Automatic evaluation system of realtime bridge safety using digital twin model, and method for the same

Abstract

A real-time bridge safety automatic evaluation system and method using a digital twin model are provided, which can shorten automated disaster response time, reduce maintenance costs, and minimize the burden on managers by moving away from existing manual analysis-based structural diagnosis methods and realizing the automation of real-time structural condition recognition and evaluation. Furthermore, by constructing an effective digital twin through response-based analysis without load conditions and precisely analyzing the structural condition solely through output responses, it offers high consistency with actual structural behavior and prevents loss of physical information. Additionally, by utilizing multi-point measurement-based high-resolution response profiles to organize rotation angle and displacement data into a continuous distribution profile, it enables high-resolution detection of asymmetric damage, local anomalies, and boundary deterioration.

Description

Technology Field

[0001] The present invention relates to a real-time automatic bridge safety evaluation system using a digital twin model, and more specifically, to a real-time automatic bridge safety evaluation system and method using a digital twin model that fuses multi-point responses (rotation angle and displacement) measured by digital sensors such as inclinometers with displacement-based structural analysis simulations at an edge terminal, calculates member forces based on responses under unknown loads, and detects and corrects anomalies using artificial intelligence (AI). Background Technology

[0002] Large structures and facilities constructed during the development of an industrial society suffer structural damage due to defects in the design and construction processes or various factors not considered at the time of design. Furthermore, as these structures gradually deteriorate over their service life, their safety is significantly threatened. For example, in the case of structures with severe structural damage, it frequently results in a shortened service life that falls far short of the planned lifespan at the time of design.

[0003] Accordingly, there is an urgent need for efforts to ensure the long-term safety and operability of architectural structures. In particular, since large structures such as buildings, bridges, and dams are continuously exposed to various operational loads, impacts from external objects, earthquakes, wind loads, wave loads, and corrosion, ensuring the safety of these structures has become a pressing issue of significant economic and social concern. For accurate safety diagnosis of these large structures, diagnostic technologies are required that utilize monitoring of structural behavior through appropriate experimental measurements, techniques for mechanically analyzing structural damage, and analysis techniques for modeling structural damage.

[0004] In Korea, social infrastructure and various industrial facilities were actively constructed starting from the 1970s and 1980s, a period of economic growth. However, the facilities built at that time have now exceeded 30 years of service life and are rapidly deteriorating.

[0005] Maintenance and safety management, including inspections and repairs / reinforcements for the recently increasing number of aging facilities, have emerged as significant social issues. In response, monitoring systems for major bridge facilities are currently being established and operated. However, the currently installed bridge monitoring systems are being applied restrictively, limited only to monitoring events that exceed management standards.

[0006] The safety of such bridges can be determined based on whether the generated member forces are within the allowable range; however, existing bridge monitoring systems have limitations in evaluating the overall safety of bridges because they install sensors only at specific locations and monitor key engineering physical quantities such as deflection and stress.

[0007] Furthermore, in order to evaluate the safety of a bridge based on digital sensing data, which is real-time measurement data of a bridge in service, it is necessary to analyze the behavior of members without sensors as well as those with sensors installed; the most reasonable and efficient method to address this is to evaluate the safety of the bridge through structural analysis that considers the measurement data.

[0008] However, installing sensors on every component is highly inefficient in terms of cost and operation. Furthermore, since existing bridge monitoring systems rely on data from specific locations, there is a problem in that it is difficult to automatically assess bridge safety through structural analysis based on measurement data.

[0009] Meanwhile, as prior art related to real-time monitoring of bridge displacement, Korean published patent number 2024-0056067 discloses an invention titled "Dual computational processing system applied to a real-time monitoring system of bridge displacement," which will be explained with reference to FIGS. 1a and 1b.

[0010] FIG. 1a is a conceptual diagram showing a bridge with a bridge displacement measuring device installed according to conventional technology, and FIG. 1b is a configuration diagram showing a dual computational processing system applied to the real-time bridge displacement monitoring system shown in FIG. 1a.

[0011] Referring to FIG. 1a, a bridge equipped with a bridge displacement measuring device according to the prior art may include a first bridge deck (11), a second bridge deck (12), a first bridge pier (13), a second bridge pier (14), and a bridge bearing (15).

[0012] The first bridge deck (11) and the second bridge deck (12) are arranged continuously along one direction, as shown in FIG. 1a. The first and second bridge decks (11, 12) may be made of reinforced concrete or steel. And an asphalt layer is provided on the upper surface of the first and second bridge decks (11, 12) so that people or vehicles can pass through.

[0013] The first pier (13) is formed so that each of the opposing ends of the first and second bridge decks (11, 12) is supported, thereby supporting the first and second bridge decks (11, 12).

[0014] The second pier (14) is formed to support the second bridge deck (12) and supports the second bridge deck (12) located on the upper side.

[0015] A bridge displacement measuring device (20) is installed on the lower part of the first and second bridge decks (11, 12) to measure the relative displacement of the first and second bridge decks (11, 12) or the displacement of each of the first and second bridge decks (11, 12). The bridge displacement measuring device (20) may be installed in the longitudinal direction of the first and second bridge decks (11, 12), as shown in the configuration arranged on the left side of FIG. 1a, or installed in the width direction of the first and second bridge decks (11, 12), as shown in the configuration arranged on the right side of FIG. 1a.

[0016] Additionally, the bridge may be equipped with a bridge bearing (15) that absorbs impact applied to the first and second bridge decks (11, 12) or transmits it to the first and second piers (13, 14) to perform a cushioning function. The bridge bearing (15) performs the role of preventing the displacement of the first and second bridge decks (11, 12) from being directly transmitted to the first and second piers (13, 14) or abutments.

[0017] Referring to FIG. 1b, the dual computational processing system applied to a bridge displacement real-time monitoring system according to the prior art is configured to include a plurality of sensors (31), a field computational device (32), a communication device (33), and a server computational device (34).

[0018] Multiple sensors (31) are positioned between the first support frame and the second support frame to connect the first support frame and the second support frame, measure linear displacement between points connected on the first support frame and the second support frame, and generate different sensing data.

[0019] The field computing device (32) is physically connected to a plurality of sensors (31) and generates first computing data by calculating the relative displacement of the first bridge deck and the second bridge deck that are spaced apart from each other using the sensing data.

[0020] The communication device (33) transmits at least one of the sensing data and the first computation data to the server computation device (34).

[0021] The server computing device (34) is physically separated from the plurality of sensors (31) and generates second computing data by calculating the relative displacement of the first bridge deck and the second bridge deck using the sensing data. At this time, the server computing device (34) can execute different functions depending on whether the first computing data and the second computing data match within a reference range.

[0022] According to the dual computational processing system applied to the real-time bridge displacement monitoring system based on conventional technology, dynamic targets are processed multiple times to accurately detect the relative displacement of the bridge and enable real-time response according to weather conditions.

[0023] In addition, multiple field computing devices operate via the Internet of Things (IoT) and can be effectively monitored and managed through a server computing device. The server computing device can monitor multiple field computing devices installed at different locations at each site in real time and manage them by applying customized standards to respond to variable weather conditions.

[0024] Meanwhile, as prior art related to bridge safety monitoring, Korean published patent number 2024-0103862 discloses an invention titled "Bridge Safety Monitoring System," which will be explained with reference to FIGS. 2a and 2b.

[0025] FIG. 2a is a drawing showing a bridge safety monitoring system according to conventional technology, and FIG. 2b is a detailed configuration diagram of a smart sensor module in the bridge safety monitoring system shown in FIG. 2a.

[0026] A bridge safety monitoring system according to the prior art may include a smart sensor module (41), a bridge data collection server (42), and a monitoring terminal device (43), as illustrated in FIGS. 2a and 2b.

[0027] The smart sensor module (41) is connected to static sensors (51) or dynamic sensors (52) installed at each location of the bridge, and after collecting the sensing values ​​detected and transmitted by the static sensors (51) or dynamic sensors (52), it calculates a result value that can determine the state of the bridge.

[0028] The sensors connected to the smart sensor module (41) are divided into static sensors (51) and dynamic sensors (52), as shown in FIG. 2b. The static sensors (51) include thermometers and displacement meters, and the dynamic sensors (52) may include stress meters, displacement meters, cable tension meters, accelerometers, wind direction and speed meters, and seismic accelerometers, and each may be installed on a bridge.

[0029] The bridge data collection server (42) is connected to a plurality of smart sensor modules (41) and collects the bridge deflection state values ​​calculated by the smart sensor modules (41).

[0030] The monitoring terminal device (43) connects to the bridge data collection server (42) to receive and display the bridge deflection status value so that an administrator can monitor it.

[0031] Specifically, the actual connection status between the static sensor (51), the dynamic sensor (52), the smart sensor module (41), the bridge data collection server (42), and the monitoring terminal device (43) is as shown in FIG. 2B. In addition, other sensors capable of monitoring bridge safety may be installed on the bridge.

