Apparatus for detecting structural condition change of marine structure and detecting method thereof

Abstract

The present invention relates to a device and method for detecting a state change of an underwater structure. A state change detection device according to one embodiment comprises: a data collection unit that collects sensing data from at least one sensor provided in an underwater structure during a preset detection period; a measurement data generation unit that generates actual pattern data, which is time-series analysis data regarding the structural behavior of an underwater structure based on the sensing data; a prediction data generation unit that generates prediction pattern data, which is time-series prediction data regarding the structural behavior of an underwater structure based on a structural behavior pattern model established to predict the structural behavior of an underwater structure; and a state change detection unit that detects a state change of an underwater structure based on the actual pattern data and the prediction pattern data.

Description

Technology Field

[0001] The present invention relates to a device for detecting changes in the state of an underwater structure, and more specifically, to a technical concept for detecting changes in the state of an underwater structure based on a sensor installed in the underwater structure and an artificial intelligence model. Background Technology

[0002] Since human access to floating underwater tunnels is very difficult and available sensors are very limited, the application of data-driven active condition assessment technology is essential for the safe use of the structure.

[0003] All structures undergo deterioration over time, and damage to major structural members may occur in succession due to repeated loading.

[0004] These floating underwater tunnels consist of the main tunnel structure and mooring lines to support it. Deterioration of key structural elements can increase the magnitude of structural responses caused by marine environmental loads and service loads, potentially leading to a significant reduction in usability. If this is not detected in advance and appropriately addressed, damage to major structural members may occur, resulting in the collapse of the structure underwater and potentially causing a major disaster. However, there is currently a lack of condition assessment technology optimized for underwater tunnels, where inspection and diagnosis by human personnel are difficult, making the development of such technology necessary. Prior art literature

[0005] Korean Registered Patent No. 10-2309077, "Photovoltaic power generation system and method applying sensor-based safety diagnosis technology" Korean Registered Patent No. 10-2139001, "Displacement measuring device for civil engineering and construction structures using a gyro sensor and method for measuring the same" The problem to be solved

[0006] The present invention aims to provide a state change detection device and a method thereof that can more easily detect changes in the state of an underwater structure, such as an underwater tunnel.

[0007] In addition, the present invention aims to provide a state change detection device and a method capable of more accurately detecting changes in the state of an underwater structure by comparing and analyzing actual pattern data and predicted pattern data in a time series. means of solving the problem

[0008] A state change detection device according to one embodiment of the present invention may include a data collection unit that collects sensing data from at least one sensor provided in an underwater structure during a preset detection period; a measurement data generation unit that generates actual pattern data, which is time-series analysis data regarding the structural behavior of an underwater structure based on the sensing data; a prediction data generation unit that generates prediction pattern data, which is time-series prediction data regarding the structural behavior of an underwater structure based on a structural behavior pattern model established to predict the structural behavior of an underwater structure; and a state change detection unit that detects a state change of an underwater structure based on the actual pattern data and the prediction pattern data.

[0009] According to one side, the underwater structure may include at least one of an underwater tunnel and a floating underwater structure.

[0010] According to one side, at least one sensor is installed inside the underwater tunnel, and may be installed at a position corresponding to at least one mooring line that has been set for the underwater tunnel.

[0011] According to one aspect, at least one sensor may include at least one of an acceleration sensor and a tilt sensor.

[0012] According to one side, the structural behavior pattern model may be a pattern recognition model based on the LSTM (Long Short Term Memory) algorithm.

[0013] According to one side, the data collection unit can collect sensing data collected from at least one sensor during a preset learning period as training data.

[0014] According to one side, the prediction data generation unit can generate a structural behavior pattern model through artificial neural network learning based on training data.

[0015] According to one side, the data collection unit can collect at least one structure data among data on material information of the underwater structure and data on geometric characteristics of the underwater structure.

[0016] According to one side, the prediction data generation unit can generate a plurality of simulation data using a pre-set structural behavior simulation model based on structural data, generate a structural behavior pattern model through first artificial neural network learning based on the plurality of simulation data, update the structural behavior pattern model through second artificial neural network learning based on the learning data, and generate prediction pattern data based on the updated structural behavior pattern model.

