable set V a The uncertainty scenario library U and the reliability-robustness assessment model N form a complete optimization model; S52: The decision variable set V a Using the uncertainty scenario library U as the solution variable, the reliability-robustness assessment result as the fitness function, and intelligent algorithms to find the optimal solution set P, the decision variable set V is determined. a The optimal value range V of each design variable in the process p .

[0042] The fitness function is: Fitness = [f N1 (X), f N2 (X), ..., f Np (X) ], Among them, f Nm (X) represents the reliability-robustness assessment result N. m The functional relationship between the design variable X and p is the number of performance objectives; when the performance objective is maximized, the corresponding f Nm Add a negative sign before (X).

[0043] The intelligent algorithm is a multi-objective intelligent optimization algorithm, including but not limited to NSGA-II, MOEA / D, and SPEA2. When the number of candidate solutions is small, manual optimization is used instead of the intelligent algorithm.

[0044] S6: Based on the optimized value range V p Based on the subjective creative needs of architecture, the decision variable set V is...a Making subjective decisions, denoted as V. p1 The set of undecided variables is denoted as V. p2 .

[0045] In the optimized value range V p Within the scope, the architect assigns a value to each design variable based on subjective judgment. This value is either a fixed value or a range of values. If the value is fixed, it indicates that the decision for that variable has been completed, and the variable is recorded in the set of decided variables V. p1 If the value is within a range, it indicates that the variable has a relatively small impact on the subjective creation of the architecture, and multiple values ​​are acceptable. This variable should be recorded in the undecided variable set V. p2 A decision will be made after the potential for performance improvement of the architectural design has approached zero. The architect's subjective judgment is based on one or more of the following factors: architectural spatial form requirements, aesthetic requirements, owner needs, and cost factors.

[0046] S7: The set of variables to be decided, V... b With the set of undecided variables V p2 Merge and return to step S3 for iterative optimization until the iteration termination condition is met, completing the decision on all design variables.

[0047] Step S7 specifically includes: S71: The set of variables to be decided in step S3, V b and the set of undecided variables V p2 Merge and redivide into a new set of decision variables V c and the new set of decision variables V d ; S72: Repeat steps S3 to S5 until all variables that significantly impact the design decision have been decided, and the potential for performance improvement of the building scheme approaches zero, at which point the iteration terminates. The quantitative criterion for termination of the iteration is: reaching the maximum number of iterations, or convergence when the optimal solution set distribution stabilizes. The maximum number of iterations is 50. S73: For variables that remain undecided at the time of the last optimization, a trade-off decision is made. This trade-off decision includes: when multiple design variables can be selected within their ranges, the architect independently selects a value for each variable within those ranges; variables with a relatively small impact on the architect's subjective creative process are comprehensively weighed after the last optimization to improve the diversity of design schemes.

[0048] Furthermore, the step-by-step decision optimization method for architectural design variables also includes the following steps: based on the values ​​of the design variables after the final decision is completed, a comprehensive evaluation of the performance of the architectural scheme is conducted, and the final optimized design scheme and its performance evaluation results are output.

[0049] The method is applied to a Building Information Modeling (BIM) environment. Some or all of the variables in the design variable set V are derived from the component parameters in the BIM model, and the performance indicators in the performance target set M are calculated through the interface between the BIM model and the simulation software.

