Traditional building micro-reconstruction performance improving system based on machine learning

Abstract

The invention discloses a traditional building micro-reconstruction performance improving system based on machine learning. According to the method, the inherent contradiction between traditional building protection and performance improvement is effectively broken through an intelligent multi-dimensional data fusion and generative design technology. A holographic information model constructed by the multi-source data acquisition module completely retains historical genes and space passwords of a building, and provides an accurate space-time reference for subsequent transformation; a two-channel analysis mechanism of the performance evaluation module brings physical structure safety and cultural value continuity into a unified evaluation system, and ensures that a transformation scheme not only meets modern use requirements but also inherits and builds wisdom. The pattern generation and space reconstruction technology of the module is creatively designed, the limitation of traditional craftsman experience inheritance is broken through, contemporary expression of traditional building languages is achieved in the digital twinning environment, and historical style protection is changed from passive limitation to active innovation.

Description

Technical Field

[0001] The present invention belongs to the technical field of traditional building renovation, and specifically is a traditional building micro-renovation performance improvement system based on machine learning. Background Art

[0002] Micro-renovation of traditional buildings to improve their performance refers to preserving the original style and structural system of traditional buildings while improving their safety, comfort, and applicability through a series of refined renovation measures. These measures generally include, but are not limited to: repairing or replacing old and damaged components to enhance structural strength; optimizing spatial layout to enhance functional use; improving lighting and ventilation conditions to enhance indoor comfort; using energy-saving and environmentally friendly materials and technologies to reduce building energy consumption; and improving fire protection, electrical and other facilities to ensure building safety. Micro-renovation emphasizes the principle of "repairing the old as it was," using traditional techniques and materials whenever possible to preserve the historical information and cultural connotations of the building, while integrating modern living needs to give old buildings new vitality. Compared with the comprehensive reconstruction or demolition and reconstruction of traditional buildings, micro-renovation has the advantages of low investment, short cycle, and low environmental impact. It is one of the important ways to achieve urban renewal and cultural heritage.

[0003] However, existing technologies usually rely on single-dimensional structural monitoring and manual experience judgment, which makes it difficult to balance cultural heritage and functional improvement in the renovation of historical buildings. At the same time, traditional design methods are limited by the efficiency of craftsmen's experience inheritance, lack the ability to generate systematic innovative solutions, and lack risk prediction and dynamic adjustment mechanisms during the renovation process. Summary of the Invention

[0004] The purpose of the present invention is to provide a traditional building micro-renovation performance improvement system based on machine learning in order to solve the above-mentioned problems.

[0005] The technical solution adopted by the present invention is as follows: a traditional building micro-renovation performance improvement system based on machine learning, the system comprising: a multi-source data acquisition module, a performance evaluation module, a renovation plan generation module, an intelligent renovation module, an effect simulation module, and a dynamic decision optimization module;

[0006] The output of the multi-source data acquisition module is connected to the heterogeneous data input of the performance evaluation module through the spatiotemporal alignment engine, and the fused holographic data packet is transmitted to the dual channels of structural safety and energy efficiency;

[0007] The digital twin output end of the performance evaluation module is directly connected to the constraint condition parser of the transformation plan generation module;

[0008] The multi-objective optimization solution output terminal of the transformation solution generation module is connected to the basic solution input interface of the intelligent transformation module, and its material matching results are synchronized to the cost prediction engine of the dynamic decision optimization module through the data bus;

[0009] The parametric design output terminal of the intelligent transformation module is seamlessly connected with the three-dimensional model import interface of the effect simulation module;

[0010] The fluid mechanics and virtual reality data output of the effect simulation module is embedded in the reinforcement learning training set of the dynamic decision optimization module, and the carbon emission prediction data is fed back to the optimization objective function of the transformation plan generation module;

[0011] The weight score output of the dynamic decision optimization module eventually flows back to the innovation intensity regulator of the intelligent transformation module, forming a closed-loop parameter calibration link across modules.

[0012] In a preferred embodiment, the intelligent transformation module is internally provided with an intelligent pattern generator, a space reconstruction engine and a light and shadow optimization system;

[0013] The vector decoration layer output terminal of the intelligent pattern generator of the intelligent transformation module is directly connected to the facade parameter calibration interface of the space reconstruction engine, and the pattern arrangement density data is injected into the geometric constraint condition library of the space reconstruction;

[0014] The parameterized three-dimensional model output terminal of the space reconstruction engine is connected to the building topology structure import port of the light and shadow optimization system through the BIM data bus, and the door and window position parameters are simultaneously written into the sunshade component calculation unit of the light and shadow system;

[0015] The dynamic lighting strategy output of the light and shadow optimization system is reversely connected to the transmittance feedback interface of the intelligent pattern generator, and its material recommendation result is transmitted to the nonlinear geometry generator of the space reconstruction engine through the feature vector mapping channel;

[0016] The cultural gene encoding library of the intelligent pattern generator and the load-bearing structure database of the spatial reconstruction engine share the same spatiotemporal alignment coordinate system, and synchronize the linkage parameters of pattern density and spatial deformation through parallel data pipelines;

[0017] The output end of the radiation field rendering engine of the light and shadow optimization system is embedded in the virtual-reality fusion display interface of the space reconstruction engine, and its energy consumption balance parameters are dynamically adjusted by the streamline analysis unit of the space reconstruction engine through the cross-module optimization loop.

