Search

KR-102963458-B1 - AI-based crack measurement and management system for construction site quality control

KR102963458B1KR 102963458 B1KR102963458 B1KR 102963458B1KR-102963458-B1

Abstract

The present invention relates to an artificial intelligence-based structural safety diagnosis system, and more specifically, to an artificial intelligence-based crack measurement and management system for construction site quality control that automatically detects cracks occurring on the surface of structures such as bridges, dams, and buildings at construction sites or in operation, goes beyond simply measuring the geometric shape of the cracks to comprehensively analyze complex physical factors applied to the structure and trends of change over time to precisely determine the potential risk of the cracks, and dynamically optimizes the monitoring cycle according to the determined risk level and the reliability of the data to efficiently manage the safety of the structures.

Inventors

  • 배인호
  • 최정현

Dates

Publication Date
20260508
Application Date
20251021

Claims (6)

  1. A shooting unit for capturing images of cracks in a structure; An image analysis module that determines the risk of cracking based on images received from the above-mentioned shooting unit and multiple sensor data; The image analysis module that dynamically controls the re-shooting cycle of the shooting unit according to the above-determined risk level; and A data management module that stores all analysis and judgment results; Includes, The above image analysis module is, In an AI-based crack measurement and management system for construction site quality control, characterized by determining the risk level through a hierarchical judgment logic that evaluates in stages multiple key risk factors representing the current state of a crack and multiple acceleration risk factors that accelerate the growth of the crack, The above-mentioned shooting unit is, High-resolution image sensor; Acceleration sensor for measuring vibrations of a structure; Temperature sensors that measure surface and ambient temperatures; and An inertial measurement unit (IMU) that measures the physical position and attitude changes of the above-mentioned imaging unit; It is a composite sensor unit including, The above image analysis module is, The present invention is characterized by receiving a data packet from the above-described shooting unit that includes image data of the image sensor, vibration data of the acceleration sensor, and temperature data of the temperature sensor, and The above image analysis module is, It includes a data validation unit that validates the validity of the image data received from the above-mentioned shooting unit, and The above data validation unit is, (a) a position alignment filtering step for determining whether the camera displacement coefficient measured by the inertial measuring device exceeds a preset position error reference value; and (b) an image quality filtering step for determining, for image data that has passed step (a) above, whether the image stability ratio, which represents the ratio of the shake correction area of the image, is less than a preset stability threshold value, or whether the boundary integrity ratio, which represents the sharpness of the crack boundary line, is less than a preset sharpness threshold value; Performing sequentially, The method is characterized by determining image data that does not meet the criteria in step (a) or (b) as 'unanalyzable data' and discarding it, thereby fundamentally preventing misanalysis that may occur in subsequent analysis steps. The comprehensive risk rating determination unit of the above image analysis module is, The above-mentioned key risk factors include a time-based growth rate calculated through comparison with past images stored in the data management module, a vibration exposure amount measured by the acceleration sensor, and a ground displacement coefficient received from an external satellite navigation system (GNSS) unit; The above acceleration risk factor is characterized by including a thermal hysteresis coefficient measured from the temperature sensor and a wet saturation index received from an external humidity sensor, and The above comprehensive risk rating division is, As a first step, the basic risk level is determined as 'Caution' by determining whether any of the time-based growth rate, the vibration exposure amount, and the ground displacement coefficient exceed a preset 'Caution' level threshold value, and at this time, if multiple key risk factors simultaneously satisfy the criteria of different levels, a highest level priority rule is applied to determine the highest risk level as the basic risk level; As a second step, only if the basic risk level was determined to be 'caution' in the first step, it is determined whether the thermal history coefficient or the wet saturation index exceeds a preset 'risk level upward adjustment' threshold value; The method is characterized by raising the final overall risk level to 'risk' when the result of the judgment in the second stage above exceeds the 'risk level upward adjustment' threshold value, thereby determining the highest risk level only when environmental factors that accelerate growth are combined with a state where crack activity is confirmed, thereby preventing excessive alarms in the system and improving the reliability of the alarm. The above data management module is, In response to a request from the above image analysis module, time series