JP3251799B2 - Equipment diagnostic equipment - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for detecting a sound or vibration generated from a device including a vibration component such as a transmission or an engine (hereinafter simply referred to as "component") or various facilities and machines. TECHNICAL FIELD The present invention relates to a device diagnostic device for detecting an abnormality of a device, and more particularly to a device diagnostic device with improved reliability of abnormality diagnosis.

[0002]

2. Description of the Related Art In general, components having a vibration source, such as components mounted on a mass-produced automobile, such as a transmission, are checked for defects in a production stage, and it is necessary to prevent defective products from leaking to the market. This is because, for example, when a defective transmission caused by defective machining or assembly of parts is mounted on an automobile and used, abnormal noise is generated from the transmission, and the transmission may be broken down. This is because there may be various inconveniences caused by the failure.

Therefore, for example, in order to cope with a case where a defective transmission occurs, it is necessary to detect an abnormality of the transmission and to take a measure in advance before the transmission is mounted on a vehicle. The above is very important for quality control of transmission production.

FIG. 18 shows, for example, the Society of Automotive Engineers of Japan.
Edited by the Committee on Vibration and Noise Section, "Report of Vehicle Noise Study, March 1992, Focusing on Passenger Car Noise" (19
FIG. 1 is a block diagram showing a conventional transmission (equipment) diagnostic apparatus inferred from the contents described in Japanese Patent Application Laid-Open No. 2001-19592. In the figure, reference numeral 1 denotes a device having a vibration source, for example, a transmission, which is constituted by the following elements 2 to 6.

[0005] Reference numeral 2 denotes a first shaft serving as an output shaft, 3 denotes a second shaft for transmitting a rotational force to the first shaft 2, and 4 denotes a first shaft 2 and a second shaft 3. A plurality of gears to be engaged at different gear ratios. Reference numeral 5 denotes, for example, a clutch provided corresponding to the input shaft, and selectively releases the engagement between the input shaft and the second shaft 3. Reference numeral 6 denotes a transmission mechanism for selecting the gear 4 of the transmission 1 by manual operation (corresponding to a shift lever).
It is.

Reference numeral 7 denotes a microphone which detects a sound S generated from the transmission 1 including the first shaft 2 to the speed change mechanism 6 and outputs an electric signal E corresponding to the detected sound S, and 9 amplifies the electric signal E. The amplifier 10 is a filter that passes a predetermined frequency component to be detected in the amplified electric signal E.
Is a sound pressure calculating device for calculating the sound pressure P from the electric signal E via the filter 10.

Reference numeral 12 denotes A for converting the sound pressure P calculated by the sound pressure calculation device 11 from an analog signal to a digital signal.
A D converter 13 for determining a malfunction of the transmission 1 based on the digitally converted sound pressure P (central processing unit);
Reference numeral 14 denotes an automatic range switching device that sets an optimum range of the AD converter 12 according to the sound pressure P output from the sound pressure calculation device 11, and 15 denotes an output device that displays or prints a determination result of the CPU 13.

Devices including a vibration source that generates noise, vibration, and the like include not only the transmission 1 but also other components such as an engine. FIG. 19 is a configuration diagram showing a general automobile engine (hereinafter, simply referred to as an engine), and 31 is an engine composed of the following 32 to 36 elements.

Reference numeral 32 denotes a crankshaft which is rotationally driven by the combustion of the fuel mixture, 33 denotes a piston connected to the crankshaft 32 and is depressurized when the fuel mixture explodes, and 34 denotes exhaust of the fuel mixture during intake and after combustion. A valve that opens and closes when gas is exhausted, 35 is a camshaft that makes one rotation while the crankshaft 32 makes two rotations, and 36 is a cam that is fixed to the camshaft 35 and controls the drive timing of the valve 34 and the like. A diagnostic device including the microphone 7 to the output device 15 in FIG. 18 may be provided for the engine 31 as shown in FIG.

Next, taking the case of the transmission 1 as an example, FIG.
The operation of the conventional device diagnostic device shown in FIG. First, when checking the transmission 1 for defects, the transmission 1
Is operated, the second shaft 3 engaged with the input shaft by operating the clutch 5 transmits the rotational force to the first shaft 2 via the gear 4 selected by the transmission mechanism 6. At this time, the sound S generated from the transmission 1 is detected by the microphone 7 and converted into an electric signal E. The electric signal E obtains an appropriate gain through the amplifier 9 and obtains an appropriate signal through the filter 10. After the frequency band is extracted, it is input to the sound pressure calculation device 11.

The sound pressure calculation device 11 calculates the effective value of the sound pressure P based on the electric signal E, and the AD converter 12 converts the sound pressure P into digital data and transfers the digital data to the CPU 13. At this time, the automatic range switching device 14
The range is automatically switched in order to execute the optimal AD conversion in 2.

The CPU 13 compares the digital data indicating the sound pressure P transferred from the AD converter 12 with a reference effective value (abnormality judgment reference) set in the CPU 13 and determines that the sound pressure P is the reference effective value. Is exceeded, it is determined that the transmission 1 has an abnormality, and the abnormality determination result is output to the output device 15. Thereby, the observer can recognize the abnormality of the transmission 1 and can deal with the abnormality before mounting the transmission 1 in the automobile.

However, the sound S generated from the transmission 1
Has a small difference between the normal state and the abnormal state. Further, the effective value of the generated sound S in the normal state of the transmission 1 may exceed the effective value in the abnormal state. Therefore, in the configuration of FIG. 18, it is difficult to accurately distinguish the normal state or the abnormal state of the transmission 1. The same applies to the case where the engine 31 (see FIG. 19) is targeted as another component.

[0014]

As described above, the conventional device diagnostic apparatus is abnormal only when the effective value of the sound S generated from the transmission 1 exceeds a certain alarm value (reference effective value), for example. Is determined, the difference between the effective value of the generated sound S in the normal state and the abnormal state is very small, and the effective value in the normal state may exceed the effective value in the abnormal state. Therefore, there is a problem that the abnormality cannot be accurately determined.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems, and distinguishes a small difference between a normal state and an abnormal state of sound or vibration generated from a device.
It is an object of the present invention to obtain a device diagnostic device with improved reliability of abnormality diagnosis.

[0016]

According to a first aspect of the present invention, there is provided a device diagnostic apparatus for detecting a sound or vibration generated from a device including a vibration element and outputting a signal corresponding to the sound or vibration. Means, a wavelet transform operation unit that performs wavelet transform on the signal and outputs transformed data, an abnormality detection processing device that detects an abnormality of the device based on the converted data, and a device state detection device that detects a predetermined state of the device.
Output means, and the signal detection means responds to a predetermined state.
Starting the operation, the abnormality detection processing device calculates a characteristic of the converted data and outputs a calculated value, a comparing unit that compares the calculated value with a preset reference value, and a comparison result of the comparing unit. Alarm means for outputting an alarm when the calculated value exceeds a predetermined relationship with respect to the reference value,
The reference value is set based on a signal obtained when the device is in a normal state.