[0018] In a preferred embodiment, the multi-source data acquisition module consists of a three-dimensional laser scanning unit, an Internet of Things sensor network, and a historical archive digitization system. The three-dimensional laser scanning unit uses a phase-shifted laser radar to achieve millimeter-level point cloud acquisition, optimizes the data density of curved surface areas through an adaptive sampling algorithm, and simultaneously uses SLAM positioning technology to generate a building point cloud skeleton. The Internet of Things sensor network deploys a six-axis vibration sensor, a multi-spectral illumination probe, and a temperature and humidity composite probe, and uses the LoRaWAN protocol to establish an ad hoc network transmission system to achieve real-time monitoring of structural micro-deformations. The historical archive digitization system is equipped with an ancient book scanner and an OCR text recognition engine to construct a timeline database of building repair records. At the same time, it tracks residents' activity trajectories through UWB positioning tags to form a space usage heat map data layer. All heterogeneous data is processed by the spatiotemporal alignment engine to output a building holographic data packet in a unified coordinate system.

[0019] In a preferred embodiment, the performance evaluation module utilizes a dual-channel analysis architecture and knowledge graph fusion technology. The structural safety channel incorporates a modified ResNet-50 network. This input receives crack morphology point cloud data, extracts multi-scale crack features through a dilated convolutional layer, and outputs wall damage grade and settlement trend predictions. The energy efficiency channel, centered on an LSTM time series network, inputs five years of temperature and humidity sensor data and meteorological station data to construct a thermal inertia matrix for the building envelope and calculate the heat gain / loss equilibrium point for different seasons.

[0020] In a preferred embodiment, the renovation solution generation module consists of a code constraint parser, a multi-objective optimization engine, and a material matching system. The code constraint parser accesses a database of cultural relics protection laws and uses the BERT model to analyze spatial modification restrictions within the legal text, automatically generating a three-dimensional geofence for the load-bearing wall protection buffer zone. The multi-objective optimization engine operates based on a modified NSGA-III algorithm. Design variables include window-to-wall ratio and sunshade tilt angle, and the objective function simultaneously optimizes simulated energy consumption, estimated renovation costs, and predicted spatial comfort.

[0021] In a preferred embodiment, the intelligent pattern generator uses an improved residual network architecture to achieve innovative design of traditional patterns. The system establishes a digital pattern library containing brackets, stained glass windows, and brick carving elements by scanning historical building components. After the designer inputs the creative sketch through the hand-drawing tablet, the generator automatically extracts the sketch outline features and performs multi-scale feature fusion with the traditional patterns. A cultural gene matching mechanism is introduced in the generation process, and a pre-trained visual semantic model is used to calculate the degree of fit between the generated pattern and the target building age and regional style. The output result retains the cultural symbols of traditional patterns and incorporates the geometric beauty of modern design language. The generator also has adaptive adjustment capabilities, which can automatically optimize the pattern arrangement density according to the actual size of the building facade, and finally outputs a vector file that can be directly imported into a CNC engraving machine and a pattern cultural value analysis report;

[0022] The dynamic equation of pattern density is:

[0023] D=D_base×log(S_ref / S_actual)^k

[0024] Where:

[0025] D represents the number of decorative elements per unit area of the generated pattern, which is used to quantify the visual density;

[0026] D_base is the traditional pattern base density derived from the ancient textual research of the "Yingzaofashi";

[0027] S_ref refers to the typical reference size of the building facade of this type, which is taken from historical building surveying data;

[0028] S_actual represents the actual measurement size of the current renovated building facade;

[0029] k is the visual correction factor, ranging from 0.6 to 1.2, and automatically adjusted according to the building type;

[0030] The innovation of this formula lies in converting ancient construction rules into mathematical constraints, balancing historical norms and modern spatial scale differences through nonlinear logarithmic functions, and introducing dynamic correction factors to achieve regional adaptive adjustment of pattern density.

[0031] In a preferred embodiment, the space reconstruction engine is based on the topological relationship of the building space and the self-attention mechanism, and realizes the intelligent reorganization of the functional space by analyzing the column grid distribution, load-bearing structure and historical protection restrictions. The system converts the plane layout of the traditional building into a three-dimensional spatial relationship map, uses the improved Transformer model to capture the mechanical transmission path between the beams and walls, and automatically generates a variety of space division schemes while ensuring structural safety. The engine has a built-in pedestrian thermal simulation module, which optimizes the traffic flow planning by combining the historical data of door and window positions, and introduces virtual-reality fusion technology to compare and display the spatial effects before and after the transformation through an augmented reality interface. During the reconstruction process, the ventilation efficiency and lighting uniformity performance indicators are dynamically calculated, and a nonlinear geometric scheme including curved walls and special-shaped ceilings is generated, and finally a three-dimensional model compatible with BIM software and a space efficiency improvement report are output.

[0032] The calculation formula of spatial variability index is:

[0033] V = S × α + H × β + F × γ;

[0034] Where V represents the quantitative value of the degree of spatial variability, ranging from 0 to 1 corresponding to complete protection to complete reconstruction;

[0035] S is the structural safety factor, which is calculated based on the modification ratio of the load-bearing wall and the material strength;

[0036] H represents the weight of historical value, which is digitally converted from the protection level designated by the cultural heritage protection department;

[0037] F is the degree of functional improvement, which is quantified by comparing the efficiency of space use before and after the renovation;

[0038] αβγ are dynamic adjustment factors, which automatically assign weights according to building types (γ<α for ancestral hall buildings, γ>α for commercial buildings);

[0039] The innovation of this formula lies in establishing a multi-dimensional quantitative evaluation system, converting highly subjective spatial transformation decisions into calculable engineering parameters, and achieving a precise balance between protection and innovation through a dynamic weight allocation mechanism.

