Bolt loosening estimation device, bolt loosening estimation method

Abstract

The present invention provides a bolt loosening estimation device that can quickly and accurately estimate the direction and distance of a loosened bolt. [Solution] The system comprises a vibration unit 15 that vibrates the steel structure 3, a sound collection unit 16 that collects sound from the observation position P toward the observation direction F, and a control unit C that creates sound collection data Ma which records the volume at the observation position P. The above problem is solved by a bolt loosening estimation device comprising: a control unit C, a sound collection data analysis unit 22 that derives the time difference between the arrival of the excitation sound and the resonance sound, and the volume difference between the volume of the resonance sound and the reference volume; a first distance derivation unit 251 that derives a first estimated distance based on the time difference; a second distance derivation unit 252 that derives a second estimated distance by inputting the volume difference into a learned distance estimation model; a third distance derivation unit 253 that derives a third estimated distance by weighting and combining the first and second estimated distances; and a notification means 12 that notifies the third estimated distance.

Description

Technical Field

[0001] The present invention relates to a bolt loosening estimation device and a bolt loosening estimation method for estimating bolt loosening, and particularly to a bolt loosening estimation device and a bolt loosening estimation method for steel structures such as iron towers.

Background Art

[0002] Conventionally, as a method for inspecting bolt loosening of steel structures such as iron towers, a method by close visual inspection such as checking the deviation of bolt mating marks has been known. However, according to such a method that relies on human senses, inspection errors may occur due to the proficiency and fatigue of the inspector. Therefore, as shown in Patent Document 1 below, a technique for vibrating a bolt with a hammer or the like and analyzing the vibration sound to detect bolt loosening (abnormality) is known.

[0003] Here, according to the technique described in Patent Document 1, bolt loosening (abnormality) can be detected without relying on human senses by analyzing the vibration sound. However, in order to analyze the vibration sound of the bolt, the inspector still has to climb to a high place of the steel structure and inspect one by one, and as a result, there are disadvantages that the work burden is large and rapid inspection cannot be performed.

[0004] Therefore, Patent Document 2 discloses a bolt loosening estimation device and a bolt loosening estimation method for estimating bolt loosening at a distance by vibrating a steel structure with a sound collecting device (microphone) having directivity and collecting the resonance sound emitted by the loosened bolt by the sound collecting device. According to the technique described in this Patent Document 2, it is possible to estimate the presence or absence of bolt loosening and the direction in which the loosened bolt exists without the inspector climbing to a high place of the steel structure and inspecting each bolt, thereby reducing the work burden.

Prior Art Documents

Patent Documents

[0005] [Patent Document 1] Japanese Patent Publication No. 2000-131195 [Patent Document 2] Japanese Patent Publication No. 2024-49982 [Overview of the project] [Problems that the invention aims to solve]

[0006] However, while the technology described in Patent Document 2 can estimate whether a bolt is loose and the direction in which the loosened bolt is located, it is difficult to accurately estimate the distance to the loosened bolt, resulting in the time and effort required to pinpoint the location of the loosened bolt. In contrast, Patent Document 2 also proposes a method to estimate the location and distance of a loosened bolt by performing vibration at multiple locations on a steel structure, collecting the vibration sound from each location using a sound collection device, deriving multiple directions in which the loosened bolt is located, and finding the intersection point. However, this method requires multiple vibrations and sound collections, making it inefficient.

[0007] Therefore, the present invention aims to solve these problems and provide a bolt loosening estimation device and bolt loosening estimation method that can quickly and accurately estimate the direction and distance of a loosened bolt, thereby reducing the effort required to identify the location of the loosened bolt. [Means for solving the problem]

[0008] To achieve the above objective, the first invention is: A vibration unit that applies a constant force to a steel structure having bolt fastenings, A sound collection unit that collects sound from the observation position toward the observation direction, The system includes a control unit that causes the excitation unit to perform excitation and collects sound using the sound collection unit to create sound collection data that records the sound volume at the observation position over time for a predetermined period of time from the excitation, The control unit includes a sound collection data analysis unit that analyzes the created sound collection data and derives the time difference between the arrival of the excitation sound and the resonant sound at the observation position, and the volume difference between the volume of the resonant sound and the reference volume, A first distance derivation unit derives a first estimated distance from the observation position to the loosened bolt based on the aforementioned time difference, A second distance derivation unit is configured to derive a second estimated distance from the observation position to the loosened bolt by inputting the sound volume difference into a trained distance estimation model that has learned the relationship between the sound volume difference and the distance from the observation position to the loosened bolt. The present invention provides a bolt loosening estimation device comprising: a third distance derivation unit that derives a third estimated distance from the observation position to the loosened bolt by weighting and combining the calculated first estimated distance and the second estimated distance at a predetermined ratio; and a notification means that notifies the inspector of the derived third estimated distance.

