Method, device and system for predicting damage of hyperbolic glass curtain wall

Abstract

The invention provides a hyperbolic glass curtain wall damage prediction method, device and system. The method comprises the following steps: acquiring sensor acquisition information of a hyperbolic glass curtain wall; a preset unmanned aerial vehicle is controlled to collect a visible light image and a thermal imaging image corresponding to the hyperbolic glass curtain wall; the unmanned aerial vehicle carries visible light equipment and a thermal infrared imager; inputting the visible light image and the thermal imaging image into a damage feature recognition model based on a convolutional neural network, and outputting image feature information; inputting information acquired by the sensor into a physical parameter analysis model based on a time sequence, and outputting parameter characteristic information; fusing the image feature information with the parameter feature information to obtain a damage feature recognition result; and constructing a health state prediction model according to the damage feature recognition, inputting the real-time sensor acquisition information, the real-time visible light image and the real-time thermal imaging image corresponding to the hyperbolic glass curtain wall into the health state prediction model, and outputting a damage prediction result corresponding to the hyperbolic glass curtain wall.

Description

Technical Field

[0001] The present application relates to the field of defect detection technology, and in particular to a method, device and system for predicting damage of a hyperbolic glass curtain wall. Background Art

[0002] With the development of modern architectural design, hyperbolic glass curtain walls, as a unique building facade material, are widely used in large buildings due to their aesthetics and unique shape. The curved shape of the hyperbolic glass curtain wall increases the complexity of its structure and puts higher requirements on damage detection. Traditional glass curtain wall detection methods, such as visual inspection, vibration detection, thermal imaging and ultrasonic detection, cannot adapt to the special geometric form of hyperbolic glass curtain walls, resulting in insufficient detection accuracy and positioning accuracy, and unable to obtain real-time and comprehensive status information of the curtain wall. It is difficult to detect early minor damage and it is difficult to accurately predict and analyze the potential damage of the curtain wall. The existing detection methods have at least the following problems:

[0003] Low detection accuracy: Hyperbolic curtain walls have complex geometric shapes. Traditional methods have low detection accuracy for damage on complex hyperbolic surfaces and it is difficult to detect subtle damage in the early stages.

[0004] Difficult to achieve real-time monitoring and trend analysis: Traditional detection methods cannot obtain real-time and continuous status information of hyperbolic glass curtain walls. It is difficult to scientifically analyze the damage development trend of hyperbolic glass curtain walls, unable to predict potential safety risks in advance, and unable to provide a strong basis for maintenance decisions.

[0005] Lack of accurate life assessment: The existing hyperbolic glass curtain wall damage prediction method is simply based on the design service life or limited experience, without fully considering the impact of various complex factors during the actual use of the curtain wall. It is difficult to accurately assess the remaining service life of the curtain wall, which is not conducive to the formulation of a reasonable curtain wall maintenance plan and replacement plan.

[0006] Therefore, a method is urgently needed to solve at least one of the above problems. Summary of the invention

[0007] This application provides a method, device and system for predicting damage of a hyperbolic glass curtain wall, aiming to solve the problem that traditional design methods usually rely on experienced engineers to manually adjust parameters, which is not only time-consuming and labor-intensive, but also difficult to ensure the optimality of the final design solution. Especially when multiple factors need to be considered, how to achieve an aesthetically pleasing and cost-effective solution while meeting functional requirements has become an urgent problem to be solved.

[0008] In a first aspect, the present application provides a method for predicting damage of a hyperbolic glass curtain wall, comprising:

[0009] Acquire sensor collection information of the hyperbolic glass curtain wall; the sensor collection information at least includes strain sensor information, temperature sensor information and displacement sensor information;

[0010] Controlling a preset drone to collect visible light images and thermal imaging images corresponding to the hyperbolic glass curtain wall; the drone is equipped with a visible light device and an infrared thermal imager;

[0011] Inputting the visible light image and the thermal imaging image into a damage feature recognition model based on a convolutional neural network, and outputting image feature information;

[0012] Inputting the sensor collected information into a time series-based physical parameter analysis model and outputting parameter characteristic information;

[0013] The image feature information is integrated with the parameter feature information to obtain a damage feature recognition result;

[0014] A health status prediction model is constructed according to the damage feature identification, so that real-time sensor collection information, real-time visible light images and real-time thermal imaging images corresponding to the hyperbolic glass curtain wall are input into the health status prediction model, and the health status prediction model outputs the damage prediction result corresponding to the hyperbolic glass curtain wall in real time.

[0015] In some embodiments, before the visible light image and thermal imaging image are input into a damage feature recognition model based on a convolutional neural network, the method further includes: preprocessing the sensor acquisition information, visible light image and thermal imaging image; the preprocessing includes at least filtering, denoising, enhancement, and cropping preprocessing; marking damage information in the preprocessed sensor acquisition information, visible light image and thermal imaging image; the damage information includes at least damage type and timing abnormality.

[0016] In some embodiments, the health status prediction model outputs in real time a damage prediction result corresponding to the hyperbolic glass curtain wall, including: the health status prediction model outputs a crack index and a deformation index corresponding to the hyperbolic glass curtain wall according to the real-time visible light image and the real-time thermal imaging image; the health status prediction model outputs a temperature influence factor and a stress distribution factor according to the real-time sensor acquisition information; and the damage prediction result is generated according to the crack index, deformation index, temperature influence factor and stress distribution factor.

[0017] Exemplarily, generating the damage prediction result according to the crack index, deformation index, temperature influence factor and stress distribution factor includes: calculating a health score index according to the crack index, deformation index, temperature influence factor and stress distribution factor; generating the damage prediction result according to the health score index; the expression of the health score index includes:

[0018] ;in, for The health score index at all times, for Moment Dynamic weight coefficient of the damage type, for Moment Injury type score, is the number of damage types, which at least include cracks and deformations, is the environmental and geometric coupling coefficient corresponding to the hyperbolic glass curtain wall, which is used to reflect the accelerating effect of seasonal changes on curtain wall damage. is the curvature change correction term, which is used to quantify the additional impact of hyperbolic surface geometric deformation on the health score.

[0019] It should be noted that, in some embodiments, the expression of the dynamic weight coefficient includes:

[0020] ;

[0021] in, for Moment Dynamic weight coefficient of the damage type, is the initial weight coefficient, For the The damage sensitivity attenuation factor of the damage index is used to control the timeliness of different damage types. For the Damage sensitivity attenuation factor of the damage index, is the number of damage types.

[0022] It should be noted that, in some embodiments, if the damage type is crack, the initial weight coefficient is 0.4, and the corresponding value range of the damage sensitivity attenuation factor is 0.05 to 0.15; if the damage type is deformation, the initial weight coefficient is 0.3, and the corresponding value range of the damage sensitivity attenuation factor is 0.03 to 0.18.

