Historical building digital monitoring protection system based on multi-scale characteristics

Abstract

The invention discloses a historical building digital monitoring protection system based on multi-scale features. The system comprises a building data acquisition module, an original data optimization module, a feature extraction model construction module, a state recognition model construction module and a building digital monitoring module. The invention relates to the technical field of historical building digital monitoring, in particular to a historical building digital monitoring protection system based on multi-scale features, and the method comprises the steps: obtaining original data through building data; a data optimization method of data alignment, data cleaning, data standardization and data set segmentation is adopted; a deep learning model is adopted as a feature extraction model, and deep correlation features of building time-varying behaviors and structural damage are learned through feature decoupling driven by time sequence modeling and physical laws; an improved clustering model is adopted as a state recognition model, and robust recognition of the building state is achieved by introducing physical threshold constraint spectral clustering, a dynamic time kernel function and a multi-scale probability fusion mechanism.

Description

Technical Field

[0001] The present invention relates to the technical field of digital monitoring of historical buildings, and in particular to a digital monitoring and protection system for historical buildings based on multi-scale features. Background Art

[0002] The digital monitoring and protection system for historical buildings uses modern information technology to conduct real-time monitoring, data collection and analysis of historical buildings. It accurately obtains changes in parameters such as the building's structure, environment, temperature and humidity through digital means, and promptly detects potential damage or danger. It can effectively extend the service life of historical buildings, avoid damage caused by natural aging or external factors, and provide strong technical support for the protection of historical culture.

[0003] However, traditional digital monitoring and protection systems for historical buildings often use fixed signal processing algorithms or shallow machine learning models to extract features, which are not adaptable enough to the nonlinear degradation process unique to buildings, make it difficult to distinguish between environmental interference and real structural damage, lack a mechanism to embed physical laws, and are easily affected by sensor noise and changes in working conditions. In terms of state recognition, traditional digital monitoring and protection systems for historical buildings cannot capture the time-varying characteristics of the evolution of building damage, ignore material constitutive relationships and environment-structure interaction effects, and are prone to forcibly classifying data with different physical mechanisms. In addition, single-scale feature analysis makes it difficult to take into account both the global macroscopic response and local microscopic damage of the building. Summary of the Invention

[0004] In view of the above situation, in order to overcome the defects of the existing technology, the present invention provides a digital monitoring and protection system for historical buildings based on multi-scale features. In view of the technical problems that traditional digital monitoring and protection systems for historical buildings mostly use fixed signal processing algorithms or shallow machine learning models to extract features, lack adaptability to the nonlinear degradation process unique to buildings, have difficulty in distinguishing environmental interference from real structural damage, lack a physical law embedding mechanism, and are easily affected by sensor noise and working condition changes, this solution creatively adopts a deep learning model as a feature extraction model, and through the decoupling of features driven by adaptive time series modeling and physical laws, automatically learns the deep correlation characteristics between the time-varying behavior of buildings and structural damage, and reflects the complex environment-structure coupling effect of buildings. And ensure the independence of features through orthogonal constraints, providing more discriminative input for subsequent state identification; in view of the fact that traditional digital monitoring and protection systems for historical buildings are unable to capture the time-varying characteristics of building damage evolution in state identification, ignore the material constitutive relationship and environment-structure interaction effects, and are prone to forcibly classify data with different physical mechanisms, and single-scale feature analysis is difficult to take into account the global macro-response and local micro-damage of the building, this scheme creatively adopts an improved clustering model as the state identification model, introduces physical threshold constraint spectral clustering, and adapts to the progressive characteristics of building damage through dynamic time kernel function, combined with multi-scale probability fusion mechanism, to achieve fine-grained perception and robust identification of building state changes.

[0005] The technical solution adopted by the present invention is as follows: the digital monitoring and protection system for historical buildings based on multi-scale features provided by the present invention includes a building data acquisition module, a raw data optimization module, a feature extraction model construction module, a state recognition model construction module and a building digital monitoring module;

[0006] The building data acquisition module obtains the original data set of historical buildings by performing data collection;

[0007] The raw data optimization module adopts data optimization methods such as data alignment, data cleaning, data standardization and data set segmentation to obtain monitoring data sets, monitoring training sets and monitoring test sets;

[0008] The feature extraction model construction module constructs a deep learning model as a feature extraction model to provide building attribute features for subsequent building digital monitoring;

[0009] The state recognition model construction module constructs an improved clustering model as a state recognition model for subsequent building digital monitoring;

[0010] The building digital monitoring module specifically performs building digital monitoring by adopting the feature extraction model in combination with the state recognition model to obtain a building digital monitoring state reference result.

[0011] Furthermore, in the building data acquisition module, the historical building original data set specifically includes the past original data set of the building and the existing original data set of the building. Both the past original data set of the building and the existing original data set of the building include building structure data, building environment data and building material performance data. The past original data set of the building also includes building damage level annotation data.

[0012] Furthermore, in the raw data optimization module, the data alignment is used to solve the problem of sensor spatiotemporal asynchrony in historical building monitoring. Specifically, by synchronizing the clocks of each device and constructing a timestamp alignment matrix, data with strict time axis synchronization is obtained.

