An intelligent interior design method and system based on AI corpus optimization

Abstract

The present invention relates to the field of architectural design technology, and specifically to an intelligent interior design method and system based on AI corpus optimization, comprising: receiving user-inputted indoor functional zoning requirements to obtain a digital protection constraint set for historical buildings; performing hierarchical filtering and weighting on a basic corpus based on an original structure retention list and building material reuse indicators to generate a dynamically optimized corpus; using a spatial generation neural network to construct a topological structure diagram using spatial morphological feature vectors, integrating structural conflict loss and material matching loss to achieve solution generation control, and embedding a two-way traceability mapping relationship between corpus and constraints; loading the interior space solution into a historical building BIM model, and achieving protection area highlighting, real-time violation detection, and traceability report output through a VR interactive interface. The present invention realizes the intelligent generation, compliance verification, and interactive correction of interior design solutions for historical buildings, effectively improving design efficiency and protection consistency.

Description

Technical Field

[0001] The present invention relates to the field of architectural design technology, and in particular to an intelligent interior design method and system based on AI corpus optimization. Background Art

[0002] As urban renewal and the reuse of historic buildings become increasingly prominent areas of focus in the architectural field, the renovation of traditional architectural spaces is increasingly facing the challenge of balancing the dual goals of "functional adaptability" and "historical preservation." With the advancement of Building Information Modeling (BIM) and artificial intelligence technologies, academia and industry are experimenting with AI-assisted generative models, architectural semantic data, and virtual reality interaction to enhance the intelligent design of historic interiors. Some existing methods express renovation intent through parametric modeling or optimize design solutions through database recommendation mechanisms, demonstrating a preliminary level of adaptive capabilities. Furthermore, methods such as interactive architectural simulations and design scenario previews, combined with VR technology, are increasingly being used in design review processes.

[0003] Current technologies still have the following significant problems: First, there is a lack of a corpus optimization mechanism for historical building protection constraints. Design schemes are often generated based on general corpus or standard style libraries, which are difficult to adapt to specific building protection items and structural constraint requirements; second, existing generation models cannot achieve component-level design-source-constraint triple traceability control, resulting in low semantic transparency in the design process and difficulty in providing clear explanations to users or regulators; third, although some systems have BIM or VR display capabilities, their functions mostly remain at the visual presentation level, failing to achieve real-time violation detection, constraint association tracing and user behavior intervention, and lack true "intelligent compliance interaction" capabilities. Summary of the Invention

[0004] The present invention provides an intelligent interior design method and system based on AI corpus optimization, and provides a systematic method that integrates AI corpus processing, protection constraint embedding, neural generation control and BIM-VR real-time feedback to realize an intelligent generation mechanism in the interior design of historical buildings with "unchanged structure, unbroken principles and controllable semantics".

[0005] An intelligent interior design method based on AI corpus optimization includes the following steps:

[0006] S1: Receive the user's input of indoor functional zoning requirements and simultaneously obtain the digital protection constraint set of the historical building;

[0007] S2: Perform hierarchical filtering on the basic corpus based on the protection constraint set to generate a dynamically optimized corpus:

[0008] S3: Inputting the dynamic optimization corpus into the spatial generative neural network, driving the spatial generative neural network to use the spatial morphological feature vector as the initial topology, and outputting an indoor space solution that satisfies all protection constraint sets, ensuring that the generation path of each design element can be traced back to the corresponding constraint;

[0009] S4: Load the interior space plan into the BIM model of the historical building, detect the compliance of user operations with the protection constraint set in real time through the VR interactive interface, and output illegal operation warnings and constraint item traceability reports.

[0010] Optionally, the digital protection constraint set includes an original structure retention list, building material reuse indicators, and spatial morphology feature vectors.

[0011] Optionally, the receiving of the indoor functional zoning requirements input by the user and the simultaneous acquisition of the digital protection constraint set of the historical building include:

[0012] S11: receiving indoor functional zoning requirements input by the user, and converting them into structured functional zoning configuration parameters through a zoning requirement parser, wherein the functional zoning configuration parameters include room type identification, area threshold, and adjacent relationship matrix;

[0013] S12: Synchronously extract digital protection constraint sets from the historical building information database, including:

[0014] Obtain a list of load-bearing components that must be retained through the BIM component library and generate a list of original structure retention;

[0015] Count the types and quantities of recyclable building materials on site and calculate the building material reuse index according to environmental protection standards;

[0016] Extract the curvature distribution and axis angle of building point cloud data to construct spatial morphological feature vectors;

[0017] S13: Synchronously bind the functional zoning configuration parameters with the digital protection constraint set to generate a combined input set of indoor functional zoning requirements and protection constraints with constraint identifiers, which serves as an input source for hierarchical filtering in S2.

[0018] Optionally, performing hierarchical filtering on the basic corpus based on the protection constraint set includes:

[0019] First-level filtering: eliminating design cases in the corpus that conflict with the original structure retention list;

[0020] Secondary weighting: Increase the weight of the corpus containing renovation plans containing old materials according to the building materials reuse indicators.

[0021] Optionally, performing hierarchical filtering on the basic corpus based on the protection constraint set to generate a dynamically optimized corpus includes:

[0022] S21: Based on the original structure retention list in the combined input set of indoor functional zoning requirements and protection constraints outputted in S1, perform a first-level filtering on the basic corpus: compare the structural component data of the corpus cases with the original structure retention list one by one, eliminate all corpus cases containing components prohibited from modification in the list, and generate a first-level filtering result;

[0023] S22: Based on the indoor functional zoning requirements and protection constraints output by S1 and the building material reuse index in the input set, perform secondary weighting on the primary filtering results to generate secondary weighted results with weight labels.

