Building design real-time optimization processing system and method based on multi-source heterogeneous data

Abstract

The invention discloses a building design real-time optimization processing system and method based on multi-source heterogeneous data, and relates to the technical field of building design optimization. A building design real-time optimization processing system based on multi-source heterogeneous data comprises a data acquisition module, a design optimization module, an alternative acquisition module, an execution analysis module, a comprehensive evaluation module, a feedback optimization module and a building repair module. According to the method, a building design optimization model based on a convolutional neural network is constructed, deep feature extraction is performed on multi-source heterogeneous data obtained in a building stage of a building object, and matching and splicing are performed in combination with a structure template and a function template, so that an initial scheme set meeting environmental conditions, specification requirements and user function requirements is generated; and the building design generation efficiency is obviously improved.

Description

Technical Field

[0001] This invention relates to the field of architectural design optimization technology, and in particular to a real-time architectural design optimization processing system and method based on multi-source heterogeneous data. Background Technology

[0002] With the increasing complexity and customization demands of construction projects, architectural design faces challenges from multiple aspects. On the one hand, during the construction phase, it is necessary to comprehensively consider a large amount of heterogeneous data, such as environmental conditions, building codes, construction progress, and user functional requirements. On the other hand, after a building is completed and put into use, its functional status and structural performance deviate over time, requiring dynamic adjustments and local repairs to maintain its safety and usability.

[0003] Traditional architectural design processes mostly rely on experience-based rules or static BIM parameter-driven methods, which are usually finalized in one go and make it difficult to respond promptly to real-time construction feedback or performance degradation during the operation phase.

[0004] Taking real-world engineering scenarios as an example, during urban renewal, complex site development, or the construction of large-scale building complexes, changes in user needs often lead to the restructuring of spatial functions; on-site construction progress deviates from the original plan; some areas have been completed but the overall scheme needs adjustment; and existing buildings suffer from structural aging, excessive energy consumption, or functional degradation. In these scenarios, the lack of a real-time optimization mechanism and feedback loop based on multi-source data can lead to design delays, uncoordinated schemes, and even resource waste and functional failure. Summary of the Invention

[0005] This invention proposes a real-time optimization method and system for building design based on multi-source heterogeneous data. To address the aforementioned issues, a design optimization model with convolutional neural networks as its core is constructed and organically integrated with functional assessment, structural strength analysis, executability matching, real-time feedback mechanisms, and repair paths during the construction phase. This provides intelligent, flexible, and controllable optimization support for building projects throughout their entire lifecycle.

[0006] Real-time optimization methods for architectural design based on multi-source heterogeneous data include: Acquire multi-source heterogeneous data of building objects, including environmental data, specification data, construction progress data and user demand data during the construction phase, or structural surveying data, condition monitoring data and renovation demand data during the completed phase; For building objects in the construction phase, their multi-source heterogeneous data are input into a design optimization model based on convolutional neural networks, and an initial set of schemes is obtained through template matching, template splicing, and functional evaluation. Obtain the structural strength evaluation value of each initial scheme and its functional evaluation value in the design optimization model. After weighted calculation, obtain the frame evaluation value. Select the three initial schemes with the top frame evaluation values ​​in the initial scheme set as candidate schemes. The candidate schemes also include three initial schemes with structural strength evaluation values ​​exceeding a preset threshold and top three functional evaluation values. The building strength values ​​of the alternative schemes are divided into three levels. The building strength values ​​consist of structural strength assessment values ​​and material strength assessment values. Based on the building strength values ​​of the three levels, the construction technology and structural materials corresponding to the material strength assessment values ​​are selected, and the feasibility assessment value of the building strength value of each level of the alternative scheme under the corresponding construction technology and structural materials is obtained. The building strength value, functional evaluation value and executability evaluation value of the alternative schemes are weighted according to user preferences to obtain a comprehensive evaluation value and obtain the final building design scheme; Based on construction progress data and on-site anomaly feedback, a process of re-acquiring architectural design schemes is triggered to achieve real-time optimization of architectural design schemes.

[0007] As a preferred embodiment of the present invention, the design optimization model based on convolutional neural networks includes: Input layer: Used to receive multi-source heterogeneous data, which is then converted into multi-source heterogeneous feature tensors through embedding encoding and normalization. Convolutional matching layer: The convolutional kernel performs feature extraction and matching operations on the multi-source heterogeneous feature tensor, and outputs each matched structural template and its sub-region and the matching feature map between the functional template, which is used to represent the structural template and its corresponding set of pre-selected functional templates. The convolutional kernel is composed of pre-trained structural templates and functional templates. Feature splicing layer: Based on the matching feature map, it combines and splices each sub-region of the structural template with its corresponding set of pre-selected functional templates to generate evaluation schemes covering all sub-regions one by one, and forms a set of evaluation schemes, which is then input into the functional evaluation layer. Functional evaluation layer: Perform independent functional evaluation, adjacent functional evaluation and global functional evaluation on each scheme to be evaluated, and calculate the functional evaluation value by weighting the evaluation values ​​of the three types of evaluations; Output layer: Used to output the schemes to be evaluated that meet the preset threshold as an initial scheme set.

