JPH0894497A - Automotive component diagnostic equipment - Google Patents

Abstract

(57) [Abstract] [Purpose] Using a relatively simple process called skewness calculation, it is possible to accurately discriminate a minute difference between normal and abnormal sounds generated by automobile components, and prevent defective products from leaking to the market. Thus, a diagnostic device for automobile components, which greatly contributes to quality improvement of automobile production and improves reliability of abnormality diagnosis, is obtained. [Structure] A microphone that detects sound generated from automobile components and outputs an electric signal according to the sound,
An abnormality detection processing device 20 for detecting an abnormality of an automobile component based on an electric signal, and the abnormality detection processing device 21 calculates a skewness computing means 21 for computing a skewness μ3 of the electrical signal.
And a comparing means 23 for comparing the skewness with a reference value μ3r. Based on the comparison result of the comparing means, when the skewness exceeds a predetermined relationship with the reference value, an alarm A is output and a detection sound is output. Is only the sound that exists in the normal state, and if the skewness is close to the normal value, the alarm is not erroneously output.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a vehicle component diagnostic apparatus for detecting a sound generated from a vehicle component such as a transmission and an engine (hereinafter simply referred to as a component) to detect an abnormality of the component, The present invention relates to an automobile component diagnostic device with improved reliability of abnormality diagnosis.

[0002]

2. Description of the Related Art In general, components mounted on mass-produced automobiles are required to be checked for defects at the production stage to prevent defective products from leaking to the market. This is because, for example, if a defective component caused by defective processing or assembly of a component is mounted on an automobile and used,
This is because abnormal noise is generated from the component, the component may be broken, and there is a risk of causing a serious accident due to the failure.

Therefore, in order to deal with the case where a defective product is produced in a component, it is necessary to detect an abnormality in the component and take a measure in advance before mounting the component on an automobile. The above is extremely important for quality control of component production.

FIG. 20 shows, for example, the Society of Automotive Engineers,
"Onomatopoeia investigation report of car, mainly on passenger car noise, March 1992", edited by Vibration and Noise Division Committee (19
It is a block diagram which shows the conventional diagnostic apparatus of the component for motor vehicles described in April 1, 1992). In the figure, reference numeral 1 denotes a vehicle component such as a transmission, which is configured by the following elements 2 to 6.

Reference numeral 2 designates, for example, a first shaft serving as an output shaft, 3 a second shaft for transmitting a rotational force to the first shaft 2, and 4 a first shaft 2 and a second shaft 3. It is a plurality of gears that are engaged at different gear ratios. Reference numeral 5 denotes a clutch provided corresponding to, for example, the input shaft, and selectively disengages the input shaft from the second shaft 3. 6 is a speed change mechanism (corresponding to a shift lever) for selecting the gear 4 of the transmission 1 by manual operation.
Is.

Reference numeral 7 denotes a microphone that detects a sound S generated from the transmission 1 including the first shaft 2 to the transmission mechanism 6 and outputs an electric signal E corresponding to the detected sound S. Reference numeral 9 amplifies the electric signal E. The amplifier 10 is a filter for passing a predetermined frequency component of the amplified electric signal E to be detected, 11
Is a sound pressure calculation device that calculates the sound pressure P from the electric signal E that has passed through the filter 10.

Reference numeral 12 is an 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 is a CPU (central processing unit) for judging an abnormality of the transmission 1 based on the digitally converted sound pressure P,
Reference numeral 14 is an automatic range switching device that sets the optimum range of the AD converter 12 according to the sound pressure P output from the sound pressure calculation device 11, and 15 is an output device that displays or prints the determination result of the CPU 13.

Further, examples of automobile components that generate noise include not only the transmission 1 but also an engine, for example. FIG. 21 is a configuration diagram showing a general automobile engine (hereinafter, simply referred to as engine).
Reference numeral 1 is an engine composed of the following 32-36 elements.

Reference numeral 32 is a crankshaft which is driven to rotate by combustion of the fuel mixture, 33 is a piston which is connected to the crankshaft 32 and is depressed when the fuel mixture explodes, and 34 is a piston after intake of the fuel mixture and after combustion. A valve that opens and closes when exhaust gas is exhausted, a cam 35 that makes one rotation while the crankshaft 32 makes two rotations, and a cam 36 that is fixed to the camshaft 35 and that controls the drive timing of the valve 34 and the like. Figure 2
For the engine 31 such as the one described above, the diagnostic device including the microphone 7 to the output device 15 in FIG. 20 may be provided.

Next, the operation of the conventional diagnostic apparatus for automobile components shown in FIG. 20 will be described by taking the case of the transmission 1 as a component as an example. First, when the transmission 1 is operated at the time of checking the transmission 1 for defects,
The second shaft 3 engaged with the input shaft by operating the clutch 5.
Through the gear 4 selected by the speed change mechanism 6 to the first shaft 2
Transmits the rotational force to. 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 an appropriate gain 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 it to the CPU 13. At this time, the automatic range switching device 14 uses the AD converter 1
In order to perform the optimum AD conversion in 2, the range is automatically switched.

The CPU 13 compares the digital data indicating the sound pressure P transferred from the AD converter 12 with the reference effective value (abnormality judgment reference) preset in the CPU 13, and the sound pressure P is determined as the reference effective value. When it exceeds, 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 deal with the abnormality before mounting the transmission 1 on the vehicle.

However, the sound S generated from the transmission 1
The difference in the effective value of is between the normal state and the abnormal state is very small. 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, with the configuration of FIG. 20, it is difficult to accurately distinguish between the normal state and the abnormal state of the transmission 1. The same applies when the engine 31 (see FIG. 21) is used as a component.

[0014]

As described above, the conventional diagnostic device for a vehicle component is limited to the case where the effective value of the sound S generated from the transmission 1 exceeds a certain alarm value (reference effective value). Since the abnormality is determined, the difference in the effective value of the generated sound S between the normal state and the abnormal state is small, and
Since the effective value in the normal state may exceed the effective value in the abnormal state, there is a problem that the abnormality cannot be accurately determined.

The present invention has been made to solve the above problems, and improves the reliability of abnormality diagnosis by discriminating a minute difference between a normal state and an abnormal state of a sound generated from a component. An object of the present invention is to obtain a diagnostic device for an automobile component.

[0016]

According to a first aspect of the present invention, there is provided a diagnostic apparatus for an automobile component, a microphone for detecting a sound generated from the automobile component and outputting an electric signal corresponding to the sound, and an electric signal. An abnormality detection processing device for detecting an abnormality of a vehicle component based on, the abnormality detection processing device, a skewness calculating means for calculating the skewness of the electrical signal, and a comparing means for comparing the skewness with a reference value Based on the comparison result of the comparison means, an alarm is output when the skewness exceeds a predetermined relationship with the reference value.

According to a second aspect of the present invention, there is provided a diagnostic apparatus for a vehicle component, which detects a sound generated from the vehicle component and outputs an electric signal corresponding to the sound, and the vehicle based on the electric signal. An abnormality detection processing device for detecting an abnormality in the component for
The abnormality detection processing device includes a kurtosis calculation means for calculating the kurtosis of the electric signal, and a comparison means for comparing the kurtosis with a reference value,
Based on the comparison result of the comparison means, an alarm is output when the kurtosis exceeds a predetermined relationship with the reference value.

According to a third aspect of the present invention, there is provided a diagnostic apparatus for a vehicle component, which detects a sound generated from the vehicle component and outputs an electric signal corresponding to the sound, and a vehicle based on the electric signal. An abnormality detection processing device for detecting an abnormality in the component for
The abnormality detection processing device includes a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal, and a comparing means for comparing the probability density distribution function with a reference value previously learned by the neural network. Based on the comparison result, an alarm is output when the probability density distribution function exceeds a predetermined relationship with the reference value.

According to a fourth aspect of the present invention, there is provided a vehicle component diagnostic apparatus, which detects a sound generated from a vehicle component and outputs an electric signal corresponding to the sound, and a vehicle based on the electric signal. An abnormality detection processing device for detecting an abnormality in the component for
The abnormality detection processing device includes a probability density distribution function calculation means for calculating a probability density distribution function of an electric signal, and a comparison means for comparing the probability density distribution function with a preset fuzzy rule reference value. Based on the comparison result,
An alarm is output when the probability density distribution function exceeds a predetermined relationship with the reference value.

According to a fifth aspect of the present invention, there is provided a vehicle component diagnostic apparatus, which detects a sound generated from a vehicle component and outputs an electric signal corresponding to the sound, and a vehicle based on the electric signal. An abnormality detection processing device for detecting an abnormality in the component for
The abnormality detection processing device includes a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal and a comparing means for comparing the probability density distribution function with a reference value of a preset pattern. Based on the result, an alarm is output when the probability density distribution function exceeds a predetermined relationship with the reference value.

According to a sixth aspect of the present invention, there is provided a vehicle component diagnostic apparatus which detects a sound generated from a vehicle component and outputs an electric signal corresponding to the sound, and a vehicle based on the electric signal. An abnormality detection processing device for detecting an abnormality in the component for
The abnormality detection processing device includes a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal, a deviation calculating means for calculating a deviation between the probability density distribution function and reference data obtained in advance, and a deviation for a neural network. And a comparison means for comparing with a pre-learned reference value,
Based on the comparison result of the comparison means, an alarm is output when the deviation exceeds a predetermined relationship with the reference value.

According to a seventh aspect of the present invention, there is provided a vehicle component diagnostic apparatus, which detects a sound generated from a vehicle component and outputs an electric signal corresponding to the sound, and a vehicle based on the electric signal. An abnormality detection processing device for detecting an abnormality in the component for
The abnormality detection processing device includes a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal, a deviation calculating means for calculating a deviation between the probability density distribution function and reference data obtained in advance, and a deviation for a neural network. And comparing means for comparing with a reference value of a preset fuzzy rule, and outputs an alarm when the deviation exceeds a predetermined relationship with the reference value based on the comparison result of the comparing means.

According to a eighth aspect of the present invention, there is provided a diagnostic apparatus for a vehicle component, which detects a sound generated by the vehicle component and outputs an electric signal corresponding to the sound, and the vehicle based on the electric signal. An abnormality detection processing device for detecting an abnormality in the component for
The abnormality detection processing device includes a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal, a deviation calculating means for calculating a deviation between the probability density distribution function and reference data obtained in advance, and a deviation pattern. It includes a pattern calculating means for obtaining and a comparing means for comparing the pattern with a reference value, and outputs an alarm when the pattern exceeds a predetermined relationship with the reference value based on the comparison result of the comparing means.

According to a ninth aspect of the present invention, there is provided a vehicle component diagnostic apparatus, which detects a sound generated from a vehicle component and outputs an electric signal corresponding to the sound, and a vehicle based on the electric signal. An abnormality detection processing device for detecting an abnormality in the component for
The abnormality detection processing device includes a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal, a deviation calculating means for calculating a deviation between the probability density distribution function and reference data obtained in advance, and a fractal dimension of the deviation. Includes a fractal dimension calculating means for obtaining the fractal dimension and a comparing means for comparing the fractal dimension with a reference value, and outputs an alarm when the fractal dimension exceeds a predetermined relationship with the reference value based on the comparison result of the comparing means. It is a thing.

According to a tenth aspect of the present invention, there is provided a diagnostic apparatus for an automobile component, which detects a sound generated from the automobile component and outputs an electric signal corresponding to the sound, and an automobile based on the electric signal. An abnormality detection processing device for detecting an abnormality of a component for use in the abnormality detection processing device, wherein the abnormality detection processing device is a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal, and a probability density distribution function and reference data obtained in advance. And a comparison means for comparing the Kullback information amount with a reference value, and a Kullback information based on the comparison result of the comparison unit. An alarm is output when the amount exceeds a predetermined relationship with the reference value.

According to an eleventh aspect of the present invention, there is provided a diagnostic apparatus for a vehicle component, which detects a sound generated from the vehicle component and outputs an electric signal corresponding to the sound, and a vehicle based on the electric signal. An abnormality detection processing device for detecting an abnormality of a component for use in the abnormality detection processing device, wherein the abnormality detection processing device is a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal, and a probability density distribution function and reference data obtained in advance. A deviation calculating means for calculating a deviation between the mean and a mean value calculating means for calculating a mean value of the deviation, and a comparing means for comparing the mean value with a reference value,
Based on the comparison result of the comparison means, an alarm is output when the average value exceeds a predetermined relationship with the reference value.

