JP2004045418A - Determination method for abnormality and device therefor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a determination method for an abnormality and a device therefor designed to be capable of stably determining various kinds of the normality and the abnormality of a product with a vibration portion, specially precisely extracting an intermittent component. <P>SOLUTION: The determination method and device comprise an analyzing means for the intermittent component analyzing an intermittent component from measured data measured by a measuring means attached to the product with the vibration portion. This analyzing means for the intermittent component includes a means pre-treating the intermittent component and a means for extracting a characteristic quantity by analyzing the pre-treated intermittent component. <P>COPYRIGHT: (C)2004,JPO

Description

The present invention relates to an abnormality determination method and apparatus for determining an abnormality of a product having a vibrating portion, and more specifically, a device having a vibrating portion such as a motor or a compressor, a product using the device having the vibrating portion, and a device using the vibrating portion. The present invention relates to an abnormality determination method and apparatus for determining whether a product driven by a device has normal or abnormal status.

In general, in a production factory, if a device having a vibrating portion such as a motor or a compressor, a product using the device having the vibrating portion, or a product driven by the device having the vibrating portion is determined to be normal or abnormal, the product The product is actually operated before shipment, and the inspector judges whether it is normal or abnormal by a so-called sensory test that listens to the ears to see if any abnormal sound is generated or touches the hand to check the vibration, This ensures quality assurance.

However, the result of the sensory test performed by the inspector as to whether the product is normal or abnormal has large variations due to individual differences or changes due to time.Furthermore, the result of the sensory test is difficult to convert into data and numerical values. There was a problem that management was also difficult.

Therefore, it has been considered to automate the determination of normality or abnormality of the above-mentioned products. As a conventional technology that enables this automation, a measuring instrument called an FFT (Fast Fourier Transform) analyzer or a computer system incorporating the same has been proposed. Have been.

FIG. 25 is a block diagram showing a processing method by a conventional computer system configured by incorporating an FFT analyzer.

In FIG. 25, an analog / digital converter (A / D converter) 1 converts a sensor signal from a sensor such as an acceleration sensor attached to a product to be inspected into a digital signal. The digital signal converted by the A / D conversion unit 1 is pre-processed by a pre-processing unit 2 based on a window function, and the pre-processed signal is converted into a frequency axis waveform by a FFT conversion unit 3 by a FFT. I do.

The component of the specific frequency band is extracted from the signal converted into the frequency axis waveform by the component extraction unit 4 of the specific frequency band, and based on the extracted component of the specific frequency band, the determination unit 5 determines whether the product is normal or abnormal. Is determined.

Here, in the processing in the transform unit 3 by FFT, as shown in FIG. 26, the time waveform χ (t) can be converted into a frequency component X (ω) by Fourier transform, and the frequency component X (ω) is The frequency analysis is performed by using the FFT that utilizes the relation that the time waveform can be converted into the time waveform χ (t) by the inverse transform. The signal preprocessed by the preprocessing unit 2 is processed by using the frequency analysis technique using the FFT. It is converted into a frequency axis waveform composed of a set of frequency components X (ω).

However, the frequency analysis using the FFT is a very effective analysis method when a wave included in the signal under test is repeatedly generated at a correct cycle. In the case of adopting the inspection, there are the following problems.

1) It is difficult to extract intermittent components.

That is, as shown in FIG. 27, the waveform generated from such a normal or abnormal test object often includes irregular or intermittently generated intermediate frequency components to high frequency components.

Here, according to the frequency analysis using the FFT, a wave having a large amplitude intermittently generated as shown in FIG. 28A and a wave having a small amplitude continuously generated as shown in FIG. Thus, the same analysis result is obtained, and the distinction cannot be made. Therefore, a non-defective product in which a continuous small-amplitude wave as shown in FIG. 28B is generated and a defective product in which an intermittent large-amplitude wave is generated as shown in FIG. It is difficult to identify.

2) Poor repeatability.

That is, in the frequency analysis using FFT, data that is actually continuous indefinitely is cut out for a certain finite time, and the data is connected before and after to process as a continuous waveform. Therefore, discontinuity of data occurs in the cut portion, which affects the analysis result.

In order to prevent this phenomenon, a method of removing discontinuities at both ends using a window function is generally adopted. However, according to the processing using this window function, an FFT is called after a forced operation. Dynamic conversion is performed, and as a result, the repetition accuracy of the FFT processing deteriorates.