[0023] In a second aspect, the present application provides a hyperbolic glass curtain wall damage prediction device, comprising:

[0024] An information acquisition module, used to acquire sensor acquisition information of the hyperbolic glass curtain wall; the sensor acquisition information at least includes strain sensor information, temperature sensor information and displacement sensor information;

[0025] An image acquisition module, used to control a preset drone to collect visible light images and thermal imaging images corresponding to the hyperbolic glass curtain wall; the drone is equipped with a visible light device and an infrared thermal imager;

[0026] An image output module, used for inputting the visible light image and the thermal imaging image into a damage feature recognition model based on a convolutional neural network, and outputting image feature information;

[0027] A feature output module, used to input the sensor collected information into a time series-based physical parameter analysis model and output parameter feature information;

[0028] A feature fusion module, used to fuse the image feature information with the parameter feature information to obtain a damage feature recognition result;

[0029] A prediction completion module is used to build a health status prediction model according to the damage feature identification, so as to input the real-time sensor collection information, real-time visible light image and real-time thermal imaging image corresponding to the hyperbolic glass curtain wall into the health status prediction model, and the health status prediction model outputs the damage prediction result corresponding to the hyperbolic glass curtain wall in real time.

[0030] In a third aspect, the present application provides a hyperbolic glass curtain wall damage prediction system, comprising: a plurality of sensors, the sensors being arranged at target points of the hyperbolic glass curtain wall, the sensors comprising at least a strain sensor, a temperature sensor and a displacement sensor; a drone, the drone being equipped with a visible light device and an infrared thermal imager; a controller, the controller comprising a memory and a processor; the memory being used to store a computer program; the processor being used to execute the computer program and to implement a method provided in any embodiment of the present application when executing the computer program.

[0031] In some embodiments, the target points include glass panel edges, connection nodes, and stress concentration areas corresponding to the hyperbolic glass curtain wall.

[0032] In some embodiments, the arrangement density of the sensors in the hyperbolic glass curtain wall is dynamically adjusted according to the curvature radius of the hyperbolic glass curtain wall. In the area where the curvature radius is less than 3m, the arrangement density is increased by a preset density coefficient compared to the area where the curvature radius is greater than or equal to 3m. The preset density coefficient corresponds to a value range of 30% to 50%.

[0033] In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program. When the computer-readable instructions are executed by the processor, one or more processors execute a method as provided in any embodiment of the present application.

[0034] The present application provides a method, device and system for predicting damage of a hyperbolic glass curtain wall, the method comprising the following technical steps:

[0035] Sensor information collection: Strain sensor: used to monitor the stress distribution of the hyperbolic glass curtain wall. Temperature sensor: used to monitor the temperature changes of the curtain wall surface and the surrounding environment. Displacement sensor: used to monitor the deformation of the curtain wall.

[0036] UAV image acquisition: Visible light equipment: carried on the drone, used to collect high-resolution visible light images of the hyperbolic glass curtain wall. Infrared thermal imager: carried on the drone, used to collect thermal imaging images of the curtain wall to detect potential temperature abnormalities.

[0037] Damage feature recognition model based on convolutional neural network (CNN): Input: visible light image and thermal imaging image. Processing: Extract feature information in the image through convolution layer, pooling layer and fully connected layer. Output: Image feature information, including but not limited to damage features such as cracks, scratches, deformation, etc.

[0038] Time series-based physical parameter analysis model: Input: strain sensor information, temperature sensor information, and displacement sensor information. Processing: Use time series analysis methods (such as ARIMA, LSTM, etc.) to model sensor data and extract key physical parameter features. Output: parameter feature information, including stress distribution, temperature change trend, displacement change trend, etc.

[0039] Feature fusion and damage identification: Fusion method: Fusion of image feature information and parameter feature information, which can be achieved by weighted fusion, principal component analysis (PCA) or deep learning methods (such as multimodal fusion network). Output: Damage feature identification results, which comprehensively reflect the health status of the curtain wall.

[0040] Health status prediction model: Input: real-time sensor information, real-time visible light images, and real-time thermal imaging images. Processing: Input the above information into a pre-trained health status prediction model, which can be based on machine learning or deep learning methods, such as support vector machine (SVM), random forest, convolutional neural network (CNN), etc. Output: Real-time output of damage prediction results of the hyperbolic glass curtain wall, including damage type, location, severity, etc.

[0041] The specific implementation methods of the above technical features may be:

[0042] Sensor deployment Install strain sensors, temperature sensors, and displacement sensors on the hyperbolic glass curtain wall and ensure that they work properly. The data collection frequency and accuracy of the sensors need to be set according to actual needs. Select a suitable drone model and equip it with visible light equipment and infrared thermal imagers. Design a flight path to ensure that all areas of the hyperbolic glass curtain wall can be fully covered. Regularly perform drone inspection tasks to collect visible light images and thermal imaging images. Collect a large amount of hyperbolic glass curtain wall damage sample data, including visible light images, thermal imaging images, and sensor data. Use convolutional neural networks (CNNs) to train image data and extract damage features. Use time series analysis methods to train sensor data and extract physical parameter features. Fusion image features and physical parameter features to train damage feature recognition models. Collect historical data of the hyperbolic glass curtain wall, including sensor data, image data, and known damage conditions. Select a suitable machine learning or deep learning algorithm to build a health status prediction model. Train and verify the model to ensure that it has high prediction accuracy. Integrate the sensor data acquisition system, drone image acquisition system, damage feature recognition model, and health status prediction model. Deployed on the cloud or local server to achieve real-time data transmission and processing. Provide a user interface to display real-time damage prediction results and historical data analysis reports.

[0043] The provided method has at least the following beneficial effects:

[0044] Improve damage prediction accuracy: By combining image data and sensor data, the damage characteristics of the hyperbolic glass curtain wall can be more comprehensively captured, improving the accuracy of the prediction.

[0045] Reduce manual intervention: Automated data collection and processing processes reduce reliance on experienced engineers and reduce labor costs.

[0046] Real-time monitoring and early warning: It can monitor the health status of the curtain wall in real time, promptly discover potential damage risks, take measures in advance for maintenance, and avoid major accidents.

[0047] Optimize design solutions: Through the analysis of large amounts of data, we can provide references for the design of hyperbolic glass curtain walls, help optimize design solutions, and improve the overall performance of buildings.

[0048] Reduce costs: By detecting and repairing damage early, the service life of the curtain wall can be effectively extended and long-term maintenance costs can be reduced.

[0049] In summary, the hyperbolic glass curtain wall damage prediction method provided in this application realizes efficient and accurate prediction of curtain wall damage by comprehensively utilizing sensor data and image processing technology, providing strong support for the safety and maintenance of buildings.

[0050] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 is a schematic flow chart of the steps of a method for predicting damage of a hyperbolic glass curtain wall provided in one embodiment of the present application;

[0053] Figure 2 It is a schematic block diagram of the structure of a hyperbolic glass curtain wall damage prediction device provided in one embodiment of the present application;

[0054] Figure 3 It is a schematic block diagram of the structure of a controller provided in one embodiment of the present application.

[0055] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0057] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0058] It should be understood that, in order to facilitate the clear description of the technical solutions of the embodiments of the present invention, in the embodiments of the present invention, the words "first", "second", etc. are used to distinguish the same items or similar items with substantially the same functions and effects. Those skilled in the art can understand that the words "first", "second", etc. do not limit the quantity and execution order, and the words "first", "second", etc. do not necessarily limit the difference.