According to a twelfth aspect of the present invention, there is provided a vehicle component diagnostic apparatus which detects a sound generated from a vehicle component and outputs an electric signal corresponding to the sound, and the vehicle based on the electric signal. An abnormality detection processing device for detecting an abnormality of a component for use in the abnormality detection processing device, wherein the abnormality detection processing device is a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal, and a probability density distribution function and reference data obtained in advance. Deviation calculating means for calculating a deviation between the deviation and, a dispersion value calculating means for calculating a dispersion value of the deviation, and a comparing means for comparing the dispersion value with a reference value,
Based on the comparison result of the comparison means, an alarm is output when the variance value exceeds a predetermined relationship with the reference value.

According to a thirteenth aspect of the present invention, there is provided a vehicle component diagnostic apparatus which detects a sound generated from a vehicle component and outputs an electric signal corresponding to the sound, and a vehicle based on the electric signal. An abnormality detection processing device for detecting an abnormality of a component for use in the abnormality detection processing device, wherein the abnormality detection processing device is a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal, and a probability density distribution function and reference data obtained in advance. A deviation calculating means for calculating a deviation between the deviation, a skewness calculating means for obtaining a skewness of the deviation, and a comparing means for comparing the skewness with a reference value. Based on the comparison result of the comparing means, the skewness is the reference value When a predetermined relationship is exceeded, an alarm is output.

According to a fourteenth aspect of the present invention, there is provided a vehicle component diagnostic apparatus which detects a sound generated from a vehicle component and outputs an electric signal corresponding to the sound, and a vehicle based on the electric signal. An abnormality detection processing device for detecting an abnormality of a component for use in the abnormality detection processing device, wherein the abnormality detection processing device is a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal, and a probability density distribution function and reference data obtained in advance. And a kurtosis calculating means for calculating the kurtosis of the deviation, and a comparing means for comparing the kurtosis with a reference value. Based on the comparison result of the comparing means, the kurtosis is the reference value. When a predetermined relationship is exceeded, an alarm is output.

According to a fifteenth aspect of the present invention, there is provided a diagnostic device for an automobile component according to any of the sixth to first aspects.
In any of the items up to 4, the reference data is set based on the electric signal obtained when the automobile component is in the normal state.

According to a sixteenth aspect of the present invention, there is provided a diagnostic device for an automobile component according to the first to the first aspects.
In any of the items up to 5, the reference value is set based on the electric signal obtained when the automobile component is in the normal state.

According to a seventeenth aspect of the present invention, there is provided a diagnostic device for an automobile component according to the first to the first aspects.
In any of the items up to 6, the object is an automobile transmission or an automobile engine as an automobile component.

[0033]

According to the first aspect of the present invention, the skewness of the electric signal of the component generated sound detected by the microphone is calculated, and an alarm is output when the skewness exceeds a predetermined relationship with respect to the reference value. . As a result, if the detected sound is only the sound existing in the normal state, the skewness becomes close to the normal value, so that the alarm is not erroneously output.

According to the second aspect of the present invention, the kurtosis of the electric signal of the component-generated sound detected by the microphone is calculated, and an alarm is issued when the kurtosis exceeds a predetermined relationship with the reference value. Output. As a result, if the detected sound is only the sound existing in the normal state, the kurtosis becomes close to the normal time, so that the alarm is not erroneously output.

In the third aspect of the present invention, the probability density distribution function of the electric signal of the component generated sound detected by the microphone is calculated, and the probability density distribution function is compared with the reference value preliminarily learned by the neural network. When a predetermined relationship is exceeded, an alarm is output.
Accordingly, if the detected sound is only the sound existing in the normal state, the probability density distribution function becomes close to the normal value, and the alarm is not erroneously output.

According to a fourth aspect of the present invention, the probability density distribution function of the electric signal of the component-generated sound detected by the microphone is calculated, and the probability density distribution function is compared with the preset fuzzy rule reference value. When a predetermined relationship is exceeded, an alarm is output. Accordingly, if the detected sound is only the sound existing in the normal state, the probability density distribution function becomes close to the normal value, and the alarm is not erroneously output.

According to a fifth aspect of the present invention, the probability density distribution function of the electric signal of the component generated sound detected by the microphone is calculated, and the probability density distribution function is compared with the reference value of the preset pattern. An alarm is output when a predetermined relationship is exceeded. Accordingly, if the detected sound is only the sound existing in the normal state, the probability density distribution function becomes close to the normal value, and the alarm is not erroneously output.

Further, in the sixth aspect of the present invention, the probability density distribution function of the electric signal of the component generated sound detected by the microphone is calculated, and the deviation between the probability density distribution function and the reference data obtained in advance is calculated. An alarm is output when the deviation exceeds a predetermined relationship with a reference value learned in advance by the neural network. Accordingly, if the detected sound is only the sound existing in the normal state, the deviation becomes close to the normal value, and the alarm is not erroneously output.

Further, in claim 7 of the present invention, the probability density distribution function of the electric signal of the component generated sound detected by the microphone is calculated, and the deviation between the probability density distribution function and the reference data obtained in advance is calculated. An alarm is output when the deviation is calculated and exceeds a predetermined relationship with a reference value of a fuzzy rule set in advance. Accordingly, if the detected sound is only the sound existing in the normal state, the deviation becomes close to the normal value, and the alarm is not erroneously output.

In the eighth aspect of the present invention, the probability density distribution function of the electric signal of the component generated sound detected by the microphone is calculated, and the deviation between the probability density distribution function and the reference data obtained in advance is calculated. The calculation is performed, and an alarm is output when the deviation exceeds a predetermined relationship with a reference value of a preset pattern.
Accordingly, if the detected sound is only the sound existing in the normal state, the deviation becomes close to the normal value, and the alarm is not erroneously output.

In the ninth aspect of the present invention, the probability density distribution function of the electric signal of the component generated sound detected by the microphone is calculated, and the deviation of the probability density distribution function from the reference data obtained in advance is calculated. The fractal dimension is calculated, and an alarm is output when the fractal dimension exceeds a predetermined relationship with the reference value. As a result, if the detected sound is only the sound that exists in the normal state,
The fractal dimension of the deviation will be close to the normal value, and the alarm will not be erroneously output.

In the tenth aspect of the present invention,
The probability density distribution function of the electrical signal of the component generated sound detected by the microphone is calculated, and the Kalback information amount of the deviation between the probability density distribution function and the reference data obtained in advance is calculated, and the Kalback information amount becomes the reference value. An alarm is output when a predetermined relationship is exceeded. As a result, if the detected sound is only the sound that exists in the normal state, the amount of deviation Kullback information is close to the normal value,
There is no false alarm output.

In the eleventh aspect of the present invention,
The probability density distribution function of the electrical signal of the component generated sound detected by the microphone is calculated, and the average value of the deviation between the probability density distribution function and the reference data obtained in advance is calculated. An alarm is output when a predetermined relationship is exceeded. As a result, if the detected sound is only the sound existing in the normal state, the average value of the deviations is close to the normal value, and the alarm is not erroneously output.

According to claim 12 of the present invention,
The probability density distribution function of the electrical signal of the component generated sound detected by the microphone is calculated, and the variance value of the deviation between the probability density distribution function and the reference data obtained in advance is calculated. An alarm is output when a predetermined relationship is exceeded. As a result, if the detected sound is only the sound existing in the normal state, the variance value of the deviation becomes close to the normal value, and the alarm is not erroneously output.

In the thirteenth aspect of the present invention,
The probability density distribution function of the electrical signal of the component generated sound detected by the microphone is calculated, and the skewness of the deviation between the probability density distribution function and the reference data obtained in advance is calculated. An alarm is output when a predetermined relationship is exceeded. As a result, if only the detected sound is a sound existing in a normal state, the skewness of the deviation is close to the normal value, and the alarm is not erroneously output.

In the fourteenth aspect of the present invention,
The probability density distribution function of the electrical signal of the component generated sound detected by the microphone is calculated, and the kurtosis of the deviation between the probability density distribution function and the reference data obtained in advance is calculated. An alarm is output when a predetermined relationship is exceeded. As a result, if the detected sound is only the sound existing in the normal state, the kurtosis of the deviation becomes close to the normal value, and the alarm is not erroneously output.

In the fifteenth aspect of the present invention,
The reference data is set based on the electrical signal of the sound generated when the automobile component is normal, and the abnormality determination of the component is ensured.

According to claim 16 of the present invention,
The reference value is set based on the electric signal of the sound generated when the automobile component is normal, and the abnormality determination of the component is ensured.

In the seventeenth aspect of the present invention,
Targeting a transmission or an engine as an automobile component, the reliability of abnormality determination of the transmission or the engine is improved.

[0050]

Embodiments 1 to 14 of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing Embodiments 1 to 14 of the present invention.
9, 10, 12, S and E are the same as described above. Hereinafter, similar to the above, a case where the transmission 1 is used as a component that generates noise will be described as an example. Note that FIG. 1 shows a schematic configuration of the entire component diagnostic apparatus, and corresponds to all of claims 1 to 17.

Yi (i = 1, 2, ..., N) is time-series data (hereinafter referred to as column data) corresponding to the voltage level of the electric signal E of the sound S detected by the microphone 7, and is an amplifier. It becomes digital data through 9, the filter 10 and the AD converter 12.

Reference numeral 20 denotes an abnormality detection processing device composed of a CPU for performing various arithmetic processing on the column data Yi,
It has two operation modes (reference value or reference data setting mode and transmission 1 diagnosis mode) that are switched by an external command based on a manual operation or the like.

The abnormality detection processing device 20 is arranged such that the column data Yi indicating the electric signal E of the detection sound S at the time of diagnosing the transmission 1.
Is taken in, and index data (to be described later) indicating the property of the column data Yi is calculated. Also, with this diagnostic data,
The data at the time of diagnosis is compared with the reference value based on the comparison result by comparing with the data (reference value or reference data obtained in advance) based on the column data Yi obtained when the transmission 1 is normal. Transmission 1 when a predetermined relationship is exceeded
Is judged and an alarm A is generated.

Example 1. (Corresponding to Claim 1, Claim 16 and Claim 17) FIG. 2 shows an abnormality detection processing apparatus 20 according to Embodiment 1 of the present invention.
21 is a functional block diagram showing the configuration of the column data 21
A skewness calculator for calculating the skewness μ3 of the electric signal E from i, 2
Reference numeral 28 is a skewness setter that stores the skewness output from the skewness calculator 21 as a reference value μ3r in the reference value setting mode, and 23 is the skewness calculator 2 in the diagnostic mode of the transmission 1.
1 is a comparator for comparing the skewness μ3 output from 1 with the reference value μ3r from the skewness setting device 22, and 24 is an alarm device for issuing an alarm A based on the comparison result of the comparator 23.

In FIG. 2, broken line arrows indicate the flow of each signal in the reference value setting mode, and solid line arrows indicate the flow of each signal in the component diagnosis mode. This also applies to each drawing described later. The skewness calculator 21 in the abnormality detection processing device 20 uses the average value μ1 obtained from the equation (1) and the variance value μ2 obtained from the equation (2) as shown below, according to the equation (3). The skewness μ3 is calculated.

Μ1 = (1 / N) Σ (Yi) (1) μ2 = √ {(1 / N) Σ (Yi−μ1) ² } (2) μ3 = (1 / N) · {1 / (Μ2) ³ } Σ (Yi−μ1) ³ (3)

In the summation term (Σ) of the above equations (1) to (3), the column data Yi is added for i = 0, 1, ..., N−1.

Next, the operation of the first embodiment of the present invention will be described with reference to FIGS. 1 and 2. First, AD
The converter 12 samples the electric signal E output via the amplifier 9 and the filter 10 at equal time intervals,
Time series N column data Yi (i = 1, 2, ..., N)
Is output.

Subsequently, the skewness calculator 21 in the abnormality detection processing device 20 performs the calculations of the equations (1) to (3) on the column data Yi to calculate the skewness μ3 of the electric signal E. At this time, in the reference value setting mode that is performed in advance when the transmission 1 is in a normal state (or using the transmission 1 that is a normal sample), the skewness μ calculated by the skewness calculator 21 is
3 is output to the skewness setting unit 22 and stored in the skewness setting unit 22 as a reference value μ3r, as indicated by a broken line in FIG.