As a conventional technique using a method other than the frequency analysis using the FFT, a so-called filter method as shown in FIG. 29 is known.

In FIG. 29, the analysis method using the filter method converts a sensor signal from a sensor such as an acceleration sensor attached to a product to be inspected into a digital signal by an analog / digital converter (A / D converter) 1. The digital signal converted by the A / D conversion unit 1 is pre-processed by a pre-processing unit 6 such as a frequency filter, and the pre-processed signal is subjected to a feature extraction by a feature extraction unit 7 using a specific frequency band component or an index function. Is extracted. Then, whether the product is normal or abnormal is determined by determining this characteristic amount by the fuzzy or neuro determination unit 8.

That is, the analysis method using the filter method obtains a characteristic amount such as a specific frequency component by performing one or more mathematical processing or filtering on the measured time signal, and determines whether the product is normal or abnormal based on the result. However, when there are a plurality of types of abnormalities to be detected, it is difficult to detect all of them by the analysis method using the filter method.

Therefore, an object of the present invention is to provide an abnormality determination method and apparatus that can stably determine various normal and abnormal states of a product having a vibrating section.

In order to achieve the above object, the invention of claim 1 is:
In an abnormality determination method for determining an abnormality of the product based on measurement data measured by a measurement unit,
A time axis waveform analysis for obtaining a time axis waveform from the measurement data and analyzing the time axis waveform;
Perform frequency axis waveform analysis to determine the frequency axis waveform from the measurement data and analyze the frequency axis waveform,
An abnormality of the product is determined from a comprehensive determination result of the time axis waveform analysis and the frequency axis waveform analysis.

The invention of claim 2 is the invention according to claim 1,
Storing the measurement data in storage means,
The time axis waveform analysis and the frequency axis waveform analysis
It is characterized by being executed in parallel based on the measurement data stored in the storage means.

The invention according to claim 3 is the invention according to claim 2,
Dividing the measurement data stored in the storage means into a plurality of data blocks,
The time axis waveform analysis and the frequency axis waveform analysis
It is characterized in that feature amount data is extracted from the analysis result in units of the divided data blocks.

The invention according to claim 4 is the invention according to claim 3,
Judgment of abnormality of the above products
It is performed based on the maximum value and the average value of the feature amount data in the data block unit.

The invention of claim 5 is the invention according to claim 3,
The above time axis waveform analysis
A frequency filter analysis for filtering the measurement data using a frequency filter and extracting a characteristic amount based on the filtered waveform;
An intermittent component analysis for extracting a feature amount based on the intermittent component of the measurement data,
A high-frequency component analysis for extracting a characteristic amount based on the high-frequency component of the measurement data,
And a low-frequency component analysis for extracting a characteristic amount based on the low-frequency component of the measurement data.

The invention according to claim 6 is the invention according to claim 5,
The frequency filter analysis above
A root mean square value, a peak value, an extreme value number, an extreme value difference, and a slope average are extracted from the waveform data obtained by the filtering using the frequency filter as the feature amount.

The invention according to claim 7 is the invention according to claim 5,
The intermittent component analysis above
The above measurement data is Hilbert-transformed, a data part before the conversion is a real part, and a data part after the conversion is an imaginary part to form a complex array, a root-mean-square array of the complex array is calculated, and a smoothing process such as a moving average is performed. After that, the data is passed through a low-pass filter, and the peak value and the extreme value difference and the difference between the maximum value and the minimum value between the data blocks are extracted from the resulting waveform data as the feature amount.

The invention according to claim 8 is the invention according to claim 5,
The high frequency component analysis is
The method is characterized in that the extreme value difference is extracted as the feature amount from the waveform data of the high-frequency component of the measurement data.

The invention of claim 9 is the invention according to claim 5,
The low frequency component analysis
A peak value and an extreme value difference and a difference between a maximum value and a minimum value between the data blocks are extracted from the waveform data of the low frequency component of the measurement data as the feature amount.

The invention according to claim 10 is the invention according to claim 3,
The frequency axis waveform analysis above
The measurement data is converted into a frequency axis waveform using fast Fourier transform, and a component in an arbitrary frequency band of the frequency axis waveform is extracted as a feature amount.