[0017]

[0018]

According to a second aspect of the present invention, there is provided an apparatus diagnostic apparatus for detecting a sound or a vibration generated from an apparatus including a vibration element and outputting a signal corresponding to the sound or the vibration; A wavelet transform arithmetic unit for performing wavelet transform on the converted data and outputting the converted data, and an abnormality detection processing device for detecting an abnormality of the device based on the converted data, and the abnormality detection processing device calculates a characteristic of the converted data. And a visualization means for visualizing at least one of the converted data and the calculated value, provided with equipment state detection means for detecting a predetermined state of the equipment, wherein the signal detection means The operation starts in response.

[0020]

According to a third aspect of the present invention, in the diagnostic device for an instrument according to the first or second aspect , the calculating means is configured to calculate an average value of a distribution of energy components in a predetermined frequency component range with respect to the converted data. As an operation value.

[0022]

According to a fourth aspect of the present invention, there is provided a diagnostic apparatus for an apparatus according to the first or second aspect, wherein the calculating means is configured to calculate a skewness of a distribution of an energy component in a predetermined frequency component range with respect to the converted data. Is output as an operation value.

According to a fifth aspect of the present invention, there is provided an apparatus for diagnosing a device according to the first or second aspect, wherein the calculating means calculates a kurtosis of a distribution of an energy component in a predetermined frequency component range with respect to the converted data. Is output as a calculation value.

According to a sixth aspect of the present invention, in the diagnostic device for an instrument according to the first or second aspect , the calculating means is configured to determine a maximum value of a distribution of an energy component in a predetermined frequency component range with respect to the converted data. As a calculated value.

According to a seventh aspect of the present invention, there is provided the diagnostic apparatus for an apparatus according to the first or second aspect, wherein the arithmetic means includes a neural network for a distribution of energy components in a predetermined frequency component range with respect to the converted data. And a neural network computing unit that outputs the learning value of as a computed value.

[0027] According to an eighth aspect of the present invention, in the device diagnostic apparatus according to the first or second aspect , the calculating means is configured to calculate a probability density of a distribution of energy components in a predetermined frequency component range with respect to the converted data. It is composed of a probability density distribution function calculator for calculating a distribution function, and a feature extractor for outputting a feature of the probability density distribution function according to a fuzzy rule as a calculation value.

According to a ninth aspect of the present invention, in the diagnosing apparatus for a device according to the first or second aspect , the arithmetic means includes a variance value of a distribution of an energy component in a predetermined frequency component range with respect to the converted data. And a variance value deviation calculator that outputs a deviation between the calculated value and the maximum value as a calculated value.

[0029]

According to the first aspect of the present invention, a signal detected in response to a sound or vibration generated from a device is subjected to a wavelet transform, and a comparison result obtained when an operation value indicating the characteristic of the converted data is compared with a reference value. And outputs an alarm when the calculated value exceeds a predetermined relationship with the reference value. As a result, if only the signal exists in a normal state, the calculated value is close to the normal value, so that no alarm is erroneously output. In addition, a reference value is set based on a signal detected when the device is in a normal state, and the abnormality determination of the device is ensured. Furthermore, when a predetermined state of the device is detected,
Signal detection means to improve signal detection repeatability
And make sure that the device is properly judged.

[0030]

[0031]

According to a second aspect of the present invention, at least one of the converted data and the calculated value is visualized and displayed to an observer, thereby improving the reliability of the abnormality determination of the device and the predetermined state of the device. When the signal is detected, the operation of the signal detecting means is started to improve the reproducibility of signal detection and to reliably determine the abnormality of the device.

According to a fifth aspect of the present invention, when a predetermined state of the device is detected, the operation of the signal detecting means is started, and the reproducibility of signal detection is improved to ensure the abnormality determination of the device.

According to a third aspect of the present invention, an average value of an energy component distribution in a predetermined frequency component region is calculated for the converted data obtained by performing a wavelet transform on the detected signal, and the average value is set as a reference value. On the other hand, an alarm is output when a predetermined relationship is exceeded. As a result, if only the signal exists in a normal state, the average value is close to the normal value, so that no alarm is erroneously output.

[0035]

According to a fourth aspect of the present invention, the skewness of the energy component distribution in a predetermined frequency component region is calculated for the converted data obtained by performing the wavelet transform on the detected signal, and the skewness is set to a reference value. On the other hand, an alarm is output when a predetermined relationship is exceeded. As a result, if the signal exists only in a normal state, the skewness is close to a normal value.
No alarm is output erroneously.

According to a fifth aspect of the present invention, the kurtosis of the energy component distribution in a predetermined frequency component region is calculated for the converted data obtained by performing the wavelet transform on the detected signal, and the kurtosis is set to a reference value. On the other hand, an alarm is output when a predetermined relationship is exceeded. As a result, if the signal exists only in a normal state, the kurtosis becomes close to the normal value.
No alarm is output erroneously.

According to a sixth aspect of the present invention, the maximum value of the energy component distribution in a predetermined frequency component region is calculated with respect to the converted data obtained by performing the wavelet transform on the detected signal, and the maximum value is set as a reference value. On the other hand, an alarm is output when a predetermined relationship is exceeded. As a result, if only the signal exists in a normal state, the maximum value is close to the normal value, and the alarm is not erroneously output.

According to a seventh aspect of the present invention, a learning value of a neural network of an energy component distribution in a predetermined frequency component region is calculated for the converted data obtained by performing a wavelet transform on the detected signal, and the learning value is calculated. An alarm is output when a predetermined relationship with the reference value is exceeded. With this, if only the signal exists in a normal state,
Since the learning value of the neural network is close to the normal value, no alarm is erroneously output.

According to the eighth aspect of the present invention, a probability density distribution function of an energy component distribution in a predetermined frequency component region is calculated for the transform data obtained by performing a wavelet transform on the detected signal, and the probability density distribution function is calculated. When the feature according to the fuzzy rule exceeds a predetermined relationship with the reference value, an alarm is output. As a result, if the signal exists only in a normal state, the characteristic of the probability density distribution function according to the fuzzy rule becomes close to the normal value, so that no alarm is erroneously output.

According to a ninth aspect of the present invention, a deviation between a variance value and a maximum value of an energy component distribution in a predetermined frequency component region is calculated for converted data obtained by performing a wavelet transform on a detected signal, An alarm is output when the variance deviation exceeds a predetermined relationship with the reference value. As a result, if the signal exists only in a normal state, the variance value deviation is close to the normal value, so that no alarm is erroneously output.

[0042]

【Example】

Embodiment 1 FIG. Hereinafter, a first embodiment of the present invention will be described with reference to the drawings. 1 to 4 are block diagrams showing a schematic configuration of a first embodiment of the present invention, and FIG. 5 is a functional block diagram showing a specific configuration example of the abnormality detection processing device in FIGS.

In each figure, 1 to 7, 9, 10, 12,
S and E are the same as described above. A microphone 7 (see FIGS. 1 and 3) that detects a sound S generated from a device including a vibration element, for example, the transmission 1 and outputs an electric signal E corresponding to the generated sound S, includes an amplifier 9 that processes the electric signal E. , The filter 10 and the AD converter 12 together constitute a signal detecting means.

Reference numeral 16 denotes an amplifier 9, a filter 10, and an AD
This is a wavelet transform operation unit that performs a wavelet transform on the electric signal E via the transformer 12, and includes three-dimensional data Yij (see FIGS. 1 and 2) or column data Yi.
(See FIGS. 3 and 4) is output as conversion data.
Reference numeral 17 denotes a parameter setting device for externally setting parameters used in the wavelet transform operation device 16 and operation modes. Wavelet transform operation unit 1
6 performs a wavelet transform according to the parameters set by the parameter setting unit 17.