[0013] The data cleaning is used to ensure the accuracy and integrity of the data, specifically by eliminating impulse noise through median filtering and removing baseline drift through empirical mode decomposition to obtain cleaned data;

[0014] The data normalization is used to eliminate dimensional distortion caused by environmental factors, specifically by correcting displacement deformation data and power-normalized vibration energy through temperature compensation, and using the Z-Score normalization method to obtain standardized data;

[0015] The data set segmentation is used to segment the data set, specifically to segment the original data set of the building to obtain a monitoring training set and a monitoring test set;

[0016] Through the data alignment, the data cleaning and the data standardization, the existing original data set of the building is optimized to obtain a monitoring data set. Through the data alignment, the data cleaning, the data standardization and the data set segmentation, the past original data set of the building is optimized to obtain a monitoring training set and a monitoring test set.

[0017] Furthermore, in the feature extraction model construction module, a model required for extracting building attribute features is constructed, specifically a deep learning model is constructed as the feature extraction model. The deep learning model is specifically a temporal-structural decoupling encoder model, including a feature extraction module, a temporal feature branch, a structural feature branch, and a decoupling fusion module;

[0018] The feature extraction model construction module specifically includes constructing a feature extraction module, constructing a temporal feature branch, constructing a structural feature branch, constructing a decoupling fusion module, and constructing and training a model;

[0019] The structural feature extraction module is used to convert the original data into a feature matrix that can be processed by the model. Specifically, it constructs the time domain, frequency domain and structural features of the building as the input of the model. The content includes:

[0020] Time domain feature extraction is used to capture the instantaneous energy changes of historical building structures. Specifically, the dynamic energy changes are calculated through the Teager energy operator to obtain the building's time domain characteristics;

[0021] Frequency domain feature extraction is used to identify changes in frequency response caused by aging of building materials. Specifically, the frequency spectrum features are extracted through Fourier transform to obtain the building frequency domain features.

[0022] Structural feature construction is used to quantify the degree of building damage. Specifically, it selects extreme values, gradients, and acceleration statistics of input data to construct features, thereby obtaining structural features that reflect structural stability.

[0023] The temporal feature construction branch is used to monitor the environmental impact of historical buildings. Specifically, the temporal features of buildings are obtained by combining a convolutional network and a temporal attention mechanism. The contents include:

[0024] Adaptive time-domain convolution is used to extract local temporal patterns. Specifically, it captures the correlation between adjacent time points by sliding the one-dimensional convolution kernel to obtain primary temporal features.

[0025] Design an activation function to enhance sensitivity to abnormal fluctuations. Specifically, a bimodal nonlinear transformation with adaptive threshold switching is used as the building attribute activation function.

[0026] Multi-scale temporal dilation is used to perceive multi-scale changes. Specifically, it captures long-term and short-term dependencies in parallel through convolutional layers with different dilation rates to obtain multi-scale temporal features.

[0027] Temporal attention is designed to focus on key time nodes. Specifically, it processes multi-scale temporal features through self-attention weights, dynamically assigns feature importance, and obtains architectural temporal features.

[0028] The structural feature branch is used to evaluate the building damage state. Specifically, it obtains the building structural state characteristics through feature transformation guided by physical constraints. The content includes:

[0029] Design physical constraints to embed the mechanical properties of the material, specifically by constraining the spatial distribution of features through constitutive equations to obtain physical structural characteristics that conform to physical laws;

[0030] Structural attention is designed to locate high-risk building components. Specifically, the physical structural features are reduced in dimension through a convolutional layer with a kernel size of 1×1. Then, the feature weights of key structural points are strengthened through attention gating to obtain the building structural status features.

[0031] The decoupling fusion module is used to achieve temporal-structural feature separation and joint expression, specifically by decoupling orthogonal constraints and feature pooling, enforcing feature independence, and constructing a joint expression to obtain building attribute features;

[0032] The constructing and training model specifically involves constructing a deep learning model through the construction feature extraction module, the construction temporal feature branch, the construction structural feature branch and the construction decoupling fusion module, training the model based on the monitoring training set, and verifying the model performance based on the monitoring test set to obtain a deep learning model as a feature extraction model.

[0033] Furthermore, in the state identification model construction module, a model required for identifying the building damage state level is constructed, specifically, an improved clustering model is constructed as a state identification model;

[0034] The state recognition model construction module specifically includes constructing a feature pyramid module, dynamic time kernel K-means clustering, physical constraint spectrum clustering, feature pyramid collaborative clustering, and model construction and training;

[0035] The feature pyramid construction module is used to generate multi-scale feature expressions to adapt to different types of historical building structures, specifically to obtain a three-level feature pyramid through linear projection transformation;

[0036] The dynamic time kernel K-means clustering is used to cluster temporal features. Specifically, adaptive wavelet kernel optimization combined with kernel K-means clustering is used to obtain category labels based on temporal features. The content includes:

[0037] Optimal scale selection is used to capture the time-varying characteristic period. Specifically, the optimal analysis scale for feature alignment is obtained through wavelet transform scale parameter optimization.