The invention of claim 11 is
In an abnormality determination device that determines an abnormality of the product based on measurement data measured by a measurement unit,
A time axis waveform analysis means for obtaining a time axis waveform from the measurement data and executing a time axis waveform analysis based on the time axis waveform;
Frequency axis waveform analysis means for obtaining a frequency axis waveform from the measurement data and executing a frequency axis waveform analysis based on the frequency axis waveform,
Comprehensive determination means for comprehensively determining an abnormality of the product from both the analysis result of the time axis waveform analysis means and the analysis result of the frequency axis waveform analysis means,
Storage means for storing the measurement data,
With
The time axis waveform analysis by the time axis waveform analysis means and the frequency axis waveform analysis by the frequency axis waveform analysis means are performed in parallel based on the measurement data stored in the storage means.

The invention according to claim 12 is the invention according to claim 11,
The time axis waveform analysis by the time axis waveform analysis means and the frequency axis waveform analysis by the frequency axis waveform analysis means are:
The measurement is performed by dividing the measurement data stored in the storage unit into a plurality of data blocks, and extracting feature amount data from an analysis result of the divided data blocks.
The above time axis waveform analysis
Frequency filter analysis means for filtering the measurement data using a frequency filter and extracting a characteristic amount based on the filtered waveform,
An intermittent component analysis means for extracting a characteristic amount based on the intermittent component of the measurement data,
High-frequency component analysis means for extracting a characteristic amount based on the high-frequency component of the measurement data,
Low frequency component analysis means for extracting a characteristic amount based on the low frequency component of the measurement data,
It is characterized by having.

The invention according to claim 13 is the invention according to claim 11,
The time axis waveform analysis by the time axis waveform analysis means and the frequency axis waveform analysis by the frequency axis waveform analysis means are:
The measurement is performed by dividing the measurement data stored in the storage unit into a plurality of data blocks, and extracting feature amount data from an analysis result of the divided data blocks.
The frequency axis waveform analysis means,
Frequency axis waveform analysis means for converting the measurement data into a frequency axis waveform using a fast Fourier transform,
Frequency axis waveform analysis feature value extraction means for extracting a component in an arbitrary frequency band of the frequency axis waveform processed by the frequency axis waveform analysis pre-processing means as a frequency axis waveform analysis feature value,
It is characterized by having.

The invention of claim 14 is
In an abnormality determination device that determines an abnormality of the product based on measurement data measured by a measurement unit,
Having intermittent component analysis means for analyzing the intermittent component of the measurement data,
The above-mentioned intermittent component analysis means,
Intermittent component analysis preprocessing means for preprocessing the intermittent component,
An intermittent component analysis feature amount extracting unit that analyzes the intermittent component preprocessed by the intermittent component analysis preprocessing unit and extracts a feature amount;
It is characterized by having.

The invention according to claim 15 is the invention according to claim 14,
The intermittent component analysis pre-processing means,
Hilbert transform means for Hilbert transforming the measurement data,
A real number part of the data before the conversion of the Hilbert conversion means, a complex number array generating means for generating a complex number array with the converted data as an imaginary part,
A root mean square array calculating means for calculating a root mean square array of the complex number array generated by the complex number array generating means,
Smoothing processing means for performing a smoothing process such as moving average of the array of the root mean square calculated by the root mean square array calculating means,
Low-pass filter means for extracting a low-frequency component from the processing result of the smoothing processing means,
It is characterized by having.

The present invention provides a time axis waveform analysis for obtaining a time axis waveform from measurement data measured by a measurement unit attached to a product having a vibrating portion, and analyzing the time axis waveform, and obtaining a frequency axis waveform from the measurement data. The frequency axis waveform analysis for analyzing the frequency axis waveform is performed, and the time axis waveform analysis and the frequency axis waveform analysis are configured to determine the abnormality of the product from the comprehensive determination result. There is an effect that it is possible to stably determine various normal and abnormal states.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a functional block diagram showing an embodiment of an abnormality determination method in an abnormality determination system configured by applying the abnormality determination method and apparatus according to the present invention.

In FIG. 1, in the abnormality determination method in the abnormality determination system, a plurality of measurement data measured by a measurement sensor attached to a product having a vibrating part are parallelly analyzed from both time axis waveform analysis and frequency axis analysis. Processing is performed, a plurality of feature amounts are calculated from the respective processed data, and normality or abnormality of the product is determined by inferring the calculated amounts.