The three-dimensional data Yij (i = 1, 2,...,
N) (j = 1, 2,..., M) is a wavelet transform of time-series data corresponding to the voltage level of the electric signal E corresponding to the sound S or the vibration detected by the microphone 7 or the vibration detector 18. The digital data is obtained by performing time-series frequency component sequence data. Note that the three-dimensional data Yij is converted into two-dimensional digital string data Yi.
Can also be output.

Reference numeral 18 denotes a vibration detector (see FIGS. 2 and 4) provided in the transmission 1 in place of the microphone 7.
A vibration generated from the transmission 1 is detected, an electric signal E corresponding to the vibration is output, and a signal detection unit is configured together with the amplifier 9, the filter 10, and the AD converter 12.

Reference numeral 19 denotes a phase detector provided on the input shaft of the transmission 1, which constitutes a device state detecting means for detecting a predetermined state of the transmission 1, and has an arbitrary position (for example, phase) of the rotational phase. The detection signal is output at an angle of 10 °) and input to the AD converter 12 in the signal detection means. AD converter 1
2 starts the AD conversion operation of the electric signal E in response to a predetermined state indicated by the detection signal from the phase detector 19.

Reference numeral 20 denotes an abnormality detection processing device for detecting an abnormality of the transmission 1 based on the converted data (three-dimensional data Yij or column data Yi), and comprises a CPU for performing various arithmetic processing on the converted data. As shown in FIG.
24 and 26 elements. FIG. 5 shows a case where the converted data is three-dimensional data Yij, but the present invention can be similarly applied to column data Yi.

In FIG. 5, reference numeral 21 denotes three-dimensional data Yij
A computing unit that computes a feature of (conversion data) and outputs a computed value μ; 22 is a reference value setting device that sets a reference value μr previously learned based on the computed value μ; and 23 is a computed value μ and a reference value μr and outputs a comparison result. The reference value μr used by the comparator 23 is set by three-dimensional data Yij based on the electric signal E detected when the transmission 1 is in a normal state, as indicated by a broken line.

Reference numeral 24 denotes an alarm which operates based on the comparison result of the comparator 23, and outputs an alarm A when the calculated value μ exceeds a predetermined relation with the reference value μr. Reference numeral 26 denotes a visualizer for visualizing and displaying various data values and the like, and visualizes at least one of conversion data, that is, three-dimensional data Yij, characteristics of the conversion data, that is, an operation value μ, and a reference value μr.

The abnormality detection processing device 20 has two operation modes (setting mode of the reference value μr and diagnosis mode of the transmission 1) which can be switched by an external command based on a manual operation or the like. In FIG. 5, arrows formed by broken lines indicate the flow of each signal in the reference value setting mode, and arrows formed by solid lines indicate the flow of each signal in the diagnostic mode of the transmission 1 (device). Is the same in each embodiment described later.

Hereinafter, referring to FIG. 1 to FIG. 5, similarly to the above, as a device having a vibration source generating noise or the like,
Taking the transmission 1 as an example, a specific operation of the first embodiment of the present invention will be described.

As a signal detector for outputting a signal corresponding to sound or vibration generated from the device, that is, an electric signal E, even if the microphone 7 is used as shown in FIGS. As described above, the vibration detector 18 may be used. The conversion data input to the abnormality detection processing device 20 includes three-dimensional data Y as shown in FIGS.
ij, the column data Yi as shown in FIGS.
May be used.

First, the AD converter 12 has a phase detector 19
In response to a detection signal (operation command) from the controller, the electric signal E via the amplifier 9 and the filter 10 is sampled at equal intervals, and the time-series N column data (digital signal)
Is output.

Subsequently, the wavelet transform calculator 16
Performs a wavelet transform on the electric signal E converted into a digital signal by the AD converter 12 as in the following equation (1), and converts three-dimensional data Yij (or column data Yi) into converted data. Output.

Φ (to, a) = (1 / √a) ∫g [(t-to) / a] · X (t) (1)

In the equation (1), however, the operation range of the integral term is -∞ to 、, to represents a time series, and a represents a frequency component sequence. X (t) corresponds to the time-series data obtained from the electric signal E, and g [(t-to) /
a] corresponds to a function called a basis function.

The wavelet transform method based on equation (1) is described, for example, in Chapter 4, “Characteristics change with time,” of “Interface” by CQ Publishing Company (January 1994, written by Keiichiro Minami and Satoshi Kawata). New Signal Analysis Method (Time-Frequency Analysis) "(pages 117 to 119). Here, the basis function can be represented using a function such as MexicanHat as in the following equation (2).

[0059] ^{g (t) = (d 2} / dt 2) exp (-t 2/2) ... (2)

Thus, the three-dimensional data Yij is obtained by performing the wavelet transform of the equation (1) on the electric signal E detected at the time of diagnosis of the transmission 1. In the equation (1), for example, a frequency component sequence a
, The three-dimensional data is converted into the column data Yi.
Can also be output as

The arithmetic unit 21 in the abnormality detection processing device 20
The wavelet-transformed three-dimensional data Yij is fetched, and index data indicating the properties of the three-dimensional data Yij is output as a calculation value μ. The index calculated value μ indicating the property of the three-dimensional data Yij is, for example, the three-dimensional data Yij
Is an energy component at a predetermined frequency when frequency analysis is performed, and is represented by the following equation (3).

[0062] μ = F (ω) = ( 1 / 2π) ∫f (t) · e -j ω t ... (3)

However, in the equation (3), the operation range of the integral term is -∞ to ∞. Further, the reference value setting unit 22 stores data (computed values) learned in advance in a reference value setting mode (see a broken line in FIG. 5) based on the three-dimensional data Yij obtained when the transmission 1 is normal. Is output as the reference value μr. The reference value μr may be stored in advance in the reference value setting unit 22 using a transmission that becomes a normal sample.

On the other hand, in the diagnostic mode of the transmission 1 (see the solid line in FIG. 5), the comparator 23 compares the operation value μ from the operation unit 21 with the reference value μr from the reference value setting unit 22. A comparison result indicating whether the transmission 1 is abnormal is output.
That is, the comparator 23 determines that the calculated value μ (diagnosis data) has a predetermined relationship with the reference value μr (for example, the reference value μr
When the value exceeds 3), a comparison result indicating an abnormality of the transmission 1 is output.

Accordingly, the alarm 24 determines an abnormality of the transmission 1 in response to the comparison result from the comparator 23, and generates an alarm A. At this time, the visualizer 26 displays at least one or all of the three-dimensional data Yij, the operation value μ, and the reference value μr to the observer, so that the contents of the detection data and the operation data can be easily grasped.

For example, when the device to be diagnosed is the engine 31 (see FIG. 19), the visualizer 26 sets the operation timing of the intake valve of the second cylinder to 0.1 from the start of data measurement.
If it is later than second, the data for only 0.02 seconds after 0.1 second of the obtained three-dimensional data Yij is displayed to the observer. A plurality of such data display portions may be set, and may be arbitrarily switched according to a request.