[0050] Example 2 like Figure 2 As shown in the figure, an embodiment of the present invention provides a step-by-step decision optimization system for architectural design variables, comprising: The data acquisition module is used to acquire the design performance target set M and design variable set V of the building scheme, and to determine the value range of each design variable; The proxy model construction module is used to establish a building performance prediction proxy model L based on the design variable set V and the design performance target set M; The sensitivity analysis and variable partitioning module is used to perform sensitivity analysis on the design variable set V based on the building performance prediction proxy model L, and, in conjunction with architectural design requirements, partition the design variable set V into a decision variable set V. a and the set of variables to be decided V b ; The uncertainty scenario generation and evaluation module is used to evaluate the set of variables to be decided, V. b Random sampling is performed to generate an uncertainty scenario library U, and a reliability-robustness assessment model N is established based on Six Sigma theory. The set of decision variables V under the uncertainty scenario library U is then calculated. a The reliability-robustness evaluation results of each candidate scheme; The optimization solution module is used to optimize the decision variable set V. a As optimization variables, the uncertainty scenario library U is used as a disturbance factor, and the reliability-robustness assessment result is used as the fitness function. An optimization algorithm is then used to solve the problem to obtain the optimal solution set P, and the decision variable set V is determined. a Optimization range of values ​​V for each variable p ; The subjective decision-making module is used to base decisions on the optimized value range V. p Based on the subjective creative needs of architecture, the decision variable set V is... a Making subjective decisions, denoted as V. p1 The set of undecided variables is denoted as V. p2 ; Iterative control module: used to process the set of variables to be decided, V b With the set of undecided variables V p2 Merge and return to the sensitivity analysis and variable partitioning module for iterative optimization until the iteration termination condition is met, thus completing the decision-making for all design variables.

[0051] Preferably, the building performance prediction proxy model construction module includes: The sampling unit is used to sample the design variable set V using a pseudo-random sampling method to generate a design scheme sample library. The simulation calculation unit is used to calculate the performance index of each sample scheme in the design scheme sample library using simulation software, and obtain a sample dataset containing design variables and performance values. The fitting unit is used to fit the quantitative relationship between the design variable V and the building performance M based on the sample dataset, so as to obtain the building performance prediction surrogate model L.

[0052] Preferably, the sensitivity analysis and variable partitioning module adopts the Sobol global sensitivity analysis method to calculate the main effect index, interaction effect index and total effect index of each design variable for each performance target, and identifies the most sensitive variable based on the cumulative variance contribution rate.

[0053] Preferably, the uncertainty scenario generation and evaluation module uses the Latin hypercube sampling method to evaluate the set of variables to be decided, V. b Sampling is performed, with the sample size being more than 20 times the number of variables to be decided; when the variables to be decided are categorical variables, all their possible values ​​are included in the uncertainty scenario library U.

[0054] Preferably, the optimization algorithm in the optimization solution module is a multi-objective intelligent optimization algorithm, including but not limited to NSGA-II, MOEA / D, and SPEA2, and the fitness function is: Fitness = [f N1 (X), f N2 (X), ..., f Np (X) ], Among them, f Nm (X) represents the reliability-robustness assessment result N. m The functional relationship between the design variable X and p, where p is the number of performance targets.

[0055] Preferably, the iteration termination condition in the iteration control module is: reaching the maximum number of iterations, or the optimal solution set distribution is stable and the potential for performance improvement of the building scheme approaches zero; the maximum number of iterations is 50 generations.

[0056] Example 3 This invention uses the low-carbon renovation and optimization of the building envelope of an office building as an example to illustrate the specific implementation of the invention.

[0057] Step S1: Define the building design performance objectives and building design variables.

[0058] a. Establish a set of building performance objectives M = {M1, M2, ..., M} p}, where p represents the number of performance indicators of the scheme. In this embodiment, p=1, and M is the building life cycle carbon emissions, including operational carbon emissions and implicit carbon emissions.

[0059] b. Establish the design variable set V = {V1, V2, ..., V...} n}, where n represents the number of variables; and specify the range of values ​​for each design variable.

[0060] In this embodiment, the performance target set M includes one performance index: building lifecycle carbon emissions, including operational carbon emissions and implicit carbon emissions. The design variable set V contains six variables: south-facing window type V1, north-facing window type V2, east-west facing window type V3, external wall insulation thickness V4, roof insulation thickness V5, and ground insulation thickness V6. The values ​​of the design variables are shown in Table 1.

