An assembled structure health monitoring method and system

Abstract

This invention belongs to the field of building technology and discloses a method and system for health monitoring of prefabricated structures. It includes the following modules: a data acquisition module for real-time acquisition of structural data in the monitoring area. This invention achieves accurate prediction of the remaining lifespan of high-risk areas by constructing a remaining lifespan prediction model, overcoming the limitation of traditional methods that only provide post-event warnings. It combines BIM 3D visualization technology to generate lifespan basin areas, clearly identifying areas of concentrated safety risks. Based on an ant colony algorithm, it generates optimal maintenance routes, balancing repair priorities and construction feasibility, thus improving maintenance efficiency. By setting a repair coefficient prediction model, it enables scientific comparison of repair schemes, ensuring that the selected scheme is both technically feasible and economically reasonable. The predicted aging rate accurately reflects the structural deterioration trend. After repair, the actual repair coefficient is compared with the predicted value to achieve iterative model optimization and scheme database accumulation.

Description

Technical Field

[0001] This invention relates to the field of building technology, and more specifically, to a method and system for health monitoring of prefabricated structures. Background Technology

[0002] With the industrialization and large-scale development of prefabricated buildings, their advantages such as efficient construction and green environmental protection have made them the core direction for the transformation and upgrading of the construction industry. However, prefabricated structures are assembled by connecting prefabricated components through nodes, and their safety performance is highly dependent on the quality of the components themselves and the reliability of the node connections.

[0003] Existing health monitoring technologies for prefabricated structures suffer from several shortcomings: traditional monitoring systems often rely on single mechanical sensors, monitoring only macroscopic parameters such as stress and displacement, failing to simultaneously cover multi-dimensional degradation indicators of components and nodes, leading to one-sided health assessments and a tendency for missed or false diagnoses; health status assessments often employ fixed threshold alarms, providing only post-event alerts and lacking accurate predictions of remaining lifespan, thus failing to provide data support for preventative maintenance; monitoring resource allocation is rigid, with insufficient monitoring frequency in high-risk areas and excessive monitoring in low-risk areas, resulting in low resource utilization efficiency; monitoring and repair are disconnected, failing to link emergency resources after early warnings, lacking a scientific system for comparing repair solutions, and making it difficult to achieve closed-loop management throughout the entire life cycle; existing technologies do not fully consider the unique characteristics of prefabricated structures, such as complex force transmission paths at nodes, the impact of component assembly deviations on structural stress, and the accelerated degradation of node connections due to environmental corrosion, resulting in insufficient targeting of monitoring solutions. Summary of the Invention

[0004] To address the problems in the background art, this invention proposes a method and system for monitoring the health of prefabricated structures.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a prefabricated structural health monitoring system, comprising the following modules: The data acquisition module is used to collect structural data of the monitoring area in real time. The health assessment module is used to generate component health scores and node health scores for the monitoring area based on the collected structural data, calculate the comprehensive health score of the monitoring area based on the component health scores and node health scores, and calculate the rate of change of the comprehensive health score of the monitoring area based on the comprehensive health score. The health grading module is used to classify the monitoring area based on the comprehensive health score and the rate of change of the comprehensive health score, and to take corresponding measures for monitoring areas of different levels. The remaining useful life prediction module is used to acquire historical remaining useful life data and build a remaining useful life prediction model. It uses the remaining useful life prediction model to predict the remaining useful life of high-risk monitoring areas, generates maintenance and repair routes based on the predicted remaining useful life of high-risk monitoring areas, and calculates the expected aging rate of high-risk monitoring areas based on the predicted remaining useful life of high-risk monitoring areas. The repair coefficient prediction module is used to acquire historical repair coefficient data and build a repair coefficient prediction model. The repair coefficient prediction model is used to predict the repair coefficient of high-risk monitoring areas, and the corresponding repair scheme is selected based on the predicted repair coefficient.

[0006] Furthermore, the prefabricated building is divided into multiple monitoring areas that match the prefabricated component units. Corresponding sensing units are deployed in each monitoring area to collect structural data of the prefabricated building. The structural data includes component body data and node connection data. Component body data includes concrete strength decay rate, steel corrosion rate, concrete carbonation depth ratio and component crack width exceeding standard ratio. Node connection data includes connection bolt preload loss rate, node weld defect length ratio, node relative displacement exceeding standard ratio and node interface crack width exceeding standard ratio. Concrete strength is tested on-site using a rebound hammer or a combined ultrasonic rebound method. Combined with long-term data from embedded strain sensors, the concrete strength attenuation rate is calculated as: (Design strength - Measured strength) / Design strength. The concrete strength attenuation rate is then normalized.

[0007] In the formula, To normalize the concrete strength degradation rate, This represents the measured concrete strength attenuation rate. and This represents the ideal optimal value and upper limit of the concrete strength degradation rate; The corrosion rate of steel bars is calculated by monitoring the polarization potential and corrosion current density using pre-embedded steel bar corrosion sensors, and then normalized.

[0008] In the formula, To normalize the steel reinforcement corrosion rate, This represents the measured steel reinforcement corrosion rate. and This represents the ideal optimal value and upper limit of the steel reinforcement corrosion rate; The concrete carbonation depth was obtained through on-site core sampling. The percentage of concrete carbonation depth was calculated as: measured carbonation depth / thickness of the protective layer. The percentage of concrete carbonation depth was then normalized.

[0009] In the formula, To normalize the percentage of concrete carbonation depth, This represents the measured percentage of concrete carbonation depth. and This represents the ideal optimal value for the proportion of concrete carbonation depth and the upper limit of deterioration. The component crack width is obtained through distributed fiber optic crack sensors and image recognition. The component crack width exceeding the standard ratio is calculated as: measured maximum crack width / maximum allowable crack width. The component crack width exceeding the standard ratio is then normalized.

[0010] In the formula, To normalize the percentage of component crack width exceeding the standard, This represents the percentage of component crack widths exceeding the standard, as measured in practice. and This represents the ideal optimal value and upper limit of the component's crack width exceeding the standard ratio; The preload of the connecting bolts is monitored by the preload stress sensor. The preload loss rate of the connecting bolts is calculated as (initial preload - measured preload) / initial preload. The preload loss rate of the connecting bolts is then normalized.

[0011] In the formula, To normalize the preload loss rate of connecting bolts, This represents the measured preload loss rate of the connecting bolts. and The ideal optimal value and upper limit of deterioration for the preload loss rate of connecting bolts; The length of defects in the joint weld was obtained by ultrasonic testing. The proportion of defect length in the joint weld was calculated as the cumulative length of defects divided by the total weld length. The proportion of defect length in the joint weld was then normalized.

[0012] In the formula, To normalize the proportion of defect length in node welds, This represents the percentage of defect length in the measured node weld. and This represents the ideal optimal value and upper limit of the proportion of defect length in the node weld; The relative displacement of the components on both sides of the node is monitored by displacement sensors. The ratio of the relative displacement exceeding the standard is calculated as: maximum measured relative displacement / allowable limit in the specification. The ratio of the relative displacement exceeding the standard is then normalized.

[0013] In the formula, This represents the normalized ratio of relative displacement exceeding the standard at nodes. This represents the measured percentage of nodal relative displacements exceeding the standard. and The ideal optimal value and upper limit of the relative displacement exceeding the standard of the node are defined; The crack width at the interface between new and old concrete is monitored using distributed fiber optic sensors. The ratio of excessive crack width at the joint interface is calculated as: (measured maximum crack width / allowable limit in the specification). The ratio of excessive crack width at the joint interface is then normalized.

[0014] In the formula, This represents the normalized ratio of excessive crack width at the node interface. This represents the percentage of joint interface crack widths exceeding the standard, as measured in practice. and This represents the ideal optimal value and upper limit of the ratio of excessive crack width at the node interface.

[0015] Furthermore, the process of generating component health scores and node health scores for the monitoring area based on the collected structural data includes: A component health score, S, is generated based on the component ontology data.

[0016] In the formula, , , and These are weighting coefficients, obtained through training based on historical data; A node health score is generated based on node connection data. Node health score Q:

[0017] In the formula, , , and These are weighting coefficients, obtained through training based on historical data.

[0018] Furthermore, the process of calculating the comprehensive health score of the monitoring area based on the component health score and the node health score includes: Overall health score P:

[0019] 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 comprehensive health score in the monitored area based on the comprehensive health score includes: A fixed monitoring period is set. After each monitoring period ends, the measured value of the comprehensive health score of the monitored area within that monitoring period is obtained. At the same time, the historical value of the comprehensive health score of the monitored area in the previous monitoring period is retrieved. The measured value of the current period is subtracted from the historical value of the previous period to obtain the absolute change in the comprehensive health score. Then, the absolute change is divided by the historical value of the comprehensive health score of the previous monitoring period to obtain the rate of change of the comprehensive health score of the monitored area in this monitoring period.

[0020] Furthermore, the process of classifying the monitoring area based on the comprehensive health score and the rate of change of the comprehensive health score includes: Set appropriate first and second comprehensive health score thresholds based on historical comprehensive health score data, where the historical comprehensive health score data refers to the dataset of comprehensive health scores of the monitored area in the past. Set an appropriate threshold for the rate of change of the comprehensive health score based on historical comprehensive health score change data, where the historical comprehensive health score change data refers to the data set of the rate of change of the comprehensive health score in the monitored area in the past. Low-risk monitoring area determination: Two conditions must be met simultaneously: first, the comprehensive health score of the monitoring area is greater than the first comprehensive health score threshold; second, the change rate of the comprehensive health score of the monitoring area is greater than or equal to the change rate threshold of the comprehensive health score. Both conditions must be met at the same time, and neither can be missing. Low-risk monitoring areas shall be subject to routine inspections and data monitoring, and do not need to be included in the repair plan.

