Building crack risk management and control method and device

Abstract

The application provides a building crack risk management method and device, which can be applied to the technical field of structural health monitoring, building protection and risk early warning. The method comprises the following steps: obtaining first state detection data, first state estimation data, a first risk management strategy and historical state detection data of a crack of a target building; obtaining a crack change trend parameter and second state estimation data based on the historical state detection data and the first state detection data; determining a first deviation type based on the first state detection data and the first state estimation data; obtaining a second detection strategy according to a crack activity and the first deviation type; processing the historical state detection data, the first state detection data, the crack change trend parameter and the first deviation type by using a risk identification model to obtain crack risk information; and obtaining a second risk management strategy for a second time period according to the crack risk information and the second detection strategy.

Description

Technical Field

[0001] This application relates to the field of building protection and risk early warning technology, and more specifically to a method and device for risk management of building cracks. Background Technology

[0002] Buildings are subjected to complex environmental factors such as continuous wet and dry cycles, temperature fluctuations, and rainwater infiltration leading to frost heave. This can cause problems such as loosening and dislodging of mortise and tenon joints in timber structures and weathering and cracking in brick and stone walls. Due to the non-renewable nature of building materials and the structural integrity of the building, crack damage is irreversible and cumulative; once a crack penetrates, it leads to permanent structural degradation. Manual visual inspection for crack detection is infrequent, highly subjective, and only records the current state without capturing subtle deformations and expansion signals, resulting in significant building management risks. Summary of the Invention

[0003] In view of the above problems, this application provides a method and equipment for building crack risk management that can reduce the cost of intervening in building cracks and optimize the allocation of building maintenance resources.

[0004] According to a first aspect of this application, a method for risk management of building cracks is provided, comprising: acquiring first state detection data, first state prediction data, a first risk management strategy, and historical state detection data for historical periods of a target building at multiple first moments in a first time period, wherein the first risk management strategy is used to maintain the target building in the first time period, and the first state prediction data characterizes the predicted crack state for the crack at multiple first moments in the first time period; performing a state assessment of the crack based on the historical state detection data and the first state detection data to obtain crack change trend parameters and second state prediction data of the crack in a second time period, wherein the second time period is later than the first time period; analyzing the deviation between the first state detection data and the first state prediction data to determine a first deviation type; adjusting the first detection strategy in the first risk management strategy according to the second state prediction data, crack activity, and the first deviation type to obtain a second detection strategy, wherein the crack activity is determined based on the crack change trend parameters and preset judgment conditions; processing the historical state detection data, the first state detection data, the crack change trend parameters, and the first deviation type using a risk identification model to obtain crack risk information; and obtaining a second risk management strategy for the second time period based on the crack risk information and the second detection strategy.

[0005] According to an embodiment of this application, a second detection strategy is obtained by adjusting the first detection strategy in the first risk control strategy based on the second state prediction data, crack activity, and a first deviation type. This includes: determining crack activity based on crack change trend parameters and a preset threshold; obtaining a priority based on the first state detection data and the second state prediction data, where the priority represents the urgency of detecting multiple cracks; initially adjusting the first detection strategy based on the priority and crack activity to obtain an initial detection strategy; and further adjusting the initial detection strategy based on the first state detection data and the first deviation type to obtain the second detection strategy.

[0006] According to an embodiment of this application, a first detection strategy includes an estimated search area, and first state detection data includes the crack extension location. The first detection strategy is adjusted based on the first state detection data and a first deviation type to obtain a second detection strategy, including: determining the first detection strategy as the second detection strategy when the crack extension location is within the estimated search area and the first deviation type is a consistent trend; adding the first state detection data to the historical state detection data for the second time period; adjusting the expansion ratio of the estimated search area in the estimation model or the initial detection strategy when the crack extension location is outside the estimated search area or the first deviation type is an accelerating trend; and clearing the first state detection data or reconstructing the data using the first state detection data as the initial point when the first deviation type is an abnormal trend.

[0007] According to an embodiment of this application, priority is obtained based on first-state detection data and second-state prediction data, including: determining the historical environmental driving intensity based on historical environmental data; extracting structural attributes from the first-state detection data to obtain the geometric expansion intensity; and determining the priority of the crack based on the historical crack risk level, second-state prediction data, historical deviation, historical environmental driving intensity, geometric expansion intensity, and estimated detection cost.

[0008] According to embodiments of this application, the first detection strategy includes at least one of a search area, an endpoint interest area, a bifurcation interest area, a local detection threshold, a manual re-inspection marker, a standard detection threshold, and a re-inspection cycle; the second state prediction data includes the estimated extension direction, estimated extension length, and position fluctuation range of the endpoint; wherein, the first detection strategy is initially adjusted according to priority and crack activity to obtain an initial detection strategy, including: when crack activity is in a stable state, reducing the search area or extending the re-inspection cycle in the first detection strategy to obtain the initial detection strategy; when crack activity is in a uniform expansion state, adjusting the search area or extending the re-inspection cycle according to the estimated extension direction, estimated extension length, and position fluctuation range of the endpoint; The search area in the first detection strategy is expanded to increase the range of positional fluctuations, resulting in an initial detection strategy. When the crack activity is in an accelerated propagation state and the environmental driving intensity is greater than the preset driving threshold, the search area in the first detection strategy is expanded according to the estimated extension length and the preset acceleration safety factor, and the local detection threshold is adjusted in the endpoint interest area and the bifurcation interest area to obtain the initial detection strategy. When the crack activity is in a decelerated propagation state, the preset benchmark detection strategy is determined as the initial detection strategy. When the crack activity is in an abnormal fluctuation state, the search area in the first detection strategy is expanded, the standard detection threshold is increased, and manual re-inspection marks are added to obtain the initial detection strategy.

[0009] According to embodiments of this application, both historical state detection data and first state detection data include geometric type data, count type data, and state type data. Specifically, based on the historical state detection data and the first state detection data, a state assessment of the crack is performed to obtain crack change trend parameters and second state prediction data for the crack in a second time period. This includes: analyzing the geometric type data using the derivative of a geometric prediction model to obtain a crack structure change trend, where the crack structure change trend includes at least one of change rate, acceleration, fluctuation coefficient, and trend fitting degree; processing the count type data using a count prediction model to obtain the probability of new bifurcation; processing the state type data using a state prediction model to obtain the probability of crack penetration state transition, where the crack change trend parameters include at least one of crack structure change trend, probability of new bifurcation, and probability of crack penetration state transition; and performing state prediction on the crack based on the geometric type data using a geometric prediction model to obtain second state prediction data, where the second state prediction data includes at least one of crack geometric prediction location, probability of bifurcation number change, bifurcation candidate region, penetration state change trend, and location fluctuation range.

[0010] According to an embodiment of this application, the geometric prediction model is determined based on the following operations: multiple candidate models are used to fit the sample geometric type data of the crack to obtain multiple sample state prediction data; the fitting residuals are obtained based on the sample geometric type data and the sample state prediction data; and the candidate model corresponding to the smallest fitting residual among the multiple fitting residuals is determined as the geometric prediction model.

[0011] According to an embodiment of this application, crack risk information is obtained by processing historical state detection data, first state detection data, crack change trend parameters, and first deviation type using a risk identification model. This includes: performing feature fusion on historical state detection data, first state detection data, and first environmental data for a first time period to obtain fused features; and processing the fused features, crack activity, and deviation using a risk identification model to obtain crack risk.

[0012] According to an embodiment of this application, feature fusion is performed on historical state detection data, first state detection data, and first environmental data of a first time period to obtain fused features, including: feature extraction on historical state detection data and first state detection data to obtain state features; feature extraction on first environmental data to obtain environmental features; coupling interaction between state features and environmental features to obtain coupling strength; processing coupling strength, state features, and environmental features using a nonlinear mapping function to obtain initial features; and modulating the initial features according to the crack material sensitivity properties to obtain fused features.

[0013] A second aspect of this application provides an electronic device, comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method described above.

[0014] According to the building crack risk management method of this application, for each crack, based on continuous multi-time period state monitoring data, the crack activity, change trend parameters, and second-time period state prediction data are estimated, thus providing a quantitative predictive basis for adjusting the detection strategy. The deviation type is determined by the discrepancy between the first-time period detection data and the first-time period state prediction data, and this deviation type is used as the basis for judging the crack evolution state, thereby achieving feedback verification of the first detection strategy's execution effect. Based on the second-time period state prediction data, crack activity, and deviation type, the first detection strategy is adaptively corrected, ensuring that the adjusted second detection strategy accurately matches the actual crack evolution trend. This realizes the conversion of the second-time period crack state prediction data into specific execution parameters for the second-time period detection strategy. A closed-loop mechanism of "detection strategy generation - detection strategy execution - deviation verification - detection strategy adjustment" has been formed. Then, by using a risk identification model to integrate multi-dimensional historical state detection data, first state detection data, crack change trend parameters, and first deviation type, crack risk information is obtained. Based on the crack risk information and the second detection strategy, a risk management strategy for the second period is generated, thereby realizing the dynamic coupling of detection resource allocation and risk status. This significantly improves the timeliness of crack risk early warning, the accuracy of detection strategy, and the utilization efficiency of detection resources. The targeted second risk management strategy in the second period reduces the building management risk caused by large-scale, unfocused detection or maintenance methods, reduces the cost of building crack intervention, and optimizes the allocation of building maintenance resources. Attached Figure Description

[0015] The above-mentioned contents, other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0016] Figure 1 The diagram illustrates an application scenario of the building crack risk management method and device according to an embodiment of this application.

[0017] Figure 2 A flowchart of a building crack risk management method according to an embodiment of this application is shown.

[0018] Figure 3 A schematic diagram illustrating the generation and adjustment of a second detection strategy in a building crack risk management method according to an embodiment of this application is shown.

[0019] Figure 4 A schematic diagram illustrating the deviation analysis of the building crack risk management method according to an embodiment of this application is shown.

[0020] Figure 5 A block diagram of an electronic device suitable for implementing a building crack risk management method according to an embodiment of this application is shown. Detailed Implementation

[0021] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0022] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0023] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0024] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0025] It should be noted that in the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, they do not mean that the applicant has used or necessarily used the solution.

[0026] In the technical solution of this application, the user information (including but not limited to user personal information, user image information, user device information, such as location information) and data (including but not limited to data used for analysis, stored data, and displayed data) involved are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with relevant laws, regulations, and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entry points for users to choose to authorize or refuse.