[0061] Table 1. Design variables and value ranges for low-carbon retrofitting of building envelope (example)

[0062] Step S2: Establish a building performance prediction proxy model L.

[0063] a. A pseudo-random sampling method is used to sample the values ​​of the design variables to generate a sample library of design schemes. In this embodiment, Latin hypercube sampling is performed using Python to generate a sample library with 2000 samples.

[0064] b. Calculate the performance of each sample scheme using simulation software to obtain a sample dataset containing design variables and performance values. In this embodiment, for each sample scheme, energy consumption simulation software is used to calculate the energy demand for cooling and heating during the building's operation phase. The carbon emission factor for each type of energy is obtained from the database. The consumption of each type of energy is multiplied by the carbon emission factor and then summed to obtain the building's operational carbon emissions. BIM and other software are used to extract the consumption of various materials. The carbon emission factor for each type of material is obtained from the database. The consumption of each type of material is multiplied by the carbon emission factor and then summed to obtain the building's implicit carbon emissions. For these 2000 samples, a complete dataset containing the building design variable V and the target value M is obtained.

[0065] c. Using the sample dataset, a quantitative model is fitted between the design variable V and the building performance M to obtain the building performance prediction surrogate model L. In this embodiment, the 2000 samples are divided into a training set, a validation set, and a test set in a 3:2:2 ratio to generate an artificial neural network model L for predicting building heating and cooling loads.

[0066] Step S3: Divide the design variables into two parts, one part being the "decision variable V".a A portion of it is considered as "the decision variable V". b ".

[0067] a. Sensitivity analysis of architectural design variables is performed based on the surrogate model L. In this embodiment, Sobol's method is used for sensitivity analysis, and the results are as follows: Figure 3 As shown.

[0068] b. Based on the sensitivity analysis results and the actual architectural design, select one or more of the most sensitive variables as "decision variables V". a The remaining variables are referred to as "decision variables V". b In this embodiment, a sensitivity analysis is performed before the first decision, and the most sensitive variable is ground insulation V6. Therefore, in the first decision, V6 is used for the decision, and V1~V5 are regarded as uncertain factors.

[0069] Step S4: Treat the "decision variable" as an uncertainty factor and make a decision on the "decision variable".

[0070] a. Regarding the "decision variable V" b Pseudo-random sampling is performed on the values ​​of 'U1' and 'U2' to generate a sample library of uncertainty factors U = {U1, U2, ..., U...}. k}, where k represents the number of uncertain factors related to the data. In this embodiment, during the first decision-making, Latin hypercube sampling is performed on V1~V5 views to generate an uncertainty scenario library U, which contains 1000 uncertainty scenarios, so k=1000.

[0071] b. Select reliability-robustness evaluation indicators and establish a reliability-robustness assessment model N for each performance aspect of the building scheme, consistent with the number of building performance targets, to obtain the scheme performance reliability-robustness assessment result N = {N1, N2, ..., N}. p}, where p represents the number of performance indicators for each scheme. For any architectural design scheme, its architectural performance is calculated under k uncertainty scenarios. In this embodiment, during the first decision, the decision variable V6 corresponds to 12 schemes, and for each scheme, there are 1000 carbon emission calculation results. In this example, "μ+6σ" is chosen to evaluate the reliability-robustness of the building scheme's carbon emissions, so N=μ+6σ, which means that under 1000 uncertainty scenarios, any design scheme has 1000 carbon emission calculation results, the average of these 1000 results is μ, and the standard deviation is σ. For any scheme, its carbon emission assessment result is N=μ+6σ.

[0072] Step S5: Establish a building performance optimization model to cope with uncertainties.

[0073] a. Integrating the design variable set Va An uncertainty scenario library U and a robust reliability assessment model N are used to form a complete optimized model. In this embodiment, Python is used to implement data transmission between the three.

