Vibration data determination device, monitoring system, and vibration data classification system

Abstract

To construct a system for extracting the output of target vibration in the output of a sensor for detecting vibration.SOLUTION: A vibration data determination device 100 includes an input unit 110 connected to an output of a vibration sensor, a calculation unit 120 connected to the input unit 110, a storage unit 130 connected to the calculation unit 120, and an output unit 140 connected to the calculation unit 120, in which the calculation unit 120 performs first order linear regression analysis on the output of the vibration sensor and determines vibration of the output of the sensor by the linear regression analysis.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a vibration data determination device for determining the output of a vibration sensor. [Background technology]

[0002] Sensors that detect vibrations by detecting acceleration are known. By detecting vibrations, it is possible to detect abnormalities in equipment and building sway, and to determine the soundness of a building based on the vibrations. A system for evaluating the soundness of a building based on the output of an acceleration sensor, particularly waveform information, has already been proposed (Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2023-159679 Summary of the Invention [Problem to be solved by the invention]

[0004] As disclosed in Patent Document 1, a structural health monitoring system is known as a system for evaluating the soundness of a building, which estimates the seismic response of the entire building during an earthquake based on information from a limited number of sensors installed in the building.

[0005] Seismometers (accelerometers) are used as sensors to detect earthquake shaking, and they measure acceleration. By installing these seismometers in buildings, it is possible to monitor shaking during an earthquake and determine the structural integrity and interior damage (damage, safety) from the waveform information of the recorded acceleration.

[0006] A system for determining the soundness of a building in the event of an earthquake must be able to correctly detect earthquake vibrations and process them correctly as normal response waveforms.

[0007] However, since vibrations are measured by sensors, vibration sensors also measure various vibration phenomena at and near the measurement location of the sensor, and may detect shaking from local vibration phenomena where the sensor is installed and output it as earthquake vibrations.

[0008] Local vibration phenomena include, for example, mechanical vibrations from equipment installed in a building, vibrations caused by indoor facility construction work, and environmental vibrations caused by nearby construction work or the movement of large vehicles.

[0009] In this case, the vibration sensor may detect vibrations other than those caused by an earthquake, which is the target of the detection, and may make an erroneous judgment on the soundness of the building based on the vibrations other than those caused by an earthquake.

[0010] Similarly, due to the effects of electromagnetic noise from lightning strikes or sensor failure, phenomena that are not normal vibration waveforms may be processed as vibration waveforms, resulting in an incorrect assessment of the soundness of the building.

[0011] In this way, to accurately determine the soundness of a building, it is necessary to correctly process the waveforms that are not responses to earthquakes, and to ensure the accuracy of the determination when operating as a monitoring system, it is necessary to improve the reliability of the entire monitoring system, including the vibration sensors.

[0012] This issue is common not only to monitoring systems for determining the soundness of buildings, but also to systems that require increased reliability in the output of vibration sensors.

[0013] The present invention has been made in consideration of the above circumstances, and aims to construct a system for extracting the output of a target vibration from the output of a sensor that detects vibration. [Means for solving the problem]

[0014] A vibration data determination device that determines an output of a vibration sensor that detects vibration, The vibration data determination device an input connected to the output of the vibration sensor; a calculation unit connected to the input unit; a storage unit connected to the calculation unit; an output unit connected to the calculation unit; Equipped with The calculation unit performs a linear regression analysis on the output of the vibration sensor, and determines the vibration of the sensor output based on the results of the linear regression analysis. [Effects of the Invention]

[0015] According to one embodiment of the present invention, it is possible to extract a target vibration from the output of a vibration sensor. [Brief explanation of the drawings]

[0016] [Figure 1] 1 is a schematic configuration diagram showing an example of a LAN-type earthquake motion evaluation system 1 according to an embodiment of the present invention. [Figure 2] 1 is a block diagram illustrating an example of a vibration determining device according to an embodiment of the present invention. [Figure 3] This is a diagram showing the waveform of vibrations caused by an earthquake and an analysis of the waveform of the vibrations. [Figure 4] This is a diagram showing an analysis of a waveform that is different from the waveform of vibrations caused by an earthquake, and the waveform of the vibration. [Figure 5] 10 is a flowchart showing a procedure for the vibration determination device to analyze vibration data. [Figure 6] This is an example in which multiple vibration sensors (301, 302, 303) are installed in a high-rise building. [Figure 7] 7 is an example of a vibration source identification device 700. [Figure 8] The feature extraction module 230 extracts features from time-series waveform data. [Figure 9] 10 is a flow chart of the operation of the classification module 240. [Figure 10] 10 is a flow chart showing the operation of the countermeasure providing module 250. [Figure 11] The idea is that the operation and management module 260 manages the entire system by managing other modules. [Figure 12] This is the flow of the building health assessment device. DETAILED DESCRIPTION OF THE INVENTION

[0017] Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings.

