Diagnostic device and diagnostic method

Abstract

The signal conversion unit (short-time Fourier transform unit and spectral flux transform unit) (42, 43) calculates an intensity signal indicating the time change of spectral intensity by time-frequency conversion at least for each time width of the acoustic signal based on the acoustic signal indicating the time change of the operation sound of the diagnosis object and one or more reinforcement signals, and generates a converted signal (spectral flux signal) by summing the intensity signals with respect to time or frequency. The one or more reinforcement signals are signals detected from the diagnosis object simultaneously with the acoustic signal according to a detection principle different from that of the acoustic signal. The feature extraction unit (44) extracts one or more feature values ​​from the converted signal. The diagnosis unit (46) diagnoses the diagnosis object based on the one or more feature values.

Description

[Technical Field]

[0001] The present disclosure relates to diagnostic devices and methods. [Background technology]

[0002] The on-load tap changer monitoring device disclosed in Japanese Patent Laid-Open Publication No. 2-19776 (Patent Document 1) comprises a sensor means, a data collection unit, a frequency analysis unit, and a signal comparison unit. The sensor means is provided on the tank wall of the on-load tap changer of the transformer or near the tank, and detects vibrations and acoustic signals generated during tap changing. The data collection unit samples the signals detected by the sensor means at regular intervals and digitizes them. The frequency analysis unit performs frequency analysis on the digital data collected by the data collection unit. The signal comparison unit compares the frequency components obtained by the frequency analysis unit with frequency components during normal operation and determines whether there is an abnormality in the on-load tap changer. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Unexamined Patent Publication No. 2-19776 Summary of the Invention [Problem to be solved by the invention]

[0004] Various diagnostic devices have been proposed that diagnose the condition of a diagnostic object by analyzing signals detected from the diagnostic object. There is a constant demand for improving the diagnostic accuracy of these diagnostic devices. Various abnormalities can occur in the diagnostic object. It is desirable to diagnose the condition of the diagnostic object using appropriate signals.

[0005] The present disclosure has been made to solve the above problems, and one of the objects of the present disclosure is to improve the diagnostic accuracy of a diagnostic subject in a diagnostic device. Another object of the present disclosure is to improve the diagnostic accuracy of a diagnostic subject in a diagnostic method. [Means for solving the problem]

[0006] A diagnostic device according to an aspect of the present disclosure includes a signal conversion unit, a feature extraction unit, and a diagnosis unit. The signal conversion unit calculates an intensity signal indicating a time change in spectral intensity by performing a time-frequency transformation on at least each time width of the acoustic signal based on an acoustic signal indicating a time change in the operation sound of the diagnostic object and one or more reinforcement signals, and generates a converted signal by summing the intensity signals with respect to time or frequency. The one or more reinforcement signals are signals that are detected from the diagnostic object simultaneously with the acoustic signal according to a detection principle different from that of the acoustic signal. The feature extraction unit extracts one or more feature values ​​from the converted signal. The diagnosis unit diagnoses the diagnostic object based on the one or more feature values.

[0007] A diagnostic method according to another aspect of the present disclosure includes first to third steps. The first step is a step of calculating an intensity signal indicating a time change in spectral intensity by performing a time-frequency transform on at least each time width of the acoustic signal based on an acoustic signal indicating a time change in the operation sound of the diagnostic object and one or more reinforcement signals, and generating a converted signal by summing the intensity signals over time or frequency. The one or more reinforcement signals are signals that are detected from the diagnostic object simultaneously with the acoustic signal according to a detection principle different from that of the acoustic signal. The second step is a step of extracting one or more feature amounts from the converted signal. The third step is a step of diagnosing the diagnostic object based on the one or more feature amounts. [Effects of the Invention]

[0008] According to the present disclosure, it is possible to improve the accuracy of diagnosis of a diagnostic subject. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is a diagram showing the overall configuration of a diagnostic system according to an embodiment; [Figure 2] FIG. 2 is a functional block diagram for explaining an outline of the functions of the diagnostic device. [Figure 3]1 is a diagram showing an overall configuration of a switching device diagnostic system according to a first embodiment. [Figure 4] FIG. 2 is a functional block diagram for explaining the function of a signal processing unit in the first embodiment. [Figure 5] FIG. 2 is a conceptual diagram for explaining signal processing by a signal processing unit. [Figure 6] FIG. 10 is a diagram showing an example of an actual waveform. [Figure 7] FIG. 1 shows a first example of changes detected in a spectral flux signal. [Figure 8] FIG. 10 shows a second example of changes detected in a spectral flux signal. [Figure 9] FIG. 2 is a diagram for explaining a first example of a feature amount. [Figure 10] FIG. 10 is a diagram for explaining a second example of feature amounts. [Figure 11] FIG. 10 is a diagram for explaining a third example of a feature amount. [Figure 12] FIG. 10 is a diagram for explaining a fourth example of feature amounts. [Figure 13] 1 is a flowchart showing a first example of a processing procedure of a process (frequency division) executed in the first embodiment. [Figure 14] 10 is a flowchart illustrating an example of a processing procedure for extracting feature amounts. [Figure 15] 10 is a flowchart illustrating an example of a procedure for diagnosing an abnormality. [Figure 16] FIG. 10 is a diagram for explaining a method for setting weighting coefficients. [Figure 17] FIG. 1 is a conceptual diagram comparing a normal spectral flux transform and a weighted spectral flux transform. [Figure 18] FIG. 1 is a conceptual diagram for explaining time division. [Figure 19] 10 is a flowchart showing a second example of the processing procedure of the processing (time division) executed in the first embodiment. [Figure 20] 10 is a flowchart showing a third example of the processing procedure of the processing (continuous operation) executed in the first embodiment. [Figure 21] 10 is a flowchart showing an example of a processing procedure for lifetime estimation according to the first embodiment. [Figure 22] FIG. 10 is a diagram for explaining a lifetime estimation method. [Figure 23] FIG. 10 is a diagram for explaining processing that can be combined with frequency division. [Figure 24] FIG. 10 is a functional block diagram for explaining the function of a signal processing unit in the second embodiment. [Figure 25] 1A to 1C are conceptual diagrams illustrating a first example of a method for generating a geometric figure and a method for extracting a feature amount. [Figure 26] 10 is a flowchart showing a first example of a processing procedure of a process (geometric processing) executed in the second embodiment. [Figure 27] FIG. 10 is a conceptual diagram for explaining a second alternative example of a method for generating a geometric figure and a method for extracting a feature amount. [Figure 28] 10 is a flowchart showing a second example of the processing procedure of the process (comparison process) executed in the second embodiment. [Figure 29] FIG. 10 is a diagram for explaining a process that can be combined with the geometric process or the comparison process. [Figure 30] FIG. 10 is a diagram showing the overall configuration of a switching device diagnostic device according to a third embodiment. [Figure 31] FIG. 11 is a functional block diagram for explaining the function of a signal processing unit in the third embodiment. [Figure 32] 11 is a flowchart showing an example of a processing procedure of a process (reinforcement process) executed in the third embodiment. [Figure 33] FIG. 10 is a diagram for explaining a process that can be combined with the reinforcement process. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. In the drawings, the same or corresponding parts are designated by the same reference numerals, and description thereof will not be repeated.

[0011] <Common configuration> 1 is a diagram showing the overall configuration of a diagnostic system according to an embodiment. The diagnostic system 100 includes a diagnostic device 1, a diagnostic target 7, and a sensor 8.

[0012] The diagnosis object 7 is typically an electrical device. As a specific example, the diagnosis object 7 may be a switchgear housed inside a power receiving and distribution facility (switchgear), as will be described later. The diagnosis object 7 may be a voltage transformer (VT), a current transformer (CT), or a fan. The type of the diagnosis object 7 is not particularly limited, and the diagnosis object 7 may be an optical device (e.g., a light-emitting element, a light-receiving element), or a mechanical mechanism (e.g., a gear, a chain, an actuator, an engine). The diagnosis object 7 may also be a living organism such as a worker working in a factory or a wild animal invading a factory.

[0013] The sensor 8 detects the state of the diagnosis object 7 (typically, a state change due to a failure, abnormality, or defect of the diagnosis object 7). The sensor 8 may be a contact sensor arranged to come into contact with the diagnosis object 7, or a non-contact sensor arranged not to come into contact with the diagnosis object 7 (but in the vicinity of the diagnosis object 7). The sensor 8 may include, for example, an acoustic sensor, a vibration sensor (including a strain sensor), a voltage sensor, a current sensor, an optical sensor (including an image sensor), a temperature sensor, a pressure sensor, a distance sensor, a speed sensor, and an acceleration sensor.

