A method and system for health assessment of historic building structures

Abstract

This invention relates to the field of historical building structural assessment technology, and discloses a method and system for health assessment of historical building structures. The method includes: using non-destructive testing technology to comprehensively monitor the historical building structure, acquiring multi-source data, and preprocessing the multi-source data; based on the preprocessed multi-source data, analyzing the service performance degradation patterns of materials and nodes, constructing a structural state evolution model, and identifying detailed information about structural damage through a trained neural network model; based on on-site monitoring data, constructing a digital twin model of the historical building, simulating structural response characteristics under different scenarios, and comparing it with actual monitoring data to determine the building's health status; and calculating a comprehensive risk assessment index and classifying the risk based on detailed information about structural damage and the building's health status. This invention achieves a systematic, intelligent, and non-destructive health assessment of historical building structures.

Description

Technical Field

[0001] This invention relates to the field of historical building structure assessment technology, and in particular to a method and system for health assessment of historical building structures. Background Technology

[0002] Historic buildings are an important part of human cultural heritage, carrying rich historical information and cultural value. However, due to their age, material aging, environmental erosion, and natural disasters, the structural safety of historic buildings often faces severe challenges. Once the structure is damaged, it will not only cause irreparable cultural loss but may also threaten public safety. Therefore, conducting scientific and accurate health assessments of historic buildings, promptly identifying potential risks, and taking reinforcement and protection measures are of paramount importance.

[0003] Existing methods for assessing the health of historical buildings mainly rely on manual visual inspection, local sampling testing, or single non-destructive testing methods. These traditional methods have many limitations: First, manual inspection is highly subjective, making it difficult to detect hidden internal damage, and it is also inefficient; second, local sampling testing is destructive, which does not conform to the principle of "minimal intervention" in the protection of historical buildings; third, the data obtained by a single testing method has limited dimensions and cannot fully reflect the complex state of the structure.

[0004] Furthermore, existing assessment systems often lack a systematic analysis of the degradation patterns of structural service performance, making it difficult to predict future structural evolution trends. While some studies have introduced numerical simulation techniques, most remain at the static analysis level, failing to effectively combine real-time monitoring data with dynamic models, resulting in assessment results lagging behind the actual structural condition. Simultaneously, existing risk assessment methods mostly employ qualitative descriptions or simple weighted scoring, lacking quantitative comprehensive assessment indicators based on multi-source data fusion and intelligent algorithms, making it difficult to accurately identify high-risk areas and guide tiered management.

[0005] With the development of IoT, big data, and AI technologies, although there have been attempts to apply these new technologies to modern building health monitoring, specialized health assessment methods for historical buildings (such as material heterogeneity, complex structural forms, and lack of original design data) are still lacking. Therefore, there is an urgent need for a health assessment method and system for historical building structures to achieve systematic, intelligent, and non-destructive health assessment of historical building structures. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for health assessment of historical building structures, aiming to solve one or more of the problems mentioned above.

[0007] This invention provides a method for health assessment of historical building structures, the method comprising: Non-destructive testing technology is used to conduct comprehensive monitoring of the structure of historical buildings, acquire multi-source data, and preprocess the multi-source data; Based on the preprocessed multi-source data, the service performance degradation law of materials and nodes is analyzed, a structural state evolution model is constructed, and detailed information of structural damage is identified through a trained neural network model. Based on on-site monitoring data, a digital twin model of the historical building is constructed to simulate the structural response characteristics under different scenarios and compare it with the actual monitoring data to determine the health status of the building. Based on the detailed information of the structural damage and the health status of the building, a comprehensive risk assessment index is calculated, and the risk is classified.

[0008] Preferably, non-destructive testing technology is used to conduct comprehensive monitoring of the historical building structure and obtain multi-source data, including: Multiple non-destructive testing technologies, including infrared thermal imaging, stress wave, ground-penetrating radar, and three-dimensional laser scanning, are used to conduct comprehensive monitoring of the structure of historical buildings and obtain multi-source data. The multi-source data includes: spatial geometry, material properties, temperature distribution, and environmental parameter data of the building surface and interior.

[0009] Preferably, the multi-source data is preprocessed, including: The multi-source data is subjected to grayscale conversion, noise reduction, and smoothing. Thresholding is performed on the processed multi-source data to identify corroded and cracked areas. A connectivity labeling algorithm is used to group connected defective pixels into the same region, generating a preliminary assessment of the health status of the building surface.

[0010] Preferably, based on preprocessed multi-source data, the service performance degradation patterns of materials and nodes are analyzed, a structural state evolution model is constructed, and a trained neural network model is used to identify detailed information about structural damage, including: Based on the preprocessed multi-source data, the service performance degradation law of materials and nodes is analyzed using linear regression or random forest models, and the structural state evolution model is constructed. Construct a weighted loss function for damage categories, and train a neural network model or support vector machine model using an optimization algorithm; Improve the balance of training data through sample balancing and background enhancement techniques; Use a trained model to identify detailed information about structural damage in a building structure.

[0011] Preferably, the detailed information on the structural damage includes the damage type, cracks, and corrosion problems.

