WO2023163524A1 - System for preventing major disaster due to building resilience collapse and method using same - Google Patents

Abstract

According to an embodiment of the present disclosure, provided is a method for preventing a serious disaster due to the collapse of a building by predicting the rigidity of the building by using information including vibration data collected by a vibration sensor (120) attached to the building, the method comprising the steps of: (a) collecting, by a training information generation module (10), vibration data and the rigidity of a building; (b) training, by a vibration characteristic training module (20), a prediction model (22); (c1) when the vibration data is input into the vibration characteristic training module (20), outputting the rigidity of the building; (d) comparing the rigidity output in the step (c1) with the size of a preset risk level; (e2) if the rigidity output in the step (d1) is less than the preset risk level, comparing an accumulated amount of an impact applied to the building during a predetermined unit period with a resilience threshold value of the building; and (f2) if the accumulated amount of the impact exceeds the resilience threshold value of the building in the step (e2), providing, by an alarm provision module (40), an alarm to a building user, a neighboring building management unit, and an infrastructure management unit.

Description

Severe disaster prevention system due to building resilience collapse and method using it

The present disclosure relates to a serious disaster prevention system and method using the same due to collapse of a building, and specifically, to a serious disaster prevention system and method for preventing disasters caused by destruction due to deterioration of a building or destruction due to demolition of a building.

The content described in this section simply provides background information for the present disclosure and does not constitute prior art.

Recently, incidents causing deaths, disappearances, and financial losses amounting to hundreds of billions of dollars have occurred frequently due to building collapse. Furthermore, numerous human casualties are occurring due to demolition of buildings and collapse of retaining walls during redevelopment.

In order to minimize the damage caused by such building collapse, research on predicting abnormal behavior of buildings and providing alarms has been increasing.

However, conventional studies have a problem in that accuracy is low because only vibration sensors are used to monitor the behavior of buildings.

In addition, there is a problem in that the accuracy is low because the building deterioration due to shocks by users of the building, abnormal weather, natural disasters, disasters, etc. is not considered.

Accordingly, it is possible to further measure the stiffness of the building, but there is a problem in that precise measurement is difficult.

In addition, there is a problem in that the accuracy of predicting the collapse of a building is low because a vast amount of information cannot be systematically classified or preprocessed.

In addition, although the stiffness of the building may be temporarily weakened, in this case, there is a problem in that a false disaster warning is provided without considering the recovery of the stiffness of the building, causing confusion.

In addition, there is a problem in that a system capable of systematically notifying the collapse of a building causes enormous damage to adjacent buildings and infrastructure.

Accordingly, an object of the present disclosure is to provide a system and method capable of predicting the stiffness of a building with high accuracy using vibration data collected in real time by a vibration sensor.

In addition, an object of the present disclosure is to provide a system and method that can be automatically collected without a user directly inputting congestion, abnormal weather, and disaster information of building users, thereby increasing convenience.

In addition, an object of the present disclosure is to provide a system and method capable of obtaining results with high reliability even under conditions in which an ultrasonic flaw detector is not applicable or a result of an ultrasonic flaw detector is difficult to trust.

In addition, an object of the present disclosure is to provide a system and method capable of obtaining high-quality learning data by excluding outliers of input information.

In addition, the present disclosure aims to provide a system and method that can take into account various events that may affect concrete.

In addition, an object of the present disclosure is to provide a system and method capable of further improving learning accuracy through labeling of input information.

In addition, an object of the present disclosure is to provide a system and method capable of preventing false alarms through measurement and determination of complex resilience indicators of building structural stiffness.

In addition, an object of the present disclosure is to provide a system and method capable of minimizing casualties and secondary damage.

The problems to be solved by the present invention are not limited to the problems mentioned above, and other problems not mentioned will be clearly understood by those skilled in the art from the description below.

According to an embodiment of the present disclosure, a method for preventing serious accidents due to collapse of a building by predicting the stiffness of the building using information including vibration data collected by the vibration sensor 120 attached to the building. (a) collecting input information including vibration data and output information including stiffness of the building by the learning information generating module 10; (b) learning, by the vibration characteristic learning module 20, the prediction model 22 using the input information and the output information; (c1) outputting the stiffness of the building when vibration data is input to the vibration characteristic learning module 20; (d) comparing the stiffness output in step (c1) with a preset risk level; (e1) repeating steps (c) to (d1) if the stiffness output in step (d1) is equal to or greater than the preset risk level; (e2) comparing the cumulative impact applied to the building for a predetermined unit period with a threshold for resilience of the building if the stiffness output in the step (d1) is less than the preset risk level; (f1) repeating steps (c1) to (e2) if the cumulative impact amount is less than or equal to the resilience threshold of the building in step (e2); And (f2) when the cumulative impact amount exceeds the resilience threshold of the building in step (e2), providing an alarm to building users, adjacent building management units and infrastructure management units by the alarm providing module 40; , provides a method.