[0032] The smart sensor module (41) of the bridge safety monitoring system may include an accelerometer (41a), a strain meter (41b), a microprocessor unit (MPU: 41c), a memory unit (41d), an AD converter (ADC: 41e), a wired communication module (41f), a power supply unit (41g), a multimedia card (41h), and a wireless communication module (41i), as shown in FIG. 2b.

[0033] Bridge measurement by the smart sensor module (41) can increase the efficiency of diagnosis by enabling quantitative evaluation of periodic diagnosis using objective and continuous data regarding the deterioration, damage, and aging of the bridge structure over a long period. In addition, the entire structure, including parts that are inaccessible or invisible, can be continuously and uninterruptedly monitored to provide an early warning of bridge safety.

[0034] The accelerometer (41a) is installed on the main tower of the bridge and on the plate girder, which is the deck installed under the support of the main tower, and acquires raw acceleration data at a set period for a set time.

[0035] A strain gauge (41b) is attached to the plate girder to measure physical strain and obtain strain gauge data.

[0036] The microprocessor unit (41c) calculates the bridge deflection state value using raw acceleration data and strain gauge data.

[0037] The memory unit (41d) stores other information about the train, such as the bridge deflection state value calculated by the microprocessor unit (41c), the location where the accelerometer (41a) is installed, and the sensing time.

[0038] The AD converter (41e) receives the sensing value of the sensor that senses data in an analog manner among the static sensor (51) and the dynamic sensor (52), and converts the received analog sensing value into digital.

[0039] The wired communication module (41f) is connected to a static sensor or a dynamic sensor via various cables including USB or HDMI, so that it can communicate via RS485 or RS232 communication methods, and receives the sensed values.

[0040] When the wireless communication module (41i) calculates a result value that can evaluate the safety of the bridge using parameters detected by the microprocessor unit (41c) from the static sensor (51) or the dynamic sensor (52), it transmits only that result value to the bridge data collection server (42).

[0041] According to a bridge safety monitoring system based on conventional technology, only the result value calculated by the smart sensor module is transmitted to the bridge data collection server, and since the transmission of large-capacity basic data required to calculate the result value is omitted, the data transmission capacity can be minimized.

[0042] In addition, if an error occurs in the central monitoring server, the result value is lost, but since the basic data for calculating the result value is stored in the field terminal, the safety of the bridge can be monitored again simply by requesting the result value from the field terminal after the error is resolved.

[0043] Meanwhile, as prior art related to artificial intelligence (AI)-based bridge load-carrying performance evaluation, Korean published patent number 2025-0103551 discloses an invention titled "Artificial Intelligence (AI)-based Bridge Load-Carrying Performance Evaluation System and Load-Carrying Performance Evaluation Method," which will be explained with reference to FIGS. 3a to 3c.

[0044] FIG. 3a is a configuration diagram of an artificial intelligence-based bridge load-bearing performance evaluation system according to conventional technology, FIG. 3b is a diagram illustrating an actual target facility and an analysis model, and FIG. 3c is a flowchart of an operation of an artificial intelligence-based bridge load-bearing performance evaluation method.

[0045] Referring to FIG. 3a, an artificial intelligence (AI)-based bridge load-bearing performance evaluation system according to the prior art may include a digital twin system (40), a target facility (61) in which a sensing unit (61a) which is an IoT sensor is installed, an analysis model (62), a model tuning unit (63), a load-bearing capacity evaluation unit (64), and an artificial intelligence model (65).

[0046] A sensing unit (61a) implemented with an IoT sensor is installed on a bridge, which is a target facility (61) for measuring and monitoring the condition of the bridge.

[0047] The analysis model (62) simulates the behavioral characteristics of the bridge, which is the target facility (61).

[0048] The model tuning unit (63) analyzes the data measured through the sensing unit and the analysis model, and updates the analysis model.

[0049] At this time, the model tuning unit (63) periodically compares and analyzes the frequency response of the measurement data and the analysis model (62), periodically updates the analysis model (62) by reflecting changes in the bridge, performs an evaluation of the bridge's load-carrying capacity by applying a virtual load, and can predict the lifespan of the bridge using a lifespan prediction algorithm, which is an artificial intelligence model, based on the load-carrying capacity obtained through repeated virtual load tests.

[0050] Here, the target facility (61) includes a building, a bridge, a retaining wall, or a temporary facility, and the sensing unit (61a) may include at least one of a displacement sensor, an accelerometer, and an inclinometer. FIG. 3b shows an example of an actual target facility (61) and an analysis model (62), where the upper figure is a bridge as the target facility (61) and the lower figure is an analysis model (62).

[0051] In addition, referring to FIG. 3c, the artificial intelligence-based bridge load-bearing performance evaluation method according to the prior art evaluates the load-bearing performance of a bridge using the artificial intelligence-based bridge load-bearing performance evaluation system illustrated in FIG. 3a.

[0052] First, apply the optimization algorithm (S11) and verify reliability (S12).

[0053] Specifically, an analysis model (62) implemented based on the geometry of the bridge is constructed. Then, the response of the target facility (61) is obtained through an IoT sensor (61a) installed on the bridge. Then, the frequency response is analyzed using the analysis model (62) and the data measured through the IoT sensor (61a). Then, the frequency response of the analysis model (62) and the measured data is compared and analyzed. Then, the analysis model (62) is updated to match the characteristics of the bridge.

[0054] Next, a virtual load is applied to the updated analysis model (62) to extract information for evaluating the load-carrying capacity of the bridge according to the applied load and to evaluate the load-carrying capacity (S14).

[0055] Next, the lifespan of the bridge is predicted using a lifespan prediction algorithm, which is an artificial intelligence model (65), based on the load-carrying capacity obtained through repeated virtual load tests (S13) (S15).

[0056] According to the AI-based bridge load-carrying performance evaluation system and method based on conventional technology, the load-carrying capacity of a bridge can be evaluated by conducting load tests without time constraints, the lifespan and durability of the bridge can be predicted more accurately, and the safety and lifespan of the bridge can be predicted by combining advanced data analysis and AI technology.

[0057] Meanwhile, as mentioned above, there is a problem in that it is difficult to automatically assess bridge safety through structural analysis based on measurement data, as the existing bridge monitoring system is configured using data from specific locations. Prior art literature

[0058] Korean Registered Patent No. 10-2824335 (Registration Date: June 19, 2025), Title of Invention: "Maintenance Management System for Bridge Structures Using an AI-Based Deterioration Model and Method Thereof" Korean Registered Patent No. 10-1431237 (Registration Date: August 11, 2014), Title of Invention: "System and Method for Detecting Abnormal Behavior of Structures and Automatically Evaluating Safety" Korean Published Patent No. 2024-0056067 (Publication Date: April 30, 2024), Title of Invention: "Dual Computation Processing System Applied to a Real-time Bridge Displacement Monitoring System" Korean Published Patent No. 2024-0103862 (Publication Date: July 4, 2024), Title of Invention: "Bridge Safety Monitoring System" Korean Published Patent No. Korean Patent Publication No. 2025-0103550 (Publication Date: July 7, 2025), Title of Invention: "Digital Twin System for Evaluating Load-carrying Capacity of Bridges and Load Test Method Using the Same" Korean Patent Publication No. 2025-0103551 (Publication Date: July 7, 2025), Title of Invention: "Artificial Intelligence-based Bridge Load-carrying Performance Evaluation System and Load-carrying Performance Evaluation Method" The problem to be solved

[0059] The technical objective of the present invention, which aims to solve the aforementioned problems, is to provide a real-time bridge safety automatic evaluation system and method using a digital twin model capable of automatically evaluating real-time bridge safety based on artificial intelligence (AI) by realizing the automation of real-time structural state recognition and evaluation using a digital twin model that fuses data obtained from actual measurements of bridge responses with structural analysis simulations.

[0060] Another technical objective of the present invention is to provide a real-time automatic bridge safety evaluation system and method using a digital twin model capable of constructing an effective digital twin through response-based analysis without load conditions and precisely analyzing the structural state solely through output responses.