[0014] The diagnostic device 1 diagnoses the diagnostic object 7 based on a signal (sensor signal) from a sensor 8 provided in the diagnostic object 7. The diagnostic device 1 may perform diagnosis based on the sensor signal itself (raw signal), or may perform diagnosis based on a processed sensor signal. The sensor signal or the processed sensor signal corresponds to the "raw signal" according to the present disclosure.

[0015] The diagnostic device 1 includes a processor 101, a memory 102, and a communication device 103. The processor 101 includes processing circuitry such as a CPU (Central Processing Unit) and an MPU (Micro Processing Unit). The memory 102 includes volatile storage devices such as DRAM (Dynamic Random Access Memory) and SRAM (Static Random Access Memory), and non-volatile storage devices such as HDD (Hard Disk Drive), SSD (Solid State Drive), and flash memory. The memory 102 stores a system program including an OS (Operating System), a control program including computer-readable code, and various parameters for diagnosing the diagnostic object 7. The processor 101 performs various arithmetic operations by reading the system program, the control program, and the parameters, expanding them into the memory 102, and executing them. The communication device 103 is configured to communicate with the outside of the diagnostic device 1 under the control of the processor 101.

[0016] Although only one processor is illustrated in FIG. 1, the diagnostic device 1 may include multiple processors. That is, the diagnostic device 1 includes one or more processors. The same applies to the memory 102. In this specification, the term "processor" is not limited to a processor in the narrow sense that executes processing using a stored program, but may also include hardwired circuits such as an ASIC (Application Specific Integrated Circuit) and an FPGA (Field-Programmable Gate Array). Therefore, the term "processor" can also be interpreted as a processing circuit (circuitry or processing circuitry) whose processing is predefined by computer-readable code and / or hardwired circuitry.

[0017] The diagnostic system 100 further includes a server 91 and a terminal 92. The server 91 includes a database that stores the diagnostic results obtained by the diagnostic device 1. The server 91 may store data (so-called logs) that indicate the progress of signal processing performed by the diagnostic device 1. The terminal 92 includes a display that displays the diagnostic results obtained by the diagnostic device 1. The terminal 92 may include an alarm (warning light, alarm, etc.) that notifies the user of any abnormality in the diagnostic object 7 that has been discovered as a result of the diagnosis performed by the diagnostic device 1. The alarm may be provided in the diagnostic device 1.

[0018] Fig. 2 is a functional block diagram for explaining an outline of the functions of diagnostic device 1. With reference to Figs. 1 and 2, diagnostic device 1 includes an input unit 11, a signal processing unit 12, and an output unit 13. Input unit 11 receives a sensor signal from sensor 8. Signal processing unit 12 performs signal processing on the sensor signal from input unit 11 in order to diagnose diagnostic object 7. Output unit 13 outputs the signal processing result (diagnosis result) by signal processing unit 12 to a server 91, a terminal 92, or the like external to diagnostic device 1.

[0019] More specifically, the signal processing unit 12 includes a filter 121, a time-frequency transform unit 122, an intensity transform unit 123, a feature extraction unit 124, and a diagnosis unit 125. Fig. 2 shows an example in which the signal processing unit 12 includes one functional block of the same type. However, as will be described later, the signal processing unit 12 may include two or more functional blocks of the same type.

[0020] The filter 121 passes signal components in a certain frequency band of the sensor signal from the input unit 11 and outputs the signal components to the time-frequency conversion unit 122. The filter 121 may be an analog filter or a digital filter.

[0021] The time-frequency transform unit 122 performs time-frequency transform on the signal that has passed through the filter 121 for each predetermined time width, thereby calculating a frequency-domain signal for each time width. The frequency-domain signal for each time width is represented by a three-dimensional waveform of time, frequency, and intensity. The time-frequency transform may be a short-time Fourier transform (STFT), a continuous wavelet transform (CWT), a discrete wavelet transform (DWT), a Stockwell transform, or the like. The time-frequency transform unit 122 outputs the frequency-domain signal for each time width to the intensity transform unit 123.

[0022] The intensity conversion unit 123 converts the frequency domain signal (three-dimensional waveform of time-frequency-intensity) for each time width into a time domain signal (two-dimensional waveform with the horizontal axis being the time axis) indicating spectral fluctuations by summing the intensities for each frequency. More specifically, the intensity conversion unit 123 generates a time domain signal indicating spectral fluctuations (intensity across the entire frequency band) by summing the intensities for all frequencies for each time width. The intensity conversion unit 123 outputs the time domain signal indicating spectral fluctuations to the feature extraction unit 124.

[0023] The intensity conversion unit 123 may convert the intensity in the frequency domain signal (three-dimensional waveform of time-frequency-intensity) for each time width into a frequency domain signal (two-dimensional waveform with the horizontal axis as the frequency axis) that indicates the time variation of the intensity by summing the intensity in the frequency domain signal for each specified time length.

[0024] The feature extraction unit 124 extracts one or more feature amounts based on the time domain signal indicating the spectral variation, and outputs the extracted one or more feature amounts to the diagnosis unit 125.

[0025] The diagnosis unit 125 diagnoses the diagnostic object 7 based on one or more feature amounts extracted by the feature amount extraction unit 124. The diagnosis unit 125 outputs the diagnosis result to the output unit 13.

[0026] The output unit 13 outputs the diagnosis result to the server 91 and the terminal 92. The output unit 13 may store the diagnosis result in the memory 102. Although not shown, data indicating the progress of the signal processing by the signal processing unit 12 (such as a frequency domain signal for each time width, a time domain signal indicating a spectral fluctuation, etc.) may be stored in the memory 102 or the server 91.

[0027] In the following, in order to facilitate understanding of the signal processing by the signal processing unit 12, a specific example will be described in which the object to be diagnosed 7 is a switching device and the sensor 8 is an acoustic sensor.

[0028] Embodiment 1 <System configuration> <Overall structure> Fig. 3 is a diagram showing the overall configuration of the switching device diagnostic system according to embodiment 1. Note that in Fig. 3, in order to prevent the page from becoming too cluttered, the server 91 and the terminal 92 (see Fig. 1) are not shown.

[0029] The switchgear 71 includes a circuit breaker 711 and a motor 712. The circuit breaker 711 is electrically connected to a power supply path to a power system (not shown). The motor 712 is an electromagnetic motor configured to open and close the circuit breaker 711 in accordance with a control command from the diagnostic device 1A. The operation of the circuit breaker 711 transitioning from a closed state to an open state is the opening operation of the switchgear 71, and the operation of the circuit breaker 711 transitioning from an open state to a closed state is the closing operation of the switchgear 71.

[0030] The circuit breaker 711 is provided with an acoustic sensor 81. The acoustic sensor 81 includes, for example, a microphone, and detects the operating sound (such as the contact sound between components) of the switchgear 71. The acoustic sensor 81 outputs an acoustic signal indicating the detection result to the diagnostic device 1A.

[0031] The diagnostic device 1A includes an input unit 11, a signal processing unit 12A, an output unit 13, and a control unit 14. The input unit 11 receives an acoustic signal from an acoustic sensor 81. The signal processing unit 12A performs signal processing on the acoustic signal to diagnose the switching device 71. The output unit 13 outputs the diagnosis result by the signal processing unit 12A to at least one of a server 91 and a terminal 92 (see FIG. 1). The output unit 13 may output the diagnosis result or an intermediate progress to a storage unit (memory 102) not shown. The control unit 14 outputs a control command to cause the switching device 71 to perform an opening operation or a closing operation.

[0032] <Function Block> Fig. 4 is a functional block diagram for explaining the function of the signal processing unit 12A in the first embodiment. Fig. 5 is a conceptual diagram for explaining signal processing by the signal processing unit 12A. The following explanation will be given taking processing on an acoustic signal as an example, but similar processing can also be performed on other types of signals (such as vibration signals). Fig. 6 is a diagram showing an example of an actual waveform.

[0033] 3 to 6, the signal processing unit 12A includes a first filter 211, a second filter 221, a first short-time Fourier transform unit 212, a second short-time Fourier transform unit 222, a first Spectral Flux (SF) transform unit 213, a second SF transform unit 223, a feature extraction unit 24, a memory unit 25, and a diagnosis unit 26.