In addition, preferably, the learning information generation module 10 according to an embodiment of the present disclosure includes a stiffness measuring unit 110, a pre-processing unit 130, an abnormal climate and disaster information input unit 140, and measurement of usage information. The unit 150 is further included, and the step (a) includes: (a1) collecting the stiffness of the building at the normal point (A) and the deterioration point (B) by the stiffness measurement unit 110; (a2) collecting vibration data of a normal point (A) and a deterioration point (B) by the vibration sensor 120; (a31) generating kurtosis by performing fast fourier transformation (FFT) on the vibration data, by the pre-processor 130; (a32) calculating, by the pre-processor 130, acceleration over time, RMS, peak, and crest factor (CF) using the vibration data; (a4) collecting weather information and disaster information by the abnormal weather and disaster information input unit 140; and (a5) collecting congestion by time slot and logistics information by time slot by the usage information measuring unit 150 .

Also, preferably, the vibration characteristic learning module 20 according to an embodiment of the present disclosure includes a learning period determining unit 24, and the step (b) includes (b1) the kurtosis, When input information including acceleration, RMS, peak, CF, congestion by time zone, logistics information by time zone, weather information, and disaster information is input, the prediction model 22 configured to output output information including the stiffness of the building vibrates. Learning by the characteristic learning module 20; and (b2) calculating, by the learning period determining unit 24, a reconstruction loss value using the input information and the output information, and adjusting a learning period using the information loss value. do.

In addition, preferably, the method according to an embodiment of the present disclosure, after the step (b2), (b3) the correction module 30 determines the stiffness of the building output in the step (b1), and the (a1) ) calculating the relational expression of the buildings collected in step; And after the step (c1), (c2) the correction module 30 correcting the stiffness of the building output in the step (c1) using the relational expression calculated in the step (b3); further comprising do.

Also, preferably, the stiffness measuring unit 110 according to an embodiment of the present disclosure is a first stiffness measuring device configured to measure a cavity formed inside a building wall, a depth of the cavity, and stiffness of the building. (112); and a second stiffness measuring device 114 configured to measure the stiffness of the building.

In addition, preferably, the first stiffness measuring device 112 according to an embodiment of the present disclosure; Is an ultrasonic flaw detector, and the second stiffness measuring device 114 is a Schmidt hammer.

In addition, preferably, when it is determined that an outlier outlier occurs among raw data of acceleration, RMS, peak, and CF calculated in step (a32) according to an embodiment of the present disclosure, data of n σ or more ( At this time, n is a positive real number) is excluded from the input information.

Also, preferably, the input information according to an embodiment of the present disclosure further includes a type of building.

Also, preferably, the cumulative impact amount according to an embodiment of the present disclosure is calculated by an impact calculation unit, and the impact calculation unit is configured by the abnormal weather and disaster information input unit 140 and the usage information measuring unit 150. The accumulated impact is calculated using the collected information.

In addition, according to an embodiment of the present disclosure, by predicting the stiffness of the building using information including vibration data collected by the vibration sensor 120 attached to the building, to prevent serious accidents due to collapse of the building A learning information generating module 10 configured to collect input information including vibration data collected by the vibration sensor 120 and output information including the stiffness of the building collected by the stiffness measurement unit 110 ); a vibration characteristic learning module 20 configured to learn a predictive model 22 configured to output the output information when the input information is input; a correction module 30 configured to correct the stiffness of the building output by the vibration characteristic learning module 20; and an alarm providing module 40 configured to determine whether or not to provide an alarm using the stiffness of the building corrected by the correction module 30 and to provide an alarm.

Also, preferably, the learning information generation module 10 according to an embodiment of the present disclosure is a pre-processing unit 130 for pre-processing the vibration data, converting the vibration data into a frequency domain to obtain Kurtosis ) FFT transform unit 132 configured to generate; and a pre-processing unit 130 including a vibration characteristic calculation unit 134 for calculating acceleration, RMS, peak, and crest factor (CF) according to time from the vibration data; An abnormal climate and disaster information input unit 140 into which weather information and disaster information are input; A usage information measuring unit 150 configured to collect usage information of a building, comprising: a user congestion measuring unit 152 configured to measure congestion by time zone; and a usage information measurement unit 150 including a logistics movement information input unit 154 configured to receive logistics information for each time period; And configured to receive and store data from the stiffness measurement unit 110, the vibration sensor 120, the pre-processing unit 130, the weather information and disaster information input unit 140, and the usage information measurement unit 150. Includes; learning information database (160).

Also, preferably, the input information according to an embodiment of the present disclosure includes the acceleration, RMS, peak, CF, kurtosis, meteorological information, disaster information, congestion by time slot, and logistics information by time slot.

Also, preferably, the vibration characteristic learning module 20 according to an embodiment of the present disclosure includes a learning period determining unit 24, and the learning period determining unit 24 is configured to connect between the input information and the output information. The learning period is adjusted using the reconstruction loss of .

Also, preferably, the stiffness measuring unit 110 according to an embodiment of the present disclosure is a first stiffness measuring device configured to measure a cavity formed inside a building wall, a depth of the cavity, and stiffness of the building. (112); and a second stiffness measuring device 114 configured to measure the stiffness of the building.

As described above, according to the present embodiment, by predicting the stiffness of the building using the information collected by the vibration sensor, it is possible to detect the abnormal behavior of the building, thereby preventing the destruction or collapse of the building. It has the effect of being able to deal with it in advance before it happens.