[0009] According to the first invention described above, the direction and distance of a loosened bolt can be estimated quickly and accurately, thereby reducing the effort required to pinpoint the location of the loosened bolt. Specifically, by collecting sound from the observation position toward one observation direction, and combining a first estimated distance based on the time difference between the excitation sound and the resonance sound with a second estimated distance based on the sound volume difference, the estimated distance in the observation direction can be accurately derived by weighting them at a predetermined ratio.

[0010] The second invention, in addition to the configuration of the first invention, The sound collection data analysis unit includes a resonance sound detection unit that determines the presence or absence of resonance sound by detecting the peak frequency band of the sound volume due to the resonance sound from the sound collection data. The resonance sound detection unit is characterized in that, when it determines that there is no resonance sound, it notifies the display unit that it is presumed that there is no loosening of the bolts in the observation direction.

[0011] The second invention provides, in addition to the effects of the first invention, If no resonance noise is generated due to vibration, the inspector can quickly obtain an estimate that there is no loosening of the bolts.

[0012] The third invention is, The vibration unit is used to vibrate a steel structure with bolt fastenings with a constant force, The sound collection unit performs the step of collecting sound in the direction of observation, The control unit causes the excitation unit to perform excitation, and the sound collection unit collects the sound, thereby creating sound collection data that records the sound volume at the observation position over time for a predetermined period of time from the excitation. The process involves analyzing the collected sound data to derive the time difference between the arrival of the excitation sound and the resonant sound at the observation position, and the volume difference between the volume of the resonant sound and the reference volume. The steps include: deriving a first estimated distance from the observation position to the loosened bolt based on the aforementioned time difference; The steps include: inputting the sound volume difference into a trained distance estimation model that has learned the relationship between the sound volume difference and the distance from the observation position to the loosened bolt, thereby deriving a second estimated distance from the observation position to the loosened bolt; The present invention provides a method for estimating bolt looseness, characterized by performing the steps of: deriving a third estimated distance from the observation position to the loosened bolt by weighting and combining the calculated first estimated distance and the second estimated distance at a predetermined ratio; and notifying the inspector of the derived third estimated distance.

[0013] According to the third invention described above, a bolt loosening estimation method can be provided that can quickly and accurately estimate the direction and distance of a loosened bolt, thereby reducing the effort required to pinpoint the location of the loosened bolt. [Effects of the Invention]

[0014] According to the present invention, a bolt loosening estimation device can be provided that can quickly and accurately estimate the direction and distance of a loosened bolt, thereby reducing the effort required to pinpoint the location of the loosened bolt. [Brief explanation of the drawing]

[0015] [Figure 1] Figure 1 is an explanatory diagram illustrating the usage state of a bolt loosening estimation device according to an embodiment of the present invention. [Figure 2] Figure 2 is a schematic diagram of the bolt loosening estimation device shown in Figure 1. [Figure 3] FIG. 3 is a block diagram centered on the control unit of the bolt loosening estimation device of FIG. 1. [Figure 4] FIG. 4 is an explanatory diagram for explaining the analysis of the sound collection data by the sound collection data analysis unit. [Figure 5] FIG. 5 is a block diagram for explaining the machine learning and input / output of the distance estimation model. [Figure 6] FIG. 6 is a flowchart showing the procedure of estimating the distance calculated by the control unit at the time of estimating the mode selection. [Figure 7] FIGS. 7(a) and 7(b) are explanatory diagrams for explaining the correlation between the volume and the distance during vibration. [Figure 8] FIG. 8 is a block diagram for explaining the machine learning and input / output of the distance estimation model according to another embodiment.

Mode for Carrying Out the Invention

[0016] <1. Method of Using Bolt Loosening Estimation Device> Hereinafter, with reference to the accompanying drawings, preferred embodiments of the present invention will be described with reference to the drawings. FIG. 1 is an explanatory diagram of the usage state of the bolt loosening estimation device 1 according to an embodiment of the present invention. First, referring to FIG. 1, the method of using the bolt loosening estimation device 1 will be described. As shown in Figure 1, the bolt loosening estimation device 1 is used by the inspector 2, who attaches the bolt loosening estimation device 1 to a predetermined observation position P on the steel structure 3 and performs a predetermined operation (for example, pressing a switch) to instruct the device 1 to start estimating whether or not there is bolt loosening and the estimated distance D. The bolt loosening estimation device 1 then vibrates the steel structure 3 and analyzes the changes in sound volume to determine whether a loose bolt 4 exists in one observation direction F determined by the orientation of the bolt loosening estimation device 1. If a loose bolt 4 exists, the estimated distance D to that loose bolt 4 (from observation position P) is calculated and displayed on the bolt loosening estimation device 1. In this embodiment, the steel structure 3 is a steel material with bolt fastenings, and more specifically, a steel tower is assumed. The observation position P is a location predetermined by the inspector 2 on a frame portion of the steel tower that extends approximately vertically near the ground. However, the observation position P is not limited to the vicinity of the ground of the steel tower. Furthermore, observation points P may be set at multiple locations on a single transmission tower (for example, at the four corners of the tower).