[0059] It should be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0060] It should also be understood that the term “and / or” used in the specification and appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0061] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0062] With the development of modern architectural design, hyperbolic glass curtain walls, as a unique building facade material, are widely used in large buildings due to their aesthetics and unique shape. The curved shape of the hyperbolic glass curtain wall increases the complexity of its structure and puts higher requirements on damage detection. Traditional glass curtain wall detection methods, such as visual inspection, vibration detection, thermal imaging and ultrasonic detection, cannot adapt to the special geometric form of hyperbolic glass curtain walls, resulting in insufficient detection accuracy and positioning accuracy, and unable to obtain real-time and comprehensive status information of the curtain wall. It is difficult to detect early minor damage and it is difficult to accurately predict and analyze the potential damage of the curtain wall. The existing detection methods have at least the following problems:

[0063] Low detection accuracy: Hyperbolic curtain walls have complex geometric shapes. Traditional methods have low detection accuracy for damage on complex hyperbolic surfaces and it is difficult to detect subtle damage in the early stages.

[0064] Difficult to achieve real-time monitoring and trend analysis: Traditional detection methods cannot obtain real-time and continuous status information of hyperbolic glass curtain walls. It is difficult to scientifically analyze the damage development trend of hyperbolic glass curtain walls, unable to predict potential safety risks in advance, and unable to provide a strong basis for maintenance decisions.

[0065] Lack of accurate life assessment: The existing hyperbolic glass curtain wall damage prediction method is simply based on the design service life or limited experience, without fully considering the impact of various complex factors during the actual use of the curtain wall. It is difficult to accurately assess the remaining service life of the curtain wall, which is not conducive to the formulation of a reasonable curtain wall maintenance plan and replacement plan.

[0066] Therefore, a method is urgently needed to solve at least one of the above problems.

[0067] To solve the above problems, please refer to Figure 1 ,like Figure 1As shown, the provided method for predicting damage of a hyperbolic glass curtain wall includes steps S101 to S106. The details are as follows:

[0068] Step S101. Acquire sensor collection information of the hyperbolic glass curtain wall; the sensor collection information at least includes strain sensor information, temperature sensor information and displacement sensor information.

[0069] Specifically, the sensors installed on the hyperbolic glass curtain wall include but are not limited to strain sensors, temperature sensors and displacement sensors. Strain sensors are used to monitor stress changes in the curtain wall structure and reflect material deformation. Temperature sensing: Measuring ambient temperature and temperature distribution on the curtain wall surface helps identify potential risks caused by temperature changes. Displacement sensors monitor the relative position changes between curtain wall components to help detect structural movement or deformation problems. The data collection frequency is set according to actual needs, usually every minute or more frequently to ensure that rapidly changing status information can be captured. Sensor data is sent to the central processing unit through wireless communication modules (such as LoRa, NB-IoT, etc.).

[0070] For example, various sensors can be arranged rationally on the hyperbolic glass curtain wall to ensure coverage of all key areas. Sensors can be calibrated regularly to ensure data accuracy. The IoT platform can be used to manage the sensor network and collect and store data from each node.

[0071] This step realizes comprehensive monitoring of the curtain wall status, especially for the parts that are difficult to directly observe. It provides rich physical parameter data to support subsequent analysis and improves the reliability and accuracy of damage detection.

[0072] Step S102: Control a preset drone to collect visible light images and thermal imaging images corresponding to the hyperbolic glass curtain wall; the drone is equipped with a visible light device and an infrared thermal imager.

[0073] Choose a commercial drone with automatic flight capabilities and equipped with a high-definition camera and infrared thermal imager. Prepare a detailed flight route map in advance to ensure that the drone can follow the predetermined trajectory to take comprehensive photos around the outside of the building. Ensure that the pictures taken are clear enough for post-processing; at the same time, adjust the thermal imager parameters so that the temperature difference is obvious and it is easy to distinguish between normal and abnormal areas.

[0074] For example, design the appropriate flight height and angle according to the size and shape of the building. Set the drone to perform tasks at a scheduled time, such as automatically taking off for inspections at a specific time every day. Use the ground station software to remotely control the drone to complete the specified action and download and save the shooting data.

[0075] High-quality visual data can be obtained without manual high-altitude operations, reducing operational difficulty and safety risks. Combining image data from multiple perspectives enhances the overall perception of complex curved structures. Infrared thermal imaging technology can effectively reveal hidden problems inside, supplementing information dimensions that are difficult to reach with traditional methods.

[0076] Step S103: Input the visible light image and the thermal imaging image into a damage feature recognition model based on a convolutional neural network, and output image feature information.

[0077] Specifically, a CNN architecture optimized with a large number of training samples is used, such as classic models such as VGGNet and ResNet. A labeled dataset containing various known damage types is used to train the model so that it can learn to identify typical defects such as cracks and fissures from different types of images. The model extracts multi-level features from the input image to form a high-level abstract representation.

[0078] The original images collected by the drone are fed into the pre-trained neural network in batches. The model automatically generates one or more feature maps, which highlight the locations where damage is suspected. The noise interference is filtered out by setting the threshold, and the truly meaningful signals are retained as the basis for the next step of analysis.

[0079] The ability to quickly locate problems from massive images has been greatly improved, saving labor costs. Compared with human visual inspection, it is more objective and fair, and less susceptible to subjective factors. Through machine learning, the algorithm performance is continuously improved, and the accuracy and robustness are continuously improved.

[0080] Step S104: Input the sensor collected information into a time series-based physical parameter analysis model, and output parameter characteristic information.

[0081] Specifically, apply long short-term memory networks (LSTM) or other recurrent neural network (RNN) variants suitable for time series data processing. Data preprocessing: Perform preprocessing steps such as normalization and denoising on the raw sensor readings to make them conform to the model input format requirements. Feature engineering: Construct new feature variables, such as rate of change, cumulative effect, etc., to enhance the learning ability of the model.

[0082] For example, you can feed the organized historical records to the LSTM model to let it learn the trend patterns that evolve over time. During the training process, you need to pay attention to the overfitting phenomenon and adjust the hyperparameters in time to avoid the model being too complex. You should update the training set regularly and introduce the latest observation data to keep the model up to date.

[0083] It can capture subtle trends and warn of possible problems in advance. It provides a strong quantitative basis for the subsequent stages and enhances the scientific nature of the overall system. Predictions based on historical data are helpful to guide future maintenance work arrangements.

[0084] Step S105: Fuse the image feature information with the parameter feature information to obtain a damage feature recognition result.

[0085] Specifically, the fusion strategy can adopt a simple splicing method or map the two to the same space and then perform weighted summation. The weight distribution dynamically adjusts the proportional coefficient according to the contribution of each part in different scenarios. Comprehensive judgment combines visual clues and physical quantity indicators to jointly determine the final conclusion.

[0086] Design a reasonable evaluation system to quantify the importance of each indicator. Develop specialized software tools to automate the integration process and reduce human intervention. Perform validation tests on the results obtained after integration to ensure their effectiveness.

[0087] It gives full play to the complementary advantages between different types of data and improves the practical value of the overall solution. It makes the basis for decision-making more comprehensive and reliable, reduces the probability of misjudgment, promotes the cross-integration of interdisciplinary knowledge, and promotes technological progress in related fields.

[0088] Step S106. A health status prediction model is constructed according to the damage feature identification, so that the real-time sensor collection information, real-time visible light image and real-time thermal imaging image corresponding to the hyperbolic glass curtain wall are input into the health status prediction model, and the health status prediction model outputs the damage prediction result corresponding to the hyperbolic glass curtain wall in real time.

[0089] Specifically, model framework: choose ensemble learning methods such as random forest, gradient boosting machine (GBM) or deep belief network (DBN). Input and output definition: clarify which are independent variables and which are target variables, and determine the logical relationship between them. Online learning mechanism: allow the model to self-adjust to new situations over time.