[0018] Embodiment 1 Fig. 1 is a schematic configuration diagram showing an example of a LAN-type earthquake motion evaluation system 1 according to an embodiment of the present invention. Fig. 2 is a block diagram of a calculation processing and determination device showing an example of a LAN-type earthquake motion evaluation system 1 according to an embodiment of the present invention.

[0019] The vibration sensor 300 detects vibrations of a monitoring target and outputs the detected vibration data to the vibration data determination device 100 via the network 200.

[0020] The vibration data input to the vibration data judgment device 100 is stored in a memory unit 130 provided in the vibration data judgment device 100. The vibration data recorded in the memory unit 130 is subjected to waveform processing in a calculation unit 120 provided in the vibration data judgment device 100.

[0021] The calculation unit 120 processes the vibration data by waveform regression analysis to obtain a regression line, which can be easily obtained by the least squares method. An example of a regression line is y=a+b·x (unit: gal, a: shift amount, b: slope, x: time).

[0022] Using the obtained regression line, the calculation unit 120 can determine whether the vibration detected by the vibration sensor 300 is the vibration that the vibration sensor 300 is trying to detect from the monitored object, or whether it is due to vibration or noise other than that of the monitored object.

[0023] In addition, the calculation unit 120 may calculate a short time average (hereinafter also referred to as "STA"), which is the average value for the short period of time when vibration occurs and triggers (exceeding the recording threshold), and a long time average (hereinafter also referred to as "LTA"), which is the average value for the period during which vibration occurs.

[0024] Furthermore, the calculation unit 120 may calculate the absolute value of the difference between the maximum and minimum values ​​of the vibration waveform at the peak.

[0025] The calculation unit 120 can determine whether the vibration data detected by the vibration sensor 300 is the vibration to be monitored based on the relationship between the calculation result and the threshold value stored as reference data in the memory unit 130.

[0026] When the calculation unit 120 analyzes and judges the vibration data using the regression line obtained by waveform processing, it can do so based on the shift amount and the slope value. If there is no influence of noise, etc., the shift amount a and the slope b will be close to 0, but if the values ​​of the shift amount a and the slope b are different from 0 and their absolute values ​​are large, it can be determined that the data includes the influence of noise, a malfunction of the vibration sensor 300, etc.

[0027] Moreover, if the difference between STA and LTA is greater than a predetermined threshold, it can be determined that the vibration data output by the vibration sensor 300 is abnormal.

[0028] Furthermore, if the difference between the maximum and minimum peak values ​​of the vibration data is smaller than a predetermined threshold, it can be determined that the vibration sensor 300 is not responding.

[0029] The threshold value used by the calculation unit 120 to determine whether the output of the sensor is being affected by noise or the like can be set appropriately depending on the monitoring target.

[0030] As described above, the calculation unit 120 can easily determine whether the vibration data is affected by noise, etc. by performing linear regression analysis on the output of the vibration sensor 300. Furthermore, a more accurate determination can be made by analyzing other vibration waveforms.

[0031] 1 shows an embodiment in which the present invention is applied to a building health monitoring system 1. Note that the application of the present invention is not limited to the building health monitoring system 1. Each component will be described below.

[0032] (Vibration sensor 300) An acceleration sensor can be used as the vibration sensor 300. As the acceleration sensor, a MEMS (Micro Electro Mechanical Systems) type acceleration sensor may be adopted and placed on the base of a column, floor, or ceiling of each floor of a building. Note that the vibration sensor 300 is not limited to an acceleration sensor, and a velocity sensor, a non-contact displacement sensor, or the like can be selected as appropriate.

[0033] The vibration sensor 300 may be installed on each floor of a building, or may be installed only on a representative floor that can indicate the state of the building. By installing the vibration sensor 300 on each floor, the vibration of each floor can be detected, and the vibration of each floor of the building can be accurately detected. In the event of a disaster such as an earthquake, damage to each floor due to vibrations caused by the earthquake can be predicted. The information can also be used to aid in the evacuation of people in the building.

[0034] Furthermore, by installing the vibration sensors 300 only on representative floors, the cost of installing the vibration sensors 300 can be reduced, and the amount of information required for calculation can be reduced, thereby increasing the speed of calculation.

[0035] The vibration sensor 300 is connected to the vibration data determination device 100 via a network 200. Vibration data detected by the vibration sensor 300 is input to an input unit 110 of the vibration data determination device 100. The connection between the vibration sensor 300 and the input unit 110 of the vibration data determination device 100 can be wired or wireless. When connected wirelessly, Bluetooth communication or Wi-Fi communication can be used.

[0036] (Vibration data determination device 100) The vibration data determination device 100 is a device that analyzes vibration data output from a vibration sensor 300. The vibration data determination device 100 includes an input unit 110, a storage unit 130, a calculation unit 120, and an output unit 140.

