A method and system for visualizing building curtain wall monitoring data processing

Abstract

The present invention discloses a method and system for visualizing building curtain wall monitoring data processing. The method uses an extreme learning machine algorithm to monitor the health of the curtain wall. By assigning dynamic weights to sensors at different locations, the significance of signals in vulnerable areas is enhanced, and the signal's responsiveness to local damage is improved. By simulating the pheromone attenuation effect, the weight and bias of each node are dynamically adjusted, thereby enhancing the model's adaptability to complex, high-dimensional data, improving noise robustness, and reducing the impact of redundant features in the data. Based on the resonance effect, the hidden layer's adaptability to nonlinear data is enhanced. By dynamically adjusting the regularization coefficient, the stability of the classification is improved. Based on the structural damage propagation consistency constraint term, the sensitivity to early local damage is enhanced, and the system's fault detection capability is improved.

Description

Technical Field

[0001] The present invention relates to the field of smart buildings, and in particular to a method and system for visualizing building curtain wall monitoring data processing. Background Art

[0002] Curtain walls, a vital component of modern architecture, are widely used on the exterior facades of high-rise buildings, commercial buildings, and more. They not only offer aesthetically pleasing exteriors but also protect the building's interior. However, as curtain walls age, they gradually develop varying degrees of damage due to factors such as the external environment, temperature fluctuations, wind, and earthquakes. If undetected, these damages can pose safety hazards, impact the building's service life, and even lead to serious accidents.

[0003] Existing technologies ignore the differences in sensitivity of sensors at different locations in the curtain wall and use simple weight initialization and processing methods, resulting in overfitting of noise and a decrease in feature extraction capabilities. Traditional activation functions cannot fully exploit the deep nonlinear structure, and fixed weight update strategies lead to poor model robustness and reduced classification accuracy. They fail to fully utilize the physical topological characteristics and damage propagation laws of building structures, resulting in weak performance in the detection and prediction of early local damage. Summary of the Invention

[0004] (1) Technical problems solved

[0005] In order to solve the above technical problems, the present invention provides a method and system for visualizing building curtain wall monitoring data processing.

[0006] (2) Technical solution

[0007] In order to solve the above-mentioned technical problems and achieve the purpose of the invention, the present invention is implemented through the following technical solutions:

[0008] A method for visualizing building curtain wall monitoring data processing includes the following steps:

[0009] S1: Data collection: real-time collection of building curtain wall health monitoring data. Specifically, the physical status information of the curtain wall is obtained through multiple sensors.

[0010] S2: Data preprocessing, including data cleaning, denoising, standardization, normalization, and missing data filling of the collected raw data;

[0011] S3: Machine learning modeling for health status classification prediction. The model uses the extreme learning machine algorithm to obtain health status classification prediction results. Specifically, an adaptive feature weighting mechanism based on the spatial location of sensors, combined with the mechanical properties of the curtain wall structure, assigns dynamic weights to sensors in different locations.

[0012] S4: Damage detection and prediction: real-time application of the trained model to detect and predict the health status of building curtain walls;

[0013] S5: Visual display, presenting monitoring data and model analysis results to users in a graphical manner;

[0014] S6: System management and monitoring, manage, control and monitor the entire monitoring system.

[0015] Furthermore, the position weights of different position sensors in step S3 are:

[0016]

[0017] Where, For the The spatial weight of each sensor; For the The Euclidean distance from each sensor to the nearest edge of the curtain wall; For the The Euclidean distance from each sensor to the nearest edge of the curtain wall; is the maximum value of all sensor edge distances; is the spatial sensitivity coefficient; is the total number of sensors.

[0018] Furthermore, the extreme learning machine algorithm initializes the weights and bias parameters of the extreme learning machine hidden layer nodes based on the biopheromone decay mechanism, dynamically adjusts the weights and biases of each node according to the characteristic distribution of the input data, and accepts random noise for correction.

[0019] Furthermore, the extreme learning machine algorithm uses nonlinear resonance constraints to the hidden layer activation function, and the activation function is expressed as:

[0020]

[0021] Where, is the output of the hidden layer activation function of the extreme learning machine, is the hidden layer activation function of the extreme learning machine; is the weight of the hidden layer node of the extreme learning machine; is the bias of the hidden layer node of the extreme learning machine; It is the input of the activation function, representing the feature vector corresponding to the input data; is the resonance effect intensity; is the resonance effect frequency.

[0022] Furthermore, the resonance effect intensity and the resonance effect frequency are expressed as follows:

[0023]

[0024]

[0025] Where, For the The intensity of the resonance effect in the iterative training controls the oscillation effect in the nonlinear mapping; For the The resonance effect frequency in the iterative training is used to control the resonance effect frequency; Tune hyperparameters for first resonance effect strength; Tuning hyperparameters for the strength of the second resonance effect; Adjust hyperparameters for the first resonance effect frequency; Adjust hyperparameters for the second resonance effect frequency; For the The actual labels of the samples; The first samples.

[0026] Furthermore, the extreme learning machine algorithm is based on a dynamic regularization mechanism of node activity, which dynamically adjusts the regularization coefficient of the hidden layer nodes by calculating the activation variance and covariance matrix of each hidden layer node.