data corresponding to specific conditions is retrieved and provided, The above image analysis module is, The system is characterized by optimizing the system so that an alert occurs at an earlier time in the future by self-strengthening the judgment criteria based on actual risk data, wherein, after the comprehensive risk level judgment unit determines a 'Caution' level, if cases of being upgraded to a 'Risk' level within an actual preset period (e.g., 30 days) accumulate in the data management module at a preset number of times (e.g., 5 times), it is determined that the current 'Caution' level judgment threshold is excessively lenient, and the 'Caution' level judgment threshold is automatically lowered by a preset ratio (e.g., 5%) relative to the previous value. The image analysis module (200) further includes a growth trend prediction unit that organically combines growth history from the past to the present and future expected environmental information to classify future crack growth risk trends into clear grades such as 'stable state', 'growth maintenance', and 'growth acceleration'. The above growth trend prediction unit receives current time-based growth rate, crack growth acceleration, expected vibration exposure amount, and expected thermal history coefficient as judgment criterion data for determining future growth risk trends, and The above growth trend prediction unit adopts the logic of 'correcting for external factors after setting a basic trend,' and (1) Based on the current time-based growth rate and the crack growth acceleration, the future basic growth trend is initially determined to be 'stable state' or 'growth maintenance', and (2) Evaluate the impact of the above-mentioned expected vibration exposure amount and the above-mentioned expected thermal history coefficient on the prediction of the above-mentioned basic trend and determine whether to finally upgrade the trend grade to 'growth acceleration', and The above estimated vibration exposure is estimated by linking with the 'weekly construction plan' data entered into the system by the site manager or with the traffic volume statistics database, and The system incorporates a database mapping 'construction equipment type-expected vibration level', collects vibration data measured by the acceleration sensor of the above-mentioned shooting unit (100), and continuously learns and corrects the database. If the determination result of the growth trend prediction unit is 'growth acceleration', the system sets the priority to 'High' in the maintenance plan and sends a 'predictive alert' including specific periods and measures, and The image analysis module (200) includes, as sub-function units, a data validation unit that validates the validity of received data, a comprehensive risk level determination unit that determines the risk level of a crack by synthesizing multiple data, a growth trend prediction unit that predicts the growth trend of a crack based on past and future information, and a dynamic re-shooting cycle control unit that determines the optimal re-shooting cycle by synthesizing all determination and prediction results. The dynamic re-imaging cycle control unit receives a plurality of judgment criterion data, including the overall risk level and the crack growth acceleration or the growth trend, and determines the next re-imaging cycle. The above dynamic re-shooting cycle control unit adopts a 'priority-based hierarchical decision-making' logic and adopts a hierarchical structure that assigns priorities in the order of 'real-time sudden events, system status, current risk level, and future risk trend', and In the top priority judgment stage, acoustic emission frequency is verified, in the second priority judgment stage, data validation results are verified, and in the third priority judgment stage, the overall risk level and crack growth acceleration are comprehensively verified, The above dynamic re-shooting cycle control unit sets different cycle modes according to the determination result, and Emergency cycle mode sets the re-shoot cycle to the minimum time configurable by the system, shortened cycle mode sets the re-shoot cycle to a time shorter than the standard, standard cycle mode maintains the routine monitoring cycle, and extended cycle mode controls to minimize energy consumption and data storage load by extending the re-shoot cycle to the maximum time. The above dynamic re-shooting cycle control unit, when the acoustic emission frequency exceeds the threshold for triggering the emergency cycle mode, ignores all other conditions and immediately sets the next re-shooting cycle to the emergency cycle mode, and If the result of the above data validation is 'unable to analyze', the dynamic re-shooting cycle control unit recognizes this as a second-priority condition and performs a re-shooting request by immediately changing the re-shooting cycle to a shortened cycle mode. The system executes a 'safety check' logic once every 7 days to check the overall status of the system by forcibly performing shooting in standard cycle mode regardless of the current cycle mode, and AI-based crack measurement and management system for construction site quality control, which considers the 'growth acceleration' prediction of the growth trend prediction unit as a condition for activating the shortened cycle mode and transmits a command to the shooting unit (100) to change the next re-shooting cycle to the shortened cycle mode.
  2. delete
  3. delete
  4. delete
  5. delete
  6. delete