[0027] In scenarios involving automated decision-making using personal information, the methods, devices, and systems provided in this application all offer users corresponding entry points for choosing to agree to or reject the automated decision-making results. If the user chooses to reject, the process proceeds to the expert decision-making stage. Here, "automated decision-making" refers to the activity of automatically analyzing and evaluating an individual's behavioral habits, interests, or economic, health, and credit status through computer programs, and then making a decision. Here, "expert decision-making" refers to the activity of making decisions by personnel who specialize in a particular field, possess specialized experience, knowledge, and skills, and have reached a certain level of professional expertise.

[0028] Buildings are subjected to complex environmental factors such as continuous wet and dry cycles, temperature fluctuations, and rainwater infiltration leading to frost heave. This can cause problems like loosening and dislodging of mortise and tenon joints in timber structures and weathering and cracking in brick and stone walls. Due to the non-renewable nature of building materials and the structural integrity of the building, crack damage is irreversible and cumulative; once it penetrates, it leads to permanent structural degradation. Traditional crack detection methods, relying on manual visual inspections, have low detection frequency and are highly subjective, only recording the current state and failing to capture the signs of micro-deformation and expansion of cracks. Therefore, deeply understanding the evolutionary patterns of building cracks and developing methods that combine crack detection strategy generation with risk warning capabilities is a necessary technical path to achieve proactive preventative protection.

[0029] In related technologies, long-term crack monitoring, trend prediction, and risk assessment methods mainly adopt the following technical approaches: 1. Crack monitoring methods based on historical change comparison or trend prediction. These methods typically compare the currently detected crack length, width, area, or rate of change with historical results to predict crack width, crack development trend, or overall structural safety status; 2. Using adaptive thresholding, adaptive segmentation, image enhancement, or noise suppression techniques, adjusting thresholds, segmentation parameters, or image preprocessing parameters in the current detection task based on current image quality, background noise, texture features, lighting conditions, or detection model confidence; 3. Introducing environmental factors to calculate the overall environmental impact coefficient, structural risk index, or overall risk level to assist in judging structural risk or crack development trend; 4. Combining environmental factors, structural status, or monitoring data for dynamic risk assessment, outputting risk level, maintenance recommendations, reinforcement recommendations, or increased monitoring recommendations.

[0030] However, the above methods still have the following technical problems: First, they usually focus on single indicator prediction or overall risk assessment. For the multi-attribute information of the same crack instance, they have not yet fully formed a unified prediction result for the next time period and transformed it into the detection strategy configuration for the next time period. Second, adaptive detection methods mostly adjust parameters for the current image. They usually do not determine the key areas of focus and re-inspection cycle in advance based on the crack prediction results before the detection of the next time period, making it difficult to achieve the active conversion from "prediction result" to "detection execution parameters". Third, when the actual detection result of the next time period deviates from the prediction result of the previous time period, the existing methods usually only treat the deviation as a prediction error or risk change result. They have not fully utilized the positional relationship between the actual value and the prediction uncertainty interval to distinguish the prediction deviation, making it difficult to form a feedback correction closed loop after detection. Fourth, although environmental factors are introduced, they do not fully combine the location of the crack in the component, indoor and outdoor exposure conditions, orientation, height, windward or leeward state, and material type for spatial alignment and sensitivity modulation, making it difficult to accurately describe the personalized environmental driving intensity corresponding to a certain crack at a certain stage of change.

[0031] In view of the above problems, embodiments of this application provide a method for risk management of building cracks that can reduce the cost of intervention in building cracks and optimize the allocation of protection resources. The method involves acquiring first-state detection data, first-state prediction data, a first risk management strategy, and historical state detection data of cracks in a target building; obtaining crack change trend parameters and second-state prediction data based on the historical state detection data and the first-state detection data; determining a first deviation type based on the first-state detection data and the first-state prediction data; obtaining a second detection strategy based on crack activity and the first deviation type; processing the historical state detection data, the first-state detection data, the crack change trend parameters, and the first deviation type using a risk identification model to obtain crack risk information; and obtaining a second risk management strategy for a second time period based on the crack risk information and the second detection strategy.

[0032] Figure 1 The diagram illustrates an application scenario of the building crack risk management method and device according to an embodiment of this application.

[0033] It is important to note that Figure 1 The examples shown are merely application scenarios that can be applied to the embodiments of this application, in order to help those skilled in the art understand the technical content of this application, but do not mean that the embodiments of this application cannot be used in other devices, systems, environments or scenarios.

[0034] like Figure 1As shown, application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0035] Users can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 via the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc. (for example only).

[0036] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0037] Server 105 can be a server that provides various services, such as a backend management server that supports websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (this is just an example). The backend management server can analyze and process data such as received user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.

[0038] It should be noted that the building crack risk management method provided in this application embodiment can generally be executed by server 105. Correspondingly, the building crack risk management device provided in this application embodiment can generally be installed in server 105. The building crack risk management method provided in this application embodiment can also be executed by a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105. Correspondingly, the building crack risk management device provided in this application embodiment can also be installed in a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105.

[0039] It should be understood that Figure 1The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0040] It should be noted that the sequence numbers of the operations in the following methods are for descriptive purposes only and should not be considered as indicating the execution order of the operations. Unless explicitly stated otherwise, the method does not need to be executed in the exact order shown.

[0041] Figure 2 A flowchart of a building crack risk management method according to an embodiment of this application is shown.

[0042] like Figure 2 As shown, the building crack risk management method of this embodiment includes operations S201 to S206.

[0043] In operation S201, acquire the first state detection data, first state prediction data, first risk control strategy, and historical state detection data of the target building's cracks at multiple first moments in the first time period.

[0044] The first risk management strategy is used to maintain the target building in the first time period, and the first state prediction data represents the predicted crack state for multiple first moments in the first time period.

[0045] The first time period can be the current time period, and the first time period includes multiple first moments.

[0046] In response to the first time period, after implementing the first risk control strategy, the first state detection data is collected. For example, the first risk control strategy is generated in the historical time period and includes crack risk information, risk trends, early warning levels, risk driving factors, detection strategies and re-inspection suggestions for the first time period. It is used to assist building maintenance personnel in crack risk control during the first time period.

[0047] The first state detection data refers to the actual observation data of the crack state. For example, the first state detection data may include at least one of the following state data: crack length, width, area, endpoint position, principal orientation angle, bifurcation state, continuity state, or time-normalized rate of change.

[0048] The first detection strategy includes at least one of the following: search area expansion direction, search area expansion ratio, local detection threshold adjustment range, predicted crack endpoint focus area, predicted crack bifurcation focus area, manual re-inspection marking, or suggested re-inspection cycle. Historical state detection data for historical periods includes state detection data for cracks from at least one period prior to the first period.

[0049] The first state prediction data is obtained by predicting the state based on historical state detection data during a historical period. The first state prediction data characterizes the predicted crack state in the first period.

[0050] For example, the first time period is from 4:00 to 5:00 on April 20th, and the closest historical time period to the first time period is from 4:00 to 5:00 on April 1st. Based on the state detection data from 4:00 to 5:00 on April 1st and the historical state detection data relative to this time period, state prediction is performed to obtain the first state prediction data for the first time period.

[0051] In operation S202, based on historical state detection data and first state detection data, the state of the crack is assessed to obtain crack change trend parameters and second state prediction data of the crack in the second time period.

[0052] The second time slot is later than the first time slot.

[0053] Both historical state detection data and first state detection data include continuous geometric attributes, count attributes, and state attributes.

[0054] For each crack, attribute classification or calculation operations are performed on historical state detection data and first state detection data to establish a prediction model. The prediction model is then used to evaluate the crack change trend parameters.

[0055] Crack change trend parameters characterize the dynamic trend of crack evolution over time. For example, crack change trend parameters may include change rate, evolution acceleration, fluctuation coefficient, or probability of new bifurcation, probability of breakthrough state transition, etc.

[0056] Based on the prediction model, the continuous attribute values, endpoint locations, probability of changes in the number of bifurcations, bifurcation candidate regions, and trends in the continuity status of cracks are predicted for one or more future time periods, and second-state prediction data is output. The second-state prediction data is generated in the first time period. For the state prediction data in the second time period, the second-state prediction data may include the predicted location of the crack, endpoint locations, probability of changes in the number of bifurcations, bifurcation candidate regions, trends in the continuity status, and location fluctuation range.

[0057] The second period is later than the first period, and the second period is one or more future periods relative to the first period.

[0058] The content items of the state prediction data correspond to the content items of the state detection data.

[0059] In operation S203, the deviation between the first state detection data and the first state prediction data is analyzed to determine the first deviation type.

[0060] The first deviation type is determined by mapping the deviation between the first state detection data and the first state prediction data to the deviation type.

[0061] For example, the first state prediction data also includes the predicted location, and the first state detection data includes the crack location and the actual crack extension area. When the actual crack extension area is located within the predicted search area, and the deviation between the crack location and the predicted location is within the location fluctuation range, the first deviation type is determined to be a normal deviation.

[0062] For example, deviation types include normal deviations, accelerating trends, or abrupt changes.

[0063] In operation S204, based on the second state prediction data, crack activity, and the first deviation type, the first detection strategy in the first risk control strategy is adjusted to obtain the second detection strategy, wherein the crack activity is determined based on the crack change trend parameter and a preset threshold.

[0064] Crack activity is classified based on crack change trend parameters and preset thresholds to determine crack activity.

[0065] For example, crack activity can include a stable state, a uniform propagation state, an accelerated propagation state, a decelerated propagation state, and an abnormal fluctuation state.

[0066] For example, preset thresholds may include preset rate thresholds for change rate, preset fluctuation thresholds for fluctuation coefficient, preset probability thresholds for new bifurcation probability and through-state transition probability, and preset negative value thresholds for evolution acceleration.

[0067] For example, for continuous geometric attribute data in the first state detection data, if the rate of change in the crack change trend parameter is lower than the preset rate threshold, the fluctuation coefficient in the crack change trend parameter is lower than the preset fluctuation threshold, and the probability of new bifurcation and the probability of transition to the through state are lower than the preset probability threshold, then the crack is determined to be in a stable state.

[0068] For example, for continuous geometric attribute data in the first state detection data, if the evolution acceleration in the crack change trend parameter is less than a preset negative threshold and the change rate in the crack change trend parameter is not lower than a preset rate threshold, then the crack is determined to be in a decelerating expansion state.

[0069] Based on the deviation between the first state detection data and the first state prediction data, after determining the first deviation type, the first detection strategy can be initially corrected based on the second state prediction data and crack activity. If the first deviation type does not meet the deviation conditions, the first detection strategy after the initial correction can be corrected again.

[0070] For example, the initial correction may include expanding the search area in the first detection strategy based on the estimated extension direction, estimated extension length, and location fluctuation range in the second state prediction data when the crack activity is in a state of uniform expansion.