Specifically, a sensor signal from a measurement sensor attached to a product having a vibrating section is converted into digital data by an analog / digital conversion section (A / D conversion section) 10 and the converted digital data (hereinafter referred to as “original data”). (Referred to as data) in the memory 20.

The original data stored in the memory 20 is divided into fixed data blocks of, for example, 1024 points in units of conversion data of the A / D conversion unit 10, and the subsequent processing, that is, the parallel preprocessing described below. The processing up to the calculation of the feature amount function is performed in units of the fixed data block.

That is, the original data stored in the memory 20 is first processed by the preprocessing 1 (filtering) unit 30-1, the preprocessing 2 (intermittent component amplification processing) unit 30-2, and the preprocessing 3 (low frequency amplification processing) unit 30 -3, a preprocessing 4 (high-frequency amplification processing) section 30-4, and a frequency conversion section 30-5 using FFT, where the preprocessing is performed in parallel.

Here, the pre-processing by the pre-processing 1 (filtering) unit 30-1 applies an analog or digital frequency filter to the original data stored in the memory 20 to remove noise components and background noise signals of the original data. This is an effective process particularly when the frequency band of the abnormal signal to be detected is known in advance.

The preprocessing 2 (intermittent component amplification processing) section 30-2 performs intermittent component amplification of the original data stored in the memory 20. Specifically, as described later,
1) Hilbert transform the original data 2) Create a complex array using the real data as the real part and the Hilbert transformed data as the imaginary part 3) Calculate the square root array of the complex array 4) Perform smoothing processing such as moving average 5) Consists of processing that passes through a low-pass filter.

The pre-processing by the pre-processing 2 (intermittent component amplification processing) unit 30-2 is effective when the original data in the normal state includes a periodic low frequency to a high frequency.

The pre-processing by the pre-processing 3 (low frequency amplification processing) unit 30-3 is as follows.
1) The original data is passed through a low-pass filter. 2) The original data passed through the low-pass filter is integrated in a short time. The pre-processing by the pre-processing 3 (low-frequency amplification processing) unit 30-3 is effective when the original data in a normal state contains some high-frequency components and an abnormal signal to be detected exists in a low-frequency region. It is.

The pre-processing by the pre-processing 4 (high-frequency amplification processing) unit 30-4 is
1) The original data is passed through a high-pass filter. 2) The original data passed through the high-pass filter is differentiated in a short time. The pre-processing by the pre-processing 4 (high-frequency amplification processing) unit 30-4 is effective when the original data in a normal state contains a certain low-frequency component and an abnormal signal to be detected exists in a high-frequency region. is there.

The pre-processing by the frequency conversion unit 30-5 using the FFT converts the time-axis waveform, which is the original data, into the frequency-axis data using the FFT (Fast Fourier Transform), which is converted by the FFT (Fast Fourier Transform). When the frequency axis data is used, a state in which an abnormal component is included can be quantified by using a component in an arbitrary frequency band as a feature amount. Although there are information lost due to FFT (Fast Fourier Transform) processing and abnormal states that cannot be detected, these abnormal states are detected based on the original data stored in the memory 20 in the above pre-processing 30-1 to 30-. This can be supplemented by analysis using the processing of No. 4.

The data pre-processed by the pre-processing 1 (filtering) unit 30-1 is passed to the feature amount calculation processing unit 40-1, where the amplitude component of the data pre-processed by filtering is extracted and the feature is obtained by mathematical processing. A quantity operation is performed.

The data preprocessed by the preprocessing 2 (intermittent component amplification process) unit 30-2 is passed to the feature value calculation processing unit 40-2, where the feature value of the data preprocessed by the intermittent component amplification process is used. An operation is performed.

The data pre-processed by the pre-processing 3 (high-frequency amplification processing) unit 30-3 is passed to the feature value calculation processing unit 40-3, where the feature value calculation of the data pre-processed by the high-frequency amplification process is performed. Be executed.

The data pre-processed by the pre-processing 4 (low-frequency amplification processing) unit 30-4 is passed to the feature-value calculation processing unit 40-4, where the feature value of the data pre-processed by the low-frequency amplification process is output. An operation is performed.

The data pre-processed by the FFT-based frequency conversion unit 30-5 is passed to the feature value calculation processing unit 40-5, where the feature value calculation of the data pre-processed by the frequency change by FFT is performed. .

