A glass curtain wall whole life cycle management method and system

Abstract

This invention belongs to the field of glass curtain wall maintenance technology. It discloses a method and system for full lifecycle management of glass curtain walls, including the following modules: a data acquisition module for collecting the status parameters of each glass panel in the glass curtain wall. This invention sets up a remaining lifespan prediction model to assign predicted lifecycle values ​​to all glass panels, and visualizes them in a three-dimensional digital twin model as a lifespan topographic map. Spatial statistical methods are used to identify clusters with significantly low remaining lifespans, i.e., lifespan depressions. By performing synergistic analysis and causal inference on the lifespan curves of the glass panels within these depressions, common causes leading to cluster performance degradation can be intelligently diagnosed. Based on this diagnosis, a unified investigation and collaborative maintenance plan for the entire depression can be generated, thereby achieving a systematic improvement from treating individual defects to enhancing regional performance. This avoids repetitive inspections and fragmented repairs, saving significant manpower and resources.

Description

Technical Field

[0001] This invention relates to the field of glass curtain wall maintenance technology, and more specifically, to a method and system for the full life cycle management of glass curtain walls. Background Technology

[0002] In modern urban architecture, glass curtain walls have become the mainstream external envelope structure for high-rise and super high-rise buildings due to their excellent lighting, light aesthetics, and rapid construction. As the frontier interface for interaction between buildings and the environment, glass curtain walls are subjected to the coupled effects of multiple environmental factors such as wind loads, temperature changes, ultraviolet radiation, and rainwater erosion over a long period of time. Their materials will inevitably experience performance degradation, leading to defects such as structural adhesive failure, glass breakage, aging and leakage of sealant, and fatigue damage of hardware.

[0003] Currently, the operation and maintenance management of glass curtain walls mainly relies on traditional regular manual inspections and post-event emergency repairs. Current detection methods are incomplete and outdated, failing to achieve real-time perception and quantitative assessment of structural health status, resulting in untimely detection of safety hazards. Maintenance plans are extensive and uneconomical, typically employing fixed time cycles for comprehensive overhauls, failing to differentiate and precisely schedule maintenance based on the actual health condition of each glass panel and the severity of its environment. This leads to under-maintenance in some areas and over-maintenance in others. Emergency response and resource allocation are inefficient; when local risks are detected, it is difficult to quickly assess their impact on the overall curtain wall system, and it is impossible to plan the optimal emergency response path and solution based on the real-time status and geographical location of maintenance resources. Currently, systems only monitor wind pressure or temperature, and some only use BIM models for visualization, failing to deeply integrate multi-source heterogeneous monitoring data, and even more so, failing to utilize artificial intelligence technology to learn the inherent laws of curtain wall aging from historical data to accurately predict its future health status and remaining lifespan. Summary of the Invention

[0004] To address the problems in the background art, this invention proposes a method and system for the full life cycle management of glass curtain walls.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a glass curtain wall full life cycle management system, comprising the following modules: The data acquisition module is used to collect the status parameters of each glass panel in the glass curtain wall; The health assessment module is used to generate a health score for the glass plate based on the collected status parameters, and to calculate the rate of change of the health score of the glass plate based on the health score. The maintenance path planning module is used to classify glass panels according to their health scores and health score change rates, and generate initial maintenance paths on the glass curtain wall based on the glass panel's level. The aging rate prediction module is used to acquire historical aging rate data and build an aging rate prediction model. The aging rate prediction model is used to predict the aging rate of the glass plate, the aging time is obtained based on the predicted aging rate, and the glass plate is marked as a glass plate that needs maintenance based on the aging time. The maintenance path optimization module is used to optimize the initial maintenance path based on the glass panel to be maintained, thereby obtaining an optimized maintenance path. The remaining life prediction module is used to acquire historical remaining life data and build a remaining life prediction model. The remaining life prediction model is used to predict the remaining life of the glass panels. Based on the predicted life, the life gaps of the glass curtain wall are identified, and unified investigation and collaborative maintenance are carried out on the glass panels in the life gaps.

[0006] Furthermore, the glass curtain wall paved on the exterior of the building is composed of multiple glass panel units spliced ​​together. The state parameters refer to the attenuation rate of the structural adhesive bonding strength of the glass panel, the aging index of the sealant, the propagation rate of microcracks on the glass surface, the cumulative fatigue damage index of the hardware, and the frame deformation compatibility index. The structural adhesive internal defects are detected by ultrasonic flaw detector, and the surface image of the adhesive joint is captured by high-definition camera. Deep learning algorithm is used to identify surface deterioration characteristics such as cracking and powdering. The structural adhesive adhesive strength decay rate is obtained by combining internal and surface data to calculate the percentage decay rate of the current adhesive strength relative to the design strength. The structural adhesive bond strength attenuation rate is determined by scoring internal defects detected using a fusion ultrasonic flaw detector. Apparent degradation scoring using deep learning in conjunction with high-definition cameras The internal defect score is derived from the overall assessment. The results are obtained based on the size, quantity, and distribution density of internal defects such as bubbles and debonding detected by ultrasonic testing. The surface image of the adhesive joint is semantically segmented and classified using a deep learning model to identify surface deterioration features such as cracking, chalking, and discoloration.

[0007] in, These are weighting coefficients, obtained through training based on historical data; The attenuation rate of structural adhesive bond strength was normalized.

[0008] In the formula, This represents the normalized structural adhesive bond strength attenuation rate. At the current intensity, For initial design strength; Infrared spectroscopy was used to analyze the chemical bonds at the sampling points of the sealant. At the same time, a Shore hardness tester was used to measure the hardness change of the colloid. The chemical aging degree and physical hardening degree were combined to calculate the sealant aging index. The sealant aging index is a measure of chemical aging as determined by infrared spectroscopy. Physical hardening degree measured by Shore hardness tester In summary, The degradation rate of chemical bonds relative to virgin adhesive was calculated by analyzing the changes in the intensity of the infrared absorption peaks of the characteristic chemical bonds in the sealant. The hardening ratio relative to the initial hardness is calculated by measuring the change in the hardness of the sealant.

[0009] in, These are weighting coefficients, obtained through training based on historical data; The aging index of the sealant was normalized:

[0010] In the formula, To normalize the aging index of the sealant, As a measure of aging, The degree of failure and aging; A laser speckle interferometer is used to perform non-contact scanning of the glass surface to acquire micron-level deformation field images. Through deep learning image recognition, the average length expansion of microcracks per unit time is calculated to obtain the microcrack propagation rate on the glass surface. The microcrack propagation rate on the glass surface is then normalized.

[0011] In the formula, To normalize the microcrack propagation rate on the glass surface, This represents the current expansion rate; Strain gauges were attached to key hardware components such as sash hinges and sliding supports to monitor their stress time history over a long period. The number of cycles at different stress amplitudes was counted using the rainflow counting method. Then, the total fatigue damage degree was calculated according to the Mainner linear cumulative damage rule to obtain the fatigue damage accumulation index of the hardware components. The fatigue damage accumulation index of the hardware components was then normalized.

[0012] In the formula, This is the normalized cumulative fatigue damage index for hardware parts. This represents the current cumulative damage level. Distributed fiber optic sensors are attached along the vertical keel and horizontal trusses of the curtain wall to monitor their strain distribution under wind load and temperature changes. The measured strain data are input into a finite element model for inversion analysis to calculate the deviation between the actual deformation and the expected design deformation. The ratio of the maximum deviation to the allowable limit is used as the frame deformation compatibility index, which is then normalized.

[0013] In the formula, This is a normalized framework deformation compatibility index. This represents the actual relative deformation. To design the maximum deformation.

[0014] Furthermore, the process of generating a health score for the glass plate based on the collected status parameters includes: Health score S:

[0015] In the formula, , , , and These are weighting coefficients, obtained through training based on historical data; The process of calculating the rate of change of the health score of the glass plate based on the health score includes: An evaluation period is set, and the health score of each glass plate is calculated once in each evaluation period. In order to accurately capture the changing trend, the change rate (ΔS / Δt) is calculated using the sliding window linear regression method. A historical sequence containing the health scores of the most recent N periods is maintained, and a univariate linear regression analysis is performed on the data points (time t, health score S) of the sequence to obtain the trend line S=a*t+b. The regression coefficient a is the change rate of the health score in that time period.