[0008] As a preferred embodiment of the present invention, the functional evaluation layer includes: Independent Function Evaluation Unit: Used to independently evaluate the functional templates of each sub-region in the evaluation scheme and obtain the corresponding functional evaluation values; sum the evaluation values ​​of similar functions to construct a functional matrix and compare it with the demand matrix extracted based on multi-source heterogeneous data; when the value of the corresponding item in the functional matrix is ​​greater than or equal to the value of the corresponding item in the demand matrix, it is considered to meet the requirements; by calculating the proportion of the number of compliant items to the total number of valid items in the demand matrix, the quantitative independent functional evaluation value is obtained. Adjacent Function Evaluation Unit: Used to evaluate the functional configuration relationship between adjacent sub-regions in the scheme to be evaluated, determine its rationality in terms of spatial layout and functional connection, and convert the evaluation results into adjacent function evaluation values; Global Functional Evaluation Unit: Used to comprehensively evaluate the energy efficiency, spatial layout coordination and functional integrity of the entire scheme to be evaluated, and output the corresponding global functional evaluation value.

[0009] As a preferred technical solution of the present invention, the structural template is composed of several sub-regions with tolerance space ranges. The pre-selected functional template is obtained by matching within the tolerance space range of the corresponding sub-region based on the input multi-source heterogeneous feature tensor. During the matching process, the functional template is filtered according to the spatial position of the sub-region and its corresponding functional preference weight. The functional preference weight is initially set according to the spatial position of the sub-region and can be adjusted according to user needs.

[0010] As a preferred technical solution of the present invention, the structural template further includes: during the process of triggering the reacquisition of the architectural design scheme, setting the sub-area that has been constructed to a fixed state and keeping its corresponding structural and functional template configuration unchanged; for the sub-area that is not under construction, continuing to perform functional template matching and screening within its original tolerance space, wherein the screening process is still based on its spatial location and functional preference weight.

[0011] As a preferred embodiment of the present invention, the acquisition of the framework evaluation value includes: For each initial scheme, a structural strength assessment is performed to obtain its structural strength assessment value, which is used to represent the strength performance coefficient of the scheme in terms of structural layout, construction rationality, and stress path. The functional assessment value of the initial scheme obtained in the design optimization model is obtained to measure the scheme's performance in terms of space utilization, functional adaptability, and overall performance. The structural strength assessment value and the functional assessment value are weighted according to preset weight coefficients to obtain the frame assessment value of the initial scheme. The weight coefficients are configured according to user preferences or building uses and support dynamic adjustment.

[0012] As a preferred embodiment of the present invention, the step of dividing the building strength values ​​of the alternative schemes into three levels includes: Three building strength levels are preset: high strength, medium strength, and low strength, each corresponding to a fixed building strength target value. For each alternative scheme, under each strength level, the required material strength assessment value is calculated by dividing the building strength target value of the scheme by the structural strength assessment value. Under the premise of meeting the material strength assessment value, the optimal combination of construction technology and structural materials is selected from the alternative construction technology and structural materials. Based on the selected combination of construction technology and structural materials, the feasibility of each alternative scheme is evaluated under different strength levels. The feasibility evaluation includes the estimation of construction period and resource utilization efficiency. The evaluation process is based on historical data fitting.

[0013] As a preferred technical solution of the present invention, the real-time optimization processing method for building design based on multi-source heterogeneous data further includes: for a building object in the completed stage, based on its structural mapping data and status monitoring data, obtaining the building design scheme and the corresponding building strength value, and inputting them into the design optimization model to obtain the corresponding functional evaluation value; based on the functional evaluation value and the building strength value, identifying sub-regions in the structural template of the building design scheme that have degraded function or insufficient strength, and performing local functional template re-matching and structural template splicing operations to generate the corresponding repair design scheme.

[0014] A real-time optimization system for building design based on multi-source heterogeneous data includes: The data acquisition module is used to acquire multi-source heterogeneous data of building objects; The design optimization module is used to input multi-source heterogeneous data of a building object into a design optimization model built on a convolutional neural network when the building object is in the construction phase, and output an initial set of schemes. The alternative acquisition module is used to obtain the structural strength evaluation value and frame evaluation value of each initial scheme, and to filter alternative schemes in combination with the functional evaluation value; The execution analysis module is used to obtain the feasibility evaluation value of the building strength value of each grade of the alternative scheme under the corresponding construction technology and structural materials; The comprehensive evaluation module is used to obtain the comprehensive evaluation value of the alternative schemes and determine the final architectural design scheme; The feedback optimization module is used to trigger the process of re-acquiring architectural design schemes based on construction progress data or on-site anomaly feedback. The repair module is used to generate corresponding repair design schemes when a building object is in the construction stage.

[0015] The present invention has the following advantages: This invention constructs an architectural design optimization model based on convolutional neural networks, performs deep feature extraction on multi-source heterogeneous data acquired during the construction phase of a building object, and combines structural templates and functional templates for matching and splicing, thereby generating an initial set of schemes that meet environmental conditions, regulatory requirements, and user functional needs, significantly improving the efficiency of architectural design generation.

[0016] This invention improves the scientific nature and overall performance of scheme selection by conducting structural strength and functional assessments on each initial scheme and performing weighted calculations to obtain the frame assessment value.

[0017] This invention divides building strength values ​​into three preset levels, then reverse-engineers the material strength assessment values ​​required to meet each strength level, and selects corresponding construction techniques and structural materials while meeting strength requirements. Combined with the results of the feasibility assessment, it realizes the selection of alternative solutions that take feasibility into consideration, thereby improving the engineering practicality and resource allocation efficiency of the design results.