[0012] Preferably, based on on-site monitoring data, a digital twin model of the historical building is constructed to simulate the structural response characteristics under different scenarios, and compared with actual monitoring data to determine the building's health status, including: Based on on-site monitoring data, a digital twin model of the historical building was constructed using a finite element software package; Simulate the structural response characteristics of a digital twin model under different loads; By comparing the structural response characteristics of actual monitoring data with those of a digital twin model using deep learning algorithms, discrepancies can be identified, and the health status of the building can be determined.

[0013] Preferably, based on detailed information about the structural damage and the building's health status, a comprehensive risk assessment index is calculated, and risk classification is performed, including: Based on the detailed information of the structural damage and the health status of the building, the surface change risk assessment index and the internal change risk assessment index are calculated respectively. The comprehensive risk assessment index of the building is calculated based on the preset weighting coefficients. Risk levels are determined based on the comprehensive risk assessment index to identify high-risk areas.

[0014] Preferably, the weighting coefficients include surface change risk weighting coefficients and internal change risk weighting coefficients, and the sum of surface change risk weighting coefficients and internal change risk weighting coefficients is 1.

[0015] Preferably, risk classification is performed based on the comprehensive risk assessment index to identify high-risk areas, including: Several risk assessment index thresholds are preset. The comprehensive risk assessment index is compared with the risk assessment index thresholds. The risk is classified according to the comparison results, and high-risk areas are determined based on the classification results.

[0016] This invention also discloses a health assessment system for historical building structures, used to apply the above-described health assessment method for historical building structures, the system comprising: The multi-source data acquisition module is configured to use non-destructive testing technology to conduct comprehensive monitoring of the structure of historical buildings, acquire multi-source data, and preprocess the multi-source data; The damage detection and identification module is configured to analyze the service performance degradation law of materials and nodes based on preprocessed multi-source data, construct a structural state evolution model, and identify detailed information of structural damage through a trained neural network model. The health status analysis module is configured to build a digital twin model of the historical building based on on-site monitoring data, simulate the structural response characteristics under different scenarios, and compare it with the actual monitoring data to determine the health status of the building. The risk assessment module is configured to calculate a comprehensive risk assessment index and classify the risk based on detailed information about the structural damage and the health status of the building.

[0017] Compared with existing technologies, the advantages of this invention lie in its ability to simultaneously acquire multi-dimensional data on the spatial geometry, material properties, temperature distribution, and environmental parameters of building surfaces and interiors by combining multiple non-destructive testing technologies such as infrared thermal imaging, stress wave detection, ground-penetrating radar, and 3D laser scanning. This overcomes the limitations of single-method detection. Furthermore, by combining grayscale conversion, noise reduction, smoothing, and thresholding processes, and utilizing a connected component labeling algorithm to merge defect pixels, data noise interference is effectively eliminated, improving the coherence and accuracy of defect area identification. This provides a high-quality data foundation for subsequent analysis and effectively solves the problems of insufficient comprehensiveness and non-destructiveness in detection.

[0018] This invention analyzes the degradation patterns of service performance by constructing a structural state evolution model. Combined with a neural network or support vector machine model optimized using sample balancing and background enhancement techniques, it can automatically and accurately identify detailed information about structural damage, including specific damage types, crack morphology, and corrosion levels. This identification method, based on weighted loss functions and deep learning, significantly reduces reliance on human experience, improves the detection rate of minor damage and complex damage patterns, and effectively solves the problems of low accuracy and automation in damage identification.

[0019] By constructing digital twin models of historical buildings and using finite element method (FEM) software to simulate structural responses under different load scenarios, this invention transforms static monitoring data into dynamic evolutionary analysis. In particular, by using deep learning algorithms to compare the response characteristics of actual monitoring data with those of the digital twin model to identify discrepancies, it can keenly capture the deviations between the actual structural state and the theoretical model, thereby more accurately determining the current health status of the building. This method not only reflects the current state but also reveals the potential responses of the structure under specific working conditions, solving the problems of insufficient dynamism and lack of predictive ability in traditional condition assessment methods.

[0020] This invention innovatively quantifies surface change risk and internal change risk separately, and calculates a comprehensive risk assessment index through preset weighting coefficients. This combined internal and external assessment method avoids the one-sidedness of traditional single-dimensional assessments. Furthermore, by comparing the comprehensive index with several preset thresholds, it achieves clear risk classification and precise location of high-risk areas. This makes the formulation of protection measures more targeted, enabling priority to be given to high-risk areas, optimizing resource allocation, and effectively solving the problem of insufficient comprehensiveness and scientific rigor in risk assessment, providing strong decision support for the preventive protection of historical buildings. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating a health assessment method for historical building structures according to the present invention. Figure 2 This is a functional block diagram of a health assessment system for historical building structures according to the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0024] like Figure 1 As shown, the present invention provides a method for health assessment of historical building structures, the method comprising: Non-destructive testing technology is used to conduct comprehensive monitoring of the structure of historical buildings, acquire multi-source data, and preprocess the multi-source data; Based on the preprocessed multi-source data, the service performance degradation law of materials and nodes is analyzed, a structural state evolution model is constructed, and detailed information of structural damage is identified through a trained neural network model. Based on on-site monitoring data, a digital twin model of the historical building is constructed to simulate the structural response characteristics under different scenarios and compare it with the actual monitoring data to determine the health status of the building. Based on the detailed information of the structural damage and the health status of the building, a comprehensive risk assessment index is calculated, and the risk is classified.