[0017] According to one side, at least one of the first artificial neural network learning and the second artificial neural network learning may be unsupervised learning.

[0018] According to one side, the state change detection unit calculates an error by comparing actual pattern data and predicted pattern data, monitors the trend of change and increase of the error, and can detect whether there is a state change based on the results of the monitoring.

[0019] A method for detecting a state change according to an embodiment of the present invention may include: a step of collecting sensing data from at least one sensor provided in an underwater structure during a preset detection period in a data collection unit; a step of generating actual pattern data, which is time-series analysis data regarding the structural behavior of an underwater structure, based on the sensing data in an actual data generation unit; a step of generating prediction pattern data, which is time-series prediction data regarding the structural behavior of an underwater structure, based on a structural behavior pattern model established to predict the structural behavior of an underwater structure in a prediction data generation unit; and a step of detecting a state change of an underwater structure based on the actual pattern data and the prediction pattern data in a state change detection unit.

[0020] According to one aspect, a state change detection method may further include the step of collecting sensing data collected from at least one sensor during a preset learning period as learning data in a data collection unit, and collecting at least one structure data among data regarding material information of an underwater structure and data regarding geometric characteristics of an underwater structure.

[0021] According to one aspect, a state change detection method may further include the step of generating a plurality of simulation data using a pre-set structural behavior simulation model based on structural data in a prediction data generation unit, generating a structural behavior pattern model through first artificial neural network learning based on the plurality of simulation data, and updating the structural behavior pattern model through second artificial neural network learning based on the learning data.

[0022] According to one side, the step of detecting a state change can generate predicted pattern data based on an updated structural behavior pattern model. Effects of the invention

[0023] According to one embodiment, the present invention can more easily detect changes in the state of an underwater structure, such as an underwater tunnel.

[0024] In addition, the present invention can more accurately detect changes in the state of underwater structures by comparing and analyzing actual pattern data and predicted pattern data in a time series. Brief explanation of the drawing

[0025] FIG. 1 is a drawing for explaining a state change detection device according to one embodiment. FIG. 2 is a diagram illustrating an example of collecting sensing data in a state change detection device according to one embodiment. FIG. 3 is a drawing for explaining a structural behavior pattern model according to one embodiment. FIG. 4 is a diagram illustrating an example of detecting a change in the state of an underwater structure in a state change detection device according to one embodiment. FIG. 5 is a diagram illustrating a method for detecting a change in state according to one embodiment. FIG. 6 is a diagram illustrating a method for updating a structural behavior pattern model according to one embodiment. Specific details for implementing the invention

[0026] Hereinafter, various embodiments of this document are described with reference to the attached drawings.

[0027] The embodiments and terms used therein are not intended to limit the technology described in this document to specific embodiments and should be understood to include various modifications, equivalents, and / or substitutions of said embodiments.

[0028] In describing various embodiments below, if it is determined that a detailed description of related known functions or configurations could unnecessarily obscure the essence of the invention, such detailed description will be omitted.

[0029] Furthermore, the terms described below are defined considering their functions in various embodiments, and these may vary depending on the intentions or practices of the user or operator. Therefore, their definitions should be based on the content throughout this specification.

[0030] In relation to the description of the drawings, similar reference numerals may be used for similar components.

[0031] A singular expression may include a plural expression unless the context clearly indicates otherwise.

[0032] In this document, expressions such as "A or B" or "at least one of A and / or B" may include all possible combinations of the items listed together.

[0033] Expressions such as "first," "second," "first," or "second" may modify the corresponding components regardless of order or importance, and are used merely to distinguish one component from another without limiting the components.

[0034] Where it is stated that a certain (e.g., first) component is "(functionally or telecommunicationally) connected" or "connected" to another (e.g., second) component, the certain component may be directly connected to the other component or connected through another component (e.g., third component).

[0035] In this specification, "configured to" may be used interchangeably with, depending on the context, for example, in hardware or software, "suitable for," "capable of," "modified to," "made to," "capable of," or "designed to."