[0027] Furthermore, the extreme learning machine algorithm performs an update calculation of the biopheromone attenuation factor based on the relative change of the loss function gradient, which is expressed as:

[0028]

[0029] Where, For the The biopheromone decay factor of the iteration; For the The biopheromone decay factor of the iteration; is the decay rate control parameter; For the extreme learning machine The gradient of the loss function at iterations; For the extreme learning machine The gradient of the loss function at iterations; is the loss function of the extreme learning machine.

[0030] Furthermore, the loss function of the extreme learning machine is calculated as follows:

[0031]

[0032] Where, is the loss function of the extreme learning machine; is the number of training samples; Represents the output value of the extreme learning machine hidden layer activation function; is the weight matrix of the hidden layer nodes of the extreme learning machine; The first samples; is the bias of the hidden layer nodes of the extreme learning machine; No. The actual labels of the samples; Characterizes the square of the prediction error of a single sample; is the dynamic regularization term; is the consistency constraint strength; is the consistency constraint term for structural damage propagation.

[0033] The present invention also provides a building curtain wall monitoring data processing and visualization system, which includes: a data acquisition unit, a data preprocessing unit, a machine learning modeling unit, a damage detection and prediction unit, a visualization display unit, and a system management and monitoring unit, wherein:

[0034] The data acquisition unit is used to collect health monitoring data of building curtain walls in real time;

[0035] The data preprocessing unit is used to perform denoising, standardization and normalization on the collected raw data;

[0036] The machine learning modeling unit is used to learn the health status pattern of the building curtain wall through training data based on the extreme learning machine algorithm, build a classifier model, and obtain the health status classification prediction results;

[0037] The damage detection and prediction unit is used to evaluate the real-time monitoring data based on the machine learning model and output the predicted results of the health status;

[0038] The visualization display unit is used to present monitoring data and model analysis results to users in a graphical manner;

[0039] The system management and monitoring unit is used to manage, control and monitor the entire monitoring system.

[0040] In addition, to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium, on which program instructions of a method for visualizing building curtain wall monitoring data processing are stored. The program instructions of a method for visualizing building curtain wall monitoring data processing can be executed by one or more processors to implement the steps of the method for visualizing building curtain wall monitoring data processing as described above.

[0041] (3) Beneficial effects

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] (1) The present invention avoids the equivalent processing of sensor data in conventional methods and significantly improves the sensitivity of early warning of damage.

[0044] (2) This invention simulates the pheromone attenuation effect in nature, enhances the model's adaptability to complex, high-dimensional data, improves noise robustness, and reduces the impact of redundant features in the data, enabling better processing of noisy and heterogeneous data. It also enhances the hidden layer's adaptability to nonlinear data, improves its sensitivity to complex patterns and subtle damage signals, and improves the performance of traditional activation functions in complex, dynamic data environments. It also avoids the impact of noise and redundant data, improving the model's robustness and classification performance.

[0045] (3) The present invention calculates the activation variance and covariance matrix of the hidden layer nodes and dynamically adjusts the regularization coefficient, effectively reducing the activation intensity of highly noise-sensitive nodes, enhancing the model's reliance on stable and effective features, and improving the stability of classification.

[0046] (4) The present invention combines the physical topological structure and damage propagation law of the building curtain wall and proposes a structural damage propagation consistency constraint term, which enhances the sensitivity to early local damage and improves the fault detection capability of the system.

[0047] (5) The present invention dynamically adjusts the biopheromone attenuation factor based on the relative change of the loss function gradient, so that the model can adapt to the time-varying characteristics of the data, effectively controls the impact of noise on model training, and enhances the stability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0049] Figure 1 Schematic diagram of a method for visualizing building curtain wall monitoring data processing according to an embodiment of the present application;

[0050] Figure 2: is a diagram showing the sensor spatial weight distribution and edge damage signal enhancement effect according to an embodiment of the present application. Figure 2 Part (a) in the figure represents the sensor spatial weight distribution. Figure 2 Part (b) shows the edge damage signal enhancement effect;

[0051] Figure 3 is a comparison chart of classification accuracy between the initialization method and the random initialization method under different noise environments according to an embodiment of the present application;

[0052] Figure 4 It is the dynamic response surface of the resonant activation function and the response surface of the Sigmoid activation function according to the embodiment of the present application, Figure 4 Part (a) shows the dynamic response surface of the resonant activation function. Figure 4 Part (b) represents the sigmoid activation function response surface;

[0053] Figure 5 is a covariance heat map according to an embodiment of the present application, Figure 5 Part (a) shows the original distribution heat map of node covariance. Figure 5 Part (b) shows the covariance distribution heat map after dynamic regularization. DETAILED DESCRIPTION

[0054] The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0055] The following describes the embodiments of the present disclosure through specific examples, and those skilled in the art can easily understand other advantages and effects of the present disclosure from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. The present disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present disclosure.

[0056] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present disclosure. The illustrations only show components related to the present disclosure and are not drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component can be changed at will, and the component layout type may also be more complicated.

[0057] See also Figure 1,A building curtain wall monitoring data processing and visualization method includes the following steps:

[0058] S1: Data acquisition: real-time collection of health monitoring data of building curtain walls. Physical status information of the curtain walls is obtained through a variety of sensors (such as temperature sensors, vibration sensors, stress and strain sensors, etc.). The data acquisition process requires high precision and high real-time performance.