[0037] The vibration data determination device 100 may be configured as a personal computer, a smartphone, etc. The vibration data determination device 100 may be installed adjacent to or integral with the vibration sensor 300, or may be installed in the building to be monitored, or may be installed in a central management center 500 that manages the building to be monitored.

[0038] (input unit 110) The vibration data determination device 100 includes an input unit 110. The input unit 110 is provided to receive vibration data from the vibration sensor 300. The vibration sensor 300 and the input unit 110 are connected via a network 200, which may be wired or wireless.

[0039] The vibration data input by the input unit 110 and output from the vibration sensor 300 is recorded in the storage unit 130. The vibration data input by the input unit may be stored in the storage unit 130 and then transmitted to the calculation unit 120, or may be transmitted directly to the calculation unit without going through the storage unit 130.

[0040] The input unit 110 may be configured to receive data or programs other than the vibration data output from the vibration sensor 300. The input unit 110 may be a device configured to allow a person to input data, such as a keyboard or a mouse. The input unit 110 may be connected to a central management center 500 that manages the building via a network 200, and configured to receive signals, data, programs, etc. from the central management center 500.

[0041] (Storage unit 130) The storage unit 130 can store data and programs input from the input unit 110. The storage unit 130 may be configured with a semiconductor memory, a magnetic storage device, an optical storage device, or the like. The storage unit 130 can store data and programs input from the input unit 110, store vibration data for analysis by the calculation unit 120, and provide the vibration data when the calculation unit 120 performs calculations.

[0042] The storage unit 130 can also store a program for the calculation unit 120 to perform calculations. The storage unit 130 can also store the relationship between vibration data and vibration, noise, failure, etc. The calculation unit 120 can determine whether the analyzed vibration data is vibration caused by an earthquake, other vibrations, or noise, etc., based on the relationship between the analyzed vibration data and the stored vibration data and vibration, noise, failure, etc.

[0043] (Computation unit 120) The vibration data determination device 100 includes a calculation unit 120. The calculation unit 120 performs waveform processing on the vibration data recorded in the storage unit 130. By performing regression analysis as waveform processing, a regression line can be obtained. The regression line can be easily obtained by the least squares method. Example of a regression line: y = a + b x (unit: gal, a: shift amount, b: slope, x: time)

[0044] The calculation unit 120 identifies the waveform shape and the maximum and minimum values ​​of the waveform from the vibration data, calculates the difference between the maximum and minimum values, and determines that the vibration sensor 300 is not detecting vibration if the difference between the maximum and minimum values ​​is smaller than a predetermined value.

[0045] The calculation unit 120 may calculate STA, which is the average value for a short period of time while vibration is occurring, and LTA, which is the average value for the entire period during which vibration is occurring. Here, the period for calculating STA to analyze vibration caused by an earthquake can be between 5 and 60 seconds, but vibration caused by an earthquake can be accurately determined by setting it to about 15 to 25 seconds, especially 20 seconds.

[0046] The calculation unit 120 may also calculate the difference in absolute value between the maximum and minimum peaks of the vibration waveform. If the difference in absolute value between the maximum and minimum peaks of the vibration waveform is equal to or greater than a predetermined value, it can be determined that the vibration waveform is not caused by an earthquake. The threshold value can be selected from values ​​between 0.3 and 0.7 gal, but a threshold value of 0.5 gal can be used to accurately determine whether the vibration waveform is caused by an earthquake.

[0047] Using the obtained regression line, the calculation unit 120 can determine whether the vibration is the one that is being detected from the monitoring target, or whether it is due to vibration or noise other than that of the monitoring target.

[0048] When the calculation unit 120 makes a judgment using the regression line obtained by calculation, the judgment can be made based on the values ​​of the shift amount and the slope. If there is no influence of noise or the like, the shift amount a and the slope b will be values ​​close to 0, but if the absolute values ​​of the shift amount a and the slope b are large, it can be determined that the data contains the influence of noise or the like. Here, if the absolute value of the shift amount a is greater than 0.1 gal, it can be determined that the waveform is abnormal. Also, if the absolute value of the slope b is greater than 0.01, it can be determined that the waveform is abnormal.

[0049] When the maximum value of the waveform is smaller than a predetermined value, the calculation unit 120 determines that the waveform is vibration caused by an earthquake. Furthermore, by performing the above calculations, the calculation unit 120 can determine whether the output of the vibration sensor 300 is vibration caused by an earthquake, vibration caused by other causes, or a waveform caused by noise.

[0050] (output unit 140) The vibration data determination device 100 includes an output unit 140 that can output the calculation results of the calculation unit 120. The output unit 140 can output the calculation results of the calculation unit 120 via a network 200 to an alarm device 400 installed in the building and to a central management center 500 that manages the building.

[0051] The alarm device 400 issues an alarm to people in the building according to the magnitude of the earthquake based on the information from the output unit 140, thereby preventing further human damage. Furthermore, even if vibrations occur in the building, if it is determined that the vibrations are not caused by an earthquake, it can notify people in the building that the vibrations are not caused by an earthquake.