[0018] This invention introduces a real-time feedback mechanism that combines construction progress data and on-site anomaly information during the construction process to trigger updates to the architectural design scheme. It keeps the constructed areas unchanged and only dynamically optimizes and adjusts the unconstructed areas, ensuring flexible response and continuous control of the architectural design scheme during the actual construction process.

[0019] This invention is applicable to the status assessment and local repair design of building objects during the construction phase. Based on structural surveying data and status monitoring data, it obtains the functional assessment value and building strength value of the current building design scheme, and then identifies sub-areas with functional degradation or insufficient structural performance. It then performs targeted local reconstruction operations of functional templates and structural templates to generate repair design schemes. This expands the application capability of this invention in the building operation and maintenance phase, and realizes intelligent optimization throughout the entire life cycle from design to operation. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are only schematic diagrams of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort. Figure 1 This is a schematic diagram of the structure of the real-time optimization processing system for building design based on multi-source heterogeneous data used in an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0022] Example 1: A real-time optimization method for building design based on multi-source heterogeneous data, comprising the following steps: Step S1: Obtain multi-source heterogeneous data of the building object, including environmental data, specification data, construction progress data and user demand data during the construction phase, or structural survey data, condition monitoring data and renovation demand data during the completed phase; The multi-source heterogeneous data refers to a data set originating from different acquisition entities, possessing multiple data structures and attribute dimensions, and used to support the architectural design optimization process. Specifically, it includes, but is not limited to, structured data, time-series data, image data, and expert rule sets. Based on the different stages of the building object, it is divided into two main categories: construction stage data and completed stage data. Multi-source heterogeneous data during the construction phase includes: Environmental data refers to natural and surrounding conditions that affect building site selection, structural design, and energy consumption performance. It is used for the generation of structural templates and the initial selection of building orientation and layout. Sources include geological survey reports, real-time or historical data interfaces from meteorological bureaus, and environmental monitoring equipment.

[0023] Example fields include: annual average wind speed, extreme wind load value (unit: m / s), peak ground acceleration, standard value of soil bearing capacity, and distribution map of surrounding obstructions (for solar radiation analysis).

[0024] Standard data refers to mandatory and recommended design standards applicable to the geographical location, building use and grade of this project. These standards are used for rule verification in functional assessment and to construct the constraint boundaries of the demand matrix. Sources include building standard databases (GB series), local implementation rules and industry guidelines.

[0025] Examples: evacuation distance requirements in building fire protection design codes; minimum clear height limits for kitchens and bathrooms in residential buildings; energy efficiency limits for office buildings.

[0026] Construction progress data refers to the progress information reflecting the current construction status of a building project. It is used to participate in the real-time adjustment mechanism of the scheme and marks the completed areas when the design scheme is re-acquired. The sources include the BIM construction management system, the site scheduling platform, and construction log records.

[0027] Example fields include: construction status by region (completed / in progress / not started); actual start and end times of construction; and logs of abnormal working conditions (such as material delays and component deviations).

[0028] User requirement data refers to the functional, layout, and performance preferences set by building users or investors. These preferences are used to convert into preference weights for functional template matching and form the reference source for the requirement matrix of the functional assessment layer. The sources are functional requirement questionnaires filled out by users, design briefs, or the results of expert interviews.

[0029] Example fields include: a list of functional spaces (e.g., 4 offices + 1 meeting room + a common tea room); the area range of functional spaces; space priority and relative location requirements (e.g., meeting rooms need to be close to the lobby); and weighted preferences for lighting, privacy, and expandability (numerical weights).

[0030] The multi-source heterogeneous data in the construction phase includes: Structural survey data: Based on the structural state information obtained from the actual measurement of the completed building, it is used to calculate the structural strength part of the building strength value, and serves as the input basis for generating repair plans. It assesses the integrity and consistency of its components. The data sources include 3D laser scanning, structural detection instruments (such as radar and infrared thermography), and BIM as-built model comparative analysis.

[0031] Example fields include: actual beam height, column cross-section dimensions, slab thickness; structural deviation (such as wall tilt angle); residual deformation value and historical deformation trend.

[0032] Status monitoring data: These are online monitoring results during building operation, used for functional degradation judgment and global assessment calculation. They are combined with functional assessment layer results to identify problem areas and originate from sensors deployed in key structural locations or functional spaces.

[0033] Example fields include: load response (stress / strain); temperature and humidity distribution; noise level; real-time energy consumption and equipment efficiency data.

[0034] Modification requirement data: User-submitted descriptions of functional changes, space reconstruction, or expansion requirements serve as the basis for triggering repair design solutions and directly participate in template re-matching. The data sources include user input systems, operation and maintenance logs, or space utilization analysis.

[0035] For example: requests to add open office areas, restrooms, or other functional changes; areas with low usage frequency are suggested to be merged or reduced; a summary of feedback on inconvenience (combined with complaint records or survey data).

[0036] Step S2: For building objects in the construction phase, input their multi-source heterogeneous data into a design optimization model based on convolutional neural networks, and obtain an initial set of schemes through template matching, template splicing and functional evaluation; The design optimization model is built upon a convolutional neural network (CNN). Its core objective is to identify the optimal structural and functional combination scheme based on multi-source heterogeneous feature inputs and output an evaluable building configuration. This model possesses end-to-end learning and feature fusion capabilities, has a clear hierarchical structure, and exhibits good generalization ability and sample adaptability.

[0037] The model input is multi-source heterogeneous data, and the output is an initial (architectural design) scheme that meets multi-dimensional performance constraints, forming a closed-loop design, evaluation, and screening process. The specific data flow order is: multi-source heterogeneous data → feature tensor → convolution matching → template mapping → splicing and combination → functional evaluation → initial scheme output.