[0021] Determination of medium-risk monitoring areas: A monitoring area can be determined as medium-risk if it meets any of the following conditions: Condition 1: The comprehensive health score of the monitoring area is within the first comprehensive health score threshold and the second comprehensive health score threshold, regardless of the value of the comprehensive health score change rate; Condition 2: The comprehensive health score of the monitoring area is greater than the first comprehensive health score threshold, but the comprehensive health score change rate is less than the comprehensive health score change rate threshold. Medium-risk monitoring areas are included in the quarterly maintenance plan and preventive maintenance and deterioration suppression measures are implemented.

[0022] High-risk monitoring area determination: If the comprehensive health score of the monitoring area is less than the second comprehensive health score threshold, it is determined to be a high-risk monitoring area. The emergency response for high-risk monitoring areas shall be activated immediately, the normal use of the corresponding area shall be suspended, and the remaining life prediction and special repair plan design shall be carried out simultaneously.

[0023] 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 high-risk monitoring areas, includes: The remaining life of high-risk monitoring areas is predicted. The remaining life refers to the remaining time before the comprehensive health score of a high-risk monitoring area drops to the danger threshold. The danger threshold is a pre-set threshold. High-risk monitoring areas with a comprehensive health score ≤ the danger threshold have reached the level of damage and collapse. Factors affecting the remaining lifespan of high-risk monitoring areas include: current comprehensive health score, rate of change of current comprehensive health score, environmental corrosion level, percentage of design service life already used, load exceedance rate, and crack propagation rate. The corrosion level of the environment refers to the classification of 1-5 (Level 1 is slightly corrosive, Level 5 is strongly corrosive) according to GB / T50046-2018 and environmental monitoring data. Percentage of designed service life already used = Years of building service life / Design service life of components; The load exceedance rate is calculated as: measured maximum load effect / design load limit, based on actual stress monitoring using strain sensors. The crack propagation rate, as monitored by distributed fiber optic sensors, is calculated as: crack width increment during the monitoring period / monitoring period duration. The historical remaining lifespan data of a single high-risk monitoring area in different monitoring cycles is obtained. The historical remaining lifespan data includes the current comprehensive health score, the rate of change of the current comprehensive health score, the corrosion level of the environment, the percentage of the design service life already used, the load exceeding the standard rate, the crack propagation rate, and the historical remaining lifespan of the single high-risk monitoring area in different monitoring cycles. Based on the current comprehensive health score, current comprehensive health score change rate, environmental corrosion level, proportion of design service life already used, load exceedance rate, crack propagation rate and corresponding historical remaining life of the high-risk monitoring area in different historical remaining life data, a remaining life 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 comprehensive health score, the rate of change of the current comprehensive health score, the corrosion level of the environment, the proportion of the design service life already used, the load exceedance rate, and the crack propagation speed from different historical remaining life data in the first training set as the input data of the first convolutional neural network, and taking the corresponding historical remaining life 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 a preset first test error threshold is used as the remaining lifetime prediction model. For each monitoring cycle, the current comprehensive health score, the rate of change of the current comprehensive health score, the corrosion level of the environment, the percentage of the design service life already used, the load exceedance rate and the crack propagation rate of each high-risk monitoring area are input into the remaining life prediction model to obtain the predicted remaining life of each high-risk monitoring area. The first convolutional neural network is a one-dimensional convolutional neural network, which includes an input layer, two convolutional layers, a pooling layer, two fully connected layers, and an output layer. The convolutional layer filter size is 3×1, and the number of filters is 32 and 16 respectively. The activation function is the ReLU function. The number of neurons in the fully connected layer is 64 and 16 respectively. The output layer uses a linear activation function. The network training hyperparameters are: learning rate 0.001, batch size 16, and the Adam optimizer. Training uses L2 regularization, Dropout layer, and early stopping mechanism to prevent overfitting. The training termination condition is that the mean absolute percentage error of the test set is ≤5%.

[0024] Furthermore, the process of generating maintenance and repair routes based on the predicted remaining lifespan of high-risk monitoring areas includes: Using the BIM 3D information model corresponding to the prefabricated building as the visualization base map, the spatial coordinates of the monitoring area and the BIM model components are matched one by one. The coordinates of the spatial center point of each monitoring area are aligned with the coordinates of the geometric center point of the corresponding prefabricated component in the BIM model to ensure that the spatial mapping accuracy error between the monitoring data and the 3D model does not exceed 5cm. After calibration, a columnar visualization model, known as a life column, is generated at the 3D coordinate position of each monitoring area, perpendicular to the horizontal plane of the building. The height of the life column is linearly positively correlated with the predicted remaining life of the corresponding high-risk monitoring area. The height of the life column is set to 0.5m / year per unit of remaining life, meaning that the height of the life column decreases by 0.5m for every year the predicted remaining life decreases. At the same time, the fill color of the life column gradually changes from green, yellow, orange to red as the predicted remaining life decreases. The diameter of the life column is negatively correlated with the comprehensive health score of the corresponding high-risk monitoring area. The lower the comprehensive health score, the larger the diameter of the life column. After generating life columns for all high-risk monitoring areas of the entire building, continuous low-lying areas are identified in the 3D map. The threshold for determining life basins is set as a predicted remaining life of ≤3 years. When three or more spatially adjacent monitoring areas have a predicted remaining life of ≤3 years and form a continuous spatial area, the continuous area is marked as a life basin area. This life basin area is a cluster of structural safety risks in prefabricated buildings, with the risk of cascading failures of components and nodes. It is listed as the highest priority in the repair priority, followed by independent high-risk monitoring areas, and finally medium-risk monitoring areas. When generating maintenance and repair routes, the route starts from the pre-set entrances for maintenance personnel and equipment within the building. Each high-risk monitoring area within the lifespan basin area is designated as a primary core node, independent high-risk monitoring areas as secondary nodes, and medium-risk monitoring areas as tertiary nodes. Path planning is performed based on the ant colony algorithm. Three constraints are incorporated into the path planning process: first, the route must avoid normal use areas, non-safe structural areas, and areas with dense electromechanical pipelines within the building; second, the route must connect nodes in descending order of priority; and third, nodes of the same priority are connected by path concatenation based on proximity. After generating the initial shortest path, the route is further optimized by considering the equipment transportation requirements and personnel workspace requirements of the maintenance construction. Path segments that cannot meet the construction access requirements are eliminated, and the optimal maintenance and repair route is finally generated. Simultaneously, this maintenance route is mapped onto a 3D map and displayed in conjunction with the lifespan basin area and monitoring areas of each risk level. The process of calculating the projected aging rate of a high-risk monitoring area based on its predicted remaining lifespan includes: For high-risk monitoring areas, first retrieve the pre-set danger threshold, the measured comprehensive health score of the current monitoring period, and the predicted remaining lifespan output by the remaining lifespan prediction model for the monitoring area. First, calculate the total decline in the health score of the high-risk monitoring area from the current health status to the danger threshold. The total decline = the measured value of the current comprehensive health score - the danger threshold. Then, divide the total decline by the predicted remaining lifespan of the high-risk monitoring area to obtain the decrease in the comprehensive health score of the monitoring area per unit time under natural deterioration without external intervention. This is the expected aging rate of the high-risk monitoring area.

[0025] Furthermore, the process of acquiring historical repair coefficient data and constructing a repair coefficient prediction model, and then using this model to predict the repair coefficient of high-risk monitoring areas, includes: The repair coefficient refers to the degree of repair effect of the high-risk monitoring area after the repair task is completed. The repair task refers to the maintenance and repair of the high-risk monitoring area. The repair coefficient ranges from [0, 1]. The closer it is to 1, the better the repair effect, the greater the improvement in the comprehensive health score after repair, and the more lasting the deterioration inhibition effect. Factors affecting the repair coefficient include: the current comprehensive health score of the high-risk monitoring area, the expected aging rate, the type of core deterioration items, the technical level of the pre-designed repair plan, the expected load changes after repair, the qualifications of the construction unit and its experience in similar projects, and the performance level and design service life of the repair materials. The core degradation item type refers to the item with the lowest value in the identification scoring system, which is divided into component body type, node connection type, and durability type. The pre-defined technical level of the repair plan refers to the four levels classified according to the depth of treatment: emergency response, routine repair, reinforcement and restoration, and replacement and reconstruction. By combining the building's future use plan, the ratio of the expected load to the design load is calculated to obtain the expected load change after the renovation. The construction unit's qualifications and experience in similar projects refer to a quantitative score based on the construction unit's qualification level and the number and pass rate of similar repair projects in the past 5 years. The performance grade and design service life of repair materials refer to the determination of the material performance grade and design service life based on national material standards and test reports; Obtain historical repair coefficient data for a single high-risk monitoring area within the repair task. The historical repair coefficient data includes the current comprehensive health score, expected aging rate, core deterioration type, technical level of the preset repair plan, expected load change after repair, construction unit qualifications and experience in similar projects, performance level and design service life of repair materials, and historical repair coefficient of the single high-risk monitoring area within different repair tasks. Based on the current comprehensive health score, expected aging rate, core deterioration type, technical level of the preset repair plan, expected load change after repair, construction unit qualifications and experience in similar projects, performance level and design service life of repair materials, and corresponding historical repair coefficients of high-risk monitoring areas in different historical repair coefficient data, a repair coefficient prediction set is generated and divided into a second training set and a second test set. A second convolutional neural network is constructed. The current comprehensive health score, expected aging rate, core deterioration item type, preset repair scheme technical level, expected load change after repair, construction unit qualification and experience in similar projects, and repair material performance level and design service life are taken from the different historical repair coefficient data in the second training set as the input data of the second convolutional neural network. The corresponding historical repair coefficients in the second training set are taken 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 repair coefficient prediction model. The current comprehensive health score, expected aging rate, core deterioration type, pre-designed repair plan technical level, expected load change after repair, construction unit qualification and experience in similar projects, and repair material performance level and design service life of each high-risk monitoring area are input into the repair coefficient prediction model to obtain the predicted repair coefficient for each high-risk monitoring area. The second convolutional neural network is a one-dimensional convolutional neural network, which includes an input layer, three convolutional layers, a pooling layer, two fully connected layers, and an output layer. The filter size of the convolutional layers is 3×1, and the number of filters is 64, 32, and 16 respectively. The activation function is the ReLU function. The number of neurons in the fully connected layers is 128 and 32 respectively. The output layer uses the Sigmoid activation function. The network training hyperparameters are: learning rate 0.0005, batch size 32, and the Adam optimizer. Training uses L2 regularization, Dropout layers, and an early stopping mechanism to prevent overfitting. The training termination condition is that the mean absolute percentage error of the test set is ≤8%.