[0017] <2. Schematic configuration of the bolt loosening estimation device> Figure 2 is a schematic diagram of the bolt loosening estimation device 1 shown in Figure 1. As shown in Figure 2, the bolt loosening estimation device 1 comprises a housing 11, a display unit 12, an operation unit 13, a magnetic attachment unit 14, an excitation unit 15, a sound collection unit 16, a positioning unit 17, and a control unit C located inside the housing 11.

[0018] The housing 11 is a box-shaped metal cover with a display unit 12 and an operating unit 13 on its front surface, and a magnetic attachment unit 14 on the opposite back surface. The hammer body 15A of the vibration unit 15 is located on the top of the housing 11, and the hammer body drive unit 15B is located inside the housing 11.

[0019] The display unit 12 is a liquid crystal display capable of displaying various information based on the video signal transmitted from the control unit C. For example, the display unit 12 displays various setting screens, the presence or absence of loose bolts 4, and the estimated distance D. This display unit 12 functions as a notification means for informing the inspector 2 of various information.

[0020] The operation unit 13 is an operation panel equipped with various switches that receive operations from the inspector 2. Information regarding the operations received by the operation unit 13 is acquired by the control unit C, which will be described later. For example, the operation unit 13 receives a predetermined operation from the inspector 2, instructing them to check for the presence or absence of loose bolts 4 and to start estimating the estimated distance D.

[0021] The magnetic attachment part 14 serves to magnetically attach and fix the bolt loosening estimation device 1 to any position on the steel structure 3, and is composed of a flat magnet (for example, a neodymium magnet). Because the magnetic attachment part 14 allows the bolt loosening estimation device 1 to be attached to and detached from the steel structure 3 with simple operation, it is convenient and allows for quick inspection work. Furthermore, it facilitates fine adjustment of the observation position P of the bolt loosening estimation device 1.

[0022] The vibration unit 15 has the function of vibrating the steel structure 3 with a constant force, and includes a hammer body 15A that vibrates by striking the steel structure 3, and a hammer body drive unit 15B that drives and rotates the hammer body 15A. The hammer body 15A includes a hammer head 151 which is the striking body, an arm 152 to which the hammer head 151 is attached at the tip, and a pivot shaft 153 provided at the base end of the arm 152, and is configured to be rotatably mounted on the upper part of the housing 11 around the pivot shaft 153.

[0023] The hammer drive unit 15B is equipped with a rotation mechanism 155 that rotates the hammer body 15A around a pivot axis 153 toward the steel structure 3 (direction F2 in Figure 2) by the drive of an electric motor 154. This allows the hammer head 151 to strike the steel structure 3 and generate vibration. The rotation mechanism 155 can be configured using a known configuration and will not be described in detail here, but please refer to the Registered Utility Model Publication No. 3171330, for example. In the following description, the operation of the vibration unit 15 that strikes the steel structure 3 with the hammer head 151 and generates vibration will be referred to as the vibration operation.

[0024] The sound collection unit 16 is a unidirectional microphone that collects sound in one observation direction F determined by the orientation of the bolt loosening estimation device 1 and performs the function of measuring the volume (also called intensity). As a result, the control unit C, which will be described later, acquires information indicating the volume measured by the sound collection unit 16 (hereinafter referred to as volume measurement information) and generates sound collection data Ma that records the volume in the observation direction F over time at the observation position P. Details of this sound collection data Ma will be described later. It is preferable that the observation direction F is set to be approximately parallel to the longitudinal direction of the steel structure 3 and pointing upward when the bolt loosening estimation device 1 is magnetically fixed, as shown in Figure 2. Alternatively, for example, the sound collection unit 16 may be attached to the housing 11 via a ball joint, and the observation direction F with respect to the orientation of the housing 11 may be configured to be adjustable by the inspector 2.

[0025] The positioning unit 17 is a positioning device that measures the position of the bolt looseness estimation device 1. For example, it is configured to include a GNSS (Global Navigation Satellite System) antenna and can perform positioning by receiving radio waves from navigation satellites S orbiting overhead. The control unit C, described later, can record the position information of the observation position P of the bolt looseness estimation device 1 by acquiring position information from the positioning unit 17. As a result, when creating sound collection data M, the control unit C is configured to acquire position information from the positioning unit 17 and store the sound collection data M together with the position information of the observation position P where the sound collection data M was created. Furthermore, if multiple sound collection data M are created at a single observation position P, it is possible to manage the sound collection data M for each observation position P by referring to the position information associated with that sound collection data M. In addition, the convenience of data management related to inspection can be improved when using (inspecting) the bolt looseness estimation device 1 at multiple observation positions P.