The above-described series of operations in the diagnostic mode of the transmission 1 (device) are repeated, for example, in a fixed cycle. Thus, the electric signal E from the transmission 1 in the normal state is
Based on the above, the operation of the reference value setting mode is executed in advance, the learned reference value μr is stored, and then the operation of the diagnosis mode is executed at the time of device diagnosis, and the operation of each transmission 1 to be diagnosed is performed. A calculated value μ indicating the nature of the sound S or the vibration is sequentially obtained, and the calculated value μ is compared with a normal reference value μr.

As a result, when the electric signal E in the diagnostic mode indicates an abnormality of the transmission 1 and the difference between the calculated value μ and the reference value μr is large (exceeds a predetermined relationship), the alarm 24 indicates the abnormality. The alarm A shown can be reliably generated. On the other hand, the electric signal E in the diagnostic mode
Is the reference value μr
(Normal time), and the alarm 24 does not generate the alarm A.

As described above, by using the converted data (Yij or Yi) obtained by performing the wavelet transform on the electric signal E for diagnosis, the sound S or the vibration is divided into the time axis and the frequency component axis to obtain the energy component distribution. be able to.

That is, a change in the frequency component of the generated energy component over time of the electric signal E, a change in the energy component over time of the predetermined frequency component, and the like are found, and a subtle difference in sound quality or vibration is calculated from the calculated value μ. Can be diagnosed.

The above diagnosis is effective when the abnormality of the electric signal E is caused not by the sound pressure of the generated sound S but by a difference in sound quality, for example, in a case where the bearing of the transmission 1 is improperly assembled. Therefore, reliable diagnosis can be performed by relatively simple processing as shown in the above equation (3).

Further, if the generated sound S is detected by using the microphone 7 as shown in FIG. 1, the sound S can be detected in a non-contact manner, and adversely affects the surface shape and material of the diagnostic object. Diagnosis can be performed without giving
On the other hand, as shown in FIG.
Is directly detected, the influence of other noise and noise (background noise) can be prevented, and the S / N of the electric signal E can be reduced.
By increasing the ratio, the reliability of diagnosis can be improved.

In this case, the microphone 7 or the vibration detector 18 is used as the signal detecting means in the case where the sound S or vibration generated from the device is detected as an example. Alternatively, the electric signal E can be obtained by detecting light or heat.

Embodiment 2 FIG. In the first embodiment, the specific processing of the arithmetic unit 21 has not been described. For example, an average value may be calculated for the converted data. FIG.
Is a functional block diagram showing a configuration of an abnormality detection processing device 20 according to Embodiment 2 of the present invention using an average value calculator 21A as a calculation means, and Yi and A are the same as those described above. Also, 21A to 24A, 26A, μ1 and μ
1r corresponds to 21 to 24, 26, μ and μr in FIG. 5, respectively.

In FIG. 6, although the column data Yi is used as the conversion data input to the abnormality detection processing device 20, three-dimensional data Yij may be used. The peripheral configuration (not shown) may be any of the configurations shown in FIGS. The same applies to other embodiments described later.

In this case, the average value calculator 21A in the abnormality detection processing device 20 calculates the average value μ1 of the distribution of the energy component in the predetermined frequency component region with respect to the column data Yi (converted data) by the calculated value (electric signal (A feature of the column data Yi based on E).

The average value setter 22A stores the average value output from the average value calculator 21A in the reference value setting mode as the reference value μ1r, and the comparator 23A stores the average value μ1 and the reference value in the diagnostic mode. μ1r,
The alarm 24A outputs an alarm A in response to the comparison result of the comparator 23A, and the visualizer 26A displays the column data Yi, the average value μ1, the reference value μ1r, and the like.

Hereinafter, a specific operation of the second embodiment of the present invention shown in FIG. 6 will be described. First, as before,
In response to the detection signal from the phase detector 19 (see FIGS. 1 to 4), the AD converter 12 generates N column data obtained by sampling at equal intervals, and performs wavelet transform operation. Reference numeral 16 denotes data obtained by performing a wavelet transform on the column data, for example, column data Yi (i = 1, 2,
.., N).

Subsequently, the average value calculator 21A in the abnormality detection processor 20 calculates the column data Yi by the following equation (4).
Is calculated.

Μ1 = (1 / N) Σ (Yi) (4)

However, in the sum term in the equation (4), the column data Yi is added for i = 0, 1,..., N−1. In the reference value setting mode, the average value μ1 thus calculated is, as shown by the broken line in FIG.
It is stored in 2A as a reference value μ1r, and is input to the comparator 23A as indicated by a solid line in the diagnostic mode.

In the diagnostic mode, the comparator 23A outputs a comparison result indicating an abnormality, for example, when the average value μ1 exceeds three times the reference value μ1r, and operates the alarm device 24A to generate an alarm A. The above operation is repeated in a fixed cycle as described above, and the display contents of the visualizer 26A can be arbitrarily selected.

The diagnosis based on the average value μ1 is generally performed when the distribution of the energy component of the column data Yi is normal as in the case where the movable part is in contact with the casing of the transmission 1, for example. This is particularly effective when moving.
Further, a diagnosis result can be obtained by relatively simple processing using the wavelet transform as in the above equation (4).

Embodiment 3 FIG. Although the average value calculator 21A is used as the calculating means in the second embodiment, a variance value may be calculated for the converted data. FIG. 7 is a functional block diagram showing a configuration of an abnormality detection processing device 20 according to a third embodiment of the present invention using a variance value calculator 21B as a calculation means. Yi and A are the same as those described above.
24B, 26B, μ2 and μ2r correspond to 21 to 24, 26, μ and μr in FIG. 5, respectively.

In this case, the variance calculator 21B in the abnormality detection processor 20 takes in, for example, the column data Yi as conversion data, and applies the variance μ2 of the distribution of the energy component in the predetermined frequency component region to the column data Yi. Is output as an operation value. Further, the variance value setting unit 22B stores the variance value in the reference value setting mode as the reference value μ2r, the comparator 23B compares the variance value μ2 with the reference value μ2r in the diagnostic mode, and the alarm device 24B compares The alarm A is output in response to the result, and the visualizer 26B displays the column data Yi, the variance μ2, the reference value μ2r, and the like.

Hereinafter, a specific operation of the third embodiment of the present invention shown in FIG. 7 will be described. Variance calculator 21B
Is obtained from the column data Yi obtained in the same manner as described above.
After calculating the average value μ1 using the above equation (4), the variance value μ2 is further calculated using the following equation (5).

Μ2 = {(1 / N)} (Yi−μ1) ² } (5)

In the total term of the above equation (5), the column data Yi is added for i = 0, 1,..., N−1. Hereinafter, in the same manner as described above, the variance value μ2 is stored in the variance value setting unit 22B in the reference value setting mode (see the broken line), and is input to the comparator 23B together with the reference value μ2r in the diagnostic mode (see the solid line). For example, the variance μ2
Exceeds three times the reference value μ2r (when an abnormality is determined), an alarm A is output.

In the diagnosis using the variance μ2 of the converted data, the distribution of the energy components of the column data Yi is dispersed as in the case of the gear eccentricity of the gear 4 in the transmission 1 (for example). This is particularly effective when the energy is large as a whole. Further, a diagnosis result can be obtained by a relatively simple process as shown in Expression (5) using the wavelet transform.