[0090] Build the initial version of the prediction system and deploy it online. Configure a feedback loop to allow users to submit corrections for continuous optimization of the algorithm. Regularly release update packages to fix known bugs and add new features.

[0091] A closed-loop control system was created, achieving one-stop service from data collection to result display to action suggestions. It provides managers with an intuitive and easy-to-understand operation interface, simplifying daily management workflows. It promotes efficient resource utilization, extends the service life of facilities, and saves long-term operating costs.

[0092] The present application provides a method, device and system for predicting damage of a hyperbolic glass curtain wall, the method comprising the following technical steps:

[0093] Sensor information collection: Strain sensor: used to monitor the stress distribution of the hyperbolic glass curtain wall. Temperature sensor: used to monitor the temperature changes of the curtain wall surface and the surrounding environment. Displacement sensor: used to monitor the deformation of the curtain wall.

[0094] UAV image acquisition: Visible light equipment: carried on the drone, used to collect high-resolution visible light images of the hyperbolic glass curtain wall. Infrared thermal imager: carried on the drone, used to collect thermal imaging images of the curtain wall to detect potential temperature abnormalities.

[0095] Damage feature recognition model based on convolutional neural network (CNN): Input: visible light image and thermal imaging image. Processing: Extract feature information in the image through convolution layer, pooling layer and fully connected layer. Output: Image feature information, including but not limited to damage features such as cracks, scratches, deformation, etc.

[0096] Time series-based physical parameter analysis model: Input: strain sensor information, temperature sensor information, and displacement sensor information. Processing: Use time series analysis methods (such as ARIMA, LSTM, etc.) to model sensor data and extract key physical parameter features. Output: parameter feature information, including stress distribution, temperature change trend, displacement change trend, etc.

[0097] Feature fusion and damage identification: Fusion method: Fusion of image feature information and parameter feature information, which can be achieved by weighted fusion, principal component analysis (PCA) or deep learning methods (such as multimodal fusion network). Output: Damage feature identification results, which comprehensively reflect the health status of the curtain wall.

[0098] Health status prediction model: Input: real-time sensor information, real-time visible light images, and real-time thermal imaging images. Processing: Input the above information into a pre-trained health status prediction model, which can be based on machine learning or deep learning methods, such as support vector machine (SVM), random forest, convolutional neural network (CNN), etc. Output: Real-time output of damage prediction results of the hyperbolic glass curtain wall, including damage type, location, severity, etc.

[0099] The specific implementation methods of the above technical features may be:

[0100] Sensor deployment Install strain sensors, temperature sensors, and displacement sensors on the hyperbolic glass curtain wall and ensure that they work properly. The data collection frequency and accuracy of the sensors need to be set according to actual needs. Select a suitable drone model and equip it with visible light equipment and infrared thermal imagers. Design a flight path to ensure that all areas of the hyperbolic glass curtain wall can be fully covered. Regularly perform drone inspection tasks to collect visible light images and thermal imaging images. Collect a large amount of hyperbolic glass curtain wall damage sample data, including visible light images, thermal imaging images, and sensor data. Use convolutional neural networks (CNNs) to train image data and extract damage features. Use time series analysis methods to train sensor data and extract physical parameter features. Fusion image features and physical parameter features to train damage feature recognition models. Collect historical data of the hyperbolic glass curtain wall, including sensor data, image data, and known damage conditions. Select a suitable machine learning or deep learning algorithm to build a health status prediction model. Train and verify the model to ensure that it has high prediction accuracy. Integrate the sensor data acquisition system, drone image acquisition system, damage feature recognition model, and health status prediction model. Deployed on the cloud or local server to achieve real-time data transmission and processing. Provide a user interface to display real-time damage prediction results and historical data analysis reports.

[0101] The provided method has at least the following beneficial effects:

[0102] Improve damage prediction accuracy: By combining image data and sensor data, the damage characteristics of the hyperbolic glass curtain wall can be more comprehensively captured, improving the accuracy of the prediction.

[0103] Reduce manual intervention: Automated data collection and processing processes reduce reliance on experienced engineers and reduce labor costs.

[0104] Real-time monitoring and early warning: It can monitor the health status of the curtain wall in real time, promptly discover potential damage risks, take measures in advance for maintenance, and avoid major accidents.

[0105] Optimize design solutions: Through the analysis of large amounts of data, we can provide references for the design of hyperbolic glass curtain walls, help optimize design solutions, and improve the overall performance of buildings.

[0106] Reduce costs: By detecting and repairing damage early, the service life of the curtain wall can be effectively extended and long-term maintenance costs can be reduced.

[0107] In summary, the hyperbolic glass curtain wall damage prediction method provided in this application realizes efficient and accurate prediction of curtain wall damage by comprehensively utilizing sensor data and image processing technology, providing strong support for the safety and maintenance of buildings.

[0108] In some embodiments, before the visible light image and thermal imaging image are input into a damage feature recognition model based on a convolutional neural network, the method further includes: preprocessing the sensor acquisition information, visible light image and thermal imaging image; the preprocessing includes at least filtering, denoising, enhancement, and cropping preprocessing; marking damage information in the preprocessed sensor acquisition information, visible light image and thermal imaging image; the damage information includes at least damage type and timing abnormality.

[0109] Preprocessing includes: Filtering: Use low-pass filters to remove high-frequency noise and ensure data smoothness. Denoising: Use median filtering, Gaussian filtering and other methods to remove random noise in the image. Enhancement: Improve the clarity of the image through contrast enhancement, brightness adjustment and other technologies. Cropping: Crop the image as needed to remove irrelevant background and retain key areas.

[0110] The damage information is labeled as follows: Annotation tools: Use professional image annotation tools (such as Labelbox, VGG ImageAnnotator) for annotation.

[0111] The annotation contents include: Damage type: including cracks, deformation, corrosion, etc. Time series abnormality: record the changes of damage over time, such as the speed of crack expansion, changes in deformation degree, etc.

[0112] For example, normalize the sensor data to ensure that the data from different sensors are on the same scale. Visible light images: filter, denoise, enhance, and crop the images, and then use annotation tools to mark the location and type of damage. Thermal imaging images: also filter, denoise, enhance, and crop, and mark the temperature anomaly areas and their types.

[0113] The preprocessing step can significantly reduce noise interference and improve the reliability and consistency of data. High-quality data input helps improve the training effect of convolutional neural networks, thereby improving the accuracy of damage identification. Detailed damage annotation provides rich supervision information for the model, which helps the model better learn and understand damage characteristics. The annotated information can be used for subsequent analysis and verification to ensure the reliability of the model output results.

[0114] In some embodiments, the health status prediction model outputs in real time a damage prediction result corresponding to the hyperbolic glass curtain wall, including: the health status prediction model outputs a crack index and a deformation index corresponding to the hyperbolic glass curtain wall according to the real-time visible light image and the real-time thermal imaging image; the health status prediction model outputs a temperature influence factor and a stress distribution factor according to the real-time sensor acquisition information; and the damage prediction result is generated according to the crack index, deformation index, temperature influence factor and stress distribution factor.

[0115] Based on real-time visible light images and thermal imaging images, the crack features are extracted through convolutional neural networks, and the crack index is calculated. Based on real-time visible light images and thermal imaging images, the deformation features are extracted through convolutional neural networks, and the deformation index is calculated.