[0059] This includes real-time data collection, preprocessing and transmission. Specific functions include: sensor signal collection, data time series synchronization, ensuring that data from various sensors are integrated in time series and ready for subsequent analysis.

[0060] It also includes detecting the working status of the sensor to ensure the reliability and stability of the data.

[0061] S2: Data preprocessing, which involves cleaning, denoising, standardizing, normalizing, and filling in missing data on the collected raw data to improve the accuracy and efficiency of subsequent analysis and ensure that the data has high-quality features before being input into the machine learning model.

[0062] S3: Machine learning model building for health status classification prediction. The model uses the extreme learning machine algorithm to learn the health status pattern of the building curtain wall through training data, build a classifier model, and obtain health status classification prediction results, including the following:

[0063] a. Adaptive feature weighting of sensor spatial location;

[0064] In building curtain wall health monitoring, sensors are usually deployed at key structural locations (such as corners, joints, and the middle). Sensors at different locations have significant differences in their sensitivity to the structural health status. For example, sensors at edges and joints are more likely to capture local deformation and stress concentration, while sensors in the middle are more sensitive to overall vibration modes.

[0065] Conventional methods treat sensor data from different locations equally, ignoring the relationship between their spatial locations and structural fragility, resulting in key signals being submerged by noise.

[0066] The present invention adopts an adaptive feature weighting mechanism based on the spatial position of sensors, combined with the mechanical properties of curtain wall structures, to assign dynamic weights to sensors at different locations to enhance the significance of signals in vulnerable areas. Specifically, the sensor position weights are defined as:

[0067]

[0068] Where, For the The spatial weight of each sensor reflects the contribution of the spatial positions of different sensors to health status monitoring; For the The Euclidean distance from each sensor to the nearest edge of the curtain wall represents the distance between the sensor and the vulnerable area of the structure (such as edges, joints, etc.). The closer the distance, the higher the structural vulnerability, and the greater the weight of the sensor should be; For the The Euclidean distance from each sensor to the nearest edge of the curtain wall (the closer the edge, the higher the structural vulnerability); is the maximum value of the edge distance of all sensors, used for normalization; The spatial sensitivity coefficient controls the rate at which the weight of sensors near the edge decays. A larger spatial sensitivity coefficient will give sensors closer to the edge a higher weight. is the total number of sensors; is an exponential function whose base is a natural constant.

[0069] Furthermore, the input data is weighted and expressed as ,in, The first samples, representing the samples weighted by the spatial weight of the sensor; is the original input feature vector; Represents element-wise multiplication; is the spatial weight vector, For the The spatial weight of each sensor.

[0070] In one embodiment, in order to verify the enhancement effect of the spatial weighting mechanism on the damage signal of the curtain wall edge area, the spatial weight distribution of the sensor and its signal enhancement effect were analyzed. The sensor deployment position was associated with the distance to the curtain wall edge, and the equivalent processing mode of the sensor data was compared with the traditional method. Figure 2 The experimental results show that by constructing a spatial weight function related to structural fragility, the present invention enables sensors near the edge of the curtain wall to obtain higher feature weights, effectively amplifying the intensity of the damage signal at the edge joint. Conventional methods, however, have difficulty extracting key features from complex background noise due to ignoring spatial position information. This indicates that the adaptive weighting mechanism based on structural mechanics significantly improves the sensitivity of early damage warning.

[0071] b. Initialize the weights and bias parameters of the extreme learning machine hidden layer nodes based on the biopheromone decay mechanism

[0072] In the classification of building curtain wall health status, data comes from various sensors installed on the curtain wall, such as temperature sensors, vibration sensors, stress and strain sensors, displacement sensors, and humidity sensors. These sensors collect physical status data of the curtain wall in real time. Due to the diversity of sensors and the complexity brought about by different types of physical quantities, the data is usually high-dimensional and contains noise. The high dimensionality and noise of the data make model training and learning more difficult.

[0073] Conventional extreme learning machines usually randomly set the weights and biases of hidden layer nodes during parameter initialization. This method lacks sensitivity to the characteristics of the data itself and easily leads to overfitting of noise during training, thereby reducing the model's generalization ability and classification performance. When faced with high-dimensional, complex and noisy sensor data, it cannot adapt to the diversity and inherent structure of the data.

[0074] The present invention initializes the weights and biases of hidden layer nodes by adopting a biopheromone decay mechanism. The biopheromone decay mechanism simulates the decay effect of pheromones in nature, dynamically adjusts the weights and biases of each node according to the characteristic distribution of the input data, and accepts random noise for correction, so that the extreme learning machine model can adapt to data changes more flexibly. During the training process, random noise terms are considered for perturbation to improve the robustness of the extreme learning machine model in complex data. Therefore, when there is noise in the input data, the model can still adapt to the noise situation, which is expressed as

[0075]

[0076]

[0077] Where, For the extreme learning machine The weights of the hidden layer nodes; For the extreme learning machine The bias of hidden layer nodes; The first samples; is the biopheromone attenuation factor, which controls the attenuation degree; The first The value of a sample after the Sigmoid function represents the nonlinear transformation of the data. is the Sigmoid activation function; is the random noise term of the weight, which obeys the normal distribution with mean 0 and variance as the unit matrix; is the biased random noise term, which obeys the normal distribution with mean 0 and variance as the unit matrix; is the number of training samples.