[0038] Key term definitions: Structural formwork: In architectural design, it serves as the spatial framework, defining the distribution logic and load-bearing relationships of functional areas. It consists of multiple sub-areas with tolerance spaces. For example, a typical residential structural formwork layout includes a central axis passageway and symmetrical functional units on both sides.

[0039] Functional templates: These are combined units that encapsulate rules for configuring spatial functions, such as office area templates (including meeting areas, desks, and storage areas) and restroom templates (including sinks, toilets, and urinals). Each template has specific space occupancy, functional adaptability, and usage requirements.

[0040] Tolerance space range: This represents the acceptable functional layout floating space for each sub-region within the structural template, supporting flexible matching of different functional templates within this range. Essentially, it is a set of spatial boundaries and directional constraints used to constrain spatial compatibility when functional modules are combined.

[0041] The design optimization model based on convolutional neural networks includes: Input layer: This layer receives multi-source heterogeneous data, which is then converted into multi-source heterogeneous feature tensors through embedding encoding and normalization. This layer uses three mechanisms—position embedding, attribute embedding, and canonical vector encoding—to fuse environmental, canonical, and user-related data into a unified tensor structure, ensuring synchronous mapping between space and semantics. After embedding, the data is uniformly converted into standard floating-point format input to the convolutional matching layer.

[0042] Convolutional Matching Layer: This layer performs feature extraction and matching operations on multi-source heterogeneous feature tensors using convolutional kernels. It outputs a matching structural template for each matched tensor, along with matching feature maps between its sub-regions and functional templates. These feature maps represent the structural template and its corresponding set of pre-selected functional templates. The convolutional kernels are composed of pre-trained structural and functional templates. A dual-channel convolution mechanism is employed: one channel loads structural template parameters, and the other channel loads functional template parameters. By performing local convolution operations on the tensor in each spatial region, spatial functional adaptation relationships are extracted, and the output matching feature maps are used for subsequent concatenation judgment.

[0043] Feature splicing layer: Based on the matching feature map, this layer combines and splices each sub-region of the structural template with its corresponding set of pre-selected functional templates, generating evaluation schemes covering all sub-regions one by one, forming a set of evaluation schemes, which is then input into the functional evaluation layer. Using the sub-regions of the structural template as axes, the matched functional templates are enumerated and spliced ​​to construct multiple complete building schemes to be evaluated. During the splicing process, boundary consistency checks are performed on the connection relationships (such as passageways and shared facilities) between functional templates to ensure the legality of the layout.

[0044] Functional evaluation layer: Independent functional evaluation, adjacent functional evaluation and global functional evaluation are performed on each scheme to be evaluated. The evaluation values ​​of the three types of evaluation are weighted to obtain the functional evaluation value. A triple path evaluation mechanism is adopted to quantify the sub-region matching quality (independence), adjacent functional coordination (adjacency) and overall layout integrity (globality). The comprehensive evaluation is carried out through adjustable weighting coefficients, and finally the set of schemes that pass the threshold conditions is selected.

[0045] Output layer: Used to output the schemes to be evaluated that meet the preset threshold as an initial scheme set.

[0046] The functional evaluation layer includes: Independent Function Evaluation Unit: Used to independently evaluate the functional templates of each sub-region in the evaluation scheme and obtain the corresponding functional evaluation values; sum the evaluation values ​​of similar functions to construct a functional matrix and compare it with the demand matrix extracted based on multi-source heterogeneous data; when the value of the corresponding item in the functional matrix is ​​greater than or equal to the value of the corresponding item in the demand matrix, it is considered to meet the requirements; by calculating the proportion of the number of compliant items to the total number of valid items in the demand matrix, the quantitative independent functional evaluation value is obtained. The adjacency function assessment unit is used to assess the adjacency of functional configuration relationships between adjacent sub-regions in the assessment scheme and obtain the corresponding adjacency function assessment value. Based on the sub-region adjacency relationship diagram in the structural template, and combined with the spatial connectivity parameters of the pre-selected functional template, it analyzes the spatial logical characteristics of adjacent sub-regions, such as accessibility, line-of-sight continuity, acoustic or airflow interference. During the assessment process, a functional connection matrix is ​​constructed and compared with the set adjacency rationality standard matrix. When the functional connection relationship matches the standard, it is judged as a reasonable connection. By calculating the proportion of reasonable connection items to the total number of connection items, the adjacency function assessment value is quantitatively output.

[0047] Global Functional Evaluation Unit: Used to perform a global performance evaluation of the entire scheme to be evaluated and obtain the corresponding global functional evaluation value; the evaluation content includes but is not limited to: overall energy efficiency of the scheme (based on heat load prediction and ventilation path simulation), spatial layout coordination (based on module hierarchy and circulation path rationality), and functional integrity (whether it includes functional areas that meet all items of the functional requirements matrix); the above indicators are normalized to form a global evaluation vector, and a global functional evaluation value is obtained based on a weighted mechanism, wherein the weights are configured according to the building use, user preferences and energy-saving policy rules.