[0061] Another technical objective of the present invention is to provide a real-time automatic bridge safety assessment system and method using a digital twin model, which can detect asymmetric damage, local anomalies, and boundary deterioration at high resolution by utilizing a multi-point measurement-based high-resolution response profile to organize rotation angle and displacement data into a continuous distribution profile. means of solving the problem

[0062] As a means to achieve the aforementioned technical challenge, a real-time automatic bridge safety evaluation system using a digital twin model according to the present invention comprises: a digital sensor installed at a predetermined location of a bridge, which is a structure to be evaluated, for measuring the response of the bridge in real time; a safety evaluation terminal that calculates a rotation angle and a displacement profile according to rotation angle data measured in real time by the digital sensor to identify the structural state, detects abnormal behavior of the bridge according to an artificial intelligence-based anomaly detection algorithm, and, upon detection of abnormal behavior of the bridge, performs a displacement-based structural analysis simulation to automatically evaluate the structural safety of the bridge; and a manager terminal that receives the structural safety evaluation results wirelessly from the safety evaluation terminal, wherein the safety evaluation terminal automates the real-time structural state recognition and evaluation according to a digital twin model that fuses the data measured in real time by the digital sensor and the structural analysis simulation, thereby automatically evaluating the real-time bridge safety based on artificial intelligence (AI); and the safety evaluation terminal includes a response data collection unit that collects rotation angle data, which is response data measured in real time through a digital sensor installed at a predetermined part of the bridge. A rotation angle and displacement profile calculation unit that calculates a rotation angle profile and a displacement profile, respectively, according to the collected rotation angle data based on the constructed digital twin model; a structural state identification unit that identifies the structural state based on the calculated rotation angle and displacement profiles; an AI-based abnormal behavior detection unit that detects abnormal behavior of the bridge according to an AI-based abnormal detection algorithm; and a structural analysis unit that performs a displacement-based structural analysis simulation when abnormal behavior of the bridge is detected.It includes a structural safety evaluation unit that automatically evaluates the structural safety of a bridge based on the results of displacement-based structural analysis simulations, wherein the digital sensor is composed of inclinometers, which are a plurality of rotation angle sensors spaced apart from each other in the span direction of the bridge, and the plurality of rotation angle sensors measure longitudinal rotation angle data of the bridge in real time while synchronized with the same sampling period, and the safety evaluation terminal maps the rotation angle data collected from the plurality of inclinometers to the span direction length coordinates of the bridge, performs interpolation for the section between the plurality of inclinometers to generate a rotation angle profile, and generates a displacement profile for the entire span section by numerically integrating the rotation angle profile and correcting the integration constant based on point displacement conditions or auxiliary displacement sensor reference values, and the artificial intelligence-based abnormal behavior detection unit detects signal errors including missing values, drift, outliers, or impact noise included in the rotation angle data, rotation angle profile, or displacement profile, and regenerates the section containing the signal error using normal data from adjacent sensors or the prediction profile of a digital twin model, and the structural analysis unit [describes] external forces or loads acting on the bridge Without directly calculating or estimating the conditions, at least one of the member forces, section forces, or stresses of the bridge is calculated by performing a displacement-based structural analysis simulation using the displacement profile and rotation angle profile as input responses to the structural analysis model.

[0063] delete

[0064] Meanwhile, as another means for achieving the aforementioned technical task, the real-time automatic bridge safety evaluation method using a digital twin model according to the present invention comprises, in the real-time automatic bridge safety evaluation method using a real-time automatic bridge safety evaluation system using the digital twin model, a) a step of measuring the response of the bridge in real time through a digital sensor installed at a predetermined part of the bridge, which is a structure to be evaluated; b) a step in which a safety evaluation terminal collects rotation angle data from the digital sensor; c) a step of calculating a rotation angle profile and a displacement profile, respectively, according to the collected rotation angle data according to the constructed digital twin model; d) a step in which an artificial intelligence-based abnormal behavior detection unit of the safety evaluation terminal detects abnormal behavior of the bridge according to an artificial intelligence-based abnormal detection algorithm; e) a step in which, upon detection of abnormal behavior of the bridge, a structural analysis unit of the safety evaluation terminal performs a displacement-based structural analysis simulation; and f) a step in which a structural safety evaluation unit of the safety evaluation terminal evaluates the structural safety of the bridge according to the result of the displacement-based structural analysis simulation. Effects of the invention

[0065] According to the present invention, by automating real-time structural status recognition and evaluation by moving away from the existing manual analysis-based structural diagnosis method, the time required for automated disaster response can be shortened, maintenance costs can be reduced, and the burden on administrators can be minimized.

[0066] According to the present invention, a digital twin can be constructed using response-based analysis without estimating load conditions, and the structural state can be precisely analyzed using only the output response; thus, there is high consistency with the actual structural behavior and no loss of physical information occurs.

[0067] According to the present invention, by organizing rotation angles and displacements into a continuous distribution type through a multi-point measurement-based high-resolution response profile, spatial non-uniformity such as asymmetric damage, local anomalies, and boundary deterioration can be detected at high resolution.

[0068] According to the present invention, by adopting an in-house analysis engine, lightweight and automated operation is possible on edge terminals, commercial license and maintenance costs are reduced, and convergence failure and delay can be minimized through a fallback (ROM → simple model) function.

[0069] According to the present invention, an artificial intelligence (AI)-based abnormal behavior detection unit can automatically detect and regenerate missing values, NaNs, drifts, impact noises, and outliers using a rule+learning ensemble to reduce false alarms and missed detections, and increase evaluation reliability by providing a refined profile for structural analysis.

[0070] According to the present invention, by reflecting structural mechanics constraints such as the kinematic relationship of θ=∂w / ∂x, boundary conditions, and modal characteristics in the correction and regeneration stage, physically consistent signal restoration is achieved, and as a result, the stability and accuracy of member force and stress calculation are improved.

[0071] According to the present invention, logs such as quality flags, calibration history, thresholds, and model versions are transparently transmitted and managed to an administrator terminal, thereby ensuring traceability and operational reliability, and can be easily expanded through modular profiles, analysis, and AI pipelines even with the expansion of sensor channels or spans. Brief explanation of the drawing

[0072] FIG. 1a is a conceptual diagram showing a bridge with a bridge displacement measuring device installed according to conventional technology, and FIG. 1b is a configuration diagram showing a dual computational processing system applied to the real-time bridge displacement monitoring system shown in FIG. 1a. FIG. 2a is a drawing showing a bridge safety monitoring system according to conventional technology, and FIG. 2b is a detailed configuration diagram of a smart sensor module in the bridge safety monitoring system shown in FIG. 2a. FIG. 3a is a configuration diagram of an artificial intelligence-based bridge load-bearing performance evaluation system according to conventional technology, FIG. 3b is a diagram illustrating an actual target facility and an analysis model, and FIG. 3c is a flowchart of an operation of an artificial intelligence-based bridge load-bearing performance evaluation method. FIG. 4 is a diagram illustrating a real-time deflection measurement view for applying a real-time bridge safety automatic evaluation system using a digital twin model according to an embodiment of the present invention. Figures 5a and 5b are drawings showing real-time deflection measurement results, respectively. Figure 6 is a table specifically showing the real-time deflection measurement results. FIG. 7 is a cross-sectional view showing a bridge to which a real-time automatic bridge safety evaluation system using a digital twin model according to an embodiment of the present invention is applied. FIG. 8 is a schematic diagram of a real-time automatic bridge safety evaluation system using a digital twin model according to an embodiment of the present invention. FIG. 9 is a diagram illustrating the concept of a digital twin model applied to a real-time bridge safety automatic evaluation system using a digital twin model according to an embodiment of the present invention. FIG. 10 is a detailed configuration diagram of a real-time bridge safety automatic evaluation system using a digital twin model according to an embodiment of the present invention. FIG. 11 is a diagram schematically illustrating the real-time bridge safety automatic evaluation process in a real-time bridge safety automatic evaluation system using a digital twin model according to an embodiment of the present invention. FIG. 12 is a diagram showing a rotation angle sensor (inclinometer), which is a digital sensor, placed on a bridge to which a real-time automatic bridge safety evaluation system using a digital twin model according to an embodiment of the present invention is applied. FIG. 13 is a diagram illustrating the types of error signals in a real-time bridge safety automatic evaluation system using a digital twin model according to an embodiment of the present invention. FIGS. 14a and FIGS. 14b are drawings illustrating an example of abnormal signal regeneration for shock noise in a real-time bridge safety automatic evaluation system using a digital twin model according to an embodiment of the present invention. FIGS. 15a and 15b are drawings illustrating an example of abnormal signal regeneration for a NaN value in a real-time bridge safety automatic evaluation system using a digital twin model according to an embodiment of the present invention. FIG. 16 is a diagram illustrating the error signal reconstruction network structure in a real-time bridge safety automatic evaluation system using a digital twin model according to an embodiment of the present invention. FIG. 17 is a flowchart illustrating a method for real-time automatic bridge safety evaluation using a digital twin model according to an embodiment of the present invention. Specific details for implementing the invention

[0073] Embodiments of the present invention are described below with reference to the attached drawings so that those skilled in the art can easily implement the invention. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the present invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification are denoted by similar reference numerals.

[0074] Throughout the specification, when a part is described as "comprising" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. Furthermore, terms such as "…part" as used in the specification refer to a unit that processes at least one function or operation, and this may be implemented in hardware, software, or a combination of hardware and software.

[0075] Hereinafter, with reference to FIGS. 4 to 16, a real-time automatic bridge safety evaluation system using a digital twin model according to an embodiment of the present invention will be described, and with reference to FIG. 17, a real-time automatic bridge safety evaluation method using a digital twin model according to an embodiment of the present invention will be described.

[0076] [Real-time automatic bridge safety evaluation system using digital twin model (100)]

[0077] The real-time automatic bridge safety evaluation system using a digital twin model according to an embodiment of the present invention is intended to utilize a bridge monitoring system for automatic safety evaluation. Unlike conventional monitoring methods that calculate only deflection at each measurement point, it continuously measures the longitudinal rotation angle of the bridge using digital sensors and converts this into an overall deflection profile in real time to calculate the deflection response.