[0038] Distance matrix calculation is used to measure the similarity of temporal features. Specifically, the time domain distance matrix is ​​obtained by calculating the Euclidean distance in wavelet space.

[0039] Kernel spatial clustering is used to obtain category labels based on temporal features. Specifically, the kernel K-means clustering method is used to iteratively optimize the cluster centers to obtain category labels based on temporal features.

[0040] The physical constraint spectral clustering is used to cluster structural features. Specifically, it constrains sample associations by physical thresholds and performs spectral clustering to obtain category labels based on structural features. The content includes:

[0041] Similarity matrix construction is used to avoid pseudo-correlation structural features. Specifically, the relationship between samples is controlled by setting physical thresholds to obtain a physical constraint similarity matrix.

[0042] Laplace matrix construction for modeling building structure damage transmission path;

[0043] Calculating eigenvalues ​​and eigenvectors, specifically by performing eigenvector decomposition on the Laplace matrix to obtain the eigenvalues ​​and eigenvectors of the Laplace matrix;

[0044] K-means clustering is performed to obtain category labels based on structural features. Specifically, the eigenvalues ​​of the Laplace matrix are sorted in ascending order, the eigenvectors corresponding to the first n eigenvalues ​​are selected, and the eigenvectors corresponding to the first n eigenvalues ​​are clustered using the K-means clustering method to obtain category labels based on structural features.

[0045] The feature pyramid collaborative clustering is used to fuse multi-scale clustering results to eliminate single-view bias. Specifically, the final recognition result is obtained through probability fusion and consensus optimization, including:

[0046] Obtain multi-scale clustering probabilities to weight the credibility of each scale. Specifically, obtain the clustering probability of each scale through temperature scaling probability conversion;

[0047] The initial consensus distribution matrix is ​​used to balance the contributions of different scales. Specifically, the initial consensus distribution matrix is ​​obtained by dynamically weighting the multi-scale clustering probability through the silhouette coefficient;

[0048] Consensus clustering optimization, specifically minimizing disagreement through Laplace regularization to obtain the final recognition result label;

[0049] The model construction and training specifically involves constructing an improved clustering model through the feature pyramid construction module, the dynamic time kernel K-means clustering, the physical constraint spectrum clustering and the feature pyramid collaborative clustering, training the model based on the monitoring training set, and verifying the model performance based on the monitoring test set to obtain an improved clustering model as a state recognition model.

[0050] Furthermore, in the building digital monitoring module, the monitoring data set is first used as the input of the feature extraction model to obtain a building attribute feature set based on the monitoring data set, and the building attribute feature set based on the monitoring data set is used as the input of the state recognition model to obtain a building digital monitoring state reference result. The building digital monitoring state reference result is specifically the final recognition result label obtained by the state recognition model.

[0051] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0052] (1) In view of the technical problems that traditional digital monitoring and protection systems for historical buildings mostly use fixed signal processing algorithms or shallow machine learning models to extract features, are not adaptable enough to the nonlinear degradation process unique to buildings, have difficulty distinguishing environmental interference from real structural damage, lack a mechanism for embedding physical laws, and are easily affected by sensor noise and changes in working conditions, this solution creatively uses a deep learning model as a feature extraction model. Through adaptive time series modeling and feature decoupling driven by physical laws, it automatically learns the deep correlation characteristics between the time-varying behavior of buildings and structural damage, reflects the complex environment-structure coupling effect of buildings, and ensures the independence of features through orthogonal constraints, providing more discriminative input for subsequent state recognition.

[0053] (2) In view of the technical problems that traditional digital monitoring and protection systems for historical buildings cannot capture the time-varying characteristics of the evolution of building damage in terms of state identification, ignore the constitutive relationship of materials and the environment-structure interaction effect, and are prone to forcibly classify data with different physical mechanisms, and single-scale feature analysis is difficult to take into account both the global macroscopic response and local microscopic damage of the building, this scheme creatively adopts an improved clustering model as a state identification model. By introducing physical threshold constrained spectral clustering and adapting to the progressive characteristics of building damage through dynamic time kernel functions, combined with a multi-scale probability fusion mechanism, it achieves fine-grained perception and robust identification of building state changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 A schematic diagram of the modules of the digital monitoring and protection system for historical buildings based on multi-scale features provided by the present invention;

[0055] Figure 2 Flowchart of the raw data optimization module;

[0056] Figure 3 Schematic diagram of the process of building modules for feature extraction model;

[0057] Figure 4 Schematic diagram of the process of building modules for the state recognition model;

[0058] Figure 5 Schematic diagram of the process of constructing the temporal feature branch in the feature extraction model building module.

[0059] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0060] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0061] In the description of the present invention, it should be understood that terms such as "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they should not be understood as limiting the present invention.