[0052] The central management center 500 can properly monitor the building based on the determination results of the vibration determination device. Furthermore, the central management center 500 can use this information to obtain vibration information about neighboring buildings and establish a management system.

[0053] An example of analyzing the vibration waveform of a specific earthquake FIG. 3 shows an example in which the output of the vibration sensor 300 detects a correct earthquake response waveform, and FIG. 4 shows an example in which the output of the vibration sensor 300 detects a waveform that is not an earthquake response waveform.

[0054] Figure 3A shows a waveform recorded as an output from a vibration sensor 300 that detected earthquake vibrations. Figure 3B shows a waveform processed by an LPF ("low pass filter": the same applies below). Figure 3C shows a waveform for the calculation interval of STA (20 seconds). Figure 3D shows a regression line obtained from the analysis. Figure 3E shows STA (20 seconds). Figure 3F shows LTA.

[0055] Figure 4A shows a waveform recorded as an output from a vibration sensor 300 that detected vibrations other than earthquakes. Figure 4B shows a waveform processed by an LPF. Figure 4C shows a waveform for the calculation interval of STA (20 seconds). Figure 4D shows a regression line obtained from the analysis. Figure 4E shows STA (20 seconds). Figure 4F shows LTA.

[0056] Figure 3A shows a waveform recorded as an output from vibration sensor 300 due to an earthquake, and the computing device processes the output from vibration sensor 300 using an LPF for analysis. In order to analyze earthquake vibrations, it is desirable to cut frequency components above 5 Hz. Figure 3B shows a waveform processed using an LPF (high-cut processing).

[0057] Figure 3C shows the waveform for the interval used to calculate STA (20 seconds). When analyzing seismic waveforms, it is desirable to calculate the average over an interval of about 20 seconds.

[0058] Figure 3D shows the regression line obtained from the analysis. From Figure 3D, it can be seen that the shift amount a and the slope b are both nearly zero.

[0059] Figure 3D shows the regression line obtained from the analysis. Figure 3E shows STA (20 seconds). Figure 3F shows LTA.

[0060] Figure 4A shows a waveform recorded as an output from the vibration sensor 300, including vibration waveforms other than those of earthquakes. The computing device processes the output from the vibration sensor 300 using an LPF for analysis. To analyze earthquake vibrations, it is desirable to cut frequency components of 5 Hz or higher. Figure 4B shows a waveform processed using the LPF (high-cut processing).

[0061] Figure 4C shows the waveform for the interval used to calculate STA (20 seconds). When analyzing seismic waveforms, it is desirable to calculate the average over an interval of about 20 seconds. Figure 4D shows the regression line obtained from the analysis. From Figure 4D, it can be seen that the shift amount and slope are both different from 0, meaning that they are significantly different from the earthquake response waveform.

[0062] FIG. 5 is a flowchart showing an example of the analysis operation. The analysis procedure will be explained with reference to the flowchart.

[0063] (Step S110) In order to perform analysis and judgment, the vibration data judgment device 100 reads vibration data (acceleration waveform) from the vibration sensor 300 via the network 200 using the input unit 110. The read vibration data is stored in the memory unit .

[0064] (Step S120) The vibration data undergoes high-cut processing before being analyzed by the calculation unit 120. In order to analyze the waveform of earthquake vibrations, high-cutting can be performed using a frequency selected from the range of 2 to 8 Hz, taking into consideration the structure of the building, etc. In this embodiment, frequency components of 5 Hz or higher are cut. Note that if the vibration data has been processed in advance using an electric circuit (LPF), etc., this step does not need to be performed.

[0065] (Step S130) The calculation unit 120 performs a first-order linear regression analysis based on the high-cut processed vibration data. The linear regression analysis can be easily performed using the least squares method. The first-order linear regression calculation can obtain y = a + bx (unit: gal, a term: shift amount, b term: slope).

[0066] (Step S140) The calculation unit 120 obtains the maximum and minimum values ​​of the waveform from the vibration data. The obtained maximum and minimum values ​​of the waveform are used to determine the vibration data.

[0067] (Step S150) The calculation unit 120 performs calculation processing of the waveform's STA (shoot time average, trigger position: average value over a 20-second interval) and LTA (long time average, average value over the entire waveform interval).

[0068] (Step S160) The calculation unit 120 calculates the difference between the maximum and minimum values ​​of the waveform obtained in step S140. If the difference between the maximum and minimum values ​​of the waveform is small, for example, 0, which means there is no difference, it is determined that the vibration sensor 300 is not responding. (Step S170)

[0069] The calculation unit 120 calculates the difference between STA and LTA obtained in step S150. If the difference between STA and LTA is large, for example, greater than 1 gal, the vibration data from the vibration sensor 300 is analyzed as a waveform other than that of an earthquake.