[0048] The structural template is a spatial structural prototype used to define the functional area division in the architectural design scheme. It consists of several sub-regions with adjustable spatial tolerance. Each sub-region contains the following information fields: spatial boundary definition, acceptable function set, initial function preference, adjacent area list, and structural rigidity level. The tolerance space range defines the allowable fluctuation range of the sub-region in terms of area, shape, opening direction, and connection boundary, which is used to support the dynamic matching needs of diverse functional combinations under the same structural template. During the functional template matching process, based on the semantic requirements, site constraints, and specification adaptation requirements in the multi-source heterogeneous feature tensor, the functional template combination scheme that meets the tolerance conditions is automatically selected. The matching strategy introduces functional preference weights to enhance the difference in functional matching priority between core areas and auxiliary areas. The initial value of the functional preference weight is based on the default setting of the structural template library and supports dynamic adjustment according to user instructions or design goals, realizing a semi-automatic functional template allocation mechanism.

[0049] During the process of triggering the reacquisition of architectural design schemes, sub-areas that have completed construction are set to a fixed state, and their corresponding structural and functional template configurations remain unchanged; for sub-areas that are not under construction, functional template matching and filtering continue to be performed within their original tolerance space, and the filtering process is still based on their spatial location and functional preference weights.

[0050] During the construction phase, if any abnormal construction progress or completion of construction in a localized area is detected, an automatic mechanism to re-acquire the design scheme is triggered. In this mechanism: the structural and functional template configurations for completed areas are locked; template matching for uncompleted areas is still performed based on the original tolerance space and preference weights; the design for the localized area is regenerated and spliced ​​with the original completed structure to ensure design continuity. This mechanism significantly improves the collaborative efficiency of the design and construction process and supports a dynamic deployment mode of designing while construction is underway.

[0051] Step S3: Obtain the structural strength evaluation value of each initial scheme and its functional evaluation value in the design optimization model. After weighted calculation, obtain the frame evaluation value. Select the top three initial schemes with the highest frame evaluation values ​​in the initial scheme set as candidate schemes. The candidate schemes also include three initial schemes with structural strength evaluation values ​​exceeding a preset threshold and top three functional evaluation values. The obtained framework evaluation values ​​include: For each initial scheme, a structural strength assessment is performed to obtain its structural strength assessment value, which is used to represent the strength performance coefficient of the scheme in terms of structural layout, construction rationality, and stress path. The functional assessment value of the initial scheme obtained in the design optimization model is obtained to measure the scheme's performance in terms of space utilization, functional adaptability, and overall performance. The structural strength assessment value and the functional assessment value are weighted according to preset weight coefficients to obtain the frame assessment value of the initial scheme. The weight coefficients are configured according to user preferences or building uses and support dynamic adjustment.

[0052] The structural strength assessment value is a comprehensive strength performance index based on the component layout and force path analysis of each sub-region in the structural template. The assessment content includes, but is not limited to: component continuity, node stiffness coordination, force transmission path stability and material compatibility. The parameters used for the strength assessment are derived from the structural configuration data output by the design optimization model, empirical coefficients of similar structures in the code database, and the simulation results of preset load conditions. A simplified structural evaluation model using finite element analysis (FEA) is employed to generate structural strength scores for each initial scheme. The scores are standardized to the [0,1] interval to represent relative strength levels. The functional evaluation value, output from the functional evaluation layer in step S2, integrates independent functions, adjacent functions, and global functions, representing the scheme's adaptability in terms of building function adaptation, spatial layout, and performance objectives. The frame evaluation value is obtained by weighting the structural strength evaluation value and the functional evaluation value proportionally. The weighting is determined using the following two methods: (1) Standard configuration based on the purpose of the building project (e.g., the ratio of structural strength to function evaluation is 3:2 for residential buildings and 4:1 for hospitals). (2) Based on dynamic input of user preferences, the system supports adjusting weight coefficients through a graphical interface and transmitting the data back to the framework evaluation module in real time. The weighting method is a linear normalized weighted model, expressed as: F frame =α×F struct +(1-α)×F func ,in α As a weight for structural strength assessment, F struct This is the structural strength assessment value. F func This is a functional evaluation value. F frame This is the framework evaluation value.

[0053] After evaluating all initial schemes within a framework, the top three schemes with the highest framework evaluation values ​​are selected as candidate schemes. Furthermore, to prevent schemes with weak structural performance but high functional scores from entering the candidate set, an additional conditional screening strategy is implemented: if a scheme's structural strength evaluation value exceeds a preset strength threshold (e.g., 0.85) and its functional evaluation value ranks among the top three, that scheme is also included in the candidate scheme set, forming a cross-selection mechanism to enhance the robustness and diversity of the evaluation system. The preset structural strength threshold is determined based on the project's importance level and recommended values ​​from specifications, empirically fitted using historical project data, and supports custom configuration by the project team.

[0054] Step S4: Divide the building strength values ​​of the candidate schemes into three levels. The building strength values ​​consist of structural strength assessment values ​​and material strength assessment values. Select the construction technology and structural materials corresponding to the material strength assessment values ​​of the three levels of building strength values, and obtain the feasibility assessment value of the building strength value of each level of the candidate scheme under the corresponding construction technology and structural materials. Definition and Grading of Building Strength Value: The building strength value is a composite strength index used to reflect the overall load-bearing capacity of the design scheme under the structural system design and material performance configuration. This value consists of the following two dimensions: structural strength assessment value, the result obtained in step S3 above, representing the stress capacity and layout rationality of the structural design itself; and material strength assessment value, representing the minimum material performance level required to meet the target building strength, generated by combining material compressive strength, elastic modulus, shear modulus, and other indicators. The division of the building strength values ​​of the alternative schemes into three levels includes: Three building strength levels are preset: high strength, medium strength, and low strength. Each level corresponds to a fixed building strength target value. The levels are revised based on actual requirements such as the project's seismic resistance level, load level, or service life. For each alternative scheme, under each strength level, the required material strength assessment value is calculated by dividing the scheme's building strength target value by the structural strength assessment value. When the calculated material strength assessment value is lower than the minimum material strength standard, the combination is automatically eliminated to prevent non-compliant items from entering the implementation assessment stage.