Embodiment 4 FIG. In the third embodiment, the variance calculator 21B is used as the calculating means, but the skewness may be calculated on the converted data. FIG. 8 is a functional block diagram showing a configuration of an abnormality detection processing device 20 according to a fourth embodiment of the present invention using a skewness calculator 21C as a calculating means. Yi and A are the same as those described above.
24C, 26C, μ3 and μ3r are respectively shown in FIG.
21 to 24, 26, μ and μr.

In this case, the skewness calculator 21C in the abnormality detection processing device 20 fetches, for example, the column data Yi as conversion data, and adds the column data Yi to the skewness μ3 of the distribution of the energy component in the predetermined frequency component region. Is output as an operation value. The skewness setting unit 22C stores the skewness in the reference value setting mode as a reference value μ3r, and
C compares the skewness μ3 with the reference value μ3r in the diagnostic mode, the alarm device 24C outputs an alarm A in response to the comparison result, and the visualizer 26C outputs the column data Yi, the skewness μ3, and the reference value μ3r. And so on.

Hereinafter, a specific operation of the fourth embodiment of the present invention shown in FIG. 8 will be described. The skewness calculator 21C is:
After calculating the average value μ1 and the variance value μ2 for the column data Yi using the above-described equations (4) and (5), the skewness μ3 is further calculated using the following equation (6).

Μ3 = (1 / N) · {1 / (μ2) ³ } (Yi−μ1) ³ (6)

In the total term of the above equation (6), the column data Yi is added for i = 0, 1,..., N−1. Hereinafter, the skewness μ3 is stored in the skewness setter 22C in the reference value setting mode (see the broken line), and is input to the comparator 23C together with the reference value μ3r in the diagnostic mode (see the solid line). When the value exceeds three times the reference value μ3r (when an abnormality is determined), an alarm A is output.

In the diagnosis using the skewness μ3 of the converted data, the distribution of the energy component of the column data Yi has a large deviation, for example, as in the case where the movable portion is in contact with the casing of the transmission 1. It is particularly effective when it occurs. Further, a diagnosis result can be obtained by relatively simple processing using the wavelet transform as in the above equation (6).

Embodiment 5 FIG. In the fourth embodiment, the skewness calculator 21C is used as the calculating means, but the kurtosis may be calculated on the converted data. FIG. 9 is a functional block diagram showing a configuration of an abnormality detection processing device 20 according to Embodiment 5 of the present invention using a kurtosis calculator 21D as a calculation means.
Yi and A are the same as described above, and 21D to 24
D, 26D, μ4 and μ4r correspond to 21 to 24, 26, μ and μr in FIG. 5, respectively.

In this case, the kurtosis calculator 21D in the abnormality detection processing device 20 fetches, for example, the column data Yi as conversion data, and adds the kurtosis μ4 of the distribution of the energy component in the predetermined frequency component region to the column data Yi. Is output as an operation value. Further, the kurtosis setting unit 22D stores the kurtosis in the reference value setting mode as the reference value μ4r, and
D compares the kurtosis μ4 with the reference value μ4r in the diagnosis mode, the alarm device 24D outputs an alarm A in response to the comparison result, and the visualizer 26D outputs the column data Yi, the kurtosis μ4, and the reference value μ4r. And so on.

The specific operation of the fifth embodiment of the present invention shown in FIG. 9 will be described below. The kurtosis calculator 21D is
After calculating the average value μ1 and the variance value μ2 for the column data Yi using the above-described expressions (4) and (5), the kurtosis μ4 is further calculated using the following expression (7).

Μ4 = (1 / N) · {1 / (μ2) ⁴ } (Yi−μ1) ³ } −3 (7)

In the total term of the above equation (7), the column data Yi is added for i = 0, 1,..., N−1. Hereinafter, the kurtosis μ4 is stored in the kurtosis setting unit 22D in the reference value setting mode (see the broken line), and is input to the comparator 23D together with the reference value μ4r in the diagnosis mode (see the solid line). An alarm A is output when the value exceeds three times the reference value μ4r (when an abnormality is determined).

In the diagnosis using the kurtosis μ4 of the converted data, the distribution of the energy component of the column data Yi has a large sharpness or spread, for example, such as the gear eccentricity of the gear 4 in the transmission 1. This is particularly effective in cases. Further, the diagnosis result can be obtained by a relatively simple process using the wavelet transform as in the above equation (7).

Embodiment 6 FIG. In the fifth embodiment, the kurtosis calculator 21D is used as the calculation means, but a maximum value may be calculated for the converted data. FIG. 10 is a functional block diagram showing a configuration of an abnormality detection processing device 20 according to a sixth embodiment of the present invention using a maximum value calculator 21E as a calculation means. Yi and A are the same as those described above.
２４24E, 26E, μ5 and μ5r correspond to 21 to 24, 26, μ and μr in FIG. 5, respectively.

Hereinafter, a sixth embodiment of the present invention shown in FIG.
A specific operation will be described. In this case, the maximum value calculator 21E in the abnormality detection processing device 20 fetches, for example, the column data Yi as the conversion data, and calculates the maximum value μ5 of the distribution of the energy component in the predetermined frequency component region for the column data Yi. , And outputs this as an operation value.

The maximum value setting unit 22E stores the maximum value obtained in the reference value setting mode as the reference value μ5r (see the broken line), and the comparator 23E stores the maximum value μ5 and the reference value in the diagnostic mode (see the solid line). μ5r, the alarm 24E indicates that the maximum value μ5 is equal to the reference value μ5r of 3
Alarm A in response to the comparison result indicating an abnormality when the number exceeds double
Is output. Further, the visualizer 26E outputs the column data Yi,
The maximum value μ5 and the reference value μ5r are displayed.

The diagnosis using such a maximum value μ5 of the converted data is performed only when the column data Yi has a large energy component only instantaneously on a long time axis, such as a dent of the gear 4 in the transmission 1. This is particularly effective in the case where the error occurs. In addition, a diagnosis result can be obtained by a relatively simple process of calculating a maximum value using a wavelet transform.

Embodiment 7 FIG. Although the maximum value calculator 21E is used as the calculating means in the sixth embodiment, a learning value of the neural network may be calculated on the converted data. FIG. 11 is a functional block diagram showing a configuration of an abnormality detection processing device 20 according to a seventh embodiment of the present invention using a neural network calculator 21F as a calculation means.
i and A are the same as those described above;
F, 26F, N1 and N1r correspond to 21 to 24, 26, μ and μr in FIG. 5, respectively.

Hereinafter, the seventh embodiment of the present invention shown in FIG.
A specific operation will be described. In this case, the neural network computing unit 21F in the abnormality detection processing device 20
Is a learning value N1 (0 <0) indicating the degree of conformity of the distribution shape of the energy component in a predetermined frequency component region using, for example, the column data Yi as conversion data and using a neural network constructed in advance for the column data Yi.
N1 <1) is output as a calculation value.

More specifically, the neural network computing unit 21F is constituted by a unit having a sigmoid characteristic with respect to a pair of column data Yi as input data and a learning value N1 as output data (teacher signal). Using a multi-layer network and a well-known method called an error back propagation learning method, considering a learning error evaluation function, a value obtained by a gradient method to obtain a connection weight that minimizes the error evaluation function is used as a learning value. The calculation is performed as N1.