[0026] Furthermore, the process of selecting the corresponding repair scheme based on the predicted repair coefficient includes: For a single high-risk monitoring area, three or more differentiated repair alternative plans are designed based on the type of core deterioration item, the level of structural importance, and on-site construction conditions. Each alternative plan has clearly defined core parameters such as the current comprehensive health score, the expected aging rate, the type of core deterioration item, the technical level of the preset repair plan, the expected load change after repair, the qualifications of the construction unit and experience in similar projects, and the performance level and design service life of the repair materials. For each alternative repair plan, the above core parameters are input into the repair coefficient prediction model to calculate the predicted repair coefficient corresponding to each alternative plan. After predicting the repair coefficients of all alternative schemes, a repair threshold is preset. Alternative schemes with predicted repair coefficients lower than the repair threshold are eliminated. These schemes cannot achieve the basic repair goal of raising the comprehensive health score of the high-risk monitoring area to above the first comprehensive health score threshold and are not feasible for engineering application. For the remaining alternative schemes with predicted repair coefficients ≥ the repair threshold, the ability to achieve the repair effect is first quantitatively calculated. The expected health score improvement value of each scheme is calculated using the formula: Expected health score improvement value = (1 - current comprehensive health score) × predicted repair coefficient. At the same time, the expected health score compliance rate is calculated using the formula: Expected health score compliance rate = (current comprehensive health score + expected health score improvement value) / repair threshold × 100%. Schemes with an expected health score compliance rate < 100% are eliminated simultaneously to ensure that the remaining alternative schemes can meet the core requirement that the comprehensive health score after repair exceeds the repair threshold. Complete data, including the scheme parameters, actual repair coefficients, and repair effects of this renovation project, were added to the training dataset of the repair coefficient prediction model. The model was iteratively optimized and trained. Every 10 sets of newly added valid repair case data were completed, and the model weights were updated to continuously improve the prediction accuracy of the model. At the same time, the verified and qualified repair schemes were stored in the scheme database to provide data support for the design of repair schemes for similar prefabricated building components and similar deterioration problems.

[0027] A method for health monitoring of prefabricated structures includes the following steps: S1: Divide the prefabricated building into monitoring areas that match the prefabricated component units. Deploy corresponding sensing units in each area to realize the real-time acquisition of component body data and node connection data, and perform normalization processing on the data. S2: Based on normalized data, the component health score and node health score are calculated using a weighted summation formula, and then the comprehensive health score is calculated. At the same time, the change rate of the comprehensive health score is calculated by the difference between the scores of the previous and subsequent periods. S3: Set the first comprehensive health score threshold, the second comprehensive health score threshold, and the comprehensive health score change rate threshold to divide the monitoring area into low, medium, and high-risk monitoring areas, and match them with control strategies for routine inspection, preventive maintenance, and emergency response, respectively. The use of the corresponding area in the high-risk monitoring area is suspended. S4: Construct a convolutional neural network remaining lifetime prediction model to predict the remaining lifetime of high-risk areas. Use the BIM model as the base map to generate life columns and life basin areas. Based on the ant colony algorithm, generate the optimal maintenance and repair route to avoid unsafe areas. S5: Construct a repair coefficient prediction model to predict the effects of different repair schemes, eliminate schemes with scores below the repair threshold, select the optimal scheme and implement it, calculate the actual repair coefficient after the repair is completed, compare it with the predicted value, iteratively optimize the model and update the scheme database.

[0028] The technical effects and advantages of the prefabricated structure health monitoring method and system of the present invention are as follows: (1) By constructing a remaining life prediction model, the remaining life of high-risk areas can be accurately predicted, which solves the limitation of traditional schemes that can only provide alarms after the fact. By combining BIM three-dimensional visualization technology to generate life basin areas, the safety risk cluster areas are identified. The optimal maintenance route is generated based on the ant colony algorithm, taking into account the repair priority and construction feasibility, which improves the efficiency of maintenance operations. By setting a repair coefficient prediction model, the repair schemes can be scientifically compared and selected to ensure that the selected schemes are both technically feasible and economically reasonable. The aging rate is accurately reflected in the trend of structural deterioration. After the repair is completed, the model is iteratively optimized and the scheme database is accumulated by comparing the actual repair coefficient with the predicted value, which effectively suppresses the rate of structural deterioration, extends the remaining life of high-risk monitoring areas, significantly reduces the safety risks and operation and maintenance costs of prefabricated buildings, and has significant economic and social benefits.

[0029] (2) By constructing a dual-dimensional health scoring system for components and nodes, a quantitative assessment of the structural health status is achieved, which solves the problem of the one-sided assessment of traditional monitoring schemes. Combined with clear classification thresholds, the monitoring area is accurately divided into low, medium and high risks. The accuracy of hazard identification is improved compared with traditional schemes. By verifying the risk coupling of adjacent areas and considering the force transmission characteristics of prefabricated structures, misjudgments caused by isolated risk assessments are avoided, making the identification of high-risk areas more comprehensive. Low, medium and high-risk areas are matched with control strategies of routine inspection, preventive maintenance and emergency response, respectively, to achieve risk classification control. The response time of structural safety hazards is greatly reduced, and the pertinence and timeliness of structural safety control are improved, providing data support for the full life cycle safety management of prefabricated buildings. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the system of the present invention; Figure 2 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0031] 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.

[0032] Reference Figure 1 A prefabricated structural health monitoring system includes the following modules: The data acquisition module is used to collect structural data of the monitoring area in real time. The health assessment module is used to generate component health scores and node health scores for the monitoring area based on the collected structural data, calculate the comprehensive health score of the monitoring area based on the component health scores and node health scores, and calculate the rate of change of the comprehensive health score of the monitoring area based on the comprehensive health score. The health grading module is used to classify the monitoring area based on the comprehensive health score and the rate of change of the comprehensive health score, and to take corresponding measures for monitoring areas of different levels. The remaining useful life prediction module is used to acquire historical remaining useful life data and build a remaining useful life prediction model. It uses the remaining useful life prediction model to predict the remaining useful life of high-risk monitoring areas, generates maintenance and repair routes based on the predicted remaining useful life of high-risk monitoring areas, and calculates the expected aging rate of high-risk monitoring areas based on the predicted remaining useful life of high-risk monitoring areas. The repair coefficient prediction module is used to acquire historical repair coefficient data and build a repair coefficient prediction model. The repair coefficient prediction model is used to predict the repair coefficient of high-risk monitoring areas, and the corresponding repair scheme is selected based on the predicted repair coefficient.

[0033] It should be further explained that, in the specific implementation process, the prefabricated building is divided into multiple monitoring areas that match the prefabricated component units. Each monitoring area is equipped with a corresponding sensing unit to collect the structural data of the prefabricated building. The structural data includes component body data and node connection data. The component body data includes concrete strength decay rate, steel corrosion rate, concrete carbonation depth ratio and component crack width exceeding standard ratio. The node connection data includes connection bolt preload loss rate, node weld defect length ratio, node relative displacement exceeding standard ratio and node interface crack width exceeding standard ratio. Concrete strength is tested on-site using a rebound hammer or a combined ultrasonic rebound method. Combined with long-term data from embedded strain sensors, the concrete strength attenuation rate is calculated as: (Design strength - Measured strength) / Design strength. The concrete strength attenuation rate is then normalized.

[0034] In the formula, To normalize the concrete strength degradation rate, This represents the measured concrete strength attenuation rate. and These represent the ideal and optimal value for the concrete strength degradation rate and the upper limit for deterioration, specifically 0% and 50%, respectively. The corrosion rate of steel bars is calculated by monitoring the polarization potential and corrosion current density using pre-embedded steel bar corrosion sensors, and then normalized.

[0035] In the formula, To normalize the steel reinforcement corrosion rate, This represents the measured steel reinforcement corrosion rate. and The ideal and optimal values ​​for steel corrosion rate and the upper limit for deterioration are 0% and 20%, respectively. The concrete carbonation depth was obtained through on-site core sampling. The percentage of concrete carbonation depth was calculated as: measured carbonation depth / thickness of the protective layer. The percentage of concrete carbonation depth was then normalized.

[0036] In the formula, To normalize the percentage of concrete carbonation depth, This represents the measured percentage of concrete carbonation depth. and The ideal and optimal values ​​for the percentage of concrete carbonation depth and the upper limit for deterioration are 0% and 100%, respectively. The component crack width is obtained through distributed fiber optic crack sensors and image recognition. The component crack width exceeding the standard ratio is calculated as: measured maximum crack width / maximum allowable crack width. The component crack width exceeding the standard ratio is then normalized.

[0037] In the formula, To normalize the percentage of component crack width exceeding the standard, This represents the percentage of component crack widths exceeding the standard, as measured in practice. and The ideal optimal value and the upper limit of deterioration for the component crack width exceeding the standard ratio are 0 and 2, respectively. The preload of the connecting bolts is monitored by the preload stress sensor. The preload loss rate of the connecting bolts is calculated as (initial preload - measured preload) / initial preload. The preload loss rate of the connecting bolts is then normalized.