[0040] In a preferred embodiment, the light and shadow optimization system combines neural radiation field rendering and physical optics simulation technology, and dynamically generates a lighting plan that takes into account both artistic effects and energy-saving needs by analyzing the ten-year sunshine trajectory data of the building's location. The system imports the three-dimensional model of the building into the radiation field rendering engine, simulates the natural light projection angle in different seasons and time periods, and automatically calculates the optimal size and position of the shading components. Under the premise of ensuring the uniformity of indoor illumination, the system recommends new materials such as translucent concrete and intelligent dimming glass, and generates a power generation-shading efficiency balance plan for photovoltaic glass and sunshades. Through real-time calculation of 24-hour light and shadow changes, the visual interface displays the dynamic projection effect of light on traditional carved window lattices, and finally outputs an optimization report containing the control logic of the intelligent shading system and the material procurement list, realizing the organic integration of the digital inheritance of the light and shadow art of historical buildings and modern energy-saving technologies.

[0041] The light-shadow material linkage equation is:

[0042] E=aL+bM+cC;

[0043] in:

[0044] E represents the comprehensive score of the lighting scheme, with a higher value indicating a more balanced artistic and functional quality.

[0045] L represents the light and shadow artistic value, which is obtained through the light and dark contrast of radiation field rendering and the projection aesthetic algorithm;

[0046] M is the material compatibility, which is calculated by the matching degree between the traditional building materials database and the new material parameters;

[0047] C is energy efficiency, which is quantified based on the net value of photovoltaic power generation minus air conditioning and cooling energy consumption;

[0048] a, b, and c are dynamic weight coefficients. The weight of a for temple-type buildings is 70%, and the weight of c for exhibition hall-type buildings is 60%.

[0049] The innovation of this formula lies in the establishment of a three-dimensional collaborative model of light and shadow aesthetics, material properties and energy consumption. For the first time, the light and shadow conception of traditional architecture is converted into quantifiable engineering parameters, and targeted optimization of different building functions is achieved through dynamic adjustment of coefficients.

[0050] In a preferred embodiment, the effect simulation module is internally equipped with a computational fluid dynamics engine, a virtual reality environment, and a full lifecycle analysis framework. The fluid dynamics engine uses the finite volume method to discretize the building microclimate model, solving the energy equation involving radiative heat transfer and natural convection, and implementing an adaptive mesh refinement strategy to capture sudden changes in airflow at door and window openings. The virtual reality environment is built on Unreal Engine, using the MotionCapture system to collect human motion data to drive virtual visitor movement routes, and using ray tracing technology to render the projection changes of carved window lattices. The full lifecycle analysis framework uses the Ecoinvent database to construct a building materials list and runs a Monte Carlo simulation to calculate the embodied carbon emissions distribution spectrum of the renovation plan over a 50-year period.

[0051] In a preferred embodiment, the dynamic decision optimization module consists of a reinforcement learning decision tree, a market forecasting model, and a risk sensitivity map. The reinforcement learning decision tree utilizes the DDPG algorithm architecture, defining solution modification actions as parameter fine-tuning in a continuous action space. The reward function integrates physical performance improvement and cultural coordination to generate immediate feedback. The market forecasting model accesses a time-series database of building material prices and uses the Prophet algorithm to decompose the seasonal and trend components of price fluctuations to predict the cost range of main material procurement during the renovation implementation phase. The risk sensitivity map utilizes graph theory algorithms to analyze the network of building component associations, calculate the strength of the structural chain reactions triggered by each renovation measure, and output a color-coded intervention risk grading heat map to support the generation of a multi-dimensional weighted scoring system for the final solution.

[0052] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0053] 1. This invention effectively resolves the inherent contradiction between traditional architectural preservation and performance improvement through intelligent multi-dimensional data fusion and generative design technology. The holographic information model constructed by the multi-source data acquisition module fully preserves the historical genes and spatial codes of the building, providing a precise spatiotemporal benchmark for subsequent renovations. The dual-channel analysis mechanism of the performance evaluation module incorporates physical structural safety and cultural value continuity into a unified evaluation system, ensuring that the renovation plan not only meets modern usage needs but also inherits construction wisdom. The pattern generation and spatial reconstruction technology of the creative design module breaks through the limitations of traditional craftsman experience inheritance, realizes the contemporary expression of traditional architectural language in the digital twin environment, and transforms the protection of historical features from passive restriction to active innovation.