[0036] In some situations, the expression "device configured to..." may mean that the device is "able to..." together with other devices or parts.

[0037] For example, the phrase “a processor configured (or set) to perform A, B, and C” may mean a dedicated processor for performing said operations (e.g., an embedded processor), or a general-purpose processor capable of performing said operations by executing one or more software programs stored in a memory device (e.g., a CPU or an application processor).

[0038] Also, the term 'or' means an inclusive or rather an exclusive or.

[0039] That is, unless otherwise noted or is not clear from the context, the expression 'x uses a or b' means any one of the natural inclusive permutations.

[0041] In the specific embodiments described above, the components included in the invention are expressed in the singular or plural according to the specific embodiments presented.

[0042] However, singular or plural expressions are selected to suit the situation presented for convenience of explanation, and the embodiments described above are not limited to singular or plural components; even if a component is expressed in the plural, it may be composed of a singular component, or even if a component is expressed in the singular, it may be composed of a plural component.

[0043] Meanwhile, although specific embodiments have been described in the description of the invention, it is obvious that various modifications are possible within the scope of the technical concept inherent in the various embodiments.

[0044] Therefore, the scope of the present invention should not be limited to the described embodiments, but should be defined by the claims set forth below as well as equivalents thereof.

[0046] FIG. 1 is a drawing for explaining a state change detection device according to one embodiment.

[0047] Referring to FIG. 1, a state change detection device (100) according to one embodiment can more easily detect a state change of an underwater structure such as an underwater tunnel.

[0048] In addition, the state change detection device (100) can more accurately detect changes in the state of an underwater structure by comparing and analyzing actual pattern data and predicted pattern data in a time series.

[0049] To this end, the state change detection device (100) may include a data collection unit (110), an actual measurement data generation unit (120), a prediction data generation unit (130), and a state change detection unit (140).

[0050] A data collection unit (110) according to one embodiment can collect sensing data from at least one sensor provided in an underwater structure during a preset detection period.

[0051] For example, at least one sensor may include at least one of an acceleration sensor and a tilt sensor.

[0052] In addition, the underwater structure may include at least one of an underwater tunnel and a floating underwater structure.

[0053] Preferably, at least one sensor is installed inside the underwater tunnel, and may be installed at a position corresponding to at least one mooring line pre-set for the underwater tunnel.

[0054] In other words, the data collection unit (110) can obtain structural responses (i.e., acceleration values ​​and / or inclination values) to major marine environmental loads such as waves and currents and other loads.

[0055] According to one aspect, the data collection unit (110) can collect sensing data collected from at least one sensor during a preset learning period as learning data, wherein the learning period may be a point in time prior to the detection period.

[0056] For example, the learning period can be set to a period of 1 to 3 years from the time the underwater structure is first installed.

[0057] According to one side, the data collection unit (110) can collect at least one structure data among data on material information of the underwater structure and data on geometric characteristics of the underwater structure.

[0058] In addition, the data collection unit (110) may also collect data on surrounding environmental factors of the underwater structure.

[0059] For example, data regarding material information may include information on the material of the underwater structure and the normal output value of the sensor according to the material, and data regarding the geometric characteristics of the underwater structure may include information on the geometric characteristics of the underwater structure and the normal output value of the sensor according to the geometric characteristics.

[0060] Additionally, data regarding geometric characteristics may include data regarding the shape of the underwater structure, and data regarding surrounding environmental factors may include at least one of wave information affecting the underwater structure, current information, traffic volume of vehicles moving through the underwater structure, and loads caused by vehicles.

[0061] For a more specific example, the data collection unit (110) can collect big data information on waves and tides in the installation area of ​​the underwater structure from the weather server as data on surrounding environmental factors.

[0062] A measurement data generation unit (120) according to one embodiment can generate measurement pattern data, which is time-series analysis data regarding the structural behavior of an underwater structure, based on sensing data.

[0063] A prediction data generation unit (130) according to one embodiment can generate prediction pattern data, which is time-series prediction data for the structural behavior of an underwater structure, based on a structural behavior pattern model established to predict the structural behavior of an underwater structure.