[0055] Under the premise of meeting the material strength assessment value, the optimal combination of construction technology and structural materials is selected from the alternative construction technology and structural materials. Based on the selected combination of construction technology and structural materials, the feasibility of each alternative scheme is evaluated at different strength levels. The feasibility evaluation includes the estimation of construction period and resource utilization efficiency. The evaluation process is based on historical data fitting.

[0056] The performance evaluation includes: schedule estimation, which is evaluated by combining historical schedule data and process interference factors in the construction progress module; resource utilization efficiency, which analyzes resource indicators such as the proportion of reusable components, material saving rate, and energy consumption assessment in the scheme; after standardizing the evaluation indicators, they are weighted and summarized to obtain the final performance score, and users can adjust the weights to adapt to actual project preferences.

[0057] Step S5: The building strength value, functional evaluation value and executability evaluation value of the alternative schemes are weighted according to user preferences to obtain a comprehensive evaluation value and obtain the final building design scheme; The building strength value, functional assessment value, and implementability assessment value are respectively derived from: building strength value ( R strength The overall load-bearing capacity and safety performance of the scheme are calculated by superimposing structural strength and material strength in step S4; the functional evaluation value ( R function The output of the functional evaluation layer in step S2 includes a weighted composite of independent functions, adjacent functions, and global functional evaluation; the executive evaluation value ( R economic The results of step S4, which evaluates the combination of construction techniques and structural materials for each strength level, cover the dimensions of construction period and resource efficiency.

[0058] The above three indicators constitute the core performance vector of the candidate building schemes { R strength ,R function , R economic All values ​​are standardized to the [0,1] interval for easier quantization calculation.

[0059] A user preference weight configuration interface is provided to allow designers, investors, or regulators to express the different levels of focus they place on different evaluation objectives. This set of weights is denoted as: W={w strength ,w function ,w economic } ,in w strength As the weight of the building strength value, w function The weights for the functional evaluation values, w economic The weights are the performance evaluation values, and the sum of the three weights is 1.

[0060] Weights can be set in the following ways: Default template configuration, providing standardized weight settings based on building type (such as residential, office, industrial, medical buildings, etc.); Manual adjustment mode, where users can set the weight of each indicator through a slider in a graphical interface and receive real-time feedback on the calculation results; Automatic learning optimization (optional): Based on the user's historical evaluation data, the optimal weight distribution is fitted through Bayesian optimization or genetic algorithm iteration.

[0061] A weighted additive approach is used to fuse multiple indicators to obtain a comprehensive evaluation value, representing the overall performance of the current architectural design scheme in terms of structural safety, functional compatibility, and feasibility. This comprehensive evaluation process can be performed separately at different intensity levels.

[0062] Based on the comprehensive evaluation scores of all candidate schemes, the scheme with the highest comprehensive evaluation score is selected as the final architectural design scheme. If multiple schemes have similar evaluation scores (the difference is less than a set threshold ε), the designers are prompted to conduct a manual comparison to assist in making a final decision from dimensions such as spatial layout, aesthetic style, or sustainability. At the same time, the unselected schemes with high scores are retained as design version history to support subsequent adjustments or rapid replacement in response to sudden changes.

[0063] Step S6: Based on the construction progress data and on-site anomaly feedback, trigger the process of re-acquiring the architectural design scheme to achieve real-time optimization of the architectural design scheme.

[0064] On-site anomaly feedback refers to problem data discovered during construction through monitoring systems or manual reporting, including: abnormal foundation settlement; environmental disturbances (such as sudden climate changes or inadequate noise control); non-compliant materials or equipment malfunctions; on-site safety risks or unforeseen structural change requirements.

[0065] Triggering strategies include, but are not limited to, the following scenarios: the construction progress of a certain sub-area is lagging behind a set threshold (e.g., ≥7 days); the actual materials used at a certain construction node are inconsistent with the material strength in the design model; abnormal feedback includes suggestions for key structural changes or spatial function changes; and external systems such as environmental impact assessments and structural monitoring push correction opinions.

[0066] When any trigger condition is met, the response process is executed. First, the structural template sub-areas of the completed construction area are frozen, i.e., set to a fixed state, to prevent them from being re-matched or replaced during subsequent optimization, ensuring that the constructed parts are not modified. For sub-areas still in the unconstructed state, based on the current input data (including the latest construction data and anomaly feedback), the following actions are performed: functional template matching; sub-area splicing; functional and strength assessment; comprehensive assessment value calculation; dynamic weight adjustment (optional). Users can reset the comprehensive assessment weights based on the priority objectives of the current construction stage (such as accelerating the schedule, reducing material usage, etc.). A new architectural design scheme is generated, and updated construction drawings and BIM instruction documents are output for rapid deployment by the construction team.

[0067] After updating the plan, the data inputs used in this round of reconstruction (such as specific lagging areas and reasons for functional changes) and reconstruction results (such as changed modules and improved executability) will be written back into the architectural design database to form a traceable design evolution path for subsequent quality analysis or data training optimization models.