[0025] This invention employs non-destructive testing technology to acquire and preprocess multi-source data, enabling the comprehensive and accurate collection of various information reflecting the structural condition of historical buildings without damaging the building structure itself. This provides a reliable data foundation for subsequent assessments. Based on the preprocessed data, a structural state evolution model and a trained neural network model can be constructed to deeply analyze the service performance degradation patterns of materials and nodes, accurately identify the location, type, and extent of structural damage, and achieve early warning and precise location of internal structural damage. Furthermore, the construction of a digital twin model, by simulating structural response characteristics under different scenarios and comparing them with actual monitoring data, can dynamically and intuitively reflect the building's health status, providing a visualized analytical tool for assessment. Finally, the comprehensive risk assessment index and risk classification calculated by combining structural damage information and health status can provide a scientific basis for decisions regarding the protection, maintenance, and management of historical buildings, effectively improving the accuracy, comprehensiveness, and intelligence level of structural health assessments, thereby better ensuring the safety and durability of historical buildings and preserving their historical and cultural value.

[0026] In some embodiments of this application, non-destructive testing techniques are used to conduct comprehensive monitoring of the historical building structure and obtain multi-source data, including: using a variety of non-destructive testing techniques such as infrared thermal imaging, stress wave, ground-penetrating radar, and three-dimensional laser scanning to conduct comprehensive monitoring of the historical building structure and obtain multi-source data; the multi-source data includes: spatial geometry, material properties, temperature distribution, and environmental parameter data of the building surface and interior.

[0027] Understandably, non-destructive methods can comprehensively and multidimensionally acquire information on the physical state of historical building structures, addressing the limitations of traditional testing methods such as blind spots, limited data, and potential damage to the artifacts themselves. By integrating multiple non-destructive testing technologies, the system can simultaneously capture the geometric morphology of the building surface, internal material defects, temperature field distribution, and surrounding environmental parameters, providing a detailed and reliable multi-source data foundation for subsequent high-precision digital twin model construction, analysis of material degradation patterns, and identification of hidden damage.

[0028] Specifically, based on the structural characteristics, material types, and specific parts to be inspected of historical buildings, one or more combinations of four technologies—infrared thermal imaging, stress wave detection, ground-penetrating radar, and 3D laser scanning—are selected for comprehensive monitoring. 3D laser scanning is primarily used to quickly acquire high-precision point cloud data of the entire historical building and its parts, reconstructing the spatial geometry of the building surface and accurately recording structural deformation, displacement, and surface morphology. Infrared thermal imaging receives infrared radiation from the surface of an object to generate temperature distribution images, effectively identifying hidden defects such as hollowing, water seepage, and peeling within walls by utilizing differences in thermal conductivity between different materials or defect areas. Ground-penetrating radar utilizes the propagation and reflection characteristics of high-frequency electromagnetic waves in a medium to penetrate the building surface and detect the distribution of internal components, crack depth, steel reinforcement corrosion, and internal voids in materials. Stress wave detection analyzes the propagation speed and attenuation patterns of stress waves in materials to assess the internal integrity and degree of mechanical property degradation of materials such as wood, stone, or brick masonry. During the aforementioned testing, environmental parameter data, including ambient temperature, humidity, wind speed, and lighting conditions, were collected simultaneously. Mathematical model compensation, dynamic calibration compensation, hardware compensation, or comparative experimental compensation methods were employed to compensate for the data and eliminate the interference of environmental factors on the test results. Finally, the data obtained through these different technical means were integrated to form a multi-source dataset containing geometric coordinates of the building surface and interior space, material physical and mechanical properties, a full-field temperature distribution cloud map, and real-time environmental parameters. This multi-source data not only covers all levels from macroscopic geometry to microscopic material properties but also takes into account both visible surface damage and hidden internal defects, ensuring the comprehensiveness and multidimensionality of the monitoring data and laying a solid foundation for subsequent data preprocessing and intelligent analysis.

[0029] Among them, the mathematical model compensation method involves using a large amount of historical detection data and corresponding environmental parameters to construct a mathematical model between environmental factors and detection errors using algorithms such as multiple regression analysis and neural networks. For example, regarding the influence of temperature on a certain detection index, a temperature-error curve equation is fitted using experimental data. During real-time detection, the current temperature value is substituted into the equation to calculate the theoretical error value at that temperature. This error value is then subtracted from the original detection data, thereby compensating for temperature interference. For other environmental parameters such as humidity, wind speed, and light intensity, models can be established separately using similar methods, or a multi-input single-output comprehensive model can be constructed to compensate for the combined interference of multiple environmental factors. The dynamic calibration compensation method involves periodically calibrating the detection system under different environmental conditions to obtain calibration coefficients for different combinations of environmental parameters. During actual detection, current environmental parameters are collected in real time. By looking up a preset calibration coefficient table or using interpolation algorithms, the calibration coefficients for the corresponding environmental conditions are obtained. Multiplying the original detection data by these calibration coefficients eliminates environmental interference. For example, in an environment with a light intensity of 5000 lux, a temperature of 25℃, and a humidity of 60%, the calibration coefficient is 1.02. When the current environmental parameters are detected to match, the original detection data is multiplied by 1.02 for compensation. Hardware compensation method: By integrating specific sensors and compensation circuits into the detection system, the changes in hardware characteristics caused by environmental factors are directly compensated. For example, temperature affects the parameters of components such as resistors and capacitors in the detection circuit. A temperature sensor and corresponding compensation circuit can be added to the circuit. When the temperature changes, the compensation circuit automatically adjusts the component parameters to keep the output of the detection circuit stable, thereby reducing the interference of temperature on the detection results. For factors such as humidity and wind speed, corresponding hardware compensation modules can also be designed according to their influence mechanism on the hardware. Comparative experiment compensation method: While conducting formal testing, a set of comparative experiments is set up that are unaffected by environmental interference or have constant environmental conditions. For example, a standard sample is placed in an ideal environment with constant temperature and humidity, no light, and no wind for testing, and its test results are used as the benchmark value. In actual testing, the test results of the sample under the current environment are compared with the benchmark value to calculate the deviation caused by environmental interference. Then, the deviation is subtracted from the actual test results to achieve interference compensation. This method can intuitively reflect the comprehensive impact of environmental factors and is especially suitable for complex testing scenarios where it is difficult to establish an accurate mathematical model.