[0064] For example, the structural behavior pattern model may be a pattern recognition model based on the LSTM (Long Short Term Memory) algorithm.

[0065] According to one side, the prediction data generation unit (130) can generate a structural behavior pattern model through artificial neural network learning based on training data.

[0066] In other words, the prediction data generation unit (130) can accumulate sensing data from at least one sensor installed at multiple locations during a certain period (i.e., a learning period), and after recognizing a structural behavior pattern from this through the learning of an artificial intelligence algorithm, can build a structural behavior pattern model and, by utilizing this, can generate prediction pattern data including acceleration values ​​and / or slope values ​​at the location where at least one sensor is installed.

[0067] According to one aspect, the prediction data generation unit (130) may generate a plurality of simulation data using a pre-set structural behavior simulation model based on structural data, generate a structural behavior pattern model through first artificial neural network learning based on the plurality of simulation data, update the structural behavior pattern model through second artificial neural network learning based on the learning data, and generate prediction pattern data based on the updated structural behavior pattern model.

[0068] For example, the structural behavior simulation model may be a model that simulates the output values ​​(i.e., acceleration values ​​and / or inclination values) of at least one sensor in response to changes in environmental factors such as wave information, current information, and traffic volume and loads caused by vehicles moving through the underwater structure.

[0069] That is, the prediction data generation unit (130) can generate simulation data using structural data collected through the data collection unit (110) and information on changes in environmental factors directly set by the user in advance as input values ​​for the structural behavior simulation model.

[0070] For example, the prediction data generation unit (130) can generate simulation data by inputting data on surrounding environmental factors actually observed around the underwater structure along with the structure data into a structural behavior simulation model, and by generating a structural behavior pattern model through learning a first artificial neural network using this simulation data as input, it can generate a structural behavior pattern model that derives a prediction result more optimized for the surrounding environment of the underwater structure.

[0071] According to one side, at least one of the first artificial neural network learning and the second artificial neural network learning may be unsupervised learning.

[0072] A state change detection unit (140) according to one embodiment can detect a state change of an underwater structure based on actual pattern data and predicted pattern data.

[0073] According to one aspect, the state change detection unit (140) calculates an error by comparing actual pattern data and predicted pattern data with each other, monitors the trend of change and increase of the error, and can detect whether there is a state change based on the result of the monitoring.

[0074] Specifically, the state change detection unit (140) monitors the change and increase trend of error due to discrepancy between the actual pattern data and the predicted pattern data through a pattern comparison process that analyzes in a time series whether the actual pattern data observed during the detection period matches the predicted pattern data derived through the structural behavior pattern model, and can detect whether there is a change in the state of the underwater structure based on the results of the monitoring.

[0075] Preferably, the state change detection unit (140) can calculate the Root Mean Square Error (RMSE) through time series analysis between actual pattern data and predicted pattern data, and detect whether there is a change in the state of the underwater structure based on the calculated Root Mean Square Error.

[0076] According to one side, the state change detection unit (140) may provide a warning message to a preset administrator terminal when the calculated value of the root mean square error exceeds a preset threshold.

[0078] FIG. 2 is a diagram illustrating an example of collecting sensing data in a state change detection device according to one embodiment.

[0079] Referring to FIG. 2, the state change detection device can collect sensing data from at least one sensor equipped in the underwater tunnel during a preset detection period.

[0080] For example, at least one sensor may be an acceleration sensor and / or a tilt sensor, and the state change detection device may collect acceleration values ​​and / or tilt values ​​from the sensor as sensing data.

[0081] In addition, at least one sensor may be installed inside the underwater tunnel, and may be installed at a position corresponding to at least one mooring line pre-set for the underwater tunnel.

[0082] According to the examples of (a), (b), and (c) of reference numeral 200, a sea tunnel with a diameter of 23 m may have multiple tethers set at intervals of 25 m, and a state change detection device may have multiple vertical acceleration sensors and lateral acceleration sensors installed at positions corresponding to each tether.