[0078] The biopheromone attenuation factor is dynamically adjusted by calculating the similarity between input samples, so that the initialization process can be adaptively adjusted according to the input data. For high-dimensional heterogeneous data from multiple types of sensors (temperature, vibration, etc.), the biopheromone attenuation factor is dynamically adjusted through similarity-driven dynamic adjustment, reducing noise weight initialization and improving the model's adaptability to the multi-physics field coupling characteristics of the curtain wall. The calculation method is expressed as:

[0079]

[0080] Where, For samples and samples The similarity between them is calculated by cosine similarity to characterize the samples and samples The degree of similarity between The first samples, which are used to measure the spatial correlation of sensor data. Due to the physical connection of curtain wall sensors (such as vibration transmitted by keels), the similarity of adjacent sensor data is high. High similarity will reduce the biopheromone attenuation factor and suppress redundant feature initialization; and The samples and The Euclidean norm of , characterizing the size and direction of the sample.

[0081] In one embodiment, the noise robustness of the biopheromone attenuation initialization mechanism is systematically verified. By constructing test environments with different noise levels and comparing them with traditional random initialization methods, the experiment simulates common noise scenarios such as temperature drift and electromagnetic interference in sensor data. The results are shown in Figure 3 The results show that in a strong noise environment, the initialization method based on biological pheromone decay effectively suppresses the weight of noise features by dynamically adjusting node parameters, maintaining stable feature extraction capabilities. In contrast, traditional random initialization methods experience feature confusion under noise interference, resulting in a sharp decline in classification performance. This technology achieves targeted enhancement of the essential characteristics of sensor data by simulating the decay pattern of natural pheromones.

[0082] c. Apply nonlinear resonance constraint to hidden layer activation function

[0083] In the task of classifying the health status of building curtain walls, the collected data has nonlinear characteristics. For example, different types of sensors may measure changes in different dimensions at the same time, and the relationship between these changes is often complex and nonlinear.

[0084] Although the hidden layer activation functions of traditional extreme learning machines, such as the Sigmoid activation function or the ReLU activation function, can capture the basic nonlinear relationships of the data, they often cannot fully express the deep nonlinear structures in the data when faced with complex and dynamic data changes.

[0085] The hidden layer activation functions of conventional extreme learning machines, such as the Sigmoid activation function or the ReLU activation function, do not consider how to further enhance the nonlinear mapping ability of the hidden layer through the resonance effect in physics, resulting in poor classification performance of the nonlinear expression ability of the extreme learning machine model when processing complex data.

[0086] The present invention uses nonlinear resonance constraints to the hidden layer activation function to imitate the resonance effect in physics to enhance the adaptability of the hidden layer to nonlinear data. This enables the activation function to not only capture the basic nonlinear mapping of the data, but also enhance the complexity of the mapping process through the resonance effect, thereby effectively improving the model's expressive power in processing complex nonlinear relationships. The activation function of the extreme learning machine is expressed as:

[0087]

[0088] Where, is the output of the hidden layer activation function of the extreme learning machine, representing the mapping of the input data after passing through the hidden layer. is the hidden layer activation function of the extreme learning machine; is the weight of the hidden layer node of the extreme learning machine; is the bias of the hidden layer node of the extreme learning machine; It is the input of the activation function, representing the feature vector corresponding to the input data; To control the resonance effect intensity, the influence of resonance is controlled. Curtain wall damage propagation has nonlinear oscillation characteristics (such as crack expansion accompanied by stress waves). When the error is high, the resonance effect intensity is enhanced to capture the mutation signal. To control the resonance effect frequency, the speed of resonance change is controlled, and the natural frequency of the curtain wall material is simulated (such as the high-frequency response of glass curtain wall and the low-frequency response of concrete), so as to enhance the ability to capture abnormalities in specific frequency bands; is a sine function.

[0089] The resonance effect intensity and resonance effect frequency are dynamically adjusted according to the error in each iterative training to ensure that the model can adjust the effect of nonlinear mapping according to the change of error during training. For the nonlinearity of curtain wall data (such as temperature-stress coupling effect), the resonance term is used to enhance the expression ability of the activation function for complex patterns and improve the classification accuracy of minor damage, which is expressed as:

[0090]

[0091]

[0092] Where, For the The intensity of the resonance effect in the iterative training controls the oscillation effect in the nonlinear mapping; For the The resonance effect frequency in the iterative training is used to control the resonance effect frequency; Tune hyperparameters for first resonance effect strength; Tuning hyperparameters for the strength of the second resonance effect; Adjust hyperparameters for the first resonance effect frequency; Adjust hyperparameters for the second resonance effect frequency; For the The actual label of each sample (target output); The first Preferably, Set to 2, Set to 0.3, Set to 3, Set to 0.5.

[0093] In one embodiment, the nonlinear expression capabilities of the resonance activation function and the traditional activation function are compared. By constructing a joint parameter space of input features and training process, the enhancement effect of the resonance mechanism on nonlinear mapping is analyzed. Figure 4 Results show that the proposed resonant activation function maintains fundamental nonlinear characteristics while generating multi-scale response patterns through dynamic adjustment of resonant parameters, effectively characterizing the stress wave oscillation characteristics of curtain wall damage propagation. Traditional activation functions exhibit a single response characteristic, making it difficult to capture high-frequency mutation signals in complex mechanical behavior. This activation mechanism, based on the principle of physical resonance, significantly improves the model's ability to analyze the coupled characteristics of multiple physical fields.