[0016] Furthermore, the process of grading glass plates based on their health scores and the rate of change in those scores includes: Set appropriate first and second health score thresholds based on historical health score data, where the historical health score data refers to the data set of previous glass plate health scores. Collect a large amount of historical health score data of glass plates and arrange them in ascending order. Set the first health score threshold to the 75% to 85% quantile of the health score distribution, which means that most healthy glass plates are above this threshold. Set the second health score threshold to the 15% to 25% quantile of the health score distribution, which means that glass plates below this threshold are in a significantly pathological state. Based on health scores, the glass plate is divided into three basic zones: When the health score is greater than or equal to the first health score threshold, the glass plate is judged to be a healthy glass plate. When the second health score threshold is less than or equal to the health score, which is less than the first health score threshold, the glass plate is judged to be a sub-healthy glass plate. When the health score is less than the second health score threshold, the glass plate is judged to be a pathological glass plate. Set an appropriate threshold for the rate of change of health scores based on historical health score change data, where the historical health score change data refers to the data set of previous health score change rates of the glass plate. Setting the threshold for the rate of change of health scores to the 10% to 20% quantile of the distribution of the rate of change of health scores indicates that glass plates with a rate of change below this threshold tend to be stable. Within each baseline interval, a secondary division is performed based on the severity of the rate of change in the health score: For healthy glass panels, if the rate of change of their health score is greater than the threshold for the rate of change of their health score, they are classified as Grade A glass panels (stable and healthy); otherwise, they are classified as Grade B glass panels (healthy but deteriorating). For sub-health glass plates, if the rate of change is greater than the threshold of the rate of change of health score, it is classified as a C-level glass plate (controllable observation); otherwise, it is classified as a D-level glass plate (requires planned maintenance). D-level glass plates are marked as glass plates that require maintenance. All damaged glass panels are directly classified as Class E glass panels (dangerous and urgent), and Class E glass panels are marked as glass panels requiring maintenance. The process of generating the initial maintenance path on the glass curtain wall based on the glass panel grade includes: First, based on the health classification results, all glass panels marked as requiring maintenance are designated as nodes to be visited. The spatial three-dimensional coordinates of these glass panels in the building information model are obtained and mapped onto a two-dimensional plan. When generating the initial path, the preset maintenance start point is used as the path starting point. The nearest neighbor greedy algorithm is used for path planning. Starting from the starting point, the algorithm always selects the glass panel that is currently unvisited and has the closest Euclidean distance to the one requiring maintenance as the next visit point, until all such glass panels are included in the visit sequence. Finally, a path is planned to return to the starting point or the specified endpoint, thus forming an initial maintenance path with the goal of minimizing the immediate movement distance.

[0017] Furthermore, the process of acquiring historical aging rate data and constructing an aging rate prediction model, and then using this model to predict the aging rate of the glass plate, includes: The aging rate refers to the aging speed of the glass plate; Factors affecting the aging rate include: current health score, rate of change of current health score, environmental severity index, material batch and initial performance coefficient, installation and construction quality score, and historical maintenance records and effectiveness score; The environmental severity index is calculated by combining real-time and historical meteorological data. Obtain material batches and initial performance coefficients from curtain wall material databases and production records; An installation quality score is derived based on inspection reports, video recordings, and initial stress test data from the construction phase. Historical maintenance records and performance evaluation refer to extracting the type, time, process, and acceptance results of each maintenance session from the maintenance management database. Obtain historical aging rate data for a single glass panel requiring maintenance in different evaluation periods. The historical aging rate data includes the current health score, current health score change rate, environmental severity index, material batch and initial performance coefficient, installation and construction quality score, historical maintenance records and effect scores, and the historical aging rate of the single glass panel requiring maintenance in different evaluation periods. Based on the current health score, current health score change rate, environmental severity index, material batch and initial performance coefficient, installation and construction quality score, historical maintenance record and effect score and corresponding historical aging rate of the glass panel to be maintained in different historical aging rate data, an aging rate prediction set is generated and divided into the first training set and the first test set. A first convolutional neural network is constructed. The first convolutional neural network is a one-dimensional convolutional neural network, which includes two convolutional layers, two pooling layers, a flattening layer, and two fully connected layers. A Dropout layer is set after the first fully connected layer. The network is trained using the Adam optimizer, the loss function is the mean squared error, and the training termination is determined by the early stopping method. The current health score, current health score change rate, environmental severity index, material batch and initial performance coefficient, installation and construction quality score, and historical maintenance record and effect score from different historical aging rate data in the first training set are used as input data for the first convolutional neural network, and the corresponding historical aging rate in the first training set is used as output data for the first convolutional neural network. The first convolutional neural network is trained to obtain the first initial convolutional neural network. The first initial convolutional neural network is validated using the first test set. The first initial convolutional neural network that outputs a value less than or equal to the preset first test error threshold is used as the aging rate prediction model. At each assessment cycle, the current health score, current health score change rate, environmental severity index, material batch and initial performance coefficient, installation and construction quality score, and historical maintenance records and effect scores of each glass panel requiring maintenance are input into the aging rate prediction model to obtain the predicted aging rate of each glass panel requiring maintenance.

[0018] Furthermore, the process of obtaining the aging time based on the predicted aging rate includes: After obtaining the predicted aging rate V based on the aging rate prediction model, a safe time window is calculated for each glass panel. This safe time window is defined as the period from the current health score... The health score has fallen to the pre-set baseline value where intervention is necessary. The required time, i.e., the aging time T = ( - ) / V, the baseline value for the health score is the second health score threshold; The process of marking a glass plate as requiring maintenance based on its aging time includes: The calculated aging time T is compared with a time threshold pre-set based on maintenance resources, weather windows, and risk tolerance. If the T of the glass plate is less than 1, then... If the glass panel is deemed to have insufficient safety time window and is at risk of deteriorating into a dangerous state before the next planned maintenance cycle, it will be marked as a glass panel to be maintained and needs to be included in the preventive maintenance plan. An early warning will be automatically generated, prompting maintenance personnel to provide the glass panel's ID, location, predicted aging time, and recommended inspection items.

[0019] Furthermore, the process of optimizing the initial maintenance path based on the glass panel requiring maintenance to obtain the optimized maintenance path includes: The glass panels to be maintained, as determined by the aging rate prediction model, are used as new potential access nodes. The optimization process employs an insertion heuristic algorithm. The algorithm traverses each path segment consisting of a pair of adjacent access nodes on the initial path. For each glass panel to be maintained, the algorithm calculates the path length increment caused by inserting it between the two endpoints of the path segment. For each glass panel to be maintained, the algorithm selects the path segment position corresponding to its minimum insertion cost. In order of increasing insertion cost, the corresponding glass panels to be maintained are inserted into the corresponding positions of the initial maintenance path. However, the constraint is that the movement distance between any adjacent maintenance points after insertion must not exceed the preset upper limit based on operational safety. This process is repeated until all glass panels to be maintained are inserted or there are no feasible insertion positions. Finally, an optimized maintenance path is generated that includes both emergency remedial points (glass panels to be maintained) and high-priority prevention points (glass panels to be maintained).