[0083] For example, 10 vertical acceleration sensors (S6 to S15) and 5 horizontal acceleration sensors (S1 to S5) may be installed in the underwater tunnel, but the number of vertical acceleration sensors and horizontal acceleration sensors is not limited to this.

[0084] In other words, the state change detection device can more accurately acquire acceleration values ​​resulting from waves affecting the underwater structure at various angles through vertical and horizontal acceleration sensors positioned at locations corresponding to the mooring line.

[0086] FIG. 3 is a drawing for explaining a structural behavior pattern model according to one embodiment.

[0087] Referring to FIG. 3, a state change detection device according to one embodiment can generate prediction pattern data, which is time-series prediction data for the structural behavior of an underwater structure based on a structural behavior pattern model, wherein the structural behavior pattern model may be a pattern recognition model based on a Long Short Term Memory (LSTM) algorithm.

[0088] The state change detection device can generate a structural behavior pattern model through artificial neural network learning based on training data collected during a preset learning period.

[0089] According to the example of reference numeral 300, the structural behavior pattern model (300) is a model that receives training data (310) collected during a training period as input and outputs predicted pattern data, and may include an LSTM layer (320-1) that receives training data (310) as input, an FC layer (fully connected layer) (320-2) that receives the output of the LSTM layer (320-1) as input, and an output layer (320-3) that receives the output of the FC layer (320-2) as input and outputs predicted pattern data.

[0090] Meanwhile, the state change detection device may also perform a weight learning process based on backpropagation during the learning process.

[0092] FIG. 4 is a diagram illustrating an example of detecting a change in the state of an underwater structure in a state change detection device according to one embodiment.

[0093] Referring to FIG. 4, the state change detection device can detect changes in the state of an underwater structure based on actual pattern data and predicted pattern data.

[0094] Specifically, the state change detection device monitors the change and increase trend of error resulting from the discrepancy between the actual pattern data and the predicted pattern data through a pattern comparison process that time-series analyzes whether the actual pattern data observed during the detection period matches the predicted pattern data derived through a structural behavior pattern model, and can detect whether there is a change in the state of the underwater structure based on the results of the monitoring.

[0095] According to the example of reference numeral 400, the state change detection device can generate pattern change predicted data for sections 420 and 430 based on a structural behavior pattern model generated using the sensing data of the sensor collected in section 410 as learning data.

[0096] In addition, the state change detection device can generate measured pattern data based on the sensing data of the sensors collected in sections 420 and 430, and detect whether there is a change in the state of the underwater structure by comparing the generated predicted pattern data with the predicted pattern data.

[0097] Specifically, the state change detection device can determine that a state change has occurred in the underwater structure by comparing the predicted pattern data with the predicted pattern data as shown in the graph above reference numeral 400, and if an error rate of 10% is found.

[0098] Preferably, the state change detection device can calculate the Root Mean Square Error (RMSE) through time series analysis between actual pattern data and predicted pattern data, as shown in the graph at the bottom of reference numeral 400, and detect whether there is a change in the state of the underwater structure based on the calculated Root Mean Square Error. Through this, the underwater structure can easily detect a change in state even when the error rate changes from 0% to 15%, and can classify the state change according to the detected error rate and provide a warning message in response.

[0100] FIG. 5 is a diagram illustrating a method for detecting a change in state according to one embodiment.

[0101] In other words, FIG. 5 is a drawing illustrating the operation method of a state change detection device according to an embodiment described through FIG. 1 to 4, and any descriptions described through FIG. 5 below that overlap with those described through FIG. 1 to 4 will be omitted.

[0102] Additionally, although Step 530 is described below as being performed after Step 520, Step 520 may be performed before Step 510, or at the same time as Step 510 or Step 520.

[0103] Referring to FIG. 5, in step 510, the detection method can collect sensing data from at least one sensor equipped in an underwater structure during a preset detection period in the data collection unit.

[0104] For example, at least one sensor may include at least one of an acceleration sensor and a tilt sensor.

[0105] In addition, the underwater structure may include at least one of an underwater tunnel and a floating underwater structure.