[0062] Example 1, see Figure 1 The digital monitoring and protection system for historical buildings based on multi-scale features provided by the present invention includes a building data acquisition module, a raw data optimization module, a feature extraction model construction module, a state recognition model construction module and a building digital monitoring module;

[0063] The building data acquisition module obtains the original data set of historical buildings by performing data collection;

[0064] The raw data optimization module adopts data optimization methods such as data alignment, data cleaning, data standardization and data set segmentation to obtain monitoring data sets, monitoring training sets and monitoring test sets;

[0065] The feature extraction model construction module constructs a deep learning model as a feature extraction model to provide building attribute features for subsequent building digital monitoring;

[0066] The state recognition model construction module constructs an improved clustering model as a state recognition model for subsequent building digital monitoring;

[0067] The building digital monitoring module specifically performs building digital monitoring by adopting the feature extraction model in combination with the state recognition model to obtain a building digital monitoring state reference result.

[0068] In the second embodiment, in the building data acquisition module, the historical building original data set specifically includes a past building original data set and a current building original data set. Both the past building original data set and the current building original data set include building structure data, building environment data, and building material performance data. The past building original data set also includes building damage level annotation data.

[0069] The building structure data specifically includes vibration acceleration data, displacement deformation data and building inclination data. The vibration acceleration data is specifically the vibration response collected by the acceleration sensor installed at the building beam and column nodes. The displacement deformation data is specifically the displacement data collected by the displacement meter installed at the key joints of the building. The building inclination data is specifically the building inclination angle collected by the inclinometer deployed at the building's load-bearing columns and bases.

[0070] The building environment data specifically includes ambient temperature and humidity data covering indoor and outdoor areas of the building and at different heights, and wind speed and direction data around the building;

[0071] The building material performance data specifically includes building material acoustic emission data and building material resistance change data;

[0072] The building damage level annotation data is specifically a building damage status level label, including intact, slightly damaged, moderately damaged and severely damaged.

[0073] Example 3, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In the raw data optimization module, the data alignment is used to solve the problem of sensor time and space asynchrony in historical building monitoring. Specifically, by synchronizing the clocks of each device and constructing a timestamp alignment matrix, data with strict time axis synchronization is obtained.

[0074] The data cleaning is used to ensure the accuracy and integrity of the data, specifically by eliminating impulse noise through median filtering and removing baseline drift through empirical mode decomposition to obtain cleaned data;

[0075] The data normalization is used to eliminate dimensional distortion caused by environmental factors. Specifically, the displacement deformation data and power-normalized vibration energy are corrected by temperature compensation, and the Z-Score normalization method is used to obtain the normalized data. The temperature compensation corrects the displacement deformation data and power-normalized vibration energy using the following formula:

[0076] ;

[0077] Where, Represents the displacement deformation data after compensation, Represents the original displacement deformation data, Te represents the material expansion coefficient, Indicates the building environment temperature change value, Represents the normalized vibration acceleration data, Represents the original vibration acceleration data, T represents the time window length, Represents the instantaneous original vibration acceleration data at time t;

[0078] The data set segmentation is used to segment the data set, specifically to segment the original data set of the building to obtain a monitoring training set and a monitoring test set;

[0079] Through the data alignment, the data cleaning and the data standardization, the existing original data set of the building is optimized to obtain a monitoring data set. Through the data alignment, the data cleaning, the data standardization and the data set segmentation, the past original data set of the building is optimized to obtain a monitoring training set and a monitoring test set.

[0080] Example 4, see Figure 1 、 Figure 3 and Figure 5 This embodiment is based on the above embodiment. In the feature extraction model construction module, a model required for extracting building attribute features is constructed, specifically a deep learning model is constructed as the feature extraction model. The deep learning model is specifically a temporal-structural decoupling encoder model, including a feature extraction module, a temporal feature branch, a structural feature branch, and a decoupling fusion module.

[0081] The feature extraction model construction module specifically includes constructing a feature extraction module, constructing a temporal feature branch, constructing a structural feature branch, constructing a decoupling fusion module, and constructing and training a model;

[0082] The structural feature extraction module is used to convert the original data into a feature matrix that can be processed by the model. Specifically, it constructs the time domain, frequency domain and structural features of the building as the input of the model. The content includes:

[0083] Temporal feature extraction is used to capture the instantaneous energy changes of historical building structures. Specifically, the dynamic energy changes are calculated through the Teager energy operator to obtain the building's temporal characteristics. The formula used is as follows:

[0084] ;

[0085] Where, Represents the time domain characteristics of the building, represents the Teager energy operator function, Represents the input data of the deep learning model;

[0086] Frequency domain feature extraction is used to identify changes in frequency response caused by aging of building materials. Specifically, the frequency spectrum features are extracted through Fourier transform to obtain the building frequency domain features.

[0087] Structural feature construction is used to quantify the degree of building damage. Specifically, it selects extreme values, gradients, and acceleration statistics of input data to construct features, thereby obtaining structural features that reflect structural stability.