[0094] d. Weight adjustment of extreme learning machine based on biopheromone decay mechanism and adaptive adjustment mechanism;

[0095] The health status data of building curtain walls often contains a large amount of sensor data, which may be affected by many factors such as environmental factors and sensor noise.

[0096] Conventional extreme learning machines adjust the weights of hidden layer nodes through a fixed weight update method, without making dynamic adjustments based on the contribution of each node. This makes it impossible for the model to effectively focus on meaningful features when faced with data complexity and noise, resulting in a decline in classification performance.

[0097] The present invention adopts a biopheromone attenuation mechanism and an adaptive adjustment mechanism. During each iterative training process, the weight is dynamically adjusted according to the contribution of each hidden layer node. This ensures that the extreme learning machine model always focuses on the most informative features during training while avoiding excessive reliance on noise and redundant data, thereby improving classification performance and robustness. It can be expressed as:

[0098]

[0099]

[0100] Where, For the extreme learning machine The hidden layer nodes are The updated weights in the training iteration; For the extreme learning machine The hidden layer nodes are The current weights in the training iteration; is the biopheromone attenuation factor, which controls the attenuation degree; For the extreme learning machine The hidden layer nodes are The amount of weight change during training iterations; The first samples; For the The adjustment factor of each sample controls the rate at which the weights are updated.

[0101] Furthermore, the change in weight can be adjusted according to the importance of each sample, so as to more accurately optimize the weight of the model. The calculation method of the sample adjustment factor is expressed as:

[0102]

[0103] Where, The first samples.

[0104] Molecular part ( ) is the derivative of the Sigmoid function and is used to represent the sensitivity of the activation function.

[0105] Regulatory Factors Essentially, it is to adjust the speed of updating the importance and influence of each sample (influence refers to the impact of sample diversity on model performance. For example, sample diversity can be measured by the variance of the sample's eigenvalues. A large variance of the sample's eigenvalues indicates a large diversity of the sample). The importance of a sample is determined by its activation value (calculated by the Sigmoid function) and sensitivity (represented by the size of the Sigmoid derivative). The more active the sample and the larger its activation amplitude (Sigmoid is close to 0.5), the greater its influence on the importance update. When a sample is in a "negative" state during training (Sigmoid is close to 0 or 1), its influence on the model is relatively small.

[0106] In addition, the derivatives of all samples are normalized in the denominator so that the adjustment factor of all samples is The scale should be within a uniform range to avoid the adjustment factors of some samples being too large or too small, which would affect the balance of model training.

[0107] Therefore, this formula adjusts the importance of samples by weighting them, so that the model pays more attention to samples that have a greater contribution to classification during training, avoiding excessive interference from noise samples or samples with weak information in the training process, thereby improving the training efficiency and robustness of the model. e. Dynamic regularization of hidden layer output and optimization of node activity

[0108] In the classification of building curtain wall health status, the high-dimensional characteristics of sensor data can lead to redundancy or over-activation in the output of hidden layer nodes. Some nodes may be sensitive to noise and produce abnormal activation values, affecting the classification accuracy.

[0109] Conventional extreme learning machines directly perform linear regression after the hidden layer output, lack dynamic regularization processing of the hidden layer output matrix, and cannot suppress the negative impact of redundant nodes.

[0110] The present invention adopts a dynamic regularization mechanism based on node activity. By calculating the activation variance and covariance matrix of each hidden layer node, the regularization coefficient is dynamically adjusted to suppress the contribution of noise-sensitive nodes and enhance robustness. The regularization term of the extreme learning machine hidden layer output matrix is defined as:

[0111]

[0112] Where, For the extreme learning machine The output vector of hidden layer nodes; For the extreme learning machine The output vector of hidden layer nodes; For the extreme learning machine Regularization coefficient of hidden layer nodes: is a dynamic regularization term that reflects the redundancy and covariance relationship between the hidden layer node outputs; is the number of hidden layer nodes; is the covariance function; is the L2 norm; For the extreme learning machine The first and The covariance penalty coefficient of the hidden layer nodes.

[0113] Furthermore, the regularization coefficient of the hidden layer nodes is calculated based on the variance ratio of the node output to ensure that the model relies on nodes with higher stability, thereby improving the robustness of the classification. The calculation method is expressed as:

[0114]

[0115] Where, For the extreme learning machine The variance of the output of the hidden layer node reflects the The sensitivity of hidden layer nodes to input changes. Nodes with large variance are sensitive to input changes (such as those affected by temperature drift). Their weights are reduced through regularization coefficients to suppress environmental noise. For the extreme learning machine The variance of the output of each hidden layer node.

[0116] Furthermore, the covariance penalty coefficient dynamically adjusts the regularization strength. The curtain wall sensor network has spatial correlation (such as synchronous vibration of adjacent sensors). High covariance nodes impose penalties on redundant features to avoid over-activation, forcing the extreme learning machine model to prioritize nodes with stable activation and low correlation with noise, thereby improving classification stability. The calculation method is expressed as:

[0117]

[0118] Where, Represents the variance function, which is used to measure the dispersion of node outputs and optimize the regularization strength between nodes.