[0020] Furthermore, the process of acquiring historical remaining lifetime data and constructing a remaining lifetime prediction model, and then using this model to predict the remaining lifetime of the glass plate, includes: The remaining lifespan refers to the remaining usage time of the glass plate under safe usage conditions; Factors affecting the remaining life of glass plates include: current aging rate, current health score, rate of change of current health score, cumulative fatigue damage of materials, creep and stress relaxation data, and systemic risk transmission coefficient; The cumulative fatigue damage of the material was calculated by applying the rainflow counting method and Mainner's linear cumulative damage rule based on long-term monitored stress time history data. Creep and stress relaxation data were obtained by fitting deformation and stress data of the adhesive joint and supporting structure through long-term monitoring. The systemic risk transmission coefficient refers to the weighting used to assess the impact of a glass panel failure on its subsystem and even the entire curtain wall system's safety and functionality. Acquire historical remaining life data of a single glass panel in different evaluation periods. The historical remaining life data includes the current aging rate, current health score, rate of change of current health score, cumulative material fatigue damage, creep and stress relaxation data, systemic risk transmission coefficient, and historical remaining life of the single glass panel in different evaluation periods. Based on the current aging rate, current health score, current health score change rate, material fatigue cumulative damage, creep and stress relaxation data, systemic risk transmission coefficient, and corresponding historical remaining life of the glass plate in different historical remaining life data, a remaining life prediction set is generated and divided into a second training set and a second test set. A second convolutional neural network is constructed. The second convolutional neural network is a one-dimensional convolutional neural network, which includes two convolutional layers, two pooling layers, a flattening layer, and two fully connected layers. A Dropout layer is set after the first fully connected layer. The network is trained using the Adam optimizer, the loss function is the mean squared error, and the training termination is determined by the early stopping method. The current aging rate, current health score, rate of change of current health score, cumulative material fatigue damage, creep and stress relaxation data, and systemic risk transmission coefficient from different historical remaining lifetime data in the second training set are used as input data for the second convolutional neural network, and the corresponding historical remaining lifetime in the second training set is used as output data for the second convolutional neural network. The second convolutional neural network is trained to obtain the second initial convolutional neural network. The second initial convolutional neural network is validated using the second test set. The second initial convolutional neural network that outputs a second test error threshold less than or equal to the preset second test error threshold is used as the remaining lifetime prediction model. At each evaluation cycle, the current aging rate, current health score, rate of change of current health score, cumulative material fatigue damage, creep and stress relaxation data, and systemic risk transmission coefficient of each glass plate within the evaluation cycle are input into the remaining life prediction model to obtain the predicted remaining life of each glass plate.

[0021] Furthermore, based on the predicted lifespan, the lifespan gaps of the glass curtain wall are identified. The process of conducting unified inspections and collaborative maintenance of the glass panels in these lifespan gaps includes: After obtaining the predicted remaining life of the glass panel through the remaining life prediction model, a life column proportional to its predicted remaining life is generated for each glass panel on a digital twin platform that integrates the building's 3D model for visualization, forming a life topography map of the glass curtain wall. This allows for the intuitive identification of clusters of areas with significantly lower lifespans, which are defined as life depressions. For the identified lifespan depressions, the average remaining lifespan of all glass plates within the depression is calculated, and the core plates of the depression with lifespans lower than the average remaining lifespan are screened out. Time-predicted lifespan curves are plotted for the core plates of the depressions, and the rate of decline and synergy of the curves are analyzed. If multiple curves are found to show a synchronous and significant increase in the rate of decline in a similar time period, it strongly suggests the existence of common pathogenic factors. By calling historical data of the core plate of the depression for correlation mining and providing possible causal hypotheses, a unified investigation and collaborative maintenance plan for the entire lifespan of the depression can be generated. It is recommended to conduct comprehensive non-destructive testing and sampling analysis in the area and formulate comprehensive governance measures including replacement, reinforcement, and protection.

[0022] A method for managing the entire lifecycle of glass curtain walls includes the following steps: S1: Real-time acquisition of the status parameters of each glass panel of the glass curtain wall through acquisition equipment, and normalization processing of the status parameters; S2: Calculate the real-time health score of each glass plate based on the processed parameters. Calculate the rate of change of the health score of the glass plate based on the continuous health score over time using the sliding window linear regression method. S3: Based on the health score and the rate of change of the health score, the glass panels are accurately classified into different levels, and the glass panels that need maintenance are marked. Using the spatial coordinates of the glass panels that need maintenance as nodes, an initial maintenance path is planned to prioritize the handling of emergency risks. S4: Build and apply an aging rate prediction model to predict the future performance aging rate of each glass panel, calculate the aging time based on the aging rate, compare it with the preset maintenance cycle threshold, mark glass panels with insufficient aging time as glass panels to be maintained, and include them in the preventive maintenance plan. S5: Based on the initial maintenance path, the marked glass panels to be maintained are added as new task points. The glass panels to be maintained are inserted into the existing path to generate an optimized maintenance path that includes both emergency remediation and cost-effective prevention tasks. S6: Construct and apply a remaining life prediction model to predict the remaining safe service life of each glass panel, visualize it in a three-dimensional digital twin model to form a life topography map, automatically identify and analyze life depressions with significantly low remaining life, and intelligently diagnose common root causes of defects by analyzing the synergy of the life curves of the glass panels in the depressions, and generate targeted cluster unified investigation and collaborative maintenance solutions.

[0023] The technical effects and advantages of the glass curtain wall full life cycle management method and system of the present invention are as follows: (1) By setting up a remaining life prediction model, all glass plates are given a predicted life value and visualized in the form of a life topographic map in a three-dimensional digital twin model. Spatial statistical methods are used to identify clusters with significantly low remaining life, namely life depressions. By conducting synergistic analysis and causal inference on the life curves of glass plates in the depressions, the common causes of cluster performance degradation can be intelligently diagnosed. Based on this diagnosis, a unified investigation and collaborative maintenance plan for the entire depression can be generated, thereby achieving a systematic improvement from treating individual defects to improving regional performance, avoiding repetitive inspections and fragmented maintenance, and saving a lot of manpower and resources.

[0024] (2) By setting a health score, the health status of each glass panel can be accurately assessed. By setting an aging rate prediction model, the future performance degradation rate of each glass panel can be accurately predicted. Based on this prediction result, the safe time window of each glass panel can be calculated, and a preventive maintenance warning can be issued proactively before its health status deteriorates to the danger threshold. This allows maintenance actions to be initiated proactively before potential failures occur. Based on the dynamic grading results of the health score and the rate of change, the shortest initial maintenance path is planned for the diseased glass panels that urgently need to be treated, ensuring that emergency risks are controlled first. The glass panels to be maintained that are about to enter the risk window and are selected by the aging rate prediction model are integrated into the initial maintenance path to form an optimized maintenance path. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the system of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0027] Reference Figure 1 A glass curtain wall full life cycle management system includes the following modules: The data acquisition module is used to collect the status parameters of each glass panel in the glass curtain wall; The health assessment module is used to generate a health score for the glass plate based on the collected status parameters, and to calculate the rate of change of the health score of the glass plate based on the health score. The maintenance path planning module is used to classify glass panels according to their health scores and health score change rates, and generate initial maintenance paths on the glass curtain wall based on the glass panel's level. The aging rate prediction module is used to acquire historical aging rate data and build an aging rate prediction model. The aging rate prediction model is used to predict the aging rate of the glass plate, the aging time is obtained based on the predicted aging rate, and the glass plate is marked as a glass plate that needs maintenance based on the aging time. The maintenance path optimization module is used to optimize the initial maintenance path based on the glass panel to be maintained, thereby obtaining an optimized maintenance path. The remaining life prediction module is used to acquire historical remaining life data and build a remaining life prediction model. The remaining life prediction model is used to predict the remaining life of the glass panels. Based on the predicted life, the life gaps of the glass curtain wall are identified, and unified investigation and collaborative maintenance are carried out on the glass panels in the life gaps.

[0028] It should be further explained that, in the specific implementation process, the glass curtain wall laid on the exterior of the building is spliced ​​together from multiple glass panel units. The state parameters refer to the attenuation rate of the structural adhesive bonding strength of the glass panel, the aging index of the sealant, the propagation rate of microcracks on the glass surface, the cumulative fatigue damage index of the hardware, and the frame deformation coordination index. The structural adhesive internal defects are detected by ultrasonic flaw detector, and the surface image of the adhesive joint is captured by high-definition camera. Deep learning algorithm is used to identify surface deterioration characteristics such as cracking and powdering. The structural adhesive adhesive strength decay rate is obtained by combining internal and surface data to calculate the percentage decay rate of the current adhesive strength relative to the design strength. The structural adhesive bond strength attenuation rate is determined by scoring internal defects detected using a fusion ultrasonic flaw detector. Apparent degradation scoring using deep learning in conjunction with high-definition cameras The internal defect score is derived from the overall assessment. The results are obtained based on the size, quantity, and distribution density of internal defects such as bubbles and debonding detected by ultrasonic testing. The surface image of the adhesive joint is semantically segmented and classified using a deep learning model to identify surface deterioration features such as cracking, chalking, and discoloration.