[0034] Each of the first filter 211 and the second filter 221 receives an acoustic signal from the acoustic sensor 81. An example of the waveform of the acoustic signal is shown in the upper part of FIG. 6. The first filter 211 is a band-pass filter that passes only signal components in a certain frequency band of the acoustic signal. The second filter 221 is a band-pass filter that passes only components in a frequency band different from that of the first filter 211. In this example, it is assumed that the first filter 211 passes low-frequency band components, and the second filter 221 passes high-frequency band components. The first filter 211 and the second filter 221 correspond to the "filter" according to the present disclosure.

[0035] The first short-time Fourier transform unit 212 calculates a signal indicating the spectral intensity for each predetermined time width (hereinafter also referred to as an “intensity signal”) according to the following equation (1), which is a short-time Fourier transform of the low-frequency band components.

number

[0036] The acoustic signal is represented by x, the window function by w, the intensity by I1, the time by t (and τ), and the frequency by f. The waveform after the short-time Fourier transform can be represented on a color map as shown in the center of FIG. 6. In the color map, the horizontal axis represents time, the vertical axis represents frequency, and the depth axis (the axis perpendicular to the paper surface) represents a color corresponding to the intensity I1. The first short-time Fourier transform unit 212 outputs the generated intensity signal I1(t, f) to the first SF conversion unit 213.

[0037] The first SF transform unit 213 generates a spectral flux signal by performing a spectral flux transform on the intensity signal I1(t, f) generated by the first short-time Fourier transform unit 212. The spectral flux transform is expressed by the following equation (2).

number

[0038] SF1 represents the spectral flux value. H is a function that performs half-wave rectification to emphasize the rising edge of the intensity signal I1, returning the value unchanged for zero or positive values ​​and zero for negative values ​​(see equation (3) below). L represents the maximum frequency (upper frequency limit).

number

[0039] As can be seen from equation (2), the spectral flux signal is calculated by calculating, for each frequency f, the change in intensity I1 over time at the same frequency f (the difference between the intensity I1(t) at a certain time and the intensity I1(t-1) at the previous time), and then summing this change over time across a set frequency band (all frequency bands in this example). This generates a signal SF1(t) indicating the change in the spectral flux value over time (hereinafter also referred to as the "spectral flux signal"). An example of the waveform of a spectral flux signal is shown at the bottom of Figure 6. In the figure, the spectral flux value is referred to as the SF value. The spectral flux signal corresponds to the "conversion signal" according to the present disclosure.

[0040] In this example, the first SF converter 213 generates a signal SF1(t) indicating a temporal change in the spectral flux value across the entire frequency band by summing the differences in intensity I1 for all frequencies f for each time interval. However, any conversion formula may be used as long as it can emphasize and detect changes in the intensity signal I1(t,f), i.e., the rising and / or falling edges of the intensity signal I1(t,f). For example, a converter (not shown) provided instead of the first SF converter 213 may sum the differences in intensity I1 between adjacent frequencies f over a specified time period. The first SF converter 213 outputs the spectral flux signal SF1(t) to the feature extractor 24.

[0041] The functions of the second short-time Fourier transform unit 222 and the second SF transform unit 223 are equivalent to the functions of the first short-time Fourier transform unit 212 and the first SF transform unit 213, respectively, except that the target frequency bands and time widths may be different, and therefore description thereof will not be repeated. The first short-time Fourier transform unit 212 and the first SF transform unit 213 correspond to the "first signal transform unit" according to the present disclosure. The second short-time Fourier transform unit 222 and the second SF transform unit 223 correspond to the "second signal transform unit" according to the present disclosure.

[0042] The feature extraction unit 24 detects changes in the spectral flux signal SF1(t) generated by the first SF conversion unit 213 and changes in the spectral flux signal SF2(t) generated by the second SF conversion unit 223. The feature extraction unit 24 may detect changes in the two spectral flux signals SF1(t) and SF2(t) separately, or may detect changes in a single signal (e.g., SF1(t)+SF2(t)) obtained by superimposing the two spectral flux signals SF1(t) and SF2(t). The feature extraction unit 24 extracts one or more features associated with the detected changes.

[0043] FIG. 7 is a diagram showing a first example of a change detected in a spectral flux signal. The horizontal axis represents time, and the vertical axis represents the spectral flux value. The same applies to the subsequent FIGS. 8 to 10. The feature extraction unit 24 may detect that a peak (especially a peak value greater than a threshold) has occurred in the spectral flux signal. In this case, the feature extraction unit 24 may extract a feature based on the peak time and / or the peak value.

[0044] 8 shows a second example of a change detected in the spectral flux signal. The feature extractor 24 may detect that the spectral flux signal has exceeded a threshold. In this case, the feature extractor 24 may extract a feature based on the time at which the spectral flux signal has exceeded the threshold (which may be the rise time or the fall time), or based on the time difference between the rise time and the fall time, or based on a spectral flux value (e.g., a peak value) greater than the threshold.

[0045] 9 is a diagram illustrating a first example of a feature. In this example, four peaks are detected in the spectral flux signal. The feature extractor 24 may extract, as the feature, time differences ΔTa, ΔTb, and ΔTc between two adjacent peaks among the four peaks.

[0046] FIG. 10 is a diagram illustrating a second example of features. The horizontal axis represents time. The upper vertical axis represents the spectral flux value, and the lower vertical axis represents the intensity of the acoustic signal. The feature extraction unit 24 may extract, as a feature, the frequency of the acoustic signal x(t) at each time when four peaks are detected. Although not shown, the feature extraction unit 24 may also extract, as a feature, the frequency (intensity on a color map) of the intensity signal I(t, f) at each time when four peaks are detected.

[0047] 11 is a diagram illustrating a third example of the feature quantity. The horizontal axis represents time, and the vertical axis represents the spectral flux value. The feature quantity extraction unit 24 may extract four peak values ​​(or the amount of change in the spectral flux value before and after the peak) ΔSFa, ΔSFb, ΔSFc, and ΔSFd as the feature quantities.

[0048] 12 is a diagram illustrating a fourth example of the feature. The horizontal axis represents time, and the vertical axis represents the intensity of the acoustic signal. The feature extractor 24 may extract, as a feature, the decay time τa or τb of the acoustic signal at the time when a peak is detected in the spectral flux signal (typically, the time required for the amplitude of the acoustic signal to become 1 / e of the amplitude before decay).

[0049] Referring again to FIG. 4, the storage unit 25 stores criteria (hereinafter referred to as "diagnostic criteria") for diagnosing the switching device 71. The diagnostic criteria are determined depending on the type of feature, and may be, for example, a reference intensity, a reference amount of change, a reference frequency, a reference time, or a reference waveform. The storage unit 25 outputs the diagnostic criteria to the diagnosing unit 26.

[0050] The diagnosis unit 26 diagnoses the diagnostic object by comparing one or more feature amounts extracted by the feature amount extraction unit 24 with the diagnostic criteria stored in the storage unit 25. The diagnostic method used by the diagnosis unit 26 will be described in detail later.

[0051] <Processing flow> <Frequency division> FIG. 13 is a flowchart showing a first example of the processing procedure of the processing (also referred to as "frequency division") executed in the first embodiment. The processing shown in this flowchart is called from a main routine (not shown) when a predetermined condition is met (for example, at every predetermined period), and executed. Each step is realized by software processing by the diagnostic device 1 (more specifically, the processor 101), but may also be realized by hardware (electrical circuitry) arranged in the diagnostic device 1. The same applies to the processing shown in the other flowcharts. Hereinafter, a step is abbreviated as S.

[0052] 1, 3 and 13, the diagnostic device 1 acquires an acoustic signal from the acoustic sensor 81 in S101.

[0053] In S102, the diagnostic device 1 typically uses a bandpass filter to divide the acoustic signal into two signals (e.g., a low-frequency signal and a high-frequency signal) according to their frequency bands. As long as the frequency bands are different, the difference in frequency bands between the two signals is not particularly limited. The diagnostic device 1 may divide the acoustic signal into three or more signals.

[0054] In S103, the diagnostic device 1 calculates an intensity signal indicating the spectral intensity for each time width by performing a short-time Fourier transform on each of the two signals divided in S102. As described above, the diagnostic device 1 may perform a continuous wavelet transform, a discrete wavelet transform, or a Stockwell transform instead of the short-time Fourier transform.

[0055] In S104, the diagnostic device 1 performs a spectral flux transform on the two intensity signals to generate a spectral flux signal that indicates a time change in the spectral flux value. The spectral flux transform is a process of summing intensity differences for all frequencies for each time interval. Alternatively, the diagnostic device 1 may sum intensity differences between adjacent frequencies over a specified time length.