[0071] For example, set the deviation conditions to normal deviation and abnormal mutation.

[0072] For example, if the first deviation type is determined to be a normal deviation, and the first deviation type meets the deviation condition, the first state detection data is added to the historical state detection data sequence of the second time period to participate in the state prediction of the second time period. At the same time, the first detection strategy after initial correction is directly used as the second detection strategy for the building maintenance personnel to execute in the second time period.

[0073] For example, if the first deviation type is determined to be an accelerating trend, and the first deviation type does not meet the deviation conditions, the prediction model is reconstructed. Based on the prediction model, the predicted crack activity and the second state prediction data are recalibrated. The first detection strategy after the initial correction is then corrected again using the adjusted crack activity and the second state prediction data to obtain the second detection strategy, which is used by building maintenance personnel to execute in the second time period.

[0074] For example, if the first deviation type is determined to be an anomalous change, and the first deviation type meets the deviation condition, the first time period is designated as an anomalous data point. The anomalous events of the anomalous data points are recorded separately. That is, the first state detection data of the first time period is not added to the historical state detection data sequence of the second time period and does not participate in the state prediction of the second time period. At the same time, the initially corrected first detection strategy is directly used as the second detection strategy for building maintenance personnel to execute in the second time period. In addition, if multiple subsequent time periods maintain the same trend as the first time period, the first time period in which the anomaly occurred is used as the initial time period for future crack state prediction, and the detection strategy is re-initialized and generated.

[0075] The re-initialization of the detection strategy may include using the first time period after the anomaly occurs as the initial time period, and generating initial detection strategy parameters based on the initial state of the crack obtained during the initial time period, the importance of the component where the crack is located, and preset rules. The preset rules may include: increasing the crack length, width, and area of ​​the initial crack state by a preset coefficient to obtain the search area in the initial detection strategy; determining the endpoint focus area in the initial detection strategy based on the endpoint positions of the initial crack state; and determining the bifurcation focus area in the initial detection strategy based on the main direction angle and bifurcation state of the initial crack state.

[0076] In one embodiment, during the initial period, initial crack state data for a wall crack is acquired: the crack length is 120 mm, the maximum width is 0.25 mm, the number of bifurcation points is 0, the continuity status is non-continuous, and the importance level of the component containing the crack is general. Since there are no historical deviations or detection strategy feedback results in the initial period, the initial estimated deviation intensity is set to 0. Based on the current crack bounding box, the length and width of the bounding box are expanded to 1.2 times the original size to obtain the search area in the initial detection strategy. The local detection threshold adjustment range is 0, meaning the local detection threshold for the endpoint focus area is the same as the standard detection threshold. For cracks with two endpoints, the area near the two endpoints is set as the endpoint focus area. No manual review marking is triggered. The re-inspection cycle is set to 30 days.

[0077] In another embodiment, during the initial period, if the acquired initial crack state indicates that the maximum crack width is not less than a preset width threshold of 0.50 mm, the number of bifurcation points is not less than one, the crack has already penetrated, or the crack is located in an important component such as a beam or column, the length and width of the current crack outer frame are expanded to 1.5 times the original size to obtain the search area in the initial detection strategy; the area near the crack endpoint and the area near the existing bifurcation points are set as the endpoint focus area; the local detection threshold is adjusted by -10%, that is, the local detection threshold of the endpoint focus area is reduced by 10% relative to the standard detection threshold, in order to improve the detection sensitivity of weak crack extension and bifurcation response; a manual review mark is triggered; and the re-inspection cycle is set to 7 days. The second detection strategy is generated in the first period and is a parameter for assisting building maintenance personnel in performing key crack detection in the second period. It includes at least one of the following: search area expansion direction, search area expansion ratio, local detection threshold adjustment, estimated crack endpoint focus area, estimated crack bifurcation focus area, manual re-inspection mark, or suggested re-inspection cycle.

[0078] In operation S205, the risk identification model is used to process historical state detection data, first state detection data, crack change trend parameters, and first deviation type to obtain crack risk information.

[0079] Based on historical state detection data, first state detection data and environmental features, a crack-environment coupled representation is constructed. The crack-environment coupled representation, crack change trend parameters and first deviation type are input into the risk identification model to obtain crack risk information.

[0080] Risk identification models can be constructed based on fully connected networks, attention networks, recurrent networks, encoder Transformers, or other nonlinear mapping structures.

[0081] Crack risk information can include crack risk level, crack risk trend, and crack early warning level.

[0082] In operation S206, based on crack risk information and the second detection strategy, a second risk control strategy for the second time period is obtained.

[0083] By integrating crack risk information and the second detection strategy, a second risk management strategy for the second time period is obtained.

[0084] The second risk management strategy should include at least one of the following: crack risk trend, crack early warning level, direction of expansion of detection area, proportion of expansion of detection area, adjustment range of local detection threshold, key areas of concern for predicted crack endpoints, key areas of concern for predicted crack bifurcation, manual re-inspection marking or recommended re-inspection cycle.

[0085] In the second phase, building maintenance personnel conducted crack condition detection according to the second risk management strategy, thereby obtaining the condition detection data for the second phase.

[0086] According to embodiments of this application, for each crack, based on continuous multi-time period state monitoring data, crack activity, trend parameters, and second-time period state prediction data are estimated, thereby providing a quantitative basis for adjusting the detection strategy. The deviation type is determined by the discrepancy between the first-time period detection data and the first-time period state prediction data, and this deviation type is used as a basis for judging the crack evolution state, thus achieving feedback verification of the first detection strategy's execution effect. Based on the second-time period state prediction data, crack activity, and deviation type, the first detection strategy is adaptively corrected, ensuring that the adjusted second detection strategy accurately matches the actual crack evolution state. This realizes the conversion of the second-time period crack state prediction data into specific execution parameters for the second-time period detection strategy, forming a… A closed-loop mechanism of "detection strategy generation - detection strategy execution - deviation verification - detection strategy adjustment" is implemented. Furthermore, a risk identification model is used to integrate multi-dimensional historical state detection data, first-state detection data, crack change trend parameters, and first deviation type to obtain crack risk information. Based on this crack risk information and a second detection strategy, a risk management strategy for the second time period is generated. This achieves dynamic coupling between detection resource allocation and risk status, significantly improving the timeliness of crack risk early warning, the accuracy of detection strategies, and the utilization efficiency of detection resources. Targeted second risk management strategies in the second time period reduce building management risks caused by large-scale, unfocused detection or maintenance methods, lower building crack intervention costs, and optimize building maintenance resource allocation.

[0087] According to embodiments of this application, both historical state detection data and first state detection data include geometric type data, count type data, and state type data. Specifically, based on the historical state detection data and the first state detection data, a state assessment of the crack is performed to obtain crack change trend parameters and second state prediction data for the crack in a second time period. This includes: analyzing the geometric type data using the derivative of a geometric prediction model to obtain a crack structure change trend, where the crack structure change trend includes at least one of change rate, acceleration, fluctuation coefficient, and trend fitting degree; processing the count type data using a count prediction model to obtain the probability of new bifurcation; processing the state type data using a state prediction model to obtain the probability of crack penetration state transition, where the crack change trend parameters include at least one of crack structure change trend, probability of new bifurcation, and probability of crack penetration state transition; and performing state prediction on the crack based on the geometric type data using a geometric prediction model to obtain second state prediction data, where the second state prediction data includes at least one of crack geometric prediction location, probability of bifurcation number change, bifurcation candidate region, penetration state change trend, and location fluctuation range.

[0088] In one embodiment, the first-state detection data and historical-state detection data of cracks in the target building can be obtained from any one or more of the following: image detection results, manual inspection records, crack meter or displacement sensor measurement results, historical monitoring database, or output results from other crack monitoring systems.

[0089] Crack status detection data includes crack instance ID, detection period, location on the component, component type, material type, crack skeleton or centerline, crack endpoint coordinates, crack main direction angle, crack length, average width, maximum width, area, number of bifurcation points, bifurcation point location, continuity status, changes in each attribute relative to historical periods, and time-normalized rate of change.

[0090] Crack condition monitoring data is categorized into geometric data, count data, and condition data based on crack attributes. Geometric data includes one or more of the following: length, average width, maximum width, area, principal orientation angle, and endpoint displacement. Count data includes the number of crack bifurcation points. Condition data includes data characterizing the crack's continuity.

[0091] In one embodiment, a geometric prediction model is established for the geometric type data of each crack.

[0092] For the The crack in the first Historical state detection data and first state detection data in the geometric attribute dimension of a geometric type of data, sequence Recorded as:

[0093] (1).

[0094] in, For the detection time, k ranges from 1 to n. For the crack in the first Geometric data values ​​for each time period, where, For the first state detection data, when k is less than n, 'c' represents historical state detection data, and 'c' represents data of geometric type.

[0095] By analyzing geometric data using the derivative of the geometric prediction model, the trend of crack structure change is obtained. The trend of crack structure change includes at least one of the following: rate of change, acceleration, fluctuation coefficient, and trend fit.

[0096] The rate of change represents how fast a geometric property changes over time, and can be obtained from the first derivative of the geometric prediction model or the difference between adjacent time periods.

[0097] Acceleration indicates whether the rate of change of a geometric property is continuously increasing or decreasing, and can be obtained from the second derivative of the geometric prediction model or the difference of the continuous rate of change.

[0098] The volatility coefficient represents the degree of volatility of the first-state detection data relative to the geometric prediction model, and can be obtained by the ratio of the residual standard deviation to the attribute mean or trend value.

[0099] Trend fit indicates the degree to which the geometric prediction model interprets historical state detection data, and can be expressed using the coefficient of determination, residual statistics, or goodness-of-fit index.

[0100] In one embodiment, a count prediction model is established for count-type data.

[0101] Using either a counting prediction model or a change probability prediction model, output the predicted number of forks or the probability of new forks in the second time period:

[0102] (2).

[0103] Where n represents the first time period, h represents the prediction step size after the first time period, and n+h represents the second time period. Indicates the first The number of bifurcation points of the crack in the first time period, where br indicates that the data is of the count type. Indicates the second time period Number of bifurcation points per predicted step size This indicates the probability of a new fork occurring in the second period. The probability of a new fork indicates the likelihood that a new fork will occur in the second period.

[0104] In one embodiment, a state prediction model is established based on state type data. The prediction of the crack penetration state transition probability, threshold discrimination, or classification model is performed to estimate the penetration state in the second time period. The result is expressed as follows:

[0105] (3).