[0106] Preferably, at least one sensor is installed inside the underwater tunnel, and may be installed at a position corresponding to at least one mooring line pre-set for the underwater tunnel.

[0107] In other words, in step 510, the detection method can obtain structural responses (i.e., acceleration values ​​and / or inclination values) caused by major marine environmental loads such as waves and currents and other loads in the data collection unit.

[0108] Next, in step 520, the detection method can generate actual pattern data, which is time-series analysis data regarding the structural behavior of an underwater structure, based on the sensing data in the actual data generation unit.

[0109] Next, in step 530, the detection method can generate prediction pattern data, which is time-series prediction data for the structural behavior of an underwater structure, based on a structural behavior pattern model established to predict the structural behavior of an underwater structure in the prediction data generation unit.

[0110] For example, the structural behavior pattern model may be a pattern recognition model based on the LSTM (Long Short Term Memory) algorithm.

[0111] Next, in step 540, the detection method can detect changes in the state of an underwater structure based on actual pattern data and predicted pattern data in the state change detection unit.

[0112] According to one side, in step 540, the detection method can calculate an error by comparing actual pattern data and predicted pattern data in a state change detection unit, monitor the trend of change and increase of the error, and detect whether there is a state change based on the result of the monitoring.

[0114] FIG. 6 is a diagram illustrating a method for updating a structural behavior pattern model according to one embodiment.

[0115] Referring to FIG. 6, in steps 610 to 640, a structural behavior pattern model can be generated through artificial neural network learning based on training data, and steps 610 to 640 can be performed at a point prior to step 510 of FIG. 5.

[0116] Specifically, in step 610, the update method can collect sensing data collected from at least one sensor during a preset learning period as learning data in the data collection unit, wherein the learning period may be a point in time earlier than the detection period.

[0117] For example, the learning period can be set to a period of 1 to 3 years from the time the underwater structure is first installed.

[0118] According to one side, in step 610, the update method may further collect at least one structure data among data on material information of the underwater structure and data on geometric characteristics of the underwater structure in the data collection unit.

[0119] For example, data regarding material information may include information on the material of the underwater structure and the normal output value of the sensor according to the material, and data regarding the geometric characteristics of the underwater structure may include information on the geometric characteristics of the underwater structure and the normal output value of the sensor according to the geometric characteristics.

[0120] Additionally, data regarding geometric characteristics may include data regarding the shape of the underwater structure, and data regarding surrounding environmental factors may include at least one of wave information affecting the underwater structure, current information, traffic volume of vehicles moving through the underwater structure, and loads caused by vehicles.

[0121] For a more specific example, in step 610, the update method can collect big data information on waves and tides in the installation area of ​​the underwater structure from the weather server as data on surrounding environmental factors in the data collection unit.

[0122] Next, in step 620, the update method can generate multiple simulation data using a pre-configured structural behavior simulation model based on structural data in the prediction data generation unit.

[0123] For example, the structural behavior simulation model may be a model that simulates the output values ​​(i.e., acceleration values ​​and / or inclination values) of at least one sensor in response to changes in environmental factors such as wave information, current information, and traffic volume and loads caused by vehicles moving through the underwater structure.

[0124] According to one side, in step 620, the update method can generate simulation data by using structural data and information on changes in environmental factors directly set by the user in advance as input values ​​for the structural behavior simulation model in the prediction data generation unit.

[0125] According to one side, in step 620, the update method can generate simulation data by inputting data on surrounding environmental factors actually observed around the underwater structure into the structural behavior simulation model along with the structure data in the prediction data generation unit.

[0126] Next, in step 630, the update method can generate a structural behavior pattern model through the first artificial neural network learning based on multiple simulation data in the prediction data generation unit.

[0127] Next, in step 640, the update method can update the structural behavior pattern model through the learning of a second artificial neural network based on the training data.

[0128] Meanwhile, the updated structural behavior pattern model can be used to generate prediction pattern data in step 530 of Fig. 5.

[0130] Ultimately, by using the present invention, changes in the state of underwater structures, such as underwater tunnels, can be detected more easily.