On the other hand, in the component diagnostic mode, the skewness μ3 calculated by the calculator 21 is output to the comparator 23 as shown by the solid line in FIG. Accordingly, the comparator 23 causes the skewness μ3 output from the skewness calculator 21 and the reference value μ3r stored in the setter 22 in the diagnostic mode.
When the skewness μ3 exceeds a predetermined relationship with the reference value μ3r (for example, three times the reference value μ3r), a comparison result indicating an abnormality is output. After receiving this comparison result,
The alarm device 24 issues an alarm A.

The series of operations in the component diagnosis mode is repeated, for example, in a fixed cycle. In this way, the operation of the reference value setting mode is executed in advance based on the electric signal E from the transmission 1 in the normal state, the reference value μ3r is stored, and then the operation of the diagnostic mode is executed at the time of component diagnosis. Then, the skewness μ3 indicating the property of the sound S generated from each transmission 1 to be diagnosed is sequentially obtained, and this skewness μ3 is compared with the reference value μ3r in the normal state.

As a result, the detection sound S in the diagnostic mode is generated.
Indicates an abnormality of the transmission 1, and the skewness μ against the reference value μ3r
When the difference of 3 is large (exceeding the predetermined relationship), the alarm device 24 can surely generate the alarm A indicating an abnormality. On the other hand, if the detected sound S in the diagnostic mode indicates the normal state of the transmission 1, the skewness μ3 is the reference value μ3r.
It is equivalent to (normal), and the alarm device 24 does not generate the alarm A.

In the diagnosis based on the skewness μ3 as described above, there is an abnormal deviation in the probability density distribution (described later) of the column data Yi such as when the casing of the transmission 1 is in contact with a movable part. It is especially effective when the features appear. Therefore, the diagnosis result can be obtained by a relatively simple process as in the equation (3).

Example 2. (Corresponding to claim 2, claim 16 and claim 17) In the first embodiment, the skewness μ3 is used as the index data indicating the property of the column data Yi corresponding to the electric signal E of the detection sound S. However, the kurtosis of the column data Yi may be used.

The second embodiment of the present invention using the kurtosis as the index data of the electric signal E will be described below with reference to the drawings. FIG. 3 is a functional block diagram showing the configuration of the abnormality detection processing device 20 according to the second embodiment of the present invention.
A corresponds to each of the skewness calculators 21 to 24 described above, and Yi and A are the same as those described above.

21A is a kurtosis calculator for calculating the kurtosis μ4 of the electric signal E from the column data Yi, and 22A is the kurtosis μ4 output from the kurtosis calculator 21A in the reference value setting mode as a reference value μ4r. The kurtosis setter 23A is the kurtosis μ4 output from the kurtosis calculator 21A and the kurtosis setter 22.
A comparator that compares the reference value μ4r in A with 24A is an alarm that generates an alarm A based on the comparison result of the comparator 23A.

In this case, the kurtosis calculator 21A uses the mean value μ1 obtained from the above equation (1) and the variance value μ2 obtained from the equation (2) to calculate the kurtosis according to the following equation (4). Calculate μ4.

Μ4 = (1 / N) {1 / (μ2) ⁴ } Σ {(Yi-μ1) ⁴ } -3 (4)

In the summation term (Σ) of the equation (4), the column data Yi is added for i = 0, 1, ..., N−1. Next, the operation of the second embodiment of the present invention will be described with reference to FIGS. First, in the same manner as described above, the AD converter 12 sets the time series N column data Yi.
Is output.

Subsequently, the kurtosis calculator 21A in the abnormality detection processing device 20 performs the calculation of (4) together with the above equations (1) and (2) on the column data Yi to calculate the kurtosis of the electric signal E. Calculate μ4. At this time, in the reference value setting mode, the kurtosis μ4 calculated by the kurtosis calculator 21A is as shown in FIG.
As indicated by the broken line in the figure, it is output to the kurtosis setter 22A,
It is stored as a reference value μ4r in the kurtosis setter 22A.

On the other hand, in the component diagnosis mode, the kurtosis μ4 calculated by the calculator 21A is output to the comparator 23A as shown by the solid line in FIG. As a result, the comparator 23A causes the kurtosis calculator 21A in the diagnostic mode.
The kurtosis μ4 output from the reference value μ4r is compared with the reference value μ4r stored in the setter 22A, and the kurtosis μ4 exceeds a predetermined relationship with the reference value μ4r (for example, three times the reference value μ4r). At this time, the comparison result indicating abnormality is output. Upon receipt of this comparison result, the alarm device 24A issues an alarm A. The series of operations in the component diagnosis mode is repeated, for example, in a fixed cycle.

In this way, the operation of the reference value setting mode is executed in advance based on the electric signal E from the transmission 1 in the normal state to store the reference value μ4r, and then the operation of the diagnostic mode is executed to make the transmission 1 Kurtosis indicating the nature of the generated sound S from
4 is sequentially obtained, and this kurtosis μ4 is compared with the reference value μ4r in the normal state.

Thus, when the detected sound S at the time of diagnosis indicates an abnormality of the transmission 1 and the difference in kurtosis μ4 with respect to the reference value μ4r is large (exceeds a predetermined relationship), the alarm device 24A indicates an abnormality. The alarm A shown can be reliably output. On the other hand, if the detected sound S at the time of diagnosis indicates the normal state of the transmission 1, the kurtosis is equal to the reference value μ4r (normal), and the alarm device 24A does not generate the alarm A. .

The diagnosis based on the kurtosis μ4 as described above is performed, for example, in the case where the gear 4 in the transmission 1 has a missing gear, an eccentricity of the gear, etc. It is especially effective when abnormal features appear in. Therefore, the diagnosis result can be obtained by a relatively simple process as shown in Expression (4).

Example 3. (Corresponding to claim 3, claim 16 and claim 17) In the second embodiment, the kurtosis μ4 is used as index data indicating the property of the column data Yi corresponding to the electric signal E of the detected sound S. However, the probability density distribution function of the column data Yi obtained by statistically processing the column data Yi may be used. In this case, the skewness μ based only on the voltage level of the column data Yi
The reliability of the diagnostic result is improved as compared with the case of using 3 (Example 1) or kurtosis μ4 (Example 2).

A third embodiment of the present invention using the probability density distribution function of the column data Yi and the value obtained by learning the same with a neural network as index data of the electric signal E will be described below with reference to the drawings. FIG. 4 is a functional block diagram showing the configuration of the abnormality detection processing device 20 according to the third embodiment of the present invention, and 23B and 24B are the comparators 23 described above.
A and alarm device 24A respectively, and Yi and A are the same as described above.

25 is a probability density distribution function calculator for calculating the probability density distribution function F1 of the electric signal E from the column data Yi. Reference numeral 26 denotes a neural network calculator that calculates the shape of the probability density distribution function F1 as a learning value N1 of the neural network, and outputs the fitness “0 to 1” by the neural network as the learning value N1. 27
Is a reference value setting device that stores the learning value N1 (0 to 1) output from the neural network calculator 26 in the reference value setting mode as the reference value N1r.

23B is a neural network computing unit 2
6 is a comparator for comparing the learning value N1 output from the reference value N1r with the reference value N1r in the reference value setting unit 22A, and 24B is a comparator 23.
It is an alarm device that issues an alarm A based on the comparison result of A.

The probability density distribution function calculator 25 uses a well-known function representing the distribution of each amplitude value (probability indicating what value the amplitude of one irregular waveform appears at a certain time).
It is calculated as a probability density distribution function F1. The probability density distribution function F1 can be obtained based on various time-series distribution characteristics regarding the column data Yi, such as the amplitude (corresponding to the voltage level) and the frequency count number.

Since the probability density distribution function calculator 25 digitally processes the column data Yi, the amplitude range is divided into equal intervals, for example, 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, it is obtained as a ratio of how many data belonging to each divided range exist in the column data Yi.

The neural network computing unit 26
For a given pair of input data and output data (teacher signal), learning is performed using a multilayer network composed of units having sigmoid characteristics and a well-known method called error backpropagation learning method. Considering the error evaluation function, the connection weight that minimizes the error evaluation function is calculated by the gradient method and is calculated as the learning value N1 of the neural network.

FIG. 5 is an explanatory diagram showing an example of a reference signal used for obtaining the probability density distribution function F1, where the horizontal axis represents time t and the vertical axis represents the count of a predetermined amplitude or the like at each time based on the column data Yi. Is a number. The count number can consist of various factors as described above.

FIG. 6 is an explanatory view showing the probability density distribution function F1 obtained based on the reference signal of FIG. 5, where the horizontal axis is the number of counts and the vertical axis is the number of appearances at each count number. Here, as the probability density distribution function F1 corresponding to the reference (normal) value N1r, the number of counts “100” is “5”.
The case where the number of "0" and the count number "90" is "10" is shown.

Next, the operation of the third embodiment of the present invention will be described with reference to FIGS. 1 and 4 to 6. First, N time-series column data Y from the AD converter 12
Based on i, the probability density distribution function calculator 25 calculates the probability density distribution function F1 for the column data Yi. Then, the neural network calculator 26 calculates the learning value N1 (0 to 1) representing the fitness by the neural network by using the neural network constructed in advance by learning the shape of the probability density distribution function F1.

In the reference value setting mode, the learning value N1 from the neural network computing unit 26 is input to the reference value setting unit 27 as shown by the broken line in FIG.
It is stored as 1r. On the other hand, in the component diagnosis mode, the learning value N1 is output to the comparator 23B as shown by the solid line in FIG.

The comparator 23B compares the learning value N1 calculated in the diagnostic mode with the reference value N1r calculated in the normal state, and the learning value N1 in the diagnosis is, for example, 1 of the reference value N1r.
When it is less than / 2 times (the degree of conformity is 1/2 or less), a comparison result indicating an abnormality is output to the alarm device 24B, and the alarm device 24B issues an alarm A in response to the comparison result. Similar to the above, the series of operations in the component diagnosis mode is repeated, for example, in a constant cycle.

As described above, the reference value N1r is stored in advance on the basis of the sound S (electrical signal E) generated from the transmission 1 at the normal time, and then the electrical signal E from the transmission 1 at the time of diagnosis is stored. The learning value N1 indicating the goodness of fit is sequentially obtained, and the learning value N
By comparing 1 with the reference value N1r, the learning value N1
Indicates that the transmission 1 is abnormal (the difference between the learned value N1 and the reference value N1r exceeds a predetermined relationship), the alarm A can be reliably output.

On the other hand, the detection sound S at the time of diagnosis is the transmission 1
, The learning value N1 is the reference value N
This is equivalent to 1r, and the alarm device 24B does not generate the alarm A. Further, the diagnosis based on the learning value N1 by such a neural network is performed by the probability density distribution function F1.
Since the signal processing can be performed only by learning the neural network from, it is not necessary to determine the signal processing or the calculation method in advance, and the calculation can be simplified.

Example 4. (Corresponding to claim 4, claim 16 and claim 17) In the third embodiment, as index data indicating the property of the column data Yi corresponding to the electric signal E of the detection sound S,
Although the learning value N1 by the neural network based on the probability density distribution function F1 of the column data Yi is used, a feature by fuzzy rules based on the probability density distribution function F1 (reference value in the membership function) may be used.

The fourth embodiment of the present invention using the feature of the fuzzy rule based on the probability density distribution function F1 as the index data of the electric signal E will be described below with reference to the drawings. Figure 7
FIG. 23 is a functional block diagram showing the configuration of the abnormality detection processing device 20 according to the fourth embodiment of the present invention.
C and 27C are the above-mentioned comparator 23B and alarm device 24.
B, the neural network computing unit 26 and the reference value setting unit 27, respectively, 25, Yi and A
Is the same as described above.

In FIG. 7, 26C is a feature extractor that digitizes and outputs the feature F2 according to the fuzzy rule of the probability density distribution function F1, F2a is a membership reference function given from the outside in the reference value setting mode, and 27C is a reference value. It is a membership function setter that stores a membership function F2r based on the characteristic (numerical value) F2 and the membership reference function F2a in the setting mode as a reference value of the fuzzy rule.

In this case, the comparator 23C is the feature extractor 2
The feature F2, which is symbolized and output by the fuzzy rule from 6C, is compared with the membership function F2r for executing the fuzzy rule. Also, the alarm device 24C is
An alarm A is issued based on the comparison result from the comparator 23C.