[0024] S23: Integrate the first-level filtering results and the second-level weighted results, reconstruct the corpus index in descending order of weight values, generate a dynamically optimized corpus and output it to the spatial generation neural network of S3.

[0025] Optionally, performing secondary weighting on the primary filtering results includes:

[0026] Calculate the matching degree between the old material utilization rate in the corpus case and the building material reuse index;

[0027] The corpus weight value is increased according to the matching ratio to generate a secondary weighted result with a weight label.

[0028] Optionally, the space generation neural network includes a corpus input layer, a topology initialization module, a constraint fusion unit, a traceability identification engine and a solution output interface, wherein;

[0029] The corpus input layer is used to receive and parse the dynamically optimized corpus output by S2, and extract the spatial layout parameters and weight labels in the corpus cases;

[0030] A topology initialization module is connected to the corpus input layer and is used to decode the spatial morphological feature vectors in the digital protection constraint set into a vectorized spatial topology structure graph;

[0031] The constraint fusion unit is connected in parallel with the topology initialization module and is used to load the digital protection constraint set in real time during the iterative process of generating the spatial generative neural network;

[0032] The retroactive identification engine is embedded in the hidden layer of the spatial generative neural network, which is used to generate at each iteration:

[0033] The corpus case ID that triggers the binding of a new wall;

[0034] Bind the corresponding constraint entry ID to the reserved component;

[0035] Generate a bidirectional traceability mapping table to record the relationship between design elements and constraints;

[0036] The plan output interface is used to output indoor space plans.

[0037] Optionally, the driving space generation neural network uses the spatial morphological feature vector as the initial topology and outputs an indoor space solution that satisfies all protection constraint sets, including:

[0038] S31: inputting the dynamically optimized corpus into the corpus input layer of the spatial generative neural network, and simultaneously inputting the spatial morphological feature vector in the digital protection constraint set into the topology initialization module;

[0039] S32: decoding the spatial morphological feature vector into a spatial topological structure graph through the topological initialization module, wherein parameters of wall curvature and column grid angle are completely inherited from geometric properties of the spatial morphological feature vector;

[0040] S33: Using the spatial topology structure graph as the initial topology, driving the constraint fusion unit of the spatial generative neural network to perform iterative generation, each round of iteration includes:

[0041] Real-time calculation of structural conflict loss: detection of the overlap between the spatial coordinates of the load-bearing components in the generated solution and the original structure retention list;

[0042] Dynamically evaluate material matching losses: compare the ratio of new and old building materials with the deviation value of building material reuse index;

[0043] S34: If the structural conflict loss and the material matching loss are both lower than a preset threshold, the traceability identification engine is activated to execute:

[0044] Bind the triggered corpus case ID to each generated wall;

[0045] Bind the corresponding constraint item ID to the retained original component;

[0046] Generate a bidirectional traceability mapping table containing all binding relationships;

[0047] S35: Output the interior space plan through the plan output interface. The plan includes a three-dimensional space layout model based on the final iteration result, a building materials list with the ratio of new and old materials marked, and a two-way traceability mapping table. The complete plan is then transmitted to the BIM model loading interface of S4.

[0048] Optionally, the S4 includes:

[0049] S41: Loading the spatial layout three-dimensional model and the building materials usage list in the interior space plan into the historical building BIM model, and simultaneously embedding the bidirectional traceability mapping table to generate a BIM model with constraint identification;

[0050] S42: Building a VR interactive interface through the Unity engine to convert the BIM model with constraint identifiers into a real-time rendered VR scene, wherein:

[0051] The load-bearing components are marked as red non-editable areas according to the original structure retention list;

[0052] The texture of old materials in the renovated area shall meet the requirements of the building material reuse index;

[0053] S43: Monitor the editing instructions of the VR interactive interface in real time during user operation, and trigger an illegal operation warning when the following behaviors are detected:

[0054] Modify the components in the red non-editable area;

[0055] The proportion of new materials used exceeds the threshold of building materials reuse index;

[0056] Destroy the curvature / angle characteristics defined by the spatial morphological feature vector;

[0057] And output a warning pop-up window containing the violation coordinates and constraint entry ID in real time;

[0058] S44: Generate a constraint item traceability report based on the bidirectional traceability mapping table. The constraint item traceability report includes:

[0059] The original structure retention list items or building material reuse index clauses associated with the illegal operation;

[0060] Design the corpus case ID and its weight value for this area;

[0061] Spatial morphology compliance deviation value calculation data;

[0062] S45: The illegal operation warning and the constraint item tracing report are superimposed and displayed on the VR interactive interface, and executable VR operation correction suggestions are output.

[0063] An intelligent interior design system based on AI corpus optimization is used to implement the above-mentioned intelligent interior design method based on AI corpus optimization, including the following modules:

[0064] Functional zoning parsing module: used to receive the indoor functional zoning requirements input by the user and generate structured functional zoning configuration parameters;

[0065] Digital conservation constraint collection module: used to extract the original structure preservation list, building material reuse index and spatial morphological feature vector from the historical building information database to form a digital conservation constraint set;

[0066] Corpus-level optimization module: used to perform primary filtering and secondary weighting on the basic corpus, and output a dynamically optimized corpus;

[0067] Spatial Generative Neural Network: This includes a corpus input layer, a topology initialization module, a constraint fusion unit, a traceability identification engine, and a solution output interface, and is used to generate indoor space solutions that meet protection constraints.

[0068] BIM loading and VR interaction module: used to load indoor space plans into BIM models with constraint identification, and detect the compliance of user operations in real time through the VR interactive interface, and generate illegal operation warnings and constraint item traceability reports.