[0116] Based on real-time temperature sensor data, the influence factor of ambient temperature on the curtain wall is calculated. Based on real-time strain sensor data, the stress distribution of the curtain wall is calculated. The health score index is calculated based on the crack index, deformation index, temperature influence factor and stress distribution factor. A comprehensive evaluation is conducted by combining multiple factors (cracks, deformation, temperature, stress), which improves the comprehensiveness and accuracy of damage prediction. Through the dynamic weight coefficient, the changes of different types of damage over time are considered, and the time adaptability of the model is enhanced. The introduction of environmental and geometric coupling coefficients considers the impact of seasonal changes and hyperbolic geometric characteristics on curtain wall damage, making the prediction more scientific and reasonable. It can output the health score index in real time, detect potential problems in time and issue early warnings, and provide strong support for maintenance decisions.

[0117] Exemplarily, generating the damage prediction result according to the crack index, deformation index, temperature influence factor and stress distribution factor includes: calculating a health score index according to the crack index, deformation index, temperature influence factor and stress distribution factor; generating the damage prediction result according to the health score index; the expression of the health score index includes:

[0118] ;in, for The health score index at all times, for Moment Dynamic weight coefficient of the damage type, for Moment Injury type score, is the number of damage types, which at least include cracks and deformations, is the environmental and geometric coupling coefficient corresponding to the hyperbolic glass curtain wall, which is used to reflect the accelerating effect of seasonal changes on curtain wall damage. is the curvature change correction term, which is used to quantify the additional impact of hyperbolic surface geometric deformation on the health score.

[0119] By combining multiple factors for comprehensive evaluation, the comprehensiveness and accuracy of damage prediction are improved. Through the dynamic weight coefficient, the changes of different types of damage over time are taken into account, and the time adaptability of the model is enhanced. The introduction of environmental and geometric coupling coefficients takes into account the impact of seasonal changes and hyperbolic geometric characteristics on curtain wall damage, making the prediction more scientific and reasonable. It can output the health score index in real time, detect potential problems in time and issue early warnings, and provide strong support for maintenance decisions.

[0120] It should be noted that, in some embodiments, the expression of the dynamic weight coefficient includes:

[0121] ;

[0122] in, for Moment Dynamic weight coefficient of the damage type, is the initial weight coefficient, For the The damage sensitivity attenuation factor of the damage index is used to control the timeliness of different damage types. For the Damage sensitivity attenuation factor of the damage index, is the number of damage types.

[0123] Through the dynamic weight coefficient, the weight can be automatically adjusted according to the changes of different damage types over time, making the model more flexible and adaptive.

[0124] The damage sensitivity attenuation factor controls the timeliness of different damage types, allowing the model to more accurately reflect the actual situation. By setting the initial weight coefficient and attenuation factor, the weight balance between different types of damage is guaranteed, improving the rationality of the overall evaluation.

[0125] It should be noted that, in some embodiments, if the damage type is crack, the initial weight coefficient is 0.4, and the corresponding value range of the damage sensitivity attenuation factor is 0.05 to 0.15; if the damage type is deformation, the initial weight coefficient is 0.3, and the corresponding value range of the damage sensitivity attenuation factor is 0.03 to 0.18.

[0126] Through specific parameter settings, the weights and timeliness of different types of damage can be controlled more accurately, improving the accuracy and robustness of the model. The setting of parameter ranges allows fine-tuning according to actual conditions, enhancing the flexibility and adaptability of the model. Reasonable parameter settings ensure the stability and reliability of the model under different time and environmental conditions.

[0127] See also Figure 2 As shown, Figure 2 1 is a schematic diagram of the structure of a hyperbolic glass curtain wall damage prediction device 200 provided in an embodiment of the present application. The hyperbolic glass curtain wall damage prediction device 200 is used to execute the steps of the hyperbolic glass curtain wall damage prediction method shown in the above embodiments. The hyperbolic glass curtain wall damage prediction device 200 can be a single server or a server cluster, or the hyperbolic glass curtain wall damage prediction device 200 can be a terminal, which can be a handheld terminal, a laptop computer, a wearable device or a robot, etc.

[0128] like Figure 2 As shown, the hyperbolic glass curtain wall damage prediction device 200 includes:

[0129] The information acquisition module 201 is used to acquire sensor acquisition information of the hyperbolic glass curtain wall; the sensor acquisition information at least includes strain sensor information, temperature sensor information and displacement sensor information;

[0130] An image acquisition module 202 is used to control a preset drone to collect visible light images and thermal imaging images corresponding to the hyperbolic glass curtain wall; the drone is equipped with a visible light device and an infrared thermal imager;

[0131] An image output module 203, used to input the visible light image and the thermal imaging image into a damage feature recognition model based on a convolutional neural network, and output image feature information;

[0132] A feature output module 204 is used to input the sensor acquisition information into a time series-based physical parameter analysis model and output parameter feature information;

[0133] A feature fusion module 205 is used to fuse the image feature information with the parameter feature information to obtain a damage feature recognition result;

[0134] The prediction completion module 206 is used to build a health status prediction model according to the damage feature identification, so as to input the real-time sensor collection information, real-time visible light image and real-time thermal imaging image corresponding to the hyperbolic glass curtain wall into the health status prediction model, and the health status prediction model outputs the damage prediction result corresponding to the hyperbolic glass curtain wall in real time.

[0135] In some embodiments, before the visible light image and thermal imaging image are input into a damage feature recognition model based on a convolutional neural network, the method further includes: preprocessing the sensor acquisition information, visible light image and thermal imaging image; the preprocessing includes at least filtering, denoising, enhancement, and cropping preprocessing; marking damage information in the preprocessed sensor acquisition information, visible light image and thermal imaging image; the damage information includes at least damage type and timing abnormality.

[0136] In some embodiments, the health status prediction model outputs in real time a damage prediction result corresponding to the hyperbolic glass curtain wall, including: the health status prediction model outputs a crack index and a deformation index corresponding to the hyperbolic glass curtain wall according to the real-time visible light image and the real-time thermal imaging image; the health status prediction model outputs a temperature influence factor and a stress distribution factor according to the real-time sensor acquisition information; and the damage prediction result is generated according to the crack index, deformation index, temperature influence factor and stress distribution factor.

[0137] Exemplarily, generating the damage prediction result according to the crack index, deformation index, temperature influence factor and stress distribution factor includes: calculating a health score index according to the crack index, deformation index, temperature influence factor and stress distribution factor; generating the damage prediction result according to the health score index; the expression of the health score index includes:

[0138] ;in, for The health score index at all times, for Moment Dynamic weight coefficient of the damage type, for Moment Injury type score, is the number of damage types, which at least include cracks and deformations, is the environmental and geometric coupling coefficient corresponding to the hyperbolic glass curtain wall, which is used to reflect the accelerating effect of seasonal changes on curtain wall damage. is the curvature change correction term, which is used to quantify the additional impact of hyperbolic surface geometric deformation on the health score.

[0139] It should be noted that, in some embodiments, the expression of the dynamic weight coefficient includes:

[0140] ;

[0141] in, for Moment Dynamic weight coefficient of the damage type, is the initial weight coefficient, For the The damage sensitivity attenuation factor of the damage index is used to control the timeliness of different damage types. For the Damage sensitivity attenuation factor of the damage index, is the number of damage types.