[0038] In the formula, To normalize the preload loss rate of connecting bolts, This represents the measured preload loss rate of the connecting bolts. and The ideal and optimal values ​​for the preload loss rate of the connecting bolts are 0% and 80%, respectively. The length of defects in the joint weld was obtained by ultrasonic testing. The proportion of defect length in the joint weld was calculated as the cumulative length of defects divided by the total weld length. The proportion of defect length in the joint weld was then normalized.

[0039] In the formula, To normalize the proportion of defect length in node welds, This represents the percentage of defect length in the measured node weld. and The ideal and optimal value for the proportion of defect length in the node weld and the upper limit for deterioration are 0% and 50%, respectively. The relative displacement of the components on both sides of the node is monitored by displacement sensors. The ratio of the relative displacement exceeding the standard is calculated as: maximum measured relative displacement / allowable limit in the specification. The ratio of the relative displacement exceeding the standard is then normalized.

[0040] In the formula, This represents the normalized ratio of relative displacement exceeding the standard at nodes. This represents the measured percentage of nodal relative displacements exceeding the standard. and The ideal optimal value and the upper limit of the deterioration of the relative displacement of the node are 0 and 3, respectively; The crack width at the interface between new and old concrete is monitored using distributed fiber optic sensors. The ratio of excessive crack width at the joint interface is calculated as: (measured maximum crack width / allowable limit in the specification). The ratio of excessive crack width at the joint interface is then normalized.

[0041] In the formula, This represents the normalized ratio of excessive crack width at the node interface. This represents the percentage of joint interface crack widths exceeding the standard, as measured in practice. and The ideal optimal value and the upper limit of the degradation of the crack width exceeding the standard at the node interface are 0 and 2.5, respectively.

[0042] It should be further explained that, in the specific implementation process, the process of generating component health scores and node health scores for the monitoring area based on the collected structural data includes: A component health score, S, is generated based on the component ontology data.

[0043] In the formula, , , and These are weighting coefficients, obtained from training on historical data, and set to 0.33, 0.27, 0.2, and 0.2 respectively. If the component data for a certain monitoring area is: It is 15%. It is 8%. It is 40%. If the value is 0.8, then the component health score S of the monitored area is 0.633; A node health score is generated based on node connection data. Node health score Q:

[0044] In the formula, , , and These are weighting coefficients, derived from historical data training, and set to 0.3, 0.25, 0.25, and 0.2 respectively. If the node connection data for a certain monitoring area is: 30%, 10%, It is 1.2. If the value is 0.5, then the node health score Q of the monitored area is 0.6975.

[0045] It should be further explained that, in the specific implementation process, the process of calculating the comprehensive health score of the monitoring area based on the component health score and the node health score includes: Overall health score P:

[0046] In the formula, and These are weighting coefficients, derived from historical data training, and set to 0.4 and 0.6 respectively; If the health score S of a certain monitoring area is 0.633 and the node health score Q is 0.6975, then the comprehensive health score P of the monitoring area is 0.6717. The process of calculating the rate of change of the comprehensive health score in the monitored area based on the comprehensive health score includes: A fixed monitoring period is set. After each monitoring period ends, the measured value of the comprehensive health score of the monitored area within that monitoring period is obtained. At the same time, the historical value of the comprehensive health score of the monitored area in the previous monitoring period is retrieved. The measured value of the current period is subtracted from the historical value of the previous period to obtain the absolute change in the comprehensive health score. Then, the absolute change is divided by the historical value of the comprehensive health score of the previous monitoring period to obtain the rate of change of the comprehensive health score of the monitored area in this monitoring period.

[0047] It should be further explained that, in the specific implementation process, the process of classifying the monitoring area based on the comprehensive health score and the rate of change of the comprehensive health score includes: Set appropriate first and second comprehensive health score thresholds based on historical comprehensive health score data, where the historical comprehensive health score data refers to the dataset of comprehensive health scores of the monitored area in the past. Set an appropriate threshold for the rate of change of the comprehensive health score based on historical comprehensive health score change data, where the historical comprehensive health score change data refers to the data set of the rate of change of the comprehensive health score in the monitored area in the past. Low-risk monitoring area determination: Two conditions must be met simultaneously: first, the comprehensive health score of the monitoring area is greater than the first comprehensive health score threshold; second, the change rate of the comprehensive health score of the monitoring area is greater than or equal to the change rate threshold of the comprehensive health score. Both conditions must be met at the same time, and neither can be missing. Low-risk monitoring areas shall be subject to routine inspections and data monitoring, and do not need to be included in the repair plan.

[0048] Determination of medium-risk monitoring areas: A monitoring area can be determined as medium-risk if it meets any of the following conditions: Condition 1: The comprehensive health score of the monitoring area is within the first comprehensive health score threshold and the second comprehensive health score threshold, regardless of the value of the comprehensive health score change rate; Condition 2: The comprehensive health score of the monitoring area is greater than the first comprehensive health score threshold, but the comprehensive health score change rate is less than the comprehensive health score change rate threshold. Medium-risk monitoring areas are included in the quarterly maintenance plan and preventive maintenance and deterioration suppression measures are implemented.

[0049] High-risk monitoring area determination: If the comprehensive health score of the monitoring area is less than the second comprehensive health score threshold, it is determined to be a high-risk monitoring area. The emergency response for high-risk monitoring areas shall be activated immediately, the normal use of the corresponding area shall be suspended, and the remaining life prediction and special repair plan design shall be carried out simultaneously.

[0050] 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 high-risk monitoring areas, includes: The remaining life of high-risk monitoring areas is predicted. The remaining life refers to the remaining time before the comprehensive health score of a high-risk monitoring area drops to the danger threshold. The danger threshold is a pre-set threshold. High-risk monitoring areas with a comprehensive health score ≤ the danger threshold have reached the level of damage and collapse. Factors affecting the remaining lifespan of high-risk monitoring areas include: current comprehensive health score, rate of change of current comprehensive health score, environmental corrosion level, percentage of design service life already used, load exceedance rate, and crack propagation rate. The corrosion level of the environment refers to the classification of 1-5 (Level 1 is slightly corrosive, Level 5 is strongly corrosive) according to GB / T50046-2018 and environmental monitoring data. Percentage of designed service life already used = Years of building service life / Design service life of components; The load exceedance rate is calculated as: measured maximum load effect / design load limit, based on actual stress monitoring using strain sensors. The crack propagation rate, as monitored by distributed fiber optic sensors, is calculated as: crack width increment during the monitoring period / monitoring period duration. The historical remaining lifespan data of a single high-risk monitoring area in different monitoring cycles is obtained. The historical remaining lifespan data includes the current comprehensive health score, the rate of change of the current comprehensive health score, the corrosion level of the environment, the percentage of the design service life already used, the load exceeding the standard rate, the crack propagation rate, and the historical remaining lifespan of the single high-risk monitoring area in different monitoring cycles. Based on the current comprehensive health score, current comprehensive health score change rate, environmental corrosion level, proportion of design service life already used, load exceedance rate, crack propagation rate and corresponding historical remaining life of the high-risk monitoring area in different historical remaining life data, a remaining life 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 comprehensive health score, the rate of change of the current comprehensive health score, the corrosion level of the environment, the proportion of the design service life already used, the load exceedance rate, and the crack propagation speed from different historical remaining life data in the first training set as the input data of the first convolutional neural network, and taking the corresponding historical remaining life in the first training set as the output data of the first convolutional neural network. The first convolutional neural network is a one-dimensional convolutional neural network adapted to one-dimensional temporal features. The overall network structure is set along the data transmission direction as follows: input layer, first convolutional layer, second convolutional layer, max pooling layer, first fully connected layer, second fully connected layer, and output layer. The input layer has a dimension of 6×1, corresponding to 6 input feature parameters. The first convolutional layer has 32 filters with a size of 3×1, a stride of 1, and uses the same padding method. The activation function is ReLU. The second convolutional layer has 16 filters with a size of 3×1, a stride of 1, and uses the same padding method. The activation function is ReLU. The max pooling layer has a pooling window size of 2×1 and a pooling stride of 2. The first fully connected layer has 64 neurons and uses the ReLU activation function. The second fully connected layer has 16 neurons and uses the ReLU activation function. The output layer has 1 neuron, corresponding to the remaining lifetime prediction value, and uses a linear activation function. The hyperparameters for network training are specifically set as follows: initial learning rate 0.001, batch size 16, maximum number of iterations 200, Adam adaptive moment estimation optimizer, mean squared error (MSE) loss function, and three measures to prevent overfitting during training: first, a Dropout layer is set after each of the two fully connected layers with a random dropout probability of 0.3; second, L2 regularization is applied to the weights of the convolutional and fully connected layers with a regularization coefficient of 0.0001; and third, an early stopping mechanism is set, which terminates training early when the test set loss value does not decrease for 10 consecutive iterations to avoid overfitting. The first test error threshold is the mean absolute percentage error (MAPE), with a threshold set to ≤5%. This threshold is determined based on the engineering accuracy requirements for predicting the remaining life of prefabricated structures. When the mean absolute percentage error of the test set prediction results is ≤5%, the model is deemed to meet the accuracy requirements for engineering applications. 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 a preset first test error threshold is used as the remaining lifetime prediction model. For each monitoring cycle, the current comprehensive health score, the rate of change of the current comprehensive health score, the corrosion level of the environment, the percentage of the design service life already used, the load exceedance rate and the crack propagation rate of each high-risk monitoring area are input into the remaining life prediction model to obtain the predicted remaining life of each high-risk monitoring area. In an embodiment of the present invention, the predicted remaining life of all high-risk monitoring areas is obtained through a remaining life prediction model. The predicted remaining life is related to the current comprehensive health score, the rate of change of the current comprehensive health score, the corrosion level of the environment, the proportion of the design service life already used, the load exceeding the standard rate, and the crack propagation speed. The current comprehensive health score directly affects the predicted life expectancy. The higher the current comprehensive health score, the longer the predicted life expectancy. Therefore, the current comprehensive health score is positively correlated with life expectancy. The rate of change of the current comprehensive health score directly affects the predicted life expectancy. The higher the rate of change of the current comprehensive health score, the slower the deterioration rate and the longer the predicted life expectancy. Therefore, the rate of change of the current comprehensive health score is positively correlated with life expectancy. The level of corrosion in the environment directly affects the predicted remaining life. The higher the level of corrosion in the environment, the more severe the corrosion of the building, and the shorter the predicted remaining life. Therefore, the level of corrosion in the environment is negatively correlated with the remaining life. The proportion of the designed service life that has been used directly affects the length of the predicted remaining life. The larger the proportion of the designed service life that has been used, the shorter the predicted remaining life. Therefore, the proportion of the designed service life that has been used is negatively correlated with the remaining life. The magnitude of the load exceedance rate directly affects the predicted remaining life. The higher the load exceedance rate, the shorter the predicted remaining life. Therefore, the load exceedance rate and the remaining life are negatively correlated. The speed of crack propagation directly affects the predicted remaining life. The faster the crack propagation, the shorter the predicted remaining life. Therefore, the load exceedance rate is negatively correlated with the remaining life.