[0119] In one embodiment, the regulation effect of the dynamic regularization mechanism on the correlation of hidden layer nodes is analyzed by covariance heat map. Figure 5The experiment compared the distribution characteristics of hidden layer node activation patterns before and after regularization. The results show that the present invention effectively suppresses the activation intensity of highly noise-sensitive nodes through the regularization coefficient adjustment based on node variance adaptation, while reducing the covariance correlation between spatially redundant nodes. Compared with traditional static regularization methods, this technology can dynamically optimize node contributions based on real-time data features, significantly improving the model's robustness to interference factors such as sensor drift while ensuring feature expression capabilities. The dynamic optimization strategy based on node activity provides stability for long-term health monitoring.

[0120] f. Error feedback driven biopheromone attenuation factor update

[0121] The building curtain wall sensor data has time-varying characteristics. The noise level and data distribution in different time periods may change. It is necessary to dynamically adjust the biopheromone attenuation factor in the initialization parameters to adapt to the time-varying characteristics of the building curtain wall sensor data.

[0122] The present invention updates the pheromone attenuation factor based on the relative change of the loss function gradient, thereby dynamically adapting to the time-varying characteristics of the data, which can be expressed as:

[0123]

[0124] Where, For the The biopheromone decay factor of the iteration; For the The biopheromone decay factor of the iteration; is an exponential function whose base is a natural constant; is the decay rate control parameter; For the extreme learning machine The gradient of the loss function at iterations; For the extreme learning machine The gradient of the loss function at iterations; is the loss function of the extreme learning machine.

[0125] Based on this, when the loss gradient increases ( ), indicating that the model is in an unstable state and the attenuation effect needs to be enhanced (reduced ) to reduce the impact of noise; when the loss gradient decreases ( ), indicating that the model tends to converge and the attenuation effect needs to be weakened (increase ) to retain valid features.

[0126] g. Calculate the loss function of the extreme learning machine;

[0127] In the health status classification of building curtain walls, structural damage often has spatial propagation characteristics (such as cracks extending from the edge to the center, and stress concentration areas causing abnormalities in adjacent sensors).

[0128] Conventional loss functions only consider single-point prediction errors and do not model the damage propagation law between sensor networks, resulting in insufficient sensitivity to early local damage.

[0129] The present invention adopts the structural damage propagation consistency constraint term in the loss function. By integrating the curtain wall physical connection topology and the damage propagation prior knowledge, the constraint model prediction results conform to the laws of engineering mechanics. Combined with the regularization term, the calculation method is:

[0130]

[0131] Where, is the loss function of the extreme learning machine, which represents the overall error between the model prediction value and the actual label; is the number of training samples, that is, the number of samples of input data; Represents the output value of the extreme learning machine hidden layer activation function, representing the The predicted value after the sample passes through the hidden layer; The weight matrix of the hidden layer nodes of the extreme learning machine is initialized and dynamically adjusted through the biopheromone decay mechanism; The first samples; The bias of the hidden layer nodes of the extreme learning machine is initialized and dynamically adjusted through the biopheromone decay mechanism; No. The actual label (target output value) of each sample, that is, the true classification result of the health status of the building curtain wall; Characterizes the square of the prediction error of a single sample, which is used to quantify the degree of deviation between the predicted value and the true value; is a dynamic regularization term used to suppress the negative impact of redundant nodes and noise-sensitive nodes; is the consistency constraint strength; is the structural damage propagation consistency constraint. Preferably, Set to 0.2.

[0132] The structural damage propagation consistency constraint encodes the mechanical properties of the curtain wall into the loss function through topological connections, forcing the model to learn prediction results that conform to the actual damage propagation pattern. The correlation constraint on the changes in the predictions of adjacent sensors can amplify the chain effect of local minor anomalies and improve the ability to detect early faults. The calculation method is expressed as:

[0133]

[0134] Where, It is the set of topological connection edges of the curtain wall sensor network, representing the physical connection relationship; Indicates the In the sample The characteristics of the sensor, Indicates the In the sample Characteristics of each sensor; The temporal gradient of the loss function of the last iteration of the extreme learning machine with respect to the sensor data is approximated by the difference of the prediction values of adjacent time steps; is the damage propagation attenuation factor, according to Node to The mechanical transfer coefficient of each node is calculated as ,in, is the L2 norm, For the The three-dimensional coordinates of the sensors, For the The three-dimensional coordinates of the sensors, For connecting edges The angle with the direction of the principal stress of the curtain wall, is the first material characteristic parameter, For the second material characteristic parameter, for example, the concrete curtain wall , .

[0135] In one embodiment, the curtain wall sensor network topology connection edge set is specifically: , where each edge Indicates that in the physical structure of the curtain wall, The sensor and There is a direct mechanical transmission path between the building components (such as glass panels, keels, connectors, etc.) where the two sensors are located. The construction method can be to directly connect the two sensors through the curtain wall load-bearing structure (such as keels, frames) according to the location of the two sensors, or to digitally generate them based on the node connection relationship in the curtain wall CAD drawings.

[0136] h. Model convergence conditions;

[0137] Considering the long-term and continuous operation of building curtain wall health monitoring, a dual-threshold convergence condition is defined, which is expressed as:

[0138]

[0139] Where, is the first error threshold, which is preset manually; For the The actual label of each sample (target output); The second error threshold is preset manually. Set to 0.01, Set to 0.02.