[0088] The temporal feature construction branch is used to monitor the environmental impact of historical buildings. Specifically, the temporal features of buildings are obtained by combining a convolutional network and a temporal attention mechanism. The contents include:

[0089] Adaptive time-domain convolution is used to extract local temporal patterns. Specifically, it captures the correlation between adjacent time points by sliding the one-dimensional convolution kernel to obtain primary temporal features. The formula used is as follows:

[0090] ;

[0091] Where, Indicates primary tense features, represents the one-dimensional convolution function, Represents the frequency domain characteristics of the building;

[0092] An activation function is designed to enhance sensitivity to abnormal fluctuations. Specifically, a bimodal nonlinear transformation with adaptive threshold switching is used as the building attribute activation function. The formula used is as follows:

[0093] ;

[0094] Where, represents the building attribute activation function, x represents the input independent variable, represents the hyperbolic tangent function, represents the sigmoid activation function, represents the SiLU activation function, 、 and are different learnable parameters;

[0095] Multi-scale temporal dilation is used to perceive multi-scale changes. Specifically, it captures long-term and short-term dependencies in parallel through convolutional layers with different dilation rates to obtain multi-scale temporal features. The formula used is as follows:

[0096] ;

[0097] Where, represents the temporal features at the first scale, represents the temporal features at the second scale, represents the temporal features at the third scale, represents the dilated convolution function with a dilation rate of 1, represents the dilated convolution function with a dilation rate of 3, represents the dilated convolution function with a dilation rate of 5, Represents multi-scale temporal features, represents a learnable multi-scale weight matrix;

[0098] Temporal attention is designed to focus on key time nodes. Specifically, it processes multi-scale temporal features through self-attention weights, dynamically assigns feature importance, and obtains architectural temporal features.

[0099] The structural feature branch is used to evaluate the building damage state. Specifically, it obtains the building structural state characteristics through feature transformation guided by physical constraints. The content includes:

[0100] Design physical constraints to embed the mechanical properties of the material. Specifically, constrain the spatial distribution of features through constitutive equations to obtain physical structural characteristics that conform to physical laws. The formula used is as follows:

[0101] ;

[0102] Where, represents the physical constraint function, Indicates the yield strength of building materials, represents the current strain of building materials, represents the strain threshold of building materials, represents the ultimate strain of building materials, Represents the structural characteristics of the building. Represents the constitutive matrix of building materials, which is a diagonal matrix composed of material parameters. Pe represents the physical structure characteristics. represents element-wise multiplication;

[0103] Structural attention is designed to locate high-risk building components. Specifically, the physical structure features are reduced in dimension through a convolution layer with a convolution kernel size of 1×1. Then, the feature weights of key structural points are strengthened through attention gating to obtain the building structure status features. The formula used is as follows:

[0104] ;

[0105] Where, represents the activation structure feature, represents the ReLU activation function, represents the convolution function with a convolution kernel size of 1×1, Indicates the structural characteristics of the building. represents the learnable structure weight, represents the tensor product;

[0106] The decoupling fusion module is used to achieve temporal-structural feature separation and joint expression. Specifically, it decouples orthogonal constraints and feature pooling, enforces feature independence, and constructs joint expression to obtain building attribute features. The formula used is as follows:

[0107] ;

[0108] Where, Represents the temporal attribute characteristics of buildings, Indicates the property characteristics of the building structure state, represents the orthogonal loss, Indicates the calculation of the Frobenius norm, represents the global average pooling function, Indicates the architectural tense characteristics, represents the maximum pooling function, Tr represents the transposition operation, and Ar represents the building attribute feature;

[0109] The constructing and training model specifically involves constructing a deep learning model through the construction feature extraction module, the construction temporal feature branch, the construction structural feature branch and the construction decoupling fusion module, training the model based on the monitoring training set, and verifying the model performance based on the monitoring test set to obtain a deep learning model as a feature extraction model.

[0110] By performing the above operations, this solution creatively uses a deep learning model as a feature extraction model to address the technical problems of traditional digital monitoring and protection systems for historical buildings, such as the use of fixed signal processing algorithms or shallow machine learning models to extract features, which are insufficiently adaptable to the nonlinear degradation process unique to buildings, making it difficult to distinguish between environmental interference and real structural damage, and lacking a mechanism for embedding physical laws, making them susceptible to sensor noise and changes in working conditions. Through adaptive time series modeling and feature decoupling driven by physical laws, this solution automatically learns the deep correlation characteristics between the time-varying behavior of buildings and structural damage, reflecting the complex environment-structure coupling effects of buildings, and ensuring the independence of features through orthogonal constraints, providing more discriminative input for subsequent state recognition.

[0111] Example 5, see Figure 1 and Figure 4 , based on the above embodiment, this embodiment is used to construct a model required for identifying the building damage status level in the state identification model construction module, specifically to construct an improved clustering model as a state identification model;

[0112] The state recognition model construction module specifically includes constructing a feature pyramid module, dynamic time kernel K-means clustering, physical constraint spectrum clustering, feature pyramid collaborative clustering, and model construction and training;

[0113] The feature pyramid construction module is used to generate multi-scale feature expressions to adapt to different types of historical building structures. Specifically, a three-level feature pyramid is obtained through linear projection transformation. The formula used is as follows:

[0114] ;