[0051] It should be further explained that, in the specific implementation process, the process of generating maintenance and repair routes based on the predicted remaining lifespan of high-risk monitoring areas includes: Using the BIM 3D information model corresponding to the prefabricated building as the visualization base map, the spatial coordinates of the monitoring area and the BIM model components are matched one by one. The coordinates of the spatial center point of each monitoring area are aligned with the coordinates of the geometric center point of the corresponding prefabricated component in the BIM model to ensure that the spatial mapping accuracy error between the monitoring data and the 3D model does not exceed 5cm. After calibration, a columnar visualization model, known as a life column, is generated at the 3D coordinate position of each monitoring area, perpendicular to the horizontal plane of the building. The height of the life column is linearly positively correlated with the predicted remaining life of the corresponding high-risk monitoring area. The height of the life column is set to 0.5m / year per unit of remaining life, meaning that the height of the life column decreases by 0.5m for every year the predicted remaining life decreases. At the same time, the fill color of the life column gradually changes from green, yellow, orange to red as the predicted remaining life decreases. The diameter of the life column is negatively correlated with the comprehensive health score of the corresponding high-risk monitoring area. The lower the comprehensive health score, the larger the diameter of the life column. After generating life columns for all high-risk monitoring areas of the entire building, continuous low-lying areas are identified in the 3D map. The threshold for determining life basins is set as a predicted remaining life of ≤3 years. When three or more spatially adjacent monitoring areas have a predicted remaining life of ≤3 years and form a continuous spatial area, the continuous area is marked as a life basin area. This life basin area is a cluster of structural safety risks in prefabricated buildings, with the risk of cascading failures of components and nodes. It is listed as the highest priority in the repair priority, followed by independent high-risk monitoring areas, and finally medium-risk monitoring areas. When generating maintenance and repair routes, the pre-set entrances for maintenance personnel and equipment within the building are used as the starting points. Each high-risk monitoring area within the lifespan basin is designated as a primary core node, independent high-risk monitoring areas as secondary nodes, and medium-risk monitoring areas as tertiary nodes. Path planning is based on the ant colony algorithm, incorporating three constraints during the path planning process. These constraints are transformed into heuristic functions and path penalty terms for the ant colony algorithm. The specific implementation is as follows: The core parameters of the ant colony algorithm are preset as follows: the pheromone heuristic factor α is set to a value range of [1, 2], the expected heuristic factor β is set to a value range of [3, 5], the pheromone volatility coefficient ρ is set to a value range of [0.1, 0.3], the pheromone intensity Q is set to a value range of [50, 200], the ant colony size m is set to a value range of [20, 50], and the maximum number of iterations i is set to a value range of [100, 300]. Algorithm transformation of constraints: ① Restricted area constraints: Mark the normal use areas, non-safety structure areas, and areas with dense electromechanical pipelines within the building as restricted grids, construct a 3D access grid map of the building, and if a direct path between any two nodes crosses a restricted grid, then the heuristic function value of that path segment is... Set it to 0, and simultaneously set the pheromone concentration of this path segment. ① Permanently set to 0 to enforce avoidance of restricted areas; ② Priority constraint: Set a mandatory rule for node access order. During path search, ants must complete the access of all first-level core nodes before accessing second-level nodes, and complete the access of all second-level nodes before accessing third-level nodes. If an ant does not access nodes in this order, a penalty coefficient C=10^6 is applied to the path, multiplying the objective function value of the path by the penalty coefficient, causing it to be eliminated in the iteration; ③ Same-priority proximity constraint: For nodes with the same priority, construct a three-dimensional Euclidean distance matrix between nodes. The heuristic function values ​​between nodes of the same priority are... Set to 1 / This causes ants to prioritize connecting paths to nodes of the same priority that are closer in distance. Initial shortest path solution: Using the shortest total path length as the objective function, execute the ant colony algorithm iteratively, and after each iteration, proceed as follows: (t+1) = (1-ρ) × (t) + Δτ updates the path pheromone concentration across the entire map, where Δ The total amount of pheromone released by the ant on path segment ij during this iteration; after reaching the maximum number of iterations, the path with the smallest objective function value is the initial shortest path. After generating the initial shortest path, the path is further optimized based on the equipment transportation and personnel workspace requirements of the renovation work. This secondary optimization employs a path pruning and node reconnection algorithm based on accessibility verification. The specific logic is as follows: First, preset access constraint thresholds: Based on the maximum equipment width and minimum safe working distance for the renovation work, set thresholds for path access width, path turning radius, and path vertical slope. Second, path segment accessibility verification: For each continuous segment of the initial shortest path, retrieve the corresponding 3D spatial parameters from the BIM model, and verify that the actual path access width is greater than or equal to the path access width threshold, the path turning radius is greater than or equal to the path turning radius threshold, and the path vertical slope is less than or equal to the path vertical slope threshold. The first step is to mark path segments that do not meet any constraint as invalid path segments; the second step is to eliminate and reconnect invalid path segments: eliminate all invalid path segments, take the first and last nodes of the invalid path segments as the start and end points, and re-execute the ant colony algorithm path search in the 3D access grid map. During the search, only path segments that meet the access constraint threshold are retained, and alternative path segments are generated and connected to the original path to complete the path reconstruction; the third step is to verify and optimize the global path: for the reconstructed complete path, the node access priority order and the requirements for avoiding prohibited areas are re-verified to ensure that the path meets all constraints. Finally, the optimal maintenance and repair route is generated, and the maintenance route is synchronously mapped to the 3D map and displayed in conjunction with the life basin area and the monitoring areas of each risk level; The process of calculating the projected aging rate of a high-risk monitoring area based on its predicted remaining lifespan includes: For high-risk monitoring areas, first retrieve the pre-set danger threshold, the measured comprehensive health score of the current monitoring period, and the predicted remaining lifespan output by the remaining lifespan prediction model for the monitoring area. First, calculate the total decline in the health score of the high-risk monitoring area from the current health status to the danger threshold. The total decline = the measured value of the current comprehensive health score - the danger threshold. Then, divide the total decline by the predicted remaining lifespan of the high-risk monitoring area to obtain the decrease in the comprehensive health score of the monitoring area per unit time under natural deterioration without external intervention. This is the expected aging rate of the high-risk monitoring area.

[0052] It should be further explained that, in the specific implementation process, the process of obtaining historical repair coefficient data and constructing a repair coefficient prediction model, and then using the repair coefficient prediction model to predict the repair coefficient of high-risk monitoring areas, includes: The repair coefficient refers to the degree of repair effect of the high-risk monitoring area after the repair task is completed. The repair task refers to the maintenance and repair of the high-risk monitoring area. The repair coefficient ranges from [0, 1]. The closer it is to 1, the better the repair effect, the greater the improvement in the comprehensive health score after repair, and the more lasting the deterioration inhibition effect. Factors affecting the repair coefficient include: the current comprehensive health score of the high-risk monitoring area, the expected aging rate, the type of core deterioration items, the technical level of the pre-designed repair plan, the expected load changes after repair, the qualifications of the construction unit and its experience in similar projects, and the performance level and design service life of the repair materials. The core degradation item type refers to the item with the lowest value in the identification scoring system, which is divided into component body type, node connection type, and durability type. The pre-defined technical level of the repair plan refers to the four levels classified according to the depth of treatment: emergency response, routine repair, reinforcement and restoration, and replacement and reconstruction. By combining the building's future use plan, the ratio of the expected load to the design load is calculated to obtain the expected load change after the renovation. The construction unit's qualifications and experience in similar projects refer to a quantitative score based on the construction unit's qualification level and the number and pass rate of similar repair projects in the past 5 years. The performance grade and design service life of repair materials refer to the determination of the material performance grade and design service life based on national material standards and test reports; Obtain historical repair coefficient data for a single high-risk monitoring area within the repair task. The historical repair coefficient data includes the current comprehensive health score, expected aging rate, core deterioration type, technical level of the preset repair plan, expected load change after repair, construction unit qualifications and experience in similar projects, performance level and design service life of repair materials, and historical repair coefficient of the single high-risk monitoring area within different repair tasks. Based on the current comprehensive health score, expected aging rate, core deterioration type, technical level of the preset repair plan, expected load change after repair, construction unit qualifications and experience in similar projects, performance level and design service life of repair materials, and corresponding historical repair coefficients of high-risk monitoring areas in different historical repair coefficient data, a repair coefficient prediction set is generated and divided into a second training set and a second test set. A second convolutional neural network is constructed. The current comprehensive health score, expected aging rate, core deterioration item type, preset repair scheme technical level, expected load change after repair, construction unit qualification and experience in similar projects, and repair material performance level and design service life are taken from the different historical repair coefficient data in the second training set as the input data of the second convolutional neural network. The corresponding historical repair coefficients in the second training set are taken as the output data of the second convolutional neural network. The second convolutional neural network is a one-dimensional convolutional neural network adapted to multi-dimensional classification and numerical mixed features. The overall network structure, along the data transmission direction, is sequentially set as follows: input layer, first convolutional layer, second convolutional layer, third convolutional layer, max pooling layer, first fully connected layer, second fully connected layer, and output layer. The input layer has a 7×1 dimension, corresponding to 7 input feature parameters. The first convolutional layer has 64 filters with a 3×1 filter size, a stride of 1, and uses same padding. The activation function is ReLU. The second convolutional layer has 32 filters with a 3×1 filter size, a stride of 1, and uses s... The system uses AME padding and ReLU activation function. The third convolutional layer has 16 filters (3×1 size) with a stride of 1 and uses SAME padding. The max pooling layer has a 2×1 pooling window and a stride of 2. The first fully connected layer has 128 neurons and ReLU activation function. The second fully connected layer has 32 neurons and ReLU activation function. The output layer has 1 neuron and corresponds to the predicted repair coefficient in the 0-1 interval, activated by the Sigmoid activation function.