[0074] b. Design variable set V a As the solution variables, the uncertainty scenario library U is used as a disturbance factor, and the reliability-robustness assessment result is used as the fitness function. An intelligent algorithm is used for optimization (manual optimization can be performed when the number of solutions is small) to find the optimal solution set P. That is, in the optimal solution set P, the design variable set V... a The range of values ​​V for each design variable p In this embodiment, during the first decision-making process, there were only 12 alternative options. Observation shows that when the floor insulation value is D2, the building carbon emission reliability-robustness is the best; followed by the floor insulation value D1.

[0075] Step S6: Subjective decision-making based on objective optimization results.

[0076] In V p Within this scope, the architect assigns a value to each design variable based on subjective judgment. This value can be a fixed value or a range of values. If the value is a fixed value, it indicates that the decision for that variable is complete; otherwise, it is considered that the variable has little impact on the architect's subjective creative process, and multiple values ​​are acceptable. The final decision is weighed when the potential for performance improvement in the architectural design approaches zero. The set of variables for which decisions have been made is denoted as V. p1 Variables that have a relatively small impact on the subjective creation of architecture are denoted as V. p2 In this embodiment, considering factors such as cost, the architect selected D1 as the optimal value for floor insulation.

[0077] Step S7: Change the "decision variable" V in S3 b and V p2 It is divided into two parts, one part being the "decision variable V". c A portion of it is considered as "the decision variable V". d "And repeat steps S3 to S5."

[0078] a. Once all variables significantly impacting the design decision have been decided, and the potential for performance improvement in the building scheme approaches zero, steps S3-S5 are no longer repeated. In this embodiment, the iterative process is as follows: Figure 3 As shown, the variable for the second decision was the east and west windows, and the final value was C7; the variable for the third decision was the external wall insulation, and the final value was Q6; the variable for the fourth decision was the south window, and the final value was C11; and the variable for the fifth decision was the north window, and the final value was C7.

[0079] b. For variables with minimal impact on the architect's subjective creative process, a trade-off decision is made upon completion of the final optimization. Specifically, these variables with less impact on subjective creative process provide greater design freedom and increase the diversity of design options. In this embodiment, roof insulation is the least sensitive variable, and its value is relatively flexible. After comprehensive consideration, the designer ultimately selected the value W6.

[0080] After the above five iterations, all design variables were decided, resulting in the final low-carbon renovation scheme for the building envelope.

[0081] Example 4 The difference between this embodiment and Embodiment 3 is that the performance objective set M includes two performance indicators: building lifecycle carbon emissions and indoor thermal comfort (PMV index). In this case, the sensitivity analysis in step S3 requires separate sensitivity analyses for each performance objective, and a comprehensive assessment of the sensitivity of each variable (using a weighted summation method, with weights determined by the architect based on project priorities). The fitness function in step S5 is a multi-objective function, and a multi-objective intelligent optimization algorithm such as NSGA-II is used to solve for the Pareto optimal solution set. The remaining steps are the same as in Embodiment 3. Finally, a set of Pareto frontier solutions is obtained, from which the architect selects the solution that meets their creative needs.

[0082] Example 5 The difference between this embodiment and Embodiment 3 is that the design variables include categorical variables (such as window type) and continuous variables (such as insulation thickness). In step S4, when the decision variable is a categorical variable, all its possible values ​​are included in the uncertainty scenario library U, that is, a corresponding uncertainty scenario is generated for each value of the categorical variable. In the optimization solution in step S5, the categorical and continuous variables adopt a mixed encoding method (the categorical variable adopts one-hot encoding, and the continuous variable adopts real number encoding), and optimization algorithms such as genetic algorithms that can handle mixed variables are used for solution. The remaining steps are the same as in Embodiment 3.