[0029] in, These are weighting coefficients, obtained through training based on historical data; The attenuation rate of structural adhesive bond strength was normalized.

[0030] In the formula, This represents the normalized structural adhesive bond strength attenuation rate. At the current intensity, The initial design strength is 1.5 MPa. Infrared spectroscopy was used to analyze the chemical bonds at the sampling points of the sealant. At the same time, a Shore hardness tester was used to measure the hardness change of the colloid. The chemical aging degree and physical hardening degree were combined to calculate the sealant aging index. The sealant aging index is a measure of chemical aging as determined by infrared spectroscopy. Physical hardening degree measured by Shore hardness tester In summary, The degradation rate of chemical bonds relative to virgin adhesive was calculated by analyzing the changes in the intensity of the infrared absorption peaks of the characteristic chemical bonds in the sealant. The hardening ratio relative to the initial hardness is calculated by measuring the change in the hardness of the sealant.

[0031] in, These are weighting coefficients, obtained through training based on historical data; The aging index of the sealant was normalized:

[0032] In the formula, To normalize the aging index of the sealant, As a measure of aging, This refers to the degree of aging and failure, specifically 80%. A laser speckle interferometer is used to perform non-contact scanning of the glass surface to acquire micron-level deformation field images. Through deep learning image recognition, the average length expansion of microcracks per unit time is calculated to obtain the microcrack propagation rate on the glass surface. The microcrack propagation rate on the glass surface is then normalized.

[0033] In the formula, To normalize the microcrack propagation rate on the glass surface, This represents the current expansion rate; Strain gauges were attached to key hardware components such as sash hinges and sliding supports to monitor their stress time history over a long period. The number of cycles at different stress amplitudes was counted using the rainflow counting method. Then, the total fatigue damage degree was calculated according to the Mainner linear cumulative damage rule to obtain the fatigue damage accumulation index of the hardware components. The fatigue damage accumulation index of the hardware components was then normalized.

[0034] In the formula, This is the normalized cumulative fatigue damage index for hardware parts. This represents the current cumulative damage level. Distributed fiber optic sensors are attached along the vertical keel and horizontal trusses of the curtain wall to monitor their strain distribution under wind load and temperature changes. The measured strain data are input into a finite element model for inversion analysis to calculate the deviation between the actual deformation and the expected design deformation. The ratio of the maximum deviation to the allowable limit is used as the frame deformation compatibility index, which is then normalized.

[0035] In the formula, This is a normalized framework deformation compatibility index. This represents the actual relative deformation. The maximum design deformation is 10mm.

[0036] It should be further explained that, in the specific implementation process, the process of generating a health score for the glass plate based on the collected status parameters includes: Health score S:

[0037] In the formula, , , , and These are weighting coefficients, obtained from training on historical data, and set to 0.3, 0.25, 0.15, 0.15, and 0.15 respectively. If the state parameters of a certain glass plate are: It is 1.2 MPa. 60%, It is 0.2 mm / year. It is 0.4. If the thickness is 8mm, then the health score S of the glass plate is 0.548; The process of calculating the rate of change of the health score of the glass plate based on the health score includes: An evaluation period is set, and the health score of each glass plate is calculated once in each evaluation period. In order to accurately capture the changing trend, the change rate (ΔS / Δt) is calculated using the sliding window linear regression method. A historical sequence containing the health scores of the most recent N periods is maintained, and a univariate linear regression analysis is performed on the data points (time t, health score S) of the sequence to obtain the trend line S=a*t+b. The regression coefficient a is the change rate of the health score in that time period.

[0038] It should be further explained that, in the specific implementation process, the process of grading the glass plates based on their health scores and the rate of change of those scores includes: Set appropriate first and second health score thresholds based on historical health score data, where the historical health score data refers to the data set of previous glass plate health scores. Collect a large amount of historical health score data of glass plates and arrange them in ascending order. Set the first health score threshold to the 75% to 85% quantile of the health score distribution, which means that most healthy glass plates are above this threshold. Set the second health score threshold to the 15% to 25% quantile of the health score distribution, which means that glass plates below this threshold are in a significantly pathological state. Based on health scores, the glass plate is divided into three basic zones: When the health score is greater than or equal to the first health score threshold, the glass plate is judged to be a healthy glass plate. When the second health score threshold is less than or equal to the health score, which is less than the first health score threshold, the glass plate is judged to be a sub-healthy glass plate. When the health score is less than the second health score threshold, the glass plate is judged to be a pathological glass plate. Set an appropriate threshold for the rate of change of health scores based on historical health score change data, where the historical health score change data refers to the data set of previous health score change rates of the glass plate. Setting the threshold for the rate of change of health scores to the 10% to 20% quantile of the distribution of the rate of change of health scores indicates that glass plates with a rate of change below this threshold tend to be stable. Within each baseline interval, a secondary division is performed based on the severity of the rate of change in the health score: For healthy glass panels, if the rate of change of their health score is greater than the threshold for the rate of change of their health score, they are classified as Grade A glass panels (stable and healthy); otherwise, they are classified as Grade B glass panels (healthy but deteriorating). For sub-health glass plates, if the rate of change is greater than the threshold of the rate of change of health score, it is classified as a C-level glass plate (controllable observation); otherwise, it is classified as a D-level glass plate (requires planned maintenance). D-level glass plates are marked as glass plates that require maintenance. All damaged glass panels are directly classified as Class E glass panels (dangerous and urgent), and Class E glass panels are marked as glass panels requiring maintenance. The process of generating the initial maintenance path on the glass curtain wall based on the glass panel grade includes: First, based on the health classification results, all glass panels marked as requiring maintenance are designated as nodes to be visited. The spatial three-dimensional coordinates of these glass panels in the building information model are obtained and mapped onto a two-dimensional plan. When generating the initial path, the preset maintenance start point is used as the path starting point. The nearest neighbor greedy algorithm is used for path planning. Starting from the starting point, the algorithm always selects the glass panel that is currently unvisited and has the closest Euclidean distance to the one requiring maintenance as the next visit point, until all such glass panels are included in the visit sequence. Finally, a path is planned to return to the starting point or the specified endpoint, thus forming an initial maintenance path with the goal of minimizing the immediate movement distance.