8 and 9 show the probability density distribution function F1.
It is explanatory drawing which shows an example of the membership function of the fuzzy rule obtained based on (refer FIG. 6), respectively.
The horizontal axis is the number of counts (90 and 100),
The vertical axis represents the membership function value (0 to 1) for determining normality or abnormality. Here, when the membership function value is 0 to 0.5, it is determined to be abnormal, and 0.5 to 1
Is judged to be normal.

FIG. 8 shows the membership function F2r (90) when the count number is "90". As is clear from FIG. 6, when the number is "10", the maximum membership function value is "1". It will be. Further, FIG. 9 shows the membership function F2r (10) when the count number is "100".
0), and when the number is “50”, the maximum membership function value is “1”.

Each membership function F2r is obtained, for example, by adding the feature F2 based on the fuzzy rule to the membership reference function F2a. The feature F2 consisting of the number is represented as a numerical value of the probability density of an appropriate amplitude energy region, for example, but any numerical value can be selected as necessary. Further, the membership reference function F2a is not limited to the shapes shown in FIGS. 8 and 9 and can be arbitrarily selected.

Next, the operation of the fourth embodiment of the present invention will be described with reference to FIGS. 1 and 7 to 9. First, N time-series column data Y from the AD converter 12
Based on i, the probability density distribution function calculator 25 calculates the probability density distribution function F1 for the column data Yi. Subsequently, the feature extractor 26C extracts the feature F2 of the probability density distribution function F1.

In the reference value setting mode, the feature F2 is input to the membership function setting unit 27C as shown by the broken line in FIG. 7, and the membership reference function F2a is input.
It is also stored as the reference value F2r. On the other hand, in the component diagnosis mode, the feature F2 is output to the comparator 23C as indicated by the solid line.

That is, in the component diagnostic mode, the probability density distribution function F1 calculated by the probability density distribution function calculator 25 is output to the feature extractor 26C,
For example, the number data (see FIG. 6) for the count numbers “100” and “90” is extracted as the feature F2.

The feature F2 thus extracted is the comparator 2
In 3C, the goodness of fit to the membership functions F2r (90) (see FIG. 8) and F2r (100) (see FIG. 9) stored in the membership function setter 27C is compared. At this time, as shown in FIGS. 8 and 9, membership functions F2r (90) and F2 obtained from the probability density distribution function F1 of the normal reference signal (see FIG. 5).
Center of the number (horizontal axis) of r (100) (feature F in normal state
2) are "10" and "50", respectively.

Now, the sound S of the transmission 1 to be diagnosed is generated.
The feature F2 extracted from the probability density distribution function F1 of is the number “8” and the count “10” for the count number “90”.
It is assumed that the number "49" for "0" is shown. Comparing these numbers “8” and “49” on the membership functions F2r (90) and F2r (100) of FIGS. 8 and 9, the function values are “0.8” and “0. 95 ".

Therefore, 0.5 <the value of the membership function F2r (90) ≦ 1.0 and 0.5 <the value of the membership function F2r (100) ≦ 1.0 are satisfied at the same time. Therefore, the transmission 1 is determined to be normal. On the contrary, when at least one of the membership function values becomes 0.5 or less, the transmission 1 to be diagnosed is determined to be abnormal, and in response to the abnormality determination result from the comparator 23C, The alarm A is issued from the alarm device 24C.

A series of operations in the component diagnosis mode is repeated, for example, in a fixed cycle. Therefore, if the operation of the reference setting mode is executed in advance using the transmission 1 in a normal state and the operation of the component diagnostic mode is executed at the time of subsequent determination, the characteristics of the sound S generated by each transmission 1 Is successively obtained and compared with the normal time, and if the difference from the normal time is large, the alarm A is output.

Therefore, if the generated sound S at the time of determination indicates an abnormality of the transmission 1, the alarm A is output, but if the generated sound S indicates a normal state, the compatibility is normal. They are equivalent and alarm A is not output.

Further, since the membership function F2r applied to the fuzzy rule performs logical signal processing and determination, reliability can be further improved.
Further, as the feature F2, the count numbers "90" and "1"
Although the number for “00” is extracted, it goes without saying that an arbitrary feature can be extracted and the number of extractions can be arbitrarily selected.

Example 5. (Corresponding to claim 5, claim 16 and claim 17) In the fourth embodiment, as index data indicating the property of the column data Yi corresponding to the electric signal E of the detection sound S,
Although the membership function F2r based on the probability density distribution function F1 of the column data Yi is used, the waveform characteristics of the probability density distribution function F1 (for example, voltage level, peak value, frequency,
(Consisting of a plurality of items such as waviness or pitch) may be used as the reference value of the pattern.

Hereinafter, as the index data of the electric signal E, the waveform feature of the probability density distribution function F1 (reference value of the pattern)
A fifth embodiment of the present invention using is described with reference to the drawings. FIG. 10 shows an abnormality detection processing device 20 according to Embodiment 5 of the present invention.
23D, 24D,
26D and 27D respectively correspond to the comparator 23C, the alarm device 24C, the feature extractor 26C and the membership function setting device 27C in FIG. 7, and 25, Yi and A are the same as those described above.

In FIG. 10, 26D is a waveform feature extractor for extracting the waveform feature F3 of the probability density distribution function F1, and 27
D is a reference value setter that stores the waveform feature F3 obtained from the normal transmission 1 in the reference setting mode as the reference value F3r, and 23D is a comparator that compares the waveform feature F3 obtained in the diagnostic mode with the reference value F3r. , 24D are alarm devices that generate an alarm A based on the comparison result of the comparator 23D.

The waveform feature F3 may include various items such as amplitude, count number and frequency.
The waveform of the probability density distribution function F1 can also be expressed as a smoothed waveform. Further, the reference value F3r may be given from the outside in the reference setting mode like the membership reference function F2r described above.

Next, referring to FIG. 1 and FIG.
The operation of the fifth embodiment of the present invention will be described. First, in the standard setting mode, the probability density distribution function calculator 25 is based on the column data Yi obtained from the normal transmission 1.
Calculates the probability density distribution function F1, and the waveform feature extractor 26
D extracts the waveform feature F3. Extracted waveform feature F3
Indicates the reference value F3r in the reference value setter 27D as indicated by the broken line.
Is stored as

On the other hand, in the component diagnosis mode, the waveform feature F3 extracted from the probability density distribution function F1.
Is compared with the reference value F3r in the comparator 23D. The transmission 1 to be diagnosed is determined to be normal if the conformity of the waveform feature F3 to the reference value F3r is high,
If the degree of compatibility is low, it is determined to be abnormal, and the alarm A is issued from the alarm device 24D. This diagnostic method is effective when an abnormal feature appears as a specific pattern in the waveform feature of the probability density distribution function F1.

Example 6. (Claim 6 and Claim 15-
In the fifth embodiment, as index data indicating the property of the column data Yi corresponding to the electric signal E of the detection sound S,
Although the waveform feature F3 of the probability density distribution function F1 of the column data Yi is used as the reference value F3r, it may be normalized using the deviation of the probability density distribution function F1 from the reference data in order to further improve the reliability of diagnosis. Good. In this case, the calculated deviation is compared with the reference value learned by the neural network.

A sixth embodiment of the present invention in which the deviation of the probability density distribution function F1 from the reference data is used as the index data of the electric signal E will be described below with reference to the drawings. FIG. 11 is a functional block diagram showing the configuration of the abnormality detection processing device 20 according to the sixth embodiment of the present invention, which is 23E, 24E, 26E.
And 27E are the comparator 23B and the alarm device 24 in FIG.
B, the neural network computing unit 26 and the reference value setting unit 27, respectively, 25, Yi and A
Is the same as described above.

In FIG. 11, 28 is a reference data setter for storing the probability density distribution function F1 output from the probability density distribution function calculator 25 in the reference setting mode as reference data F1r, and 29 is the probability density distribution function F1. The deviation calculator is a subtracter that calculates a deviation ΔF between the column data and the reference data F1r.

26E is a neural network calculator for learning the shape of the calculation result (column data) of the deviation calculator 29, and 27E is a learning value N2 of 0 to 1 output from the neural network calculator 26E in the reference setting mode.
Is a reference value setting unit for storing the learning value N2 (0 to 1) output from the neural network computing unit 26E as a reference value N2r in the reference value setting unit 27E.
Reference numeral 24E is a comparator for comparing with 2r, and reference numeral 24E is an alarm device for issuing an alarm A based on the comparison result of the comparator 23E.

Next, referring to FIG. 1 and FIG.
The operation of the sixth embodiment of the present invention will be described. First, the probability density distribution function calculator 25 calculates a probability density distribution function F1 of a signal based on the column data Yi, and outputs it as column data representing the probability density distribution function F1.

In the standard setting mode, the probability density distribution function F1 represented by column data is input to the standard data setter 28 as indicated by the broken line in FIG.
It is stored as 1r. Then, the sampling of the column data Yi is performed again, and the column data of the probability density distribution function F1 output from the probability density distribution function calculator 25 is input to the deviation calculator 29, and the reference data in the reference data setter 28 is input. The deviation ΔF from the data F1r (deviation for each corresponding individual piece of column data) ΔF is calculated.

Deviation Δ of the calculated probability density distribution function F1
F is input to the neural network calculator 26E, and the shape of the deviation ΔF is learned. In the reference setting mode, the learning value N2 (0 to 1) calculated by the neural network calculator 26E is input to the reference value setter 27E as shown by the broken line in FIG. 11 and stored as the reference value N2r.

On the other hand, in the diagnostic mode, the column data of the probability density distribution function F1 is always input to the deviation calculator 29 and the deviation ΔF from the reference data in the reference data setter 28 is set.
Is calculated. Subsequently, the deviation ΔF of the probability density distribution function F1 output from the deviation calculator 29 is input to the neural network calculator 26E, and the neural network calculates a learning value N2 representing the degree of conformity 0 to 1.

The learning value N2 calculated by the neural network calculator N2 is compared with the reference value N2r in the reference value setting unit 27E in the comparator 23E. Comparator 23
E is abnormal in the alarm device 24E when the learning value N2 output from the neural network computing unit 26E falls below, for example, 1/2 times the reference value N2r (degree of conformity) stored in the reference value setting unit 27E. Is output, and the alarm device 24E issues an alarm in response to the comparison result.

Since a series of operations in the component diagnosis mode is repeated, for example, in a fixed cycle, the reference data F1r and the reference value N2r using the normal transmission 1 are set in the reference setting mode, and thereafter, In the determination mode of 1, the diagnosis of the transmission 1 can be executed.

As a result, the probability density distribution function F1 and the probability density distribution function F1 showing the characteristics of the sound S generated by each transmission 1 are obtained.
Using the deviation ΔF from the reference data F1r in the normal state, the learning value N2 (fitness) of the neural network is sequentially obtained. Then, in the comparator 23E, the learning value N2
Is compared with the reference value N2r in the normal state, and if the difference from the reference value N2r in the normal state is large, an alarm A is issued.

Therefore, similarly to the above, if the generated sound S at the time of determination indicates an abnormality of the transmission 1, the alarm A is issued.
Is generated, but if the generated sound S indicates a normal state,
The learning value N2 representing the goodness of fit becomes a value equivalent to the reference value N2r in the normal state, and the alarm A is not issued.

According to the component diagnosing method of the sixth embodiment of the present invention, the neural network computing unit 26E performs the signal processing only by learning the deviation ΔF between the probability density distribution function F1 and the reference data F1r in the neural network. Therefore, it is not necessary to determine the signal processing and calculation method in advance.

Example 7. (Claim 7 and Claim 15-
In the sixth embodiment, the deviation ΔF between the probability density distribution function F1 and the reference data F1r is learned by a neural network and the learning value N2 is compared with the reference value N2r. The feature may be extracted and compared with the reference value (membership function) of the fuzzy rule.

Hereinafter, the deviation ΔF of the probability density distribution function F1 with respect to the reference data F1r is used as the index data of the electric signal E, and the feature of the deviation ΔF is compared with the reference value of the fuzzy rule. 7 will be described with reference to the drawings.