[0056] In S105, the diagnostic device 1 extracts one or more feature amounts from the two spectral flux signals (feature amount extraction), as will be described with reference to FIG.

[0057] In S106, the diagnostic device 1 diagnoses whether or not there is an abnormality in the switching device 71 based on one or more feature amounts extracted by the feature amount extraction, as will be described with reference to FIG. 15 (abnormality diagnosis).

[0058] Feature Extraction 1, 3, and 14, in S51, the diagnostic device 1 extracts, as a feature, a time difference between a plurality of peaks detected in each of the two spectral flux signals (see FIG. 9).

[0059] In S52, the diagnostic device 1 extracts, as a feature, the frequency of the acoustic signal at the time when a peak is detected in each of the two spectral flux signals (see FIG. 10).

[0060] In S53, the diagnostic device 1 extracts one or more peak values ​​in each of the two spectral flux signals as feature quantities (see FIG. 11).

[0061] In S54, the diagnostic device 1 extracts, as a feature, the decay time of the acoustic signal at the time when the peak is detected in each of the two spectral flux waveforms (see FIG. 12).

[0062] The order of the four processes of S51 to S54 is not particularly limited, and the four processes can be executed in any order. Instead of all of the four processes, only one, two, or three of the four processes may be executed. When the feature extraction process is completed, the diagnostic device 1 returns to the frequency division process (see FIG. 13).

[0063] <Abnormality diagnosis> FIG. 15 is a flowchart showing an example of a processing procedure for abnormality diagnosis. With reference to FIGS. 1, 3, and 15, in S61, the diagnostic device 1 determines whether each of one or more feature quantities extracted by the feature extraction process satisfies a predetermined diagnostic criterion. If at least one feature quantity satisfies the diagnostic criterion (YES in S61), the diagnostic device 1 diagnoses that an abnormality associated with the feature quantity has occurred in the switching device 71 (S62). The diagnostic device 1 then notifies an external server 91 or terminal 92 of the type of abnormality that has occurred (S63). On the other hand, if all feature quantities do not satisfy the diagnostic criterion (NO in S61), the diagnostic device 1 diagnoses that the switching device 71 is normal (S64). Although not shown, the diagnostic device 1 may notify the external server 91 or terminal 92 of the diagnosis result that the switching device 71 is normal.

[0064] When a physical quantity indicating a time change of an object to be diagnosed is detected by a sensor, a signal that is originally desired to be acquired (e.g., a signal on the order of kHz) may be superimposed with another signal (e.g., a signal on the order of Hz) having a different frequency band from the original signal. Furthermore, noise may be superimposed on the original signal. In such cases, simply using the detected physical quantity for diagnosis may result in interference with signals other than the original signal, reducing the diagnostic accuracy of the object to be diagnosed. In the first embodiment, a sensor signal indicating a time change of an object to be diagnosed (in this example, switching device 71) is divided into a first time-domain signal and a second time-domain signal having different frequency bands. This makes it possible to analyze phenomena occurring in the object to be diagnosed separately according to the frequency band. Therefore, according to the first embodiment, the diagnostic accuracy of the object to be diagnosed can be improved.

[0065] <Explanation of various variations> <Weighted Spectral Flux Transform> Instead of the normal spectral flux transform shown in the above formula (2), a weighted spectral flux transform as shown in the following formula (4) may be used. The weighted spectral flux transform is a function H for emphasizing the onset of an audio signal, with a weighting coefficient Kf This differs from the ordinary spectral flux transform in that it is multiplied by the weighting coefficient K f is a set of weighting coefficients determined for each frequency f.

number

[0066] Figure 16 shows the weighting coefficient K f 16 is a diagram for explaining a method for setting the frequency f. The horizontal axis represents frequency f. The vertical axis represents the difference in spectral intensity at a certain time difference. In this example, the difference between the spectral intensity I(t,f) at time t and the spectral intensity I(t-1,f) at time t-1 is used on the vertical axis. By calculating the difference in spectral intensity for various time differences, many frequency distributions such as those shown in FIG. 16 can be obtained. The average of many frequency distributions is used as the weighting coefficient K f Alternatively, a weighting coefficient K can be calculated by machine learning (deep learning, etc.) on a large number of frequency distributions. f It is also possible to set

[0067] Figure 17 is a conceptual diagram comparing the ordinary spectral flux transform and the weighted spectral flux transform. The spectral flux values ​​calculated by the ordinary spectral flux transform can contain not only a few high peaks but also a large number of low peaks. In this case, additional processing is required to separate the high peaks from the low peaks (i.e., to extract the high peaks). In contrast, with the weighted spectral flux transform, the low peaks are suppressed because the weighting coefficients corresponding to the low peaks are small, while the high peaks are emphasized because the weighting coefficients corresponding to the high peaks are large. As a result, peak detection becomes easier.

[0068] <Time division> So far, we have explained "frequency division," which generates two signals by dividing an acoustic signal according to frequency bands. Below, we will explain an example of dividing an acoustic signal in time in addition to frequency bands. This process is called "time division."

[0069] FIG. 18 is a conceptual diagram illustrating time division. Time division can be contrasted with the frequency division shown in FIG. 5. First, an acoustic signal is frequency-divided to obtain a low-frequency signal and a high-frequency signal. Next, in this example, the low-frequency signal is time-divided into two, and the high-frequency signal is time-divided into two. Each signal may be time-divided into two or more signals, for example, based on the time when the intensity of the acoustic signal is higher than a reference value. If an acoustic signal intensity lower than the reference value continues for a long period of time, it is possible that the acoustic signal that is actually to be acquired has not occurred during that period, and only noise has occurred. Therefore, acoustic signals during periods (noise periods) in which the intensity lower than the reference value continues for longer than a specified length may be removed.

[0070] 18, a short-time Fourier transform is then performed on each of the four divided signals, followed by a spectral flux transform. The processing details of the short-time Fourier transform and the spectral flux transform are the same as those described above, and therefore are not shown in detail.

[0071] 19 is a flowchart showing a second example of the processing procedure of the processing (time division) executed in Embodiment 1. With reference to FIGS. 1, 3 and 19, the diagnostic device 1 acquires an acoustic signal from the acoustic sensor 81 in S201.

[0072] In S202, the diagnostic device 1 typically uses a band-pass filter to divide the acoustic signal into two signals according to frequency bands (frequency division).

[0073] In S203, the diagnostic device 1 divides each of the two frequency-divided signals in time (time division). As described above, the diagnostic device 1 may compare the intensity with a reference value and divide each signal into two or more signals using the reference value as a boundary. The diagnostic device 1 may also divide each signal into three or more signals. The diagnostic device 1 may also divide only some of the two or more frequency-divided signals in time.

[0074] When the object to be diagnosed is a switching device 71 including contacts as in this example, the diagnostic device 1 may receive a contact operation signal (a signal indicating the operation of the contacts) from the switching device 71. The diagnostic device 1 may time-divide each of the two frequency-divided signals based on the time at which the contact operation signal changes.

[0075] In S204, the diagnostic device 1 removes noise periods, during which the intensity is lower than the reference value and continues for a longer period than a specified length, from each signal obtained by time division. For example, when the opening and closing operations of the switching device 71 are alternately repeated (in the case of continuous operation, which will be described later), a waiting time occurs between the opening and closing operations during which no operation is performed. Therefore, by removing the noise periods, it is possible to remove environmental noise between the opening and closing operations. As a result, signals including a period during which the intensity is higher than the reference value (a period during which a meaningful signal is generated) are selectively retained.

[0076] The processes from S205 onwards are the same as the corresponding processes in the flowcharts shown in FIGS. 13 to 15, and therefore detailed description thereof will not be repeated.

[0077] As described above, time division makes it possible to remove noise periods. Additionally, time division makes it possible to adjust the conditions of the short-time Fourier transform for each signal, thereby improving the frequency resolution of each signal. More specifically, in general, an analysis window function is used in the short-time Fourier transform, which has a trade-off between frequency resolution and time resolution. The analysis window function is a function that has a non-zero value within the range 0≦t≦N−1 (N is the frame length) and is zero outside this range. Removing noise periods through time division lengthens the frame length N. This reduces the time resolution but increases the frequency resolution. Therefore, time division can further improve the diagnostic accuracy of the switching device 71.

[0078] <Continuous operation> If there is a time interval between the opening operation and closing operation of the switchgear 71, the power supply to the power system located downstream of the switchgear 71 may be cut off, causing a power outage in the power system. Power outages in the power system can be prevented by performing a closing operation immediately after the opening operation of the switchgear 71. Hereinafter, performing the opening operation and closing operation of the switchgear 71 in a short time period that can prevent a power outage in the power system, and diagnosing the switchgear 71 during that time, will be referred to as the "continuous operation" of the switchgear 71.