[0053] The hyperparameters for network training are specifically set as follows: initial learning rate 0.0005, batch size 32, maximum number of iterations 300, Adam adaptive moment estimation optimizer, and binary cross-entropy loss function. Three measures are implemented to prevent overfitting during training: first, Dropout layers are added after each of the two fully connected layers, with a random dropout probability of 0.4; second, L2 regularization is applied to the weights of both convolutional and fully connected layers, with a regularization coefficient of 0.0001; and third, an early stopping mechanism is implemented, terminating training early when the test set loss value does not decrease for 15 consecutive iterations to avoid overfitting. The second test error threshold uses the mean absolute percentage error (MAPE), set to ≤8%. This threshold is determined based on the engineering accuracy requirements for predicting the effects of prefabricated structure repair schemes. When the mean absolute percentage error of the test set prediction results is ≤8%, the model is deemed to meet the engineering application accuracy requirements. 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 repair coefficient prediction model. The current comprehensive health score, expected aging rate, core deterioration type, pre-designed repair plan technical level, expected load change after repair, construction unit qualification and experience in similar projects, and repair material performance level and design service life of each high-risk monitoring area are input into the repair coefficient prediction model to obtain the predicted repair coefficient for each high-risk monitoring area. In an embodiment of the present invention, the predicted repair coefficient of all high-risk monitoring areas is obtained through the repair coefficient prediction model. The predicted repair coefficient is related to the current comprehensive health score, the expected aging rate, the type of core deterioration item, the technical level of the preset repair plan, the expected load change after repair, the qualifications of the construction unit and experience in similar projects, and the performance level and design service life of the repair materials. The current comprehensive health score directly affects the predicted repair coefficient. The higher the current comprehensive health score, the larger the predicted repair coefficient. Therefore, the current comprehensive health score and the repair coefficient are positively correlated. The predicted aging rate directly affects the predicted repair coefficient; the higher the predicted aging rate, the lower the predicted repair coefficient. Therefore, the predicted aging rate and the repair coefficient are negatively correlated. The core degradation item type reflects the core damage type. The maturity of repair technology varies significantly among different types. The more mature the repair technology for a degradation item, the greater the repair coefficient. The technical level of the pre-designed repair plan directly affects the magnitude of the predicted repair coefficient. The higher the technical level of the pre-designed repair plan, the larger the predicted repair coefficient. Therefore, the technical level of the pre-designed repair plan and the repair coefficient are positively correlated. The magnitude of the expected load change after repair directly affects the magnitude of the predicted repair coefficient. The higher the expected load change after repair, the smaller the predicted repair coefficient. Therefore, the expected load change after repair is negatively correlated with the repair coefficient. The qualifications of the construction unit and its experience in similar projects reflect its ability to ensure the quality of repair and construction. The stronger the construction unit's capabilities, the greater the repair coefficient. The performance grade and design service life of repair materials directly affect the predicted repair coefficient. The higher the performance grade and design service life of repair materials, the larger the predicted repair coefficient. Therefore, the performance grade and design service life of repair materials are positively correlated with the repair coefficient.

[0054] It should be further explained that, in the specific implementation process, the process of selecting the corresponding repair scheme based on the predicted repair coefficient includes: For a single high-risk monitoring area, three or more differentiated repair alternative plans are designed based on the type of core deterioration item, the level of structural importance, and on-site construction conditions. Each alternative plan has clearly defined core parameters such as the current comprehensive health score, the expected aging rate, the type of core deterioration item, the technical level of the preset repair plan, the expected load change after repair, the qualifications of the construction unit and experience in similar projects, and the performance level and design service life of the repair materials. For each alternative repair plan, the above core parameters are input into the repair coefficient prediction model to calculate the predicted repair coefficient corresponding to each alternative plan. After predicting the repair coefficients of all alternative schemes, a repair threshold is preset. Alternative schemes with predicted repair coefficients lower than the repair threshold are eliminated. These schemes cannot achieve the basic repair goal of raising the comprehensive health score of the high-risk monitoring area to above the first comprehensive health score threshold and are not feasible for engineering application. For the remaining alternative schemes with predicted repair coefficients ≥ the repair threshold, the ability to achieve the repair effect is first quantitatively calculated. The expected health score improvement value of each scheme is calculated using the formula: Expected health score improvement value = (1 - current comprehensive health score) × predicted repair coefficient. At the same time, the expected health score compliance rate is calculated using the formula: Expected health score compliance rate = (current comprehensive health score + expected health score improvement value) / repair threshold × 100%. Schemes with an expected health score compliance rate < 100% are eliminated simultaneously to ensure that the remaining alternative schemes can meet the core requirement that the comprehensive health score after repair exceeds the repair threshold. Complete data, including the scheme parameters, actual repair coefficients, and repair effects of this renovation project, were added to the training dataset of the repair coefficient prediction model. The model was iteratively optimized and trained. Every 10 sets of newly added valid repair case data were completed, and the model weights were updated to continuously improve the prediction accuracy of the model. At the same time, the verified and qualified repair schemes were stored in the scheme database to provide data support for the design of repair schemes for similar prefabricated building components and similar deterioration problems.

[0055] A method for health monitoring of prefabricated structures includes the following steps: S1: Divide the prefabricated building into monitoring areas that match the prefabricated component units. Deploy corresponding sensing units in each area to realize the real-time acquisition of component body data and node connection data, and perform normalization processing on the data. S2: Based on normalized data, the component health score and node health score are calculated using a weighted summation formula, and then the comprehensive health score is calculated. At the same time, the change rate of the comprehensive health score is calculated by the difference between the scores of the previous and subsequent periods. S3: Set the first comprehensive health score threshold, the second comprehensive health score threshold, and the comprehensive health score change rate threshold to divide the monitoring area into low, medium, and high-risk monitoring areas, and match them with control strategies for routine inspection, preventive maintenance, and emergency response, respectively. The use of the corresponding area in the high-risk monitoring area is suspended. S4: Construct a convolutional neural network remaining lifetime prediction model to predict the remaining lifetime of high-risk areas. Use the BIM model as the base map to generate life columns and life basin areas. Based on the ant colony algorithm, generate the optimal maintenance and repair route to avoid unsafe areas. S5: Construct a repair coefficient prediction model to predict the effects of different repair schemes, eliminate schemes with scores below the repair threshold, select the optimal scheme and implement it, calculate the actual repair coefficient after the repair is completed, compare it with the predicted value, iteratively optimize the model and update the scheme database.

[0056] 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.

[0057] 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 fabricated structural health monitoring system, characterized by, Includes the following modules: The data acquisition module is used to collect structural data of the monitoring area in real time. The health assessment module is used to generate component health scores and node health scores for the monitoring area based on the collected structural data, calculate the comprehensive health score of the monitoring area based on the component health scores and node health scores, and calculate the rate of change of the comprehensive health score of the monitoring area based on the comprehensive health score. The health grading module is used to classify the monitoring area based on the comprehensive health score and the rate of change of the comprehensive health score, and to take corresponding measures for monitoring areas of different levels. The remaining useful life prediction module is used to acquire historical remaining useful life data and build a remaining useful life prediction model. It uses the remaining useful life prediction model to predict the remaining useful life of high-risk monitoring areas, generates maintenance and repair routes based on the predicted remaining useful life of high-risk monitoring areas, and calculates the expected aging rate of high-risk monitoring areas based on the predicted remaining useful life of high-risk monitoring areas. The repair coefficient prediction module is used to acquire historical repair coefficient data and build a repair coefficient prediction model. The repair coefficient prediction model is used to predict the repair coefficient of high-risk monitoring areas, and the corresponding repair scheme is selected based on the predicted repair coefficient.