[0106] in, Indicates the first The penetration state of the crack in the first time period n. Indicates the first The crack was in a continuous state during the second time period n+h. Indicates the first The status type data in the historical status detection data of the crack. Indicates the first The environmental feature vector corresponding to each crack, where "through" represents state-type data.

[0107] The probability of a crack transitioning from a non-through state to a through state or remaining through a crack represents the likelihood of the crack transitioning from a non-through state to a through state or remaining through a crack.

[0108] In one embodiment, based on the prediction model, one or more of the following are predicted: geometric value of the crack in the second period or multiple periods later than the first period: predicted location of the crack endpoint, predicted value of bifurcation, trend of change of penetration state and range of position fluctuation, to obtain second state prediction data.

[0109] For example, geometric prediction values ​​may include geometric prediction location, geometric prediction quantity, geometric prediction length, and geometric prediction width; crack endpoint prediction location may include prediction extension direction and prediction extension length; and bifurcation prediction values ​​may include the probability of change in the number of bifurcations and bifurcation candidate regions.

[0110] Geometric location estimation involves extrapolating the attribute values ​​of the h-th prediction step in the second time period using a geometric estimation model based on geometrically typed data, to obtain the geometric estimate. .

[0111] The estimated location of crack endpoints includes, for cracks with two endpoints, allocating the estimated length increment to both endpoints. Let the... The two ends of the crack are respectively and The estimated increase in the length of the second period is The extension weights at the two endpoints are and ,and . No. The first crack Endpoint extension weights of each endpoint It can be determined by a combination of historical endpoint displacement, local endpoint orientation, weak crack response in grayscale or texture near the endpoint, and environmental exposure orientation:

[0112] (4).

[0113] in, Indicates the first The first crack Endpoint feature vectors of each endpoint Indicates the first The first crack Endpoint feature vectors of each endpoint This represents the weight parameter vector corresponding to the endpoint feature vector.

[0114] For example, Indicates the first The first crack Endpoint feature vectors of each endpoint This represents the corresponding weight parameter vector.

[0115] Predicted location of the second phase endpoint of the crack Represented as:

[0116] (5).

[0117] in, Indicates the first The first crack The location of the crack endpoints in the first time period of each endpoint. Indicates the first The first crack Endpoint extension weights of each endpoint Indicates the first The estimated increase in the length of the crack from the first time period to the second time period. Indicates the first The first crack The local extension direction unit vector of each endpoint.

[0118] Based on the probability of changes in the number of bifurcations, candidate bifurcation areas, and trends in the continuity status, the candidate areas for new bifurcations that may occur in the second period are estimated, considering the probability of new bifurcations, historical locations of new bifurcations, the morphology near the current endpoint, and the response to weak cracks. Based on the probability of continuity status transition, the distance from the endpoint to the edge of the component, and the continuity determination threshold, the trend of continuity status changes is predicted.

[0119] Predicting the range of location fluctuations includes determining the range of location fluctuations based on historical residuals, residual distributions, quantile regression, Bayesian estimation, model ensemble variance, or a combination thereof. As an optional method, the range of location fluctuations... Represented as:

[0120] (6).

[0121] Where n represents the first time period, h represents the prediction step size after the first time period, and n+h represents the second time period. Indicates the first Crack No. Geometric data in the second time period The positional fluctuation range of each predicted step size. For the first Crack No. Geometric estimates of geometric data at the h-th prediction step in the second time period, where c represents data of geometric type. The standard deviation of historical residuals. and These are the lower and upper bound coefficients that vary with the prediction step size.

[0122] According to the embodiments of this application, multiple attributes of the same crack instance are jointly modeled, and geometric prediction model, count prediction model and state prediction model are constructed for geometric type data, count type data and state type data. These models can respectively predict the output geometric attribute prediction value, the probability of bifurcation number change, the trend of penetration state change and the location fluctuation range, providing a data foundation for subsequent generation of detection strategies and realizing multi-attribute state prediction of crack instances.

[0123] According to an embodiment of this application, the geometric prediction model is determined based on the following operations: multiple candidate models are used to fit the sample geometric type data of the crack to obtain multiple sample state prediction data; the fitting residuals are obtained based on the sample geometric type data and the sample state prediction data; and the candidate model corresponding to the smallest fitting residual among the multiple fitting residuals is determined as the geometric prediction model.

[0124] In one embodiment, multiple candidate models are fitted to the geometric type data of the historical state detection data, and the candidate models include two or more of the following: linear model, exponential model, logarithmic model, piecewise linear model, multinomial model, or smooth time series model.

[0125] Multiple candidate models are used to fit the sample geometric data of cracks to obtain multiple sample state prediction data. Based on the sample geometric data and sample state prediction data, the fitting residuals are obtained.

[0126] The candidate model with the smallest fit residual among multiple fit residuals is determined as the geometric prediction model. That is, the optimal candidate model among the candidate models for geometric data is selected as the geometric prediction model using the root mean square error (RMSE) of the fit residuals.

[0127] (7).

[0128] in, Indicates the first One candidate model, Indicates the first Crack No. The geometric type data in the first The first time period adopts the first The fitted values ​​of each candidate model, Characterizing the first The root mean square of the fitting residuals of each candidate model.

[0129] In one embodiment, determining the geometric prediction model from candidate models may also include prioritizing the model with lower model complexity when the RMSE difference between two candidate models is less than a preset threshold.

[0130] According to embodiments of this application, by determining the optimal candidate model as the geometric prediction model, the sum of squared residuals is minimized, the prediction error is reduced, and the prediction accuracy of the trend of geometric properties is improved.

[0131] According to an embodiment of this application, a second detection strategy is obtained by adjusting the first detection strategy in the first risk control strategy based on the second state prediction data, crack activity, and a first deviation type. This includes: determining crack activity based on crack change trend parameters and a preset threshold; obtaining a priority based on the first state detection data and the second state prediction data, where the priority represents the urgency of detecting multiple cracks; initially adjusting the first detection strategy based on the priority and crack activity to obtain an initial detection strategy; and further adjusting the initial detection strategy based on the first state detection data and the first deviation type to obtain the second detection strategy.

[0132] In one embodiment, the crack activity is determined by combining one or more crack change trend parameters, such as the crack structure change trend, the probability of new bifurcation, and the probability of crack penetration state transition, output by the prediction model, with a preset threshold.

[0133] Crack activity states include stable state, uniform propagation state, accelerated propagation state, decelerated propagation state, and abnormal fluctuation state.

[0134] A stable state is a crack state where the rate of change of geometric data is lower than a preset rate threshold, the fluctuation coefficient is lower than a preset fluctuation threshold, and the probability of new bifurcation of count data and the probability of through-state transition of state data are both lower than preset probability thresholds.

[0135] The uniform expansion state is the crack state when the rate of change of geometric data is not lower than a preset rate threshold and the absolute value of acceleration is lower than a preset acceleration threshold.

[0136] The accelerated expansion state is a crack state in which the acceleration of geometric data exceeds a preset positive threshold, or in which either the probability of a new bifurcation in count data or the probability of a through state transition in state data exceeds a preset probability threshold.

[0137] The deceleration expansion state is a crack state in which the acceleration of the geometric data is less than a preset negative threshold and the rate of change is not lower than a preset rate threshold.

[0138] For example, using three consecutive detection periods as the state determination window; when the time interval between adjacent detection periods is the same, or when the crack change rate is time-normalized, if the length growth rate of a certain crack in the three consecutive detection periods is 4.8%, 2.6%, and 1.2% respectively, and the maximum width growth rate is 3.5%, 1.8%, and 0.8% respectively, it indicates that the crack's change rate is continuously decreasing. Simultaneously, the positional fluctuation range of the length and maximum width does not exceed 10% of the corresponding estimated value, the probability of a new bifurcation is less than 0.2, the probability of a through-state transition is less than 0.1, and the first state detection data for the three consecutive detection periods are all within the corresponding positional fluctuation range. When the above conditions are met, the crack activity is determined to be in a decelerating expansion state.

[0139] Abnormal fluctuation state refers to the crack state when the fluctuation coefficient of geometric type data is not lower than the preset fluctuation threshold, the trend fitting degree is lower than the preset fitting degree threshold, or the prediction results of one or more of the count type data and state type data are significantly inconsistent with the historical change pattern.

[0140] The aforementioned preset thresholds can be determined based on historical state detection data statistics, sample quantiles, engineering experience, component type, material type, or adaptive learning methods.

[0141] In one embodiment, a first detection strategy is executed to obtain first-state detection data, and a second-state prediction data and crack activity are obtained through a prediction model. A priority is then determined based on the first-state detection data and the second-state prediction data.

[0142] Based on priority and crack activity, the first detection strategy is initially adjusted to obtain the initial detection strategy.

[0143] The first detection strategy includes one or more of the following: search area expansion direction, search area expansion ratio, local detection threshold adjustment range, key areas of focus for predicted crack endpoints, key areas of focus for predicted crack bifurcation, manual re-inspection marking, or suggested re-inspection cycle.

[0144] The initial detection strategy is based on the first detection strategy, and the detection execution parameters are adjusted according to priority and crack activity. The adjustment operations include whether to prioritize detection, whether to expand the search area, whether to increase the search range in the endpoint direction or vertical direction, whether to adjust the local detection threshold, whether to trigger manual review, and whether to shorten or extend the re-inspection cycle.

[0145] A deviation analysis is performed based on the discrepancy between the first-state detection data and the first-state prediction data to determine the first deviation type. Based on the first-state detection data and the first deviation type, the initial detection strategy is adjusted to obtain a second detection strategy. This adjustment includes revising the crack prediction model to address the first deviation type.

[0146] For example, if the first deviation type is determined to be a normal deviation, the first state detection data is added to the historical state detection data, and the second state prediction data for the second time period is not updated; if the first deviation type is determined to be an accelerating trend, the crack state prediction model is evaluated, and if necessary, the prediction model is redefined, the prediction parameters are recalibrated, and then the adjusted prediction model is used to make a new prediction to obtain the updated second state prediction data for the second time period; if the first deviation type is determined to be an abnormal change, the abnormal data points are recorded separately as abnormal events and are not added to the historical state detection data. If the new trend of change is maintained in multiple subsequent time periods, the time period in which the abnormality occurred is taken as the first time period of the new crack state prediction, and the second state prediction data for the second time period is not updated.

[0147] When the second-state predicted data is updated, the priority is updated to obtain the updated priority, and the initial detection strategy is adjusted to obtain the second detection strategy.

[0148] If the predicted data for the second state is not updated, the initial detection strategy is determined to be the second detection strategy.