[0054] 2. In the present invention, the controllability and adaptability of the renovation project are significantly improved through a dynamic closed-loop optimization system. The multi-physics field coupling deduction of the effect simulation module can predict in advance the long-term impact of renovation measures on the building microclimate and structural life, effectively avoiding hidden risks in actual construction; the intelligent weight distribution mechanism of the decision optimization module transforms complex humanistic values, economic costs and engineering feasibility into quantifiable decision parameters, realizing a fundamental shift from experience-driven to data-driven. The synergistic effect of the material matching system and the risk sensitivity map ensures that each micro-renovation can find the optimal balance between traditional craftsmanship and modern technology, ultimately forming a replicable, traceable and iterative intelligent renovation paradigm, providing a systematic solution for the sustainable use of historical buildings. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 is a block diagram of the overall system of the present invention;

[0056] Figure 2 This is a system block diagram of the intelligent transformation module in the present invention. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0058] Example:

[0059] Reference Figure 1-2 ,

[0060] A machine learning-based system for improving the performance of traditional buildings through micro-renovation. The system includes: a multi-source data acquisition module, a performance evaluation module, a renovation plan generation module, an intelligent renovation module, an effect simulation module, and a dynamic decision optimization module.

[0061] The output of the multi-source data acquisition module is connected to the heterogeneous data input of the performance evaluation module through the spatiotemporal alignment engine, and the fused holographic data packet is transmitted to the dual channels of structural safety and energy efficiency;

[0062] The constraint condition parser of the digital twin output direct connection transformation plan generation module of the performance evaluation module;

[0063] The multi-objective optimization solution output of the transformation solution generation module is connected to the basic solution input interface of the intelligent transformation module. At the same time, its material matching results are synchronized to the cost prediction engine of the dynamic decision optimization module through the data bus.

[0064] The parametric design output of the intelligent transformation module is seamlessly connected with the 3D model import interface of the effect simulation module;

[0065] The fluid mechanics and virtual reality data outputs of the effect simulation module are embedded in the reinforcement learning training set of the dynamic decision optimization module, while the carbon emission prediction data is fed back into the optimization objective function of the transformation plan generation module.

[0066] The weight score output of the dynamic decision optimization module eventually flows back to the innovation intensity regulator of the intelligent transformation module, forming a closed-loop parameter calibration link across modules.

[0067] The intelligent transformation module is equipped with an intelligent pattern generator, a space reconstruction engine and a light and shadow optimization system;

[0068] The vector decoration layer output of the intelligent pattern generator in the intelligent transformation module is directly connected to the facade parameter calibration interface of the spatial reconstruction engine, injecting pattern arrangement density data into the geometric constraint library of spatial reconstruction;

[0069] The parameterized 3D model output of the spatial reconstruction engine is connected to the building topology structure import port of the light and shadow optimization system through the BIM data bus, and the door and window position parameters are simultaneously written into the sunshade component calculation unit of the light and shadow system;

[0070] The dynamic lighting strategy output of the light and shadow optimization system is reversely connected to the transmittance feedback interface of the intelligent pattern generator, and its material recommendation results are transmitted to the nonlinear geometry generator of the space reconstruction engine through the feature vector mapping channel;

[0071] The cultural gene coding library of the intelligent pattern generator and the load-bearing structure database of the spatial reconstruction engine share the same space-time alignment coordinate system, and synchronize the linkage parameters of pattern density and spatial deformation through parallel data pipelines;

[0072] The output end of the radiation field rendering engine of the light and shadow optimization system is embedded in the virtual-reality fusion display interface of the space reconstruction engine. At the same time, its energy balance parameters are dynamically adjusted by the streamline analysis unit of the space reconstruction engine through the cross-module optimization loop.

[0073] The multi-source data acquisition module consists of a 3D laser scanning unit, an IoT sensor network, and a historical archive digitization system. The 3D laser scanning unit uses a phase-shifted laser radar to capture millimeter-level point clouds, optimizing data density in curved areas through an adaptive sampling algorithm. Simultaneously, it incorporates SLAM positioning technology to generate a building point cloud skeleton. The IoT sensor network deploys a six-axis vibration sensor, a multispectral illumination probe, and a temperature and humidity composite probe. Using the LoRaWAN protocol, it establishes an ad hoc network transmission system for real-time monitoring of structural micro-deformations. The historical archive digitization system, equipped with an ancient book scanner and an OCR text recognition engine, constructs a timeline database of building renovation records. It also uses UWB positioning tags to track resident activity, creating a spatial usage heat map data layer. All heterogeneous data is processed by a spatiotemporal alignment engine, outputting a unified coordinate system for building holographic data.

[0074] The performance evaluation module utilizes a dual-channel analysis architecture and knowledge graph fusion technology. The structural safety channel incorporates an improved ResNet-50 network. Its input receives crack morphology point cloud data, extracts multi-scale crack features through a dilated convolutional layer, and outputs wall damage grade and settlement trend predictions. The energy efficiency channel, centered around an LSTM time series network, inputs five years of temperature and humidity sensor data and meteorological station data to construct a thermal inertia matrix for the building envelope and calculate the heat gain / loss balance point for different seasons. The cultural value assessment channel utilizes an architectural style knowledge graph based on the Neo4j graph database. Using a graph attention network, it extracts 36 eigenvalues, such as eaves rise ratio and bracket combination patterns. Ultimately, it fuses physical performance with cultural indicators to generate a digital twin data model.

[0075] The renovation plan generation module consists of a code constraint parser, a multi-objective optimization engine, and a material matching system. The code constraint parser accesses a database of cultural relics protection law provisions and uses the BERT model to parse the spatial modification restrictions contained in the legal text, automatically generating a three-dimensional geofence for the load-bearing wall protection buffer zone. The multi-objective optimization engine operates based on an improved NSGA-III algorithm. Design variables include 18 parameters, such as window-to-wall ratio and sunshade tilt angle. The objective function simultaneously optimizes simulated energy consumption, estimated renovation costs, and predicted spatial comfort. The material matching system constructs a feature vector space for traditional building materials, maps modern material parameters to this space using transfer learning techniques, and recommends alternative energy-saving material combinations through cosine similarity calculations.