[0079] FIG. 20 is a flowchart showing a third example of the processing procedure of the processing (continuous operation) executed in the first embodiment. With reference to FIGS. 1, 3, and 20, in S301, the diagnostic device 1 outputs a control command to the switching device 71 to cause the switching device 71 to perform an opening operation. Subsequently, the diagnostic device 1 acquires an acoustic signal accompanying the opening operation of the switching device 71 from the acoustic sensor 81 (S302). The diagnostic device 1 divides the acoustic signal into two signals according to the frequency band, for example, by using a band-pass filter (S303). The diagnostic device 1 calculates an intensity signal by performing a short-time Fourier transform on each of the two frequency-divided signals (S304). The diagnostic device 1 generates a spectral flux signal by performing a spectral flux transform on the two intensity signals (S305). Each of the processing steps from S302 to S305 is equivalent to the corresponding processing in the flowchart shown in FIG.

[0080] In S306, the diagnostic device 1 extracts feature quantities from the spectral flux signal based on the acoustic signal accompanying the opening operation of the switching device 71. The feature quantity extraction can be performed using a process equivalent to that shown in the flowchart of FIG.

[0081] Next, in S307, the diagnostic device 1 outputs a control command to the switching device 71 to cause the switching device 71 to perform a closing operation. The time from when the control command to cause the switching device 71 to perform an opening operation is output in S301 to when the control command to cause the switching device 71 to perform a closing operation is output in S307 is set to be sufficiently short (for example, less than one second) to prevent a power outage in the power grid. The subsequent processes of S308 to S311 are equivalent to the processes of S302 to S305.

[0082] In S312, the diagnostic device 1 extracts feature quantities from the spectral flux signal based on the acoustic signal accompanying the closing operation of the switching device 71. The feature quantity extraction can be performed using a process equivalent to that shown in the flowchart of FIG.

[0083] In S313, the diagnostic device 1 executes an abnormality diagnosis and determines whether or not predetermined diagnostic criteria are satisfied for both the feature amount associated with the opening operation of the switching device 71 and the feature amount associated with the closing operation of the switching device 71. Although the feature amount used for the abnormality diagnosis is different, the abnormality diagnosis is the same as the processing shown in the flowchart of FIG.

[0084] In this example, the short-time Fourier transform (S304), spectral flux transform (S305), and feature extraction (S306) processes related to the opening operation of the switchgear 71 are executed between the opening operation and closing operation of the switchgear 71. However, the processes of S304 to S306 may be executed after acquiring an acoustic signal accompanying the closing operation of the switchgear 71 (after the process of S308). In other words, the diagnostic device 1 may first acquire an acoustic signal accompanying the opening operation of the switchgear 71 and an acoustic signal accompanying the closing operation of the switchgear 71, and then execute signal processing (the short-time Fourier transform, spectral flux transform, and feature extraction processes) on these signals afterwards.

[0085] As described above, continuous operation makes it possible to diagnose the switching gear 71 while suppressing power outages in the power grid. Normally, opportunities to diagnose the switching gear 71 are only available under circumstances where power outages in the power grid are not a problem. In contrast, continuous operation makes it easy to ensure opportunities to diagnose the switching gear 71, making it possible to diagnose the switching gear 71 at a desired timing or more frequently.

[0086] ≪Lifespan estimation≫ An example has been described in which the presence or absence of an abnormality in the switching device 71 is diagnosed based on the feature amount (abnormality diagnosis). The lifespan of the switching device 71 may also be estimated. This process is referred to as "lifespan estimation." In the following example, lifespan estimation is performed in addition to abnormality diagnosis (see FIG. 15). However, lifespan estimation may also be performed instead of abnormality diagnosis.

[0087] Fig. 21 is a flowchart showing an example of a processing procedure for lifetime estimation according to embodiment 1. With reference to Figs. 1, 3 and 21, in S71, diagnostic device 1 determines whether each of a plurality of feature amounts extracted by feature amount extraction (see Fig. 14) satisfies a diagnostic criterion.

[0088] If at least one feature quantity satisfies the diagnostic criterion (YES in S71), the diagnostic device 1 diagnoses that an abnormality associated with the feature quantity has occurred in the switching device 71 (S72). Then, the diagnostic device 1 notifies the external server 91 or terminal 92 of what kind of abnormality has occurred (S73).

[0089] On the other hand, if all the feature amounts do not satisfy the diagnostic criteria (NO in S71), the diagnostic device 1 diagnoses that the switching device 71 is normal (S74).

[0090] In S75, the diagnostic device 1 estimates the lifespan of the switching device 71 by extrapolating a feature that has been previously associated with a lifespan from among the multiple feature values. The diagnostic device 1 then notifies an external server 91 or terminal 92 of the estimated lifespan of the switching device 71 along with the diagnosis result that the switching device 71 is normal. The diagnostic device 1 further notifies the server 91 or terminal 92 of a recommended update time for the switching device 71 that is determined based on the estimated lifespan (S76). The recommended update time can be determined a specified period before the estimated lifespan. Note that estimating the "lifespan" of the switching device 71 can include estimating the "remaining lifespan" of the switching device 71.

[0091] FIG. 22 is a diagram illustrating a lifespan estimation method. The horizontal axis represents elapsed time, and the vertical axis represents the feature value. In this example, a threshold value for the feature value is set in advance. The feature value reaching the threshold value signifies that the lifespan of the switching device 71 has expired. Although the feature value gradually increases over time, it has not yet reached the threshold value at the present time (time t5). The diagnostic device 1 estimates the time (time t6) at which the feature value will reach the threshold value by extrapolation (typically, extrapolation of a regression curve) based on the transition of the feature value over a period from the past to the present (the period from time t1 to time t5). The diagnostic device 1 may determine the time at which the feature value reaches the threshold value as the estimated lifespan of the switching device 71. Alternatively or additionally, the diagnostic device 1 may determine the length of the period from the present time to the time at which the feature value reaches the threshold value (the period between time t5 and time t6) as the estimated remaining lifespan of the switching device 71.

[0092] As described above, the lifespan estimation allows the user (administrator, etc.) of the switching device 71 to understand the estimated lifespan of the switching device 71. This enables the user to revise the maintenance plan for the switching device 71 or update (inspect, service, or replace) the switching device 71.

[0093] <Processing Combination> FIG. 23 is a diagram illustrating processes that can be combined with frequency division. In the diagram, a check mark next to time division, continuous operation, or lifetime estimation indicates that the process can be combined with frequency division. Frequency division can be performed alone (No. 1), combined with time division only (No. 2), combined with time division and continuous operation (No. 3), combined with time division and lifetime estimation (No. 4), combined with all of time division, continuous operation, and lifetime estimation (No. 5), combined with continuous operation only (No. 6), combined with continuous operation and lifetime estimation (No. 7), or combined with lifetime estimation only (No. 8).

[0094] Embodiment 2 In the second embodiment, a process of recognizing a spectral flux signal as a geometric figure (typically a polygon) and extracting feature quantities from the geometric figure will be described. In the second embodiment, the diagnostic device includes a signal processing unit 12B that is different from the signal processing unit 12A (see FIG. 4) in the first embodiment. The overall configuration of the diagnostic system in the second embodiment is similar to the configuration shown in FIGS. 1 to 3, and therefore description thereof will not be repeated.

[0095] <System configuration> 24 is a functional block diagram for explaining the function of the signal processing unit in the second embodiment. Signal processing unit 12B includes a filter 31, a short-time Fourier transform unit 32, a spectral flux transform unit 33, a geometric figure generation unit 34, a feature extraction unit 35, a storage unit 36, and a diagnosis unit 37. That is, signal processing unit 12B differs from signal processing unit 12A in the first embodiment in that signal processing unit 12B includes one filter, one short-time Fourier transform unit, and one spectral flux transform unit, and further includes a geometric figure generation unit 34. Filter 31, short-time Fourier transform unit 32, spectral flux transform unit 33, storage unit 36, and diagnosis unit 37 are equivalent to first filter 211, first short-time Fourier transform unit 212, first SF transform unit 213, storage unit 36, and diagnosis unit 37 in the first embodiment, respectively. The short-time Fourier transform unit 32 and the spectral flux transform unit 33 correspond to the "signal conversion unit" according to the present disclosure.