[0078] The deflection response calculated in this way is utilized as an input file for a structural analysis engine tuned with an in-house structural model using measurement data. While external forces such as vehicle loads and wind loads are typically input in structural analysis, the loads acting on the bridge in real time cannot be accurately known; therefore, in this embodiment, the deflection caused by unknown input is used as an input variable, and the structural state is precisely analyzed solely based on the output response.

[0079] Figure 4 shows a real-time deflection measurement view to which the present system is applied.

[0080] In the verification stage, for example, as shown in Fig. 4, LVDTs (Linear Variable Differential Transformers) were fixedly installed at 1 / 4L, 1 / 2L, and 3 / 4L positions along the span, and the sensor tips were vertically contacted to the upper surface of the test member to continuously measure the reference deflection.

[0081] In the central part, stepwise displacement was introduced using a micro-loading device to reproduce the reference deflection of the first, second, and third cases (2.54 mm, 5.22 mm, and 7.99 mm, respectively).

[0082] All LVDTs were zeroed with a calibration block before testing, and the time series was synchronized by setting the acquisition period to be the same as that of the digital inclinometer, and high-frequency noise was removed with a low-pass filter.

[0083] Accordingly, as shown in FIGS. 5a, 5b, and 6, the deflection based on the LVDT and the deflection calculated by rotation angle integration were compared with the time average value and the profile by position, and quantitatively evaluated using relative error and RMSE.

[0084] Figures 5a and 5b illustrate real-time deflection measurement results, respectively, with Figure 5a showing deflection history data and Figure 5b showing deflection profile data. Additionally, Figure 6 is a table summarizing system results relative to LVDT reference values, presenting reference deflection and relative error for each case to verify quantitative accuracy.

[0085] That is, Figures 5a and 5b illustrate real-time deflection results obtained simultaneously during the verification phase.

[0086] FIG. 5a is a time domain deflection history, in which the reference deflection of the LVDT installed at span points 1 / 4L, 1 / 2L, and 3 / 4L and the deflection converted by numerical integration in the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention are compared as a time series with the rotation angle of the inclinometer placed in the same section.

[0087] The target deflection level for each case was reproduced by sequentially applying step loads, and the transient sections occurring during load transitions were preprocessed using a low-pass filter and offset removal. It was confirmed that the average deflection value in the stabilization section was close to the LVDT reference value.

[0088] Figure 5b is a spatial area deflection profile.

[0089] The real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention integrates the continuous rotation angle θ(x) in the span direction to restore the deflection with the relationship w′(x)=θ(x) and w(x)=∫θ(x)dx+constant, and calculates the entire span deflection curve by correcting the integration constant with the support displacement condition (e.g., w(0)=w(L)=0) and the LVDT reference value.

[0090] The central reference deflection of the first, second, and third cases shows an increasing trend, with approximately 2.54 mm, 5.22 mm, and 7.99 mm, respectively.

[0091] FIG. 6 is a table showing a quantitative comparison of results relative to the LVDT reference value. The relative error with respect to the deflection calculated in the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention is as follows.

[0092] Case 1 (Reference 2.54 mm): Analog 2.48 mm, error 3.1%, Digital 2.47 mm, error 3.0%. Case 2 (Reference 5.22 mm): Analog 5.06 mm, error 3.1%, Digital 5.08 mm, error 2.8%. Case 3 (Reference 7.99 mm): Analog 7.70 mm, error 3.6%, Digital 7.80 mm, error 2.4%.

[0093] Accordingly, the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention measures the rotation angle of the bridge (200) in real time, converts it into a deflection profile for the entire span, calculates member forces using this, and automatically evaluates safety by comparing them with allowable values. The procedure is performed in the order of rotation angle measurement (inclinometer) → deflection calculation (integral-based) → analysis model input → member force calculation → comparison with allowable values ​​→ result notification.

[0094] The real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention acquires a response by continuously arranging digital sensors (110, inclinometers) in the longitudinal direction of the bridge and calculates the response by generating a deflection profile in real time.

[0095] The structural analysis input is the deflection under unknown loads rather than directly estimating external forces, and the system is designed to analyze the structural state in high resolution using only the output response.

[0096] In a verification test, the analog-digital processing results of the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention were compared with the LVDT reference value and confirmed repeatability and reproducibility with an error of up to several percent.

[0097] Through such simulation based on response under unknown load, the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention analytically derives the total member force of the bridge from the actual deflection and automatically determines whether the calculated member force exceeds the allowable range, thereby providing a series of processes for measurement, analysis, and evaluation.

[0098] FIG. 7 is a cross-sectional view showing a bridge to which a real-time automatic bridge safety evaluation system using a digital twin model according to an embodiment of the present invention is applied.

[0099] FIG. 7 shows a conceptual cross-sectional view of a bridge for applying a real-time bridge safety automatic evaluation system (100) using a digital twin model according to an embodiment of the present invention. The real-time bridge safety automatic evaluation system (100) using a digital twin model according to the present invention installs a digital sensor (110) (inclinometer) at a predetermined location on the bridge (200), for example, at a certain interval on the bottom surface of the deck or on the girder web, to continuously measure the rotation angle response.

[0100] Each digital sensor (110) performs zero point correction and slope baseline alignment before installation and is synchronized with the same sampling period to generate a time-simultaneous data stream.

[0101] The safety evaluation terminal (120) (edge ​​terminal) of the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention is deployed near the bridge, such as at a field control box, and collects measured rotation angle data, performs preprocessing (offset removal, low-pass filtering), and then calculates the full span displacement profile in real time using an integration-based algorithm.

[0102] The calculated displacement profile is corrected by point conditions and reference measurements and converted into input for structural analysis.

[0103] The real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention immediately performs displacement-based structural analysis simulation when abnormal behavior is detected to calculate responses such as member forces, section forces, and stresses, and automatically evaluates safety by comparing them with preset allowable values. The evaluation results are classified into normal, warning, danger, etc., and transmitted to an administrator terminal, and if necessary, notifications and report generation are performed in parallel.

[0104] The safety evaluation terminal (120) of the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention operates in an edge computing manner to directly process real-time data from the on-site constant monitoring system and improves the accuracy of the analysis by using the entire span displacement profile as a simulation input.

[0105] Although commercial programs such as MIDAS, RM, and ANSYS may be used as structural analysis engines, the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention adopts an in-house analysis engine to ensure lightweighting and automation are implemented for real-time operation, taking into account management costs.

[0106] The above-mentioned in-house analysis engine is implemented based on, for example, the finite element method (FEM), and the main structural members of the bridge can be modeled as three-dimensional frame elements.

[0107] For example, Ernst cable elements are applied to the cables of cable-stayed bridges and the hanger cables of suspension bridges, and elastic catenary cable elements can be applied to the main cables of suspension bridges.

[0108] The behavior of the support section can be configured to reflect horizontal, vertical, and rotational stiffness by modeling the support stiffness of bridge bearings, etc., as spring elements.

[0109] The above-mentioned in-house analysis engine includes, but is not limited to, the following analysis modules.

[0110] First, the linear analysis module is applied to general bridge structures to calculate static responses and section forces.

[0111] Second, the geometric nonlinear analysis module is applied to cable-dominated structures such as cable-stayed and suspension bridges to consider large deformation effects and cable nonlinearity.

[0112] Third, the eigenvalue analysis module evaluates dynamic characteristics by calculating natural vibration modes and natural frequencies.

[0113] Fourth, the dynamic analysis module analyzes the time domain or frequency domain response to seismic loads and vehicle moving loads, etc. If necessary, the analysis module may be configured to be selected, added, or excluded in consultation with the user (manager) to suit the type, scale, and usage conditions of the target structure.

[0114] The above-mentioned in-house analysis engine is lightweight and automated so that it can be operated in real time at the safety evaluation terminal (120) of the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention, and the analysis results such as calculated member force, stress, and displacement are provided to the structural safety evaluation unit (126) so that structural safety is automatically evaluated through comparison with allowable values.

[0115] The real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention can construct a digital twin model that faithfully reflects actual behavior by tuning the stiffness, boundary conditions, material parameters, etc. of an in-house structural model using measurement data, and can continuously update it.

[0116] In addition, the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention applies artificial intelligence techniques to correct or regenerate data when abnormal measurement data occurs and detects pattern-based abnormal behavior. When abnormal behavior is confirmed, the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention analyzes and evaluates the safety performance of the bridge through real-time automatic simulation and immediately notifies the evaluation results to the administrator terminal.

[0117] Meanwhile, FIG. 8 is a schematic diagram of a real-time bridge safety automatic evaluation system using a digital twin model according to an embodiment of the present invention, and FIG. 9 is a diagram explaining the concept of a digital twin model applied to a real-time bridge safety automatic evaluation system using a digital twin model according to an embodiment of the present invention.