[0149] According to embodiments of this application, a first detection strategy is adjusted based on calculated crack priority and crack activity, and a second detection strategy is generated by combining a first deviation type. This achieves differentiated detection strategies for cracks in different states, converting the prediction results into detection execution parameters for the second time period. By adjusting the prediction method and thus the detection strategy through the deviation between the first state detection data and the first state prediction data, the utilization rate of the deviation between the first state detection data and the first state prediction data is improved.

[0150] According to an embodiment of this application, priority is obtained based on first-state detection data and second-state prediction data, including: determining the historical environmental driving intensity based on historical environmental data; extracting structural attributes from the first-state detection data to obtain the geometric expansion intensity; and determining the priority of the crack based on the historical crack risk level, second-state prediction data, historical deviation, historical environmental driving intensity, geometric expansion intensity, and estimated detection cost.

[0151] In one embodiment, the priority of each crack is calculated based on the first state detection data and the second state prediction data, including: determining the priority based on historical environmental data, the first state detection data, historical crack risk level, the second state prediction data, historical deviation, historical environmental driving intensity, geometric propagation intensity, and estimated detection cost.

[0152] In one embodiment, the first Crack priority The calculation formula is:

[0153] (8).

[0154] in, Indicates the first The historical crack risk level that can be obtained in the first time period is dimensionless. Indicates the first The second-state prediction data for the crack is dimensionless. Indicates the first The historical deviation of the crack is generated based on condition monitoring data and condition prediction data for historical periods, and is dimensionless. Indicates the first The historical environmental driving intensity of the crack is dimensionless. Indicates the first The geometric propagation strength of the crack is dimensionless; Indicates the first The estimated detection cost of the crack before generating the second detection strategy is dimensionless. , , , , , These are the weighting coefficients.

[0155] Historical crack risk level This includes crack risk level and crack warning level conversion values ​​from crack risk information output in historical time periods, or the initial risk level obtained from crack activity status, continuity status, component importance, and preset rules when the first time period is the first detection period.

[0156] Second state prediction data It is determined by one or more of the following: geometric estimate, estimated location of crack endpoint, estimated bifurcation estimate, trend of change in penetration status, and range of location fluctuation.

[0157] In one embodiment, the second state prediction data Represented as:

[0158] (9).

[0159] in, This represents the geometric estimate. This indicates the estimated value of the fork. This indicates the trend of changes in the connected state. and These represent the weights of the estimated bifurcation value and the trend of the through-through state, respectively.

[0160] Geometric Prediction It can be represented as:

[0161] (10).

[0162] in, This indicates the number of geometric attribute dimensions in geometric data types. Indicates the first Weights of each geometric attribute dimension, Indicates the first Crack No. The width of the positional fluctuation range of each geometric attribute dimension Indicates the first The historical mean or current value of the geometric attribute in each geometric attribute dimension. To prevent constants with a denominator of zero.

[0163] Position fluctuation range width The upper bound of the position fluctuation range can be determined by and the lower realm Sure:

[0164] (11).

[0165] When the location fluctuation range adopts the formula When expressing, the width of the positional fluctuation range It can also be expressed as:

[0166] (12).

[0167] in, and These are the lower and upper bound coefficients that vary with the prediction step size. For the first Crack No. The historical residual standard deviation of each geometric attribute dimension.

[0168] Fork Estimated Value The probability of a new bifurcation can be determined by the entropy, variance, model ensemble variance, or the degree to which the model output probability approaches the bifurcation determination threshold.

[0169] Trend of changes in the through-connection status It can be determined by the entropy, variance, model integration variance, or the degree to which the breakthrough probability approaches the breakthrough judgment threshold.

[0170] In one embodiment, the fork estimate Trend of changes in the connected state It can be represented as:

[0171] (13).

[0172] (14).

[0173] in, This indicates the probability of a new fork occurring in the second time period. This indicates the crack penetration status in the second time period.

[0174] Historical bias Can be from the most recent The calculation of the extent to which the forecast deviation exceeds the limit for each historical period is expressed as follows:

[0175] (15).

[0176] in, Indicates the first Crack No. The geometric attribute dimension in the th dimension The deviation between the first state detection data and the second state prediction data for each time period This indicates the width of the fluctuation range at the corresponding location. Indicates the first Deviation weights for each geometric attribute dimension This indicates the number of geometric attribute dimensions in geometric data types. To prevent constants with a denominator of zero. Historical bias. The larger the value, the more unstable the historical prediction results or the more the crack evolution deviates from the original trend; when it is necessary to consider the number of bifurcations or the continuity status at the same time, the deviation of the probability of change in the number of bifurcations and the deviation of the trend of change in the continuity status can be included as additional deviation terms. .

[0177] Historical environment driving intensity The data is determined by a combination of second-period environmental forecast data, first-period environmental data, historical environmental data, the spatial location of the component containing the crack, and the material sensitivity coefficient. The first-period environmental data and historical environmental data are used to align with the time interval of crack changes and include at least one or more of the following: temperature, humidity, rainfall, freeze-thaw cycle, solar radiation intensity, wind speed, wind direction, and wind load. The second-period environmental forecast data is used for predicting the risk trend in the second period. For example, the second-period environmental forecast data includes forecasts for second-period temperature, humidity, rainfall, or wind load.

[0178] In one embodiment, historical environment-driven intensity Represented as:

[0179] (16).

[0180] in, Indicates the number of categories of environmental factors. Indicates the first Environmental factor weights for each geometric attribute dimension Indicates the relationship with the first The first crack after spatiotemporal alignment Normalized historical environmental data features or second-period environmental forecast data features based on geometric attribute dimensions. Indicates the first The crack is related to the first Material sensitivity coefficient, spatial exposure coefficient, or a combination of both, for environmental factors in a geometric dimension.

[0181] Geometric extension strength It is obtained by extracting structural attributes from the first state detection data, or determined based on one or more of the following: estimated length change, width change, area change, endpoint displacement, number of bifurcations, or through-state change.

[0182] The constraints on inspection resources include available inspection time, number of inspection equipment, manual verification capability, estimated inspection cost per crack, estimated area cost of the inspection area, or the resource occupancy of intensified monitoring.

[0183] Estimated testing cost The priority is determined by one or more of the following factors: current crack geometry, historical search area, conventional detection method, equipment occupancy, and preliminary manual review requirements. This priority is used to calculate the priority before generating the second detection strategy.

[0184] In one embodiment, the estimated detection cost Represented as:

[0185] (17).

[0186] in, Indicates according to the first The scalar value of the basic search area estimated from the outer region of the crack, the historical search region, or the conventional expansion region is dimensionless; Indicates the first The cost scalar of the detection method used for each crack is dimensionless. Indicates the first The crack is initially marked with a 0 or 1 to determine whether manual verification may be required based on the rules. , , This represents the weighting coefficient. After generating the second detection strategy configuration for the second time period, the execution cost can be recalculated based on the actual search area and manual re-inspection markers for subsequent resource statistics or strategy feedback.

[0187] According to embodiments of this application, by calculating priorities based on estimated detection costs, resource utilization is improved when detection resources are limited, avoiding omissions in detecting or intervening in high-risk cracks, and optimizing the allocation of building maintenance resources. Utilizing historical crack risk levels to calculate priorities achieves the technical effect of interconnecting crack risk information with specific detection execution parameters for the second time period, enabling dynamic adjustment of detection execution parameters based on environmental monitoring data.

[0188] Figure 3 A schematic diagram illustrating the generation and adjustment of a second detection strategy in a building crack risk management method according to an embodiment of this application is shown.

[0189] like Figure 3 As shown, the first state detection data 302 and the historical state detection data 303 are input into the prediction model 305 to obtain the second state prediction data 307 and the crack change trend parameter 313, and the crack activity 308 is determined according to the crack change trend parameter 313 and the preset threshold.

[0190] Based on historical environmental data, historical deviations, estimated detection costs 306, second-state estimated data 307, and first-state detection data 302, priority 309 is generated. Historical environmental data is used to generate historical environmental driving strength, and first-state detection data 302 is used to extract geometric expansion strength.

[0191] Based on priority 309 and crack activity 308, the first detection strategy 310 is adjusted to obtain the initial detection strategy 311. The first detection strategy 310 is used to execute detection parameters to obtain first-state detection data 302. The first detection strategy includes one or more of the following: search area expansion direction, search area expansion ratio, local detection threshold adjustment range, estimated crack endpoint focus area, estimated crack bifurcation focus area, manual re-inspection marking, or suggested re-inspection cycle.

[0192] Based on the deviation between the first state prediction data 301 and the first state detection data 302, the first deviation type 304 is determined.

[0193] Based on the first deviation type 304, adjust the prediction model 305, update the priority 309 and crack activity 308.

[0194] In one embodiment, if the first deviation type 304 is determined to be a normal deviation, the first state detection data 302 is added to the historical state detection data sequence and used as the historical state detection data for the second time period for state prediction, and the initial detection strategy 311 is directly determined as the second detection strategy 312.

[0195] In one embodiment, if the first deviation type 304 is determined to be an accelerating trend, the crack state prediction model 305 is evaluated, the prediction model 305 is redefined, the prediction parameters are recalibrated, and then the adjusted prediction model 305 is used to make a new prediction, resulting in updated second state prediction data 307 and crack activity 308. Based on the updated second state prediction data 307, the first deviation type 304 and priority 309 are updated, and then based on the updated crack activity 308 and priority 309, the initial detection strategy 311 is further adjusted based on the updated first deviation type 304 to obtain the second detection strategy 312.

[0196] In one embodiment, if the first deviation type 304 is determined to be an abnormal mutation, the first state detection data 302 of the abnormal data point is recorded separately as an abnormal event and is not added to the historical state detection data 303. The initial detection strategy 311 is directly determined as the second detection strategy 312. In addition, if the new change trend is maintained in multiple subsequent time periods, the time period in which the abnormality occurred is taken as the starting time period of the new crack state prediction, and the initialized second detection strategy 312 is regenerated.

[0197] According to embodiments of this application, the first detection strategy includes at least one of a search area, an endpoint interest area, a bifurcation interest area, a local detection threshold, a manual re-inspection marker, a standard detection threshold, and a re-inspection cycle; the second state prediction data includes the estimated extension direction, estimated extension length, and position fluctuation range of the endpoint; wherein, the first detection strategy is initially adjusted according to priority and crack activity to obtain an initial detection strategy, including: when crack activity is in a stable state, reducing the search area or extending the re-inspection cycle in the first detection strategy to obtain the initial detection strategy; when crack activity is in a uniform expansion state, adjusting the search area or extending the re-inspection cycle according to the estimated extension direction, estimated extension length, and position fluctuation range of the endpoint; The search area in the first detection strategy is expanded to increase the range of positional fluctuations, resulting in an initial detection strategy. When the crack activity is in an accelerated propagation state and the environmental driving intensity is greater than the preset driving threshold, the search area in the first detection strategy is expanded according to the estimated extension length and the preset acceleration safety factor, and the local detection threshold is adjusted in the endpoint interest area and the bifurcation interest area to obtain the initial detection strategy. When the crack activity is in a decelerated propagation state, the preset benchmark detection strategy is determined as the initial detection strategy. When the crack activity is in an abnormal fluctuation state, the search area in the first detection strategy is expanded, the standard detection threshold is increased, and manual re-inspection marks are added to obtain the initial detection strategy.