[0115] Where, Indicates the architectural attribute characteristics at the first scale, Represents the architectural attribute characteristics at the second scale, Indicates the architectural attribute characteristics at the third scale, represents the learnable weight matrix of the first scale, represents the learnable weight matrix of the second scale, Represents the learnable weight matrix of the third scale;

[0116] The dynamic time kernel K-means clustering is used to cluster temporal features. Specifically, adaptive wavelet kernel optimization combined with kernel K-means clustering is used to obtain category labels based on temporal features. The content includes:

[0117] The optimal scale selection is used to capture the time-varying characteristic period. Specifically, the optimal analysis scale for feature alignment is obtained by optimizing the wavelet transform scale parameter. The formula used is as follows:

[0118] ;

[0119] Where, represents the optimal analysis scale, Represents the scale parameter of wavelet transform, num represents the total number of samples, represents the continuous wavelet transform function, represents the architectural temporal attribute characteristics of the i-th sample, represents the architectural temporal attribute characteristics of the jth sample, Indicates calculation of L2 norm;

[0120] Distance matrix calculation is used to measure the similarity of temporal features. Specifically, the time domain distance matrix is ​​obtained by calculating the Euclidean distance in wavelet space. The formula used is as follows:

[0121] ;

[0122] Where, The elements of the time domain distance matrix are the Euclidean distances between the i-th sample and the j-th sample in wavelet space;

[0123] Kernel spatial clustering is used to obtain category labels based on temporal features. Specifically, the kernel K-means clustering method is used to iteratively optimize the cluster centers to obtain category labels based on temporal features.

[0124] The physical constraint spectral clustering is used to cluster structural features. Specifically, it constrains sample associations by physical thresholds and performs spectral clustering to obtain category labels based on structural features. The content includes:

[0125] The similarity matrix is ​​constructed to avoid pseudo-correlation structural features. Specifically, the physical constraint similarity matrix is ​​obtained by setting a physical threshold to control the relationship between samples. The formula used is as follows:

[0126] ;

[0127] Where, Represents the elements of the physical constraint similarity matrix, pa represents the proportional parameter that controls the similarity decay rate, represents the building structure state attribute characteristics of the i-th sample, represents the building structure state attribute characteristics of the jth sample, represents the kth dimension of the building structure state attribute characteristics of the i-th sample, represents the kth dimension of the building structure state attribute characteristics of the jth sample, A preset physical threshold value representing the kth dimension of the building structure state attribute characteristics;

[0128] The Laplace matrix is ​​constructed to model the damage transmission path of the building structure. The formula used is as follows:

[0129] ;

[0130] Where Lm represents the Laplace matrix, D represents the degree matrix of the physical constraint similarity matrix, represents the physical constraint similarity matrix;

[0131] Calculating eigenvalues ​​and eigenvectors, specifically by performing eigenvector decomposition on the Laplace matrix to obtain the eigenvalues ​​and eigenvectors of the Laplace matrix;

[0132] K-means clustering is performed to obtain category labels based on structural features. Specifically, the eigenvalues ​​of the Laplace matrix are sorted in ascending order, the eigenvectors corresponding to the first n eigenvalues ​​are selected, and the eigenvectors corresponding to the first n eigenvalues ​​are clustered using the K-means clustering method to obtain category labels based on structural features.

[0133] The feature pyramid collaborative clustering is used to fuse multi-scale clustering results to eliminate single-view bias. Specifically, the final recognition result is obtained through probability fusion and consensus optimization, including:

[0134] Obtain multi-scale clustering probabilities to weight the credibility of each scale. Specifically, the clustering probability of each scale is obtained by temperature scaling probability conversion. The formula used is as follows:

[0135] ;

[0136] Where, represents the clustering probability of the mth scale, represents the Mahalanobis distance calculation function, represents the building attribute characteristics at the mth scale, represents the pth fusion cluster center, tem represents the temperature parameter, represents the qth fusion cluster center, represents the dynamic time kernel K-means clustering center transformation weight, represents the physical constraint spectral clustering center transformation weight, represents the pth dynamic time kernel K-means cluster center, represents the pth physical constraint spectrum cluster center;

[0137] The initial consensus distribution matrix is ​​used to balance the contributions of different scales. Specifically, the initial consensus distribution matrix is ​​obtained by dynamically weighting the multi-scale clustering probability through the silhouette coefficient. The formula used is as follows:

[0138] ;

[0139] Where, represents the initial consensus distribution matrix, Represents the weighted weight of the silhouette coefficient of the mth scale, and the consensus distribution matrix has a size of ,in Indicates the total number of tags;

[0140] Consensus clustering optimization, specifically minimizing divergence through Laplace regularization, obtains the final recognition result label. The formula used is as follows:

[0141] ;

[0142] In the formula, Fun represents the objective function to be optimized, U represents the consensus distribution matrix, and it is updated iteratively through the gradient descent method. represents the regularization parameter, Represents the trace calculation function of the matrix, La represents the final recognition result label, represents the elements of the consensus allocation matrix;

[0143] The model construction and training specifically involves constructing an improved clustering model through the feature pyramid construction module, the dynamic time kernel K-means clustering, the physical constraint spectrum clustering and the feature pyramid collaborative clustering, training the model based on the monitoring training set, and verifying the model performance based on the monitoring test set to obtain an improved clustering model as a state recognition model.