[0142] It should be noted that, in some embodiments, if the damage type is crack, the initial weight coefficient is 0.4, and the corresponding value range of the damage sensitivity attenuation factor is 0.05 to 0.15; if the damage type is deformation, the initial weight coefficient is 0.3, and the corresponding value range of the damage sensitivity attenuation factor is 0.03 to 0.18.

[0143] It should be noted that those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working processes of the hyperbolic glass curtain wall damage prediction device and each module described above can refer to the corresponding processes in the embodiments of the hyperbolic glass curtain wall damage prediction method described in the above embodiments, and will not be repeated here.

[0144] The above-mentioned double-curved glass curtain wall damage prediction method can be implemented in the form of a computer program. The computer program can be used in Figure 2 Run on the device shown.

[0145] The present application provides a hyperbolic glass curtain wall damage prediction system, comprising a plurality of sensors, wherein the sensors are arranged at target points of the hyperbolic glass curtain wall, and the sensors include at least a strain sensor, a temperature sensor, and a displacement sensor; an unmanned aerial vehicle, wherein the unmanned aerial vehicle is equipped with a visible light device and an infrared thermal imager; a controller, wherein the controller includes a memory and a processor; the memory is used to store a computer program; and the processor is used to execute the computer program and implement the hyperbolic glass curtain wall damage prediction method provided in any embodiment of the present application when executing the computer program.

[0146] Types of sensors: strain sensors, temperature sensors and displacement sensors. Layout: Sensors are installed at key points of the hyperbolic glass curtain wall to ensure coverage of all important areas. For example, sensors are placed at the four corners, center points and joints that may be subject to high stress. Function: Strain sensor: monitors stress changes in the curtain wall structure and reflects material deformation. Temperature sensor: measures ambient temperature and temperature distribution on the curtain wall surface, which helps identify potential risks caused by temperature changes. Displacement sensor: monitors the relative position changes between curtain wall components to help detect structural movement or deformation problems.

[0147] Configuration of the drone: Equipped with visible light equipment (such as high-definition cameras) and infrared thermal imagers. Function: Visible light image acquisition: Take high-resolution visible light images of the hyperbolic glass curtain wall for visual inspection. Thermal imaging image acquisition: Use infrared thermal imagers to obtain thermal imaging images of the curtain wall, detect temperature abnormalities, and reveal hidden problems inside. Flight path planning: Pre-set a detailed flight route map to ensure that the drone can take comprehensive photos around the outside of the building according to the predetermined trajectory. Data transmission: The data collected by the drone is transmitted to the controller in real time through a wireless communication module (such as Wi-Fi, 4G / 5G).

[0148] Hardware of the controller: Memory: used to store computer programs, sensor data, image data collected by drones, and model parameters. Processor: high-performance processor used to execute computer programs and perform data processing and analysis. Software: Operating system: operating system that supports multi-tasking, such as Linux. Application: including data acquisition module, image processing module, feature extraction module, health status prediction module, etc. Database: stores historical data and data sets required for model training. Function: Data acquisition and storage: collect data from sensors and drones and store them in memory. Data preprocessing: perform preprocessing steps such as filtering, denoising, enhancement, and cropping on sensor data and image data. Feature extraction: use convolutional neural networks to extract damage features from visible light images and thermal imaging images, and use time series models to extract physical parameter features from sensor data. Feature fusion: fuse image feature information and physical parameter features to generate a comprehensive feature vector. Health status prediction: based on the fused feature vector, use the health status prediction model to output the damage prediction results of the hyperbolic glass curtain wall in real time, including crack index, deformation index, temperature influence factor, and stress distribution factor. Health score calculation: The health score index is calculated based on the crack index, deformation index, temperature influence factor and stress distribution factor, and the final damage prediction result is generated. Alarm and notification: When an abnormal situation is detected, the system automatically sends an alarm notification to the maintenance personnel so that timely measures can be taken.

[0149] The specific implementation of the system can be: Sensor installation and wiring: Install strain sensors, temperature sensors and displacement sensors at key points of the hyperbolic glass curtain wall. Transmit sensor data to the controller in real time through wireless communication modules (such as LoRa, NB-IoT).

[0150] Drone configuration and flight path planning: Configure the drone and ensure it is equipped with a high-definition camera and infrared thermal imager. Use ground station software to design a detailed flight route map to ensure that the drone can fully cover all areas of the curtain wall. Set the drone to perform tasks at a scheduled time, such as automatically taking off for inspections at a specific time every day.

[0151] Controller setup and software deployment: Configure the controller hardware, including high-performance processors and large-capacity memory. Install the operating system and related applications, and configure the database. Develop data acquisition modules, image processing modules, feature extraction modules, feature fusion modules, and health status prediction modules. Integrate all modules into a unified application to ensure smooth data flow between modules.

[0152] Data acquisition and processing: The sensor collects strain, temperature and displacement data in real time and transmits them to the controller through the wireless communication module. The drone regularly performs inspection tasks, collects visible light images and thermal imaging images, and transmits them to the controller through the wireless communication module. The collected data is pre-processed, including filtering, denoising, enhancement and cropping.

[0153] Feature extraction and fusion: Use convolutional neural networks to extract damage features from visible light images and thermal imaging images. Use time series models (such as LSTM) to extract physical parameter features from sensor data. Fusion of image feature information and physical parameter features to generate a comprehensive feature vector.

[0154] Health status prediction and score calculation: Based on the fused feature vector, the health status prediction model is used to output the crack index, deformation index, temperature influence factor and stress distribution factor of the hyperbolic glass curtain wall in real time. The health score index is calculated based on these factors, and the final damage prediction result is generated. When an abnormal situation is detected, the system automatically sends an alarm notification to the maintenance personnel.

[0155] System maintenance and updates: Regularly calibrate sensors to ensure data accuracy. Update the flight path planning of drones to adapt to changes in buildings. Regularly update the training set and introduce the latest observation data to keep the model up to date. Release update packages to fix known bugs and add new features.

[0156] The system has at least the following beneficial effects:

[0157] Comprehensive monitoring: Through the combination of multiple sensors and drones, comprehensive monitoring of the hyperbolic glass curtain wall is achieved, especially for the parts that are difficult to directly observe. It provides rich physical parameter data to support subsequent analysis and improve the reliability and accuracy of damage detection.

[0158] Real-time monitoring and early warning: The real-time monitoring of the state of the hyperbolic glass curtain wall is realized, and early minor damage can be discovered in time. Through the health status prediction model, potential safety risks can be predicted in advance, providing a strong basis for maintenance decisions.

[0159] Accurate assessment: By comprehensively considering multiple factors (cracks, deformation, temperature, stress), a more accurate damage assessment is provided. The introduction of dynamic weight coefficients and environmental and geometric coupling coefficients makes the assessment results more scientific and reasonable.

[0160] Automation and intelligence: The system is highly automated, reducing manual intervention, operational difficulty and safety risks. The algorithm performance is continuously improved through machine learning, and the accuracy and robustness are continuously improved.

[0161] Efficient use of resources: Through real-time monitoring and early warning, problems can be discovered and handled in a timely manner, extending the service life of facilities, saving long-term operating costs and improving resource utilization efficiency.

[0162] Easy to manage: It provides managers with an intuitive and easy-to-understand operation interface, simplifying daily management workflow. Through a closed-loop control system, it achieves a one-stop service from data collection to result display to action suggestions.