[0026] Furthermore, if there are multiple observation positions P, the control unit C may be configured to create a database for each observation position P containing the collected sound data M for that observation position P, and store it in an external server 19, as described later. When inspector 2 performs inspection work at a specific observation position P, the control unit C may be configured to refer to the location information of that specific observation position P, retrieve the collected sound data M associated with the corresponding location information from the external server 19, and make it available for use (for example, by comparing and analyzing the collected sound data M from the previous work). In addition, the control unit C may be configured to allow inspector 2 to record (register) the location information of the observation position P through a predetermined operation, and when performing inspection work, it may refer to the location information of the observation position P at the time of inspection work, and output a warning if the location information is not recorded (registered). This prevents errors in the observation position P during inspection work by registering the observation position P in advance.

[0027] The communication module 18 is a transmitting and receiving device that accesses the network NW and communicates with other devices via this network NW. The control unit C, which will be described later, is connected to the network NW by this communication module 18 and is configured to send and receive various information with the external server 19 (see Figure 3).

[0028] <3-1. Configuration of the Control Unit> Figure 3 is a block diagram centered on the control unit C of the bolt loosening estimation device 1 shown in Figure 1. Control Unit C is an information processing device (computer) used to control various configurations and processes. For example, it may have a CPU or GPU, and various functions are realized when programs stored inside Control Unit C are executed using RAM or similar as the working area.

[0029] As shown in Figure 3, the control unit C has an operation unit 13, a sound collection unit 16, and a positioning unit 17 connected to its input side, and a vibration excitation unit 15 and a display unit 12 connected to its output side. The control unit C is also connected to a network NW via a communication module 18 and is configured to send and receive various information with an external server 19. The external server 19 is an example of a computer device and includes at least a processor (not shown), a storage device, and a transmitting / receiving unit. The network NW refers to a telecommunications network, such as the Internet or a LAN (local area network). Known wired or wireless communication methods can be used for communication on the network NW.

[0030] Furthermore, the control unit C includes a sound collection data creation unit 21, a sound collection data analysis unit 22, a learning unit 23, a storage unit 24, and a distance estimation unit 25. As a result, the control unit C creates and analyzes sound collection data Ma using the sound collection data creation unit 21 and the sound collection data analysis unit 22. The control unit C is also configured to switch between a learning mode, which is used for machine learning with the created sound collection data M, and an estimation mode, which is used for determining the presence or absence of loose bolts 4 and calculating the estimated distance D. In the learning mode, processing is mainly performed by the learning unit 23, and in the estimation mode, processing is mainly performed by the distance estimation unit 25. The storage unit 24 is a storage medium having a storage area capable of storing various information, and information (data) necessary for various processes, such as the sound collection data Ma, the correct distance data Mc, the distance learning data Md, and the distance estimation model 24b, which will be described later, is appropriately stored there.

[0031] The sound collection data creation unit 21 has the function of acquiring sound volume measurement information from the sound collection unit 16 and creating sound collection data Ma, and includes an excitation execution unit 21a that causes the excitation unit 15 to perform an excitation operation (in other words, to execute excitation), and a recording execution unit 21b that records the sound volume measurement information and creates sound collection data Ma. When an inspector performs a predetermined operation to instruct the operation unit 13 to start creation, the sound collection data creation unit 21 starts creating the sound collection data Ma.

[0032] When the operation unit 13 receives a predetermined operation from the inspector 2 to check for bolt looseness and to begin estimating the estimated distance D, and an instruction command is issued to the sound collection data creation unit 21 to create sound collection data Ma, the sound collection data creation unit 21 starts recording sound volume measurement information (i.e., the sound volume measured by the sound collection unit 16) using the recording execution unit 21b, and then causes the vibration unit 15 to perform an excitation operation using the vibration execution unit 21a. The recording of sound volume measurement information by the recording execution unit 21b is performed for a predetermined recording time Te (for example, 5 seconds), during which the measured sound volume is recorded over time. In other words, how the sound volume changes over time before and after the excitation operation by the vibration unit 15 is recorded. When the recording is completed after the predetermined recording time has elapsed, the recorded sound volume measurement information is stored as sound collection data Ma in the various data storage units 24a of the storage unit 24.

[0033] The sound collection data Ma created in this way records the volume of the excitation sound generated when the excitation unit 15 is excited by the excitation execution unit 21a, and the volume of the resonance sound, which is the sound generated when the loose bolts 4 resonate with the vibration of the steel structure 3 due to the excitation, if there are loose bolts.

[0034] <3-2. Analysis of collected sound data> The sound collection data analysis unit 22 performs the function of analyzing the sound collection data M created by the sound collection data creation unit 21, and includes a resonance sound detection unit 22a that determines the presence or absence of resonance sound from the change in volume contained in the sound collection data, a time difference calculation unit 22b that calculates the time difference between when the excitation sound and the resonance sound arrive at the observation position P, and a volume difference calculation unit 22c that calculates the volume difference between the volume of the resonance sound at the observation position P and the reference volume.

[0035] Here, with reference to Figure 4, we will explain the analysis of the sound collection data Ma by the sound collection data analysis unit 22. Figure 4 is an explanatory diagram illustrating the analysis of the sound-collected data Ma by the sound-collected data analysis unit 22, and visually shows the data content of the sound-collected data Ma. Specifically, the vertical axis in Figure 4 represents volume (v), and the horizontal axis represents time (t). The waveform W shows the change in volume (v) over elapsed time (t) as measured and recorded by the sound-collecting unit 16.