[0068] Step S7: For buildings in the completed stage, based on their structural survey data and status monitoring data, obtain the building design scheme and the corresponding building strength value, and input them into the design optimization model to obtain the corresponding functional evaluation value; based on the functional evaluation value and the building strength value, identify the sub-regions in the structural template of the building design scheme that have degraded function or insufficient strength, and perform local functional template re-matching and structural template splicing operations to generate the corresponding repair design scheme.

[0069] Functional assessment: Through the functional assessment layer, the existing spatial distribution is compared with the functional requirements to determine whether there is insufficient functional load (such as a serious shortage of classroom capacity in the teaching building), whether the spatial adjacency configuration is unreasonable (such as the surgical area being close to the pollution source), and whether the energy efficiency deviates from the original setting.

[0070] Strength assessment: Based on condition monitoring data and the initial building design strength target, the current building strength value is recalculated. If the actual structural performance of a local area is found to be lower than the original design safety margin threshold (e.g., lower than 80% of the design strength), it is marked as a weak area.

[0071] The above two types of assessment results are combined to construct a structural formwork risk marking map, highlighting areas of functional or strength degradation. For the marked degradation sub-regions, the following optimization process is performed only on these areas: retain the fixed sub-regions in the original structural formwork frame; rematch the functionally degraded sub-regions with the most suitable set of functional formwork for the current usage requirements; recalculate the acceptable material strength threshold for sub-regions with insufficient strength, and recommend local reinforcement strategies (such as adding steel structures, carbon fiber reinforcement, etc.) with reference to the latest construction technology; automatically assess the structural compatibility between the new formwork and the surrounding existing components before splicing the new formwork to ensure the stability of the new and old connection nodes; and integrate the new formwork into the original structural formwork to form a new repair design scheme.

[0072] After the repair design is generated, the system outputs an updated structural template model and functional layout diagram; a construction process specification for the repaired area; a material requirement forecast table; and a report comparing the functional and strength improvements before and after the repair, quantifying the repair benefits. All repair data is synchronously written to the historical version database for subsequent building lifecycle management (such as operation and maintenance planning, secondary renovation) or for training and optimizing models.

[0073] Example 2: Real-time optimization processing system for building design based on multi-source heterogeneous data (see [link]). Figure 1 As shown, it includes the following modules: The data acquisition module is used to acquire multi-source heterogeneous data of building objects; The design optimization module is used to input multi-source heterogeneous data of a building object into a design optimization model built on a convolutional neural network when the building object is in the construction phase, and output an initial set of schemes. The alternative acquisition module is used to obtain the structural strength evaluation value and frame evaluation value of each initial scheme, and to filter alternative schemes in combination with the functional evaluation value; The execution analysis module is used to obtain the feasibility evaluation value of the building strength value of each grade of the alternative scheme under the corresponding construction technology and structural materials; The comprehensive evaluation module is used to obtain the comprehensive evaluation value of the alternative schemes and determine the final architectural design scheme; The feedback optimization module is used to trigger the process of re-acquiring architectural design schemes based on construction progress data or on-site anomaly feedback. The repair module is used to generate corresponding repair design schemes when a building object is in the construction stage.

[0074] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for real-time optimization of architectural design based on multi-source heterogeneous data, characterized in that, include: Acquire multi-source heterogeneous data of building objects, including environmental data, specification data, construction progress data and user demand data during the construction phase, or structural surveying data, condition monitoring data and renovation demand data during the completed phase; For building objects in the construction phase, their multi-source heterogeneous data are input into a design optimization model based on convolutional neural networks, and an initial set of schemes is obtained through template matching, template splicing, and functional evaluation. Obtain the structural strength evaluation value of each initial scheme and its functional evaluation value in the design optimization model. After weighted calculation, obtain the frame evaluation value. Select the three initial schemes with the top frame evaluation values ​​in the initial scheme set as candidate schemes. The candidate schemes also include three initial schemes with structural strength evaluation values ​​exceeding a preset threshold and top three functional evaluation values. The building strength values ​​of the alternative schemes are divided into three levels. The building strength values ​​consist of structural strength assessment values ​​and material strength assessment values. Based on the building strength values ​​of the three levels, the construction technology and structural materials corresponding to the material strength assessment values ​​are selected, and the feasibility assessment value of the building strength value of each level of the alternative scheme under the corresponding construction technology and structural materials is obtained. The building strength value, functional evaluation value and executability evaluation value of the alternative schemes are weighted according to user preferences to obtain a comprehensive evaluation value and obtain the final building design scheme; Based on construction progress data and on-site anomaly feedback, a process of re-acquiring architectural design schemes is triggered to achieve real-time optimization of architectural design schemes.

2. The method for real-time optimization of architectural design based on multi-source heterogeneous data according to claim 1, characterized in that, The design optimization model based on convolutional neural networks includes: Input layer: Used to receive multi-source heterogeneous data, which is then converted into multi-source heterogeneous feature tensors through embedding encoding and normalization. Convolutional matching layer: The convolutional kernel performs feature extraction and matching operations on the multi-source heterogeneous feature tensor, and outputs each matched structural template and its sub-region and the matching feature map between the functional template, which is used to represent the structural template and its corresponding set of pre-selected functional templates. The convolutional kernel is composed of pre-trained structural templates and functional templates. Feature splicing layer: Based on the matching feature map, it combines and splices each sub-region of the structural template with its corresponding set of pre-selected functional templates to generate evaluation schemes covering all sub-regions one by one, and forms a set of evaluation schemes, which is then input into the functional evaluation layer. Functional evaluation layer: Perform independent functional evaluation, adjacent functional evaluation and global functional evaluation on each scheme to be evaluated, and calculate the functional evaluation value by weighting the evaluation values ​​of the three types of evaluations; Output layer: Used to output the schemes to be evaluated that meet the preset threshold as an initial scheme set.