[0198] In one embodiment, the second-time detection strategy It can be represented as:

[0199] (18).

[0200] in, Indicates the direction of search area expansion. Indicates the expansion ratio of the search area. This indicates the adjustment range of the local detection threshold. This indicates areas of particular interest. Indicates manual re-inspection mark. This indicates the recommended re-inspection period. Furthermore, It can be along the main direction of the crack, the local direction at the end point, perpendicular to the main direction, or a combination of multiple directions; An expansion ratio greater than or equal to 1; This refers to the adjustment amount or percentage relative to the standard detection threshold. This may include the estimated endpoint attention area, the estimated bifurcation attention area, or the high uncertainty area; Take 0 or 1; The unit can be day, week, or month.

[0201] An initial detection strategy is generated based on priority and crack activity. The initial detection strategy is generated differently based on priority and crack activity, specifically including the following cases.

[0202] When the crack activity is stable, for cracks with low prediction uncertainty, the search area is slightly expanded based on the historical external rotation frame, historical crack spatial range, historical detection frame, historical crack area outline, or crack area output by the upstream detection system, while maintaining the standard detection threshold or appropriately extending the re-inspection cycle.

[0203] When the crack activity is in a state of uniform expansion, the search area is expanded according to the estimated extension direction and estimated extension length and position fluctuation range in the estimated location of the crack endpoint, and the detection threshold is adjusted in the area near the estimated endpoint to improve the ability to capture the weak extension of the endpoint.

[0204] When the crack activity is in an accelerated propagation state and the environmental driving intensity is greater than the preset driving threshold, the search area is expanded by multiplying the estimated extension length by the preset acceleration safety factor. At the same time, the search width perpendicular to the main direction of the crack is increased, and the local detection threshold is adjusted in the estimated endpoint area and the estimated bifurcation area. Meanwhile, when the priority exceeds the preset threshold, manual re-inspection marking is triggered or the re-inspection cycle is shortened.

[0205] When the crack activity is in a decelerating propagation state, for cracks with a continuously decreasing rate of change and low prediction uncertainty, the execution parameters of the initial detection strategy are gradually restored to the preset benchmark detection strategy, while the monitoring of abnormal width increases and changes in the continuity state is retained.

[0206] For example, a preset benchmark detection strategy may include: based on the historical detection frame of the crack, expanding the length and width of the historical detection frame to 1.1 times the original size to obtain the search area in the preset benchmark detection strategy, without adding additional search ranges in the direction of crack endpoint extension or perpendicular to the main direction of the crack; maintaining the standard detection threshold, i.e., the local detection threshold adjustment relative to the standard detection threshold is 0; not triggering manual review marking; and setting the re-inspection cycle to 30 days. Simultaneously, in subsequent time periods, it continues to detect whether the maximum width of the crack increases abnormally and whether the continuity status changes.

[0207] When the crack activity is in an abnormal fluctuation state or the historical deviation continues to exceed the width of the location fluctuation range, the search area is expanded to cover multiple possible change directions, the standard detection threshold is maintained or the threshold adjustment constraint is increased to reduce false detections and trigger manual re-inspection marking.

[0208] The aforementioned expansion ratio, safety factor, acceleration factor, and threshold adjustment range can be determined through preset rules, historical statistics, or adaptive learning.

[0209] According to the embodiments of this application, the initial detection strategy for cracks in the second time period is determined by priority and crack activity. Differentiated detection execution parameters are generated for each crack instance, providing engineering-practical detection suggestions such as whether to prioritize detection, whether to expand the search area, whether to increase the search range in the endpoint direction or vertical direction, whether to adjust the local detection threshold, whether to trigger manual review, and whether to shorten or extend the re-inspection cycle, thereby improving the rationality and utilization of resource allocation.

[0210] According to an embodiment of this application, a deviation analysis is performed on the deviation between the first state detection data and the first state prediction data to identify normal deviation, trend acceleration, or abnormal sudden change events, and to determine the first deviation type.

[0211] Figure 4 A schematic diagram illustrating the deviation analysis of the building crack risk management method according to an embodiment of this application is shown.

[0212] like Figure 4 As shown, there is a deviation between the first state prediction data 401 and the first state detection data 402 after actual detection. For example, the detection endpoint is not at the predicted endpoint position, and the newly added fork exceeds the position fluctuation range.

[0213] In one embodiment, the deviation between the first state detection data 402 and the first state prediction data 401 is calculated for each attribute dimension of each crack. :

[0214] (19).

[0215] in, Indicates the first Crack No. The geometric attribute dimension in the th dimension Deviation over a period of time, Indicates the first Crack No. The geometric attribute dimension in the th dimension The first state detection data for each time period Indicates the first Crack No. The geometric attribute dimension in the th dimension First state prediction data for each time period.

[0216] Calculate the positional relationship between the deviation and the range of positional fluctuation to determine the first deviation type 403.

[0217] If the first state detection data falls within the position fluctuation range, then the first deviation type is determined to be a normal deviation type, indicating that the crack evolution still conforms to the original trend.

[0218] If the first state detection data exceeds the position fluctuation range and the deviation direction is consistent with the original trend direction, then the first deviation type is determined to be the trend acceleration type, indicating that the crack development speed exceeds the original prediction.

[0219] If the first state detection data exceeds the position fluctuation range and the deviation direction is opposite to the original trend direction, or multiple key attributes show inconsistent deviations at the same time, then the first deviation type is determined to be an abnormal mutation type.

[0220] Abnormal mutation events of the deviation type include one or more of the following: accelerated deterioration, sudden convergence, morphological mutation, new bifurcation anomaly, and through state mutation. The abnormal event type, occurrence time, involved attributes, and deviation magnitude are output to the risk identification model.

[0221] The identified first deviation type can be used to adjust various detection execution parameters in the first detection strategy 404, such as crack change trend parameters, search area expansion ratio, local detection threshold adjustment range, manual review mark, priority, and re-inspection cycle.

[0222] According to an embodiment of this application, a first detection strategy includes an estimated search area, and first state detection data includes the crack extension location. The first detection strategy is adjusted based on the first state detection data and a first deviation type to obtain a second detection strategy, including: determining the first detection strategy as the second detection strategy when the crack extension location is within the estimated search area and the first deviation type is a consistent trend; adding the first state detection data to the historical state detection data for the second time period; adjusting the expansion ratio of the estimated search area in the estimation model or the initial detection strategy when the crack extension location is outside the estimated search area or the first deviation type is an accelerating trend; and clearing the first state detection data or reconstructing the data using the first state detection data as the initial point when the first deviation type is an abnormal trend.

[0223] In one embodiment, when the actual crack extension location of the first state detection data is within the predicted search area of ​​the first state prediction data, and the first deviation type is a normal deviation type, the first detection strategy configuration is determined to be effective, the initial detection strategy is determined to be the second detection strategy, and the first state detection data is added to the historical state detection data for the second time period, and the prediction model is refitted for the second time period.

[0224] If the crack extends outside the estimated search area, the new bifurcation appears in an uncovered area, or the first deviation type is a trend acceleration type, the search area expansion is deemed insufficient. In subsequent periods, the prediction uncertainty weight of the crack is increased or the estimated search area expansion ratio in the initial detection strategy is increased, or the candidate model is re-evaluated. If necessary, the model is switched to an exponential model, a piecewise linear model, or a higher-order trend model, and the prediction model is redefined. The second-state prediction data and the initial detection strategy are then regenerated.

[0225] When the first deviation type is a trend anomaly, the abnormal data points are not directly involved in the prediction model fitting, but are recorded separately as abnormal events; if the new change level is maintained in multiple subsequent periods, the first period in which the anomaly occurred is taken as the starting point of the new trend to reconstruct the prediction model.

[0226] When the number of false detections generated after adjusting the local detection threshold exceeds the preset threshold, it is determined that the threshold adjustment range is too large, and the threshold adjustment constraint is increased in the initial detection strategy to reduce the threshold adjustment range in low confidence regions.

[0227] When manual verification confirms that a crack has abnormally expanded, changed its continuity, or undergone a sudden change in morphology, the event is input into the risk identification model as an abnormal event, and the priority of the crack is increased in subsequent time periods.

[0228] Based on the deviation analysis, the initial detection strategy, whether adjusted or not, is determined as the second detection strategy.

[0229] According to the embodiments of this application, the detection execution parameters in the detection strategy are dynamically adjusted by utilizing the deviation between the first state detection data and the first state prediction data, thereby improving the utilization rate of the prediction deviation and forming a closed-loop mechanism of "detection strategy generation - actual state detection data verification - detection strategy adjustment". This continuously improves the accuracy of crack prediction results and enables effective early intervention in cracks.

[0230] According to an embodiment of this application, before generating crack risk information using a risk identification model, the method further includes preprocessing historical state detection data, first state detection data, and environmental monitoring data. The preprocessing operations include time alignment, spatial alignment, data cleaning, missing data handling, and sequence standardization.

[0231] In one embodiment, time alignment specifically includes determining a first time period corresponding to the crack state detection data. For the change in crack state detection data corresponding to the crack in the first time period, the environmental data time interval is preset to be the interval Gn between two detection timestamps, which is the time interval between the end time of the historical time period and the start time of the first time period.

[0232] The continuous environmental monitoring data within the time interval Gn is aggregated into the initial environmental data for the first time period. The aggregation operation includes window average, window maximum, window minimum, window fluctuation amplitude, cumulative amount, duration, and extreme event count.

[0233] When the crack detection interval is long, the interval can be further divided into multiple sub-windows to calculate the early, middle and late environmental characteristics respectively, which can be used to describe the stages of environmental impact.

[0234] In one embodiment, spatial alignment specifically includes associating the crack with the nearest environmental monitoring point, the environmental sensor in the area, or the corresponding spatial environmental grid, based on the location of the crack on the component.

[0235] The environmental sensitivity coefficient is set according to the spatial location and exposure conditions of the component. The spatial location includes indoor, outdoor, sunny side, shady side, high position, low position, windward side or leeward side.