In addition, the usage information measurement unit and the abnormal weather and disaster information input unit are formed to collect information using previously entered data or in collaboration with an external database. It works.

In addition, since the Schmidt hammer is further used in an environment in which the reliability of the result of the ultrasonic flaw detector is lowered, there is an effect that the reliability of the result can be increased.

In addition, by selecting the Schmidt hammer as an alternative in an environment where the use of an ultrasonic flaw detector is difficult, there is an effect that the internal strength of concrete can be measured in any environment.

In addition, if an outlier outlier occurs among the input data, acceleration, RMS, peak, and CF, values greater than 1σ are excluded, so data that can be treated as noise is excluded from the obtained information, and high-quality training data is obtained. there is

In addition, the present disclosure has an effect of predicting the stiffness of a building with higher accuracy in consideration of climate change or disasters that may affect concrete.

In addition, since it considers the users who can affect the concrete, it has the effect of predicting the stiffness of the building with higher accuracy.

In addition, since the structure load according to the load and flow of the loading logistics and transport system that can affect the concrete is also considered, there is an effect that the stiffness of the building can be predicted with higher accuracy.

In addition, through vibration data labeling, there is an effect of further improving learning accuracy.

In addition, since it is judged that the temporary weakening of the stiffness of the building does not truly cause the destruction or collapse of the building, it is possible to prevent false alarms by considering the resilience of the stiffness of the building, thereby preventing unnecessary evacuation. , there is an effect that the formation of anxiety can be prevented.

In addition, due to the alarm provided to the users using the building, the users can escape the building in advance before the collapse of the building, thereby minimizing human casualties.

In addition, an alarm can be provided to users and residents of adjacent buildings around the building to be measured, so there is an effect that secondary damage can be minimized.

In addition, there is an effect that secondary damage can be minimized by providing an alarm to institutions and departments such as telecommunications companies, electrical construction companies, fire departments, Korea Electric Power Corporation, and City Gas Corporation.

1 is a block diagram of a serious accident prevention system according to an embodiment of the present disclosure.

Figure 2 is a flow chart of a serious accident prevention method according to an embodiment of the present disclosure.

3 is a block diagram of a learning information generation module according to an embodiment of the present disclosure.

4 illustrates the flow of data according to an embodiment of the present disclosure.

5 is for explaining data processed by a vibration sensor and a pre-processing unit according to an embodiment of the present disclosure.

6 is for explaining data processed by a usage information measurement unit according to an embodiment of the present disclosure.

7 is for explaining Kurtosis according to an embodiment of the present disclosure.

Figure 8 is a comparison of the predicted value and the actual value by the serious accident prevention system according to an embodiment of the present disclosure.

9 is for explaining a range in which an alarm providing module provides an alarm according to an embodiment of the present disclosure.

Hereinafter, some embodiments of the present disclosure will be described in detail through exemplary drawings. In adding reference numerals to components of each drawing, it should be noted that the same components have the same numerals as much as possible, even if they are displayed on different drawings. In addition, in describing the present disclosure, if it is determined that a detailed description of a related known configuration or function may obscure the gist of the present disclosure, the detailed description will be omitted.

In describing the components of the embodiment according to the present disclosure, symbols such as first, second, i), ii), a), and b) may be used. These codes are only for distinguishing the component from other components, and the nature or sequence or order of the corresponding component is not limited by the codes. In the specification, when a part is said to 'include' or 'include' a certain component, it means that it may further include other components, not excluding other components unless explicitly stated otherwise. .

In the present disclosure, "stiffness of a building" means the internal strength of a vicinity where a vibration sensor is attached. Since buildings are generally constructed from concrete, in this disclosure it is preferable to understand that the stiffness of a building is the strength of the concrete at the point where the vibration sensor is attached. In the present disclosure, the description is given on the premise that the unit of stiffness of a building is Mpa.

In addition, in the present disclosure, “resilience of a building” means a property in which the stiffness of a building temporarily decreases and then recovers to its original stiffness after a concentrated force is applied to a local area of the building for a short period of time. Hereinafter, the cumulative impact amount beyond the resilience and no longer recovering to its original stiffness is referred to as "the resilience threshold of the building".

In addition, in the present disclosure, the severe disaster prevention system 1 may be applied to concrete facilities including retaining walls, bridges, and dams in addition to buildings, but for convenience of explanation, a case where it is applied to a building will be described.

Severe Accident Prevention System Overview

1 is a block diagram of a serious accident prevention system according to an embodiment of the present disclosure. Figure 2 is a flow chart of a serious accident prevention method according to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, when vibration data collected from one or more vibration sensors attached to a building is input to the major disaster prevention system 1, the major disaster prevention system 1 determines the stiffness of the building, more specifically It is configured to output the internal strength of the concrete constituting the building (unit is preferably MPa). In addition, when a major disaster is expected because the predicted stiffness of the building is determined to be higher than the risk level, an alarm can be provided to minimize victims from the major disaster. That is, according to the present disclosure, by predicting the stiffness of a building using information collected by a vibration sensor, it is possible to detect an abnormal behavior of a building, and through this, it is possible to deal with it in advance before destruction or collapse of the building occurs. There is an effect that there is.