[0039] It should be further explained that, in the specific implementation process, the process of acquiring historical aging rate data and constructing an aging rate prediction model, and then using the aging rate prediction model to predict the aging rate of the glass plate, includes: The aging rate refers to the aging speed of the glass plate; Factors affecting the aging rate include: current health score, rate of change of current health score, environmental severity index, material batch and initial performance coefficient, installation and construction quality score, and historical maintenance records and effectiveness score; The environmental severity index is calculated by combining real-time and historical meteorological data (annual average temperature difference, ultraviolet radiation, acid rain frequency, and maximum wind pressure). Obtain material batches and initial performance coefficients from curtain wall material databases and production records; An installation quality score is derived based on inspection reports, video recordings, and initial stress test data from the construction phase. Historical maintenance records and performance evaluation refer to extracting the type, time, process, and acceptance results of each maintenance session from the maintenance management database. Obtain historical aging rate data for a single glass panel requiring maintenance in different evaluation periods. The historical aging rate data includes the current health score, current health score change rate, environmental severity index, material batch and initial performance coefficient, installation and construction quality score, historical maintenance records and effect scores, and the historical aging rate of the single glass panel requiring maintenance in different evaluation periods. Based on the current health score, current health score change rate, environmental severity index, material batch and initial performance coefficient, installation and construction quality score, historical maintenance record and effect score and corresponding historical aging rate of the glass panel to be maintained in different historical aging rate data, an aging rate prediction set is generated and divided into the first training set and the first test set. Construct a first convolutional neural network, taking the current health score, current health score change rate, environmental severity index, material batch and initial performance coefficient, installation and construction quality score and historical maintenance record and effect score from different historical aging rate data in the first training set as the input data of the first convolutional neural network, and taking the corresponding historical aging rate in the first training set as the output data of the first convolutional neural network. The first convolutional neural network is trained to obtain the first initial convolutional neural network. The first initial convolutional neural network is validated using the first test set. The first initial convolutional neural network that outputs a value less than or equal to the preset first test error threshold is used as the aging rate prediction model. The first convolutional neural network adopts a one-dimensional convolutional neural network structure. Its input layer receives six feature parameters: current health score, current health score change rate, environmental severity index, material batch and initial performance coefficient, installation and construction quality score, and historical maintenance record and effect score. The input dimension is 6×1. The network contains two convolutional layers. The first convolutional layer uses 32 convolutional kernels of size 3 with ReLU activation function and same padding. The second convolutional layer uses 64 convolutional kernels of size 3 with ReLU activation function and same padding. Each convolutional layer is followed by a max pooling layer with a pooling window size of 2. With a stride of 2, the pooled feature maps are flattened and then fed into two fully connected layers. The first fully connected layer has 128 neurons with ReLU activation function, followed by a Dropout layer with a dropout rate of 0.5 to prevent overfitting. The output layer has one neuron (linear output) corresponding to the predicted aging rate. The network is trained using the Adam optimizer with an initial learning rate of 0.001, a batch size of 32, and a loss function of mean squared error. Early stopping is used, and training is terminated when the validation set loss no longer decreases for 10 consecutive rounds. The preset first test error threshold is mean absolute percentage error ≤ 5%. The model is considered to have passed validation if it reaches this metric on the test set. At each assessment cycle, the current health score, current health score change rate, environmental severity index, material batch and initial performance coefficient, installation and construction quality score, and historical maintenance records and effect scores of each glass panel requiring maintenance are input into the aging rate prediction model to obtain the predicted aging rate of each glass panel requiring maintenance. In an embodiment of the present invention, the predicted aging rate of all glass panels requiring maintenance is obtained through an aging rate prediction model. The predicted aging rate is related to the current health score, the rate of change of the current health score, the environmental severity index, the material batch and initial performance coefficient, the installation and construction quality score, and the historical maintenance records and effect scores. The current health score directly affects the predicted rate of aging. The higher the current health score, the slower the predicted rate of aging. Therefore, the current health score is negatively correlated with the rate of aging. The rate of change in the current health score directly affects the rate of aging prediction. The higher the rate of change in the current health score, the slower the rate of aging prediction. Therefore, the rate of change in the current health score is negatively correlated with the rate of aging. The level of the environmental severity index directly affects the speed of the predicted aging rate. The higher the environmental severity index, the greater the accelerating effect of the external environment on material aging, and the faster the predicted aging rate. Therefore, the environmental severity index and the aging rate are positively correlated. The material batch and initial performance coefficient directly affect the predicted aging rate. The higher the material batch and initial performance coefficient, the better the initial performance can slow down the aging rate and the slower the predicted aging rate. Therefore, the material batch and initial performance coefficient are negatively correlated with the aging rate. The installation quality score directly affects the predicted aging rate. The lower the installation quality score, the more the initial defects or residual stress introduced by poor installation will significantly accelerate the aging process and the faster the predicted aging rate. Therefore, the installation quality score is negatively correlated with the aging rate. The quality of historical maintenance records and effectiveness scores directly affects the predicted aging rate. Higher historical maintenance records and effectiveness scores indicate that effective preventative maintenance can reduce the aging rate and the predicted aging rate is slower. Therefore, historical maintenance records and effectiveness scores are negatively correlated with the aging rate.

[0040] It should be further explained that, in the specific implementation process, the process of obtaining the aging time based on the predicted aging rate includes: After obtaining the predicted aging rate V based on the aging rate prediction model, a safe time window is calculated for each glass panel. This safe time window is defined as the period from the current health score... The health score has fallen to the pre-set baseline value where intervention is necessary. The required time, i.e., the aging time T = ( - ) / V, the baseline value for the health score is the second health score threshold; The process of marking a glass plate as requiring maintenance based on its aging time includes: The calculated aging time T is compared with a time threshold pre-set based on maintenance resources, weather windows, and risk tolerance. If the T of the glass plate is less than 1, then... If the glass panel is deemed to have insufficient safety time window and is at risk of deteriorating into a dangerous state before the next planned maintenance cycle, it will be marked as a glass panel to be maintained and needs to be included in the preventive maintenance plan. An early warning will be automatically generated, prompting maintenance personnel to provide the glass panel's ID, location, predicted aging time, and recommended inspection items.

[0041] It should be further explained that, in the specific implementation process, the process of optimizing the initial maintenance path based on the glass panel requiring maintenance to obtain the optimized maintenance path includes: The glass panels to be maintained, as determined by the aging rate prediction model, are used as new potential access nodes. The optimization process employs an insertion heuristic algorithm. The algorithm traverses each path segment consisting of a pair of adjacent access nodes on the initial path. For each glass panel to be maintained, the algorithm calculates the path length increment caused by inserting it between the two endpoints of the path segment. For each glass panel to be maintained, the algorithm selects the path segment position corresponding to its minimum insertion cost. In order of increasing insertion cost, the corresponding glass panels to be maintained are inserted into the corresponding positions of the initial maintenance path. However, the constraint is that the movement distance between any adjacent maintenance points after insertion must not exceed the preset upper limit based on operational safety. This process is repeated until all glass panels to be maintained are inserted or there are no feasible insertion positions. Finally, an optimized maintenance path is generated that includes both emergency remedial points (glass panels to be maintained) and high-priority prevention points (glass panels to be maintained).