[0236] The material sensitivity coefficient is set according to the material type of the component. The material type includes one or more of the following: wood, brick, rammed earth, mortar, stone or mixed materials.

[0237] By combining the initial environmental data of the first time period with the spatial sensitivity coefficient and the material sensitivity coefficient, the first environmental data of the first time period corresponding to a specific crack instance is obtained.

[0238] In one embodiment, data cleaning and sequence standardization includes: interpolating, filling in adjacent time periods, statistically filling, or labeling missing values ​​in the crack attribute sequence and environment sequence; and removing, truncating, or labeling outliers that are significantly outside the physical range of the sensor or the reasonable detection range.

[0239] Crack attributes and environmental features of different dimensions are standardized or normalized to make them suitable for input into subsequent coupled feature learning models.

[0240] By preserving anomaly labeling information, outliers can be used as auxiliary inputs for risk assessment without disrupting model training.

[0241] According to an embodiment of this application, crack risk information is obtained by processing historical state detection data, first state detection data, crack change trend parameters, and first deviation type using a risk identification model. This includes: performing feature fusion on historical state detection data, first state detection data, and first environmental data for a first time period to obtain fused features; and processing the fused features, crack activity, and deviation using a risk identification model to obtain crack risk information.

[0242] In one embodiment, features extracted from historical state detection data, first state detection data, and first environmental data of the first time period are fused to obtain fused features.

[0243] The crack activity level is encoded to obtain crack activity level encoded features.

[0244] The fusion features, crack activity encoding features, and bias are concatenated into an input vector. :

[0245] (20).

[0246] in, Indicates fusion features, This represents the coding feature of crack activity. This indicates the deviation between the first-state detection data and the first-state prediction data.

[0247] The input vector is fed into the risk identification model, which includes a risk value regression branch and a risk trend classification branch.

[0248] Risk value regression branch outputs normalized crack risk level:

[0249] (twenty one).

[0250] in, For risk level calculation function, As the input vector for risk prediction, For logical functions, crack risk level The closer it is to 1, the higher the risk.

[0251] Risk trend classification branch output crack risk trend :

[0252] (twenty two).

[0253] in, A function is used to calculate risk trends. As the input vector for risk prediction, For the normalized exponential function, the crack risk trend Values ​​between 0 and 1 indicate a trend of crack risk. The system determines the trend category based on a preset trend judgment threshold. For example, if p is 0.8, which is greater than the preset trend judgment threshold of 0.7, the trend category is determined as 1, representing an upward trend in risk. The trend category can be 0, 1, or 2, representing three trend categories: rising risk, stable risk, and falling risk, respectively.

[0254] The risk identification model is constructed based on recurrent networks, temporal convolutional networks, Transformers, or other temporal modeling networks.

[0255] The second state prediction data also includes risk trend prediction results. These risk trend prediction results include the predicted trend of crack evolution itself, the predicted trend driven by environmental factors, and the predicted trend of the first type of deviation.

[0256] For example, the predicted trend of crack evolution itself includes the direction of change in rate of change, the sign of acceleration, the trend of endpoint extension, the trend of bifurcation, and the change of continuity.

[0257] For example, environmentally driven forecast trends include second-period forecast data for temperature, humidity, rainfall, freeze-thaw cycles, sunshine, or wind loads, as well as key environmental drivers identified by the coupled weight matrix.

[0258] For example, the first type of deviation is the predicted trend, including whether the actual detected value continuously exceeds the position fluctuation range, whether the deviation converges back to the position fluctuation range, and whether the deviation direction remains consistent.

[0259] The early warning level is determined by classifying the crack risk level, crack risk trend, and risk trend prediction results. This early warning level is then used to assist maintenance personnel in conducting inspections during the second period.

[0260] The early warning level classification adopts a three-level judgment mechanism of "baseline level of crack risk level + crack risk trend adjustment + risk trend prediction result correction".

[0261] The baseline warning level is determined based on the risk level of cracks. The baseline warning level includes normal attention, general warning, relatively serious warning and severe warning.

[0262] The baseline level is adjusted based on the crack risk trend. When the crack risk trend is rising, the warning level is raised by one level; when the crack risk trend is stable, the warning level remains unchanged; when the crack risk trend is falling and continues to do so for a preset time window number of times, the warning level is lowered by one level.

[0263] The warning level is adjusted based on the risk trend forecast results. When the deviation analysis detects an abnormal event, the warning level is at least a general warning. When the crack penetration status changes, the warning level is at least a severe warning. When multiple cracks in the same area are simultaneously under severe or serious warning, the overall warning level of the area is raised by one level.

[0264] In one embodiment, the warning level threshold and level adjustment rules can be preset or dynamically configured according to the building type, component importance, material sensitivity, and protection requirements.

[0265] The crack risk level, crack risk trend, risk trend prediction results, and crack early warning level are output to obtain crack risk information.

[0266] According to the embodiments of this application, the risk trend prediction result and crack warning level for the second time period are obtained by using spatiotemporally aligned crack state data and environmental data. This achieves the technical effect of implementing personalized and fine-grained early warning for cracks. At the same time, the second risk trend prediction result is generated based on the first deviation type, which improves the accuracy of crack dynamic prediction under different environmental and spatial conditions.

[0267] According to an embodiment of this application, feature fusion is performed on historical state detection data, first state detection data, and first environmental data of a first time period to obtain fused features, including: feature extraction on historical state detection data and first state detection data to obtain state features; feature extraction on first environmental data to obtain environmental features; coupling interaction between state features and environmental features to obtain coupling strength; processing coupling strength, state features, and environmental features using a nonlinear mapping function to obtain initial features; and modulating the initial features according to the crack material sensitivity properties to obtain fused features.

[0268] In one embodiment, feature extraction is performed on historical state detection data, first state detection data, and first environmental data for the first time period to obtain state features and environmental features. The environmental feature vector can be constructed according to statistical features, cumulative features, extreme event features, directional relationship features, and sensitivity modulation feature categories.

[0269] For example, statistical characteristics include the average, maximum, minimum, standard deviation, and fluctuation range of environmental variables such as temperature, humidity, and wind speed within a time window.

[0270] For example, cumulative characteristics include cumulative rainfall, cumulative sunshine duration, cumulative high humidity duration, and cumulative strong wind duration.

[0271] For example, characteristics of extreme events include the number of freeze-thaw cycles, the number of consecutive high humidity events, the duration of extreme high temperatures, the duration of extreme low temperatures, and the number of rainstorm events.

[0272] For example, directional relationship characteristics include the angle between the prevailing wind direction and the main direction of the crack, the relationship between the direction of rainfall exposure and the orientation of the component, and the relationship between the direction of sunlight and the orientation of the component.

[0273] For example, sensitivity modulation features include environmental sensitivity coefficients derived from component material, spatial location, and exposure conditions.

[0274] After extracting the state and environmental features, the crack state feature vector is input into the crack change encoder to obtain the crack latent vector. :

[0275] (twenty three).

[0276] in, This represents the feature vector of the crack state. This is a crack change encoder.

[0277] The environmental feature vector is input into the environmental factor encoder to obtain the environmental latent vector. :

[0278] (twenty four).

[0279] in, For environmental feature vectors, An environmental factor encoder.

[0280] In one embodiment, the crack variation encoder and the environmental factor encoder may employ a fully connected network, attention network, recurrent network, Transformer, or other nonlinear mapping structure.

[0281] The interaction relationship between the crack latent vector and the environment latent vector is calculated by the coupling interaction module, and the coupling strength scalar is obtained. and initial features :

[0282] (25).

[0283] (26).

[0284] in, Scalar representing the coupling strength between the crack latent vector and the environment latent vector. Indicates initial features, Represents element-wise product. The coupling weight matrix is ​​a learnable matrix. It is a nonlinear mapping function. The coupling strength scalar. It can be used as a concatenation of one-dimensional features and vector features as input to a nonlinear mapping function. (Frost hidden vector) Latent vectors of the environment It can be adjusted to the same dimension via linear mapping. Dimensions and crack latent vectors Latent vectors of the environment The dimensions match; when the crack latent vector and environment latent vectors When dimensions are inconsistent, the dimensions are first unified through a projection layer, and then element-wise multiplication and coupling strength calculation are performed.

[0285] By introducing a material sensitivity coefficient or a material embedding vector to modulate the initial features, the final fused coupled representation, i.e., the fused features, is obtained. :

[0286] (27).

[0287] in, This represents a sensitivity modulation vector related to material type and component exposure conditions. As initial features, and Having the same dimensions, or adjusted to the same dimensions after linear mapping. Same dimension.

[0288] According to the embodiments of this application, by fusing crack state data parameters with environmental monitoring data, coupling strength and fusion characteristics are obtained, which solves the problem in related technologies that it is difficult to accurately describe the personalized environmental characteristics and state characteristics of a crack at a certain time period. The use of fusion characteristics avoids the problem of using only overall environmental indicators, and provides an accurate data basis for crack early warning.

[0289] According to embodiments of this application, the crack-environment coupling representation model and the risk identification model can be implemented using a trainable method, specifically including processes such as constructing training samples, constructing training labels, inputting them into the model for prediction, and training the model.

[0290] For example, constructing training samples includes building features for each training sample, such as crack change features, environmental features, material / space sensitivity features, crack activity status, and prediction bias features within a time window.

[0291] For example, constructing training labels includes constructing one or more training labels, including risk level labels, risk trend category labels, and abnormal event labels; wherein, risk level labels can be obtained from manual assessment results, historical maintenance records, changes in connectivity status, abnormal expansion events, or expert rules.

[0292] For example, inputting the model for prediction involves feeding training samples into a crack change encoder, an environmental factor encoder, a coupled interaction module, and a risk prediction network to obtain the predicted risk value. Risk trend category probability and probability of abnormal events .

[0293] For example, training the model includes using a joint loss function, with the combined scalar loss value L being:

[0294] (28).

[0295] in, Indicates risk level label, Indicates risk trend category label, Indicates an exception event label, Represents cross-entropy loss, Represents the set of model parameters. , , , To lose weight, To estimate the risk value, For risk trend category probability, This represents the probability of an abnormal event.

[0296] In subsequent monitoring cycles, abnormal crack expansion, changes in continuity, repair and treatment results, or new risk assessment results that have been manually verified can be added to the training samples to periodically update the model or fine-tune it online.

[0297] When there are insufficient training samples or when neural network training is not used, risk prediction can also be implemented using rule-based methods that do not rely on training data, specifically including the following steps.

[0298] Normalized scores were applied to the crack activity state, first deviation degree, environmental driving intensity, probability of change in penetration state, and geometric propagation intensity.