To this end, the serious accident prevention system 1 according to an embodiment of the present disclosure includes all or all of the learning information generation module 10, the vibration characteristic learning module 20, the correction module 30, and the alarm providing module 40 includes some

The learning information generation module 10 is configured to generate information necessary for the vibration characteristic learning module 20 to learn. To this end, the learning information generating module 10 includes a stiffness measuring unit 110, a vibration sensor 120, a preprocessing unit 130, an abnormal weather and disaster information input unit 140, a usage information measuring unit 150, and a learning information database. Includes all or part of (160). A detailed description related to the learning information generation of the learning information generation module 10 will be described in detail with reference to FIGS. 3 to 7 .

The vibration characteristic learning module 20 is configured to learn a prediction model capable of predicting the stiffness of a building using the information generated and collected by the learning information generating module 10 . The configuration and function of the vibration characteristic learning module 20 will be described in detail with reference to FIG. 4 .

The correction module 30 uses the correlation between the stiffness of the building output by the vibration characteristic learning module 20 and the stiffness of the actual building to correct the output stiffness of the building to be close to the stiffness of the actual building. It consists of The function of the correction module 30 will be described in detail with reference to FIG. 8 .

The alarm providing module 40 is configured to provide an alarm to building users, management offices of neighboring buildings, and infrastructure facilities associated with the building to be measured, when it is determined that the corrected stiffness of the building has reached a predetermined threshold. . The alarm providing module 40 will be described in detail with reference to FIG. 9 .

Referring to FIG. 2, the overall sequence of the major disaster prediction method using the major disaster prediction system according to an embodiment of the present disclosure will be described.

Prediction model input information is collected by the vibration sensor 120, the usage information measurement unit 150, and the abnormal climate and disaster information input unit 140, and the prediction model output information (by the stiffness measurement unit 110) output information) is collected (S200).

The vibration characteristic learning module 20 learns the predictive model 22 (see FIG. 4) using all or part of the information collected in step S220 (S210). At this time, the prediction model 22 is formed to output the stiffness of the building when information including information sensed by the vibration sensor 120 is input.

When learning is completed for a predetermined learning period, new sensing information not used for learning the prediction model by the vibration sensor 120 is input to the prediction model 22, and the stiffness of the building (specifically, the internal strength of concrete) is input. ) is output (S220). On the other hand, the usage information measurement unit 150 and the abnormal weather and disaster information input unit 140 are formed to collect information using previously entered data or in collaboration with an external database (eg, the Korea Meteorological Administration database). There is an effect that the user only needs to input the vibration data collected by the vibration sensor. At this time, preferably, the stiffness of the outputted building may be corrected by the correction module 30, but it should be noted that this can be appropriately selected according to the designer's needs.

Then, the output and/or calibrated stiffness of the building is compared to a preset threshold. If it is determined that the stiffness of the outputted building in step S230 is greater than or equal to the preset risk level, the process returns to step S220. That is, since it is greater than the risk value, this is to prevent the alarm from ringing.

If it is determined that the output stiffness of the building is lower than a predetermined threshold, the alarm providing module 40 compares the cumulative impact accumulated during the unit period with the building's resilience threshold (S240). If it is determined in step S240 that the cumulative impact is less than the resilience threshold of the building, it returns to step S220.

If it is determined in step S240 that the cumulative impact is greater than the resilience of the building, the alarm providing module 40 provides an alarm (S250). In this case, the alarm providing module 40 may provide alarms to building users, management offices of neighboring buildings, and infrastructure facilities related to the building to be measured.

Hereinafter, steps S200 to S220 will be described in detail.

Data collection and preprocessing for training

3 is a block diagram of a learning information generation module according to an embodiment of the present disclosure. 5 is for explaining data processed by a vibration sensor and a pre-processing unit according to an embodiment of the present disclosure. FIG. 6 is for explaining data processed by the usage information measurer according to an embodiment of the present disclosure. FIG. 7 is for explaining kurtosis according to an embodiment of the present disclosure.

Referring to FIGS. 3 and 5 to 7, types of data for learning will be described.

Referring to FIG. 3, the learning information generation module 10 according to an embodiment of the present disclosure includes a stiffness measuring unit 110, a vibration sensor 120, a pre-processing unit 130, and an abnormal weather and disaster information input unit 140. , Includes all or part of the usage information measurement unit 150 and the learning information database 160.

The stiffness measurement unit 110 is installed or attached to a portion of a building to measure the stiffness of the building.

The stiffness measuring unit 110 may be attached to a normal point (A) and a deterioration point (B), respectively. Here, the normal point (A) and the deterioration point (B) are points determined by the precise safety diagnosis result. Precise safety inspection means a building safety inspection conducted according to the standards provided by the law. The results of the detailed safety diagnosis are stored in an external database (not shown) that can be accessed by the building owner or a building stakeholder. The stiffness measuring unit 110 will be attached to each of the deterioration points (B) determined to be .

The stiffness measuring unit 110 includes an ultrasonic flaw detector 112 and a Schmidt hammer 114.