[0069] Beneficial effects of the present invention:

[0070] This invention introduces a digital protection constraint set consisting of an original structure preservation list, building material reuse indicators, and spatial morphological feature vectors. By performing hierarchical filtering and weighting on the basic corpus in S2, it effectively eliminates design corpora that conflict with architectural protection principles, thereby improving the green transformation adaptability of the generated corpus. In S3, a spatial generative neural network is constructed with spatial morphological feature vectors as the initial topology, and structural conflict loss and material matching loss are introduced to control the direction of spatial generation. This achieves full-process constraint transmission from data corpora to structural components, effectively avoiding design deviations, misjudgments, and repeated revisions in traditional manual modeling, significantly improving the compliance and speed of generated solutions.

[0071] The present invention embeds a traceability identification engine in the spatial generation neural network. Based on the generation process, it binds the corpus case ID to the newly created wall, binds the protection constraint item ID to the retained component, and generates a bidirectional traceability mapping table. In S4, this mapping table is embedded in the historical building BIM model, allowing users to click on any component through the VR interactive interface to obtain its generation basis and protection reasons in real time. This mechanism enables the system to have a precise "component-constraint-corpus" triple binding capability, greatly improving the traceability and controllability of the interior design process, and effectively supporting application scenarios such as design approval, compliance review, and historical record archiving.

[0072] This invention uses S4 to load interior space plans into the BIM model of a historic building and builds a VR interactive interface using the Unity engine. This significantly enhances the user's immersive design experience while also enabling triple violation detection for structural protection and green indicators (i.e., damage to load-bearing components, deviation from material proportions, and damage to spatial form). The system triggers violation alerts based on real-time editing commands and automatically generates constraint item traceability reports and correction suggestions, effectively preventing users from unknowingly touching protective components or violating green renovation goals. This creates a closed-loop interactive system of "front-end alerts + mid-stage guidance + back-end traceability," driving the transformation of traditional buildings towards compliance, efficiency, and intelligence. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0074] Figure 1 Schematic diagram of a method flow in an embodiment of the present invention;

[0075] Figure 2 Schematic diagram of the system flow of an embodiment of the present invention. DETAILED DESCRIPTION

[0076] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. It is also noted that, to provide a more detailed description, the following embodiments are best and preferred embodiments, and those skilled in the art may employ alternative methods for implementing certain known technologies. Furthermore, the accompanying drawings are intended only to provide a more detailed description of the embodiments and are not intended to limit the present invention.

[0077] like Figure 1 As shown, an intelligent interior design method based on AI corpus optimization includes the following steps:

[0078] S1: Receive the user's input of indoor functional zoning requirements and simultaneously obtain the digital protection constraint set of the historical building. The protection constraint set includes the original structure preservation list, building material reuse index, and spatial morphological feature vector;

[0079] S2: Perform hierarchical filtering on the base corpus based on the protection constraint set:

[0080] First-level filtering: eliminating design cases in the corpus that conflict with the original structure retention list;

[0081] Secondary weighting: Increase the weight of renovation plans containing old materials according to the building materials reuse index;

[0082] Generate dynamically optimized corpus;

[0083] S3: Input the dynamically optimized corpus into the spatial generative neural network, drive the network to use the spatial morphological feature vector as the initial topology, and output an indoor space solution that satisfies all protection constraints, ensuring that the generation path of each design element can be traced back to the corresponding constraint;

[0084] S4: Load the interior space plan into the BIM model of the historical building, and use the VR interactive interface to detect the compliance of user operations with the protection constraint set in real time, and output illegal operation warnings and constraint item traceability reports.

[0085] S1 includes:

[0086] S11: First, the system receives user input of indoor functional zoning requirements through the user interface. Users typically enter their space usage requirements for various functional areas (such as bedroom, kitchen, living room, study, etc.) through graphical layout, text description, or preset templates. The system calls the embedded zoning requirement parser to structure this input information and automatically extract three core parameters:

[0087] Room type identification: Standardize the coding of all functional areas to form semantic labels that can be matched with the corpus;

[0088] Area threshold: Convert the user's expected area value into an area threshold with an upper and lower tolerance range to adapt to dynamic allocation during space generation;

[0089] Adjacency Matrix: Based on the spatial connectivity relationships set by the user, a Boolean relationship matrix is ​​generated to describe the rooms that should maintain direct connections or functional adjacencies.

[0090] The above three types of information are integrated into structured functional partition configuration parameters to provide precise guidance for subsequent corpus screening and neural network input.

[0091] The zoning requirement parser is used to perform standardized parsing of the indoor functional zoning requirements input by the user. The zoning requirement parser includes the following subunits:

[0092] Semantic recognition unit: Based on natural language processing algorithms and a preset spatial function dictionary, this unit identifies text content or graphic symbols in user input regarding the purpose of a room and converts them into standardized room type identifiers, such as "bedroom," "kitchen," and "study."

[0093] Area Intent Extraction Unit: This unit extracts user-specified room area specifications, including numerical ranges, minimum area limits, or relative area ratios. All area expressions are uniformly converted into area thresholds with upper and lower tolerances.

[0094] Spatial connection structured unit: This unit recognizes the user's description of the spatial connection relationship between functional areas, such as "the dining room is connected to the kitchen" and "the bathroom is not adjacent to the bedroom". By constructing a Boolean adjacency relationship matrix, it clarifies the relative layout constraints in the spatial arrangement process.

[0095] The above three sub-units work together to convert user input into structured functional zoning configuration parameters, providing accurate guidance for subsequent design generation.

[0096] S12: After the user's zoning requirements are determined, the system will simultaneously extract the digital conservation constraint set associated with the target building from the connected historical building information database to ensure that the design results meet the user's functional expectations while not destroying the conservation elements of the existing building. This process includes the following three contents:

[0097] Generate a list of original structures to be preserved: The system searches the BIM component library of the historic building to identify all structural components that must be absolutely preserved, especially load-bearing walls, column grid systems, and floor units. This list, which indicates component location, size, and connection logic, serves as a key constraint in the subsequent filtering process.