[0070] (Step S180) The calculation unit 120 calculates the difference between the maximum and minimum peak differences of the waveform obtained in step S140. If the difference between the maximum and minimum values ​​of the waveform is large, for example, if the difference is greater than 0.5 gal, the vibration data is analyzed as a waveform other than that of an earthquake.

[0071] (Step S190) The calculation unit 120 may analyze the vibration data as a waveform other than that of an earthquake when the slope b of the equation y=a+b·x (unit: gal, a: shift amount, b: slope, x: time) obtained by the first-order linear regression calculation obtained in step S130 is large, for example, when it is larger than 0.01, or when it is larger than 0.001. Also, when the shift amount a is large, the calculation unit 120 may analyze the vibration data as a waveform other than that of an earthquake.

[0072] (Step S200) The calculation unit 120 can determine whether the vibration is due to an earthquake or other vibrations based on the maximum value of the waveform. For example, if the maximum value of the waveform is 1 gal or less, the calculation unit 120 analyzes the vibration data as a waveform other than an earthquake. The threshold value can be set depending on the situation where the building is located.

[0073] (Step S210) Waveforms that do not fall under the criteria of steps S160 to S200 can be determined to be waveforms caused by earthquakes. (Step S220) A waveform that falls under any of the above steps S160 to S200 can be determined to be a waveform other than that caused by an earthquake.

[0074] By analyzing the vibration data output by the vibration sensor 300 using the above analysis procedure, it is possible to easily determine whether the vibration data output by the vibration sensor 300 is due to vibration caused by an earthquake, vibration other than an earthquake, a malfunction of the vibration sensor 300, the effects of electromagnetic noise, etc.

[0075] In the above embodiment, if any of steps S160 to S200 is true, the vibration data is judged to be a waveform other than an earthquake waveform. However, it is also possible to configure the system to analyze all of steps S160 to S200 and make a judgment based on the combined analysis results.

[0076] A system for monitoring the health of a building is preferably a monitoring system comprising a vibration sensor that detects vibrations and a vibration data determination device that determines the output of the vibration sensor, wherein the vibration sensor outputs vibration data, and the vibration data determination device comprises an input unit connected to the output of the vibration sensor, a calculation unit connected to the input unit, a memory unit connected to the calculation unit, and an output unit connected to the calculation unit, and wherein the calculation unit performs a first-order linear regression analysis of the output of the vibration sensor and determines whether the output of the vibration sensor is vibration caused by an earthquake based on the results of the linear regression analysis.

[0077] Furthermore, in the system for monitoring the health of a building, it is preferable that the calculation unit obtains the equation y=a+b·x (unit: gal, a: shift amount, b: gradient, x: time) by the linear regression analysis, and if the absolute value of the gradient b in the equation is greater than 0.01, judge that the output of the vibration sensor is a waveform caused by a cause other than an earthquake.

[0078] Furthermore, in the system for monitoring the health of a building, it is preferable that the calculation unit calculates a short-time average, which is the average value for a short period of time while vibration is occurring, a long-time average, which is the average value while vibration is occurring, and the absolute value of the difference between the maximum and minimum values ​​of the vibration waveform at their peaks, and if the difference between the short-time average and the long-time average is greater than 0.3 gals, it determines that the vibration data output by the vibration sensor is a waveform caused by a cause other than an earthquake, if the difference between the maximum and minimum values ​​of the vibration data at their peaks is less than 0.5 gals, it determines that the vibration sensor is not responding, and if the maximum value of the waveform is 1 gal or less, it determines that the vibration is a vibration other than an earthquake.

[0079] Furthermore, it is preferable that the system for monitoring the soundness of a building further has a learning function.

[0080] (Other embodiments) Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments, and can be modified as appropriate within the scope of the technical concept of the present invention.

[0081] For example, the analysis of vibration waveforms may be performed using other methods. The present invention performs a first-order linear regression analysis of vibration data from the vibration sensor 300, and the cause of vibration in the vibration data can be identified using AI for the regression line obtained by the analysis.

[0082] The calculation unit 120 of the vibration determination device performs steps S110, S120, S130, S140, and S140, which were performed in the above embodiment, and determines the difference between STA and LTA, etc., thereby making it possible to determine whether the output of the vibration sensor 300 is due to vibration or electromagnetic noise, etc., identify the source of vibration, and determine whether the vibration sensor is faulty.

[0083] When it is determined that the output of the vibration sensor 300 is due to vibration, the vibration can be identified. For example, the output of the vibration sensor 300 can be configured to identify various vibration sources, such as those due to an earthquake, those associated with the rotation of a motor, and those due to a fall.

[0084] When the vibration determination device is employed to detect automobile collisions, it may be configured to be able to distinguish between vibrations caused by normal driving, vibrations caused by opening and closing doors, vibrations caused by sudden acceleration or braking of the automobile, and vibrations caused by traffic accidents.

[0085] (Other embodiments) The central management center 500 may be connected to the information collection device 600 via a network 200. The vibration data acquired by the central management center 500 may be acquired by the information collection device 600 and further analyzed in the information collection device 600.