The ultrasonic flaw detector 112 is formed to transmit ultrasonic waves to the attached building and receive the reflected ultrasonic waves. In addition, a cavity formed inside a building wall can be detected by comparing transmitted ultrasound with reflected ultrasound. In addition, the depth of the cavity formed inside the building wall and the stiffness of the building can be measured.

Schmidt Hammer (114) is configured to test the internal strength of concrete. To this end, the Schmidt hammer is configured to estimate the internal strength by applying an impact to the attached wall surface and measuring the degree of resilience.

Meanwhile, the stiffness measuring unit 110 according to the present disclosure basically includes the ultrasonic flaw detector 112, but may optionally or additionally include the ultrasonic flaw detector 112 or the Schmidt hammer 114. This is because there is an effect that the internal strength of concrete can be measured in any environment by alternatively selecting the Schmidt hammer 114 in an environment where the use of the device 112 is difficult. In general, although the ultrasonic flaw detector 112 has a higher accuracy than the Schmidt hammer 114, the Schmidt hammer 114 has an advantage in that it can be applied to many places due to its simple operation method and low price. Therefore, in some limited environments, the Schmidt hammer 114 may be used instead of the ultrasonic flaw detector 112.

The vibration sensor 120 is formed to collect vibration data, and for this purpose is attached to a normal point (A) and a deterioration point (B), respectively.

The pre-processing unit 130 is configured to process vibration data collected by the vibration sensor 120 . To this end, the pre-processing unit 130 includes an FFT conversion unit 132 and a vibration characteristic calculation unit 134.

Vibration data is information about acceleration over time (see FIG. 5). The FFT conversion unit 132 converts vibration data from a time domain to a frequency domain to generate frequency data. The frequency data processed by the FFT converter 132 may be acceleration information with respect to frequency or amplitude information with respect to frequency.

Acceleration information with respect to frequency may be displayed in a shape such as a graph shown under the FFT conversion unit 132 of FIG. 5 . Amplitude information with respect to frequency may be displayed in a kurtosis shape as shown in the graph of FIG. 7 .

The vibration characteristic calculation unit 134 is configured to calculate an acceleration, a root mean square (RMS), a peak, and a crest factor (CF) using vibration data. Acceleration, RMS, peak, and CF can be understood using the graph shown below the vibration characteristic calculating unit 134 of FIG. 5 .

If the vibration characteristic calculating unit 134 determines that data that is close to noise in the collected acceleration, RMS, peak, and CF, that is, outliers are present in the data, it is possible to construct high-quality data by excluding them through a specific range. there is. Acceleration, RMS, peak, and CF are information identified from vibration data collected during the second period in units of the first period, and values outside a specific range are preferably excluded. Here, the first period may be, for example, 10 seconds, the second period may be 10 minutes, and the specific range may be 1σ. However, it should be noted that these are only exemplary values, and the lengths and specific ranges of the first period and the second period may be appropriately designed and changed by the user. Since values over a specific range (eg, 1σ) among acceleration, RMS, peak, and CF are excluded, data that can be treated as noise among the obtained information, that is, outliers, are excluded, and high-quality training data is obtained. It has an effect.

The abnormal weather and disaster information input unit 140 is configured to input weather changes and disasters that affect the deterioration of concrete. When events such as severe cold, heat waves, heavy rain, earthquakes, typhoons, etc., which were not considered at the time of design, occur, concrete structures are affected. Weather change and disaster information is input to the abnormal weather and disaster information input unit 140 based on a predetermined period, for example, a week or a quarter. The present disclosure has an effect of being able to predict the stiffness of a building with higher accuracy in consideration of climate change or disasters that may affect concrete.

The usage information measurement unit 150 is configured to input usage information according to the time of the building. To this end, the usage information measurement unit 150 includes a user congestion measurement unit 152 and a logistic movement information input unit 154.

The user congestion measuring unit 152 is configured to measure congestion by time zone (see (a) in FIG. 6). Figure 6 (a) shows the average number of Incheon Airport users by time period from July to August 2022. Referring to (a) of FIG. 6, it can be seen that the largest number of users is from 8:00 to 20:00. The impact of user loads can also affect the stiffness of a building. The present disclosure has an effect of being able to predict the stiffness of a building with higher accuracy because it also considers users who may affect concrete.

The logistics movement information input unit 154 is configured to input logistics information (see FIG. 6(b)) for each time period. Figure 6 (b) shows the average weight of cargoes disposed at Incheon Airport for each time period from July to August 2022. Referring to (b) of FIG. 6, it can be seen that the cargo weight is the heaviest from 9 o'clock to 20 o'clock. In other words, for the vibration data observed between 9:00 and 20:00, the weight of cargo entering and exiting the building may not be important if it is an office or residential building, but the weight of cargo entering and exiting the building is the stiffness of the building if it is a warehouse or airport. will be able to affect The present disclosure has an effect of predicting the stiffness of a building with higher accuracy because it also considers the weight of logistics that may affect concrete.