[0076] The intelligent pattern generator uses an improved residual network architecture to achieve innovative design of traditional patterns. The system establishes a digital pattern library containing elements such as brackets, stained glass windows, and brick carvings by scanning historical building components. After the designer inputs the creative sketch through the hand-drawing tablet, the generator automatically extracts the sketch outline features and integrates the multi-scale features with the traditional patterns. A cultural gene matching mechanism is introduced in the generation process, and a pre-trained visual semantic model is used to calculate the degree of fit between the generated pattern and the target building age and regional style. The output result retains the cultural symbols of traditional patterns and incorporates the geometric beauty of modern design language. The generator also has adaptive adjustment capabilities, which can automatically optimize the pattern arrangement density according to the actual size of the building facade. The final output is a vector file that can be directly imported into a CNC engraving machine and a pattern cultural value analysis report;

[0077] The dynamic equation of pattern density is:

[0078] D=D_base×log(S_ref / S_actual)^k

[0079] Where:

[0080] D represents the number of decorative elements per unit area of the generated pattern, which is used to quantify the visual density;

[0081] D_base is the traditional pattern base density derived from the ancient textual research of the "Yingzaofashi";

[0082] S_ref refers to the typical reference size of the building facade of this type, which is taken from historical building surveying data;

[0083] S_actual represents the actual measurement size of the current renovated building facade;

[0084] k is the visual correction factor, ranging from 0.6 to 1.2, and automatically adjusted according to the building type;

[0085] The innovation of this formula lies in converting ancient construction rules into mathematical constraints, balancing historical norms and modern spatial scale differences through nonlinear logarithmic functions, and introducing dynamic correction factors to achieve regional adaptive adjustment of pattern density.

[0086] The spatial reconstruction engine, based on the topological relationships of architectural spaces and a self-attention mechanism, intelligently reorganizes functional spaces by analyzing column grid distribution, load-bearing structures, and historical protection restrictions. The system transforms the plan layout of traditional buildings into a three-dimensional spatial relationship map, using an improved Transformer model to capture the mechanical transmission paths between beams and walls, automatically generating multiple spatial partitioning schemes while ensuring structural safety. The engine has a built-in pedestrian thermal simulation module that optimizes traffic flow planning based on historical data on door and window locations. It also introduces virtual-reality fusion technology to compare and display the spatial effects before and after the renovation through an augmented reality interface. During the reconstruction process, ventilation efficiency and lighting uniformity performance indicators are dynamically calculated, and nonlinear geometric schemes are generated that include curved walls and special-shaped ceilings. The final output is a three-dimensional model and spatial efficiency improvement report compatible with BIM software.

[0087] The calculation formula of spatial variability index is:

[0088] V = S × α + H × β + F × γ;

[0089] Where V represents the quantitative value of the degree of spatial variability, ranging from 0 to 1, corresponding to complete protection to complete reconstruction;

[0090] S is the structural safety factor, which is calculated based on the modification ratio of the load-bearing wall and the material strength;

[0091] H represents the weight of historical value, which is digitally converted from the protection level designated by the cultural heritage protection department;

[0092] F is the degree of functional improvement, which is quantified by comparing the efficiency of space use before and after the renovation;

[0093] αβγ are dynamic adjustment factors, which automatically assign weights according to building types (γ<α for ancestral hall buildings, γ>α for commercial buildings);

[0094] The innovation of this formula lies in establishing a multi-dimensional quantitative evaluation system, converting highly subjective spatial transformation decisions into calculable engineering parameters, and achieving a precise balance between protection and innovation through a dynamic weight allocation mechanism.

[0095] The light and shadow optimization system combines neural radiation field rendering and physical optics simulation technology. By analyzing ten years of sunlight trajectory data for the building's location, it dynamically generates a lighting plan that balances artistic effects and energy conservation needs. The system imports the building's three-dimensional model into the radiation field rendering engine, simulates the natural light projection angles in different seasons and time periods, and automatically calculates the optimal size and position of the shading components. While ensuring uniform indoor illumination, the system recommends new materials such as translucent concrete and intelligent dimming glass, and generates a power generation-shading efficiency balance plan for photovoltaic glass and sunshades. By calculating the changes in light and shadow in real time over 24 hours, the visual interface displays the dynamic projection effect of light on traditional carved window lattices. The final output includes an optimization report containing the control logic of the intelligent shading system and a material procurement list, realizing the organic integration of the digital inheritance of the light and shadow art of historical buildings and modern energy-saving technology.

[0096] The light-shadow material linkage equation is:

[0097] E=aL+bM+cC;

[0098] in:

[0099] E represents the comprehensive score of the lighting scheme, with a higher value indicating a more balanced artistic and functional quality.

[0100] L represents the light and shadow artistic value, which is obtained through the light and dark contrast of radiation field rendering and the projection aesthetic algorithm;

[0101] M is the material compatibility, which is calculated by the matching degree between the traditional building materials database and the new material parameters;

[0102] C is energy efficiency, which is quantified based on the net value of photovoltaic power generation minus air conditioning and cooling energy consumption;

[0103] a, b, and c are dynamic weight coefficients. The weight of a for temple-type buildings is 70%, and the weight of c for exhibition hall-type buildings is 60%.