FIG. 12 is a functional block diagram showing the structure of the abnormality detection processing device 20 according to the seventh embodiment of the present invention.
3F, 24F, 26F, 27F and F4a respectively correspond to the comparator 23C, the alarm device 24C, the feature extractor 26C, the membership function setting device 27C and the membership reference function F2a in FIG. Two
9, Yi and A are the same as described above.

In FIG. 12, 26F is a deviation calculator 29.
A feature extractor that digitizes or symbolizes the feature F4 from the deviation ΔF calculated in step S27, and 27F is a fuzzy rule membership function F obtained in the reference setting mode.
4r is a membership function setting device for storing 4r.

Reference numeral 23F is a comparator for judging whether the transmission 1 to be diagnosed is correct or not. The feature F4 extracted by the feature extractor 26F is compared with the membership function F4r in the membership function setting unit 27F. By executing the fuzzy rules. Reference numeral 24F is an alarm device for issuing an alarm A based on the comparison result of the comparator 52.

Next, referring to FIG. 1 and FIG.
The operation of the seventh embodiment of the present invention will be described. First, similarly to the above, the deviation calculator 29 calculates the deviation ΔF between the probability density distribution function F1 based on the column data Yi and the reference data F1r in the reference data setter 28.

The deviation ΔF of the probability density distribution function F1 is input to the feature extractor 26F, and the feature F4 represented by a numerical value is extracted. In the reference setting mode, the feature extractor 26
The feature F4 extracted by F is input to the membership function setter 27F as shown by a broken line and becomes a membership function F4r together with the membership reference function F4a given from the outside, which serves as a reference value for executing the fuzzy rule. It is stored in the ship function setter 27F.

On the other hand, in the component diagnostic mode, the column data of the probability density distribution function F1 calculated by the probability density distribution function calculator 25 is always input to the deviation calculator 29, and the deviation ΔF from the reference data F1r is calculated. To be done. Subsequently, the deviation ΔF of the probability density distribution function F1 is input to the feature extractor 26F, and the feature F4 represented as a numerical value of the probability density in an appropriate amplitude energy region is extracted.

The extracted feature F4 is compared with the membership function F4r in the comparator 23F, and, for example,
If the membership function value is 0.5 or less and the compatibility is low, the transmission 1 to be diagnosed is determined to be abnormal.
At this time, the comparator 23F outputs a comparison result indicating an abnormality, and the alarm device 24F issues an alarm A in response to the comparison result.

Note that, as described above, a series of operations in the diagnostic mode is repeated, for example, in a fixed cycle, the fitness is sequentially obtained, and the alarm A is issued only when the difference from the normal state is large. . Further, the reliability of the diagnosis can be improved by performing the logical signal processing and the determination according to the fuzzy rule and using the deviation ΔF of the probability density distribution function F1.

Example 8. (Claims 8 and 15 to
In the seventh embodiment, the feature F4 of the deviation ΔF of the probability density distribution function F1 is extracted and compared with the membership function F4r of the fuzzy rule, but the waveform feature of the deviation ΔF is extracted as a pattern. However, it may be compared with the reference value of the pattern.

An eighth embodiment of the present invention will be described below in which the deviation ΔF of the probability density distribution function F1 is used as the index data of the electric signal E and the waveform characteristic of the deviation ΔF is compared with the reference value. .

FIG. 13 is a functional block diagram showing the structure of the abnormality detection processing device 20 according to the eighth embodiment of the present invention.
3G, 24G, 26G and 27G correspond to the comparator 23D, the alarm device 24D, the waveform feature extractor 26D and the reference value setting device 27D in FIG.
8, 29, Yi and A are the same as described above.

In FIG. 13, 26G is a deviation calculator 29.
The waveform feature extractor that extracts the waveform feature F5 (for example, consisting of a plurality of items) as a digitized pattern from the deviation ΔF calculated in step 27G stores the waveform feature F5 obtained in the reference setting mode as the reference value F5r. The reference value setter 23G is a comparator for comparing the waveform feature F5 output from the waveform feature extractor 26G with the reference value F5r in the reference value setter 27G, and 24G is a response to the comparison result of the comparator 24G. It is an alarm device that generates an alarm A.

Next, referring to FIG. 1 and FIG.
The operation of the eighth embodiment of the present invention will be described. In this case, the probability density distribution function F1 calculated by the deviation calculator 29
Deviation ΔF is input to the waveform feature extractor 26G, and the waveform feature F5 is extracted. In the reference setting mode, the waveform feature F5 is stored as the reference value F5r in the reference value setter 27G as indicated by the broken line.

On the other hand, in the component diagnosis mode, the waveform feature F5 having the deviation ΔF extracted by the waveform feature extractor 26G is always input to the comparator 23G and the reference value F5 is input.
is compared to r. The waveform feature F5 can also be expressed as a waveform obtained by smoothing the waveform of the probability density distribution function F1.

The comparator 23G has a waveform feature F5 and a reference value F.
Transmission 1 to be diagnosed when the compatibility with 5r is high
Is determined to be normal, and conversely, when the conformity between the waveform feature F5 and the reference value F5r is low, the transmission 1 to be diagnosed is determined to be abnormal. Then, in response to the comparison result indicating the abnormality from the comparator 23G, the alarm device 24G issues an alarm A.

Note that, as described above, a series of operations in the component diagnostic mode is repeated, for example, in a fixed cycle, the conformity is sequentially obtained, and the alarm A is issued only when the difference from the normal state is large. To do. In this case, the degree of complexity of the waveform amplitude (noise waveform including high frequency, etc.) occupied in the plane space can be obtained from the difference between the waveform feature F5 and the reference value F5r.

Example 9. (Claim 9 and Claim 15 to
In the eighth embodiment, although the waveform feature F5 of the deviation ΔF of the probability density distribution function F1 is extracted as a pattern and compared with the reference value F5r of the pattern, the fractal dimension of the deviation ΔF is calculated. May be compared with a reference value.

A ninth embodiment of the present invention in which the deviation ΔF of the probability density distribution function F1 is used as index data of the electric signal E and the fractal dimension of the deviation ΔF is compared with a reference value will be described with reference to the drawings. .

FIG. 14 is a functional block diagram showing the structure of the abnormality detection processing device 20 according to the ninth embodiment of the present invention.
3H, 24H, 26H and 27H correspond to the comparator 23G, the alarm device 24G, the waveform feature extractor 26G and the reference value setting device 27G in FIG.
8, 29, Yi and A are the same as described above.

In FIG. 14, 26H is the deviation calculator 29.
The fractal dimension calculator for calculating the fractal dimension F6 of the deviation ΔF calculated in step 27H is a fractal dimension setter for storing the fractal dimension F6 obtained in the reference setting mode as the reference value F6r, and 23H is output from the fractal dimension calculator 26H. The fractal dimension F6 is compared with a reference value F6r in the fractal dimension setter 27H, and 24H is an alarm that generates an alarm A in response to the comparison result of the comparator 24H.

Next, referring to FIG. 1 and FIG.
The operation of the ninth embodiment of the present invention will be described. In this case, the deviation ΔF (N column data) calculated by the deviation calculator 29 is input to the fractal dimension calculator 26H to calculate the fractal dimension F6. Fractal dimension F6
In the reference setting mode, is input to the fractal dimension setting device 27H as indicated by the broken line and stored as the reference value F6r.

Here, the fractal dimension F6 is, for example, a well-known function representing the property of a point sequence in a two-dimensional space, and a radius centered on any one of the points forming the point sequence. When the number of points in the circle of r is N (r), the number of points N (r) is represented by an index ν that satisfies the following expression (5). Therefore, if there is a time-series data part, the fractal dimension F6 can be obtained by considering this as a point in the two-dimensional space.

N (r) ≈rν (5)

However, in the formula (5), the variable ν corresponding to the fractal dimension F6 is an exponent, which is expressed in the form of multiplication (rν) with respect to the radius r for convenience.

On the other hand, in the component diagnostic mode, the fractal dimension F6 calculated by the fractal dimension calculator 26H is always input to the comparator 23H as indicated by the solid line, and the comparator 23H causes the reference value F6r of the fractal dimension to be calculated.
Compared to. In the comparator 23H, for example, the fractal dimension F6 calculated by the fractal dimension calculator 26H is
Fractal dimension F6r in fractal dimension setter 27H
When it exceeds three times the (reference value), the comparison result indicating the abnormality is output to the alarm device 24H. In response to this, the alarm device 2
4H issues an alarm A.

A series of operations in the component diagnosis mode is repeated, for example, in a fixed cycle, and the fractal dimension F6 of the deviation ΔF of the probability density distribution function F1 is sequentially obtained. The fractal dimension F6 is the fractal dimension (reference value F
6r), the fractal dimension F6 and the reference value F6r
Alarm A is issued only when the difference between Also in this case, the degree of complexity such as the waveform amplitude can be obtained from the difference between the fractal dimension F6 and the reference value F6r.

Example 10. (Claim 10 and Claim 1
5 to claim 17) In the ninth embodiment, the fractal dimension F6 of the deviation ΔF of the probability density distribution function F1 is calculated and compared with the reference value F6r. However, the Kalback information amount of the deviation ΔF is calculated. You may compare with a reference value.

A tenth embodiment of the present invention will be described below in which the deviation ΔF of the probability density distribution function F1 is used as the index data of the electric signal E and the Kullback information amount of the deviation ΔF is compared with the reference value. To do.

FIG. 15 is a functional block diagram showing the structure of the abnormality detection processing device 20 according to the tenth embodiment of the present invention.
23I, 24I, 26I, and 27I are the comparator 23H, the alarm device 24H, and the fractal dimension calculator 26 in FIG.
The H and fractal dimension setting devices 27H correspond to 25, 28, 29, Yi, and A, respectively.

In FIG. 15, 26I is the deviation calculator 29.
27I is a Kullback information amount setting unit that stores the Kullback information amount F7 obtained in the reference setting mode as the reference value F7r, and 23I is the Kullback information amount. The Kullback information amount F7 output from the computing unit 26I is set to the reference value F7 in the Kullback information amount setting unit 27I.
A comparator 24I for comparing with r is an alarm device for issuing an alarm A in response to the comparison result of the comparator 24I.

Next, referring to FIG. 1 and FIG.
The operation of the tenth embodiment of the present invention will be described. In this case, the deviation ΔF (N column data) calculated by the deviation calculator 29 is input to the Kullback information amount calculator 26I to calculate the Kullback information amount F7. In the reference setting mode, the Kullback information amount F7 is input to the Kullback information amount setting unit 27I as shown by the broken line and stored as the reference value F7r.

Here, the Kullback information amount F7 is one of the measures for comparing the two probability density distributions p (x) and q (x), and is the probability density distribution of the irregular signal at the normal time. By modeling the distribution situation, it is possible to determine whether the measurement signal is normal or abnormal by using a statistical hypothesis test method. Therefore, it becomes possible to detect an abnormality at a certain superior level. Further, since it is not necessary to know the signal characteristic at the time of abnormality in advance and only a plurality of signals at the time of normal need be measured, the present invention can be applied without depending on the target plant.

Here, the Kullback information amount F7 is set to I [p
(X), q (x)], if the column data p (x) and q (x) of the two probability density distributions are given, the sum operation is performed on the variable x in the range of −∞ to ∞. Due to
It can be expressed by the following equation (6).

I [p (x), q (x)] = Σ [p (x) · log {q (x) / p (x]} ... (6)

On the other hand, in the component diagnosis mode, the Kullback information amount F7 calculated by the Kullback information amount calculating unit 26I is always input to the comparator 23I as indicated by the solid line, and the comparator 23I indicates the Kullback information amount reference value F7r. Compared to. The comparator 23I indicates an abnormality when, for example, the Kullback information amount F7 calculated by the Kullback information amount calculation unit 26I exceeds three times the Kullback information amount F7r (reference value) in the Kullback information amount setting unit 27I. The comparison result shown is output to the alarm device 24I. In response to this, the alarm device 24I issues an alarm A.

A series of operations in the component diagnostic mode are repeated, for example, in a fixed cycle, and the Kullback information amount F7 of the deviation ΔF of the probability density distribution function F1 is sequentially obtained. Then, this Kullback information amount F7 is compared with the Kullback information amount (reference value F7r) in the normal state, and the alarm A is issued only when the difference between the Kullback information amount F7 and the reference value F7r is large. In this case, the processing can be simplified.