[0144] By performing the above operations, this solution creatively adopts an improved clustering model as a state recognition model to address the technical problems of traditional digital monitoring and protection systems for historical buildings, such as their inability to capture the time-varying characteristics of building damage evolution in terms of state recognition, ignoring material constitutive relationships and environment-structure interaction effects, and easily forcibly classifying data with different physical mechanisms. In addition, single-scale feature analysis is difficult to take into account both the global macro-response and local micro-damage of the building. By introducing physical threshold constrained spectral clustering and adapting to the progressive characteristics of building damage through dynamic time kernel functions, combined with a multi-scale probability fusion mechanism, it achieves fine-grained perception and robust recognition of building state changes.

[0145] Example 6, see Figure 1 This embodiment is based on the above embodiment. In the building digital monitoring module, specifically, the monitoring data set is first used as the input of the feature extraction model to obtain a building attribute feature set based on the monitoring data set, and the building attribute feature set based on the monitoring data set is used as the input of the state recognition model to obtain a building digital monitoring state reference result. The building digital monitoring state reference result is specifically the final recognition result label obtained by the state recognition model.

[0146] 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 terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0147] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions, and alterations can be made to the embodiments without departing from the principles and spirit of the invention.

[0148] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. A digital monitoring and protection system for historical buildings based on multi-scale features, characterized by: The system includes a building data acquisition module, a raw data optimization module, a feature extraction model construction module, a state recognition model construction module and a building digital monitoring module; The building data acquisition module obtains the original data set of historical buildings by performing data collection, and the original data set of historical buildings specifically includes the original data set of past buildings and the original data set of current buildings; The raw data optimization module adopts data optimization methods such as data alignment, data cleaning, data standardization and data set segmentation to obtain monitoring data sets, monitoring training sets and monitoring test sets; The feature extraction model construction module is used to construct a model required for extracting building attribute features, specifically to construct a deep learning model as a feature extraction model. The deep learning model is specifically a temporal-structural decoupling encoder model, including a feature extraction module, a temporal feature branch, a structural feature branch, and a decoupling fusion module; The state recognition model construction module is used to construct the model required to identify the building damage state level, specifically to construct an improved clustering model as a state recognition model. The specific content includes constructing a feature pyramid module, dynamic time kernel K-means clustering, physical constraint spectrum clustering, feature pyramid collaborative clustering, and model construction and training; The building digital monitoring module specifically first uses the monitoring data set as the input of the feature extraction model to obtain a building attribute feature set based on the monitoring data set, and then uses the building attribute feature set based on the monitoring data set as the input of the state recognition model to obtain a building digital monitoring state reference result. The building digital monitoring state reference result is specifically the final recognition result label obtained by the state recognition model.

2. The multi-scale feature-based digital monitoring and protection system for historical buildings according to claim 1 is characterized by: The feature extraction model construction module specifically includes constructing a feature extraction module, constructing a temporal feature branch, constructing a structural feature branch, constructing a decoupling fusion module, and constructing and training a model.

3. The multi-scale feature-based digital monitoring and protection system for historical buildings according to claim 2 is characterized by: The structural feature extraction module is used to convert the original data into a feature matrix that can be processed by the model. Specifically, it constructs the time domain, frequency domain and structural features of the building as the input of the model. The content includes: Time domain feature extraction is used to capture the instantaneous energy changes of historical building structures. Specifically, the dynamic energy changes are calculated through the Teager energy operator to obtain the building's time domain characteristics; Frequency domain feature extraction is used to identify changes in frequency response caused by aging of building materials. Specifically, the frequency spectrum features are extracted through Fourier transform to obtain the building frequency domain features. Structural feature construction is used to quantify the degree of building damage. Specifically, it selects extreme values, gradients, and acceleration statistics of input data to construct features, thereby obtaining structural features that reflect structural stability. The temporal feature construction branch is used to monitor the environmental impact of historical buildings. Specifically, the temporal features of buildings are obtained by combining a convolutional network and a temporal attention mechanism. The contents include: Adaptive time-domain convolution is used to extract local temporal patterns. Specifically, it captures the correlation between adjacent time points by sliding the one-dimensional convolution kernel to obtain primary temporal features. Design an activation function to enhance sensitivity to abnormal fluctuations. Specifically, a bimodal nonlinear transformation with adaptive threshold switching is used as the building attribute activation function. Multi-scale temporal dilation is used to perceive multi-scale changes. Specifically, it captures long-term and short-term dependencies in parallel through convolutional layers with different dilation rates to obtain multi-scale temporal features. Temporal attention is designed to focus on key time nodes. Specifically, it processes multi-scale temporal features through self-attention weights, dynamically assigns feature importance, and obtains architectural temporal features. The structural feature branch is used to evaluate the building damage state. Specifically, it obtains the building structural state characteristics through feature transformation guided by physical constraints. The content includes: Design physical constraints to embed the mechanical properties of the material, specifically by constraining the spatial distribution of features through constitutive equations to obtain physical structural characteristics that conform to physical laws; Structural attention is designed to locate high-risk building components. Specifically, the physical structural features are reduced in dimension through a convolutional layer with a kernel size of 1×1. Then, the feature weights of key structural points are strengthened through attention gating to obtain the building structural status features. The decoupling fusion module is used to achieve temporal-structural feature separation and joint expression, specifically by decoupling orthogonal constraints and feature pooling, enforcing feature independence, and constructing a joint expression to obtain building attribute features; The constructing and training model specifically involves constructing a deep learning model through the construction feature extraction module, the construction temporal feature branch, the construction structural feature branch and the construction decoupling fusion module, training the model based on the monitoring training set, and verifying the model performance based on the monitoring test set to obtain a deep learning model as a feature extraction model.