[0163] In summary, the hyperbolic glass curtain wall damage prediction system realizes comprehensive, real-time and accurate monitoring and evaluation of the hyperbolic glass curtain wall through multi-sensor information fusion and UAV image acquisition, combined with deep learning models for feature extraction and fusion, providing strong support for maintenance decisions.

[0164] In some embodiments, the target points include glass panel edges, connection nodes, and stress concentration areas corresponding to the hyperbolic glass curtain wall.

[0165] Target point selection: Glass plate edge: Install sensors at the edge of each glass of the hyperbolic glass curtain wall. These locations are usually stress concentration areas and are prone to cracks and deformation. Connection nodes: Install sensors at each connection node of the curtain wall. Connection nodes are key parts of the structure and are subject to large stress and displacement changes. Stress concentration areas: Based on design and engineering analysis, determine the stress concentration areas on the curtain wall and arrange sensors in these areas. For example, areas with large curvature changes and uneven stress.

[0166] Sensor types and functions: Strain sensors: monitor stress changes at the edges of glass panels, connection nodes and stress concentration areas. Temperature sensors: monitor temperature distribution in these key areas, especially in seasons or time periods with large temperature changes. Displacement sensors: monitor relative displacement changes in these areas, especially at connection nodes.

[0167] Conduct a comprehensive on-site survey of the hyperbolic glass curtain wall to determine the specific locations of all target points. Install corresponding sensors at each target point and ensure that they are firm and reliable. Transmit sensor data to the controller in real time through wireless communication modules (such as LoRa, NB-IoT). Perform initial calibration of the installed sensors to ensure data accuracy.

[0168] By focusing on the edges of glass panels, connection nodes and stress concentration areas, we have achieved comprehensive coverage of the key parts of the hyperbolic glass curtain wall. These areas are usually where damage is most likely to occur. Focused monitoring can detect minor damage earlier and improve detection accuracy. Timely discovery and treatment of problems in these key areas can effectively prevent major accidents and improve the safety of buildings. Through focused monitoring, maintenance plans can be formulated more targeted, unnecessary inspections and repairs can be reduced, and maintenance efficiency can be improved.

[0169] In some embodiments, the arrangement density of the sensors in the hyperbolic glass curtain wall is dynamically adjusted according to the curvature radius of the hyperbolic glass curtain wall. In the area where the curvature radius is less than 3m, the arrangement density is increased by a preset density coefficient compared to the area where the curvature radius is greater than or equal to 3m. The preset density coefficient corresponds to a value range of 30% to 50%.

[0170] According to the design drawings of the hyperbolic glass curtain wall, the curtain wall is divided into areas with different curvature radii. The area with a curvature radius less than 3m is defined as a high curvature area, and the area with a curvature radius greater than or equal to 3m is defined as a low curvature area. Sensor layout density adjustment: High curvature area: In areas with a curvature radius less than 3m, increase the layout density of sensors. Specifically, the layout density increases the preset density coefficient compared to the low curvature area, and the value range of the preset density coefficient is 30% to 50%. Low curvature area: In areas with a curvature radius greater than or equal to 3m, maintain the standard sensor layout density.

[0171] The specific steps corresponding to the embodiment may be: Curvature analysis: using computer-aided design (CAD) software or three-dimensional scanning technology, a detailed curvature analysis is performed on the hyperbolic glass curtain wall to determine the curvature radius of each area.

[0172] Density calculation: Based on the division results of the curvature radius, the sensor layout density in the high curvature area and the low curvature area is calculated.

[0173] Sensor installation: Arrange sensors at increased density in areas of high curvature and at standard density in areas of low curvature.

[0174] Data transmission and calibration: The sensor data is transmitted to the controller in real time through the wireless communication module, and the installed sensors are initially calibrated.

[0175] Increasing the density of sensor layout in high curvature areas can more carefully monitor the state changes in these areas and improve monitoring accuracy. By dynamically adjusting the density of sensor layout, it is possible to reasonably allocate sensor resources and avoid waste while ensuring the monitoring effect. For the complex geometric shapes of hyperboloids, the division of curvature radius and density adjustment can better meet the monitoring needs of different areas. High curvature areas are usually more prone to problems such as stress concentration and material fatigue. Increasing the density of sensors can detect these problems earlier and improve the reliability of the system. Through reasonable sensor layout, excessive layout of sensors in low curvature areas is avoided, thereby reducing the overall cost of the system.

[0176] See also Figure 3 , Figure 3 : is a schematic block diagram of the structure of a controller provided in an embodiment of the present application. The controller includes a processor, a memory and a network interface connected via a device bus, wherein the memory may include a storage medium and an internal memory.

[0177] The storage medium can store an operating device and a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any method for predicting damage of a hyperbolic glass curtain wall.

[0178] The processor is used to provide computing and control capabilities and support the operation of the entire controller.

[0179] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any one of the hyperbolic glass curtain wall damage prediction methods.

[0180] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the terminal to which the scheme of the present application is applied. The specific controller may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0181] It should be understood that the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0182] In one embodiment, the processor is used to run a computer program stored in the memory to implement the following steps:

[0183] Acquire sensor collection information of the hyperbolic glass curtain wall; the sensor collection information at least includes strain sensor information, temperature sensor information and displacement sensor information;

[0184] Controlling a preset drone to collect visible light images and thermal imaging images corresponding to the hyperbolic glass curtain wall; the drone is equipped with a visible light device and an infrared thermal imager;

[0185] Inputting the visible light image and the thermal imaging image into a damage feature recognition model based on a convolutional neural network, and outputting image feature information;

[0186] Inputting the sensor collected information into a time series-based physical parameter analysis model and outputting parameter characteristic information;

[0187] The image feature information is integrated with the parameter feature information to obtain a damage feature recognition result;

[0188] A health status prediction model is constructed according to the damage feature identification, so that real-time sensor collection information, real-time visible light images and real-time thermal imaging images corresponding to the hyperbolic glass curtain wall are input into the health status prediction model, and the health status prediction model outputs the damage prediction result corresponding to the hyperbolic glass curtain wall in real time.

[0189] In some embodiments, before the visible light image and thermal imaging image are input into a damage feature recognition model based on a convolutional neural network, the method further includes: preprocessing the sensor acquisition information, visible light image and thermal imaging image; the preprocessing includes at least filtering, denoising, enhancement, and cropping preprocessing; marking damage information in the preprocessed sensor acquisition information, visible light image and thermal imaging image; the damage information includes at least damage type and timing abnormality.

[0190] In some embodiments, the health status prediction model outputs in real time a damage prediction result corresponding to the hyperbolic glass curtain wall, including: the health status prediction model outputs a crack index and a deformation index corresponding to the hyperbolic glass curtain wall according to the real-time visible light image and the real-time thermal imaging image; the health status prediction model outputs a temperature influence factor and a stress distribution factor according to the real-time sensor acquisition information; and the damage prediction result is generated according to the crack index, deformation index, temperature influence factor and stress distribution factor.

[0191] Exemplarily, generating the damage prediction result according to the crack index, deformation index, temperature influence factor and stress distribution factor includes: calculating a health score index according to the crack index, deformation index, temperature influence factor and stress distribution factor; generating the damage prediction result according to the health score index; the expression of the health score index includes:

[0192] ;in, for The health score index at all times, for Moment Dynamic weight coefficient of the damage type, for Moment Injury type score, is the number of damage types, which at least include cracks and deformations, is the environmental and geometric coupling coefficient corresponding to the hyperbolic glass curtain wall, which is used to reflect the accelerating effect of seasonal changes on curtain wall damage. is the curvature change correction term, which is used to quantify the additional impact of hyperbolic surface geometric deformation on the health score.