Example 11. (Claim 11 and Claim 1
5 to claim 17) In the tenth embodiment, the Kalback information amount F7 of the deviation ΔF of the probability density distribution function F1 is calculated to obtain the reference value F7r.
However, the average value of the deviation ΔF may be calculated and compared with the reference value.

Hereinafter, the deviation ΔF of the probability density distribution function F1 is used as the index data of the electric signal E, and the average value of the deviation ΔF is compared with the reference value.
1 will be described with reference to the drawings.

FIG. 16 is a functional block diagram showing the structure of the abnormality detection processing device 20 according to the eleventh embodiment of the present invention.
23J, 24J, 26J and 27J are the comparator 23I, the alarm device 24I and the Kullback information amount calculator 2 in FIG.
6I and Kullback information amount setting device 27I respectively, and 25, 28, 29, Yi and A are the same as those described above.

In FIG. 16, 26J is the deviation calculator 29.
An average value calculator for calculating the average value F8 of the deviation ΔF calculated in step 27, an average value setter 27J for storing the average value F8 obtained in the reference setting mode as the reference value F8r, 23
J is a comparator for comparing the average value F8 calculated by the average value calculator 26J with the reference value F8r in the average value setter 27J, 2
Reference numeral 4J is an alarm device which issues an alarm A in response to the comparison result of the comparator 24J.

Next, referring to FIG. 1 and FIG.
The operation of the eleventh embodiment of the present invention will be described. In this case, the deviation ΔF (N column data) calculated by the deviation calculator 29 is input to the average value calculator 26J and the average value F8 is calculated.
Is calculated. In the reference setting mode, the average value F8 is input to the average value setter 27J as a broken line and stored as the reference value F8r. The average value F8 is calculated as μ1 by the above-mentioned formula (1).

On the other hand, in the component diagnosis mode, the average value F8 calculated by the average value calculator 26J is always input to the comparator 23J as shown by the solid line, and the comparator 23J
Is compared with the average reference value F8r. Comparator 23
J is, for example, when the average value F8 calculated by the average value calculator 26J exceeds three times the average value F8r (reference value) in the average value setter 27J, a comparison result indicating an abnormality is given by an alarm device. Output to 24J. In response to this, the alarm device 24J issues an alarm A.

A series of operations in the component diagnosis mode are repeated, for example, in a constant cycle, and the average value F8 of the deviation ΔF from the normal of the probability density distribution function F1 is sequentially obtained. And this average value F
8 is compared with the average value (reference value F8r) in the normal state, and the alarm A is issued only when the difference between the average value F8 and the reference value F8r is large. Also in this case, the processing can be simplified.

Example 12. (Claim 12 and Claim 1
5 to claim 17) In the eleventh embodiment, the average value F8 of the deviation ΔF of the probability density distribution function F1 is calculated and compared with the reference value F8r. You may compare it with the value.

Hereinafter, the deviation ΔF of the probability density distribution function F1 is used as the index data of the electric signal E, and the variance value of the deviation ΔF is compared with the reference value.
2 will be described with reference to the drawings.

FIG. 17 is a functional block diagram showing the structure of the abnormality detection processing device 20 according to the twelfth embodiment of the present invention.
Reference numerals 23K, 24K, 26K and 27K correspond to the comparator 23J, the alarm device 24J, the average value calculator 26J and the average value setter 27J in FIG.
8, 29, Yi and A are the same as described above.

In FIG. 17, 26K is the deviation calculator 29.
A variance value calculator for calculating the variance value F9 of the deviation ΔF calculated in step 27K is a variance value setter for storing the variance value F9 obtained in the reference setting mode as a reference value F9r, 23K
K is a comparator for comparing the variance value F9 calculated by the variance value calculator 26K with the reference value F9r in the variance value setter 27K, 2
Reference numeral 4K is an alarm device which issues an alarm A in response to the comparison result of the comparator 24K.

Next, referring to FIG. 1 and FIG.
The operation of the twelfth embodiment of the present invention will be described. In this case, the deviation ΔF (N number of column data) calculated by the deviation calculator 29 is input to the variance value calculator 26K and the variance value F9 is input.
Is calculated. In the reference setting mode, the variance value F9 is input to the variance value setter 27K as indicated by the broken line and stored as the reference value F9r. The variance value F9 is calculated as μ2 by the above-mentioned formula (2).

On the other hand, in the component diagnostic mode, the variance value F9 calculated by the variance value calculator 26K is always input to the comparator 23K as shown by the solid line, and the comparator 23K.
Is compared with the reference value F9r of the variance value. Comparator 23
K is, for example, when the variance value F9 calculated by the variance value calculator 26K exceeds three times the variance value F9r (reference value) in the variance value setting unit 27K, a comparison result indicating an abnormality is output as an alarm device. Output to 24K. In response to this, the alarm device 24K issues an alarm A.

A series of operations in the component diagnosis mode are repeated, for example, in a fixed cycle, and the variance value F9 of the deviation ΔF from the normal of the probability density distribution function F1 is sequentially obtained. Then, the variance value F
9 is compared with the variance value (reference value F9r) at the normal time, and the alarm A is issued only when the difference between the variance value F9 and the reference value F9r is large. Also in this case, the processing can be simplified.

Example 13 (Claim 13 and Claim 1
5 to Claim 16) In the twelfth embodiment, the variance value F9 of the deviation ΔF of the probability density distribution function F1 is calculated and compared with the reference value F9r. You may compare it with the value.

In the following, the deviation ΔF of the probability density distribution function F1 is used as the index data of the electric signal E, and the skewness of the deviation ΔF is compared with a reference value.
Will be described with reference to FIG.

FIG. 18 is a functional block diagram showing the structure of the abnormality detection processing device 20 according to the thirteenth embodiment of the present invention.
Reference numerals 23L, 24L, 26L and 27L correspond to the comparator 23K, the alarm device 24K, the variance value calculator 26K and the variance value setter 27K in FIG.
8, 29, Yi and A are the same as described above.

In FIG. 18, 26L is a deviation calculator 29.
The skewness calculator for calculating the skewness μ5 of the deviation ΔF calculated in step 27L is the skewness μ5 obtained in the reference setting mode.
Is stored as a reference value μ5r, and 23L is a skewness setter 27 that sets the skewness μ5 calculated by the skewness calculator 26L.
A comparator for comparing with the reference value μ5r in L, 24L is an alarm device for generating an alarm A in response to the comparison result of the comparator 24L.

Next, referring to FIG. 1 and FIG.
The operation of the thirteenth embodiment of the present invention will be described. In this case, the deviation ΔF (N column data) calculated by the deviation calculator 29 is input to the skewness calculator 26L and the skewness μ5 is calculated. In the reference setting mode, the skewness μ5 is input to the skewness setter 27L as indicated by the broken line, and the reference value μ5r
Is stored as It should be noted that the skewness μ5 is calculated by the above equation (1).
~ Is calculated as µ3 by the formula (3).

On the other hand, in the component diagnostic mode, the skewness μ5 calculated by the skewness calculator 26L is always input to the comparator 23L as indicated by the solid line, and is compared with the reference value μ5r of the skewness by the comparator 23L. It The comparator 23L is
For example, the skewness μ5 calculated by the skewness calculator 26L is
When the skewness μ5r (reference value) in the skewness setter 27L exceeds three times, the comparison result indicating abnormality is output to the alarm device 24L. In response to this, the alarm device 24L issues an alarm A.

A series of operations in the component diagnostic mode are repeated, for example, in a fixed cycle, and the skewness μ5 of the deviation ΔF from the normal state of the probability density distribution function F1 is sequentially obtained. And this skewness μ5
Is compared with the skewness (reference value μ5r) under normal conditions,
The alarm A is issued only when the difference between the skewness μ5 and the reference value μ5r is large. Also in this case, the processing can be simplified.

Example 14. (Corresponding to Claims 14 to 17) In the above-described Embodiment 13, the skewness μ5 of the deviation ΔF of the probability density distribution function F1 is calculated and compared with the reference value μ5r, but the kurtosis of the deviation ΔF is calculated. Then, it may be compared with the reference value.

Hereinafter, the deviation ΔF of the probability density distribution function F1 is used as the index data of the electric signal E, and the kurtosis of the deviation ΔF is compared with the reference value.
Will be described with reference to FIG.

FIG. 19 is a functional block diagram showing the structure of the abnormality detection processing device 20 according to the fourteenth embodiment of the present invention.
23M, 24M, 26M and 27M correspond to the comparator 23L, the alarm device 24L, the skewness calculator 26L and the skewness setter 27L in FIG. 18, respectively, 25, 28,
29, Yi and A are the same as described above.

In FIG. 19, 26M is a deviation calculator 29.
The kurtosis calculator for calculating the kurtosis μ6 of the deviation ΔF calculated in step 27M is the kurtosis μ6 obtained in the reference setting mode.
Is stored as a reference value μ6r, 23M is a kurtosis setter 27 for the kurtosis μ6 calculated by the kurtosis calculator 26M.
A comparator for comparing with a reference value μ6r in M, and 24M is an alarm device for generating an alarm A in response to the comparison result of the comparator 24M.

Next, referring to FIG. 1 and FIG.
The operation of the thirteenth embodiment of the present invention will be described. In this case, the deviation ΔF (N column data) calculated by the deviation calculator 29 is input to the kurtosis calculator 26M to calculate the kurtosis μ6. In the reference setting mode, the kurtosis μ6 is input to the kurtosis setter 27M as indicated by a broken line and the reference value μ6r
Is stored as Note that the kurtosis μ6 is calculated by the above-mentioned equation (1).
Is calculated as μ4 by the equation (4).

On the other hand, in the component diagnostic mode, the kurtosis μ6 calculated by the kurtosis calculator 26M is always input to the comparator 23M as indicated by the solid line, and is compared with the reference value μ6r of the kurtosis by the comparator 23M. It The comparator 23M is
For example, the kurtosis μ6 calculated by the kurtosis calculator 26M is
When it exceeds three times the kurtosis μ6r (reference value) in the kurtosis setting device 27M, the comparison result indicating an abnormality is output to the alarm device 24M. In response to this, the alarm device 24M issues an alarm A.

A series of operations in the component diagnosis mode is repeated, for example, in a fixed cycle, and the kurtosis μ6 of the deviation ΔF from the normal of the probability density distribution function F1 is sequentially obtained. And this kurtosis μ6
Is compared with the kurtosis (reference value μ6r) in the normal state,
The alarm A is issued only when the difference between the kurtosis μ6 and the reference value μ6r is large. Also in this case, the processing can be simplified.

In each of the above embodiments, the transmission 1 (see FIG. 1) is targeted as a component that generates noise, but other components such as an engine (see FIG. 21) may be targeted. Needless to say, the same effect is achieved.

[0191]

As described above, according to the first aspect of the present invention, a microphone for detecting a sound generated from an automobile component and outputting an electric signal corresponding to the sound, and a microphone for the automobile based on the electric signal An abnormality detection processing device for detecting an abnormality of a component, wherein the abnormality detection processing device includes a skewness computing means for computing the skewness of the electric signal and a comparing means for comparing the skewness with a reference value. Based on the comparison result of, the warning is output when the skewness exceeds a predetermined relationship with the reference value, and the warning is issued if the detected sound is only the sound that exists in the normal state and the skewness is close to the normal value. There is no erroneous output. Therefore, by using a relatively simple process called skewness calculation, it is possible to accurately determine the minute difference between the normal state and the abnormal state of the sound generated by the automobile component, and prevent the outflow of defective products to the market. As a result, it is possible to greatly contribute to the improvement of the quality of automobile production, and it is possible to obtain a diagnostic device for automobile components with improved reliability of abnormality diagnosis.

Further, according to claim 2 of the present invention, a microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality of the automobile component is detected based on the electric signal. And an abnormality detection processing device, wherein the abnormality detection processing device includes a kurtosis calculation means for calculating the kurtosis of the electric signal and a comparison means for comparing the kurtosis with a reference value, and based on the comparison result of the comparison means. , If the kurtosis exceeds a predetermined value with respect to the reference value, an alarm is output, and if the detected sound is only the sound that exists in a normal state and the kurtosis is close to the normal value, it is not possible to erroneously output the alarm. Absent. Therefore, by using a relatively simple process called kurtosis calculation, it is possible to accurately determine the minute difference between the normal state and the abnormal state of the sound generated by the automobile component, and prevent the defective product from leaking to the market. As a result, it is possible to greatly contribute to the improvement of the quality of automobile production, and it is possible to obtain a diagnostic device for automobile components with improved reliability of abnormality diagnosis.