The reference value setting unit 22F of the neural network stores the learning value of the neural network in the reference value setting mode as the reference value N1r (see the broken line). The comparator 23F compares the learning value N1 with the reference value N1r in the diagnostic mode (see the solid line), and, for example, when the learning value N1 falls below 1/2 times the reference value N1r (the degree of matching is less than 1/2). At the time, a comparison result indicating an abnormality is output.

In response to the comparison result, the alarm 24F outputs the alarm A, and the visualizer 26F displays the column data Yi, the learning value N1 of the neural network, the reference value N1r, and the like. The diagnosis using the learning value N1 of the neural network for such conversion data is performed by using the column data Y
Since it is only necessary to learn the neural network from i, there is no need to determine the signal processing and calculation method in advance,
The arithmetic processing can be simplified, and the reliability of the diagnosis result can be improved.

Embodiment 8 FIG. In the seventh embodiment, the learning value N1 of the neural network is calculated by using the neural network calculator 21F as the calculating means. However, the probability density distribution function is calculated on the transformed data, and the fuzzy function based on the probability density distribution function is calculated. A feature according to a rule (a reference value in a membership function) may be used.

FIG. 12 is a functional block diagram showing a configuration of an abnormality detection processing device 20 according to Embodiment 8 of the present invention using a probability density distribution function calculator 25G and a feature extractor 21G as calculation means. Same as above, 21G-24G, 26G, F2 and F2r
Are 21 to 24, 26, μ and μ in FIG. 5, respectively.
r.

A probability density distribution function calculator 25G takes in the column data Yi as conversion data.
, The probability density distribution function F1 of the energy component distribution in the predetermined frequency component region is calculated. Reference numeral 21G denotes a feature extractor which digitizes the feature F2 of the probability density distribution function F1 according to the fuzzy rule and outputs the result as an operation value, and constitutes an arithmetic unit together with the probability density distribution function calculator 25G.

F2a is a membership reference function given externally in the reference value setting mode (see the broken line),
2G is a membership function setting unit that stores a membership function (reference value of fuzzy rule) F2r based on the feature F2 in the reference value setting mode and the membership reference function F2a.

The probability density distribution function calculator 25G calculates, for example, a function representing the distribution of each amplitude value (probability indicating what value the amplitude of one irregular waveform appears at a certain time, etc.) The calculation is performed as the density distribution function F1. Further, the probability density distribution function F1 can be obtained based on various time-series distribution characteristics regarding the column data Yi, such as, for example, amplitude (corresponding to a voltage level) and frequency count.

Further, the probability density distribution function calculator 25G is
Since the column data Yi is digitally processed, the range of the amplitude is divided into, for example, equal intervals, and the probability that the amplitude value is included in each divided range is obtained. As a result, the data of the probability density distribution function F1 becomes a data string corresponding to the number of divisions of the amplitude range. Specifically, for example, as a ratio of the number of data belonging to each division range in the column data Yi, Desired.

In this case, the comparator 23G includes the feature extractor 2
Feature F symbolized by fuzzy rules in 1G
2 and the membership function F2r for executing the fuzzy rule, and the alarm 24G
Alarm A is generated based on the comparison result of
Displays column data Yi, probability density distribution function F1, feature F2, reference value F2r, and the like.

FIG. 13 is an explanatory diagram showing an example of a reference signal used for obtaining the probability density distribution function F1, wherein the horizontal axis represents time t, and the vertical axis represents a count of a predetermined amplitude or the like for each time based on column data Yi. Is a number. Note that the count number can be composed of various elements as described above.

FIG. 14 is an explanatory diagram showing the probability density distribution function F1 obtained based on the reference signal shown in FIG. 13. The horizontal axis is the count number, and the vertical axis is the number of appearances for each count number. Here, as the probability density distribution function F1 corresponding to the reference value (normal value) N1r, a case where the number of counts “100” is “50” and the number of counts “90” is “10” is shown.

FIGS. 15 and 16 show the probability density distribution function F
FIG. 15 is an explanatory diagram showing an example of a membership function F2r of a fuzzy rule obtained based on No. 1 (see FIG. 14);
The horizontal axis represents the number of counts (90 and 100), and the vertical axis represents the membership function value (0 to 1) for determining normal or abnormal. Here, when the membership function value is 0 to 0.5, it is determined to be abnormal,
When it is 0.5 to 1, it is determined to be normal.

FIG. 15 shows the membership function F2r (90) when the count number is “90”. As is apparent from FIG. 14, when the number is “10”, the membership function value becomes the maximum value “1”. ". FIG. 16 shows the membership function F2r (1) when the count number is “100”.
00), and when the number is “50”, the membership function value becomes the maximum value “1”.

Each membership function F2r is obtained, for example, by adding a feature F2 according to a fuzzy rule to the membership criterion function F2a. The number of features F2 is represented, for example, as a numerical value of the probability density of an appropriate amplitude energy region, but an arbitrary numerical value can be selected as needed. Further, the membership criterion function F2a is not limited to the shapes shown in FIGS. 15 and 16, and may be arbitrarily selected.

Next, the operation of the eighth embodiment of the present invention shown in FIG. 12 will be described with reference to FIGS. 1 to 4 and FIGS. First, the probability density distribution function calculator 2
5G calculates a probability density distribution function F1 for the wavelet-transformed time-series column data Yi,
The feature extractor 21G calculates the feature F2 of the probability density distribution function F1.
Is extracted.

In the reference value setting mode, the extracted feature F2 is input to the membership function setting unit 22G together with the membership reference function F2a (see the broken line in FIG. 12), and the membership function, that is, the reference of the fuzzy rule is set. Stored as value F2r. In the diagnostic mode, the signal is input to the comparator 23G (see the solid line in FIG. 12).

In the diagnostic mode of the device, the feature extractor 2
1G extracts, for example, the number data (see FIG. 14) for the count numbers “100” and “90” from the probability density distribution function F1 as the feature F2. Subsequently, the comparator 23G includes membership functions F2r (90) (see FIG. 15) and F2r (100) serving as reference values of the fuzzy rules stored in the membership function setting unit 22G.
The degree of conformity of the feature F2 to (see FIG. 16) is compared.

At this time, a normal reference signal (see FIG. 13)
(The horizontal axis) of the membership functions F2r (90) and F2r (100) (see FIGS. 15 and 16) obtained from the probability density distribution function F1 of FIG.
2) is “10” and “50”, respectively.

Now, the sound S generated by the transmission 1 to be diagnosed will be described.
The feature F2 extracted from the probability density distribution function F1 of the number “8” and the count “10”
It is assumed that the number “49” for “0” is shown. These numbers “8” and “49” are converted to the membership functions F2r (90) and F2r (10
0) When compared above, the function values are respectively “0.8”
And "0.95".

Therefore, 0.5 <“the value of the membership function F2r (90)” ≦ 1.0 and 0.5 <
“Value of membership function F2r (100)” ≦ 1.
0 at the same time, the transmission 1 to be diagnosed is
Is determined to be normal. On the other hand, if at least one of the membership function values is 0.5 or less,
The transmission 1 to be diagnosed is determined to be abnormal, and the
In response to the abnormality determination result from G, an alarm A is generated from the alarm device 24G.