[0096] The geometric figure generation unit 34 receives the spectral flux signal from the spectral flux conversion unit 33 and generates a geometric figure from the spectral flux signal. The geometric figure generation unit 34 outputs the generated geometric figure to the feature extraction unit 35. The feature extraction unit 35 extracts features based on the geometric figure. As described below, the signal processing unit 12B executes "geometric processing" or "comparison processing." Each process will be described in detail.

[0097] <Conceptual diagram and processing flow> <Geometric processing> FIG. 25 is a conceptual diagram illustrating a first example of a geometric figure generation method and a feature extraction method. With reference to FIGS. 24 and 25, the geometric figure generation unit 34, for example, identifies multiple extreme values ​​(maximum and minimum values) of the spectral flux signal as multiple vertices for generating a geometric figure (typically a polygon). The geometric figure generation unit 34 then generates the geometric figure by connecting some or all of the multiple vertices. As a result, three triangles are generated in the example shown in FIG. 25. The geometric figure is not limited to a triangle. The geometric figure generation unit 34 may also generate a geometric figure with four or more vertices. For example, if all vertices are connected, a heptagon may be generated, as shown enclosed by a dashed line. Furthermore, when the spectral flux signal has an intersection with the time axis (a point where the spectral flux value is 0) due to correction of the intensity of the spectral flux signal, the geometric figure generation unit 34 may generate a geometric figure that includes the intersection of the spectral flux signal with the time axis as part of multiple vertices in addition to the extreme values ​​of the spectral flux signal.

[0098] The feature extraction unit 35 extracts features based on a geometric figure. For example, the feature extraction unit 35 may extract the center of gravity of each of three triangles as a feature. The feature extraction unit 35 may also extract the center of gravity of a geometric figure (a heptagon in the example shown within the dashed line) generated by connecting all of the vertices as a feature.

[0099] 26 is a flowchart showing a first example of a processing procedure of a process (geometric processing) executed in Embodiment 2. With reference to FIGS. 1, 3, 25 and 26, the diagnostic device 1 acquires an acoustic signal from the acoustic sensor 81 in S401.

[0100] In S402, the diagnostic device 1 calculates an intensity signal by short-time Fourier transform of the acoustic signal. As in the first embodiment, the diagnostic device 1 may perform continuous wavelet transform, discrete wavelet transform, or Stockwell transform.

[0101] In S403, the diagnostic device 1 generates a spectral flux signal by spectral flux transforming the intensity signal. The spectral flux transform may be a sum of intensity differences for all frequencies for each time width, or a sum of intensity differences between adjacent frequencies over a specified time length.

[0102] In S404, the diagnostic device 1 identifies a plurality of extrema of the spectral flux signal as a plurality of vertices for generating a geometric figure.

[0103] In S405, the diagnostic device 1 generates a geometric figure by connecting some or all of the vertices identified in S404. The diagnostic device 1 may generate a plurality of triangles as shown in Fig. 25, or may generate one or more n-gons (n ​​is a natural number equal to or greater than 4).

[0104] In S406, the diagnostic device 1 extracts the center of gravity of the geometric figure generated in S405 as a feature amount (feature amount extraction). The diagnostic device 1 may extract the vertices of the geometric figure as feature amounts.

[0105] In S407, the diagnostic device 1 diagnoses whether or not there is an abnormality in the switching device 71 based on the feature amounts extracted by the feature amount extraction (abnormality diagnosis). This process is similar to the process described with reference to Fig. 15, and therefore description thereof will not be repeated.

[0106] <<Comparison process>> 27 is a conceptual diagram illustrating a second example of a method for generating a geometric figure and a method for extracting a feature quantity. In this example, a spectral flux signal (corresponding to a "normal signal" according to the present disclosure) acquired from a normal switching device 71 is pre-stored in, for example, the geometric figure generator 34. The geometric figure generator 34 may acquire the spectral flux signal from a storage unit (memory 102 in FIG. 1) not shown in FIG. 27.

[0107] The geometric figure generation unit 34 compares the currently acquired spectral flux signal with a spectral flux signal acquired from a normal switching device 71 and identifies a region where the two signals differ. The geometric figure generation unit 34 extracts signal components from the currently acquired spectral flux signal in the identified region. The geometric figure generation unit 34 then identifies multiple extreme values ​​of the extracted signal components as multiple vertices for generating a geometric figure. The geometric figure generation unit 34 generates the geometric figure by connecting some or all of the multiple vertices. The geometric figure generation unit 34 then outputs the generated geometric figure to the feature extraction unit 35.

[0108] The feature extraction unit 35 extracts the center of gravity or the vertex of the geometric figure as the feature, similarly to the example shown in FIG.

[0109] Fig. 28 is a flowchart showing a second example of the processing procedure of the processing (comparison processing) executed in Embodiment 2. With reference to Figs. 1, 3, 27 and 28, the processing in S501 to S503 is equivalent to the processing in S401 to S403 shown in Fig. 26.

[0110] In S504, the diagnostic device 1 reads out from the memory 102 the spectral flux signal acquired from the normal switching device 71.

[0111] In S505, the diagnostic device 1 compares the spectral flux signal calculated in S503 with the spectral flux signal read out in S504 to identify a region where the two spectral flux signals differ from each other. The diagnostic device 1 identifies the region where the two spectral flux signals differ, for example, by using the difference between the two spectral flux signals. The diagnostic device 1 then extracts signal components in the identified region from the spectral flux signal calculated in S503.

[0112] In S506, the diagnostic device 1 identifies a plurality of extreme values ​​(which may be inflection points or the like) in the signal components extracted in S505 as a plurality of vertices for generating a geometric figure.

[0113] The subsequent processing of S507 to S509 is the same as the processing of S405 to S407 shown in FIG.

[0114] The waveform shape of the spectral flux signal may change due to various factors. In the second embodiment, instead of directly extracting feature values ​​from the spectral flux signal, a geometric figure is generated from the spectral flux signal, and feature values ​​are extracted based on the generated geometric figure. By generating a geometric figure from the spectral flux signal, it is possible to highlight the range in which some fluctuation (rising edge, falling edge, etc.) occurs in the spectral flux signal. Furthermore, by extracting the center of gravity of the geometric figure or the like as a feature value, it is possible to ignore minute changes in the spectral flux signal waveform, while making substantial changes in the spectral flux signal clear. Therefore, according to the second embodiment, it is possible to improve the diagnostic accuracy of the switching device 71.

[0115] In addition, the amount of information (data size) of a geometric figure is smaller than the amount of information of a spectral flux signal. Therefore, when data (log) showing the diagnostic progress of switchgear 71 is stored in memory 102 or server 91, the storage capacity can be reduced.

[0116] <Processing Combination> 29 is a diagram illustrating processes that can be combined with geometric processing or comparison processing. Geometric processing can be performed alone (No. 1), combined with time division only (No. 2), combined with time division and continuous motion (No. 3), combined with time division and lifespan estimation (No. 4), combined with all of time division, continuous motion, and lifespan estimation (No. 5), combined with continuous motion only (No. 6), combined with continuous motion and lifespan estimation (No. 7), or combined with lifespan estimation only (No. 8). The same applies to comparison processing (Nos. 9 to 16).

[0117] The frequency division in the first embodiment may be combined with the geometric processing or comparison processing in the second embodiment. That is, the geometric processing or comparison processing may be performed on each of two or more frequency-divided spectral flux signals.

[0118] Embodiment 3 In the third embodiment, a configuration will be described in which the diagnostic accuracy is improved by combining an acoustic signal with another signal. In the third embodiment, the diagnostic target is one or more electrical devices housed inside the power receiving and distribution facility. The one or more electrical devices include a switching device 71.

[0119] <System configuration> Fig. 30 is a diagram showing the overall configuration of a switchgear diagnostic device according to embodiment 3. In embodiment 3, as in embodiment 1 (see Fig. 3), an acoustic sensor 81 is provided in a circuit breaker 711 of a switchgear 71. The acoustic sensor 81 includes, for example, a microphone, and detects the operating sound of the switchgear 71. The acoustic sensor 81 outputs an acoustic signal indicating the detection result to a diagnostic device 1C.