According to claim 3 of the present invention, a microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality of the automobile component is detected based on the electric signal. And a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal, and the probability density distribution function is compared with a reference value learned in advance by a neural network. Including comparison means,
Based on the comparison result of the comparison means, an alarm is output when the probability density distribution function exceeds a predetermined relationship with the reference value, and the detected sound is only the sound existing in the normal state and the probability density distribution function is normal. If it is close to the value, the alarm will not be erroneously output.
Therefore, it is possible to accurately discriminate a minute difference between the normal state and the abnormal state of the sound generated by the automobile component by learning processing using the neural network without preliminarily determining the calculation method. There is an effect that it is possible to prevent the outflow to the market and greatly contribute to the improvement of the quality of automobile production, and to obtain a diagnostic device for automobile components that improves reliability of abnormality diagnosis.

Further, according to claim 4 of the present invention, a microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality of the automobile component is detected based on the electric signal. And a probability density distribution function computing means for computing a probability density distribution function of an electric signal, and the probability density distribution function is compared with a preset fuzzy rule reference value. A comparison means, and based on the comparison result of the comparison means, outputs an alarm when the probability density distribution function exceeds a predetermined relationship with respect to the reference value, and the detected sound is only the sound existing in the normal state. If the probability density distribution function is close to the normal value, the alarm will not be erroneously output. Therefore, the logical processing using the fuzzy rules can accurately determine the minute difference between the normal state and the abnormal state of the sound generated by the automobile components, and prevent the defective products from leaking to the market. There is an effect that a diagnostic device for automobile components can be obtained, which can greatly contribute to quality improvement of automobile production and improve reliability of abnormality diagnosis.

Further, according to claim 5 of the present invention, a microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality of the automobile component is detected based on the electric signal. And a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal, and a comparison for comparing the probability density distribution function with a reference value of a preset pattern. Based on the comparison result of the comparison means, a warning is output when the probability density distribution function exceeds a predetermined relationship with the reference value, and the probability that the detected sound is only the sound existing in the normal state If the density distribution function is close to the normal value, the alarm will not be erroneously output. Therefore, especially when a characteristic appears in the waveform pattern of the electric signal, it is possible to accurately determine the minute difference between the normal state and the abnormal state of the sound generated by the automobile component, and prevent the defective product from leaking to the market. As a result, it is possible to greatly contribute to the improvement of the quality of automobile production, and it is possible to obtain a diagnostic device for automobile components with improved reliability of abnormality diagnosis.

Further, according to claim 6 of the present invention, a microphone that detects a sound generated from the automobile component and outputs an electric signal corresponding to the sound, and detects an abnormality of the automobile component based on the electric signal. And an abnormality detection processing device, the abnormality detection processing device calculates a deviation between a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal and a reference data obtained in advance. A deviation calculation means and a comparison means for comparing the deviation with a reference value learned in advance by a neural network are included, and an alarm is issued when the deviation exceeds a predetermined relationship with the reference value based on the comparison result of the comparison means. If the detected sound is only the sound existing in the normal state and the deviation is close to the normal value, the alarm is not erroneously output. Therefore, the learning process using the neural network based on the deviation of the probability density distribution function can accurately determine the minute difference between the normal state and the abnormal state of the sound generated by the automobile component without determining the calculation method in advance. Therefore, it is possible to prevent defective products from leaking to the market, greatly contribute to improving the quality of automobile production, and obtain a diagnostic device for automobile components with improved reliability of abnormality diagnosis.

Further, according to claim 7 of the present invention, a microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality of the automobile component is detected based on the electric signal. And an abnormality detection processing device, the abnormality detection processing device calculates a deviation between a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal and a reference data obtained in advance. When the deviation exceeds a predetermined relationship with the reference value based on the comparison result of the comparison means, including deviation calculation means and comparison means for comparing the deviation with a reference value of a fuzzy rule preset in the neural network. If the detected sound is only the sound existing in the normal state and the deviation is close to the normal value, the alarm is not erroneously output. Therefore, by a logical process using a fuzzy rule based on the deviation of the probability density distribution function, it is possible to accurately discriminate a minute difference between the normal state and the abnormal state of the sound generated by the automobile component. There is an effect that it is possible to prevent the outflow to the market and greatly contribute to the improvement of the quality of automobile production, and to obtain a diagnostic device for automobile components that improves reliability of abnormality diagnosis.

Further, according to claim 8 of the present invention, a microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality of the automobile component is detected based on the electric signal. And an abnormality detection processing device, the abnormality detection processing device calculates a deviation between a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal and a reference data obtained in advance. Deviation calculation means, pattern calculation means for obtaining a deviation pattern, and comparison means for comparing the pattern with a reference value, and based on the comparison result of the comparison means,
An alarm is output when the pattern exceeds a predetermined relationship with respect to the reference value, and the alarm is not erroneously output if the detected sound is only a sound existing in a normal state and the deviation is close to the normal value. Therefore, by using the pattern based on the deviation of the probability density distribution function, the complexity of the frequency components included in the electric signal is obtained, and the slight difference between the normal state and the abnormal state of the sound generated by the automobile component is accurately determined. Therefore, it is possible to prevent defective products from leaking to the market, greatly contribute to improving the quality of automobile production, and obtain a diagnostic device for automobile components with improved reliability of abnormality diagnosis.

According to claim 9 of the present invention, a microphone that detects a sound generated from a vehicle component and outputs an electric signal corresponding to the sound, and detects an abnormality of the vehicle component based on the electric signal. And an abnormality detection processing device, the abnormality detection processing device calculates a deviation between a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal and a reference data obtained in advance. It includes a deviation calculating means, a fractal dimension calculating means for obtaining a fractal dimension of the deviation, and a comparing means for comparing the fractal dimension with a reference value. Based on the comparison result of the comparing means, the fractal dimension has a predetermined relationship with the reference value. When it exceeds the threshold, an alarm is output, and if the detected sound is only the sound that exists in the normal state and the fractal dimension of the deviation is close to the normal value, an alarm is issued. There is no possibility that erroneous output. Therefore, by using the fractal dimension based on the deviation of the probability density distribution function, the complexity of the frequency components included in the electric signal is obtained, and the slight difference between the normal state and the abnormal state of the sound generated by the automobile component is calculated. There is an effect that it can be accurately discriminated, it can contribute to the quality improvement of automobile production by preventing the outflow of defective products to the market, and the diagnostic device for automobile components can improve the reliability of abnormality diagnosis. .

According to the tenth aspect of the present invention, a microphone that detects a sound generated from a vehicle component and outputs an electric signal corresponding to the sound, and an abnormality of the vehicle component is detected based on the electric signal. And an abnormality detection processing device, the abnormality detection processing device calculates a deviation between a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal and a reference data obtained in advance. A deviation calculating means, a kullback information amount calculating means for obtaining a kullback information amount of the deviation, and a comparing means for comparing the kullback information amount with a reference value, and based on the comparison result of the comparing means, the kullback information amount with respect to the reference value. Alarm is output when a predetermined relationship is exceeded, the detected sound is only the sound that exists in the normal state, and the amount of deviation Kalback information becomes the normal value. Does not output false alarm if Kere. Therefore, using a relatively simple process of calculating the Kalback information amount based on the deviation of the probability density distribution function,
It is possible to accurately determine the slight difference between the normal state and abnormal state of the sound generated by automobile components, prevent defective products from leaking to the market, and contribute significantly to improving the quality of automobile production. There is an effect that a diagnostic device for an automobile component with improved reliability can be obtained.

According to the eleventh aspect of the present invention, a microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality of the automobile component is detected based on the electric signal. And an abnormality detection processing device, the abnormality detection processing device calculates a deviation between a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal and a reference data obtained in advance. A deviation calculating means, an average value calculating means for obtaining an average value of the deviations, and a comparing means for comparing the average value with a reference value are included. Based on the comparison result of the comparing means, the average value has a predetermined relationship with the reference value. When it exceeds, the alarm is output, and if the detected sound is only the sound existing in the normal state and the average deviation is close to the normal value, the alarm is not erroneously output. Therefore, it is possible to accurately determine the minute difference between the normal state and the abnormal state of the sound generated by the automobile component by using the relatively simple process of calculating the average value based on the deviation of the probability density distribution function. There is an effect that a diagnostic device for automobile components, which can prevent the outflow of non-defective products to the market and greatly contribute to the quality improvement of automobile production, and improve the reliability of abnormality diagnosis.

According to the twelfth aspect of the present invention, a microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality of the automobile component is detected based on the electric signal. And an abnormality detection processing device, the abnormality detection processing device calculates a deviation between a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal and a reference data obtained in advance. A deviation calculating means, a variance value calculating means for obtaining a variance value of the deviation, and a comparing means for comparing the variance value with a reference value, and the variance value has a predetermined relationship with the reference value based on the comparison result of the comparing means. If the detected sound is only the sound existing in the normal state and the variance value of the deviation is close to the normal value, the alarm is not erroneously output. Therefore, it is possible to accurately determine the minute difference between the normal state and the abnormal state of the sound generated by the automobile component by using the relatively simple process of calculating the variance value based on the deviation of the probability density distribution function. There is an effect that a diagnostic device for automobile components, which can prevent the outflow of non-defective products to the market and greatly contribute to the quality improvement of automobile production, and improve the reliability of abnormality diagnosis.

According to the thirteenth aspect of the present invention, a microphone for detecting a sound generated from the automobile component and outputting an electric signal corresponding to the sound, and detecting an abnormality of the automobile component based on the electric signal. And an abnormality detection processing device, the abnormality detection processing device calculates a deviation between a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal and a reference data obtained in advance. A deviation calculating means, a skewness calculating means for obtaining a skewness of the deviation, and a comparing means for comparing the skewness with a reference value, and the skewness has a predetermined relationship with the reference value based on a comparison result of the comparing means. If the detected sound is only a sound existing in a normal state and the skewness of the deviation is close to the normal value, the alarm is not erroneously output. Therefore, by using a relatively simple process of skewness calculation based on the deviation of the probability density distribution function, it is possible to accurately determine the minute difference between the normal state and the abnormal state of the sound generated by the automobile component. There is an effect that a diagnostic device for automobile components, which can prevent the outflow of non-defective products to the market and greatly contribute to the quality improvement of automobile production, and improve the reliability of abnormality diagnosis.

According to the fourteenth aspect of the present invention, a microphone that detects a sound generated from a vehicle component and outputs an electric signal corresponding to the sound, and an abnormality of the vehicle component is detected based on the electric signal. And an abnormality detection processing device, the abnormality detection processing device calculates a deviation between a probability density distribution function calculating means for calculating a probability density distribution function of an electric signal and a reference data obtained in advance. The deviation calculation means, the kurtosis calculation means for obtaining the kurtosis of the deviation, and the comparison means for comparing the kurtosis with the reference value are included. Based on the comparison result of the comparison means, the kurtosis has a predetermined relationship with the reference value. When the difference exceeds the normal value, the alarm is output, and if the detected sound is only the sound existing in the normal state and the kurtosis of the deviation is close to the normal value, the alarm is not erroneously output. Therefore, by using a relatively simple process of kurtosis calculation based on the deviation of the probability density distribution function, it is possible to accurately discriminate a minute difference between the normal state and the abnormal state of the sound generated by the automobile component. There is an effect that a diagnostic device for automobile components, which can prevent the outflow of non-defective products to the market and greatly contribute to the quality improvement of automobile production, and improve the reliability of abnormality diagnosis.

According to a fifteenth aspect of the present invention, in any one of the sixth to fourteenth aspects, the reference data is set based on an electric signal obtained when the automobile component is in a normal state. So I decided to
There is an effect that a diagnostic device for an automobile component that reliably determines an abnormality in the automobile component can be obtained.

According to a sixteenth aspect of the present invention, in any one of the first to fifteenth aspects, the reference value is set based on an electric signal obtained when the automobile component is in a normal state. As a result, there is an effect that a diagnostic device for an automobile component that reliably determines an abnormality in the automobile component can be obtained.

According to the seventeenth aspect of the present invention, in any one of the first to sixteenth aspects, since the transmission or the engine is used as the automobile component, it is possible to determine the abnormality of the transmission or the engine. There is an effect that a diagnostic device for an automobile component with improved reliability can be obtained.