As described above, a series of operations in the diagnostic mode are repeated in a fixed cycle, and the fitness indicating the nature of the sound S generated by each transmission 1 is sequentially determined. At this time, since the logical signal processing and the determination are performed by the membership function F2r applied to the fuzzy rule, the reliability can be further improved. Here, as the feature F2, the numbers corresponding to the count numbers “90” and “100” are extracted. However, it is needless to say that other arbitrary features can be extracted and the number of extractions can be arbitrarily selected.

Embodiment 9 FIG. In the eighth embodiment, as index data of the column data Yi based on the electric signal E,
Although the feature of the fuzzy rule based on the probability density distribution function is used, a deviation between the variance and the maximum value of the column data Yi may be used.

FIG. 17 is a functional block diagram showing the configuration of an abnormality detection processing device 20 according to the ninth embodiment of the present invention using a variance value deviation calculator 21H as a calculation means. Yi and A are the same as those described above. Yes, 21H-24H, 26
H, μ6 and μ6r are 21 to 2 in FIG. 5, respectively.
4, 26, μ and μr.

The ninth embodiment of the present invention shown in FIG.
A specific operation will be described. In this case, the variance deviation calculator 21H calculates the variance deviation μ6 between the variance μ2 and the maximum value μ5 of the distribution of the energy components in the predetermined frequency component range with respect to the column data Yi (converted data) by the following equation ( 8), and outputs this as an operation value.

Μ6 = | μ5 | −μ2 (8)

In the above equation (8), the variance μ2 is obtained from the above equations (4) and (5), and the maximum value μ5 of the column data Yi is also obtained as described above.

Subsequently, the reference value setting unit 22H compares the dispersion value deviation obtained in the reference value setting mode with the reference value μ6r
(See the broken line), and the comparator 23H outputs the variance deviation μ6 and the reference value μ6r in the diagnostic mode (see the solid line).
The alarm 24H outputs, for example, the variance value deviation μ6
Is greater than three times the reference value μ6r, an alarm A is output in response to the comparison result indicating an abnormality. Also, visualizer 26H
Represents column data Yi, variance value deviation μ6, and reference value μ6r
And so on.

Diagnosis using the variance value μ6 of the converted data is based on the fact that the column data Yi indicates that the average energy on a long time axis, such as when the gear 4 of the transmission 1 has a defect or the like. This is particularly effective when a large energy component is generated only instantaneously as compared with the generation of. Further, a diagnosis result can be obtained by a relatively simple process such as Expression (8) using the wavelet transform.

In each of the above embodiments, the visualizer (26, 26A to 26H) converts the wavelet-transformed data (Yi or Yi) based on the electric signal E.
j) and operation values (μ, μ1 to
μ6, N1 or F2) and reference values (μr, μ1r to μ
6r, N1r or F2r), any number of data may be selectively displayed.

Embodiment 10 FIG. Further, by displaying the conversion data and the arithmetic unit, the comparators (23, 23A to 23A) are displayed.
H) and alarms (24, 24A to 24H) are omitted,
It is also possible to leave the judgment of the equipment abnormality to the observer.

In each of the above embodiments, the transmission 1 is used as a diagnosis target, and the generated sound S is used as a parameter. However, other devices such as the engine 31 (see FIG. 19) are used as targets, and other parameters (vibration, It is needless to say that the same effect can be obtained when detecting an electromagnetic wave or the like.

[0140]

As described above, according to the first aspect of the present invention, signal detecting means for detecting a sound or vibration generated from a device including a vibration element and outputting a signal corresponding to the sound or vibration, A wavelet transform operation unit that performs wavelet transform on a signal and outputs transformed data, and an abnormality detection processing device that detects a device abnormality based on the transformed data
And a device state detecting means for detecting a predetermined state of the device , wherein the signal detecting means starts operation in response to the predetermined state.
The abnormality detection processing device computes a characteristic of the converted data and outputs a computed value, a comparing unit that compares the computed value with a preset reference value, and computes based on a comparison result of the comparing unit. Alarm means for outputting an alarm when the value exceeds a predetermined relationship with respect to the reference value, and outputting an alarm only when the calculated value indicating the characteristic of the converted data indicates an abnormality, and erroneously outputting the alarm And the reference value is set based on the signal obtained when the device is in a normal state .
In addition, it is possible to obtain a diagnostic device for a device in which a minute difference in a state of sound or vibration generated from the device is determined, reliability of abnormality diagnosis is improved, and abnormality determination is reliably performed.

[0141]

[0142]

According to a second aspect of the present invention, a signal detecting means for detecting a sound or a vibration generated from a device including a vibration element and outputting a signal corresponding to the sound or the vibration, and a wavelet for the signal. A wavelet transform operation unit that performs conversion and outputs the converted data; and an abnormality detection processing device that detects an abnormality of the device based on the converted data. The abnormality detection processing device calculates characteristics of the converted data and calculates an operation value. An arithmetic unit for outputting, and a visualizing unit for visualizing at least one of the converted data and the calculated value, enable an observer to determine an abnormality based on the contents of the converted data and the calculated value, and change a predetermined state of the device. The device state detecting means for detecting is provided, and the signal detecting means starts operation in response to a predetermined state, so that the circuit configuration is simplified and abnormality determination is performed. Increasing reliability, diagnostic apparatus of equipment to ensure detection reproducibility improves abnormality determination signal is the effect obtained.

[0144]

According to a third aspect of the present invention, in the first or second aspect , the arithmetic means is configured to calculate an average value of a distribution of energy components in a predetermined frequency component range with respect to the converted data as an arithmetic value. Since it is constituted by the output average value calculator, there is an effect that a diagnostic device for equipment with improved reliability of abnormality diagnosis can be obtained.

[0146]

According to a fourth aspect of the present invention, in the first or second aspect , the arithmetic means is configured to calculate a skewness of a distribution of an energy component in a predetermined frequency component range with respect to the converted data as an arithmetic value. Since it is constituted by the skewness calculator that outputs, there is an effect that a diagnostic device for equipment with improved reliability of abnormality diagnosis can be obtained.

According to a fifth aspect of the present invention, in the first or second aspect , the arithmetic means is configured to calculate a kurtosis of an energy component distribution in a predetermined frequency component range with respect to the converted data as an arithmetic value. Since it is constituted by the output kurtosis calculator, there is an effect of obtaining a device diagnostic apparatus with improved reliability of abnormality diagnosis.

According to a sixth aspect of the present invention, in the first or second aspect , the arithmetic means is configured to calculate a maximum value of a distribution of energy components in a predetermined frequency component range with respect to the converted data as an arithmetic value. Since it is constituted by the maximum value calculator for outputting, there is an effect that a diagnostic device for a device with improved reliability of abnormality diagnosis can be obtained.

According to a seventh aspect of the present invention, in the first or the second aspect , the arithmetic means is arranged to calculate a learning value of a neural network distribution of an energy component in a predetermined frequency component range with respect to the converted data. Since it is constituted by the neural network computing unit that outputs as a computed value, there is an effect that a diagnostic device for equipment with improved reliability of abnormality diagnosis can be obtained.

According to an eighth aspect of the present invention, in the first or second aspect , the calculating means calculates a probability density distribution function of a distribution of an energy component in a predetermined frequency component range with respect to the converted data. And a feature extractor that outputs the fuzzy rule of the probability density distribution function as a calculation value, so that a diagnostic device for a device with improved reliability of abnormality diagnosis can be obtained. There is.