[0120] In the third embodiment, in addition to the acoustic sensor 81, one or more other sensors are provided in the circuit breaker 711. The other sensors are used to reinforce the diagnosis based on the acoustic signal from the acoustic sensor 81, and are therefore also referred to as "reinforcing sensors" hereinafter. The reinforcing sensors may detect a state of the switchinggear 71 different from that detected by the acoustic sensor 81, among various states of the switchinggear 71 that can be used to diagnose the switchinggear 71. The reinforcing sensors may detect the state of the switchinggear 71 according to a different principle than that used by the acoustic sensor 81. This is because the accuracy of diagnosing the switchinggear 71 can be improved compared to when the acoustic sensor 81 is used alone. Furthermore, it may also be possible to diagnose the switchinggear 71 from a different perspective than that used by the acoustic sensor 81, or for a different range than that used by the acoustic sensor 81.

[0121] More specifically, the acoustic sensor 81 is provided for diagnosis based on the frequency of the operating sound of the opening and closing device 71. The acoustic sensor 81 is placed at a position a certain distance away from the opening and closing device 71 (in other words, non-contact), and because the speed of sound is slower than the speed of light, there may be a slight time delay (propagation delay) before the operating sound of the opening and closing device 71 is detected by the acoustic sensor 81. In addition, noise due to sound diffraction or the like may be superimposed when the acoustic signal propagates through the air.

[0122] In view of the characteristics of the acoustic sensor 81, the reinforcing sensor may have a different detection target than the acoustic sensor 81 (for diagnosing something other than the frequency of the operating sound of the opening and closing device 71), may have a large or negligible time delay, or may be placed in contact with the opening and closing device 71.

[0123] In this example, the reinforcing sensors include at least one of a vibration sensor 82, a voltage sensor 83, and a current sensor 84. The vibration sensor 82 includes, for example, a piezoelectric element and detects vibrations of the switchgear 71. The vibration sensor 82 may be a strain sensor. The voltage sensor 83 detects the voltage applied to the switchgear 71 (circuit breaker 711). The current sensor 84 detects the current flowing through the switchgear 71 (circuit breaker 711). Each sensor outputs a signal indicating its detection result to the diagnostic device 1C.

[0124] The detection targets of the vibration sensor 82, voltage sensor 83, and current sensor 84 are different from those of the acoustic sensor 81. The vibration sensor 82, voltage sensor 83, and current sensor 84 are all arranged in contact with the switchgear 71, and the time delay with respect to the operation of the switchgear 71 is negligibly small. By using the vibration sensor 82, it is possible to detect abnormal vibrations that occur when the switchgear 71 operates. By using the voltage sensor 83, it is possible to diagnose whether or not insulation deterioration has occurred in the circuit breaker 711 based on the voltage phase. By using the current sensor 84, it is possible to diagnose whether or not an arc has occurred in the circuit breaker 711 based on the current phase.

[0125] The reinforcing sensor is not limited to the above three types of sensors. The reinforcing sensor may be an optical sensor (photodetector) or an image sensor. The optical sensor or image sensor has a smaller time delay with respect to the operation of the switching device 71 than the acoustic sensor 81. By using the optical sensor or image sensor, it becomes possible to detect light (such as an arc) emitted by the switching device 71 (circuit breaker 711). Note that the image sensor may be used as a surveillance camera to monitor whether there are any suspicious persons, wild animals, etc. around the switching device 71 based on the captured images.

[0126] The reinforcing sensor may be a temperature sensor. The reinforcing sensor may be a contact-type temperature sensor (such as a thermocouple) or a non-contact-type temperature sensor (such as a radiation thermometer). The reinforcing sensor may be a thermograph (thermal camera). The temperature sensor outputs a signal (an image in the case of a thermal camera) indicating a change in temperature of the switchgear 71. By using a temperature sensor, it becomes possible to diagnose abnormal heat generation due to welding of the contacts of the circuit breaker 711, the possibility of freezing of the circuit breaker 711, etc.

[0127] The reinforcing sensor may be appropriately selected depending on the diagnostic target. The reinforcing sensor may be a sensor that detects mechanical movement (such as a distance (displacement) sensor, a velocity sensor, or an acceleration sensor). The reinforcing sensor may be a pressure sensor. The reinforcing sensor may be a smoke sensor or an odor sensor.

[0128] Diagnostic device 1C differs from diagnostic device 1A in embodiment 1 (see FIG. 4) in that diagnostic device 1C includes signal processing unit 12C instead of signal processing unit 12A. Other configurations of diagnostic device 1C are equivalent to the corresponding configurations of diagnostic device 1A.

[0129] 31 is a functional block diagram for explaining the functions of a signal processing unit 12C in Embodiment 3. The signal processing unit 12C includes a filter 41, a short-time Fourier transform unit 42, a spectral flux unit 43, a feature extraction unit 44, a storage unit 45, and a diagnosis unit 46.

[0130] The filter 41 simultaneously receives the acoustic signal and the reinforcement signal. In other words, the filter 41 receives the acoustic signal and the reinforcement signal so that the reinforcement signal and the acoustic signal are synchronized with each other. The filter 41 passes only signal components of a certain frequency band of the acoustic signal and the reinforcement signal and outputs them to the short-time Fourier transform unit 42. The functions of the subsequent short-time Fourier transform unit 42, spectral flux unit 43, feature extraction unit 44, storage unit 45, and diagnosis unit 46 are equivalent to the functions of the corresponding functional blocks in the first embodiment (see FIG. 4).

[0131] Although not shown, a filter suitable for the acoustic signal and a filter suitable for the support signal may be provided separately. Similarly, the short-time Fourier transform unit, the spectral flux unit, and the feature extraction unit may each have a functional block for the acoustic signal and a functional block for the support signal.

[0132] The filter for the reinforcement signal, the short-time Fourier transform unit, the spectral flux unit, and the feature extraction unit may not be provided. The reinforcement signal may be, for example, a contact signal (a signal indicating contact or disconnection) acquired from a switching device, or an operation signal for operating the switching device. These signal waveforms do not oscillate. Therefore, for the contact signal or operation signal, it is not necessary to calculate an intensity signal indicating the time change in spectral intensity. For the contact signal or operation signal, the raw waveform can be used as the reinforcement signal. By using the contact signal or operation signal, it is possible to determine which time period of the waveform of the acoustic signal should be subjected to the short-time Fourier transform, the spectral flux transform, and the feature extraction (the appropriate timing for signal processing).

[0133] Alternatively, feature extraction can be performed on a waveform generated by superimposing, in time or frequency, the contact signal or operation signal and an intensity signal obtained by short-time Fourier transforming the waveform of the acoustic signal. This allows a point where the contact signal or operation signal changes without a peak to be identified as a feature point. Even for a peak-containing reinforcement signal, if the signal does not require a short-time Fourier transform (e.g., a signal measuring the operation of a switch contact), a spectral flux signal can be generated by summing, in time or frequency, the raw waveform and the intensity signal obtained by short-time Fourier transforming the waveform of the acoustic signal (spectral flux transform), and the spectral flux signal can be used for feature extraction.

[0134] <Processing flow> Fig. 32 is a flowchart showing an example of a processing procedure of a process (reinforcement process) executed in embodiment 3. With reference to Fig. 1, Fig. 30 and Fig. 32, in S501, diagnostic device 1C acquires an acoustic signal from acoustic sensor 81, and also acquires a reinforcement signal from a reinforcement sensor (here, any one of vibration sensor 82, voltage sensor 83 and current sensor 84) in synchronization therewith.

[0135] In S502, the diagnostic device 1C calculates an intensity signal by performing a short-time Fourier transform on the acoustic signal and the reinforcement signal (a single signal in which the acoustic signal and the reinforcement signal are synchronized with each other). As in the first embodiment, the diagnostic device 1 may perform a continuous wavelet transform, a discrete wavelet transform, or a Stockwell transform. As described above, depending on the type of reinforcement signal, the reinforcement signal (such as a contact signal of a switching device or an operation signal of a switching device) may not be subject to the short-time Fourier transform (and the subsequent spectral flux transform and extraction of feature amounts).

[0136] In S503, the diagnostic device 1C generates a spectral flux signal by spectral flux transforming the acoustic signal and the support signal. The spectral flux transform may be a sum of intensity differences for all frequencies for each time width, or a sum of intensity differences between adjacent frequencies over a specified time length.

[0137] In S504, diagnostic apparatus 1C extracts one or more feature amounts from the spectral flux signal generated in S503 (feature amount extraction). This process is equivalent to the process described with reference to Fig. 14, and therefore description thereof will not be repeated.

[0138] In S505, the diagnostic device 1C diagnoses whether or not there is an abnormality in the switching device 71 based on one or more feature amounts extracted by the feature amount extraction (abnormality diagnosis). This process is the same as the process described with reference to Fig. 15, and therefore description thereof will not be repeated.