[0083] Example 6 This embodiment focuses on optimizing the complex shapes of large public buildings (such as airport terminals and stadiums). Performance targets include annual energy consumption, natural lighting uniformity, and structural steel consumption. Design variables include surface morphology control parameters (such as coordinates of Bézier surface control points), curtain wall grid dimensions, and structural mesh density, totaling 12 variables. Due to the complex building shape, a random forest (500 decision trees) was used as the surrogate model, increasing the sample size to 5000. Sensitivity analysis showed that the height of the surface control points and the curtain wall transmittance were the most sensitive variables. After six iterations, the main morphological parameters and building envelope parameters were gradually determined. The final solution reduced energy consumption by 7.2% while improving lighting uniformity by 12%, and increased structural steel consumption by only 2.3%, demonstrating the applicability and robustness of the method for complex building types.

[0084] Example 7 This invention provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the steps of the step-by-step decision optimization method for building design variables as described above. The computer program runs on a general-purpose computer, workstation, or cloud server. Program modules are stored in memory, executed by the processor, and interact with Building Information Modeling (BIM) software and energy consumption simulation software through input / output interfaces.

[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A step-by-step decision optimization method for architectural design variables, characterized in that, Includes the following steps: S1: Obtain the design performance target set M and design variable set V of the building scheme, and determine the value range of each design variable; S2: Based on the design variable set V and the design performance target set M, establish a building performance prediction proxy model L; S3: Based on the building performance prediction proxy model L, perform sensitivity analysis on the design variable set V, and in conjunction with architectural design requirements, divide the design variable set V into a decision variable set V. a and the set of variables to be decided V b ; S4: For the set of variables to be decided, V b Random sampling is performed to generate an uncertainty scenario library U. A reliability-robustness assessment model N is established based on Six Sigma theory. The set of decision variables V under the uncertainty scenario library U is then calculated. a The reliability-robustness evaluation results of each candidate scheme; S5: Transfer the decision variable set V a As optimization variables, the uncertainty scenario library U is used as a disturbance factor, and the reliability-robustness assessment result is used as the fitness function. An optimization algorithm is then used to solve the problem to obtain the optimal solution set P, and the decision variable set V is determined. a Optimization range of values ​​V for each variable p ; S6: Based on the optimized value range V p Based on the subjective creative needs of architecture, the decision variable set V is... a Making subjective decisions, denoted as V. p1 The set of undecided variables is denoted as V. p2 ; S7: The set of variables to be decided, V... b With the set of undecided variables V p2 Merge and return to step S3 for iterative optimization until the iteration termination condition is met, completing the decision on all design variables.

2. The step-by-step decision optimization method for architectural design variables according to claim 1, characterized in that, The design performance target set M = {M1, M2, ..., M} p }, where p represents the number of performance indicators; the design variable set V={V1, V2, ...,V n } where n represents the number of variables; the architectural design variables include one or more of the following: building layout and orientation, building dimensions, structural selection and material usage, window-to-wall ratio, window selection and window usage, exterior wall construction selection and material usage, roof construction selection and material usage, floor construction selection and material usage, interior wall parameters and material usage, and equipment selection and material usage; the types of the design variables include categorical variables and continuous variables; The continuous variables include variables that can take values ​​continuously within a certain interval and variables that take values ​​at a finite number of discrete points.

3. The step-by-step decision optimization method for architectural design variables according to claim 1, characterized in that, Step S2 specifically includes: S21: Use a pseudo-random sampling method to sample the design variable set V to generate a design scheme sample library; S22: Use simulation software to calculate the performance of each sample scheme and obtain a sample dataset containing design variables and performance values; S23: Using the sample dataset, fit a quantitative model between the design variable V and the building performance M to obtain the building performance prediction proxy model L.

4. The step-by-step decision optimization method for architectural design variables according to claim 1, characterized in that, Step S3 specifically includes: S31: Based on the building performance prediction proxy model L, perform sensitivity analysis on the design variable set V; S32: Based on the sensitivity analysis results and the actual situation of the architectural design, one or more of the most sensitive variables are selected as decision variables, and the remaining variables are selected as variables to be decided. At the same time, architects are allowed to include factors that have a greater impact on subjective creation into decision variables according to their subjective creative needs.