[0118] Referring to FIG. 8, a real-time bridge safety automatic evaluation system (100) using a digital twin model according to an embodiment of the present invention may include a digital sensor (110), a safety evaluation terminal (120), and an administrator terminal (130).

[0119] A digital sensor (110) is installed at a predetermined location on a bridge (200), which is a structure to be evaluated, to measure the response of the bridge (200) in real time. That is, the digital sensor (110) is implemented as an inclinometer, which is a rotation angle sensor, and upon installation, zero point correction and reference axis alignment are performed, and the response of the bridge (200) is measured while synchronized with the same sampling period.

[0120] The safety evaluation terminal (120) calculates the rotation angle and displacement profiles based on the rotation angle data measured in real time by the digital sensor (110) to determine the structural state, detects abnormal behavior of the bridge (200) according to an artificial intelligence-based anomaly detection algorithm, and automatically evaluates the structural safety of the bridge by performing a displacement-based structural analysis simulation when abnormal behavior of the bridge (200) is detected.

[0121] The administrator terminal (130) receives the structural safety evaluation results wirelessly from the safety evaluation terminal (120).

[0122] In addition, as shown in Fig. 9, the digital twin model is a virtual model that replicates a real-world physical object (thing, system, etc.) in the same structure in a virtual world, is updated in real time through sensor data, and is used for prediction, analysis, monitoring, and optimization of the actual object in operation.

[0123] These virtual models maintain consistency throughout the lifecycle of physical objects and play an important role in solving complex real-world problems by utilizing technologies such as simulation and machine learning.

[0124] Specifically, the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention according to an embodiment of the present invention acquires response data based on multi-point sensing.

[0125] That is, it is composed of a plurality of digital sensors (110) installed on the main members of the bridge (200) or structure, and the digital sensors (110) are configured to simultaneously measure vertical displacement and rotation angle data at each point, and each digital sensor (110) is arranged at regular intervals throughout the entire span to collect continuous responses of the bridge (200) which is the structure to be evaluated.

[0126] In addition, the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention according to an embodiment of the present invention generates a rotation angle profile and a displacement profile, respectively, as response profiles.

[0127] That is, multi-point deflection and rotation angle data collected from a digital sensor (110) are input, and rotation angles are integrated and corrected with boundary conditions to calculate rotation profiles and displacement profiles, which are static or quasi-static response curves of the structure at a certain point in time.

[0128] Here, the displacement profile and rotation angle profile reflect the overall shape behavior of the structure and are quantified to include segment-by-segment response distribution, continuity, and slope changes; furthermore, the rotation angle data is combined with the slope change of the deflection profile to reflect even the differential characteristics of the response curve.

[0129] In addition, the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention according to an embodiment of the present invention is implemented to be linked with a structural analysis simulation.

[0130] In other words, the displacement profile and rotation angle profile as the above response profiles are input into an in-house simulation engine and utilized to analyze the structural state even without load conditions.

[0131] In particular, the displacement profile and rotation angle profile curves are directly compared with the analysis results of the digital twin structural model and can be used as reference indicators for evaluating model consistency and automatic tuning.

[0132] In addition, the displacement profile and rotation angle profile can determine whether there are abnormalities in a specific section (e.g., excessive deflection, rotation angle displacement, etc.) through similarity analysis or threshold comparison with the reference response curve of the design or initial state.

[0133] In addition, the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention according to an embodiment of the present invention can be linked with an artificial intelligence model or a time-series anomaly detection algorithm.

[0134] Meanwhile, FIG. 10 is a detailed configuration diagram of a real-time bridge safety automatic evaluation system using a digital twin model according to an embodiment of the present invention.

[0135] Referring to FIG. 10, the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention includes a digital sensor (110), a safety evaluation terminal (120), and a manager terminal (130). The safety evaluation terminal (120) may include, but is not limited to, a response data collection unit (121), a rotation angle and displacement profile calculation unit (122), a structural status identification unit (123), an artificial intelligence-based abnormal behavior detection unit (124), a structural analysis unit (125), a structural safety evaluation unit (126), and an evaluation result transmission unit (127).

[0136] The response data collection unit (121) collects real-time rotation angle data measured from a digital sensor (110) (inclinometer) installed at a specific part of the bridge (200). Data from an auxiliary displacement meter may be collected together as needed, and zero point correction, sampling period synchronization, offset removal, and noise filtering are performed during the collection process.

[0137] The rotation angle and displacement profile calculation unit (122) calculates the rotation angle profile and displacement profile, respectively, by integrating the rotation angle data collected according to the constructed digital twin model and correcting it with boundary conditions. The calculated profiles are stored in a time-synchronized form to represent the continuous response of the entire span.

[0138] The structural state identification unit (123) identifies the structural state based on the calculated rotation angle profile and displacement profile. Specifically, it generates a response profile that quantifies the response distribution, continuity, and slope change of the entire bridge section using multi-point response data, and then provides it as a key input value for anomaly detection and structural analysis.

[0139] The AI-based abnormal behavior detection unit (124) detects abnormal behavior of the bridge (200) using an AI-based abnormal behavior detection algorithm. Specifically, it analyzes abnormal patterns of the rotation angle profile and displacement profile in real time and automatically determines abnormal behavior such as sensor abnormalities, local damage, excessive sagging, and sudden change responses in a learning-based or threshold-based manner.

[0140] To this end, the AI ​​training data generation unit generates training data, and the AI-based data learning unit uses the data to train an anomaly detection model.

[0141] The structural analysis unit (125) performs displacement-based structural analysis simulation when abnormal behavior of the bridge (200) is detected. That is, it performs structural analysis using in-house analysis software, and when an abnormality occurs, a lightweight analysis engine is automatically executed, and even without directly estimating the input load, it inversely estimates abnormalities in member forces, strength indices, and boundary conditions from the actual response.

[0142] The structural safety evaluation unit (126) automatically evaluates the structural safety of the bridge based on the results of a displacement-based structural analysis simulation. The analysis results are automatically graded by comparing them with the allowable deflection, allowable stress, and member force criteria, and the evaluation results are immediately prepared for real-time response.

[0143] The evaluation result transmission unit (127) transmits the evaluation results of the structural safety evaluation unit (126) to the administrator terminal (130) via wireless communication. The transmitted items may include grade, location, time of occurrence, excess indicator, recommended measures, etc.

[0144] The real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention is configured to enable rapid response to dangerous sections and to enable repeated analysis of the same section and tracking of long-term conditions.

[0145] To this end, the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention generates a behavior profile based on multi-point measurement by distributing a plurality of digital sensors (110) to major members and connections on the bridge (200) or structure, and captures the deformation and response of the structure at high resolution.

[0146] In addition, the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention links artificial intelligence-based abnormal behavior detection and structural analysis. Displacement profiles and rotation angle profiles are transmitted to an artificial intelligence-based abnormal behavior detection unit (124), and the artificial intelligence-based abnormal behavior detection unit (124) automatically detects abnormal patterns such as abnormal signs, local damage, sudden change responses, and sensor malfunctions within the input profile, and operates flexibly in a threshold-based or pre-learning-based manner.

[0147] In addition, the structural analysis results derived from the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention are compared and analyzed with predefined allowable deflection, stress, member force criteria, etc., so that the condition of the structure is graded and the presence or absence of abnormalities is determined.

[0148] When a risk is detected, the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention transmits a real-time alarm to the administrator terminal (130) to enable rapid maintenance and response.

[0149] Meanwhile, FIG. 11 is a diagram schematically illustrating the real-time bridge safety automatic evaluation process in a real-time bridge safety automatic evaluation system using a digital twin model according to an embodiment of the present invention, wherein the top of FIG. 11 shows the generation of a vertical displacement profile and a rotation angle profile, the middle of FIG. 11 shows a bridge that has undergone displacement-based structural analysis, and the bottom of FIG. 11 shows the evaluation of structural safety.

[0150] In the case of the real-time bridge safety automatic evaluation system using the digital twin model of the present invention, as shown in FIG. 11, by performing a simulation using the deflection caused by an unknown input as an input to the structural analysis model, the total member forces, section forces, and stress distribution of the bridge can be analytically calculated from the deflection.

[0151] The calculated results automatically determine whether a predefined allowable range has been exceeded to determine the current safety of the bridge, and provide a continuous process of measurement, analysis, and evaluation.

[0152] Here, the real-time response of the bridge (200), such as deflection and tilt (rotation angle), is collected through a rotation angle sensor, which is a digital sensor (110) installed on the bridge (200), and the collected measurements are processed in real-time by an edge computing device.

[0153] For example, as illustrated at the top of Fig. 11, a vertical displacement profile represented in orange and a rotation angle profile represented in blue are generated, and the two profiles are synchronized on the same time axis and used as key inputs for the displacement-based structural analysis of the interruption.

[0154] Meanwhile, when the real-time automatic bridge safety evaluation system using the digital twin model of the present invention uses a commercial structural analysis program, continuous costs associated with program licensing and maintenance are incurred, and the system specifications for real-time processing become excessively high, limiting its general application to field edge terminals.