By using the degree of congestion for each time zone, the early morning time zone (Ta) with the smallest congestion and the daytime time zone (Tb) with the highest congestion may be extracted. According to the example shown in FIG. 6 , Ta may range from 5 o'clock to 6 o'clock, and Tb may range from 17 o'clock to 18 o'clock. Considering the corresponding time zone, vibration data in the corresponding time zone may be labeled and input to the predictive model 22 . Similarly, vibration data may be additionally labeled using logistic information for each time zone. Vibration data labeling has an effect of further improving learning accuracy.

Meanwhile, the information shown in FIG. 6 is exemplary for explaining user congestion and logistics movement information, and the information obtained may be different depending on the type of building.

The learning information database 160 receives data collected by the stiffness measuring unit 110, the vibration sensor 120, the pre-processing unit 130, the abnormal weather and disaster information input unit 140, and the usage information measuring unit 150. and store. At this time, preferably, the learning information database 160 converts the received information into a dataset and stores it. For example, the learning information database 160 may vectorize and store inputted information.

The type of building may be further input into the learning information database 160 . The type of building may be, for example, a general building, a department store, a mart, a public facility for citizens (for example, a sports facility, a library, etc.), an airport, a school, a concrete retaining wall, an industrial complex infrastructure facility, and the like.

The learning information database 160 may transmit the stored data set to the vibration characteristic learning module 20 upon request from the vibration characteristic learning module 20 .

Input data, output data, and predicted results of the learning model

4 illustrates the flow of data according to an embodiment of the present disclosure.

Referring to FIG. 4 , the vibration characteristic learning module 20 is configured to learn the prediction model 22 . The prediction model 22 is the acceleration collected or processed by the vibration sensor 120 and the pre-processing unit 130, RMS, peak, CF and Kurtosis, and the time period measured by the user congestion measuring unit 152 Congestion, logistic information by time zone entered into the logistics movement information input unit 154, weather information and disaster information input into the abnormal weather and disaster information input unit 140 are input data, and the building measured by the stiffness measurement unit 110 The stiffness of is used as output data.

Preferably, the prediction model 22 may be implemented as an unsupervised learning or reinforcement learning model. However, it is not necessarily limited thereto, and artificial neural networks (ANNs), convolutional neural networks (CNNs), recurrent neural networks (RNNs), Bayesian networks, long short-term memory (LSTMs), generative neural networks (GANs) It can be implemented by any one of adversarial network), Transformer, and decision tree, and is not necessarily limited to the above model, and may be implemented with any kind of predictive model.

Anomaly detection of time-series vibration data detects anomalies in observed values, point anomalies, and anomalies caused by changes in data tendencies. The method of measuring the score can be divided into categories such as 'Reconstruction error', 'Prediction error' and 'Dissimilarity', and here, the anomaly score that is optimal for each prediction model is measured.

For example, in case of using LSTM and autoencoder method, reconstruction error is measured by encoding the time-series data into a low-dimensional space and then decoding the encoded vector again. Anomaly data detection is performed on the assumption that the reconstruction error measured in this way is small in case of normal data and large in case of abnormal data.

The vibration characteristic learning module 20 may arbitrarily set a learning period, but may preferably adjust the learning period to obtain a more accurate result. To this end, the vibration characteristic learning module 20 may further include a learning period determining unit 24 .

The learning period determining unit 24 takes into account the stiffness of the building output by the prediction model 22 and the reconstruction loss of the difference between the input data and the output data input to the prediction model 22. The period can be reset. Specifically, the learning period may be adjusted until the anomaly of the information loss value falls within a preset threshold range. Because of this, no matter what kind of learning model the predictive model 22 is implemented, an appropriate learning period can be set, and the compatibility of already given data with various models can be increased.

Figure 8 is a comparison of the predicted value and the actual value by the serious accident prevention system according to an embodiment of the present disclosure.

Referring to FIG. 8 , it is confirmed that the stiffness of the actual building (refer to the actual axis) and the stiffness of the building output by the vibration characteristic learning module 20 (refer to the predicted axis) have a high correlation.

However, since there may be some difference between the stiffness of the actual building and the stiffness of the output building, it is necessary to correct the stiffness of the building output by the prediction model 22. To this end, the serious accident prevention system 1 of the present disclosure may correct the output stiffness of the building so that the stiffness of the outputted building through the correction module 30 approaches the stiffness of the actual building. For example, the correction module 30 may correct the stiffness of the outputted building using linear regression. However, it should be noted that this is only an example, and an appropriate correction equation may be selected by the user.

Hereinafter, a method of determining whether to provide an alarm using the output stiffness of the building will be described in detail (see steps S230 to S250 of FIG. 2 ).

How to provide alarm

According to step S230 of FIG. 2 , the alarm providing module 40 may determine whether the stiffness of the outputted building is greater than a preset risk level (threshold). At this time, the preset risk level may be the structural safety standard strength specified in the rules on the structural standards of buildings. However, it is not necessarily limited thereto and may be adjusted according to the management level of the safety manager or the building owner.

If it is determined that the output stiffness of the building is greater than or equal to a preset risk level, the alarm provision module 40 determines not to provide an alarm. Therefore, it returns to the step of collecting vibration data using the vibration sensor 120.