[0042] It should be further explained that, in the specific implementation process, the process of acquiring historical remaining lifetime data and constructing a remaining lifetime prediction model, and then using the remaining lifetime prediction model to predict the remaining lifetime of the glass plate, includes: The remaining lifespan refers to the remaining usage time of the glass plate under safe usage conditions; Factors affecting the remaining life of glass plates include: current aging rate, current health score, rate of change of current health score, cumulative fatigue damage of materials, creep and stress relaxation data, and systemic risk transmission coefficient; The cumulative fatigue damage of the material was calculated by applying the rainflow counting method and Mainner's linear cumulative damage rule based on long-term monitored stress time history data. Creep and stress relaxation data were obtained by fitting deformation and stress data of the adhesive joint and supporting structure through long-term monitoring. The systemic risk transmission coefficient refers to the weighting used to assess the impact of a glass panel failure on its subsystem and even the entire curtain wall system's safety and functionality. Acquire historical remaining life data of a single glass panel in different evaluation periods. The historical remaining life data includes the current aging rate, current health score, rate of change of current health score, cumulative material fatigue damage, creep and stress relaxation data, systemic risk transmission coefficient, and historical remaining life of the single glass panel in different evaluation periods. Based on the current aging rate, current health score, current health score change rate, material fatigue cumulative damage, creep and stress relaxation data, systemic risk transmission coefficient, and corresponding historical remaining life of the glass plate in different historical remaining life data, a remaining life prediction set is generated and divided into a second training set and a second test set. A second convolutional neural network is constructed, and the current aging rate, current health score, current health score change rate, material fatigue cumulative damage, creep and stress relaxation data and systemic risk transmission coefficient in different historical remaining lifetime data in the second training set are used as the input data of the second convolutional neural network, and the corresponding historical remaining lifetime in the second training set is used as the output data of the second convolutional neural network. The second convolutional neural network is trained to obtain the second initial convolutional neural network. The second initial convolutional neural network is validated using the second test set. The second initial convolutional neural network that outputs a second test error threshold less than or equal to the preset second test error threshold is used as the remaining lifetime prediction model. The second convolutional neural network adopts the same structure as the first convolutional neural network, but its input layer receives six feature parameters: current aging rate, current health score, rate of change of current health score, cumulative material fatigue damage, creep and stress relaxation data, and systemic risk transmission coefficient. The input dimension is 6×1. The network also contains two convolutional layers (32 3×1 convolutional kernels in the first layer and 64 3×1 convolutional kernels in the second layer, both using ReLU activation function and same padding), two max pooling layers (2×2 window, stride 2), a flattening layer, and two fully connected layers (128 neurons in the first layer with ReLU activation and Dropout 0.5, and 1 neuron in the second layer with linear output). The hyperparameter settings are the same as the first convolutional neural network: Adam optimizer (learning rate 0.001), batch size 32, loss function MSE, and early stopping (stopping if the validation set loss does not decrease for 10 consecutive rounds). The second test error threshold is also set to mean absolute percentage error ≤ 5%. The model is considered to have passed validation if it reaches this metric on the test set. At each evaluation cycle, the current aging rate, current health score, rate of change of current health score, cumulative material fatigue damage, creep and stress relaxation data, and systemic risk transmission coefficient of each glass plate within the evaluation cycle are input into the remaining life prediction model to obtain the predicted remaining life of each glass plate. In an embodiment of the present invention, the predicted remaining life of all glass plates is obtained through a remaining life prediction model. The predicted remaining life is related to the current aging rate, current health score, rate of change of current health score, cumulative material fatigue damage, creep and stress relaxation data, and systemic risk transmission coefficient. The current aging rate directly affects the predicted remaining lifespan. The faster the current aging rate, the faster the glass plate ages and the shorter the predicted remaining lifespan. Therefore, the current aging rate and the remaining lifespan are negatively correlated. The current health score directly affects the predicted life expectancy. The higher the current health score, the better the glass plate condition, and the greater the predicted life expectancy. Therefore, the current health score is positively correlated with the life expectancy. The rate of change of the current health score directly affects the amount of life remaining predicted. The higher the rate of change of the current health score, the better the condition of the glass plate, and the greater the life remaining predicted. Therefore, the rate of change of the current health score is positively correlated with life remaining. The degree of cumulative fatigue damage of materials directly affects the predicted remaining life. The higher the degree of cumulative fatigue damage of materials, the more irreversible damage caused by cyclic loading of the glass plate, and the lower the predicted remaining life. Therefore, the degree of cumulative fatigue damage of materials is negatively correlated with the remaining life. The level of creep and stress relaxation data directly affects the predicted remaining life. The higher the creep and stress relaxation data, the more significant the creep of the glass plate will be, which will shorten the structural life and reduce the predicted remaining life. Therefore, creep and stress relaxation data are negatively correlated with remaining life. The level of the systemic risk transmission coefficient directly affects the amount of predicted remaining life. The higher the systemic risk transmission coefficient, the less acceptable remaining life may be, and the less predicted remaining life may be. Therefore, the systemic risk transmission coefficient is negatively correlated with remaining life.

[0043] It should be further explained that, in the specific implementation process, based on the predicted lifespan to identify the lifespan gaps in the glass curtain wall, the process of conducting unified inspections and collaborative maintenance of the glass panels in these gaps includes: After obtaining the predicted remaining life of the glass panel through the remaining life prediction model, a life column proportional to its predicted remaining life is generated for each glass panel on a digital twin platform that integrates the building's 3D model for visualization, forming a life topography map of the glass curtain wall. This allows for the intuitive identification of clusters of areas with significantly lower lifespans, which are defined as life depressions. For the identified lifetime depressions, the average remaining lifetime of all glass panels within the depression is calculated, and the core panels of the depression with lifetimes lower than the average remaining lifetime are selected. Time-predicted lifetime curves are plotted for the core panels of the depressions, and the rate of decline and synergy of the curves are analyzed. If multiple curves are found to show a synchronous and significant increase in the rate of decline in a similar time period, it strongly suggests the existence of common pathogenic factors (such as shared structural defects, common environmental erosion sources, batch material problems). By calling historical data (construction records, maintenance archives, and environmental monitoring data) of the core slab of the depression for correlation mining and providing possible causal hypotheses, a unified investigation and collaborative maintenance plan for the entire lifespan of the depression can be generated. It is recommended to conduct comprehensive non-destructive testing and sampling analysis in the area and to formulate comprehensive treatment measures including replacement, reinforcement, and protection.

[0044] A method for managing the entire lifecycle of glass curtain walls includes the following steps: S1: Real-time acquisition of the status parameters of each glass panel of the glass curtain wall through acquisition equipment, and normalization processing of the status parameters; S2: Calculate the real-time health score of each glass plate based on the processed parameters. Calculate the rate of change of the health score of the glass plate based on the continuous health score over time using the sliding window linear regression method. S3: Based on the health score and the rate of change of the health score, the glass panels are accurately classified into different levels, and the glass panels that need maintenance are marked. Using the spatial coordinates of the glass panels that need maintenance as nodes, an initial maintenance path is planned to prioritize the handling of emergency risks. S4: Build and apply an aging rate prediction model to predict the future performance aging rate of each glass panel, calculate the aging time based on the aging rate, compare it with the preset maintenance cycle threshold, mark glass panels with insufficient aging time as glass panels to be maintained, and include them in the preventive maintenance plan. S5: Based on the initial maintenance path, the marked glass panels to be maintained are added as new task points. The glass panels to be maintained are inserted into the existing path to generate an optimized maintenance path that includes both emergency remediation and cost-effective prevention tasks. S6: Construct and apply a remaining life prediction model to predict the remaining safe service life of each glass panel, visualize it in a three-dimensional digital twin model to form a life topography map, automatically identify and analyze life depressions with significantly low remaining life, and intelligently diagnose common root causes of defects by analyzing the synergy of the life curves of the glass panels in the depressions, and generate targeted cluster unified investigation and collaborative maintenance solutions.

[0045] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0046] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A glass curtain wall full life cycle management system, characterized in that, Includes the following modules: The data acquisition module is used to collect the status parameters of each glass panel in the glass curtain wall; The health assessment module is used to generate a health score for the glass plate based on the collected status parameters, and to calculate the rate of change of the health score of the glass plate based on the health score. The maintenance path planning module is used to classify glass panels according to their health scores and health score change rates, and generate initial maintenance paths on the glass curtain wall based on the glass panel's level. The aging rate prediction module is used to acquire historical aging rate data and build an aging rate prediction model. The aging rate prediction model is used to predict the aging rate of the glass plate, the aging time is obtained based on the predicted aging rate, and the glass plate is marked as a glass plate that needs maintenance based on the aging time. The maintenance path optimization module is used to optimize the initial maintenance path based on the glass panel to be maintained, thereby obtaining an optimized maintenance path. The remaining life prediction module is used to acquire historical remaining life data and build a remaining life prediction model. The remaining life prediction model is used to predict the remaining life of the glass panels. Based on the predicted life, the life gaps of the glass curtain wall are identified, and unified investigation and collaborative maintenance are carried out on the glass panels in the life gaps.