[0030] Specifically, an infrared thermal imager (FLIR E75, resolution 640×480, thermal sensitivity 0.03℃) was used to perform a full-range scan of the surface of the historical building to obtain surface temperature distribution data. At the same time, a ground-penetrating radar (Ingersoll Rand SR-16, center frequency 160kHz, power 160W) was used to perform a three-dimensional scan of the building surface to obtain surface height and roughness data. Finally, a vibration sensor (PCI-2, sensitivity 2.5V / g) and an environmental monitoring sensor (SHT30, accuracy ±0.3℃, ±2%RH) were deployed to monitor environmental parameters.

[0031] In some embodiments of this application, the multi-source data is preprocessed, including: grayscale conversion, noise reduction, and smoothing of the multi-source data; thresholding of the processed multi-source data to identify corrosion areas and crack areas; and merging connected defect pixels into the same region using a connectivity labeling algorithm to generate a preliminary assessment result of the building surface health status.

[0032] Understandably, by eliminating redundant information and errors caused by environmental interference, equipment noise, or unstable signal transmission in multi-source detection data, the quality and signal-to-noise ratio of the data can be improved. Through image segmentation and feature extraction techniques, defect areas representing corrosion and cracks can be accurately separated from complex backgrounds. By using connected component analysis to integrate discrete defect pixels into physically meaningful independent defect objects, a quantitative description of surface damage to historical buildings can be achieved, providing standardized and highly usable preliminary assessment data for subsequent structural state evolution analysis and neural network model training.

[0033] Specifically, the collected multi-source raw data is converted to grayscale, transforming color images or multi-channel data into single-channel grayscale images to reduce data dimensionality and highlight texture and intensity features, facilitating subsequent calculations. Subsequently, filtering algorithms (such as Gaussian filtering, median filtering, or wavelet denoising) are used to denoise and smooth the grayscale data, effectively filtering out random noise points, smoothing image edges, and preserving key features of defect areas to prevent misjudgments due to noise interference. After basic denoising, the processed data is thresholded. Based on the significant differences in grayscale values ​​or feature values ​​between historical building materials (such as brick, stone, and wood) and defects (such as cracks and corrosion pits), appropriate global or adaptive thresholds are selected (global threshold selection methods include histogram thresholding, maximum inter-class variance, and fixed thresholding; adaptive threshold selection methods include local adaptive thresholding, edge-based thresholding, and iterative thresholding). Image pixels are divided into foreground (defects) and background (normal structures), thus initially identifying and extracting binary images of corrosion and crack areas. Finally, for potentially fragmented and discrete defect pixels in the binarized image, a connected component labeling algorithm is used for post-processing. This algorithm traverses the image pixels, grouping spatially adjacent defect pixels with the same attributes into a single connected region and assigning it a unique label ID. Through this process, the originally scattered pixels are integrated into complete defect patches, enabling accurate calculation of geometric parameters such as area, perimeter, and centroid of each defect region. It also eliminates minor pseudo-defects caused by residual noise, ultimately generating a preliminary assessment of the building surface health status, including the location, shape, and distribution of defects, serving as input for subsequent in-depth analysis.

[0034] Specifically, in addition to image data, for the spatial geometric data of building surfaces and interiors, coordinate normalization is used to uniformly transform 3D point cloud or BIM model data from different acquisition coordinate systems to a preset reference coordinate system, eliminating scale differences and positional offsets. For missing geometric data points, nearest neighbor interpolation or surface fitting algorithms are used to complete them, ensuring the integrity of the spatial topology. For material property data (such as density, thermal conductivity, compressive strength, etc.), outlier detection is first performed, using the Z-score method or box plot method to identify and remove outliers that exceed reasonable ranges. Then, data standardization (such as Min-Max normalization) is used to transform material parameters of different dimensions to the [0,1] interval, facilitating the fusion analysis of multi-attribute data. During the preprocessing of temperature distribution data, the temperature values ​​collected by infrared thermal imaging or sensors are first spatiotemporally aligned to ensure the comparability of temperature data at different time points in the same monitoring area. For local abrupt noise that may exist in the temperature data, moving average filtering or Kalman filtering is used for smoothing. At the same time, spatial interpolation (such as Kriging interpolation) is used to convert discrete temperature sampling points into continuous temperature field distribution images. Environmental parameter data (such as temperature, humidity, light intensity, wind speed, etc.) need to be time-series aligned and missing values ​​filled. Linear interpolation or LSTM neural network prediction models are used to supplement discontinuous data, and Fourier transform or wavelet transform is used to extract the periodic variation characteristics of environmental parameters, so as to provide a data foundation for subsequent analysis of the impact of environmental factors on the health of building structures.