[0086] The analysis results collected by the central management center 500 and the results of newly performed analyses may be output to the central management center 500 via the network 200. The central management center 500 can update the reference data used for judgment in the vibration data judgment device 100 via the network 200.

[0087] By updating the reference data of the vibration data judgment device 100, analysis can be performed based on the latest data, and the accuracy of the vibration data judgment device 100 can be improved.

[0088] Furthermore, when the vibration sensor 300 installed in a building is updated to a different model, the reference data can be updated to correspond to the updated vibration sensor 300, making it easier to update the vibration sensor 300.

[0089] The vibration determination device 100 compares the determination result based on the vibration output by the vibration sensor with earthquake information from the Japan Meteorological Agency and other sources, thereby improving the accuracy of the monitoring system.

[0090] When it is determined that a vibration waveform that the vibration determination device 100 has determined to be caused by an earthquake is not caused by an earthquake, or when it is determined that a vibration waveform that the vibration determination device 100 has determined to be caused by something other than an earthquake is not caused by an earthquake, the device can learn based on the results and improve the accuracy of subsequent determinations.

[0091] <Other embodiments>

[0092] FIG. 6 shows an example in which multiple vibration sensors (301, 302, 303) are arranged in a multi-story building.

[0093] The plurality of vibration sensors (301, 302, 303) may be arranged such that, for example, the vibration sensor (301) is located on the lowest floor, the vibration sensor (303) is located on the top floor, and the vibration sensor (302) is located on an intermediate floor.

[0094] Vibrations detected by multiple vibration sensors (301, 302, 303) are recorded by a data measurement logger, and the vibration data judgment device analyzes, interprets, judges, and displays the information, and an alarm is issued by an alarm device if necessary.

[0095] Although vibration detection is possible with a single vibration sensor (300), vibration detection can be performed with higher accuracy by using multiple vibration sensors (301, 302, 303).

[0096] When multiple vibration sensors (301, 302, 303) detect vibrations, for example, when the vibrations are detected in the order of the vibration sensor (301) located on the lowest floor, the vibration sensor (302) located on an intermediate floor, and the vibration sensor (303) located on the highest floor, it can be analyzed that the cause of the vibrations is vibrations transmitted to the building on higher floors via the ground. Vibrations occurring on the ground can be considered to be caused by earthquakes, traffic vibrations, etc.

[0097] When multiple vibration sensors (301, 302, 303) detect vibrations, for example, when the vibrations are detected in the order of the vibration sensor (303) located on the top floor, the vibration sensor (302) located on an intermediate floor, and the vibration sensor (301) located on the bottom floor, it can be analyzed that the cause of the vibrations is vibrations that have been transmitted to the vibration sensors (301, 302, 303) on each floor through the entire building. Vibrations occurring throughout the building can be considered to be caused by strong winds.

[0098] 6 shows an example in which multiple vibration sensors (301, 302, 303) are arranged, with the vibration sensor (301) on the lowest floor, the vibration sensor (303) on the top floor, and the vibration sensor (302) on an intermediate floor, but multiple vibration sensors (300) may be arranged on the same floor. By arranging multiple vibration sensors (300) on the same floor and detecting the speed (timing) at which each of the multiple vibration sensors (300) detects vibration and the magnitude of the vibration of each vibration sensor (300), it is possible to predict or detect where on the same floor a vibration has occurred.

[0099] The vibration sensors (301, 302, 303) may be arranged on multiple floors, and multiple sensors may be arranged on the same floor. By arranging the vibration sensors (300) three-dimensionally and determining the order (timing) of vibration detection by the multiple vibration sensors (300), the location of the cause of vibration can be identified, enabling highly accurate vibration detection.

[0100] In order to enable the monitoring system to make a quick and accurate judgment, a vibration source identification device 700 can be used to determine from what vibration source a vibration waveform is generated and to which the vibration is responding.

[0101] 7 shows a vibration source identification device 700. The vibration source identification device 700 is composed of a data collection module 210, a preprocessing module 220, a feature extraction module 230, a classification module 240, a countermeasure provision module 250, and an operation and monitoring module.

[0102] Vibrations detected by the vibration sensor are analyzed by a data collection module 210, a pre-processing module 220, a feature extraction module 230, a classification module 240, a countermeasure provision module 250, and an operation and management module 260, and the results are reported to the outside via the output unit as notifications or alarms.

[0103] <Data Collection Module 210> The data collection module 210 uses various sensors to collect real-time vibration waveform data from earthquakes, mechanical vibrations, traffic vibrations, construction vibrations, etc. The data is sent to a cloud server and may be managed by a central management device or a local PC. The data collection module 210 is connected to the vibration sensor via a wired or wireless connection, and vibration-related data from the vibration sensor (300) is sent to the pre-processing module 220 via the data collection module 210.