According to step S240 of FIG. 2 , when it is determined that the output stiffness of the building is less than a preset risk level, the alarm providing module 40 determines whether the cumulative impact is greater than the resilience threshold of the building.

After a short-term, concentrated force is applied to a local area of a building, the building may temporarily lose its stiffness. When these shocks accumulate, the building can lose its resilience. Here, the magnitude of the impact at which the building loses its resilience is called the resilience threshold. The magnitude of the accumulated impact is the accumulation of the magnitude of the impact applied to the building in a preset period, that is, a unit period. To this end, a separately provided impact calculation unit (not shown) may calculate the accumulated impact using information measured or input by the abnormal weather and disaster information input unit 140 and the usage information measuring unit 150.

For example, if it is confirmed that an earthquake has occurred in the vicinity of the building to be measured by the abnormal weather and disaster information input unit 140, the impact calculation unit calculates the amount of impact applied to the building using the magnitude of the earthquake, the distance to the epicenter, and the duration of the earthquake size can be calculated. Alternatively, when more than twice as many people and logistics as the number of available people in the building are detected by the usage information measuring unit 150 for 1 hour, the impact calculation unit can calculate the size of the impact applied to the building by the weight of the users and the logistics there is.

When it is determined in step S240 that the magnitude of the cumulative impact is less than the resilience threshold of the building, the alarm providing module 40 determines not to provide an alarm. Therefore, it returns to the step of collecting vibration data using the vibration sensor 120. That is, since it is determined that the result of the temporary weakening of the rigidity of the building does not truly cause the building to be destroyed or collapsed, a false alarm can be prevented. Due to this, there is an effect that unnecessary evacuation, anxiety, etc. can be prevented.

When it is determined in step S240 that the magnitude of the cumulative impact is greater than the threshold for resilience of the building, the alarm provision module 40 determines to provide an alarm as in step S250.

Here, the alarm provided to the alarm provision module 40 includes an alarm provided to users in the building, an alarm provided to an adjacent building management unit, and an alarm provided to infrastructure facilities associated with the building.

An alarm provided to users in a building may be provided through, for example, a speaker or a display. Alternatively, a push alarm or the like may be provided through an application (for example, a department store application) installed in a terminal of a user in a building. Due to this, there is an effect that users using the building can escape the building in advance before the collapse of the building, thereby minimizing human casualties.

The alarm provided to the management unit of the adjacent building may be provided through a terminal provided to the management unit provided in the adjacent building. Here, the adjacent buildings may include buildings and facilities facing the building to be measured, as shown in FIG. 9 . If the building to be measured collapses, secondary damage may occur to neighboring buildings due to falling rocks and dust generated during the collapse. On the other hand, if an alarm is provided to the management unit of an adjacent building, it is possible to prepare for damage caused by falling rocks and impact, and an alarm can also be provided to users and residents of the adjacent building, thereby minimizing secondary damage.

Alarms provided to infrastructure facilities associated with buildings may be provided to agencies, ministries, and facilities that manage overhead high-voltage lines, communication facilities, high-pressure tanks, substations, city gas, and water supply, for example. In the case of damage to infrastructure such as electricity, communication, gas, water, etc., extensive damage may occur. Therefore, there is an effect that secondary damage can be minimized by providing an alarm to agencies and ministries such as telecommunications companies, electrical construction companies, fire departments, Korea Electric Power Corporation, and City Gas Corporation.

The above description is merely an example of the technical idea of the present embodiment, and various modifications and variations can be made to those skilled in the art without departing from the essential characteristics of the present embodiment. Therefore, the present embodiments are not intended to limit the technical idea of the present embodiment, but to explain, and the scope of the technical idea of the present embodiment is not limited by these embodiments. The scope of protection of this embodiment should be construed according to the claims below, and all technical ideas within the scope equivalent thereto should be construed as being included in the scope of rights of this embodiment.

[Description of code]

1: Severe Accident Prevention System

10: learning information generation module

20: vibration characteristic learning module

30: correction module

40: alarm providing module

110: stiffness measuring unit

120: vibration sensor

130: pre-processing unit

140: abnormal weather and disaster information input unit

150: usage information measuring unit

160: learning information database

Claims (14)

1. As a method for preventing serious accidents due to the resilience collapse of a building by predicting the stiffness of the building using information including vibration data collected by the vibration sensor 120 attached to the building,

   (a) collecting input information including vibration data and output information including stiffness of a building by the learning information generation module 10;

   (b) learning, by the vibration characteristic learning module 20, the prediction model 22 using the input information and the output information;

   (c1) outputting the stiffness of the building when vibration data is input to the vibration characteristic learning module 20;

   (d) comparing the stiffness output in step (c1) with a preset risk level;

   (e1) repeating steps (c) to (d1) if the stiffness output in step (d1) is equal to or greater than the preset risk level;

   (e2) comparing the cumulative impact applied to the building for a predetermined unit period with a threshold for resilience of the building if the stiffness output in the step (d1) is less than the preset risk level;

   (f1) repeating steps (c1) to (e2) if the cumulative impact amount is less than or equal to the resilience threshold of the building in step (e2); and