[0155] Therefore, in the real-time automatic bridge safety evaluation system using the digital twin model of the present invention, it is desirable to adopt a configuration that utilizes an in-house analysis engine suitable for lightweighting and automation.

[0156] FIG. 12 is a diagram showing a rotation angle sensor (inclinometer), which is a digital sensor, placed on a bridge to which a real-time automatic bridge safety evaluation system using a digital twin model according to an embodiment of the present invention is applied.

[0157] In the example of FIG. 12, digital sensors (110a, 110b, 110c, 110d, 110e) are arranged at equal intervals along the span length L, and the support points at both ends are depicted as bridge piers or abutments (210a, 210b).

[0158] In the real-time bridge safety automatic evaluation system using the digital twin model of the present invention, the detailed specifications of the digital sensor (110) are as follows.

[0159] The rotation angle sensor (inclinometer), which is a digital sensor (110), is applied in a digital manner rather than an analog manner, and the precision may be ±0.0025° and the resolution may be 0.001°.

[0160] When installing, the sensor reference axis is aligned with the longitudinal direction of the bridge, and zero correction and sampling period synchronization are performed to generate data on the same time axis.

[0161] If necessary, the application of a temperature compensation factor and cable / power integrity checks may be included, but are not limited to.

[0162] As shown in FIG. 12, the installation interval of the digital sensors (110) is such that five or more inclinometers are placed at equal intervals within the span, and inclinometers may also be placed near both support points (piers or abutments). For example, when five inclinometers are installed, the installation interval is L / 4 (sensor positions: 0, L / 4, L / 2, 3L / 4, L), where L represents the span length.

[0163] In addition, when n inclinometers are installed, the installation interval is L / (n-1), where n represents the number of installed inclinometers. The installation location is preferably a site where structural continuity is ensured, such as the bottom surface of the deck or the girder web.

[0164] In the real-time bridge safety automatic evaluation system using the digital twin model of the present invention, the sampling rate must exhibit a performance of 10 Hz to 100 Hz so that the natural frequency of the fifth vibration mode or higher of the bridge (200) being evaluated can be observed.

[0165] The sampling rate is selected based on the length, stiffness, expected traffic load, and wind response characteristics of the target bridge, and all sensors operate at the same sampling period.

[0166] In addition, in the real-time automatic bridge safety evaluation system using the digital twin model of the present invention, the rotation angle profile generation method applies interpolation using n inclinometer data.

[0167] The above interpolation method is incorporated into a rotation angle profile generation module so that the user can select linear, Lagrange, cubic spline, B-spline, Newton, Bayesian, Akima, etc. The interpolation result provides a distribution of longitudinal rotation angles θ(x) of the interval between sensors and can be corrected with boundary conditions (e.g., point rotation constraints) if necessary.

[0168] In the case of the displacement profile generation method, the rotation angle profile is generated by numerically integrating it. Similar to the rotation angle profile, the displacement profile generation module is equipped with a numerical integration method that allows the user to select options such as rectangular, trapezoidal, Simpson, Romberg, and Gaussian quadrature. The integration constant is recalibrated based on the point displacement conditions or the reference values ​​of the auxiliary displacement gauge to restore the absolute deflection.

[0169] In addition, in the real-time automatic bridge safety evaluation system using the digital twin model of the present invention, a noise filtering module is provided to allow the user to select a method, incorporating Low-pass, High-pass, Band-pass, and Moving Average noise filtering techniques. Filter application is performed during the preprocessing stage and may include spike removal and offset correction functions.

[0170] Meanwhile, FIG. 13 is a diagram illustrating the types of error signals in a real-time bridge safety automatic evaluation system (100) using a digital twin model according to an embodiment of the present invention.

[0171] Figure 13 shows a) a normal signal, b) a signal including a missing interval, c) a drift signal where the baseline moves slowly, and d) an outlier signal where a momentary spike occurs.

[0172] As the digital sensor (110) installed on the bridge (200) operates in an outdoor environment for a long period of time, such error signals may frequently occur due to power instability, signal line disconnection / poor contact, temperature changes, electronic interference, etc.

[0173] If these error signals are used without preprocessing, data reliability is reduced, and distortion may occur in the calculation of the rotation angle-based deflection profile.

[0174] Accordingly, the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention automatically detects the type shown in FIG. 13 through the signal quality evaluation and regeneration module within the artificial intelligence-based abnormal behavior detection unit (124) of the safety evaluation terminal (120), and corrects and regenerates it as follows.

[0175] Missing intervals: After masking the missing intervals based on time synchronization, they are replaced with interpolation methods (linear, cubic spline, B-spline, Akima, etc.) or digital twin predictions.

[0176] Drift: Remove drift components by estimating baselines by segment (low-behind trend removal, higher-order regression, HP filter, etc.), and record the removal amount as a quality flag.

[0177] Outlier: Identify outliers using Hampel filters, interquartile ranges (IQR), thresholds, or learning-based detection, and mask or replace them.

[0178] Power Anomaly / Disconnection: Assigns a status flag occurring simultaneously with consecutive missing values ​​to invalidate the corresponding section and generate an administrator notification.

[0179] The regenerated signal is stored along with quality flags and correction history, and is subsequently used in the calculation of rotation angle and displacement profiles, displacement-based structural analysis, and structural safety assessment. The raw data is preserved separately and managed to enable verification and traceability.

[0180] Meanwhile, FIGS. 14a and 14b are drawings illustrating an example of abnormal signal regeneration for shock noise in a real-time bridge safety automatic evaluation system using a digital twin model according to an embodiment of the present invention.

[0181] The top of FIG. 14a illustrates the time series of five digital sensors (110) (inclinometers) installed on a simple beam, in which saturation and clipping due to impact noise occur in a specific section of sensor number 2, and the abnormal section is marked as a highlighted area. For the same point in time (e.g., time 150 seconds), the spatial diagram at the bottom shows that raw data containing abnormal signals distorts the sagging profile.

[0182] The real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention automatically detects abnormal sections based on the rapid slope change, abnormal amplitude, saturation duration, etc. characteristic of impact noise in the artificial intelligence-based abnormal behavior detection unit (124) of the safety evaluation terminal (120).

[0183] The detected segment is masked by being assigned a quality flag, and the regeneration pipeline is executed immediately. Regeneration is performed in the order of combining the normal rotation angle time series of adjacent sensors (e.g., 1, 3, 4, 5) with the digital twin prediction profile, interpolating and fusing (selectively linear, cubic spline, Akima, etc.), and then aligning the absolute standard through boundary condition correction.

[0184] The top of Fig. 14b shows the time series results after regeneration, and it can be seen that the impulse noise section is removed and replaced, and the phase and amplitude between channels are consistently restored.

[0185] The spatial diagram at the bottom presents the deflection profile at the same point in time, demonstrating that the profile after regeneration (gray solid line / marker) closely matches the AI-based predicted profile (red solid line), showing that the entire profile is stably restored even in the event of a single sensor failure.

[0186] As such, the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention is configured to detect abnormal signals in real time when impact noise occurs, and to maintain the reliability of the rotation angle profile and displacement profile calculation, displacement-based structural analysis, and structural safety evaluation stages by regenerating data using adjacent channels and the digital twin model.

[0187] The regeneration status and correction history are recorded as quality metadata, enabling future verification and auditing.

[0188] In addition, FIGS. 15a and 15b are drawings illustrating an example of an abnormal signal regeneration for a NaN (Not a Number) value in a real-time bridge safety automatic evaluation system using a digital twin model according to an embodiment of the present invention.

[0189] As illustrated in FIG. 15a and FIG. 15b, a NaN value occurred in a certain time interval at the fourth digital sensor (110), and the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention automatically detected the interval, assigned a quality flag, and then regenerated the signal using the normal time series of the adjacent channel and the digital twin prediction profile.

[0190] After regeneration, the deflection profile at the same point in time regains continuity and is corrected so as not to interfere with the calculation of the rotation angle profile and displacement profile.

[0191] Meanwhile, FIG. 16 is a diagram illustrating the error signal reconstruction network structure in a real-time bridge safety automatic evaluation system using a digital twin model according to an embodiment of the present invention.

[0192] In the case of the real-time bridge safety automatic evaluation system using the digital twin model of the present invention, as shown in FIG. 16, the error signal reconstruction network is designed based on an autoencoder and consists of an encoder and a decoder.

[0193] The encoder compresses the input vector into a low-dimensional latent representation to extract key features, and can apply a 128-dimensional fully connected layer and ReLU activation.

[0194] The decoder reconstructs the original signal from the latent representation using a fully connected layer that generates an output of the same dimension.

[0195] This network uses input dimensions equal to the number of measurement sensors and simultaneously reconstructs outputs of the same size. Sections masked by anomaly detection are replaced with the output of this network, and the replacement history and quality flags are stored.

[0196] In the case of the real-time bridge safety automatic evaluation system using the digital twin model of the present invention, the analysis engine is configured based on FEM, and three-dimensional frame elements are applied to the bridge members.