[0104] <Pre-processing module 220> The preprocessing module 220 performs preprocessing such as noise removal, normalization, and missing value completion on the data received by the data collection module 210. In addition, a Butterworth filter or the like may be used to remove noise.

[0105] <Feature Extraction Module 230> The feature extraction module 230 uses the Tsfresh library to extract features from the time-series data, and may also use a Butterworth filter or the like to remove noise. <Classification Module 240> The classification module 240 classifies the vibration data into different types using a machine learning model, such as a random forest, a deep learning model, etc. The classification results are displayed in real time on a monitor in a central control room, and appropriate countermeasures can be provided.

[0106] <Measurement provision module 250> Based on the classification results by the classification module 240, the countermeasure provision module 250 can provide appropriate countermeasures, such as a report on the earthquake healthiness of the building, evacuation instructions if the building is in danger, and in the case of vibrations other than earthquakes, for example, instructions for equipment maintenance when an abnormality in equipment is detected.

[0107] <Operation and Monitoring Module 260> The operation and monitoring module 260 can operate and monitor models using MLflow and Flask. The operation and monitoring module updates models and improves accuracy every time new data is collected. MLflow is a library designed for machine learning lifecycle management (MLOps) and is an application primarily used for data management. The operation and monitoring module 260 can execute and manage requirements definition, data analysis, feature engineering, model selection, and learning (tuning).

[0108] FIG. 8 shows an image of how the feature extraction module 230 extracts features from time-series waveform data.

[0109] Figure 8A shows time series data of vibration, and using the Tsfresh library, feature points can be extracted as trend data shown in Figure 8B, periodic data shown in Figure 8C, spike data shown in Figure 8D, and absolute quantity data shown in Figure 8E.

[0110] In FIG. 8, trend data, periodic data, spike data (noise), and absolute amount data are extracted as feature points, but it is not necessary to extract these four points, and feature points from other perspectives may also be extracted.

[0111] The classification module 240 performs machine learning based on the amount of information extracted by the feature extraction module 230, and classifies the vibration types such as earthquake response waveforms.

[0112] FIG. 9 shows the flow of operation of the classification module 240. <Step S310> In step S310, the input data is set as features extracted by the feature extraction module 230. Trend data, periodicity data, spike data (noise), and absolute amount data are input as feature points, but other data may also be used.

[0113] <Step S320> In step S320, a learning model is generated. Examples of the learning model include random forest and deep learning.

[0114] <Step S330> In step S330, the learning model is trained, predicted, and evaluated. The training, prediction, and evaluation of the learning model can be performed appropriately using well-known methods.

[0115] <Step S340> In step S340, the type of vibration is determined based on the output of step S330, whether the vibration is due to an earthquake or other vibration.

[0116] The countermeasure provision module 250 performs subsequent processing based on the results of the evaluation by the classification module 240. If the classification module 240 determines that the vibration is an earthquake response, the module proceeds to the building health assessment process and displays the results. At the same time, the module also sends necessary email notifications and issues commands such as machine control required in the event of a disaster.

[0117] FIG. 10 shows the flow of the operation of the measure providing module 250.

[0118] <Step S410> In step S410, the classification of the vibration is set based on the output of the classification module 240. Vibrations of a building can be classified into those caused by earthquakes, vibrations caused by equipment installed in the building, construction work near the building, vibrations caused by large vehicles, etc.

[0119] <Step S420> In step S420, countermeasures against the vibrations are selected based on the vibrations classified in step S410. If the vibrations are due to an earthquake, the need to conduct a building integrity assessment to determine whether damage has occurred to the building or to notify the building that an earthquake has occurred is considered, and if the vibrations are due to a malfunction of equipment in the building, the need to repair the equipment is considered.

[0120] <Step S430> In step S430, the countermeasure selected in step S420 is notified. Methods for notifying the countermeasure include displaying the results of the countermeasure, sending emails to relevant parties, making announcements in the building, or stopping the operation of equipment that is thought to have malfunctioned.

[0121] FIG. 11 shows an image of the operation and management module 260 managing the entire system by managing other modules.

[0122] According to the above embodiment, multiple sensors are installed in a building to collect vibration data such as earthquake, mechanical vibration, traffic vibration, and construction vibration. The data is transmitted to a cloud server, where it can be denoised and normalised by the pre-processing module 220. Furthermore, the Tsfresh library can be used to extract features from time-series data and store them in a database. Furthermore, a machine learning model can be used to classify the vibration data into different types, and appropriate countermeasures can be notified to a central control room or users based on the classification results.

[0123] Furthermore, according to the above embodiment, it is possible to collect vibration data over a wider area in a network environment by adding an IoT device to the data collection module 210. In addition, it is possible to improve the classification accuracy by type using a deep learning model, and the model can be operated and monitored by MLflow, allowing the model to be continuously updated.

[0124] The flow of the building health assessment device is shown in Figure 12. The flow of the building health assessment device determining whether damage has occurred in a building in which an earthquake has been detected will be explained.