4. The multi-scale feature-based digital monitoring and protection system for historical buildings according to claim 1 is characterized by: In the state recognition model construction module, the feature pyramid construction module is used to generate multi-scale feature expressions to adapt to different types of historical building structures, specifically to obtain a three-level feature pyramid through linear projection transformation; The dynamic time kernel K-means clustering is used to cluster temporal features. Specifically, adaptive wavelet kernel optimization combined with kernel K-means clustering is used to obtain category labels based on temporal features. The content includes: Optimal scale selection is used to capture the time-varying characteristic period. Specifically, the optimal analysis scale for feature alignment is obtained through wavelet transform scale parameter optimization. Distance matrix calculation is used to measure the similarity of temporal features. Specifically, the time domain distance matrix is ​​obtained by calculating the Euclidean distance in wavelet space. Kernel spatial clustering is used to obtain category labels based on temporal features. Specifically, the kernel K-means clustering method is used to iteratively optimize the cluster centers to obtain category labels based on temporal features. The physical constraint spectral clustering is used to cluster structural features. Specifically, it constrains sample associations by physical thresholds and performs spectral clustering to obtain category labels based on structural features. The content includes: Similarity matrix construction is used to avoid pseudo-correlation structural features. Specifically, the relationship between samples is controlled by setting physical thresholds to obtain a physical constraint similarity matrix. Laplace matrix construction for modeling building structure damage transmission path; Calculating eigenvalues ​​and eigenvectors, specifically by performing eigenvector decomposition on the Laplace matrix to obtain the eigenvalues ​​and eigenvectors of the Laplace matrix; K-means clustering is performed to obtain category labels based on structural features. Specifically, the eigenvalues ​​of the Laplace matrix are sorted in ascending order, the eigenvectors corresponding to the first n eigenvalues ​​are selected, and the eigenvectors corresponding to the first n eigenvalues ​​are clustered using the K-means clustering method to obtain category labels based on structural features. The feature pyramid collaborative clustering is used to fuse multi-scale clustering results to eliminate single-view bias. Specifically, the final recognition result is obtained through probability fusion and consensus optimization, including: Obtain multi-scale clustering probabilities to weight the credibility of each scale. Specifically, obtain the clustering probability of each scale through temperature scaling probability conversion; The initial consensus distribution matrix is ​​used to balance the contributions of different scales. Specifically, the initial consensus distribution matrix is ​​obtained by dynamically weighting the multi-scale clustering probability through the silhouette coefficient; Consensus clustering optimization, specifically minimizing disagreement through Laplace regularization to obtain the final recognition result label; The model construction and training specifically involves constructing an improved clustering model through the feature pyramid construction module, the dynamic time kernel K-means clustering, the physical constraint spectrum clustering and the feature pyramid collaborative clustering, training the model based on the monitoring training set, and verifying the model performance based on the monitoring test set to obtain an improved clustering model as a state recognition model.

5. The digital monitoring and protection system for historical buildings based on multi-scale features according to claim 1 is characterized by: In the building data acquisition module, the past original data set of the building and the current original data set of the building both include building structure data, building environment data and building material performance data. The past original data set of the building also includes building damage level annotation data.

6. The multi-scale feature-based digital monitoring and protection system for historical buildings according to claim 1 is characterized by: In the raw data optimization module, data alignment is used to solve the problem of sensor spatiotemporal asynchrony in historical building monitoring. Specifically, by synchronizing the clocks of each device and constructing a timestamp alignment matrix, data with strict time axis synchronization is obtained. The data cleaning is used to ensure the accuracy and integrity of the data, specifically by eliminating impulse noise through median filtering and removing baseline drift through empirical mode decomposition to obtain cleaned data; The data normalization is used to eliminate dimensional distortion caused by environmental factors, specifically by correcting displacement deformation data and power-normalized vibration energy through temperature compensation, and using the Z-Score normalization method to obtain standardized data; The data set segmentation is used to segment the data set, specifically to segment the original data set of the building to obtain a monitoring training set and a monitoring test set; Through the data alignment, the data cleaning and the data standardization, the existing original data set of the building is optimized to obtain a monitoring data set. Through the data alignment, the data cleaning, the data standardization and the data set segmentation, the past original data set of the building is optimized to obtain a monitoring training set and a monitoring test set.

Images