[0036] The resonance sound detection unit 22a refers to the created sound collection data Ma and detects the peak frequency band E11 of the volume due to the excitation sound from the recorded volume changes, and then determines the presence or absence of resonance sound by detecting the peak frequency band E12 of the volume due to the resonance sound from the recorded volume changes. Here, the method for detecting the peak frequency bands E11 and E12 of the volume of the excitation sound and resonance sound may be configured such that when the volume first exceeds a predetermined threshold, the peak frequency band E11 of the volume due to the excitation sound is detected, and then when the volume first exceeds a predetermined threshold (a value greater than the normally observed volume, i.e., the volume of ambient sound, is set), the peak frequency band E12 of the volume of the resonance sound is detected. In this case, the range of peak frequency bands E11 and E12 is the range in which the volume exceeds the predetermined threshold.

[0037] Furthermore, the resonance sound detection unit 22a may detect the peak frequency band E11 of the excitation sound volume and the peak frequency band E12 of the resonance sound volume by pattern matching processing of the waveform W. Alternatively, the peak frequency band E12 of the resonance sound volume may be detected by a learning model that has learned the waveforms when resonance sound occurs and when it does not. In this way, when the resonance sound detection unit 22a detects the peak frequency band E12 of the resonance sound volume, it determines that there is resonance sound, and when it does not detect the peak frequency band E12 of the resonance sound volume, it determines that there is no resonance sound.

[0038] Next, when the resonance sound detection unit 22a determines that a resonance sound is present, the sound collection data analysis unit 22 calculates the time difference between the arrival of the excitation sound and the resonance sound at the observation position P using the time difference calculation unit 22b. Specifically, it calculates the time difference Td(t2-t1) (seconds) between the time t1 when the maximum sound volume was measured in the peak band E11 of the sound volume due to the excitation sound and the time t2 when the maximum sound volume was measured in the peak band E12 of the sound volume due to the resonance sound. Information regarding the calculated time difference Td is appropriately stored in the various data storage units 24a of the storage unit 24.

[0039] Furthermore, when the sound collection data analysis unit 22 determines that a resonance sound is present using the resonance sound detection unit 22a, the volume difference calculation unit 22c calculates the volume difference between the volume of the resonance sound at the observation position P and the reference volume. Here, the reference volume can be, for example, the maximum volume v1 of the peak band E11 of the volume caused by the resonance sound. Alternatively, a predetermined volume set in advance can be used as the reference volume. In this embodiment, the case in which the maximum volume v1 of the peak band E11 of the volume caused by the resonance sound is used as the reference volume will be described. Specifically, as shown in Figure 4, the volume difference Vd (=v1-v2) is calculated by calculating the difference between the maximum volume v1 of the peak band E11 of the volume caused by the resonance sound and the maximum volume v2 of the peak band E12 of the volume caused by the resonance sound. Information regarding the calculated volume difference Vd is appropriately stored in the various data storage units 24a of the storage unit 24.

[0040] <3-3. Configuration of the Learning Unit and Distance Estimation Model> Returning to Figure 3, the learning unit 23 performs the function of executing machine learning and includes a learning data creation unit 23a that creates distance learning data and a machine learning execution unit 23b that generates a trained distance estimation model 24b by executing machine learning using the distance learning data. The distance estimation model 24b is composed of a neural network capable of machine learning using deep learning and has an input layer, a hidden layer, and an output layer (not shown). Next, the machine learning and input / output of the distance estimation model 24b will be explained below with reference to Figure 5.

[0041] Figure 5 is a block diagram illustrating the machine learning and input / output of the distance estimation model 24b. The learning unit 23 trains the distance estimation model 24b using supervised learning methods. Therefore, the learning data creation unit 23a obtains the input data and correct labels during machine learning and creates distance learning data Md by pairing these. Specifically, the inspector 2 prepares loose bolts 4 in the steel structure 3 in advance for data creation for machine learning, selects the learning mode through a predetermined operation, and then uses the sound collection data creation unit 21 to perform an excitation operation on the excitation unit 15, thereby performing the procedure for creating sound collection data Ma. The sound collection data Ma created in this way is then acquired by the sound collection data analysis unit 22 (step #101). The sound collection data analysis unit 22 analyzes the sound collection data Ma and calculates the volume difference Vd (step #102), and passes the volume difference data Mb, which shows the calculated volume difference Vd, to the learning data creation unit 23a (step #103).

[0042] At this time, the learning data creation unit 23a obtains the correct distance data Mc, which is a pre-prepared correct label and indicates the correct distance from the observation position P to the loosened bolt 4 (step #104). The correct distance data Mc may be stored in the various data storage units 24 of the storage unit 24 in advance by the inspector and retrieved from there, or the correct distance data Mc may be stored in the external server 19 in advance and retrieved from the external server 19. As a result, the learning data creation unit 23a creates distance learning data Md, which is a pair of sound volume difference data Mb and the correct distance data Mc (step #105), and passes it to the machine learning execution unit 23b (step #106).