[0035] Specifically, Matlab R2020a software was used to perform grayscale processing on the acquired infrared thermal imaging data to remove noise; a median filtering algorithm was used to smooth the data with a filter kernel size of 5×5; a threshold of 60℃ was set, and the processed data was binarized to identify temperature anomaly areas; a connectivity labeling algorithm was used to classify connected anomaly areas into the same category to form a preliminary assessment result of the building surface health status.

[0036] In some embodiments of this application, based on preprocessed multi-source data, the service performance degradation patterns of materials and nodes are analyzed, a structural state evolution model is constructed, and detailed information on structural damage is identified using a trained neural network model. This includes: analyzing the service performance degradation patterns of materials and nodes using linear regression or random forest models based on preprocessed multi-source data, and constructing the structural state evolution model; constructing a weighted loss function for damage categories, and training a neural network model or support vector machine model using optimization algorithms; improving the balance of training data through sample balancing and background enhancement techniques; and using the trained model to identify detailed information on structural damage in the building structure.

[0037] In some embodiments of this application, the details of the structural damage include damage type, cracks, and corrosion problems.

[0038] Understandably, by deeply exploring the performance evolution mechanisms of historical building materials and nodes during long-term service, the approach upgrades from static data detection to dynamic pattern analysis. By constructing a structural state evolution model, the degradation trend of material properties over time or environmental changes can be quantified. Simultaneously, an intelligent recognition model trained with special optimization strategies (such as weighted loss functions, sample balancing, and background enhancement) addresses the recognition challenges caused by scarce historical building damage samples, class imbalance, and complex backgrounds. This enables automated and high-precision extraction of detailed information such as damage type, crack morphology, and corrosion degree, providing accurate damage quantification indicators for subsequent digital twin modeling and risk assessment.

[0039] Specifically, based on preprocessed multi-source data, linear regression or random forest models are selected as analysis tools. Environmental exposure time, temperature and humidity history, and stress levels of the materials are used as input features, while performance indicators such as material strength, elastic modulus, or node connection stiffness are used as output targets. Degradation curves or mapping relationships between material and node service performance are fitted, thereby constructing a structural state evolution model that reflects the evolution of the structural state over time or under different operating conditions to predict future trends in structural performance. Secondly, to train a high-precision damage identification model, a weighted loss function for each damage category is constructed. Given the scarcity of severe damage samples and the abundance of minor damage samples in historical buildings, rare or important damage categories are assigned higher weights, forcing the model to focus more on difficult-to-identify damage features during training. Based on this, gradient descent and other optimization algorithms are used to iteratively train neural network models (such as convolutional neural networks) or support vector machines, continuously adjusting model parameters to minimize the weighted loss. During the training data preparation phase, sample balancing and background enhancement techniques are implemented. By sampling minority class samples or undersampling majority class samples, the uneven distribution of damage types in the training data is addressed, preventing the model from biased towards predicting common categories. Simultaneously, background enhancement techniques such as rotation, scaling, adding noise, or altering lighting conditions are used to expand the diversity of the training dataset, simulating the complex surface textures and variable on-site environments of historical buildings, thus improving the model's generalization ability and robustness. Finally, the trained model is used to perform a full-element scan and identification of the historical building structure. The model automatically outputs detailed information on structural damage, including: identifying the specific type of damage (e.g., stress cracks, temperature cracks, weathering spalling, insect infestation, decay, etc.); quantifying the geometric characteristics of cracks (e.g., length, width, direction, and distribution density); and assessing the severity of corrosion problems (e.g., corrosion area percentage, depth estimation, and material loss rate). This detailed information will serve as key inputs for refining the digital twin model parameters and calculating subsequent risk assessment indices.

[0040] Specifically, based on the acquired multi-source data, the LinearRegression model from the scikit-learn library in Python was used to analyze the service performance degradation patterns of materials and nodes, constructing a model of the evolution of the state of historical building structures. A weighted loss matrix for damage types (no damage, cracks, corrosion, etc.) was constructed, and the Adam optimization algorithm was used to train a neural network model. The model structure is a 5-layer convolutional neural network with 64 hidden layer units. Random sampling was used to balance the dataset, making the ratio of samples of each type close to 1:1. Image enhancement techniques were used to expand the dataset and increase the number of training samples. The trained model was then used to identify the acquired test data and obtain detailed information on structural damage.

[0041] In some embodiments of this application, a digital twin model of a historical building is constructed based on on-site monitoring data to simulate the structural response characteristics under different scenarios and compare it with actual monitoring data to determine the health status of the building. This includes: constructing a digital twin model of the historical building using a finite element software package based on on-site monitoring data; simulating the structural response characteristics of the digital twin model under different loads; and comparing the structural response characteristics of the actual monitoring data and the digital twin model using a deep learning algorithm to identify areas of difference and determine the health status of the building.