5. The step-by-step decision optimization method for architectural design variables according to claim 1, characterized in that, Step S4 specifically includes: S41: For the set of variables to be decided V b Pseudo-random sampling is performed on the values ​​to generate an uncertainty scenario library U={U1,U2, ..., U...} k }, where k represents the number of uncertain scenarios; S42: Based on the 6σ theory, establish a reliability-robustness evaluation model N for each performance of the building scheme. The reliability evaluation result of any scheme i and any target performance m is N. mi Under p performance objectives, the reliability-robustness evaluation result N of scheme i is... i ={N 1i N 2i , ..., N pi } 6. The step-by-step decision optimization method for architectural design variables according to claim 1, characterized in that, The mathematical expression for the reliability-robustness assessment model N is: N mi = m mi + 6p mi Where, μ mi and σ mi Let M represent the performance of any architectural design scheme i under k uncertain scenarios. m The mean and standard deviation; M represents the performance of any architectural design scheme i under any uncertainty scenario j. m ; M represents the performance of scheme i. m The results of the reliability and robustness assessment.

7. The step-by-step decision optimization method for architectural design variables according to claim 1, characterized in that, Step S5 specifically includes: S51: Integrate the decision variable set V a The uncertainty scenario library U and the reliability-robustness assessment model N form a complete optimization model; S52: The decision variable set V a As the solution variables, the uncertainty scenario library U is used as a disturbance factor, and the reliability-robustness assessment result is used as the fitness function. An intelligent algorithm is used to find the optimal solution set P, and the decision variable set V is determined. a The optimal value range V of each design variable in the middle p .

8. The step-by-step decision optimization method for architectural design variables according to any one of claims 1 to 7, characterized in that, Step S7 specifically includes: S71: The set of variables to be decided in step S3, V b and the set of undecided variables V p2 Merge and redivide into a new set of decision variables V c and the new set of decision variables V d ; S72: Repeat steps S3 to S5 until all variables that have a significant impact on the design decision have been decided and the potential for performance improvement of the building scheme approaches zero, at which point the iteration is terminated. S73: Make a trade-off decision for variables that are still undecided when the last optimization is completed.

9. A step-by-step decision optimization system for architectural design variables, characterized in that, include: The data acquisition module is used to acquire the design performance target set M and design variable set V of the building scheme, and to determine the value range of each design variable; The proxy model construction module is used to establish a building performance prediction proxy model L based on the design variable set V and the design performance target set M; The sensitivity analysis and variable partitioning module is used to perform sensitivity analysis on the design variable set V based on the building performance prediction proxy model L, and, in conjunction with architectural design requirements, partition the design variable set V into a decision variable set V. a and the set of variables to be decided V b ; The uncertainty scenario generation and evaluation module is used to evaluate the set of variables to be decided, V. b Random sampling is performed to generate an uncertainty scenario library U, and a reliability-robustness assessment model N is established based on Six Sigma theory. The set of decision variables V under the uncertainty scenario library U is then calculated. a The reliability-robustness evaluation results of each candidate scheme; The optimization solution module is used to optimize the decision variable set V. a As optimization variables, the uncertainty scenario library U is used as a disturbance factor, and the reliability-robustness assessment result is used as the fitness function. An optimization algorithm is then used to solve the problem to obtain the optimal solution set P, and the decision variable set V is determined. a Optimization range of values ​​V for each variable p ; The subjective decision-making module is used to base decisions on the optimized value range V. p Based on the subjective creative needs of architecture, the decision variable set V is... a Making subjective decisions, denoted as V. p1 The set of undecided variables is denoted as V. p2 ; Iterative control module: used to process the set of variables to be decided, V b With the set of undecided variables V p2 Merge and return to the sensitivity analysis and variable partitioning module for iterative optimization until the iteration termination condition is met, thus completing the decision-making for all design variables.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the stepwise decision optimization method for architectural design variables as described in any one of claims 1 to 8.

Images