[0193] It should be noted that, in some embodiments, the expression of the dynamic weight coefficient includes:

[0194] ;

[0195] in, for Moment Dynamic weight coefficient of the damage type, is the initial weight coefficient, For the The damage sensitivity attenuation factor of the damage index is used to control the timeliness of different damage types. For the Damage sensitivity attenuation factor of the damage index, is the number of damage types.

[0196] It should be noted that, in some embodiments, if the damage type is crack, the initial weight coefficient is 0.4, and the corresponding value range of the damage sensitivity attenuation factor is 0.05 to 0.15; if the damage type is deformation, the initial weight coefficient is 0.3, and the corresponding value range of the damage sensitivity attenuation factor is 0.03 to 0.18.

[0197] It should be noted that technical personnel in the relevant field can clearly understand that, for the convenience and conciseness of description, the specific working processes of the controller and each module described above can refer to the corresponding processes in the method embodiments described in the above embodiments, and will not be repeated here.

[0198] The present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the processor implements the steps of the hyperbolic glass curtain wall damage prediction method provided in any embodiment of the present application.

[0199] The computer-readable storage medium may be an internal storage unit of the controller described in the foregoing embodiment, such as a hard disk or memory of the controller. The computer-readable storage medium may also be an external storage device of the controller, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc., equipped on the controller.

[0200] It should be noted that technical personnel in the relevant field can clearly understand that, for the convenience and conciseness of description, the specific working processes of the storage medium and each module described above can refer to the corresponding processes in the method embodiments described in the above embodiments, and will not be repeated here.

[0201] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for predicting damage of a hyperbolic glass curtain wall, characterized in that: include: Acquire sensor collection information of the hyperbolic glass curtain wall; the sensor collection information at least includes strain sensor information, temperature sensor information and displacement sensor information; Controlling a preset drone to collect visible light images and thermal imaging images corresponding to the hyperbolic glass curtain wall; the drone is equipped with a visible light device and an infrared thermal imager; Inputting the visible light image and the thermal imaging image into a damage feature recognition model based on a convolutional neural network, and outputting image feature information; Inputting the sensor collected information into a time series-based physical parameter analysis model and outputting parameter characteristic information; The image feature information is integrated with the parameter feature information to obtain a damage feature recognition result; A health status prediction model is constructed according to the damage feature identification, so that real-time sensor collection information, real-time visible light images and real-time thermal imaging images corresponding to the hyperbolic glass curtain wall are input into the health status prediction model, and the health status prediction model outputs the damage prediction result corresponding to the hyperbolic glass curtain wall in real time.

2. The method according to claim 1, characterized in that: Before inputting the visible light image and the thermal imaging image into the damage feature recognition model based on the convolutional neural network, the method further includes: Preprocessing the sensor collected information, visible light image and thermal imaging image; the preprocessing at least includes filtering, denoising, enhancing and cropping preprocessing; The damage information is marked in the pre-processed sensor acquisition information, visible light image and thermal imaging image; the damage information at least includes the damage type and timing abnormality.

3. The method according to claim 1, characterized in that The health status prediction model outputs the damage prediction result corresponding to the hyperbolic glass curtain wall in real time, including: The health status prediction model outputs the crack index and deformation index corresponding to the hyperbolic glass curtain wall according to the real-time visible light image and the real-time thermal imaging image; the health status prediction model outputs the temperature influence factor and the stress distribution factor according to the real-time sensor collection information; The damage prediction result is generated according to the crack index, deformation index, temperature influence factor and stress distribution factor.

4. The method according to claim 3, characterized in that Generating the damage prediction result according to the crack index, deformation index, temperature influence factor and stress distribution factor includes: Calculating a health score index according to the crack index, deformation index, temperature influence factor and stress distribution factor; The injury prediction result is generated according to the health score index; the expression of the health score index includes: ； in, for The health score index at all times, for Moment Dynamic weight coefficient of the damage type, for Moment Injury type score, is the number of damage types, which at least include cracks and deformations, is the environmental and geometric coupling coefficient corresponding to the hyperbolic glass curtain wall, which is used to reflect the accelerating effect of seasonal changes on curtain wall damage. is the curvature change correction term, which is used to quantify the additional impact of hyperbolic surface geometric deformation on the health score.

5. The method according to claim 4, characterized in that The expression of the dynamic weight coefficient includes: ； in, for Moment Dynamic weight coefficient of the damage type, is the initial weight coefficient, For the The damage sensitivity attenuation factor of the damage index is used to control the timeliness of different damage types. For the Damage sensitivity attenuation factor of the damage index, is the number of damage types.

6. The method according to claim 5, characterized in that If the damage type is a crack, the initial weight coefficient is 0.4, and the corresponding value range of the damage sensitivity attenuation factor is 0.05 to 0.15; If the damage type is deformation, the initial weight coefficient is 0.3, and the corresponding value range of the damage sensitivity attenuation factor is 0.03 to 0.

18.

7. A hyperbolic glass curtain wall damage prediction device, characterized in that: include: An information acquisition module, used to acquire sensor acquisition information of the hyperbolic glass curtain wall; the sensor acquisition information at least includes strain sensor information, temperature sensor information and displacement sensor information; An image acquisition module, used to control a preset drone to collect visible light images and thermal imaging images corresponding to the hyperbolic glass curtain wall; the drone is equipped with a visible light device and an infrared thermal imager; An image output module, used for inputting the visible light image and the thermal imaging image into a damage feature recognition model based on a convolutional neural network, and outputting image feature information; A feature output module, used to input the sensor collected information into a time series-based physical parameter analysis model and output parameter feature information; A feature fusion module, used to fuse the image feature information with the parameter feature information to obtain a damage feature recognition result; A prediction completion module is used to build a health status prediction model according to the damage feature identification, so as to input the real-time sensor collection information, real-time visible light image and real-time thermal imaging image corresponding to the hyperbolic glass curtain wall into the health status prediction model, and the health status prediction model outputs the damage prediction result corresponding to the hyperbolic glass curtain wall in real time.

8. A hyperbolic glass curtain wall damage prediction system, characterized in that: include A plurality of sensors, wherein the sensors are arranged at target points of the hyperbolic glass curtain wall, and the sensors include at least a strain sensor, a temperature sensor and a displacement sensor; Unmanned aerial vehicles, without any visible light device or infrared thermal imager installed on board; A controller comprising a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and implement the method according to any one of claims 1 to 6 when executing the computer program.

9. The hyperbolic glass curtain wall damage prediction system according to claim 8, characterized in that: The target points include glass panel edges, connection nodes and stress concentration areas corresponding to the hyperbolic glass curtain wall.

10. The hyperbolic glass curtain wall damage prediction system according to claim 8, characterized in that: The arrangement density of the sensors in the hyperbolic glass curtain wall is dynamically adjusted according to the curvature radius of the hyperbolic glass curtain wall. In the area where the curvature radius is less than 3m, the arrangement density is increased by a preset density coefficient compared with the area where the curvature radius is greater than or equal to 3m. The value range corresponding to the preset density coefficient is 30% to 50%.

Images