2. The glass curtain wall full life cycle management system according to claim 1, characterized in that, The glass curtain wall on the exterior of the building is composed of multiple glass panel units spliced ​​together. The state parameters refer to the attenuation rate of the structural adhesive bonding strength of the glass panel, the aging index of the sealant, the propagation rate of microcracks on the glass surface, the cumulative fatigue damage index of the hardware, and the frame deformation compatibility index. The structural adhesive internal defects are detected by ultrasonic flaw detector, and the surface image of the adhesive joint is captured by high-definition camera. Deep learning algorithm is used to identify surface deterioration characteristics such as cracking and powdering. The structural adhesive adhesive strength decay rate is obtained by combining internal and surface data to calculate the percentage decay rate of the current adhesive strength relative to the design strength. The structural adhesive bond strength attenuation rate is determined by scoring internal defects detected using a fusion ultrasonic flaw detector. Apparent degradation scoring using deep learning in conjunction with high-definition cameras The internal defect score is derived from the overall assessment. The results are obtained based on the size, quantity, and distribution density of internal defects such as bubbles and debonding detected by ultrasonic testing. The surface image of the adhesive joint is semantically segmented and classified using a deep learning model to identify surface deterioration features such as cracking, chalking, and discoloration. in, These are weighting coefficients, obtained through training based on historical data; The attenuation rate of structural adhesive bond strength was normalized. In the formula, This represents the normalized structural adhesive bond strength attenuation rate. At the current intensity, For initial design strength; Infrared spectroscopy was used to analyze the chemical bonds at the sampling points of the sealant. At the same time, a Shore hardness tester was used to measure the hardness change of the colloid. The chemical aging degree and physical hardening degree were combined to calculate the sealant aging index. The sealant aging index is a measure of chemical aging as determined by infrared spectroscopy. Physical hardening degree measured by Shore hardness tester In summary, The degradation rate of chemical bonds relative to virgin adhesive was calculated by analyzing the changes in the intensity of the infrared absorption peaks of the characteristic chemical bonds in the sealant. The hardening ratio relative to the initial hardness is calculated by measuring the change in the hardness of the sealant. in, These are weighting coefficients, obtained through training based on historical data; The aging index of the sealant was normalized: In the formula, To normalize the aging index of the sealant, As a measure of aging, The degree of failure and aging; A laser speckle interferometer is used to perform non-contact scanning of the glass surface to acquire micron-level deformation field images. Through deep learning image recognition, the average length expansion of microcracks per unit time is calculated to obtain the microcrack propagation rate on the glass surface. The microcrack propagation rate on the glass surface is then normalized. In the formula, To normalize the microcrack propagation rate on the glass surface, This represents the current expansion rate; Strain gauges were attached to key hardware components such as sash hinges and sliding supports to monitor their stress time history over a long period. The number of cycles at different stress amplitudes was counted using the rainflow counting method. Then, the total fatigue damage degree was calculated according to the Mainner linear cumulative damage rule to obtain the fatigue damage accumulation index of the hardware components. The fatigue damage accumulation index of the hardware components was then normalized. In the formula, This is the normalized cumulative fatigue damage index for hardware parts. This represents the current cumulative damage level. Distributed fiber optic sensors are attached along the vertical keel and horizontal trusses of the curtain wall to monitor their strain distribution under wind load and temperature changes. The measured strain data are input into a finite element model for inversion analysis to calculate the deviation between the actual deformation and the expected design deformation. The ratio of the maximum deviation to the allowable limit is used as the frame deformation compatibility index, which is then normalized. In the formula, This is a normalized framework deformation compatibility index. This represents the actual relative deformation. To design the maximum deformation.

3. The glass curtain wall full life cycle management system according to claim 2, characterized in that, The process of generating a health score for a glass plate based on the collected status parameters includes: Health score S: In the formula, , , , and These are weighting coefficients, obtained through training based on historical data; The process of calculating the rate of change of the health score of the glass plate based on the health score includes: An evaluation period is set, and the health score of each glass plate is calculated once in each evaluation period. In order to accurately capture the changing trend, the change rate (ΔS / Δt) is calculated using the sliding window linear regression method. A historical sequence containing the health scores of the most recent N periods is maintained, and a univariate linear regression analysis is performed on the data points (time t, health score S) of the sequence to obtain the trend line S=a*t+b. The regression coefficient a is the change rate of the health score in that time period.

4. The glass curtain wall full life cycle management system according to claim 3, characterized in that, The process of grading glass plates based on their health scores and the rate of change of those scores includes: Set appropriate first and second health score thresholds based on historical health score data, where the historical health score data refers to the data set of previous glass plate health scores. Collect a large amount of historical health score data of glass plates and arrange them in ascending order. Set the first health score threshold to the 75% to 85% quantile of the health score distribution, which means that most healthy glass plates are above this threshold. Set the second health score threshold to the 15% to 25% quantile of the health score distribution, which means that glass plates below this threshold are in a significantly pathological state. Based on health scores, the glass plate is divided into three basic zones: When the health score is greater than or equal to the first health score threshold, the glass plate is judged to be a healthy glass plate. When the second health score threshold is less than or equal to the health score, which is less than the first health score threshold, the glass plate is judged to be a sub-healthy glass plate. When the health score is less than the second health score threshold, the glass plate is judged to be a pathological glass plate. Set an appropriate threshold for the rate of change of health scores based on historical health score change data, where the historical health score change data refers to the data set of previous health score change rates of the glass plate. Setting the threshold for the rate of change of health scores to the 10% to 20% quantile of the distribution of the rate of change of health scores indicates that glass plates with a rate of change below this threshold tend to be stable. Within each baseline interval, a secondary division is performed based on the severity of the rate of change in the health score: For healthy glass panels, if the rate of change of their health score is greater than the threshold for the rate of change of their health score, they are classified as Grade A glass panels (stable and healthy); otherwise, they are classified as Grade B glass panels (healthy but deteriorating). For sub-health glass plates, if the rate of change is greater than the threshold of the rate of change of health score, it is classified as a C-level glass plate (controllable observation); otherwise, it is classified as a D-level glass plate (requires planned maintenance). D-level glass plates are marked as glass plates that require maintenance. All damaged glass panels are directly classified as Class E glass panels (dangerous and urgent), and Class E glass panels are marked as glass panels requiring maintenance. The process of generating the initial maintenance path on the glass curtain wall based on the glass panel grade includes: First, based on the health classification results, all glass panels marked as requiring maintenance are designated as nodes to be visited. The spatial three-dimensional coordinates of these glass panels in the building information model are obtained and mapped onto a two-dimensional plan. When generating the initial path, the preset maintenance start point is used as the path starting point. The nearest neighbor greedy algorithm is used for path planning. Starting from the starting point, the algorithm always selects the glass panel that is currently unvisited and has the closest Euclidean distance to the one requiring maintenance as the next visit point, until all such glass panels are included in the visit sequence. Finally, a path is planned to return to the starting point or the specified endpoint, thus forming an initial maintenance path with the goal of minimizing the immediate movement distance.

5. The glass curtain wall full life cycle management system according to claim 4, characterized in that, The process of acquiring historical aging rate data and constructing an aging rate prediction model, and then using that model to predict the aging rate of the glass plate, includes: The aging rate refers to the aging speed of the glass plate; Factors affecting the aging rate include: current health score, rate of change of current health score, environmental severity index, material batch and initial performance coefficient, installation and construction quality score, and historical maintenance records and effectiveness score; The environmental severity index is calculated by combining real-time and historical meteorological data. Obtain material batches and initial performance coefficients from curtain wall material databases and production records; An installation quality score is derived based on inspection reports, video recordings, and initial stress test data from the construction phase. Historical maintenance records and performance evaluation refer to extracting the type, time, process, and acceptance results of each maintenance session from the maintenance management database. Obtain historical aging rate data for a single glass panel requiring maintenance in different evaluation periods. The historical aging rate data includes the current health score, current health score change rate, environmental severity index, material batch and initial performance coefficient, installation and construction quality score, historical maintenance records and effect scores, and the historical aging rate of the single glass panel requiring maintenance in different evaluation periods. Based on the current health score, current health score change rate, environmental severity index, material batch and initial performance coefficient, installation and construction quality score, historical maintenance record and effect score and corresponding historical aging rate of the glass panel to be maintained in different historical aging rate data, an aging rate prediction set is generated and divided into the first training set and the first test set. A first convolutional neural network is constructed. The first convolutional neural network is a one-dimensional convolutional neural network, which includes two convolutional layers, two pooling layers, a flattening layer, and two fully connected layers. A Dropout layer is set after the first fully connected layer. The network is trained using the Adam optimizer, the loss function is the mean squared error, and the training termination is determined by the early stopping method. The current health score, current health score change rate, environmental severity index, material batch and initial performance coefficient, installation and construction quality score, and historical maintenance record and effect score from different historical aging rate data in the first training set are used as input data for the first convolutional neural network, and the corresponding historical aging rate in the first training set is used as output data for the first convolutional neural network. The first convolutional neural network is trained to obtain the first initial convolutional neural network. The first initial convolutional neural network is validated using the first test set. The first initial convolutional neural network that outputs a value less than or equal to the preset first test error threshold is used as the aging rate prediction model. The first convolutional neural network is trained using the first training set, and the trained network is validated using the first test set. When the average absolute percentage error of the prediction is less than or equal to a preset first test error threshold, the aging rate prediction model is obtained. At each assessment cycle, the current health score, current health score change rate, environmental severity index, material batch and initial performance coefficient, installation and construction quality score, and historical maintenance records and effect scores of each glass panel requiring maintenance are input into the aging rate prediction model to obtain the predicted aging rate of each glass panel requiring maintenance.