[0098] Building material reuse index calculation: Through the construction site feedback system, the types and quantities of recyclable building materials (such as old wood, bricks, stone, doors and windows, etc.) in the current building are counted. Based on national and industry green building standards, the system combines material performance and reuse level to calculate a set of material reuse indexes to guide the allocation of data weights and scheme evaluation;

[0099] Constructing spatial morphological feature vectors: The system uses high-precision point cloud data from historical buildings to calculate parameters such as curvature distribution, boundary continuity, and structural axis angles on major surfaces such as walls and ceilings. This extracts a set of feature vectors that reflect the geometric style of the architectural space. These feature vectors serve as the initial spatial topology reference during the neural network generation phase.

[0100] S13: Finally, the system synchronizes the extracted structural functional zoning configuration parameters with the digital protection constraint set. This process uses a label coupling mechanism to map each functional requirement to the protection constraints it may be affected, thereby generating a joint input set of indoor functional zoning requirements and protection constraints with constraint identifiers. This input set not only preserves the integrity of the user's spatial intent but also embeds control information about protection restrictions. It serves as the basic input source for performing hierarchical corpus filtering in step S2, ensuring that the subsequently generated design solutions not only meet user needs but also comply with the principles of historical building protection.

[0101] S2 includes:

[0102] To achieve intelligent generation of interior space solutions based on historical building protection constraints, we first construct a basic corpus for semantic filtering and space generation. Its definition and construction method are as follows:

[0103] The basic corpus refers to a collection of interior design cases with structured characteristics. Each piece of corpus contains multi-dimensional parameters such as spatial configuration, component configuration, material properties and functional adaptation. It is used to achieve responsive filtering of the digital protection constraint set of historical buildings while meeting the user's indoor functional zoning requirements.

[0104] The basic corpus mainly includes the following contents:

[0105] Spatial configuration data: records room types, scale parameters, bay-to-depth ratios, lighting orientations, and zoning combinations, providing a basis for judging the rationality of the space;

[0106] Component configuration data: describes the spatial layout and connection relationships of building components such as doors, windows, walls, beams and columns, and is used for one-to-one comparison with the "original structure retention list";

[0107] Material attribute data: Mark whether the materials used in each component are recycled materials, record material categories, reusability levels and environmental certification parameters, etc., to assess the reuse potential of building materials;

[0108] Functional adaptation labels: including the building function type applicable to the corpus, user satisfaction labels, historical renovation evaluation results, etc., which are used for semantic recommendation and sorting priority judgment.

[0109] The construction of the basic corpus adopts the following method to carry out multi-source fusion:

[0110] Project archiving and collection: Extracting real renovation data from various historical building renovation projects, inferring spatial structure and component parameters through BIM models, and forming standard corpus units;

[0111] Drawing and data translation: Collect classic architectural atlases, design case manuals, renovation guides, and other materials, and combine manual annotation with image recognition technology to extract structured design content;

[0112] Parametric simulation expansion: With the help of the building simulation platform and parametric generation engine, characteristic spatial samples are generated in batches based on controllable structural boundary conditions to fill the gaps in special scenario corpus.

[0113] The resulting basic corpus is stored as a structured database, supporting index retrieval, semantic comparison, and weighted sorting. This corpus serves as the input source for step S2, and will be screened and weighted in the subsequent first-level filtering and second-level weighting processes based on the "original structure retention list" and "building material reuse index," respectively.

[0114] S21: Structural comparison of all design cases in the base corpus is performed based on the original structural retention list in the joint input set. This list typically consists of load-bearing components that must be retained, including beams, columns, load-bearing walls, and specific floor units in the building, whose retention properties are non-violable.

[0115] The system traverses each design case in the corpus, extracting the structural component layout data involved and comparing it item by item with the component position, size, and connection type in the inventory. If a case in the corpus modifies, obscures, or interferes with the structure (such as removal, replacement, or relocation) of any component in the inventory, the system immediately marks the case as prohibited and removes it.

[0116] The processing results form a first-level filtered result set, in which all corpora are guaranteed not to violate the hard constraints of the original structure retention list.

[0117] For example: If the list states "retain the east load-bearing wall", and a design case in the corpus breaks through it to form an open kitchen, the case will be identified and eliminated.

[0118] S22: On the basis of ensuring that the structural retention requirements are not violated, the system further performs weighted processing on the design cases in the first-level filtering results according to the building material reuse index to improve the recommendation priority of the green renovation plan.

[0119] The specific method is as follows:

[0120] (1) The system traverses each corpus case in the first-level filtering result and calculates the utilization rate of old materials used in the case, that is, the proportion of recycled building materials contained in its design;

[0121] (2) Compare with the extracted building material reuse index to evaluate the matching degree between the two and calculate the weighted coefficient using the following formula : ;

[0122] in, For the The weighting coefficient of the corpus case, For the The actual utilization rate of old materials in the corpus case, Recommended values ​​for building material reuse indicators in the joint input set.

[0123] (3) The weighting coefficient is added as a weight label to each corpus case to form a secondary weighted result.

[0124] Example: If the recommended value of the building material reuse index is 60%, and a corpus design scheme uses 65% of old building materials, its weighting coefficient is , the system will appropriately increase the sorting priority of the solution.

[0125] S23: After obtaining the first-level filtering results and the corresponding weighted values, the system sorts all retained corpora in descending order according to the weighted values ​​and reconstructs the corpus index structure. Corpora with higher rankings represent higher compliance and optimization value in both structural preservation and building material reuse.

[0126] The resulting corpus is a dynamically optimized corpus. This corpus retains a subset of data that meets the requirements of the original structural preservation list and further optimizes its ranking priority through building material reuse indicators, demonstrating excellent green adaptability and generative adaptability. This corpus will serve as the input source for the spatial generative neural network in the next step, S3, driving the neural network to generate innovative interior space solutions that meet various constraints.