Description

AI-based crack measurement and management system for construction site quality control The present invention relates to an artificial intelligence-based structural safety diagnosis system, and more specifically, to an artificial intelligence-based crack measurement and management system for construction site quality control that automatically detects cracks occurring on the surface of structures such as bridges, dams, and buildings at construction sites or in operation, goes beyond simply measuring the geometric shape of the cracks to comprehensively analyze complex physical factors applied to the structure and trends of change over time to precisely determine the potential risk of the cracks, and dynamically optimizes the monitoring cycle according to the determined risk level and the reliability of the data to efficiently manage the safety of the structures. Recently, as the aging of social infrastructure accelerates and external environmental uncertainties such as earthquakes and climate change increase, the importance of maintenance technologies to ensure structural safety in advance is being highlighted. In particular, technology for detecting and managing cracks—the most critical indicator for assessing the deterioration of durability in concrete structures—occupies a core part of structural safety diagnosis. Conventional crack management technology has primarily relied on manual methods in which managers visit the site in person and measure the width and length of cracks using crack gauges or microscopes. This method has problems such as difficulty in ensuring data consistency and objectivity due to the high possibility of subjective bias on the part of the measurer, inability to conduct constant inspections of high-altitude or hazardous areas that are difficult to access, and inefficiency due to the large amount of manpower and time required. To address these issues, technologies have recently been proposed that automatically detect cracks and measure their size by analyzing images captured by cameras mounted on drones or robots using artificial intelligence deep learning. However, these image-based conventional technologies focus solely on statically measuring the current 'state' of the cracks, and thus have clear limitations in evaluating the dynamic risk regarding why the cracks occurred and how they might change in the future. For example, even though the risk level differs significantly between a crack in a bridge deck exposed to continuous vehicle vibration and a crack in the interior wall of a building with almost no external load, conventional technologies fail to distinguish between the two, even if both have the same width of 0.2 mm. Furthermore, simple monitoring methods that involve shooting at fixed time intervals or shortening the shooting cycle when deemed dangerous caused unnecessary data accumulation and energy waste during stable conditions, while potentially missing the critical moment when rapid changes occur just before destruction. Figure 1 illustrates an overall relationship diagram according to the present invention. Figure 2 illustrates a flowchart between all components according to the present invention. Figure 3 illustrates a flowchart of Process 1 according to the present invention. Figure 4 illustrates a flowchart of process 2 according to the present invention. Figure 5 illustrates a flowchart of process 3 according to the present invention. Figure 6 illustrates a flowchart of process 4 according to the present invention. Hereinafter, various embodiments are described in more detail with reference to the attached drawings. The embodiments described in this specification may be modified in various ways. Specific embodiments may be depicted in the drawings and described in detail in the detailed description. However, specific embodiments disclosed in the attached drawings are intended only to facilitate understanding of various embodiments. Accordingly, the technical concept is not limited by specific embodiments disclosed in the attached drawings, and it should be understood that it includes all equivalents or substitutions that fall within the spirit and scope of the invention. Terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but these components are not limited by the aforementioned terms. The aforementioned terms are used solely for the purpose of distinguishing one component from another. Functions related to artificial intelligence according to the present disclosure are operated through a processor and memory. The processor may be composed of one or more processors. In this case, the one or more processors may be general-purpose processors such as CPUs, APs, and DSPs (Digital Signal Processors), graphics-dedicated processors such as GPUs and VPUs (Vision Processing Units), or artificial intelligence-dedicated processors such as NPUs. The one or more processors control the processing of input data according t