[0197] Ernst cable elements are applied to cable-stayed bridge cables and suspension bridge hanger cables, and elastic catenary cable elements are applied to main suspension bridge cables. Support stiffness, such as bridge bearings, is modeled using spring elements.

[0198] In terms of analysis modules, linear analysis is applied to general bridge structures, while geometric nonlinear analysis is applied in cases requiring large deformation effects, such as cable-stayed or suspension bridges.

[0199] Natural vibration modes and frequencies are calculated through eigenvalue analysis, and seismic loads or moving vehicle loads can be applied during dynamic analysis. If necessary, the system is configured to allow analysis modules to be added or excluded in consultation with the user to suit the target structure.

[0200] In the case of the real-time automatic bridge safety evaluation system using the digital twin model of the present invention, the initial setup method of the digital twin model is as follows. First, a basic analysis model is constructed based on design or as-built drawings.

[0201] Subsequently, the basic analysis model is updated based on measurement data, utilizing actual responses such as deflection and natural frequency for the update.

[0202] The updated interpretation model is installed in the real-time automatic bridge safety evaluation system using the digital twin model of the present invention and applied to the real-time automatic safety evaluation.

[0203] In the case of the real-time automatic bridge safety evaluation system using the digital twin model of the present invention, the automatic structural safety evaluation indicators are set according to member materials and performance standards. For steel members, the generated stress is compared against the allowable stress, and for concrete members, the generated strength is compared against the design strength to determine safety.

[0204] In the case of deflection, it is compared with the allowable deflection of the design standards, but the system is configured to allow managers or users to define threshold values ​​according to site conditions. Such evaluation settings enable the selection and application of appropriate methods considering the actual conditions of the target structure.

[0205] Ultimately, according to an embodiment of the present invention, by moving away from the conventional manual analysis-based structural diagnosis method and automating real-time structural status recognition and evaluation, disaster response time can be shortened, maintenance costs reduced, and the burden on administrators minimized.

[0206] Furthermore, according to an embodiment of the present invention, an effective digital twin is constructed through response-based analysis without load conditions, and since the structural state can be precisely analyzed solely through the output response, there is high consistency with actual behavior and no loss of physical information occurs.

[0207] In addition, according to an embodiment of the present invention, by utilizing a high-resolution response profile based on multi-point measurement to configure rotation angle and displacement data into a continuous distribution profile, asymmetric damage, local anomalies, and boundary deterioration can be detected at high resolution.

[0208] Finally, according to an embodiment of the present invention, judgment errors can be minimized by fusing artificial intelligence-based anomaly detection and structural analysis simulation, and misjudgments caused by sensor malfunctions or communication failures can be prevented by including an automatic filtering and correction function for abnormal data.

[0209] [Real-time Automatic Bridge Safety Assessment Method Using Digital Twin Models]

[0210] FIG. 17 is a flowchart illustrating a method for real-time automatic bridge safety evaluation using a digital twin model according to an embodiment of the present invention.

[0211] Referring to FIGS. 10 and FIGS. 17, the real-time automatic bridge safety evaluation method using a digital twin model according to an embodiment of the present invention is as follows.

[0212] First, the rotation angle response of the bridge (200) is measured in real time through a digital sensor (110) installed on a specific part of the bridge (200) which is the structure to be evaluated (S110).

[0213] Here, the digital sensor (110) is an inclinometer that is a rotation angle sensor and is synchronized with, for example, the same sampling period of 10 to 100 Hz, and upon installation, reference axis alignment, zero point correction, temperature correction and timestamp synchronization (NTP / PTP) are performed, and metadata such as sensor coordinates, installation surface, and point conditions are stored together.

[0214] Next, the safety evaluation terminal (120) collects rotation angle data from the digital sensor (110) (S120).

[0215] Here, the safety evaluation terminal (120) automatically detects and masks missing values / NaN, drift, outliers, impact noise, etc. simultaneously with streaming collection, performs interpolation / regeneration using adjacent channels and digital twin prediction profiles (linear, cubic spline, Akima, autoencoder output), and logs all raw values, correction values, and quality flags.

[0216] Next, a rotation angle profile and a displacement profile are calculated from the rotation angle data according to the constructed digital twin model (S130).

[0217] Here, the multi-point rotation angle is mapped to the length coordinate to interpolate θ(x,t), and w(x,t) is restored through integration and boundary condition correction (w(0)=w(L)=0 or auxiliary displacement gauge reference value), and the two profiles are synchronized to the same time axis and feature quantities such as continuity, slope change, and curvature κ(x)=∂θ / ∂x are calculated and stored together.

[0218] Next, the artificial intelligence-based abnormal behavior detection unit (124) of the safety evaluation terminal (120) detects abnormal behavior using the profile as input (S140).

[0219] Here, threshold comparison (allowable sag / slope change limit), change point detection (CUSUM), frequency domain residual (modal component comparison), autoencoder reconstruction error, etc. are applied as an ensemble to determine signs of sensor error, excessive sag, and local damage, and branch results are produced along with event type, location, severity, and confidence.

[0220] Next, when abnormal behavior of the bridge (200) is detected, the structural analysis unit (125) of the safety evaluation terminal (120) performs a displacement-based structural analysis simulation (S150).

[0221] Here, 3D frame / shell elements and cable elements (Ernst, Elastic catenary) and support springs are used as an in-house analysis engine, and geometric nonlinear analysis is applied if necessary to calculate member forces (M, V, N), section forces {Mx, My, Mxy, Nx, Ny, Nxy} and stress / strain using a w(x,t)·θ(x,t) window as input, and in the event of convergence failure, fallback is performed in the order of reduced ROM → simple beam analysis.

[0222] Next, the structural safety evaluation unit (126) of the safety evaluation terminal (120) evaluates the structural safety of the bridge based on the displacement-based structural analysis simulation results (S160).

[0223] Here, utilization rates relative to standards are calculated for allowable deflection, allowable steel stress, concrete design strength, and fatigue accumulation indices (rainflow and Miner's rule); normal / caution / warning / danger grades, items and locations exceeding limits, and recommended actions (deceleration, shutdown, inspection) are automatically determined; and digital twin model alignment evaluation and minor parameter tuning are performed in parallel.

[0224] Next, the safety evaluation terminal (120) wirelessly transmits the structural safety evaluation result to the administrator terminal (130) (S170).

[0225] Here, the transmission packet includes the event time and location, grade, supporting indicators (profile snapshot, utilization, stress map), applicable criteria, quality flag, correction history, and recommended actions; for grades above a critical threshold, an immediate notification is sent, and all records are stored for audit trails.

[0226] Here, the safety evaluation terminal (120) can automate structural state recognition and evaluation according to a digital twin model that fuses data measured in real time by a digital sensor (110) with a structural analysis simulation, and can automatically evaluate real-time bridge safety based on artificial intelligence (AI).

[0227] The following is an example scenario for explaining the operation of a real-time bridge safety automatic evaluation system (100) using a digital twin model of the present invention.

[0228] Assume a situation in which a 30-ton truck passes over a bridge (200) with a single span L = 40 m during nighttime hours. Five digital sensors (110) (inclinometers) are installed on the bridge (200) at span directions 0, L / 4, L / 2, 3L / 4, and L, and are synchronized with a 50 Hz identical sampling period to transmit rotation angle time series in real time to a safety evaluation terminal (120).

[0229] The allowable deflection is set to L / 800 (= 50 mm), and the allowable stress of the steel member is set to 150 MPa.

[0230] At the time when the truck enters (t = 152 s), the safety evaluation terminal (120) preprocesses the rotation angle data collected from the digital sensor (110) (zero point / temperature correction, spike removal, synchronization) and then calculates the rotation angle profile and the displacement profile.

[0231] At this time, the change in slope of the rotation angle profile in the L / 4~L / 2 section is Δθ = 0.004 rad, and the minimum deflection of the displacement profile is evaluated as w_min = -18.6 mm (additional deflection -6.6 mm compared to the reference state).

[0232] These profiles and feature quantities are provided to the structural state identification unit (123), and then the artificial intelligence-based abnormal behavior detection unit (124) determines the abnormal state through an ensemble judgment that considers the allowable slope change limit and the learning-based reconstruction error together.

[0233] In this example, signal quality degradation such as missing values, NaN, and impulse noise is not detected, so the regeneration procedure is not performed.

[0234] When abnormal behavior is confirmed, the structural analysis unit (125) of the safety evaluation terminal (120) immediately performs a displacement-based structural analysis simulation. The analysis uses an in-house analysis engine, applies three-dimensional frame elements and support springs, and considers geometric nonlinearity if necessary.

[0235] With w(x, t), θ(x, t) and boundary conditions as input, the maximum bending moment is calculated as M_max = 3.2 MN·m, the maximum shear force as V_max = 0.65 MN, and the maximum fiber stress as σ_max = 130 MPa. These correspond to utilization rates of 0.91, 0.72, and 0.87, respectively, relative to predefined allowable values.