2. The prefabricated structure health monitoring system according to claim 1, characterized in that, The prefabricated building is divided into multiple monitoring areas that match the prefabricated component units. Corresponding sensing units are deployed in each monitoring area to collect structural data of the prefabricated building. The structural data includes component body data and node connection data. Component body data includes concrete strength decay rate, steel corrosion rate, concrete carbonation depth ratio and component crack width exceeding standard ratio. Node connection data includes connection bolt preload loss rate, node weld defect length ratio, node relative displacement exceeding standard ratio and node interface crack width exceeding standard ratio. Concrete strength is tested on-site using a rebound hammer or a combined ultrasonic rebound method. Combined with long-term data from embedded strain sensors, the concrete strength attenuation rate is calculated as: (Design strength - Measured strength) / Design strength. The concrete strength attenuation rate is then normalized. In the formula, To normalize the concrete strength degradation rate, This represents the measured concrete strength attenuation rate. and This represents the ideal optimal value and upper limit of the concrete strength degradation rate; The corrosion rate of steel bars is calculated by monitoring the polarization potential and corrosion current density using pre-embedded steel bar corrosion sensors, and then normalized. In the formula, To normalize the steel reinforcement corrosion rate, This represents the measured steel reinforcement corrosion rate. and This represents the ideal optimal value and upper limit of the steel reinforcement corrosion rate; The concrete carbonation depth was obtained through on-site core sampling. The percentage of concrete carbonation depth was calculated as: measured carbonation depth / thickness of the protective layer. The percentage of concrete carbonation depth was then normalized. In the formula, To normalize the percentage of concrete carbonation depth, This represents the measured percentage of concrete carbonation depth. and This represents the ideal optimal value for the proportion of concrete carbonation depth and the upper limit of deterioration. The component crack width is obtained through distributed fiber optic crack sensors and image recognition. The component crack width exceeding the standard ratio is calculated as: measured maximum crack width / maximum allowable crack width. The component crack width exceeding the standard ratio is then normalized. In the formula, To normalize the percentage of component crack width exceeding the standard, This represents the percentage of component crack widths exceeding the standard, as measured in practice. and This represents the ideal optimal value and upper limit of the component's crack width exceeding the standard ratio; The preload of the connecting bolts is monitored by the preload stress sensor. The preload loss rate of the connecting bolts is calculated as (initial preload - measured preload) / initial preload. The preload loss rate of the connecting bolts is then normalized. In the formula, To normalize the preload loss rate of connecting bolts, This represents the measured preload loss rate of the connecting bolts. and The ideal optimal value and upper limit of deterioration for the preload loss rate of connecting bolts; The length of defects in the joint weld was obtained by ultrasonic testing. The proportion of defect length in the joint weld was calculated as the cumulative length of defects divided by the total weld length. The proportion of defect length in the joint weld was then normalized. In the formula, To normalize the proportion of defect length in node welds, This represents the percentage of defect length in the measured node weld. and This represents the ideal optimal value and upper limit of the proportion of defect length in the node weld; The relative displacement of the components on both sides of the node is monitored by displacement sensors. The ratio of the relative displacement exceeding the standard is calculated as: maximum measured relative displacement / allowable limit in the specification. The ratio of the relative displacement exceeding the standard is then normalized. In the formula, This represents the normalized ratio of relative displacement exceeding the standard at nodes. This represents the measured percentage of nodal relative displacements exceeding the standard. and The ideal optimal value and upper limit of the relative displacement exceeding the standard of the node are defined; The crack width at the interface between new and old concrete is monitored using distributed fiber optic sensors. The ratio of excessive crack width at the joint interface is calculated as: (measured maximum crack width / allowable limit in the specification). The ratio of excessive crack width at the joint interface is then normalized. In the formula, This represents the normalized ratio of excessive crack width at the node interface. This represents the percentage of joint interface crack widths exceeding the standard, as measured in practice. and This represents the ideal optimal value and the upper limit of the degradation for the ratio of excessive crack width at the node interface.

3. The prefabricated structure health monitoring system according to claim 2, characterized in that, The process of generating component health scores and node health scores for the monitoring area based on the collected structural data includes: A component health score, S, is generated based on the component ontology data. In the formula, , , and These are weighting coefficients, obtained through training based on historical data; A node health score is generated based on node connection data. Node health score Q: In the formula, , , and These are weighting coefficients, obtained through training based on historical data.

4. The prefabricated structure health monitoring system according to claim 3, characterized in that, The process of calculating the comprehensive health score of the monitoring area based on the component health score and the node health score includes: Overall health score P: 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 comprehensive health score in the monitored area based on the comprehensive health score includes: A fixed monitoring period is set. After each monitoring period ends, the measured value of the comprehensive health score of the monitored area within that monitoring period is obtained. At the same time, the historical value of the comprehensive health score of the monitored area in the previous monitoring period is retrieved. The measured value of the current period is subtracted from the historical value of the previous period to obtain the absolute change in the comprehensive health score. Then, the absolute change is divided by the historical value of the comprehensive health score of the previous monitoring period to obtain the rate of change of the comprehensive health score of the monitored area in this monitoring period.

5. The prefabricated structure health monitoring system according to claim 4, characterized in that, The process of classifying monitoring areas based on comprehensive health scores and the rate of change in comprehensive health scores includes: Set appropriate first and second comprehensive health score thresholds based on historical comprehensive health score data, where the historical comprehensive health score data refers to the dataset of comprehensive health scores of the monitored area in the past. Set an appropriate threshold for the rate of change of the comprehensive health score based on historical comprehensive health score change data, where the historical comprehensive health score change data refers to the data set of the rate of change of the comprehensive health score in the monitored area in the past. Low-risk monitoring area determination: Two conditions must be met simultaneously: first, the comprehensive health score of the monitoring area is greater than the first comprehensive health score threshold; second, the change rate of the comprehensive health score of the monitoring area is greater than or equal to the change rate threshold of the comprehensive health score. Both conditions must be met at the same time, and neither can be missing. Low-risk monitoring areas shall be subject to routine inspections and data monitoring, and do not need to be included in the repair plan. Determination of medium-risk monitoring areas: A monitoring area can be determined as medium-risk if it meets any of the following conditions: Condition 1: The comprehensive health score of the monitoring area is within the first comprehensive health score threshold and the second comprehensive health score threshold, regardless of the value of the comprehensive health score change rate; Condition 2: The comprehensive health score of the monitoring area is greater than the first comprehensive health score threshold, but the comprehensive health score change rate is less than the comprehensive health score change rate threshold. Medium-risk monitoring areas are included in the quarterly maintenance plan and preventive maintenance and deterioration suppression measures are implemented. High-risk monitoring area determination: If the comprehensive health score of the monitoring area is less than the second comprehensive health score threshold, it is determined to be a high-risk monitoring area. The emergency response for high-risk monitoring areas shall be activated immediately, the normal use of the corresponding area shall be suspended, and the remaining life prediction and special repair plan design shall be carried out simultaneously.

6. The prefabricated structure health monitoring system according to claim 5, 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 high-risk monitoring areas, includes: The remaining life of high-risk monitoring areas is predicted. The remaining life refers to the remaining time before the comprehensive health score of a high-risk monitoring area drops to the danger threshold. The danger threshold is a pre-set threshold. High-risk monitoring areas with a comprehensive health score ≤ the danger threshold have reached the level of damage and collapse. Factors affecting the remaining lifespan of high-risk monitoring areas include: current comprehensive health score, rate of change of current comprehensive health score, environmental corrosion level, percentage of design service life already used, load exceedance rate, and crack propagation rate. The corrosion level of the environment refers to the classification of 1-5 (Level 1 is slightly corrosive, Level 5 is strongly corrosive) according to GB / T50046-2018 and environmental monitoring data. Percentage of designed service life already used = Years of building service life / Design service life of components; The load exceedance rate is calculated as: measured maximum load effect / design load limit, based on actual stress monitoring using strain sensors. The crack propagation rate, as monitored by distributed fiber optic sensors, is calculated as: crack width increment during the monitoring period / monitoring period duration. The historical remaining lifespan data of a single high-risk monitoring area in different monitoring cycles is obtained. The historical remaining lifespan data includes the current comprehensive health score, the rate of change of the current comprehensive health score, the corrosion level of the environment, the percentage of the design service life already used, the load exceeding the standard rate, the crack propagation rate, and the historical remaining lifespan of the single high-risk monitoring area in different monitoring cycles. Based on the current comprehensive health score, current comprehensive health score change rate, environmental corrosion level, proportion of design service life already used, load exceedance rate, crack propagation rate and corresponding historical remaining life of the high-risk monitoring area in different historical remaining life data, a remaining life 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 comprehensive health score, the rate of change of the current comprehensive health score, the corrosion level of the environment, the proportion of the design service life already used, the load exceedance rate, and the crack propagation speed from different historical remaining life data in the first training set as the input data of the first convolutional neural network, and taking the corresponding historical remaining life 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 a preset first test error threshold is used as the remaining lifetime prediction model. For each monitoring cycle, the current comprehensive health score, the rate of change of the current comprehensive health score, the corrosion level of the environment, the percentage of the design service life already used, the load exceedance rate and the crack propagation rate of each high-risk monitoring area are input into the remaining life prediction model to obtain the predicted remaining life of each high-risk monitoring area. The first convolutional neural network is a one-dimensional convolutional neural network, which includes an input layer, two convolutional layers, a pooling layer, two fully connected layers, and an output layer. The convolutional layer filter size is 3×1, and the number of filters is 32 and 16 respectively. The activation function is the ReLU function. The number of neurons in the fully connected layer is 64 and 16 respectively. The output layer uses a linear activation function. The network training hyperparameters are: learning rate 0.001, batch size 16, and the Adam optimizer. Training uses L2 regularization, Dropout layer, and early stopping mechanism to prevent overfitting. The training termination condition is that the mean absolute percentage error of the test set is ≤5%.