[0127] S3 includes:

[0128] The constructed spatial generative neural network is invoked, taking the dynamically optimized corpus and the digital conservation constraint set as its input data sources. Through multiple rounds of iterative generation, a spatial layout model that meets both conservation constraints and design requirements is generated. This spatial generative neural network has five core components: a corpus input layer, a topology initialization module, a constraint fusion unit, a traceability identification engine, and a solution output interface. Its overall structure ensures the logical controllability, constraint consistency, and corpus traceability of the spatial generation results.

[0129] The spatial generative neural network architecture includes:

[0130] Corpus input layer: This layer receives and parses the dynamically optimized corpus. It uses semantic parsing to extract spatial layout parameters (such as partitioning, streamline organization, bay-to-depth ratio, etc.) and their associated weight labels from each corpus case, providing semantic reference for subsequent network decisions.

[0131] Topology initialization module: This module is connected to the corpus input layer and is used to receive the spatial morphological feature vectors in the digital protection constraint set and decode them into a standardized vectorized spatial topology structure diagram. This structure diagram fully preserves the original axis curvature distribution and column grid angle structural relationship of the historical building, ensuring that the spatial generation starts from the original appearance of the building.

[0132] Constraint fusion unit: continuously loads the digital protection constraint set in each network generation iteration and calculates the following two types of loss values:

[0133] Structural conflict loss: Detects whether the newly added or adjusted components in the current iterative generation scheme overlap with the load-bearing components specified in the "Original Structure Retention List";

[0134] Material matching loss: Dynamically evaluate the deviation between the ratio of new and old materials used in the generated plan and the "building material reuse index" to ensure consistency with green building goals;

[0135] Traceability Identification Engine: Embedded in the hidden layer of the neural network, it is used to bidirectionally trace and bind design behavior to data sources during each iteration. Its main functions include:

[0136] Bind the triggered corpus case ID to each newly created wall;

[0137] Bind each retained original component with its corresponding protection constraint entry ID;

[0138] Automatically generate a bidirectional traceability mapping table between "design elements-constraints" to ensure subsequent explainability and compliance audit capabilities;

[0139] Solution Output Interface: When all constraint loss values ​​converge to below a preset threshold, the system outputs the final solution through this interface. This solution includes: a 3D interior space layout model; a building materials list with the ratio of new and old materials; and a complete two-way traceability mapping table.

[0140] The S3 execution process includes:

[0141] S31: First, the dynamically optimized corpus generated by S2 is fed into the corpus input layer of the spatial generative neural network. The system automatically extracts the layout logic, key dimensions, and design preference labels for each corpus. Simultaneously, the spatial morphological feature vectors output by S1 are fed into the topology initialization module to provide geometric initial conditions for spatial structure generation.

[0142] S32: The system uses the topology initialization module to analyze the spatial morphological feature vectors, converting the geometric information recorded therein, such as wall curvature and column grid angles, into a spatial topological structure diagram. This diagram serves as the initial network structure foundation for subsequent generation processes. Its configuration is fully inherited from the original form of the historical building itself, ensuring that the generation process does not deviate from the building's geometric logical framework.

[0143] S33: Using this topology as the initial input topology of the neural network, the system starts the constraint fusion unit and performs spatial solution generation guided by the constraint conditions. In each generation iteration, the system completes the following two loss assessments in real time:

[0144] Calculation of Structural Conflict Loss: The system detects whether generated spatial components, such as walls and door openings, intersect or overlap in 3D coordinates with load-bearing components specified in the original structural retention list. If a conflict exists, the system applies a negative feedback weight to the area containing the component, guiding the network to make local structural adjustments.

[0145] Calculation of material matching loss: The system calculates the ratio of new and old materials used in the current solution and compares it with the recommended ratio in the "Building Material Reuse Index." If the deviation is large, the system reduces the probability of generating the solution and guides the material distribution to converge toward the target ratio.

[0146] For example, if the recommended reuse ratio for building materials is 70%, but the actual ratio in the generated plan is 60%, then the loss value is set to the absolute value of the deviation |60%-70%|=10%, and this value will be used as a negative guiding factor in the next round of generation.

[0147] S34: When the system detects that both loss values ​​in the current iteration result are lower than the preset threshold (e.g., structural conflict loss is less than 5%, and material matching deviation is less than 8%), the system activates the traceability identification engine. The engine completes the following three binding operations:

[0148] For each newly created wall, record which corpus case triggered its creation and bind the corresponding corpus case ID;

[0149] For each retained component, bind the protection constraint entry to which it is restricted and record the constraint entry ID;

[0150] Output a bidirectional traceability mapping table between each spatial component and its source corpus or protection restrictions to ensure bidirectional traceability between users and the system.

[0151] S35: Finally, the system outputs a complete interior space plan through the plan output interface, which includes the following three parts:

[0152] A 3D spatial layout model for BIM modeling and subsequent scene loading;

[0153] A list of building materials, with the proportion of new and old materials marked by component, for green building evaluation;

[0154] A bidirectional traceability mapping table is used for constraint consistency verification and user behavior prompts in subsequent operations.

[0155] The spatial plan will be transmitted to the BIM model loading interface in step S4 to achieve visualization, interactivity and compliance monitoring of the design plan with protection constraints.