[0104] The innovation of this formula lies in the establishment of a three-dimensional collaborative model of light and shadow aesthetics, material properties and energy consumption. For the first time, the light and shadow conception of traditional architecture is converted into quantifiable engineering parameters, and targeted optimization of different building functions is achieved through dynamic adjustment of coefficients.

[0105] The simulation module incorporates a computational fluid dynamics engine, a virtual reality environment, and a full lifecycle analysis framework. The fluid dynamics engine discretizes the building's microclimate model using the finite volume method, solving energy equations that incorporate both radiative heat transfer and natural convection. An adaptive mesh refinement strategy is employed to capture sudden airflow changes at door and window openings. The virtual reality environment, built on Unreal Engine, uses MotionCapture to capture human motion data to drive virtual visitor movement, while ray tracing technology is used to render the projections of carved window lattices. The full lifecycle analysis framework utilizes the Ecoinvent database to construct a building materials inventory and runs Monte Carlo simulations to calculate the embodied carbon emissions profile of the renovation plan over a 50-year period.

[0106] The dynamic decision optimization module consists of a reinforcement learning decision tree, a market forecasting model, and a risk sensitivity map. The reinforcement learning decision tree utilizes the DDPG algorithm architecture, defining solution modification actions as parameter fine-tuning in a continuous action space. The reward function integrates physical performance improvement and cultural coordination to generate immediate feedback. The market forecasting model integrates a time-series database of building material prices and uses the Prophet algorithm to decompose the seasonal and trend components of price fluctuations to predict the cost range of main material procurement during the renovation implementation phase. The risk sensitivity map utilizes graph theory algorithms to analyze the network of building component connections, calculate the strength of the structural chain reactions triggered by each renovation measure, and output a color-coded intervention risk grading heat map, supporting the generation of a multi-dimensional weighted scoring system for the final solution.

[0107] From the above we can know:

[0108] This invention effectively resolves the inherent contradiction between traditional architectural preservation and performance improvement through intelligent multi-dimensional data fusion and generative design technology. The holographic information model constructed by the multi-source data acquisition module fully preserves the historical genes and spatial codes of the building, providing an accurate spatiotemporal benchmark for subsequent renovations. The dual-channel analysis mechanism of the performance evaluation module incorporates the safety of the physical structure and the continuity of cultural values into a unified evaluation system, ensuring that the renovation plan not only meets modern usage needs but also inherits construction wisdom. The pattern generation and spatial reconstruction technology of the creative design module breaks through the limitations of traditional craftsman experience inheritance, realizes the contemporary expression of traditional architectural language in the digital twin environment, and transforms the protection of historical features from passive restrictions to active innovation.

[0109] In the present invention, the controllability and adaptability of the renovation project are significantly improved through a dynamic closed-loop optimization system. The multi-physics field coupling deduction of the effect simulation module can predict in advance the long-term impact of renovation measures on the building microclimate and structural life, effectively avoiding hidden risks in actual construction; the intelligent weight distribution mechanism of the decision optimization module transforms complex humanistic values, economic costs and engineering feasibility into quantifiable decision parameters, realizing a fundamental shift from experience-driven to data-driven. The synergistic effect of the material matching system and the risk sensitivity map ensures that each micro-renovation can find the optimal balance between traditional craftsmanship and modern technology, ultimately forming a replicable, traceable and iterative intelligent renovation paradigm, providing a systematic solution for the sustainable use of historical buildings.

[0110] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprises" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements that are not explicitly listed, or also includes elements that are inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device that includes the element.

[0111] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A machine learning-based system for improving the performance of traditional buildings through micro-renovation, characterized by: The system includes: a multi-source data acquisition module, a performance evaluation module, a transformation plan generation module, an intelligent transformation module, an effect simulation module and a dynamic decision optimization module; The output of the multi-source data acquisition module is connected to the heterogeneous data input of the performance evaluation module through the spatiotemporal alignment engine, and the fused holographic data packet is transmitted to the dual channels of structural safety and energy efficiency; The digital twin output terminal of the performance evaluation module is directly connected to the constraint condition parser of the transformation plan generation module; The multi-objective optimization solution output terminal of the transformation solution generation module is connected to the basic solution input interface of the intelligent transformation module, and its material matching results are synchronized to the cost prediction engine of the dynamic decision optimization module through the data bus; The parametric design output terminal of the intelligent transformation module is seamlessly connected with the three-dimensional model import interface of the effect simulation module; The fluid mechanics and virtual reality data output of the effect simulation module is embedded in the reinforcement learning training set of the dynamic decision optimization module, and the carbon emission prediction data is fed back to the optimization objective function of the transformation plan generation module; The weight score output of the dynamic decision optimization module eventually flows back to the innovation intensity regulator of the intelligent transformation module, forming a closed-loop parameter calibration link across modules.