3. The method for real-time optimization of building design based on multi-source heterogeneous data according to claim 2, characterized in that, The functional evaluation layer includes: Independent Function Evaluation Unit: Used to independently evaluate the functional templates of each sub-region in the evaluation scheme and obtain the corresponding functional evaluation values; sum the evaluation values ​​of similar functions to construct a functional matrix and compare it with the demand matrix extracted based on multi-source heterogeneous data; when the value of the corresponding item in the functional matrix is ​​greater than or equal to the value of the corresponding item in the demand matrix, it is considered to meet the requirements; by calculating the proportion of the number of compliant items to the total number of valid items in the demand matrix, the quantitative independent functional evaluation value is obtained. Adjacent Function Evaluation Unit: Used to evaluate the functional configuration relationship between adjacent sub-regions in the scheme to be evaluated, determine its rationality in terms of spatial layout and functional connection, and convert the evaluation results into adjacent function evaluation values; Global Functional Evaluation Unit: Used to comprehensively evaluate the energy efficiency, spatial layout coordination and functional integrity of the entire scheme to be evaluated, and output the corresponding global functional evaluation value.

4. The method for real-time optimization of building design based on multi-source heterogeneous data according to claim 2, characterized in that, The structural template consists of several sub-regions with tolerance space ranges. The pre-selected functional templates are obtained by matching within the tolerance space range of the corresponding sub-regions based on the input multi-source heterogeneous feature tensor. During the matching process, the functional templates are filtered according to the spatial position of the sub-regions and their corresponding functional preference weights. The functional preference weights are initially set according to the spatial position of the sub-regions and can be adjusted according to user needs.

5. The method for real-time optimization of building design based on multi-source heterogeneous data according to claim 4, characterized in that, The structural template also includes: during the process of triggering the reacquisition of architectural design schemes, setting the sub-areas that have been constructed to a fixed state and keeping their corresponding structural and functional template configurations unchanged; for sub-areas that are not under construction, continuing to perform functional template matching and filtering within their original tolerance space, the filtering process is still based on their spatial location and functional preference weights.

6. The method for real-time optimization of building design based on multi-source heterogeneous data according to claim 1, characterized in that, The obtained framework evaluation values ​​include: For each initial scheme, a structural strength assessment is performed to obtain its structural strength assessment value, which is used to represent the strength performance coefficient of the scheme in terms of structural layout, construction rationality, and stress path. The functional assessment value of the initial scheme obtained in the design optimization model is obtained to measure the scheme's performance in terms of space utilization, functional adaptability, and overall performance. The structural strength assessment value and the functional assessment value are weighted according to preset weight coefficients to obtain the frame assessment value of the initial scheme. The weight coefficients are configured according to user preferences or building uses and support dynamic adjustment.

7. The method for real-time optimization of architectural design based on multi-source heterogeneous data according to claim 1, characterized in that, The division of the building strength values ​​of the alternative schemes into three levels includes: Three building strength levels are preset: high strength, medium strength, and low strength, each corresponding to a fixed building strength target value. For each alternative scheme, under each strength level, the required material strength assessment value is calculated by dividing the building strength target value of the scheme by the structural strength assessment value. Under the premise of meeting the material strength assessment value, the optimal combination of construction technology and structural materials is selected from the alternative construction technology and structural materials. Based on the selected combination of construction technology and structural materials, the feasibility of each alternative scheme is evaluated under different strength levels. The feasibility evaluation includes the estimation of construction period and resource utilization efficiency. The evaluation process is based on historical data fitting.

8. The method for real-time optimization processing of building design based on multi-source heterogeneous data according to claim 1, characterized in that, Also includes: For buildings in the completed stage, based on their structural survey data and condition monitoring data, the building design scheme and corresponding building strength value are obtained and input into the design optimization model to obtain the corresponding functional evaluation value. Based on functional assessment values ​​and building strength values, sub-regions with functional degradation or insufficient strength in the structural template of the building design scheme are identified, and local functional template re-matching and structural template splicing operations are performed to generate corresponding repair design schemes.

9. A real-time optimization processing system for architectural design based on multi-source heterogeneous data, characterized in that, The system applies any one of the real-time optimization processing methods for building design based on multi-source heterogeneous data as described in claims 1 to 8, including: The data acquisition module is used to acquire multi-source heterogeneous data of building objects; The design optimization module is used to input multi-source heterogeneous data of a building object into a design optimization model built on a convolutional neural network when the building object is in the construction phase, and output an initial set of schemes. The alternative acquisition module is used to obtain the structural strength evaluation value and frame evaluation value of each initial scheme, and to filter alternative schemes in combination with the functional evaluation value; The execution analysis module is used to obtain the feasibility evaluation value of the building strength value of each grade of the alternative scheme under the corresponding construction technology and structural materials; The comprehensive evaluation module is used to obtain the comprehensive evaluation value of the alternative schemes and determine the final architectural design scheme; The feedback optimization module is used to trigger the process of re-acquiring architectural design schemes based on construction progress data or on-site anomaly feedback. The repair module is used to generate corresponding repair design schemes when a building object is in the construction stage.

Images