[0042] Understandably, constructing a digital twin that maps and interacts dynamically with the physical historical building in real time is a way to address the problem that traditional static assessment methods cannot reflect the true response of the structure under complex working conditions. Through finite element modeling and multi-scenario simulation, the mechanical behavior of the building under different load environments is reproduced. Deep learning algorithms are used to compare theoretical simulation values ​​with actual monitoring values ​​with high precision, automatically identifying the deviation areas between the two. This allows for a deeper understanding of the structure's intrinsic health status beyond superficial data, enabling early warning of potential hazards and dynamic assessment of overall safety.

[0043] Specifically, based on the on-site monitoring data obtained in the preceding steps (including spatial geometric dimensions, material property parameters, boundary conditions, and environmental loads), a high-fidelity digital twin model of the historical building is established using professional finite element software packages (such as ANSYS and ABAQUS). This model not only accurately reproduces the building's geometric topology but also corrects the constant model parameters according to the measured material degradation laws, ensuring a high degree of consistency between the digital model and the physical entity in the initial state. Subsequently, based on the constructed digital twin model, numerical simulations are performed under various typical and extreme conditions. These conditions cover daily dead loads, live loads, as well as wind loads, seismic effects, and sudden temperature changes. Through calculation, structural response characteristic data such as displacement, stress, strain, and vibration frequency of key parts (such as nodes, supports, and crack tips) of the digital twin model under the above various loads are obtained, forming a benchmark database of theoretical responses. Finally, deep learning algorithms (such as Long Short-Term Memory Networks (LSTM) or Convolutional Neural Networks (CNN)) are introduced as comparative analysis tools. The algorithm inputs real-time monitoring data sequences collected by actual sensors and theoretical response characteristic sequences output by the digital twin model under corresponding working conditions. The algorithm automatically learns the data mapping relationship under normal conditions and calculates the residuals between actual and theoretical values ​​in real time. When the residual exceeds a preset confidence interval, the algorithm marks it as a discrepancy region. These discrepancy regions often correspond to unforeseen damage expansion, loosening of connections, or abrupt changes in material properties. By analyzing the distribution range, amplitude, and evolution trend of these discrepancy regions, the algorithm comprehensively determines the current health status of the building (e.g., healthy, sub-healthy, damaged, or dangerous), thereby achieving an intelligent leap from "data monitoring" to "condition diagnosis."

[0044] Specifically, the criteria for comprehensively determining the current health status of a building are as follows: Using deep learning algorithms (such as LSTM or CNN), the real-time monitoring data sequences collected by actual sensors are compared and analyzed with the theoretical response characteristic sequences output by the digital twin model under corresponding operating conditions, and the residuals between the two are calculated. When the residual exceeds a preset confidence interval, it is marked as a difference region. A comprehensive judgment is then made based on the distribution range, amplitude, and evolution trend of these difference regions. Specifically, difference regions correspond to situations such as damage expansion, loosening of connections, or abrupt changes in material properties that the model did not foresee. Analyzing their distribution range can determine the spatial impact of the damage; analyzing the amplitude can assess the severity of the damage; and analyzing the evolution trend can predict the speed of damage development and potential risks. Finally, combining these three aspects of information, the building is comprehensively determined to be in a healthy, sub-healthy, damaged, or dangerous state.

[0045] Specifically, based on on-site monitoring data, a digital twin model of the historical building was constructed using the PyFEM open-source finite element software package in Python to simulate the structural response characteristics under different loads. By comparing the response characteristics of the virtual model, a deep learning-based difference recognition algorithm was used to identify the areas of difference between the actual structure and the virtual model. Combined with structural health monitoring data, the response characteristics of the virtual model were compared to identify the areas of difference between the actual structure and the virtual model, thus determining the health status of the building.

[0046] In some embodiments of this application, a comprehensive risk assessment index is calculated and risk classification is performed based on detailed information about the structural damage and the health status of the building. This includes: calculating a surface change risk assessment index and an internal change risk assessment index based on detailed information about the structural damage and the health status of the building, respectively; calculating a comprehensive risk assessment index for the building based on preset weighting coefficients; and classifying the risk based on the comprehensive risk assessment index to identify high-risk areas.

[0047] In some embodiments of this application, the weighting coefficients include surface change risk weighting coefficients and internal change risk weighting coefficients, and surface change risk weighting coefficients + internal change risk weighting coefficients = 1.

[0048] In some embodiments of this application, risk classification is performed based on the comprehensive risk assessment index to identify high-risk areas, including: presetting several risk assessment index thresholds, comparing the comprehensive risk assessment index with the risk assessment index thresholds, classifying the risk based on the comparison results, and determining high-risk areas based on the classification results.

[0049] Understandably, establishing a quantitative, scientific, and comprehensive risk assessment system for historical building structures can address the shortcomings of existing assessment methods, such as excessive qualitative descriptions, insufficient quantitative indicators, and fragmented evaluation of surface damage and internal hazards. By calculating risk indices for both the surface and interior surfaces and assigning appropriate weights, a comprehensive risk assessment index can be constructed. This transforms complex, multi-dimensional testing data into intuitive risk values. Furthermore, standardized risk grading can be performed based on this index, accurately identifying high-risk areas and providing clear priority guidance and scientific basis for differentiated protection, reinforcement decisions, and preventative maintenance of historical buildings. Weighting coefficients can be preset based on historical data, experimental verification, and expert experience.