2. The system for improving performance of traditional buildings through micro-renovation based on machine learning as claimed in claim 1, characterized in that: The intelligent transformation module is internally provided with an intelligent pattern generator, a space reconstruction engine and a light and shadow optimization system; The vector decoration layer output terminal of the intelligent pattern generator of the intelligent transformation module is directly connected to the facade parameter calibration interface of the space reconstruction engine, and the pattern arrangement density data is injected into the geometric constraint condition library of the space reconstruction; The parameterized three-dimensional model output terminal of the space reconstruction engine is connected to the building topology structure import port of the light and shadow optimization system through the BIM data bus, and the door and window position parameters are simultaneously written into the sunshade component calculation unit of the light and shadow system; The dynamic lighting strategy output of the light and shadow optimization system is reversely connected to the transmittance feedback interface of the intelligent pattern generator, and its material recommendation result is transmitted to the nonlinear geometry generator of the space reconstruction engine through the feature vector mapping channel; The cultural gene encoding library of the intelligent pattern generator and the load-bearing structure database of the spatial reconstruction engine share the same spatiotemporal alignment coordinate system, and synchronize the linkage parameters of pattern density and spatial deformation through parallel data pipelines; The output end of the radiation field rendering engine of the light and shadow optimization system is embedded in the virtual-reality fusion display interface of the space reconstruction engine, and its energy consumption balance parameters are dynamically adjusted by the streamline analysis unit of the space reconstruction engine through the cross-module optimization loop.

3. The system for improving performance of traditional buildings through micro-renovation based on machine learning as claimed in claim 1, characterized in that: The multi-source data acquisition module consists of a three-dimensional laser scanning unit, an Internet of Things sensor network and a historical archive digitization system; the three-dimensional laser scanning unit uses a phase-type laser radar to achieve millimeter-level point cloud acquisition, optimizes the data density of the curved area through an adaptive sampling algorithm, and simultaneously uses SLAM positioning technology to generate a building point cloud skeleton; the Internet of Things sensor network deploys a six-axis vibration sensor, a multi-spectral light probe and a temperature and humidity composite probe, and uses the LoRaWAN protocol to form an ad hoc network transmission system to achieve real-time monitoring of structural micro-deformations; the historical archive digitization system is equipped with an ancient book scanner and an OCR text recognition engine to build a timeline database of building repair records, and at the same time track residents' activity trajectories through UWB positioning tags to form a space usage heat map data layer; all heterogeneous data are processed by the spatiotemporal alignment engine to output a building holographic data packet in a unified coordinate system.

4. The system for improving performance of traditional buildings through micro-renovation based on machine learning as claimed in claim 1, characterized in that: The performance evaluation module adopts a dual-channel analysis architecture and knowledge graph fusion technology; the structural safety channel has a built-in improved ResNet-50 network, which receives crack morphology point cloud data at the input, extracts multi-scale crack features through the void convolution layer, and outputs wall damage level and settlement trend predictions; the energy efficiency channel uses the LSTM timing network as the core, inputs temperature and humidity sensor data and meteorological station information from the past five years, constructs the thermal inertia matrix of the building envelope structure, and calculates the heat gain / loss balance point in different seasons.

5. The system for improving performance of traditional buildings through micro-renovation based on machine learning as claimed in claim 1, characterized in that: The renovation plan generation module consists of a regulatory constraint parser, a multi-objective optimization engine, and a material matching system. The regulatory constraint parser accesses a database of cultural relics protection law clauses and uses a BERT model to parse the spatial modification restrictions in the legal text, automatically generating a three-dimensional geo-fence for the load-bearing wall protection buffer zone. The multi-objective optimization engine is based on an improved NSGA-III algorithm. Design variables include window-to-wall ratio and sunshade tilt angle. The objective function simultaneously optimizes the simulated energy consumption value, the estimated renovation cost value, and the predicted spatial comfort value.

6. The system for improving performance of traditional buildings through micro-renovation based on machine learning as claimed in claim 1, characterized in that: The intelligent pattern generator uses an improved residual network architecture to achieve innovative design of traditional patterns. The system scans historical architectural components to create a digital pattern library containing brackets, stained glass windows, and brick carvings. Designers can input creative sketches using a hand-drawn tablet, and the generator automatically extracts the sketch contour features and integrates them with the traditional pattern through multi-scale feature fusion. A cultural gene matching mechanism is introduced into the generation process, using a pre-trained visual semantic model to calculate the compatibility of the generated pattern with the target building's age and regional style. The output retains the cultural symbolism of traditional patterns while incorporating the geometric beauty of modern design language. The generator also has adaptive adjustment capabilities, automatically optimizing the pattern arrangement density based on the actual size of the building facade. The final output is a vector file that can be directly imported into a CNC engraving machine, as well as a pattern cultural value analysis report. The dynamic equation of pattern density is: D=D_base×log(S_ref / S_actual)^k Where: D represents the number of decorative elements per unit area of the generated pattern, which is used to quantify the visual density; D_base is the traditional pattern base density derived from the ancient textual research of the "Yingzaofashi"; S_ref refers to the typical reference size of the building facade of this type, which is taken from historical building surveying data; S_actual represents the actual measurement size of the current renovated building facade; k is the visual correction factor, ranging from 0.6 to 1.2, and automatically adjusted according to the building type; The innovation of this formula lies in converting ancient construction rules into mathematical constraints, balancing historical norms and modern spatial scale differences through nonlinear logarithmic functions, and introducing dynamic correction factors to achieve regional adaptive adjustment of pattern density.