[0131] In addition, by using the present invention, changes in the state of underwater structures can be detected more accurately by comparing and analyzing actual pattern data and predicted pattern data in a time series.

[0133] Although the embodiments have been described above with reference to the limited drawings, those skilled in the art can make various modifications and variations from the description above. For example, suitable results can be achieved even if the described techniques are performed in a different order than described, and / or the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents.

[0134] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below. Explanation of the symbols

[0135] 100: State change detection device 110: Data collection unit 120: Actual data generation unit 130: Predicted data generation unit 140: State change detection unit

Claims

Claim 1 A data collection unit that collects sensing data from at least one sensor equipped in an underwater structure during a preset detection period; A measurement data generation unit that generates actual pattern data, which is time-series analysis data regarding the structural behavior of the underwater structure, based on the sensing data; a prediction data generation unit that generates prediction pattern data, which is time-series prediction data regarding the structural behavior of the underwater structure, based on a pre-established structural behavior pattern model to predict the structural behavior of the underwater structure; and a state change detection unit that detects a state change of the underwater structure based on the actual pattern data and the prediction pattern data, wherein the data collection unit collects sensing data collected from at least one sensor during a pre-set learning period as learning data, and collects at least one structural data among data regarding material information of the underwater structure and data regarding geometric characteristics of the underwater structure, and the prediction data generation unit generates a plurality of simulation data using a pre-set structural behavior simulation model based on the structural data, generates the structural behavior pattern model through first artificial neural network learning based on the plurality of simulation data, updates the structural behavior pattern model through second artificial neural network learning based on the learning data, and generates the prediction pattern data based on the updated structural behavior pattern model. Detector. Claim 2 In claim 1, the underwater structure is a state change detection device comprising at least one of an underwater tunnel and a floating underwater structure. Claim 3 In paragraph 2, the state change detection device wherein the at least one sensor is installed inside the underwater tunnel and is installed at a position corresponding to at least one mooring line pre-set for the underwater tunnel. Claim 4 In claim 1, the state change detection device comprising at least one of an acceleration sensor and an inclination sensor, wherein the at least one sensor comprises at least one of an acceleration sensor and an inclination sensor. Claim 5 In claim 1, the structural behavior pattern model is a state change detection device that is a pattern recognition model based on an LSTM (Long Short Term Memory) algorithm. Claim 6 delete Claim 7 In claim 1, the prediction data generation unit is a state change detection device that generates the structural behavior pattern model through artificial neural network learning based on the training data. Claim 8 delete Claim 9 In claim 1, a state change detection device in which at least one of the first artificial neural network learning and the second artificial neural network learning is unsupervised learning. Claim 10 In claim 1, the state change detection unit calculates an error by comparing the actual pattern data and the predicted pattern data with each other, monitors the trend of change and increase of the error, and detects whether there is a state change based on the result of the monitoring. Claim 11 In the data collection unit, a step of collecting sensing data from at least one sensor equipped in the underwater structure during a preset detection period; The method further comprises: a step of generating actual pattern data, which is time-series analysis data regarding the structural behavior of the underwater structure, based on the sensing data in the actual data generation unit; a step of generating prediction pattern data, which is time-series prediction data regarding the structural behavior of the underwater structure, based on a pre-established structural behavior pattern model for predicting the structural behavior of the underwater structure in the prediction data generation unit; and a step of detecting a change in the state of the underwater structure based on the actual pattern data and the prediction pattern data in the state change detection unit; a step of collecting sensing data collected from at least one sensor during a pre-set learning period as learning data in the data collection unit, and collecting at least one structural data among data regarding material information of the underwater structure and data regarding geometric characteristics of the underwater structure in the prediction data generation unit, generating a plurality of simulation data using a pre-set structural behavior simulation model based on the structural data, generating the structural behavior pattern model through first artificial neural network learning based on the plurality of simulation data, and updating the structural behavior pattern model through second artificial neural network learning based on the learning data. The step of detecting a state change is a state change detection method that generates the predicted pattern data based on the updated structural behavior pattern model. Claim 12 delete

Images