6. The glass curtain wall full life cycle management system according to claim 5, characterized in that, The process of obtaining aging time based on the predicted aging rate includes: After obtaining the predicted aging rate V based on the aging rate prediction model, a safe time window is calculated for each glass panel. This safe time window is defined as the period from the current health score... The health score has fallen to the pre-set baseline value where intervention is necessary. The required time, i.e., the aging time T = ( - ) / V, the baseline value for the health score is the second health score threshold; The process of marking a glass plate as requiring maintenance based on its aging time includes: The calculated aging time T is compared with a time threshold pre-set based on maintenance resources, weather windows, and risk tolerance. If the T of the glass plate is less than 1, then... If the glass panel is deemed to have insufficient safety time window and is at risk of deteriorating into a dangerous state before the next planned maintenance cycle, it will be marked as a glass panel to be maintained and needs to be included in the preventive maintenance plan. An early warning will be automatically generated, prompting maintenance personnel to provide the glass panel's ID, location, predicted aging time, and recommended inspection items.

7. The glass curtain wall full life cycle management system according to claim 6, characterized in that, The process of optimizing the initial maintenance path based on the glass panel requiring maintenance to obtain the optimized maintenance path includes: The glass panels to be maintained, as determined by the aging rate prediction model, are used as new potential access nodes. The optimization process employs an insertion heuristic algorithm. The algorithm traverses each path segment consisting of a pair of adjacent access nodes on the initial path. For each glass panel to be maintained, the algorithm calculates the path length increment caused by inserting it between the two endpoints of the path segment. For each glass panel to be maintained, the algorithm selects the path segment position corresponding to its minimum insertion cost. In order of increasing insertion cost, the corresponding glass panels to be maintained are inserted into the corresponding positions of the initial maintenance path. However, the constraint is that the movement distance between any adjacent maintenance points after insertion must not exceed the preset upper limit based on operational safety. This process is repeated until all glass panels to be maintained are inserted or there are no feasible insertion positions. Finally, an optimized maintenance path is generated that includes both emergency remedial points (glass panels to be maintained) and high-priority prevention points (glass panels to be maintained).

8. The glass curtain wall full life cycle management system according to claim 7, characterized in that, The process of acquiring historical remaining lifetime data and constructing a remaining lifetime prediction model, and then using that model to predict the remaining lifetime of the glass plate, includes: The remaining lifespan refers to the remaining usage time of the glass plate under safe usage conditions; Factors affecting the remaining life of glass plates include: current aging rate, current health score, rate of change of current health score, cumulative fatigue damage of materials, creep and stress relaxation data, and systemic risk transmission coefficient; The cumulative fatigue damage of the material was calculated by applying the rainflow counting method and Mainner's linear cumulative damage rule based on long-term monitored stress time history data. Creep and stress relaxation data were obtained by fitting deformation and stress data of the adhesive joint and supporting structure through long-term monitoring. The systemic risk transmission coefficient refers to the weighting used to assess the impact of a glass panel failure on its subsystem and even the entire curtain wall system's safety and functionality. Acquire historical remaining life data of a single glass panel in different evaluation periods. The historical remaining life data includes the current aging rate, current health score, rate of change of current health score, cumulative material fatigue damage, creep and stress relaxation data, systemic risk transmission coefficient, and historical remaining life of the single glass panel in different evaluation periods. Based on the current aging rate, current health score, current health score change rate, material fatigue cumulative damage, creep and stress relaxation data, systemic risk transmission coefficient, and corresponding historical remaining life of the glass plate in different historical remaining life data, a remaining life prediction set is generated and divided into a second training set and a second test set. A second convolutional neural network is constructed. The second convolutional neural network is a one-dimensional convolutional neural network, which includes two convolutional layers, two pooling layers, a flattening layer, and two fully connected layers. A Dropout layer is set after the first fully connected layer. The network is trained using the Adam optimizer, the loss function is the mean squared error, and the training termination is determined by the early stopping method. The current aging rate, current health score, rate of change of current health score, cumulative material fatigue damage, creep and stress relaxation data, and systemic risk transmission coefficient from different historical remaining lifetime data in the second training set are used as input data for the second convolutional neural network, and the corresponding historical remaining lifetime in the second training set is used as output data for the second convolutional neural network. The second convolutional neural network is trained to obtain the second initial convolutional neural network. The second initial convolutional neural network is validated using the second test set. The second initial convolutional neural network that outputs a second test error threshold less than or equal to the preset second test error threshold is used as the remaining lifetime prediction model. At each evaluation cycle, the current aging rate, current health score, rate of change of current health score, cumulative material fatigue damage, creep and stress relaxation data, and systemic risk transmission coefficient of each glass plate within the evaluation cycle are input into the remaining life prediction model to obtain the predicted remaining life of each glass plate.

9. The glass curtain wall full life cycle management system according to claim 8, characterized in that, Based on the predicted lifespan, the lifespan gaps of the glass curtain wall are identified. The process of unified inspection and collaborative maintenance of the glass panels in these gaps includes: After obtaining the predicted remaining life of the glass panel through the remaining life prediction model, a life column proportional to its predicted remaining life is generated for each glass panel on a digital twin platform that integrates the building's 3D model for visualization, forming a life topography map of the glass curtain wall. This allows for the intuitive identification of clusters of areas with significantly lower lifespans, which are defined as life depressions. For the identified lifespan depressions, the average remaining lifespan of all glass plates within the depression is calculated, and the core plates of the depression with lifespans lower than the average remaining lifespan are screened out. Time-predicted lifespan curves are plotted for the core plates of the depressions, and the rate of decline and synergy of the curves are analyzed. If multiple curves are found to show a synchronous and significant increase in the rate of decline in a similar time period, it strongly suggests the existence of common pathogenic factors. By calling historical data of the core plate of the depression for correlation mining and providing possible causal hypotheses, a unified investigation and collaborative maintenance plan for the entire lifespan of the depression can be generated. It is recommended to conduct comprehensive non-destructive testing and sampling analysis in the area and formulate comprehensive governance measures including replacement, reinforcement, and protection.

10. A method for managing the entire life cycle of a glass curtain wall, implemented based on the glass curtain wall life cycle management system according to any one of claims 1-9, characterized in that, Includes the following steps: S1: Real-time acquisition of the status parameters of each glass panel of the glass curtain wall through acquisition equipment, and normalization processing of the status parameters; S2: Calculate the real-time health score of each glass plate based on the processed parameters. Calculate the rate of change of the health score of the glass plate based on the continuous health score over time using the sliding window linear regression method. S3: Based on the health score and the rate of change of the health score, the glass panels are accurately classified into different levels, and the glass panels that need maintenance are marked. Using the spatial coordinates of the glass panels that need maintenance as nodes, an initial maintenance path is planned to prioritize the handling of emergency risks. S4: Build and apply an aging rate prediction model to predict the future performance aging rate of each glass panel, calculate the aging time based on the aging rate, compare it with the preset maintenance cycle threshold, mark glass panels with insufficient aging time as glass panels to be maintained, and include them in the preventive maintenance plan. S5: Based on the initial maintenance path, the marked glass panels to be maintained are added as new task points. The glass panels to be maintained are inserted into the existing path to generate an optimized maintenance path that includes both emergency remediation and cost-effective prevention tasks. S6: Construct and apply a remaining life prediction model to predict the remaining safe service life of each glass panel, visualize it in a three-dimensional digital twin model to form a life topography map, automatically identify and analyze life depressions with significantly low remaining life, and intelligently diagnose common root causes of defects by analyzing the synergy of the life curves of the glass panels in the depressions, and generate targeted cluster unified investigation and collaborative maintenance solutions.

Images