[0139] Although acoustic signals are suitable for diagnosing the switching device 71 based on the frequency of the operating sound of the switching device 71, there may be abnormalities in the switching device 71 that are difficult to diagnose sufficiently using only acoustic signals. In the third embodiment, a reinforcement signal is used in addition to the acoustic signal. The reinforcement signal is based on a different detection principle than the acoustic signal and detects a different target than the acoustic signal. The acoustic signal and the reinforcement signal detect different states of the switching device 71 and detect different time delays. Therefore, by using both the acoustic signal and the reinforcement signal, the accuracy of diagnosing the switching device 71 can be improved compared to using only the acoustic signal.

[0140] <Processing Combination> 33 is a diagram illustrating processes that can be combined with the reinforcement process. The reinforcement process may be performed alone (No. 1), combined with only time division (No. 2), combined with time division and continuous operation (No. 3), combined with time division and lifespan estimation (No. 4), combined with all of time division, continuous operation, and lifespan estimation (No. 5), combined with only continuous operation (No. 6), combined with continuous operation and lifespan estimation (No. 7), or combined with only lifespan estimation (No. 8).

[0141] The reinforcement processing in the third embodiment may be combined with the frequency division in the first embodiment. The acoustic signal and the reinforcement signal may be frequency divided. Furthermore, the reinforcement processing in the third embodiment may be combined with the geometric processing or the comparison processing in the second embodiment. The geometric processing or the comparison processing may be performed on the spectral flux signal generated based on the acoustic signal and the reinforcement signal.

[0142] The embodiments disclosed herein should be considered to be illustrative in all respects and not restrictive. The scope of the present disclosure is defined by the claims, not by the description of the above embodiments, and is intended to include all modifications within the meaning and scope of the claims. [Explanation of symbols]

[0143] 100 diagnostic system, 1, 1A, 1C diagnostic device, 101 processor, 102 memory, 103 communication equipment, 3 controller, 7 diagnostic target, 8 sensor, 11 input unit, 12 signal synthesis unit, 121 filter, 122 time-frequency transformation unit, 123 signal transformation unit, 124 feature extraction unit, 125 diagnostic unit, 12A signal processing unit, 13 output unit, 14 control unit, 211 first filter, 212 first short-time Fourier transform unit, 213 first spectral flux transformation unit, 223 second spectral flux transformation unit, 221 second filter, 222 second short-time Fourier transform unit, 24 feature extraction unit, 25 memory unit, 26 diagnostic unit, 12B signal processing unit, 31 filter, 32 short-time Fourier transform unit, 33 spectral flux transformation unit, 34 geometric figure generation unit, 35 Feature extraction unit, 36 memory unit, 37 diagnosis unit, 12C signal processing unit, 41 filter, 42 short-time Fourier transform unit, 43 spectral flux unit, 44 feature extraction unit, 45 memory unit, 46 diagnosis unit, 711 circuit breaker, 712 motor, 71 switchgear, 81 acoustic sensor, 82 vibration sensor, 83 voltage sensor, 84 current sensor, 91 server, 92 terminal.

Claims

1. a signal conversion unit that generates a converted signal based on an acoustic signal indicating a time change in the operating sound of the diagnostic object and one or more reinforcement signals that are detected from the diagnostic object simultaneously with the acoustic signal according to a detection principle different from that of the acoustic signal; a feature extraction unit that extracts one or more feature values ​​from the converted signal; a diagnosis unit that diagnoses the diagnostic object based on the one or more feature amounts, The signal conversion unit calculating an intensity signal indicating a time change in spectral intensity by performing a time-frequency transform for each time width on both the acoustic signal and the one or more enhancement signals; A diagnostic device that generates the transformed signal by summing the intensity signals over time or frequency.

2. the signal conversion unit acquires the one or more reinforcement signals from a sensor placed in contact with the diagnostic object; 2. The diagnostic device of claim 1, wherein the one or more augmentation signals include at least one of a vibration signal indicating vibration of the diagnostic object, a voltage signal indicating a voltage applied to the diagnostic object, and a current signal indicating a current flowing through the diagnostic object.

3. the one or more reinforcing signals include the voltage signal; The diagnostic device according to claim 2 , wherein the diagnosing unit diagnoses whether the object to be diagnosed has insulation degradation based on the one or more feature amounts extracted from the phase or amplitude of the voltage signal.

4. the one or more augmentation signals include the current signal; The diagnostic unit according to claim 2 , wherein the diagnostic unit diagnoses whether an arc has occurred in the diagnostic object based on the one or more feature amounts extracted from the phase or amplitude of the current signal. Diagnostic equipment.

5. the signal conversion unit acquires the one or more reinforcement signals from a sensor that has a smaller time delay with respect to the motion of the diagnosis target than the acoustic signal; The diagnostic device of claim 1 , wherein the one or more augmentation signals include a signal indicative of light emitted by the diagnostic object.

6. The diagnostic device of claim 1 , wherein the one or more augmenting signals include a signal indicative of a temperature change of the diagnostic object.

7. The diagnostic device according to claim 1 , wherein the one or more augmentation signals include a signal indicative of any one of a displacement, a velocity, and an acceleration of the diagnostic object.

8. The diagnostic device of claim 1 , wherein the one or more augmentation signals include a signal indicative of a pressure change in the diagnostic subject.

9. The diagnostic device of claim 1 , wherein the one or more augmentation signals include a signal indicative of smoke or an odor generated in the diagnostic object.

10. the diagnosis target is a switching device including a contact, The diagnostic device of claim 1 , wherein the one or more augmenting signals include a signal indicative of operation of the contacts.

11. a filter that frequency-divides each of the acoustic signal and the one or more enhancement signals into two time-domain signals having different frequency bands; The diagnostic device according to claim 1 , wherein the feature extracting unit extracts the one or more feature amounts from the frequency-divided acoustic signal and the one or more enhancement signals.

12. The diagnostic device according to claim 11 , wherein the filter divides the frequency of the one or more support signals based on the frequency of the one or more support signals when the diagnostic object is normal.

13. The diagnostic device according to claim 1 , wherein the feature extracting unit extracts, as the one or more feature amounts, a peak value of the converted signal or an extreme value of the converted signal that is greater than a threshold value.

14. the diagnosis target is one or more electrical devices housed inside the power receiving and distribution equipment, the one or more electrical devices including a switchgear, The diagnostic device according to claim 1 , further comprising a control unit that controls the opening and closing device.

15. the one or more feature amounts are a plurality of feature amounts, The diagnostic device according to claim 1 , wherein the diagnostic unit estimates the lifetime of the diagnostic object by extrapolating the plurality of feature amounts.

16. The diagnostic device according to claim 15 , wherein the diagnostic unit notifies a recommended update time for the diagnostic target, which is determined based on the lifespan.

17. generating a conversion signal based on an acoustic signal indicating a time change in the operating sound of the diagnostic object and one or more reinforcement signals detected from the diagnostic object simultaneously with the acoustic signal according to a detection principle different from that of the acoustic signal; extracting one or more features from the transformed signal; diagnosing the diagnostic object based on the one or more feature amounts; The generating step includes: calculating an intensity signal indicating a time change in spectral intensity by performing a time-frequency transform for each time width on both the acoustic signal and the one or more enhancement signals; and generating the transformed signal by summing the intensity signals over time or frequency.

18. a signal conversion unit that calculates an intensity signal that indicates a time change in spectral intensity by performing a time-frequency conversion for each time width on an acoustic signal that indicates a time change in the operation sound of an object to be diagnosed, and generates a converted signal by superimposing the intensity signal and one or more reinforcement signals in terms of time or frequency, wherein the one or more reinforcement signals are signals that are detected from the object to be diagnosed simultaneously with the acoustic signal and in accordance with a detection principle different from that of the acoustic signal; and a feature extraction unit that extracts one or more feature values ​​from the converted signal; a diagnostic unit that diagnoses the diagnostic object based on the one or more feature amounts.

19. a signal conversion unit that calculates an intensity signal that indicates a time change in spectral intensity by performing a time-frequency conversion for each time width on an acoustic signal that indicates a time change in the operation sound of an object to be diagnosed, and generates a converted signal by summing the intensity signal and one or more reinforcement signals with respect to time or frequency, wherein the one or more reinforcement signals are signals that are detected from the object to be diagnosed simultaneously with the acoustic signal and in accordance with a detection principle different from that of the acoustic signal; and a feature extraction unit that extracts one or more feature values ​​from the converted signal; a diagnostic unit that diagnoses the diagnostic object based on the one or more feature amounts.

Images