[0050] Specifically, based on the detailed structural damage information (such as crack width and corrosion area) identified in the previous steps and the determined building health status (such as displacement deviation and stress concentration), calculation models for surface change risk assessment index and internal change risk assessment index are constructed respectively. The surface change risk assessment index primarily quantifies the impact of visible surface defects on the structural integrity and durability; the internal change risk assessment index focuses on quantifying the potential threats to the overall structural load-bearing capacity and stability posed by hidden internal defects, material performance degradation, and joint connection failures. Subsequently, preset weighting coefficients are introduced to weight and fuse the two sub-indices, calculating the building's comprehensive risk assessment index. These weighting coefficients include a surface change risk weighting coefficient and an internal change risk weighting coefficient. These two coefficients are preset based on the historical building's structural type, material characteristics, and protection requirements, and strictly satisfy the constraint that their sum equals 1. This constraint ensures the mathematical normalization of the comprehensive index while allowing assessors to flexibly adjust the emphasis on surface aesthetics and internal safety according to actual conditions, achieving a dynamic balance of assessment dimensions. Finally, risk grading is performed based on the calculated comprehensive risk assessment index. Several risk assessment index thresholds are pre-set (e.g., low-risk, medium-risk, high-risk thresholds). The comprehensive index of each monitored area or component is compared with these thresholds one by one. Based on the comparison results, the building structure is divided into different risk levels (e.g., Level 1 Safety, Level 2 Caution, Level 3 Warning, Level 4 Danger, etc.). For areas determined to be high-risk, the system highlights them on the spatial model or outputs specific coordinates, thereby accurately identifying high-risk areas that urgently require intervention. This process not only achieves quantitative risk classification but also directly outputs specific control targets, guiding managers to prioritize reinforcement or repair measures for high-risk areas, optimize resource allocation, and ensure the safe preservation of historical buildings.

[0051] Specifically, the threshold setting for the risk assessment index mainly includes the following aspects: First, it is based on the ultimate limit of bearing capacity and the benchmark value of durability design of historical building structures. This is the fundamental bottom line for ensuring the safety of building structures. The threshold setting must be lower than the critical value that may lead to structural instability or functional failure, so as to provide early warning of potential risks. Second, it refers to the risk accident case data accumulated in the long-term monitoring of similar historical buildings, and analyzes the distribution characteristics of the comprehensive index corresponding to different risk levels, so that the threshold division conforms to actual engineering experience. At the same time, combined with the protection level and restoration priority of historical buildings, for buildings with important historical and cultural value, the high-risk threshold can be appropriately lowered to conduct risk management with stricter standards; while for minor components or areas with lower restoration difficulty, the threshold range can be appropriately relaxed. In addition, the aging rate of structural materials and environmental corrosion factors must also be considered. For example, in a humid environment, the corrosion risk threshold of concrete structures should be lower than that in a dry environment to adapt to the degree of impact of different environmental conditions on structural risks. Finally, the threshold setting must be consistent with existing building safety assessment standards and industry specifications to ensure that the risk classification results are authoritative and comparable, so as to facilitate managers to formulate subsequent protection and restoration strategies based on general standards.

[0052] For example, based on detailed information about structural damage and the building's health status, the risk weight for surface changes is set to 0.4, and the risk weight for internal changes is set to 0.6. The surface change risk assessment index and the internal change risk assessment index are calculated. Based on the weight coefficients of each assessment index, the building's comprehensive risk assessment index is calculated. The risk is classified according to the risk assessment index, and areas with a risk assessment index greater than 0.7 are identified as high-risk areas.

[0053] Understandably, by transforming continuously quantified comprehensive risk assessment indices into discrete, standardized risk levels, the problem of assessment results being difficult to directly guide engineering decisions can be addressed. Through a pre-set threshold comparison mechanism, a unified risk judgment benchmark is established, enabling rapid classification and grading of the safety status of different parts of historical buildings. Furthermore, based on the grading results, high-risk areas can be precisely identified, and key areas requiring priority intervention can be clearly defined. This provides an intuitive and actionable basis for developing differentiated repair plans, allocating maintenance resources, and implementing tiered management.

[0054] like Figure 2 As shown, the present invention also discloses a health assessment system for historical building structures, used to apply the above-described health assessment method for historical building structures, the system comprising: The multi-source data acquisition module is configured to use non-destructive testing technology to conduct comprehensive monitoring of the structure of historical buildings, acquire multi-source data, and preprocess the multi-source data; The damage detection and identification module is configured to analyze the service performance degradation law of materials and nodes based on preprocessed multi-source data, construct a structural state evolution model, and identify detailed information of structural damage through a trained neural network model. The health status analysis module is configured to build a digital twin model of the historical building based on on-site monitoring data, simulate the structural response characteristics under different scenarios, and compare it with the actual monitoring data to determine the health status of the building. The risk assessment module is configured to calculate a comprehensive risk assessment index and classify the risk based on detailed information about the structural damage and the health status of the building.