[0156] Example of a bidirectional traceability mapping table:

[0157] Element Type Component number Generate source type Association ID Related instructions New wall W-101 Corpus Case Case_A013 Generate a solution based on the "separation of living room and study" in corpus case A013 New partition D-304 Corpus Case Case_B227 Derived from the functional partition optimization logic in Case B227 Retain column components C-017 Protection constraint entries Struct_Keep_005 This item belongs to the "Second Floor Load-Bearing Column C-017" item in the original structure preservation list. Retain load-bearing walls W-002 Protection constraint entries Struct_Keep_001 Located on the south wall of the building, cannot be moved New window openings WD-406 Corpus Case Case_C145 From the "Improve Ventilation - Renovate Windows" suggestion in Case C145 Retain beam components B-003 Protection constraint entries Struct_Keep_009 Associated with the floor load system and cannot be removed or modified

[0158] S4 includes:

[0159] This paper adopts a historical building BIM model based on the IFC (Industry Foundation Classes) standard. Its overall structure is divided into the following four layers:

[0160] Building layer: defines the total spatial boundary and positioning coordinates of the entire historical building entity, including global attributes such as building height benchmark, orientation, and terrain contact relationship;

[0161] Floor level: subdivided into structural units, each floor contains its own structural components (such as floor slabs, beams, columns) and use areas (such as rooms, corridors);

[0162] Component layer: includes beams, columns, walls, slabs, doors, windows, stairs and other physical components. All components have the following data fields:

[0163] Geometric properties (length, thickness, height, curvature, slope, etc.);

[0164] Structural role attributes (whether load-bearing, removable);

[0165] Material properties (whether it is old material, recycling level, thermal conductivity, etc.);

[0166] Space association attributes (room ID, connection object ID);

[0167] Logical relationship layer: This layer is responsible for recording the connection logic and semantic constraints between components, including load-bearing and force-transmitting relationships, spatial adjacency relationships, generation source identification, protection item binding, etc.

[0168] To ensure the system can make real-time judgments on conservation constraints and structural compliance, a dual mapping mechanism is used for the connection relationships between all structural components in the BIM model of historical buildings, as follows:

[0169] (1) Structural connection mapping:

[0170] For all load-bearing components (such as columns, load-bearing walls, beams, etc.), the system constructs a "force transmission path diagram" to clarify their upstream and downstream load-bearing relationships;

[0171] Each beam component needs to define the start and end column numbers of the connection, and each floor slab needs to be associated with its supporting beam number and the four boundary wall IDs;

[0172] The components are modeled using a directed graph structure to form a topological chain of load-bearing and loaded relationships.

[0173] (2) Spatial adjacency mapping:

[0174] The locations of boundary walls, bays, and door openings between all rooms are marked using the spatial adjacency matrix;

[0175] The system uses rooms as nodes and connecting components as edges to construct a spatial adjacency graph to support space generation and semantic navigation.

[0176] (3) Protection logo binding:

[0177] The reserved field in each component stores a Boolean label indicating whether it belongs to the “original structure retention list”;

[0178] At the same time, bind the "Protection Constraint Entry ID" of the corresponding entry for quick indexing and tracing in subsequent operations.

[0179] The historical building BIM model in this invention supports the following dynamic embedding functions to ensure its responsiveness and interactivity throughout the entire intelligent interior design process:

[0180] Embed corpus to generate components: Generated 3D components (such as new walls, partitions, and window openings) can be loaded into the model while preserving the historical component structure and automatically integrated with the existing topology;

[0181] Embedded building material usage list: Each new component comes with the ratio of new and old materials, and is mapped to the rendering material through material properties;

[0182] Embed a bidirectional traceability mapping table: Bind the corpus case ID or protected item ID through the component ID to achieve subsequent visual traceability, violation prompts and report generation.

[0183] S41: First, the system loads the 3D spatial layout model and building materials list output from S3 into the BIM model of the historic building, achieving spatial integration through component coordinates, topological connections, and material properties. Simultaneously, the system embeds the binding information from the bidirectional traceability mapping table, assigning constraint tags to each component in the model, including the corresponding corpus case ID or the original structure preservation list item ID, forming a constraint-labeled BIM model with complete data binding relationships.

[0184] This embedding process is automatically completed based on the mapping relationship between component ID and location, ensuring that each design element has a queryable traceability identifier in the VR scene.

[0185] S42: The system calls the Unity engine to build a VR interactive interface and imports the BIM model with constraint identification into a real-time rendered VR scene. To enhance user perception and interactive recognition capabilities, the system visualizes the following protection information in the VR interface:

[0186] Load-bearing components are highlighted in red in the VR scene and set as non-editable areas according to the marking rules in the "Original Structure Preservation List";

[0187] Different material textures are attached to the renovable areas according to the building material usage properties. Rough or mottled material maps are used in the old material areas, and the building material reuse index requirements are loaded in the property panel.

[0188] The corpus trigger component can retrieve its corresponding corpus case ID, weight value and design intent description in the detail view.

[0189] This process enables users to identify component properties and their corresponding protection constraints or design logic in real time while in VR editing mode.

[0190] S43: When the user edits, replaces, deletes, or creates a new model using the VR controller, the system monitors and analyzes all editing commands in real time. Based on the loaded original structure retention list, building material reuse indicators, and spatial morphology feature vectors, the system performs the following three types of violation detection logic:

[0191] Structural conflict detection: If a user attempts to modify, delete, or pass through a load-bearing component (such as a beam, column, or load-bearing wall) in a red, non-editable area, the system will immediately identify it as a violation of the structural retention requirement.

[0192] Material ratio verification: Real-time statistics of the material types of newly added material components are collected. When the cumulative proportion of new materials exceeds the threshold set in the building material reuse index (for example, the proportion of old materials must not be less than 60%), the system automatically identifies it as a green index violation;

[0193] Morphological feature destruction detection: If the user's editing behavior causes a drastic change in the curvature of the spatial surface, or the original column grid angle structure is destroyed, the system determines whether it exceeds the morphological deviation tolerance value based on the deviation calculation method of the spatial morphological feature vector.