[0043] The machine learning execution unit 23b inputs the acquired distance learning data Md into the distance estimation model 24b and performs machine learning (step #107). The distance estimation model 24b, trained using a large amount of distance learning data Md, is stored in the memory unit 24. As a result, when the estimation mode is selected, the distance estimation model 24b is configured to receive volume difference data Mb calculated by analysis from the sound collection data Ma (step #108) and output estimated distance data Me indicating the estimated distance from the observation position P to the loosened bolt 4 (step #109).

[0044] <4-1. Procedure for calculating the estimated distance by the control unit> Returning to Figure 3, when an estimation mode is selected, the distance estimation unit 25 calculates the estimated distance D from the observation position P in the observation direction F to the loosened bolt when the sound collection data analysis unit 22 (more specifically, the resonance sound detection unit 22a) determines that a resonance sound is present. The distance estimation unit 25 includes a first distance calculation unit 251 that calculates the first estimated distance, a second distance calculation unit 252 that calculates the second estimated distance, and a third distance calculation unit 253 that calculates the third estimated distance based on the calculated first and second estimated distances. Next, the procedure for calculating the estimated distance D by the control unit C when an estimation mode is selected will be explained with reference to Figure 6.

[0045] Figure 6 is a flowchart showing the procedure for calculating the estimated distance by the control unit C when an estimation mode is selected. As a prerequisite for starting the procedure shown in Figure 6, it is assumed that the bolt loosening estimation device 1 is attached to the observation position P of the steel structure 3, the estimation mode is selected for the bolt loosening estimation device 1, the learned distance estimation model 24b is stored, and the inspector 2 has performed a predetermined operation (for example, pressing the switch on the operation unit 13) to start the process related to the calculation of the estimated distance.

[0046] Upon the inspection operator's (2) predetermined operation, the sound collection data creation unit 21 causes the excitation unit 15 to perform an excitation operation (step #1) and creates sound collection data Ma (step #2). Subsequently, the sound collection data analysis unit 22 analyzes the created sound collection data Ma (step #3), and the resonance sound detection unit 22a determines that there is resonance sound if it detects a peak frequency band E12 due to the resonance sound by analyzing the sound collection data Ma, and determines that there is no resonance sound if it does not detect a peak frequency band E12 due to the resonance sound (step #4).

[0047] If it is determined that there is no resonance sound (N in step #4), the control unit C (resonance sound detection unit 22a of the sound collection data analysis unit 22) displays the estimation result on the display unit 12 indicating that it is estimated that there is no loosening of the bolts in the observation direction F (step #5). For example, it displays "No detection". This allows the inspector 2 to quickly know the estimation result that there is no loosening of the bolts if no resonance sound is generated due to vibration.

[0048] On the other hand, when it is determined that there is a resonant sound (Y in step #4), the distance estimation unit 25 calculates a first estimated distance d1 using the first distance calculation unit 251 (step #6), calculates a second estimated distance d2 using the second distance calculation unit 252 (step #7), and calculates a third estimated distance d3 using the third distance calculation unit 253 based on the calculated first and second estimated distances (step #8). Then, the control unit C notifies the inspector 2 by displaying the calculated third estimated distance as the final estimated distance D on the display unit 12 (step #9). The calculation methods for the first to third estimated distances will be explained below in order.

[0049] <4-2. Method for Calculating the First Estimated Distance> The first distance calculation unit 251 includes a time difference distance conversion unit 251a that converts the time difference Td calculated by the time difference calculation unit 22b of the sound collection data analysis unit 22 into distance (see Figure 3). The time difference distance conversion unit 251a acquires the time difference Td information calculated by the time difference calculation unit 22b, multiplies the acquired time difference Td by the speed of sound 340 (meters / second), and calculates the first estimated distance d1. According to this calculation method, the estimated distance from the observation position P to the loosened bolt can be obtained by utilizing the time difference between the generation of excitation sound and resonance sound. Note that the transmission speed of vibrations in the steel structure 3 is overwhelmingly faster than the transmission speed of air vibrations (speed of sound) and is therefore ignored.

[0050] <4-3. Method for Calculating the Second Estimated Distance> The second distance calculation unit 252 includes a volume difference distance conversion unit 252a that converts the volume difference Vd calculated by the volume difference calculation unit 22c of the sound collection data analysis unit 22 into a distance (see FIG. 3). The volume difference distance conversion unit 252a acquires information on the volume difference Vd (volume difference data Mb) calculated by the volume difference calculation unit 22c, inputs the acquired information on the volume difference Vd into the distance estimation model 24b, and obtains an output of estimated distance data Me indicating the estimated distance from the observation position P to the loose bolt. The second distance calculation unit 252 sets the estimated distance indicated by the estimated distance data Me calculated based on the output of the distance estimation model 24b as the second estimated distance d2.