7. The prefabricated structure health monitoring system according to claim 6, characterized in that, The process of generating maintenance and repair routes based on the predicted remaining lifespan of high-risk monitoring areas includes: Using the BIM 3D information model corresponding to the prefabricated building as the visualization base map, the spatial coordinates of the monitoring area and the BIM model components are matched one by one. The coordinates of the spatial center point of each monitoring area are aligned with the coordinates of the geometric center point of the corresponding prefabricated component in the BIM model to ensure that the spatial mapping accuracy error between the monitoring data and the 3D model does not exceed 5cm. After calibration, a columnar visualization model, known as a life column, is generated at the 3D coordinate position of each monitoring area, perpendicular to the horizontal plane of the building. The height of the life column is linearly positively correlated with the predicted remaining life of the corresponding high-risk monitoring area. The height of the life column is set to 0.5m / year per unit of remaining life, meaning that the height of the life column decreases by 0.5m for every year the predicted remaining life decreases. At the same time, the fill color of the life column gradually changes from green, yellow, orange to red as the predicted remaining life decreases. The diameter of the life column is negatively correlated with the comprehensive health score of the corresponding high-risk monitoring area. The lower the comprehensive health score, the larger the diameter of the life column. After generating life columns for all high-risk monitoring areas of the entire building, continuous low-lying areas are identified in the 3D map. The threshold for determining life basins is set as a predicted remaining life of ≤3 years. When three or more spatially adjacent monitoring areas have a predicted remaining life of ≤3 years and form a continuous spatial area, the continuous area is marked as a life basin area. This life basin area is a cluster of structural safety risks in prefabricated buildings, with the risk of cascading failures of components and nodes. It is listed as the highest priority in the repair priority, followed by independent high-risk monitoring areas, and finally medium-risk monitoring areas. When generating maintenance and repair routes, the route starts from the pre-set entrances for maintenance personnel and equipment within the building. Each high-risk monitoring area within the lifespan basin area is designated as a primary core node, independent high-risk monitoring areas as secondary nodes, and medium-risk monitoring areas as tertiary nodes. Path planning is performed based on the ant colony algorithm. Three constraints are incorporated into the path planning process: first, the route must avoid normal use areas, non-safe structural areas, and areas with dense electromechanical pipelines within the building; second, the route must connect nodes in descending order of priority; and third, nodes of the same priority are connected by path concatenation based on proximity. After generating the initial shortest path, the route is further optimized by considering the equipment transportation requirements and personnel workspace requirements of the maintenance construction. Path segments that cannot meet the construction access requirements are eliminated, and the optimal maintenance and repair route is finally generated. Simultaneously, this maintenance route is mapped onto a 3D map and displayed in conjunction with the lifespan basin area and monitoring areas of each risk level. The process of calculating the projected aging rate of a high-risk monitoring area based on its predicted remaining lifespan includes: For high-risk monitoring areas, first retrieve the pre-set danger threshold, the measured comprehensive health score of the current monitoring period, and the predicted remaining lifespan output by the remaining lifespan prediction model for the monitoring area. First, calculate the total decline in the health score of the high-risk monitoring area from the current health status to the danger threshold. The total decline = the measured value of the current comprehensive health score - the danger threshold. Then, divide the total decline by the predicted remaining lifespan of the high-risk monitoring area to obtain the decrease in the comprehensive health score of the monitoring area per unit time under natural deterioration without external intervention. This is the expected aging rate of the high-risk monitoring area.

8. The prefabricated structure health monitoring system according to claim 7, characterized in that, The process of acquiring historical repair coefficient data and constructing a repair coefficient prediction model, and then using this model to predict the repair coefficient of high-risk monitoring areas, includes: The repair coefficient refers to the degree of repair effect of the high-risk monitoring area after the repair task is completed. The repair task refers to the maintenance and repair of the high-risk monitoring area. The repair coefficient ranges from [0, 1]. The closer it is to 1, the better the repair effect, the greater the improvement in the comprehensive health score after repair, and the more lasting the deterioration inhibition effect. Factors affecting the repair coefficient include: the current comprehensive health score of the high-risk monitoring area, the expected aging rate, the type of core deterioration items, the technical level of the pre-designed repair plan, the expected load changes after repair, the qualifications of the construction unit and its experience in similar projects, and the performance level and design service life of the repair materials. The core degradation item type refers to the item with the lowest value in the identification scoring system, which is divided into component body type, node connection type, and durability type. The pre-defined technical level of the repair plan refers to the four levels classified according to the depth of treatment: emergency response, routine repair, reinforcement and restoration, and replacement and reconstruction. By combining the building's future use plan, the ratio of the expected load to the design load is calculated to obtain the expected load change after the renovation. The construction unit's qualifications and experience in similar projects refer to a quantitative score based on the construction unit's qualification level and the number and pass rate of similar repair projects in the past 5 years. The performance grade and design service life of repair materials refer to the determination of the material performance grade and design service life based on national material standards and test reports; Obtain historical repair coefficient data for a single high-risk monitoring area within the repair task. The historical repair coefficient data includes the current comprehensive health score, expected aging rate, core deterioration type, technical level of the preset repair plan, expected load change after repair, construction unit qualifications and experience in similar projects, performance level and design service life of repair materials, and historical repair coefficient of the single high-risk monitoring area within different repair tasks. Based on the current comprehensive health score, expected aging rate, core deterioration type, technical level of the preset repair plan, expected load change after repair, construction unit qualifications and experience in similar projects, performance level and design service life of repair materials, and corresponding historical repair coefficients of high-risk monitoring areas in different historical repair coefficient data, a repair coefficient prediction set is generated and divided into a second training set and a second test set. A second convolutional neural network is constructed. The current comprehensive health score, expected aging rate, core deterioration item type, preset repair scheme technical level, expected load change after repair, construction unit qualification and experience in similar projects, and repair material performance level and design service life are taken from the different historical repair coefficient data in the second training set as the input data of the second convolutional neural network. The corresponding historical repair coefficients in the second training set are taken 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 repair coefficient prediction model. The current comprehensive health score, expected aging rate, core deterioration type, pre-designed repair plan technical level, expected load change after repair, construction unit qualification and experience in similar projects, and repair material performance level and design service life of each high-risk monitoring area are input into the repair coefficient prediction model to obtain the predicted repair coefficient for each high-risk monitoring area. The second convolutional neural network is a one-dimensional convolutional neural network, which includes an input layer, three convolutional layers, a pooling layer, two fully connected layers, and an output layer. The filter size of the convolutional layers is 3×1, and the number of filters is 64, 32, and 16 respectively. The activation function is the ReLU function. The number of neurons in the fully connected layers is 128 and 32 respectively. The output layer uses the Sigmoid activation function. The network training hyperparameters are: learning rate 0.0005, batch size 32, and the Adam optimizer. Training uses L2 regularization, Dropout layers, and an early stopping mechanism to prevent overfitting. The training termination condition is that the mean absolute percentage error of the test set is ≤8%.

9. The prefabricated structure health monitoring system according to claim 8, characterized in that, The process of selecting the corresponding repair scheme based on the predicted repair coefficient includes: For a single high-risk monitoring area, three or more differentiated repair alternative plans are designed based on the type of core deterioration item, the level of structural importance, and on-site construction conditions. Each alternative plan has clearly defined core parameters such as the current comprehensive health score, the expected aging rate, the type of core deterioration item, the technical level of the preset repair plan, the expected load change after repair, the qualifications of the construction unit and experience in similar projects, and the performance level and design service life of the repair materials. For each alternative repair plan, the above core parameters are input into the repair coefficient prediction model to calculate the predicted repair coefficient corresponding to each alternative plan. After predicting the repair coefficients of all alternative schemes, a repair threshold is preset. Alternative schemes with predicted repair coefficients lower than the repair threshold are eliminated. These schemes cannot achieve the basic repair goal of raising the comprehensive health score of the high-risk monitoring area to above the first comprehensive health score threshold and are not feasible for engineering application. For the remaining alternative schemes with predicted repair coefficients ≥ the repair threshold, the ability to achieve the repair effect is first quantitatively calculated. The expected health score improvement value of each scheme is calculated using the formula: Expected health score improvement value = (1 - current comprehensive health score) × predicted repair coefficient. At the same time, the expected health score compliance rate is calculated using the formula: Expected health score compliance rate = (current comprehensive health score + expected health score improvement value) / repair threshold × 100%. Schemes with an expected health score compliance rate < 100% are eliminated simultaneously to ensure that the remaining alternative schemes can meet the core requirement that the comprehensive health score after repair exceeds the repair threshold. Complete data, including the scheme parameters, actual repair coefficients, and repair effects of this renovation project, were added to the training dataset of the repair coefficient prediction model. The model was iteratively optimized and trained. Every 10 sets of newly added valid repair case data were completed, and the model weights were updated to continuously improve the prediction accuracy of the model. At the same time, the verified and qualified repair schemes were stored in the scheme database to provide data support for the design of repair schemes for similar prefabricated building components and similar deterioration problems.

10. A method for monitoring the health of prefabricated structures, implemented based on the prefabricated structure health monitoring system according to any one of claims 1-9, characterized in that, Includes the following steps: S1: Divide the prefabricated building into monitoring areas that match the prefabricated component units. Deploy corresponding sensing units in each area to realize the real-time acquisition of component body data and node connection data, and perform normalization processing on the data. S2: Based on normalized data, the component health score and node health score are calculated using a weighted summation formula, and then the comprehensive health score is calculated. At the same time, the change rate of the comprehensive health score is calculated by the difference between the scores of the previous and subsequent periods. S3: Set the first comprehensive health score threshold, the second comprehensive health score threshold, and the comprehensive health score change rate threshold to divide the monitoring area into low, medium, and high-risk monitoring areas, and match them with control strategies for routine inspection, preventive maintenance, and emergency response, respectively. The use of the corresponding area in the high-risk monitoring area is suspended. S4: Construct a convolutional neural network remaining lifetime prediction model to predict the remaining lifetime of high-risk areas. Use the BIM model as the base map to generate life columns and life basin areas. Based on the ant colony algorithm, generate the optimal maintenance and repair route to avoid unsafe areas. S5: Construct a repair coefficient prediction model to predict the effects of different repair schemes, eliminate schemes with scores below the repair threshold, select the optimal scheme and implement it, calculate the actual repair coefficient after the repair is completed, compare it with the predicted value, iteratively optimize the model and update the scheme database.

Images