[0194] Once any of the above violation conditions is met, the system will immediately pop up a warning window for illegal operations, prompting the user the spatial location of the violation, the violation type, and the associated constraint entry ID.

[0195] Example: When a user attempts to create a window opening on a red load-bearing wall, the system will pop up a prompt in the VR interface: "Modification prohibited: Load-bearing wall W-002 is a component defined by Struct_Keep_001 item and cannot be changed."

[0196] S44: To help users understand the source and impact of violations, the system automatically generates a constraint entry traceability report based on the bidirectional traceability mapping table. The report includes:

[0197] The original structure retention list item ID directly associated with the violation, or the building material reuse index item code;

[0198] The corpus case ID and weight value that triggered the generation of the regional component are used to analyze why the design result occurs;

[0199] The spatial form compliance deviation value caused by the current user behavior, such as the change in wall curvature, axis deflection angle, etc.

[0200] The report is stored as structured data and can be archived for subsequent review or used to generate AI recommendation feedback.

[0201] S45: The system will highlight the violation location with a flashing light effect in the VR view and display a complete violation warning and constraint item traceability report in the sidebar. At the same time, based on the violation type, the system provides optional correction suggestions, such as:

[0202] If the protection of load-bearing components is violated, users are advised to try arranging it adjacent to non-load-bearing walls;

[0203] If the proportion of building materials is not in compliance, it is recommended to replace some new material components with old materials;

[0204] If the spatial curvature deviation is too large, the system will provide a "Restore to initial topology" option.

[0205] Users can directly click on the suggested items to correct the model with one click, or return to the historical version to undo the operation, ensuring that the entire operation is visible, controllable, and traceable within the protection framework.

[0206] like Figure 2 As shown, an intelligent interior design system based on AI corpus optimization is used to implement the above-mentioned intelligent interior design method based on AI corpus optimization, including the following modules:

[0207] Functional zoning parsing module: used to receive the indoor functional zoning requirements input by the user and generate structured functional zoning configuration parameters;

[0208] Digital conservation constraint collection module: used to extract the original structure preservation list, building material reuse index and spatial morphological feature vector from the historical building information database to form a digital conservation constraint set;

[0209] Corpus-level optimization module: used to perform primary filtering and secondary weighting on the basic corpus, and output a dynamically optimized corpus;

[0210] Spatial Generative Neural Network: This includes a corpus input layer, a topology initialization module, a constraint fusion unit, a traceability identification engine, and a solution output interface, and is used to generate indoor space solutions that meet protection constraints.

[0211] BIM loading and VR interaction module: used to load indoor space plans into BIM models with constraint identification, and detect the compliance of user operations in real time through the VR interactive interface, and generate illegal operation warnings and constraint item traceability reports.

[0212] The present invention encompasses any alternatives, modifications, equivalents, and solutions that fall within the spirit and scope of the present invention. To provide a thorough understanding of the present invention, specific details are described in detail below in connection with the preferred embodiments of the present invention, but those skilled in the art will be able to fully understand the present invention without these detailed descriptions. Furthermore, to avoid unnecessary confusion regarding the essence of the present invention, well-known methods, processes, procedures, components, and circuits have not been described in detail.

[0213] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. An intelligent interior design method based on AI corpus optimization, characterized in that: The following steps are involved: S1: Receive the user's input of indoor functional zoning requirements and simultaneously obtain the digital protection constraint set of the historical building; S2: Perform hierarchical filtering on the basic corpus based on the protection constraint set to generate a dynamically optimized corpus: S3: Inputting the dynamic optimization corpus into the spatial generative neural network, driving the spatial generative neural network to use the spatial morphological feature vector as the initial topology, and outputting an indoor space solution that satisfies all protection constraint sets, ensuring that the generation path of each design element can be traced back to the corresponding constraint; S4: Load the interior space plan into the BIM model of the historical building, detect the compliance of user operations with the protection constraint set in real time through the VR interactive interface, and output illegal operation warnings and constraint item traceability reports.

2. The intelligent interior design method based on AI corpus optimization according to claim 1 is characterized in that: The digital protection constraint set includes an original structure retention list, building material reuse indicators, and spatial morphology feature vectors.

3. The intelligent interior design method based on AI corpus optimization according to claim 2 is characterized in that: The receiving of the indoor functional zoning requirements input by the user and the simultaneous acquisition of the digital protection constraint set of the historical building include: S11: receiving indoor functional zoning requirements input by the user, and converting them into structured functional zoning configuration parameters through a zoning requirement parser, wherein the functional zoning configuration parameters include room type identification, area threshold, and adjacent relationship matrix; S12: Synchronously extract digital protection constraint sets from the historical building information database, including: Obtain a list of load-bearing components that must be retained through the BIM component library and generate a list of original structure retention; Count the types and quantities of recyclable building materials on site and calculate the building material reuse index according to environmental protection standards; Extract the curvature distribution and axis angle of building point cloud data to construct spatial morphological feature vectors; S13: Synchronously bind the functional zoning configuration parameters with the digital protection constraint set to generate a combined input set of indoor functional zoning requirements and protection constraints with constraint identifiers, which serves as an input source for hierarchical filtering in S2.

4. The intelligent interior design method based on AI corpus optimization according to claim 3 is characterized in that: The performing hierarchical filtering on the basic corpus based on the protection constraint set includes: First-level filtering: eliminating design cases in the corpus that conflict with the original structure retention list; Secondary weighting: Increase the weight of the corpus containing renovation plans containing old materials according to the building materials reuse indicators.