[Brief description of drawings]

FIG. 1 is a block diagram showing a schematic configuration of embodiments 1 to 14 (corresponding to claims 1 to 17) of the present invention.

FIG. 2 is a functional block diagram showing a configuration of an abnormality detection processing device according to a first embodiment (corresponding to claim 1, claim 16 and claim 17) of the present invention.

FIG. 3 is a functional block diagram showing a configuration of an abnormality detection processing device according to a second embodiment (corresponding to claim 2, claim 16 and claim 17) of the present invention.

FIG. 4 is a functional block diagram showing a configuration of an abnormality detection processing device according to a third embodiment (corresponding to claim 3, claim 16 and claim 17) of the present invention.

5 is a waveform diagram showing an example of a reference signal at a normal time used in the probability density distribution function calculator in FIG.

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

FIG. 7 is a functional block diagram showing a configuration of an abnormality detection processing device according to a fourth embodiment (corresponding to claims 4, 16 and 17) of the present invention.

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

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

FIG. 10 is a fifth embodiment of the present invention (claims 5 and 16).
And (corresponding to claim 17) is a functional block diagram showing a configuration of an abnormality detection processing device.

FIG. 11 is a functional block diagram showing a configuration of an abnormality detection processing device according to a sixth embodiment (corresponding to claim 6 and claims 15 to 17) of the present invention.

FIG. 12 is a functional block diagram showing a configuration of an abnormality detection processing device according to a seventh embodiment (corresponding to claim 7 and claims 15 to 17) of the present invention.

FIG. 13 is a functional block diagram showing a configuration of an abnormality detection processing device according to an eighth embodiment (corresponding to claim 8 and claims 15 to 17) of the present invention.

FIG. 14 is a functional block diagram showing the configuration of an abnormality detection processing device according to a ninth embodiment (corresponding to claims 9 and 15 to 17) of the present invention.

FIG. 15 is a functional block diagram showing a configuration of an abnormality detection processing device according to Embodiment 10 (corresponding to claims 10 and 15 to 17) of the present invention.

FIG. 16 is a functional block diagram showing a configuration of an abnormality detection processing device according to an eleventh embodiment (corresponding to claim 11 and claims 15 to 17) of the present invention.

FIG. 17 is a functional block diagram showing a configuration of an abnormality detection processing device according to a twelfth embodiment of the invention (corresponding to claims 12 and 15 to 17).

FIG. 18 is a functional block diagram showing a configuration of an abnormality detection processing device according to a thirteenth embodiment (corresponding to claims 13 and 15 to 17) of the present invention.

FIG. 19 is a functional block diagram showing a configuration of an abnormality detection processing device according to a fourteenth embodiment (corresponding to claims 14 to 16) of the present invention.

FIG. 20 is a block diagram showing a schematic configuration of a conventional vehicle component diagnostic device.

FIG. 21 is a configuration diagram showing another example of a general automobile component.

[Explanation of symbols]

1 Automotive transmission (automotive component), 7
Microphone, 20 abnormality detection processing device, 21, 26L
Skewness calculator, 21A, 26M Kurtosis calculator, 22, 27
L skewness setting device, 22A, 27M kurtosis setting device, 23,
23A-23M comparator, 24, 24A-24M alarm device, 25 probability density distribution function calculator, 26, 26E neural network calculator, 26C feature extractor, 2
6D waveform feature extractor, 26H fractal dimension calculator,
26I Kullback information amount calculator, 26J average value calculator, 26K variance value calculator, 27, 27D, 27E, 2
7G reference value setting device, 27C, 27F membership function setting device, 27H fractal dimension setting device, 27I Kullback information amount setting device, 27J average value setting device, 27K
Variance value setter, 28 reference data setter, 29 deviation calculator, 31 automobile engine (vehicle component), a alarm, E electric signal, F1 probability density distribution function, F2 feature, F3, F5 waveform feature (pattern) ,
F6 fractal dimension, F7 Kullback information amount, F8
Average value, F9 variance value, F1r reference data, ΔF
Deviation, F2r, F4r membership function (reference value of fuzzy rule), F3r, F5r pattern reference value, F6r fractal dimension reference value, F7r Kullback information amount reference value, F8r average value reference value, F9r
Variance value reference value, N1, N2 learning value, S generated sound,
Yi column data, μ3, μ5 skewness, μ4, μ6 kurtosis, N1r, N2r, μ3r, μ4r, μ5r, μ6r
Standard value.

Continuation of front page (51) Int.Cl. ⁶ Identification code Internal reference number FI technical display location // G06F 15/18 540 A 8837-5L 560 G 8837-5L

Claims (17)

[Claims] 1. A microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality detection processing device that detects an abnormality of the automobile component based on the electric signal. The abnormality detection processing device includes a skewness computing unit that computes the skewness of the electric signal, and a comparing unit that compares the skewness with a reference value. Based on a comparison result of the comparing unit, the strain A diagnostic device for a vehicle component, which outputs an alarm when the degree exceeds a predetermined relationship with the reference value. 2. A microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality detection processing device that detects an abnormality of the automobile component based on the electric signal. The abnormality detection processing device includes a kurtosis calculation unit that calculates a kurtosis of the electric signal, and a comparison unit that compares the kurtosis with a reference value, and the kurtosis is based on a comparison result of the comparison unit. A diagnostic device for a vehicle component, which outputs an alarm when the degree exceeds a predetermined relationship with the reference value. 3. A microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality detection processing device that detects an abnormality of the automobile component based on the electric signal. The abnormality detection processing device comprises a probability density distribution function calculating means for calculating a probability density distribution function of the electric signal, and a comparing means for comparing the probability density distribution function with a reference value previously learned by a neural network. A diagnostic device for an automobile component, comprising: outputting an alarm when the probability density distribution function exceeds a predetermined relationship with respect to the reference value based on a comparison result of the comparison means. 4. A microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality detection processing device that detects an abnormality of the automobile component based on the electric signal. Comprising, the abnormality detection processing device, a probability density distribution function calculation means for calculating a probability density distribution function of the electrical signal, and a comparison means for comparing the probability density distribution function with a reference value of a preset fuzzy rule. A diagnostic device for an automobile component, comprising: outputting an alarm when the probability density distribution function exceeds a predetermined relationship with respect to the reference value based on a comparison result of the comparison means. 5. A microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality detection processing device that detects an abnormality of the automobile component based on the electric signal. The abnormality detection processing device includes a probability density distribution function calculation unit that calculates a probability density distribution function of the electric signal, and a comparison unit that compares the probability density distribution function with a reference value of a preset pattern. A diagnostic device for an automobile component, which outputs an alarm when the probability density distribution function exceeds a predetermined relationship with respect to the reference value based on a comparison result of the comparison means. 6. A microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality detection processing device that detects an abnormality of the automobile component based on the electric signal. The abnormality detection processing device comprises a probability density distribution function calculating means for calculating a probability density distribution function of the electric signal, and a deviation calculating means for calculating a deviation between the probability density distribution function and reference data obtained in advance. Comparing the deviation with a reference value learned in advance by a neural network, based on a comparison result of the comparing means, an alarm is issued when the deviation exceeds a predetermined relationship with the reference value. A diagnostic device for an automobile component, which is characterized by outputting. 7. A microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality detection processing device that detects an abnormality of the automobile component based on the electric signal. The abnormality detection processing device comprises a probability density distribution function calculating means for calculating a probability density distribution function of the electric signal, and a deviation calculating means for calculating a deviation between the probability density distribution function and reference data obtained in advance. A comparison means for comparing the deviation with a reference value of a fuzzy rule preset in the neural network, based on a comparison result of the comparison means, when the deviation exceeds a predetermined relationship with the reference value. A diagnostic device for a vehicle component, which outputs an alarm to a vehicle. 8. A microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality detection processing device that detects an abnormality of the automobile component based on the electric signal. The abnormality detection processing device comprises a probability density distribution function calculating means for calculating a probability density distribution function of the electric signal, and a deviation calculating means for calculating a deviation between the probability density distribution function and reference data obtained in advance. A pattern calculation unit for obtaining the deviation pattern and a comparison unit for comparing the pattern with a reference value, and the pattern exceeds a predetermined relationship with the reference value based on a comparison result of the comparison unit. A diagnostic device for automobile components, which is characterized by outputting an alarm at times. 9. A microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality detection processing device that detects an abnormality of the automobile component based on the electric signal. The abnormality detection processing device comprises a probability density distribution function calculating means for calculating a probability density distribution function of the electric signal, and a deviation calculating means for calculating a deviation between the probability density distribution function and reference data obtained in advance. A fractal dimension calculating means for obtaining a fractal dimension of the deviation, and a comparing means for comparing the fractal dimension with a reference value, based on a comparison result of the comparing means, the fractal dimension is a predetermined value with respect to the reference value. A diagnostic device for an automobile component, which outputs an alarm when the relationship is exceeded. 10. A microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality detection processing device that detects an abnormality of the automobile component based on the electric signal. The abnormality detection processing device comprises a probability density distribution function calculating means for calculating a probability density distribution function of the electric signal, and a deviation calculating means for calculating a deviation between the probability density distribution function and reference data obtained in advance. , A Kalback information amount calculating unit for obtaining the Kullback information amount of the deviation, and a comparing unit for comparing the Kullback information amount with a reference value, and based on the comparison result of the comparing unit, the Kullback information amount is equal to the reference value. A diagnostic device for an automobile component, which outputs an alarm when a predetermined relationship is exceeded. 11. A microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality detection processing device that detects an abnormality of the automobile component based on the electric signal. The abnormality detection processing device comprises a probability density distribution function calculating means for calculating a probability density distribution function of the electric signal, and a deviation calculating means for calculating a deviation between the probability density distribution function and reference data obtained in advance. An average value calculating means for obtaining an average value of the deviations, and a comparing means for comparing the average value with a reference value, based on a comparison result of the comparing means, the average value is a predetermined value with respect to the reference value. A diagnostic device for an automobile component, which outputs an alarm when the relationship is exceeded. 12. A microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality detection processing device that detects an abnormality of the automobile component based on the electric signal. The abnormality detection processing device comprises a probability density distribution function calculating means for calculating a probability density distribution function of the electric signal, and a deviation calculating means for calculating a deviation between the probability density distribution function and reference data obtained in advance. A variance value calculating means for obtaining a variance value of the deviation, and a comparing means for comparing the variance value with a reference value. Based on a comparison result of the comparing means, the variance value is predetermined with respect to the reference value. A diagnostic device for an automobile component, which outputs an alarm when the relationship is exceeded. 13. A microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality detection processing device that detects an abnormality of the automobile component based on the electric signal. The abnormality detection processing device comprises a probability density distribution function calculating means for calculating a probability density distribution function of the electric signal, and a deviation calculating means for calculating a deviation between the probability density distribution function and reference data obtained in advance. A skewness calculating unit that obtains a skewness of the deviation, and a comparing unit that compares the skewness with a reference value. Based on a comparison result of the comparing unit, the skewness is predetermined with respect to the reference value. A diagnostic device for an automobile component, which outputs an alarm when the relationship is exceeded. 14. A microphone that detects a sound generated from an automobile component and outputs an electric signal corresponding to the sound, and an abnormality detection processing device that detects an abnormality of the automobile component based on the electric signal. The abnormality detection processing device comprises a probability density distribution function calculating means for calculating a probability density distribution function of the electric signal, and a deviation calculating means for calculating a deviation between the probability density distribution function and reference data obtained in advance. A kurtosis calculation unit that obtains the kurtosis of the deviation, and a comparison unit that compares the kurtosis with a reference value. Based on the comparison result of the comparison unit, the kurtosis is predetermined with respect to the reference value. A diagnostic device for an automobile component, which outputs an alarm when the relationship is exceeded. 15. The automobile according to claim 6, wherein the reference data is set based on the electric signal obtained when the automobile component is in a normal state. Component diagnostic equipment. 16. The method according to claim 1, wherein the reference value is set based on the electric signal obtained when the vehicle component is in a normal state.
A diagnostic device for any of the automobile components up to 5. 17. The diagnostic device for an automobile component according to claim 1, wherein the automobile component is an automobile transmission or an automobile engine.