According to a ninth aspect of the present invention, in the first or second aspect , the arithmetic means is arranged to determine, with respect to the converted data, a variance value and a maximum value of a distribution of an energy component in a predetermined frequency component range. Is provided by a variance value deviation calculator that outputs the deviation of the calculated value as a calculation value. Therefore, there is an effect that a diagnostic device for a device with improved reliability of abnormality diagnosis can be obtained.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a schematic configuration when a microphone is used as signal detection means and three-dimensional data is used as conversion data in Embodiment 1 of the present invention.

FIG. 2 is a block diagram showing a schematic configuration when a vibration detector is used as signal detection means and three-dimensional data is used as conversion data in the first embodiment of the present invention.

FIG. 3 is a block diagram showing a schematic configuration when a microphone is used as signal detection means and column data is used as conversion data in the first embodiment of the present invention.

FIG. 4 is a block diagram showing a schematic configuration when a vibration detector is used as signal detection means and column data is used as conversion data in the first embodiment of the present invention.

FIG. 5 is a functional block diagram illustrating a configuration example of an abnormality detection processing device according to the first embodiment of the present invention.

FIG. 6 is a functional block diagram illustrating a configuration example of an abnormality detection processing device according to a second embodiment of the present invention.

FIG. 7 is a functional block diagram illustrating a configuration example of an abnormality detection processing device according to a third embodiment of the present invention.

FIG. 8 is a functional block diagram illustrating a configuration example of an abnormality detection processing device according to a fourth embodiment of the present invention.

FIG. 9 is a functional block diagram illustrating a configuration example of an abnormality detection processing device according to a fifth embodiment of the present invention.

FIG. 10 is a functional block diagram illustrating a configuration example of an abnormality detection processing device according to a sixth embodiment of the present invention.

FIG. 11 is a functional block diagram illustrating a configuration example of an abnormality detection processing device according to a seventh embodiment of the present invention.

FIG. 12 is a functional block diagram illustrating a configuration example of an abnormality detection processing device according to an eighth embodiment of the present invention.

FIG. 13 is a waveform chart showing an example of a reference signal in a normal state used in the probability density distribution function calculator in FIG.

14 is a waveform diagram showing an example of a probability density distribution function in a normal state obtained from the reference signal of FIG.

15 is a waveform diagram showing an example of a membership function of a fuzzy rule used in the comparator in FIG.

16 is a waveform chart showing an example of a membership function of a fuzzy rule used in the comparator in FIG.

FIG. 17 is a functional block diagram illustrating a configuration example of an abnormality detection processing device according to a ninth embodiment of the present invention.

FIG. 18 is a block diagram showing a schematic configuration of a conventional device diagnostic device.

FIG. 19 is a configuration diagram showing another example (engine) of a general device.

[Explanation of symbols]

1 transmission (equipment), 7 microphone (signal detection means), 12 A / D converter (signal detection means), 16 wavelet transform calculator, 18 vibration detector (signal detection means), 19 phase detector (equipment state detection means) ), 20
Anomaly detection processor, 21 calculator, 21A average calculator, 21B variance calculator, 21C skewness calculator, 21
D kurtosis calculator, 21E maximum value calculator, 21F neural network calculator, 21G feature extractor, 21
H dispersion value deviation calculator, 22, 22A-22H reference value setting device, 23, 23A-23H comparator, 24, 24A
-24H alarm, 25 probability density distribution function calculator, 2
6, 26A-26H visualizer, 31 engine (equipment), A alarm, E electric signal, F1 probability density distribution function, F2 feature, learning value of N1 neural network, S generated sound, Yi column data, Yij three-dimensional data, μ operation value, μ1 average value, μ2 variance value, μ3
Skewness, μ4 kurtosis, μ5 maximum value, μ6 variance deviation,
F2r, N1r, μr, μ1r to μ6r Reference values.

──────────────────────────────────────────────────続 き Continued on the front page (58) Fields surveyed (Int. Cl. ⁷ , DB name) G01M 19/00 G01H 17/00 G01M 17/007

Claims (9)

(57) [Claims] 1. A signal detecting means for detecting a sound or a vibration generated from a device including a vibration element and outputting a signal corresponding to the sound or the vibration, and performing a wavelet transform on the signal to output conversion data. to a wavelet transform operator, an abnormality detection apparatus for detecting an abnormality of the equipment based on the converted data, and a device state detection means for detecting a predetermined state of said apparatus, said signal detecting means, the predetermined Open action in response to status
First, the abnormality detection processing device calculates a characteristic of the conversion data and outputs a calculated value, a calculating unit that compares the calculated value with a preset reference value, and a comparison of the comparing unit. Alarm means for outputting an alarm when the calculated value exceeds a predetermined relationship with the reference value based on the result , wherein the reference value is a signal obtained when the device is in a normal state.
A diagnostic device for a device, which is set based on a signal. 2. A sound generated from a device including a vibrating element or
Detects vibration and outputs the sound or a signal corresponding to the vibration.
A signal detection means for outputting, and wavelet transform for the signal, and transform data
A wavelet transform arithmetic unit for outputting an error, and an error detecting an abnormality of the device based on the converted data.
A normal detection processing device, wherein the abnormality detection processing device calculates a characteristic of the conversion data and outputs a calculated value.
Means and at least one of the conversion data and the operation value
And visualization means for visualizing, and device state detection means for detecting a predetermined state of the device.
The signal detecting means operates in response to the predetermined state.
Diagnostic device of the equipment, characterized in that to start. 3. The computing means according to claim 1 , wherein
The distribution of the energy components in the predetermined frequency component range.
Structure the average value calculator for outputting average value as the calculated value
3. The method according to claim 1 or 2,
Diagnostic apparatus of the mounting of the device. 4. The calculation means according to claim 1 , wherein
The distortion of the energy component distribution in the predetermined frequency component range.
And a skewness calculator that outputs the degree as the calculated value.
Diagnostic device apparatus according to claim 1 or claim 2, characterized in that the. 5. The computing means according to claim 1 , wherein
The peak of the energy component distribution in the predetermined frequency component range
A kurtosis calculator that outputs the degree as the calculated value.
3. The device diagnosis apparatus according to claim 1, wherein 6. The calculation means according to claim 1 , wherein
The distribution of the energy components in the predetermined frequency component range.
The maximum value calculator outputs a large value as the calculated value.
The apparatus diagnosis apparatus according to claim 1 , wherein the apparatus diagnosis is performed. 7. The computing means according to claim 1 , wherein
The energy component distribution in the predetermined frequency component range.
Output the learning value of the neural network as the calculated value
Of the neural network
Diagnostic device apparatus according to claim 1 or claim 2, wherein the door. Wherein said computing means, to said conversion data, error in the predetermined frequency component area
Probability density to compute probability density distribution function of distribution of energy component
The degree distribution function calculator and the characteristics of the probability density distribution function based on fuzzy rules are described above.
And a feature extractor that outputs the calculated value .
3. The device diagnostic apparatus according to claim 1, wherein 9. The calculation means according to claim 1 , wherein
The distribution of the energy component in the predetermined frequency component range.
Variance that outputs the deviation between the scattered value and the maximum value as the calculated value
A value deviation calculator.
An apparatus for diagnosing a device according to claim 1 .