[0299] Risk level is calculated using weighted rules. :

[0300] (29).

[0301] in, This indicates the crack activity level score. Indicating historical bias, Indicates the intensity of environmental driving force. Indicates geometric extension strength, This represents the probability of a change in the connected state. to These are the weighting coefficients. This indicates that the result will be restricted to... Within the range.

[0302] Based on the direction of risk level changes, the number of times the risk level exceeds the range of location fluctuations consecutively, and the direction of changes in the intensity of environmental drivers, the risk trend is determined to be upward, stable, or downward.

[0303] According to the embodiments of this application, crack risk information and a second detection strategy are integrated to output a second risk management strategy. The second risk management strategy includes at least: a first crack risk level, a first crack early warning level, a second risk trend prediction result, the crack activity of each crack instance, crack change trend parameters of each geometric attribute dimension, second state prediction data, location fluctuation range, priority, abnormal event markers, risk driving factors, a second detection strategy, priority re-inspection components or areas sorted by the first crack risk level, recommended re-inspection cycle, recommended intensified monitoring measures, and manual review prompts.

[0304] The second detection strategy includes at least one of the following: search area range, search area expansion direction, search area expansion ratio, estimated endpoint attention area, estimated bifurcation attention area, local detection threshold adjustment range, manual re-inspection marking, and manual review of key locations.

[0305] According to embodiments of this application, the building crack risk management method of the present invention can also be used in conjunction with an upstream crack detection method and a crack cross-time period change quantification method. The crack instances, structured crack attributes, cross-time period correlations, and time-normalized changes output by the upstream method can be used as inputs to this method; the second time period detection strategy, re-inspection area, and detection threshold configuration output by this method can be fed back to the upstream crack detection method.

[0306] Figure 5 A block diagram of an electronic device suitable for implementing a building crack risk management method according to an embodiment of this application is shown. Figure 5 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0307] like Figure 5 As shown, an electronic device 500 according to an embodiment of this application includes a processor 501, which can perform various appropriate actions and processes according to a program stored in ROM 502 (Read-Only Memory) or a program loaded from storage portion 508 into RAM 503 (Random Access Memory). The processor 501 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 501 may also include onboard memory for caching purposes. The processor 501 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of this application.

[0308] RAM 503 stores various programs and data required for the operation of electronic device 500. Processor 501, ROM 502, and RAM 503 are interconnected via bus 504. Processor 501 executes various operations of the method flow according to embodiments of this application by executing programs in ROM 502 and / or RAM 503. It should be noted that programs may also be stored in one or more memories other than ROM 502 and RAM 503. Processor 501 may also execute various operations of the method flow according to embodiments of this application by executing programs stored in one or more memories.

[0309] According to embodiments of this application, the electronic device 500 may further include an input / output (I / O) interface 505, which is also connected to a bus 504. The electronic device 500 may also include one or more of the following components connected to the input / output (I / O) interface 505: an input section 506 including a keyboard, mouse, etc.; an output section 507 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN card, modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to the input / output (I / O) interface 505 as needed. A removable medium 511, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 510 as needed so that computer programs read from it can be installed into the storage section 508 as needed.

[0310] This application also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of this application.

[0311] According to embodiments of this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium, such as including but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this application, the computer-readable storage medium may include ROM 502 and / or RAM 503 and / or one or more memories other than ROM 502 and RAM 503 described above.

[0312] Embodiments of this application also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code enables the computer system to implement the building crack risk management method provided in the embodiments of this application.

[0313] When the computer program is executed by the processor 501, it performs the functions defined in the system / apparatus of this application embodiment. According to the embodiments of this application, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0314] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 509, and / or installed from a removable medium 511. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0315] In such an embodiment, the computer program can be downloaded and installed from a network via communication section 509, and / or installed from removable medium 511. When the computer program is executed by processor 501, it performs the functions defined in the system of this application embodiment. According to embodiments of this application, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0316] According to embodiments of this application, program code for executing the computer programs provided in the embodiments of this application can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0317] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0318] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application.

[0319] The embodiments of this application have been described above. However, these embodiments are merely illustrative and not intended to limit the scope of this application. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Without departing from the scope of this application, those skilled in the art can make various substitutions and modifications, all of which should fall within the scope of this application.

Claims

1. A method for risk management of building cracks, characterized in that, The method includes: The system acquires first state detection data, first state prediction data, first risk control strategy, and historical state detection data of the target building at multiple first moments in the first time period. The first risk control strategy is used to maintain the target building in the first time period, and the first state prediction data represents the predicted crack state for the crack at multiple first moments in the first time period. Based on the historical state detection data and the first state detection data, the state of the crack is assessed to obtain crack change trend parameters and second state prediction data of the crack in a second time period, which is later than the first time period; Analyze the deviation between the first state detection data and the first state prediction data to determine the first deviation type; Based on the second state prediction data, crack activity, and the first deviation type, the first detection strategy in the first risk control strategy is adjusted to obtain the second detection strategy, wherein the crack activity is determined based on the crack change trend parameter and a preset threshold. The historical state detection data, the first state detection data, the crack change trend parameters, and the first deviation type are processed using a risk identification model to obtain crack risk information. Based on the crack risk information and the second detection strategy, a second risk management strategy is obtained for the second time period.

2. The method according to claim 1, characterized in that, The step of adjusting the first detection strategy in the first risk management strategy based on the second state prediction data, crack activity, and the first deviation type to obtain a second detection strategy includes: The activity level of the crack is determined based on the crack change trend parameters and the preset threshold. Based on the first state detection data and the second state prediction data, a priority is obtained, wherein the priority represents the urgency of detecting multiple cracks; Based on the priority and the crack activity, the first detection strategy is initially adjusted to obtain an initial detection strategy. Based on the first state detection data and the first deviation type, the initial detection strategy is adjusted to obtain the second detection strategy.

3. The method according to claim 2, characterized in that, The first detection strategy includes estimating the search area, and the first state detection data includes the crack extension location; The step of adjusting the initial detection strategy based on the first state detection data and the first deviation type to obtain the second detection strategy includes: When the crack extension location is within the estimated search area and the first deviation type is a trend-consistent type, the initial detection strategy is determined as the second detection strategy, and the first state detection data is added to the historical state detection data for the second time period. If the crack extension location is outside the estimated search area or the first deviation type is a trend acceleration type, adjust the expansion ratio of the estimated search area in the prediction model or the initial detection strategy. If the first deviation type is a trend anomaly type, the first state detection data is cleared or reconstructed using the first state detection data as the initial point.

4. The method according to claim 2, characterized in that, The step of obtaining the priority based on the first state detection data and the second state prediction data includes: Determine the intensity of historical environmental driving forces based on historical environmental data; Structural properties are extracted from the first state detection data to obtain the geometric extension strength; The priority of cracks is determined based on historical crack risk levels, second state prediction data, historical deviations, historical environmental driving strength, geometric propagation strength, and estimated detection costs.

5. The method according to claim 2, characterized in that, The first detection strategy includes at least one of the following: search area, endpoint focus area, bifurcation focus area, local detection threshold, manual re-inspection mark, standard detection threshold, and re-inspection cycle; the second state prediction data includes the predicted extension direction, predicted extension length, and position fluctuation range of the endpoint. The step of initially adjusting the first detection strategy based on the priority and the crack activity to obtain an initial detection strategy includes: When the crack activity is in a stable state, the search area in the first detection strategy is reduced or the re-inspection cycle is extended to obtain the initial detection strategy. When the crack activity is in a state of uniform expansion, the search area in the first detection strategy is expanded according to the estimated extension direction, estimated extension length and position fluctuation range to obtain the initial detection strategy. When the crack activity is in an accelerated propagation state and the environmental driving intensity is greater than the preset driving threshold, the search area in the first detection strategy is expanded according to the estimated extension length and the preset acceleration safety factor, and the local detection threshold is adjusted in the endpoint interest area and the bifurcation interest area to obtain the initial detection strategy. When the crack activity is in a decelerated propagation state, the preset benchmark detection strategy is determined as the initial detection strategy; When the crack activity is in an abnormally fluctuating state, the search area in the first detection strategy is expanded, the standard detection threshold is increased, and manual re-inspection marks are added to obtain the initial detection strategy.

6. The method according to claim 1, characterized in that, Both the historical state detection data and the first state detection data include geometric type data, count type data, and state type data. The step of assessing the state of the crack based on the historical state detection data and the first state detection data to obtain crack change trend parameters and second state prediction data of the crack in the second time period includes: The derivative of the geometric prediction model is used to analyze geometric data to obtain the crack structure change trend, which includes at least one of the following: change rate, acceleration, fluctuation coefficient, and trend fit degree. The probability of new bifurcation is obtained by processing count-type data using a count prediction model. The state type data is processed using a state prediction model to obtain the crack breakthrough state transition probability. The crack change trend parameter includes at least one of the crack structure change trend, the probability of new bifurcation, and the crack breakthrough state transition probability. Based on the geometric data, the state of the crack is estimated using the geometric prediction model to obtain the second state prediction data. The second state prediction data includes at least one of the following: crack geometric prediction location, probability of change in the number of bifurcations, bifurcation candidate region, trend of change in continuity state, and range of location fluctuation.

7. The method according to claim 6, characterized in that, The geometric prediction model is determined based on the following operations: Multiple candidate models were used to fit the sample geometric data of the cracks to obtain multiple sample state prediction data. The fitting residuals are obtained based on the sample geometry data and the sample state prediction data; The candidate model corresponding to the smallest fitting residual among the multiple fitting residuals is determined as the geometric prediction model.

8. The method according to claim 1, characterized in that, The process of using a risk identification model to process the historical state detection data, the first state detection data, the crack change trend parameters, and the first deviation type to obtain crack risk information includes: The historical state detection data, the first state detection data, and the first environmental data of the first time period are fused to obtain fused features; The risk identification model is used to process the fusion features, the crack activity, and the deviation to obtain crack risk information.

9. The method according to claim 8, characterized in that, The feature fusion of the historical state detection data, the first state detection data, and the first environmental data of the first time period to obtain fused features includes: Feature extraction is performed on the historical state detection data and the first state detection data to obtain state features; Feature extraction is performed on the first environmental data to obtain environmental features; The coupling strength is obtained by coupling the state features and the environment features. The initial features are obtained by processing the coupling strength, the state characteristics, and the environmental characteristics using a nonlinear mapping function; The initial features are modulated based on the material-sensitive properties of the crack to obtain the fused features.

10. An electronic device, comprising: One or more processors; Memory, used to store one or more computer programs. The characteristic feature is that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 9.

Images