[0125] <Step S510> In step S510, data relating to vibration is received from the vibration sensor 300.

[0126] <Step S520> In step S520, it is determined whether the waveform detected by the vibration sensor 300 is a waveform related to an earthquake.

[0127] <Step S530> In step S530, if it is determined that the waveform detected by the vibration sensor 300 is a waveform related to an earthquake, it is determined whether or not the building has been damaged by the vibrations of the earthquake, and whether or not to issue an earthquake warning to people staying there.

[0128] <Step S540> In step S540, depending on the content determined in step S530, a message is sent to a required location by e-mail or an alarm is issued by an alarm device.

[0129] Although the example of the vibration source identification device 700 focuses on earthquakes, this embodiment is not limited to earthquakes. It can detect vibrations in equipment, and based on the detection, it can detect abnormalities in the equipment, issue an alarm, or prompt for maintenance depending on the detected vibrations.

[0130] By installing a vibration sensor in a car and detecting the vibrations of the car, it is possible to detect abnormalities in the car. For example, by detecting the vibrations of the brakes while the car is running, it is possible to detect abnormalities in the brake pads, and by detecting the vibrations of the engine, it is possible to detect the need to change the engine oil or the wear of engine parts. [Explanation of symbols]

[0131] 1. Monitoring system (earthquake motion evaluation system) 100...Vibration data determination device 110...input section 120...Arithmetic section 130...Storage section 140...Output section 200…Network 210...Data collection module 220...Pretreatment module 230...Feature extraction module 240...Classification module 250…Countermeasure provision module 260...Operation and management module 300...Vibration sensor 301...Vibration sensor 302...Vibration sensor 303...Vibration sensor 400…Alarm device 500...Central Management Center 600...Information gathering device 700…Vibration source identification device

Claims

1. A vibration data determination device that determines an output of a vibration sensor that detects vibration, The vibration data determination device an input connected to the output of the vibration sensor; a calculation unit connected to the input unit; a storage unit connected to the calculation unit; an output unit connected to the calculation unit; Equipped with the calculation unit performs a linear regression analysis on the output of the vibration sensor, and determines the vibration of the sensor output based on the result of the linear regression analysis. Vibration data determination device.

2. the calculation unit determines whether the output of the sensor is due to vibration by the linear regression analysis. The vibration data determination device according to claim 1 .

3. The calculation unit calculates, by the linear regression analysis, The equation y = a + b x (unit: gal, a: shift amount, b: gradient, x: time) is obtained, The value of b is used to determine whether the output of the vibration sensor is due to vibration or other causes. The vibration determination device according to claim 2 .

4. The calculation unit calculates a short time average, which is an average value for a short time while vibration is occurring, a long time average, which is an average value for a period during which vibration is occurring, and The absolute value of the difference between the maximum and minimum peak values ​​of the vibration waveform, Calculate If the difference between the short-time average and the long-time average is large, the vibration data output by the vibration sensor is determined to be abnormal; If the difference between the maximum and minimum peak values ​​of the vibration data is small, it is determined that the vibration sensor is not responding. The vibration determination device according to claim 3 .

5. a vibration sensor for detecting vibrations; a vibration data determination device for determining an output of the vibration sensor; A monitoring system comprising: The vibration sensor outputs vibration data. The vibration data determination device an input connected to the output of the vibration sensor; a calculation unit connected to the input unit; a storage unit connected to the calculation unit; an output unit connected to the calculation unit; Equipped with the calculation unit performs a linear regression analysis on the output of the vibration sensor, and determines whether the output of the vibration sensor is vibration caused by an earthquake based on the result of the linear regression analysis. Monitoring system.

6. The calculation unit calculates, by the linear regression analysis, The equation y = a + b x (unit: gal, a: shift amount, b: gradient, x: time) is obtained, If the absolute value of the slope b in the equation is greater than 0.01, the output of the vibration sensor is determined to be a waveform caused by a cause other than an earthquake. The monitoring system of claim 5.

7. The calculation unit calculates a short time average, which is an average value for a short time while vibration is occurring, a long time average, which is an average value while vibration is occurring, and The absolute value of the difference between the maximum and minimum peak values ​​of the vibration waveform, Calculate If the difference between the short-time average and the long-time average is greater than 0.3 gal, the vibration data output by the vibration sensor is determined to be a waveform caused by a cause other than an earthquake, If the difference between the maximum and minimum peak values ​​of the vibration data is less than 0.5 gal, it is determined that the vibration sensor is not responding. If the maximum value of the waveform is 1 gal or less, the vibration is determined to be vibration other than an earthquake. The monitoring system of claim 6.

8. 6. The monitoring system according to claim 5, which has a learning function.

9. a feature extraction module that collects vibration data and extracts features from time-series data of the vibration data; a classification module that uses a machine learning model to classify the vibration data into different classes; A vibration data classification system comprising:

Images