[0051] Here, referring to FIGS. 7(a) and 7(b), the correlation between the volume difference and the distance will be described. FIGS. 7(a) and 7(b) are explanatory diagrams for explaining the correlation between the volume and the distance during vibration. More specifically, FIG. 7(a) shows the time elapsed (t) and the volume (v) during vibration at the observation position P when there is a loose bolt 4 at a specific point A, and FIG. 7(b) shows an example of the relationship between the time elapsed (t) and the volume (v) during vibration at the observation position P when there is a loose bolt 4 at a point B closer to the observation position P than the specific point A.

[0052] Comparing the graphs of FIGS. 7(a) and 7(b), the peak bands E21 and E31 of the volume due to the vibration sound have the same waveforms W1 and W2, but the peak bands E22 and E32 of the volume due to the resonance sound appear earlier (t22 < t21) in the peak band E32 of the volume due to the resonance sound in FIG. 7(b) than in the peak band E22 of the volume due to the resonance sound in FIG. 7(a) because the point B is closer to the observation position P than the point A. Here, since the resonance sound attenuates in proportion to the distance from the observation position P, which is the vibration location, the volume difference Vd1 in FIG. 7(a) is larger than the volume difference Vd2 in FIG. 7(b), and the volume difference Vd increases as the loose bolt 4 moves away from the observation position P. Thus, there is a correlation between the distance from the observation position P, which is the vibration location, to the loose bolt 4 and the volume difference Vd. Therefore, by inputting information on the volume difference Vd into the distance estimation model 24b, the estimated distance to the loose bolt 4 can be suitably obtained.

[0053] <4-4. Method for Calculating the Third Estimated Distance> The third distance calculation unit 253 includes a weighting processing unit 253a that calculates a third estimated distance d3 by weighting the first estimated distance d1 and the second estimated distance d2, and an estimated distance output unit 253b that displays the third estimated distance d3 calculated by the weighting processing unit 253a as the estimated distance D on the display unit 12.

[0054] The weighting processing unit 253a obtains information on the calculated first estimated distance d1 and second estimated distance d2, and calculates a third estimated distance d3, for example, using the following equation (1).

[0055] The third estimated distance d3 = α × the first estimated distance d1 + β × the second estimated distance d2 ... (1) Note that α and β are weighting coefficients, and α + β = 1. The values ​​of α and β may be configured to be settable by the inspector. For example, in areas with extreme temperature fluctuations, the setting value of α can be made relatively smaller to lower the evaluation based on the time difference Td and to overestimate the evaluation based on the sound difference Vd in order to calculate the estimated distance. The third estimated distance d3 calculated in this way is displayed on the display unit 12 by the estimated distance output unit 253b as the estimated distance D to the loosened bolt 4 in the observation direction F of the bolt loosening estimation device 1. This allows the inspector 2 to know the estimated distance D. In this way, the bolt loosening estimation device 1 according to the present invention can quickly and accurately estimate the distance to the loosened bolt 4 located in the observation direction F by calculating the estimated distance using information from the time difference Td and information from the sound difference Vd.

[0056] As described above, according to the bolt loosening estimation device 1 and bolt loosening estimation method of the present invention, the bolt loosening estimation device 1 collects sound from the observation position P toward one observation direction F, and accurately calculates the estimated distance D in the observation direction F by weighting the first estimated distance d1 based on the time difference Td between the excitation sound and the resonance sound and the second estimated distance d2 based on the sound volume difference Vd at a predetermined ratio and summing them up. As a result, the direction and distance of the loosened bolt 4 can be estimated quickly and accurately, thereby reducing the effort required to identify the location of the loosened bolt 4.

[0057] The embodiments of the present invention have been described above. The present invention is not limited to the embodiments described above. It goes without saying that modifications can be made as appropriate within the scope of the technical idea. Another embodiment will be described below.

[0058] <5. Regarding another embodiment> Figure 8 is a block diagram illustrating the machine learning and input / output of a distance estimation model according to another embodiment. In the embodiment shown in Figure 5, an example is shown in which the volume difference data Mb is input to the learning data creation unit 23a in learning mode and to the distance estimation model 24b in estimation mode. In contrast, in another embodiment shown in Figure 8, the time difference data Mt and the volume difference data Mb are input to the learning data creation unit 23a (steps #201 to #205), and these are paired with the ground truth distance data Mc to create distance learning data Md (steps #206 to #208). This is then used to train the distance estimation model 24b using the machine learning execution unit 23b (step #209). In estimation mode, the time difference data Mt and the volume difference data Mb are input to the trained distance estimation model 24b (steps #210, #211), and the output of distance estimation data Me is obtained (step #212).

[0059] In this configuration, instead of steps #6 to #8 in the flowchart of Figure 6, the time difference data Mt and sound difference data Mb are input to the trained distance estimation model 24b, the output of distance estimation data Me is obtained, and the estimated distance indicated by the distance estimation data Me is displayed as the final estimated distance D (step #9 in Figure 6). This makes it possible to obtain distance estimation data Me based on the time difference Td and sound difference Vd, and to calculate the estimated distance D with greater accuracy.