7. The system for improving performance of traditional buildings through micro-renovation based on machine learning as claimed in claim 1, characterized in that: The spatial reconstruction engine, based on the topological relationship of architectural space and the self-attention mechanism, achieves intelligent reorganization of functional space by analyzing column grid distribution, load-bearing structure, and historical protection restrictions. The system converts the plane layout of traditional buildings into a three-dimensional spatial relationship map, using an improved Transformer model to capture the mechanical transmission path between beams and walls, and automatically generates multiple space division schemes while ensuring structural safety. The engine has a built-in pedestrian thermal simulation module, which optimizes traffic flow planning by combining historical data on door and window positions. At the same time, it introduces virtual-reality fusion technology to compare and display the spatial effects before and after the renovation through an augmented reality interface. During the reconstruction process, ventilation efficiency and lighting uniformity performance indicators are dynamically calculated, and nonlinear geometric schemes including curved walls and special-shaped ceilings are generated, ultimately outputting a three-dimensional model compatible with BIM software and a spatial efficiency improvement report. The calculation formula of spatial variability index is: V = S × α + H × β + F × γ; Where V represents the quantitative value of the degree of spatial variability, ranging from 0 to 1, corresponding to complete protection to complete reconstruction; S is the structural safety factor, which is calculated based on the modification ratio of the load-bearing wall and the material strength; H represents the weight of historical value, which is digitally converted from the protection level designated by the cultural heritage protection department; F is the degree of functional improvement, which is quantified by comparing the efficiency of space use before and after the renovation; αβγ are dynamic adjustment factors, which automatically assign weights according to building types (γ<α for ancestral hall buildings, γ>α for commercial buildings); The innovation of this formula lies in establishing a multi-dimensional quantitative evaluation system, converting highly subjective spatial transformation decisions into calculable engineering parameters, and achieving a precise balance between protection and innovation through a dynamic weight allocation mechanism.

8. The system for improving performance of traditional buildings through micro-renovation based on machine learning as claimed in claim 1, characterized in that: The light and shadow optimization system combines neural radiation field rendering and physical optics simulation technology. By analyzing ten years of sunlight trajectory data for the building's location, it dynamically generates a lighting plan that balances artistic effects and energy conservation needs. The system imports the building's three-dimensional model into the radiation field rendering engine, simulates the natural light projection angles in different seasons and time periods, and automatically calculates the optimal size and position of shading components. While ensuring uniform indoor illumination, the system recommends new materials such as translucent concrete and intelligent dimming glass, and generates a power generation-shading efficiency balance plan for photovoltaic glass and sunshades. By calculating the changes in light and shadow in real time over 24 hours, a visual interface displays the dynamic projection effect of light on traditional carved window lattices, and ultimately outputs an optimization report containing the control logic of the intelligent shading system and a material procurement list, achieving an organic integration of the digital inheritance of the light and shadow art of historical buildings and modern energy-saving technologies. The light-shadow material linkage equation is: E=aL+bM+cC; in: E represents the comprehensive score of the lighting scheme, with a higher value indicating a more balanced artistic and functional quality. L represents the light and shadow artistic value, which is obtained through the light and dark contrast of radiation field rendering and the projection aesthetic algorithm; M is the material compatibility, which is calculated by the matching degree between the traditional building materials database and the new material parameters; C is energy efficiency, which is quantified based on the net value of photovoltaic power generation minus air conditioning and cooling energy consumption; a, b, and c are dynamic weight coefficients. The weight of a for temple-type buildings is 70%, and the weight of c for exhibition hall-type buildings is 60%. The innovation of this formula lies in the establishment of a three-dimensional collaborative model of light and shadow aesthetics, material properties and energy consumption. For the first time, the light and shadow conception of traditional architecture is converted into quantifiable engineering parameters, and targeted optimization of different building functions is achieved through dynamic adjustment of coefficients.

9. The system for improving performance of traditional buildings through micro-renovation based on machine learning as claimed in claim 1, characterized in that: The effect simulation module is internally equipped with a computational fluid dynamics engine, a virtual reality environment, and a full life cycle analysis framework. The fluid dynamics engine uses the finite volume method to discretize the building microclimate model, solves the energy equation involving radiation heat transfer and natural convection, and sets an adaptive mesh refinement strategy to capture sudden changes in airflow at door and window openings. The virtual reality environment is built on Unreal Engine, using the MotionCapture system to collect human motion data to drive the virtual visitor movement route, and uses ray tracing technology to render the projection changes of carved window lattices. The full life cycle analysis framework uses the Ecoinvent database to construct a building materials list and runs a Monte Carlo simulation to calculate the embodied carbon emission distribution spectrum of the renovation plan over a 50-year period.

10. The system for improving performance of traditional buildings through micro-renovation based on machine learning as claimed in claim 1, characterized in that: The dynamic decision optimization module consists of a reinforcement learning decision tree, a market forecasting model, and a risk sensitivity map. The reinforcement learning decision tree uses the DDPG algorithm architecture, defining solution modification actions as parameter fine-tuning in a continuous action space. The reward function integrates physical performance improvement and cultural coordination to generate instant feedback. The market forecasting model is connected to a time-series database of building material prices, and uses the Prophet algorithm to decompose the seasonal and trend components of price fluctuations to predict the main material procurement cost range during the renovation implementation phase. The risk sensitivity map uses graph theory algorithms to analyze the relationship network of building components, calculate the intensity of the structural chain reaction caused by each renovation measure, and output a color-coded intervention risk classification heat map to support the generation of a multi-dimensional weighted scoring system for the final plan.

Images