   (f2) in the step (e2), when the cumulative impact exceeds the resilience threshold of the building, the alarm providing module 40 provides an alarm to building users, adjacent building management units, and infrastructure management units; Including,

   method.
2. According to claim 1,

   The learning information generation module 10 further includes a stiffness measuring unit 110, a preprocessing unit 130, an abnormal climate and disaster information input unit 140, and a usage information measuring unit 150,

   In step (a),

   (a1) collecting the stiffness of a building at a normal point (A) and a deterioration point (B) by the stiffness measuring unit 110;

   (a2) collecting vibration data of a normal point (A) and a deterioration point (B) by the vibration sensor 120;

   (a31) generating kurtosis by performing fast fourier transformation (FFT) on the vibration data, by the pre-processor 130;

   (a32) calculating, by the pre-processor 130, acceleration over time, RMS, peak, and crest factor (CF) using the vibration data;

   (a4) collecting weather information and disaster information by the abnormal weather and disaster information input unit 140; and

   (a5) collecting congestion by time slot and logistics information by time slot by the usage information measuring unit 150;

   method.
3. According to claim 2,

   The vibration characteristic learning module 20 includes a learning period determining unit 24,

   In step (b),

   (b1) When input information including the Kurtosis, acceleration, RMS, peak, CF, congestion by time zone, logistics information by time zone, weather information, and disaster information is input, output information including the stiffness of the building is output Learning the prediction model 22 configured to be performed by the vibration characteristic learning module 20; and

   (b2) the learning period determination unit 24 calculating a reconstruction loss using the input information and the output information, and adjusting the learning period using the information loss value; ,

   method.
4. According to claim 3,

   After step (b2), (b3) calculating, by the correction module 30, a relational expression between the stiffness of the building output in step (b1) and the building collected in step (a1); and

   After the step (c1), (c2) the correction module 30 correcting the stiffness of the building output in the step (c1) using the relational expression calculated in the step (b3); further comprising ,

   method.
5. According to claim 2,

   The stiffness measuring unit 110 includes a first stiffness measuring device 112 configured to measure a cavity formed inside a building wall, a depth of the cavity, and stiffness of the building; and

   Including a second stiffness measuring device 114 configured to measure the stiffness of the building,

   method.
6. According to claim 5,

   The first stiffness measuring device 112; is an ultrasonic flaw detector,

   The second stiffness measuring device 114 is a Schmidt hammer,

   method.
7. According to claim 2,

   If it is determined that an outlier occurs among the raw data of acceleration, RMS, peak, and CF calculated in step (a32), data of n σ or more (where n is a positive real number) is excluded from the input information. felled,

   method.
8. According to claim 2,

   The input information further includes the type of building,

   method.
9. According to claim 2,

   The cumulative impact amount is calculated by an impact calculation unit,

   The impact calculation unit calculates the cumulative impact amount using information collected by the abnormal weather and disaster information input unit 140 and the usage information measuring unit 150,

   method.
10. As a device for preventing serious accidents due to the resilience collapse of a building by predicting the stiffness of the building using information including vibration data collected by the vibration sensor 120 attached to the building,

    A learning information generating module 10 configured to collect input information including vibration data collected by the vibration sensor 120 and output information including the stiffness of the building collected by the stiffness measuring unit 110;

    a vibration characteristic learning module 20 configured to learn a predictive model 22 configured to output the output information when the input information is input;

    a correction module 30 configured to correct the stiffness of the building output by the vibration characteristic learning module 20; and

    An alarm providing module 40 configured to determine whether to provide an alarm using the stiffness of the building corrected by the correction module 30 and provide an alarm;

    Device.
11. According to claim 10,

    The learning information generation module 10,

    As a pre-processing unit 130 for pre-processing the vibration data,

    an FFT conversion unit 132 configured to generate kurtosis by converting the vibration data into a frequency domain; and

    A pre-processing unit 130 including a vibration characteristic calculation unit 134 for calculating acceleration according to time, RMS, peak, and CF (crest factor) from the vibration data;

    An abnormal climate and disaster information input unit 140 into which weather information and disaster information are input;

    As a usage information measuring unit 150 configured to collect usage information of a building,

    a user congestion measuring unit 152 configured to measure congestion by time zone; and

    a usage information measurement unit 150 including a logistics movement information input unit 154 configured to receive logistics information by time slot; and

    Learning configured to receive and store data from the stiffness measurement unit 110, the vibration sensor 120, the preprocessor 130, the weather information and disaster information input unit 140, and the usage information measurement unit 150 Including; information database 160;

    Device.
12. According to claim 11,

    The input information includes the acceleration, RMS, peak, CF, kurtosis, weather information, disaster information, congestion by time zone, and logistics information by time zone.

    Device.
13. According to claim 12,

    The vibration characteristic learning module 20 includes a learning period determining unit 24,

    The learning period determining unit 24 adjusts the learning period using a reconstruction loss between the input information and the output information.

    Device.
14. According to claim 10,

    The stiffness measuring unit 110 includes a first stiffness measuring device 112 configured to measure a cavity formed inside a building wall, a depth of the cavity, and stiffness of the building; and

    Including a second stiffness measuring device 114 configured to measure the stiffness of the building,

    Device.

Images