[0055] This invention achieves non-destructive and comprehensive data acquisition and preprocessing of historical building structures through a multi-source data acquisition module, providing a high-quality data foundation for subsequent assessments. The damage detection and identification module uses a neural network model to accurately identify structural damage details and combines this with a structural state evolution model to deeply understand the degradation patterns of material and node performance. The health status analysis module simulates structural responses under different scenarios using a digital twin model and compares them with actual monitoring data, enabling dynamic and accurate judgment of the building's health status. The risk assessment module integrates damage information and health status to calculate a risk assessment index and classify it, providing a scientific basis for decisions on the protection and restoration of historical buildings. This effectively improves the accuracy, comprehensiveness, and intelligence level of historical building structural health assessments, contributing to the refined protection and sustainable utilization of historical buildings.

[0056] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program goods. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program goods embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0057] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program goods according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0058] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0059] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for health assessment of historical building structures, characterized in that, The method includes: Non-destructive testing technology is used to conduct comprehensive monitoring of the structure of historical buildings, acquire multi-source data, and preprocess the multi-source data; Based on the preprocessed multi-source data, the service performance degradation law of materials and nodes is analyzed, a structural state evolution model is constructed, and detailed information of structural damage is identified through a trained neural network model. Based on on-site monitoring data, a digital twin model of the historical building is constructed to simulate the structural response characteristics under different scenarios and compare it with the actual monitoring data to determine the health status of the building. Based on the detailed information of the structural damage and the health status of the building, a comprehensive risk assessment index is calculated, and the risk is classified.

2. The health assessment method for historical building structures according to claim 1, characterized in that, Non-destructive testing technology was used to conduct comprehensive monitoring of the historical building structure, acquiring multi-source data, including: Multiple non-destructive testing technologies, including infrared thermal imaging, stress wave, ground-penetrating radar, and three-dimensional laser scanning, are used to conduct comprehensive monitoring of the structure of historical buildings and obtain multi-source data. The multi-source data includes: spatial geometry, material properties, temperature distribution, and environmental parameter data of the building surface and interior.

3. The health assessment method for historical building structures according to claim 1, characterized in that, Preprocessing the multi-source data includes: The multi-source data is subjected to grayscale conversion, noise reduction, and smoothing. Thresholding is performed on the processed multi-source data to identify corroded and cracked areas. A connectivity labeling algorithm is used to group connected defective pixels into the same region, generating a preliminary assessment of the health status of the building surface.

4. The health assessment method for historical building structures according to claim 1, characterized in that, Based on preprocessed multi-source data, the service performance degradation patterns of materials and nodes are analyzed, a structural state evolution model is constructed, and a trained neural network model is used to identify detailed information about structural damage, including: Based on the preprocessed multi-source data, the service performance degradation law of materials and nodes is analyzed using linear regression or random forest models, and the structural state evolution model is constructed. Construct a weighted loss function for damage categories, and train a neural network model or support vector machine model using an optimization algorithm; Improve the balance of training data through sample balancing and background enhancement techniques; Use a trained model to identify detailed information about structural damage in a building structure.

5. The health assessment method for historical building structures according to claim 1, characterized in that, Detailed information on the structural damage includes the type of damage, cracks, and corrosion issues.

6. The health assessment method for historical building structures according to claim 1, characterized in that, Based on on-site monitoring data, a digital twin model of the historical building is constructed to simulate its structural response characteristics under different scenarios. This model is then compared with actual monitoring data to determine the building's health status, including: Based on on-site monitoring data, a digital twin model of the historical building was constructed using a finite element software package; Simulate the structural response characteristics of a digital twin model under different loads; By comparing the structural response characteristics of actual monitoring data with those of a digital twin model using deep learning algorithms, discrepancies can be identified, and the health status of the building can be determined.

7. The health assessment method for historical building structures according to claim 1, characterized in that, Based on the detailed information of the structural damage and the health status of the building, a comprehensive risk assessment index is calculated, and risk classification is performed, including: Based on the detailed information of the structural damage and the health status of the building, the surface change risk assessment index and the internal change risk assessment index are calculated respectively. The comprehensive risk assessment index of the building is calculated based on the preset weighting coefficients. Risk levels are determined based on the comprehensive risk assessment index to identify high-risk areas.

8. The health assessment method for historical building structures according to claim 7, characterized in that, The weighting coefficients include surface change risk weighting coefficients and internal change risk weighting coefficients, and the sum of surface change risk weighting coefficients and internal change risk weighting coefficients is 1.

9. The health assessment method for historical building structures according to claim 7, characterized in that, Risk is classified according to the comprehensive risk assessment index to identify high-risk areas, including: Several risk assessment index thresholds are preset. The comprehensive risk assessment index is compared with the risk assessment index thresholds. The risk is classified according to the comparison results, and high-risk areas are determined based on the classification results.

10. A health assessment system for historical building structures, used for applying the health assessment method for historical building structures as described in any one of claims 1-9, characterized in that, The system includes: The multi-source data acquisition module is configured to use non-destructive testing technology to conduct comprehensive monitoring of the structure of historical buildings, acquire multi-source data, and preprocess the multi-source data; The damage detection and identification module is configured to analyze the service performance degradation law of materials and nodes based on preprocessed multi-source data, construct a structural state evolution model, and identify detailed information of structural damage through a trained neural network model. The health status analysis module is configured to build a digital twin model of the historical building based on on-site monitoring data, simulate the structural response characteristics under different scenarios, and compare it with the actual monitoring data to determine the health status of the building. The risk assessment module is configured to calculate a comprehensive risk assessment index and classify the risk based on detailed information about the structural damage and the health status of the building.

Images