[0140] When the dual-threshold convergence condition is met, the iteration stops, indicating that the model training is completed.

[0141] S4: Damage detection and assessment: The trained model is applied in real time to assess the health status of the building curtain wall and determine whether there is any damage.

[0142] Damage detection and assessment performs real-time evaluation of monitoring data based on machine learning models and outputs health status assessment results.

[0143] In one embodiment, the prediction results are different health status levels, which are divided into 5 levels. The higher the level, the better the health status.

[0144] It should be noted that the category of the prediction result corresponds to the classification category of the extreme learning machine and depends on the labeled category of the training samples used to train the extreme learning machine model. For example, the labeled categories are different health status levels, which are divided into 5 levels. The higher the level, the better the health status, and the type of the prediction result is the same.

[0145] S5: Visual display, presenting monitoring data and model analysis results to users in a graphical manner.

[0146] The visualization display unit is responsible for presenting real-time health monitoring data, prediction results, and the operating status of the model to users in the form of easy-to-understand charts, heat maps, trend charts, etc., helping users to intuitively understand the health status of the building curtain wall.

[0147] The functions of the visualization display unit include graphical display of real-time data streams, dynamic display of prediction results, statistical analysis and trend display of historical data, etc. Through the visualization display unit, users can easily view and analyze the health status of building curtain walls and make maintenance decisions.

[0148] S6: System management and monitoring, manage, control and monitor the entire monitoring system.

[0149] This includes managing the system's hardware and software components, monitoring the system's operational status, and ensuring system stability and efficiency. This includes device status monitoring, data flow management, system logging, error alarms, and repairs.

[0150] In this embodiment, by assigning dynamic weights to sensors at different locations, the significance of signals in vulnerable areas is enhanced, especially by assigning higher weights to sensors near the edges and joints of building curtain walls, thereby improving the signal's responsiveness to local damage, avoiding the equivalent processing of sensor data in conventional methods, and significantly improving the sensitivity of early warning of damage. By simulating the pheromone attenuation effect in nature and dynamically adjusting the weight and bias of each node, the model's adaptability to complex, high-dimensional data is enhanced, noise robustness is improved, and the impact of redundant features in the data is reduced. Compared with traditional random initialization methods, it can better handle noise and heterogeneous data. In addition, the present invention imitates the resonance effect in physics and introduces it into the hidden layer activation function of the extreme learning machine, thereby enhancing the hidden layer's adaptability to nonlinear data. It not only captures the basic nonlinear mapping of the data, but also improves the sensitivity to complex patterns and small damage signals through the resonance effect, improving the performance of traditional activation functions in complex, dynamic data environments. By combining a biopheromone attenuation mechanism with an adaptive adjustment mechanism, the weights are dynamically adjusted during each iteration based on the contribution of hidden layer nodes, focusing on meaningful features and avoiding the influence of noise and redundant data, thereby improving the model's robustness and classification performance. By calculating the activation variance and covariance matrices of the hidden layer nodes and dynamically adjusting the regularization coefficient, the negative impact of redundant and noise-sensitive nodes is suppressed, effectively reducing the activation intensity of highly noise-sensitive nodes, strengthening the model's reliance on stable and effective features, and improving classification stability. Incorporating the physical topology and damage propagation laws of building curtain walls, a structural damage propagation consistency constraint is proposed, ensuring that the model's predictions account for the spatial propagation characteristics between sensors, enhancing sensitivity to early localized damage, and improving the system's fault detection capabilities. Dynamically adjusting the biopheromone attenuation factor based on the relative change in the loss function gradient enables the model to adapt to the time-varying nature of the data, effectively controlling the impact of noise on model training, and enhancing model stability.

[0151] The embodiment of the present invention further provides a building curtain wall monitoring data processing and visualization system, including a data acquisition unit, a data preprocessing unit, a machine learning modeling unit, a damage detection and prediction unit, a visualization display unit, and a system management and monitoring unit, wherein:

[0152] The data acquisition unit is used to collect health monitoring data of building curtain walls in real time and supports various data upload methods such as database, file, HTTP, etc.

[0153] The data preprocessing unit is used to perform denoising, standardization and normalization on the collected raw data, and also has the functions of high-speed caching and multi-dimensional analysis.

[0154] The machine learning unit uses machine learning algorithms to build predictive models to assess building health and detect damage;

[0155] The machine learning modeling unit is used to build a prediction model based on the extreme learning machine algorithm to assess the building health and detect damage. It learns the health status pattern of the building curtain wall through training data, builds a classifier model, and obtains the health status classification prediction results;

[0156] The damage detection and prediction unit is used to evaluate the real-time monitoring data based on the machine learning model and output the predicted results of the health status;

[0157] The visualization display unit is used to present monitoring data and model analysis results to users in a graphical manner. Visual elements include interactive dashboards, reports, and web interfaces to help users clearly understand the health status of the building;

[0158] The system management and monitoring unit is responsible for ensuring the normal operation of the entire system by monitoring the health of hardware and software, managing data flow and performing daily maintenance.

[0159] In addition, an embodiment of the present invention also proposes a computer-readable storage medium, on which program instructions of a method for visualizing building curtain wall monitoring data processing are stored. The program instructions for visualizing building curtain wall monitoring data processing can be executed by one or more processors to implement the steps of the method for visualizing building curtain wall monitoring data processing as described above.