5. The intelligent interior design method based on AI corpus optimization according to claim 4 is characterized in that: The step of performing hierarchical filtering on the basic corpus based on the protection constraint set to generate a dynamically optimized corpus includes: S21: Based on the original structure retention list in the combined input set of indoor functional zoning requirements and protection constraints outputted in S1, perform a first-level filtering on the basic corpus: compare the structural component data of the corpus cases with the original structure retention list one by one, eliminate all corpus cases containing components prohibited from modification in the list, and generate a first-level filtering result; S22: Based on the indoor functional zoning requirements and protection constraints output from S1 and the building material reuse index in the input set, perform secondary weighting on the primary filtering results to generate secondary weighted results with weight labels; S23: Integrate the first-level filtering results and the second-level weighted results, reconstruct the corpus index in descending order of weight values, generate a dynamically optimized corpus and output it to the spatial generation neural network of S3.

6. The intelligent interior design method based on AI corpus optimization according to claim 5 is characterized in that: The performing secondary weighting on the primary filtering result includes: Calculate the matching degree between the old material utilization rate in the corpus case and the building material reuse index; The corpus weight value is increased according to the matching ratio to generate a secondary weighted result with a weight label.

7. The intelligent interior design method based on AI corpus optimization according to claim 6 is characterized in that: The spatial generation neural network includes a corpus input layer, a topology initialization module, a constraint fusion unit, a traceability identification engine and a solution output interface, wherein; The corpus input layer is used to receive and parse the dynamically optimized corpus output by S2, and extract the spatial layout parameters and weight labels in the corpus cases; A topology initialization module is connected to the corpus input layer and is used to decode the spatial morphological feature vectors in the digital protection constraint set into a vectorized spatial topology structure graph; The constraint fusion unit is connected in parallel with the topology initialization module and is used to load the digital protection constraint set in real time during the iterative process of generating the spatial generative neural network; The retroactive identification engine is embedded in the hidden layer of the spatial generative neural network, which is used to generate at each iteration: The corpus case ID that triggers the binding of a new wall; Bind the corresponding constraint entry ID to the reserved component; Generate a bidirectional traceability mapping table to record the relationship between design elements and constraints; The plan output interface is used to output indoor space plans.

8. The intelligent interior design method based on AI corpus optimization according to claim 7 is characterized in that: The driving space generation neural network uses the spatial morphological feature vector as the initial topology and outputs an indoor space solution that satisfies all protection constraint sets, including: S31: inputting the dynamically optimized corpus into the corpus input layer of the spatial generative neural network, and simultaneously inputting the spatial morphological feature vector in the digital protection constraint set into the topology initialization module; S32: decoding the spatial morphological feature vector into a spatial topological structure graph through the topological initialization module, wherein parameters of wall curvature and column grid angle are completely inherited from geometric properties of the spatial morphological feature vector; S33: Using the spatial topology structure graph as the initial topology, driving the constraint fusion unit of the spatial generative neural network to perform iterative generation, each round of iteration includes: Real-time calculation of structural conflict loss: detection of the overlap between the spatial coordinates of the load-bearing components in the generated solution and the original structure retention list; Dynamically evaluate material matching losses: compare the ratio of new and old building materials with the deviation value of building material reuse index; S34: If the structural conflict loss and the material matching loss are both lower than a preset threshold, the traceability identification engine is activated to execute: Bind the triggered corpus case ID to each generated wall; Bind the corresponding constraint item ID to the retained original component; Generate a bidirectional traceability mapping table containing all binding relationships; S35: Output the interior space plan through the plan output interface. The plan includes a three-dimensional space layout model based on the final iteration result, a building materials list with the ratio of new and old materials marked, and a two-way traceability mapping table. The complete plan is then transmitted to the BIM model loading interface of S4.

9. The intelligent interior design method based on AI corpus optimization according to claim 8, characterized in that: The S4 includes: S41: Loading the spatial layout three-dimensional model and the building materials usage list in the interior space plan into the historical building BIM model, and simultaneously embedding the bidirectional traceability mapping table to generate a BIM model with constraint identification; S42: Building a VR interactive interface through the Unity engine to convert the BIM model with constraint identifiers into a real-time rendered VR scene, wherein: The load-bearing components are marked as red non-editable areas according to the original structure retention list; The texture of old materials in the renovated area shall meet the requirements of the building material reuse index; S43: Monitor the editing instructions of the VR interactive interface in real time during user operation, and trigger an illegal operation warning when the following behaviors are detected: Modify the components in the red non-editable area; The proportion of new materials used exceeds the threshold of building materials reuse index; Destroy the curvature / angle characteristics defined by the spatial morphological feature vector; And output a warning pop-up window containing the violation coordinates and constraint entry ID in real time; S44: Generate a constraint item traceability report based on the bidirectional traceability mapping table. The constraint item traceability report includes: The original structure retention list items or building material reuse index clauses associated with the illegal operation; Design the corpus case ID and its weight value for this area; Spatial morphology compliance deviation value calculation data; S45: The illegal operation warning and the constraint item tracing report are superimposed and displayed on the VR interactive interface, and executable VR operation correction suggestions are output.

10. An intelligent interior design system based on AI corpus optimization, used to implement an intelligent interior design method based on AI corpus optimization according to any one of claims 1 to 9, characterized in that: Includes the following modules: Functional zoning parsing module: used to receive the indoor functional zoning requirements input by the user and generate structured functional zoning configuration parameters; Digital conservation constraint collection module: used to extract the original structure preservation list, building material reuse index and spatial morphological feature vector from the historical building information database to form a digital conservation constraint set; Corpus-level optimization module: used to perform primary filtering and secondary weighting on the basic corpus, and output a dynamically optimized corpus; Spatial Generative Neural Network: This includes a corpus input layer, a topology initialization module, a constraint fusion unit, a traceability identification engine, and a solution output interface, and is used to generate indoor space solutions that meet protection constraints. BIM loading and VR interaction module: used to load indoor space plans into BIM models with constraint identification, and detect the compliance of user operations in real time through the VR interactive interface, and generate illegal operation warnings and constraint item traceability reports.

Images