[0060] Furthermore, as a variation of the above configuration, the distance learning data Md may include not only time difference data Mt and sound difference data Mb, but also information on temperature (air temperature) and sunlight. The bolt loosening estimation device 1 may also be equipped with a thermometer to measure temperature (air temperature) and an illuminometer to measure sunlight. By training the distance estimation model 24b with information on temperature (air temperature) and sunlight, and then inputting this information into the trained distance estimation model 24b to obtain distance estimation data Me, it is possible to obtain distance estimation data Me that takes into account changes in the speed of sound due to temperature (air temperature) and sunlight, as well as the thermal expansion of the steel structure 3, and to calculate the estimated distance D with even greater accuracy.

[0061] Furthermore, although the above embodiment shows a configuration in which the control unit C comprises a learning unit 23 and a distance estimation model 24b, these may also be configured to be placed on an external server 19 by sending and receiving necessary data with the external server 19. This allows for a simpler configuration of the bolt loosening estimation device 1.

[0062] Alternatively, when the learning unit 23 learns the volume difference data Mb, it may be configured to learn the volume generated at the loosened bolt 4 as ground truth data, thereby estimating the volume of the resonant sound generated at the location of the loosened bolt 4, and using the estimated volume of the resonant sound as a reference volume, the volume difference calculation unit 22c may calculate the volume difference between that volume and the volume of the resonant sound at the observation position P. [Explanation of Symbols]

[0063] 1. Looseness estimation device 2. Inspector 3. Steel structures (transmission towers) 4. Loose bolts 11 cabinets 12 Display section 13 Control section 14 Magnetized part 15 Vibration section 15A Hammer Body 15B Hammer body drive unit 16 Sound collection section 17 Positioning Unit 18 Communication Module 19 External Servers 151 Hammerhead 152 Arm 153 Rotary shaft 154 Electric motor 155 Rotating Mechanism C control section D Estimated distance E11, E21, E31 Peak frequency band (excitation sound) E12, E22, E32 Peak frequency band (resonant sound) F Observation direction Te Recording time Td time difference P Observation location NW Network Vd volume difference Waveforms showing the changes in volume levels for W, W1, and W2.

Claims

1. A vibration unit that applies a constant force to a steel structure having bolt fastenings, A sound collection unit that collects sound from the observation position toward the observation direction, The system includes a control unit that causes the excitation unit to perform excitation and collects sound using the sound collection unit to create sound collection data that records the sound volume at the observation position over time for a predetermined period of time from the excitation, The control unit includes a sound collection data analysis unit that analyzes the created sound collection data and derives the time difference between the arrival of the excitation sound and the resonant sound at the observation position, and the volume difference between the volume of the resonant sound and the reference volume, A first distance derivation unit derives a first estimated distance from the observation position to the loosened bolt based on the aforementioned time difference, A second distance derivation unit is configured to derive a second estimated distance from the observation position to the loosened bolt by inputting the sound volume difference into a trained distance estimation model that has learned the relationship between the sound volume difference and the distance from the observation position to the loosened bolt. A bolt loosening estimation device comprising: a third distance derivation unit that derives a third estimated distance from the observation position to the loosened bolt by weighting and combining the calculated first estimated distance and the second estimated distance at a predetermined ratio; and a notification means that notifies the inspector of the derived third estimated distance.

2. The sound collection data analysis unit includes a resonance sound detection unit that determines the presence or absence of resonance sound by detecting the peak frequency band of the sound volume due to the resonance sound from the sound collection data. The bolt loosening estimation device according to claim 1, characterized in that when the resonance sound detection unit determines that there is no resonance sound, the notification means notifies that it is estimated that there is no bolt loosening in the observation direction.

3. The vibration unit is used to vibrate a steel structure with bolt fastenings with a constant force, The sound collection unit performs the step of collecting sound in the direction of observation, The control unit causes the excitation unit to perform excitation, and the sound collection unit collects the sound, thereby creating sound collection data that records the sound volume at the observation position over time for a predetermined period of time from the excitation. The process involves analyzing the collected sound data to derive the time difference between the arrival of the excitation sound and the resonant sound at the observation position, and the volume difference between the volume of the resonant sound and the reference volume. The steps include: deriving a first estimated distance from the observation position to the loosened bolt based on the aforementioned time difference; The steps include: inputting the sound volume difference into a trained distance estimation model that has learned the relationship between the sound volume difference and the distance from the observation position to the loosened bolt, thereby deriving a second estimated distance from the observation position to the loosened bolt; A method for estimating bolt looseness, characterized by performing the steps of: deriving a third estimated distance from the observation position to the loosened bolt by weighting and combining the calculated first estimated distance and the second estimated distance at a predetermined ratio; and notifying the inspector of the derived third estimated distance.

Images