[0236] The structural safety evaluation unit (126) performs a grading by comparing the above analysis results with the allowable deflection, allowable stress, and member force standards. In this example, the deflection is within the allowable limit and the stress-moment utilization rate is evaluated to be around 90%, so a “caution” grade is automatically assigned.

[0237] The evaluation result transmission unit (127) wirelessly transmits a packet to an administrator terminal (130) that includes time, location (span center ±5 m), grade, basis indicators (Δθ, w_min, utilization rate), rotation angle / displacement profile snapshot, quality flag (normal), and recommended action (deceleration pass, recommendation to inspect the center plate and branch bearings). Simultaneously with transmission, an event log and correction history (not applicable) are stored, enabling subsequent verification and audit tracking.

[0238] By the above procedure, the real-time bridge safety automatic evaluation system (100) using the digital twin model of the present invention can detect abnormalities based solely on actual response in actual operating situations such as truck passage, and quantify member forces and stress levels through displacement-based structural analysis to immediately grade and notify, thereby supporting rapid and accurate maintenance decision-making at the site.

[0239] In addition, the safety evaluation terminal (120) can measure the longitudinal rotation angle of the bridge using a digital sensor (110) and convert it into a deflection profile of the entire span in real time to calculate the deflection response.

[0240] In addition, the safety evaluation terminal (120) can automatically determine the safety of the current bridge by performing a displacement-based simulation using the deflection caused by an unknown input as input to the structural analysis model, analytically calculating the total member forces of the bridge, and checking whether the calculated member forces exceed the allowable range.

[0241] Ultimately, according to an embodiment of the present invention, measurement data is actively utilized for automated safety evaluation beyond simple monitoring, and rapid and accurate maintenance can be performed by integrating artificial intelligence and digital twin technology to process the condition analysis and diagnosis of a bridge into a single automated evaluation system.

[0242] In addition, since bridge maintenance can be performed more efficiently through real-time automatic safety assessment technology that integrates data and simulation, the efficiency and reliability of related processes, such as existing precision safety inspections, can be improved.

[0243] The foregoing description of the present invention is for illustrative purposes only, and those skilled in the art will understand that other specific forms can be easily modified without altering the technical spirit or essential features of the present invention. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form.

[0244] The scope of the present invention is defined by the claims set forth below rather than by the detailed description above, and all modifications or variations derived from the meaning and scope of the claims and equivalent concepts thereof should be interpreted as being included within the scope of the present invention. Explanation of the symbols

[0245] 100: Real-time Automatic Bridge Safety Assessment System 200: Bridge (Structure under evaluation) 110: Digital sensor 120: Safety evaluation terminal 121: Response data collection unit 122: Rotation Angle and Displacement Profile Calculation Unit 123: Structure Status Assessment Unit 124: AI-based anomaly detection unit 125: Structural Analysis Section 126: Structural Safety Evaluation Department 127: Evaluation Result Transmission Unit 130: Administrator Terminal

Claims

Claim 1 In a real-time automatic bridge safety evaluation system, a digital sensor (110) installed at a predetermined location on a bridge (200) which is a structure to be evaluated, and which measures the response of the bridge (200) in real time; and a safety evaluation terminal (120) that calculates a rotation angle and a displacement profile according to the rotation angle data measured in real time by the digital sensor (110) to determine the structural state, detects abnormal behavior of the bridge (200) according to an artificial intelligence-based anomaly detection algorithm, and, when abnormal behavior of the bridge (200) is detected, performs a displacement-based structural analysis simulation to automatically evaluate the structural safety of the bridge. The system includes a manager terminal (130) that receives the structural safety evaluation results wirelessly from the safety evaluation terminal (120), wherein the safety evaluation terminal (120) automatically recognizes and evaluates the real-time structural state according to a digital twin model that fuses data measured in real-time by the digital sensor (110) and a structural analysis simulation, thereby automatically evaluating the real-time bridge safety based on artificial intelligence (AI), and the safety evaluation terminal (120) comprises: a response data collection unit (121) that collects rotation angle data, which is response data measured in real-time through a digital sensor (110) installed at a predetermined part of the bridge (200); a rotation angle and displacement profile calculation unit (122) that calculates a rotation angle profile and a displacement profile, respectively, according to the collected rotation angle data according to the constructed digital twin model; and a structural state identification unit (123) that identifies the structural state according to the calculated rotation angle and displacement profiles. An AI-based abnormal behavior detection unit (124) that detects abnormal behavior of the bridge (200) according to an AI-based abnormal behavior detection algorithm; a structural analysis unit (125) that performs a displacement-based structural analysis simulation when abnormal behavior of the bridge (200) is detected;A structural safety evaluation unit (126) that automatically evaluates the structural safety of a bridge based on the results of a displacement-based structural analysis simulation is included, wherein the digital sensor (110) is composed of inclinometers, which are a plurality of rotation angle sensors spaced apart from each other in the span direction of the bridge (200), and the plurality of rotation angle sensors measure the longitudinal rotation angle data of the bridge (200) in real time while synchronized with the same sampling period, and the safety evaluation terminal (120) maps the rotation angle data collected from the plurality of inclinometers to the span direction length coordinates of the bridge (200), performs interpolation for the interval between the plurality of inclinometers to generate a rotation angle profile, and generates a displacement profile for the entire span by numerically integrating the rotation angle profile and correcting the integration constant based on the point displacement condition or the auxiliary displacement sensor reference value, and the artificial intelligence-based abnormal behavior detection unit (124) detects signals including missing values, drift, outliers, or impact noise included in the rotation angle data, rotation angle profile, or displacement profile. A real-time bridge safety automatic evaluation system using a digital twin model that detects errors, regenerates the section containing the signal error using normal data from adjacent sensors or a prediction profile of a digital twin model, and the structural analysis unit (125) does not directly calculate or estimate the external force or load conditions acting on the bridge (200), but performs a displacement-based structural analysis simulation using the displacement profile and rotation angle profile as input responses to the structural analysis model, thereby calculating at least one of member force, section force, or stress of the bridge (200). Claim 2 delete Claim 3 A real-time bridge safety automatic evaluation system using a digital twin model, characterized in that, in claim 1, the safety evaluation terminal (120) performs a simulation of deflection caused by an unknown load in a structural analysis model to analytically determine the total member force of the bridge caused by the deflection, and checks whether the member force calculated therefrom exceeds an allowable range to automatically determine the current safety of the bridge. Claim 4 delete Claim 5 A real-time bridge safety automatic evaluation system using a digital twin model, further comprising, in claim 1, an evaluation result transmission unit (127) that wirelessly transmits the structural safety evaluation result of the structural safety evaluation unit (126) to an administrator terminal (130). Claim 6 delete Claim 7 delete Claim 8 A real-time automatic bridge safety evaluation system using a digital twin model, characterized in that, in claim 1, the rotation angle data is combined with the slope change of the deflection profile, and the displacement profile and rotation angle profile reflect the overall shape behavior of the structure and are quantified including the response distribution by section, continuity, and slope change. Claim 9 A real-time bridge safety automatic evaluation system using a digital twin model, characterized in that, in claim 8, the displacement profile and rotation angle profile are utilized to analyze the structural state without load conditions by linking with a structural analysis simulation, and the displacement profile and rotation angle profile are directly compared with the analysis results of the digital twin model and used as reference indicators for model consistency evaluation and automatic tuning. Claim 10 A real-time automatic bridge safety evaluation system using a digital twin model, characterized in that, in claim 8, the displacement profile and rotation angle profile determine whether there is an anomaly in a specific section through similarity analysis or threshold comparison with a reference response curve of the design or initial state. Claim 11 In a method for real-time automatic safety evaluation of a bridge using a real-time automatic bridge safety evaluation system using a digital twin model of claim 1, the method comprises: a) a step of measuring the response of the bridge (200) in real time through a digital sensor (110) installed at a predetermined part of the bridge (200) which is the structure to be evaluated; b) a step in which a safety evaluation terminal (120) collects rotation angle data from the digital sensor (110); c) a step of calculating a rotation angle profile and a displacement profile, respectively, according to the collected rotation angle data according to a constructed digital twin model; d) a step in which an artificial intelligence-based abnormal behavior detection unit (124) of the safety evaluation terminal (120) detects abnormal behavior of the bridge (200) according to an artificial intelligence-based abnormal behavior detection algorithm; e) a step in which, when abnormal behavior of the bridge (200) is detected, a structural analysis unit (125) of the safety evaluation terminal (120) performs a displacement-based structural analysis simulation. and f) a method for real-time automatic bridge safety evaluation using a digital twin model, comprising the step of the structural safety evaluation unit (126) of the safety evaluation terminal (120) evaluating the structural safety of the bridge according to the displacement-based structural analysis simulation results. Claim 12 In claim 11, the real-time automatic bridge safety evaluation method using a digital twin model further comprises the step of the safety evaluation terminal (120) wirelessly transmitting the structural safety evaluation result to the administrator terminal (130). Claim 13 delete Claim 14 delete Claim 15 delete Claim 16 delete Claim 17 delete Claim 18 delete Claim 19 delete

Images