[0160] The embodiments described above are merely descriptions of preferred implementations of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should fall within the scope of protection determined by the claims of the present invention.

Claims

1. A method for visualizing building curtain wall monitoring data processing, characterized in that: The steps include: S1: Data collection: real-time collection of building curtain wall health monitoring data. Specifically, the physical status information of the curtain wall is obtained through multiple sensors. S2: Data preprocessing, including data cleaning, denoising, standardization, normalization, and missing data filling of the collected raw data; S3: Machine learning modeling for health status classification prediction. The model uses the extreme learning machine algorithm to obtain health status classification prediction results. Specifically, an adaptive feature weighting mechanism based on the spatial location of sensors, combined with the mechanical properties of the curtain wall structure, assigns dynamic weights to sensors in different locations. The position weights of sensors at different positions are: ； Where, For the The spatial weight of each sensor; For the The Euclidean distance from each sensor to the nearest edge of the curtain wall; For the The Euclidean distance from each sensor to the nearest edge of the curtain wall; is the maximum value of all sensor edge distances; is the spatial sensitivity coefficient; is the total number of sensors; The extreme learning machine algorithm uses nonlinear resonance constraints to the hidden layer activation function, and the activation function is expressed as: ； Where, is the output of the hidden layer activation function of the extreme learning machine, is the hidden layer activation function of the extreme learning machine; is the weight of the hidden layer node of the extreme learning machine; is the bias of the hidden layer node of the extreme learning machine; It is the input of the activation function, representing the feature vector corresponding to the input data; is the resonance effect intensity; is the resonance effect frequency; S4: Damage detection and prediction: real-time application of the trained model to detect and predict the health status of building curtain walls; S5: Visual display, presenting monitoring data and model analysis results to users in a graphical manner; S6: System management and monitoring, manage, control and monitor the entire monitoring system.

2. The method for visualizing building curtain wall monitoring data processing according to claim 1, characterized in that: The extreme learning machine algorithm initializes the weight and bias parameters of the extreme learning machine hidden layer nodes based on the biopheromone decay mechanism, dynamically adjusts the weight and bias of each node according to the characteristic distribution of input data, and accepts random noise for correction.

3. The method for visualizing building curtain wall monitoring data processing according to claim 1, characterized in that: The resonance effect intensity and the resonance effect frequency are expressed as follows: ； ； Where, For the The intensity of the resonance effect in the iterative training controls the oscillation effect in the nonlinear mapping; For the The resonance effect frequency in the iterative training is used to control the resonance effect frequency; Tune hyperparameters for first resonance effect strength; Tuning hyperparameters for the strength of the second resonance effect; Adjust hyperparameters for the first resonance effect frequency; Adjust hyperparameters for the second resonance effect frequency; For the The actual labels of the samples; The first samples.

4. The method for visualizing building curtain wall monitoring data processing according to claim 1, characterized in that: The extreme learning machine algorithm is based on a dynamic regularization mechanism of node activity, which dynamically adjusts the regularization coefficient of the hidden layer nodes by calculating the activation variance and covariance matrix of each hidden layer node.

5. The method for visualizing building curtain wall monitoring data processing according to claim 1, characterized in that: The extreme learning machine algorithm performs an update calculation of the biopheromone attenuation factor based on the relative change of the loss function gradient, which is expressed as: ； Where, For the The biopheromone decay factor of the iteration; For the The biopheromone decay factor of the iteration; is the decay rate control parameter; For the extreme learning machine The gradient of the loss function at iterations; For the extreme learning machine The gradient of the loss function at iterations; is the loss function of the extreme learning machine.

6. The method for visualizing building curtain wall monitoring data processing according to claim 1, characterized in that: The loss function of the extreme learning machine is calculated as follows: ； Where, is the loss function of the extreme learning machine; is the number of training samples; Represents the output value of the extreme learning machine hidden layer activation function; is the weight matrix of the hidden layer nodes of the extreme learning machine; The first samples; is the bias of the hidden layer nodes of the extreme learning machine; No. The actual labels of the samples; Characterizes the square of the prediction error of a single sample; is the dynamic regularization term; is the consistency constraint strength; is the consistency constraint term for structural damage propagation.

7. A building curtain wall monitoring data processing and visualization system, based on the building curtain wall monitoring data processing and visualization method according to any one of claims 1 to 6, comprising a data acquisition unit, a data preprocessing unit, a machine learning modeling unit, a damage detection and prediction unit, a visualization display unit, and a system management and monitoring unit, wherein: The data acquisition unit is used to collect health monitoring data of building curtain walls in real time; The data preprocessing unit is used to perform denoising, standardization and normalization on the collected raw data; The machine learning modeling unit is used to learn the health status pattern of the building curtain wall through training data based on the extreme learning machine algorithm, build a classifier model, and obtain the health status classification prediction results; The damage detection and prediction unit is used to evaluate the real-time monitoring data based on the machine learning model and output the predicted results of the health status; The visualization display unit is used to present monitoring data and model analysis results to users in a graphical manner; The system management and monitoring unit is used to manage, control and monitor the entire monitoring system.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores program instructions for a method for visualizing building curtain wall monitoring data processing. The program instructions for visualizing building curtain wall monitoring data processing can be executed by one or more processors to implement the steps of the method for visualizing building curtain wall monitoring data processing as described in any one of claims 1-6.

Images