Here, the feature value computations of the feature value computation processing units 40-1 to 40-5 are performed in parallel in the same manner as the preprocessing 30-1 to 30-5.

The feature value calculation processing in the feature value calculation processing units 40-1 to 40-5 is for quantifying the feature of the measurement data, and a specific function is performed based on the data preprocessed in the preprocessing 30-1 to 30-5. By executing the calculation, a feature amount as information indicating the degree of abnormality is extracted.

特 徴 As the feature value calculation processing for extracting the feature value, a calculation using a mathematical function such as RMS (square root) calculation, averaging calculation, or a statistical function, which is generally used conventionally, can be used.

In the abnormality determination system according to the present embodiment, in addition to the mathematical functions such as the RMS (square root) calculation and the average calculation, and the calculation using the statistical function, the calculation processing using the following calculation function is adopted.

1) Feature value calculation function for finding the number of extreme values The feature value calculation function for finding the number of extreme values executes a calculation for detecting the number of extreme values of the waveform from the original data array or the data array after preprocessing.

The number of extrema of the waveform of the original data array or the pre-processed data array is determined as an extremum when the following equation is satisfied with respect to data χi (where i = 1, 2,...) Of the data array. , The value obtained by counting the number of extreme values in the data array is the number of extreme values in the data array.

That is, for three consecutive vibration data χm-1, χm, χm + 1 in the data array,
a) | χm-χm-1 | ≦ α
b) | χm + 1-χm | ≦ α
c) (χm-χm-1) × (χm + 1-χm) <0
When the above three conditions are simultaneously satisfied, the vibration data Δm is determined to be an extreme value. Here, α is a variable. The closer the variable α is to “0”, the more the extreme value can be detected from a minute change. If the variable α is set to a relatively large value, the extreme value ignoring an erroneous change can be detected. .

The feature value calculation function for calculating the number of extremums is calculated by the feature value calculation processing unit 40-4 which extracts a feature value from the data array preprocessed by the preprocessing 4 (low frequency amplification process) unit 30-4. Effective in processing. In other words, in a state where the low-frequency component of the measurement data has an abnormality, the low-frequency component is amplified by the pre-processing by the pre-processing 4 (low-frequency amplification processing) unit 30-4, so that the noise-like small extreme value And the number of extreme values becomes smaller than that in a normal state, so that it is easy to grasp the feature amount.

2) Feature value calculation function for finding an extreme value difference The feature value calculation function for finding an extreme value difference executes a calculation for detecting the number of extreme values of a waveform from an original data array or a data array after preprocessing.

In the feature value calculation function for finding the extreme value difference, for the extreme values in the data array obtained by the feature value calculation function for finding the number of extreme values, an array of the absolute value of the difference between the preceding and following extreme values is obtained. Then, the value calculated by the following procedure is calculated as an extreme value difference.

That is, assuming that the data in the array obtained by calculating the absolute value of the difference between the preceding and following extreme values is D1, D2,... Dn, k data are extracted in descending order from the data, and the averaged value is defined as the extreme value. Value difference. Here, k is a variable.

That is, in the feature value calculation function for calculating the extreme value difference, by appropriately setting the variable k, the difference cannot be detected compared with the normal state by the average value or the RMS calculation within a certain time, but the difference is temporarily stored in the measurement data. This is effective for detecting a normal or abnormal state when a high-frequency level change is included periodically or intermittently.

3) Function for calculating the difference between the maximum value and the minimum value between the data blocks The function for calculating the characteristic value for calculating the difference between the maximum and minimum values between the data blocks is a function for calculating the difference between the original data array or the pre-processed data array. The difference between the maximum value and the minimum value, that is, the calculation of (maximum value)-(minimum value) is executed.

The feature value calculation function for calculating the difference between the maximum value and the minimum value between the data blocks cannot detect the difference in the average value or the RMS calculation within a certain period of time as compared with the normal state. This is effective for detecting normality and abnormality when a high-frequency level change is included.

4) Feature amount calculation function for obtaining frequency band peak value The feature value calculation function for obtaining frequency band peak value is set from the data array converted into the frequency axis signal in the preprocessing of the frequency conversion unit 30-5 by FFT. The calculation for obtaining the peak value of the data corresponding to the frequency is executed.

That is, in the feature value calculation function for obtaining the frequency band peak value, assuming that the corresponding data is P1, P2,... Pn, s pieces of data are extracted from the data in descending order, and the average is obtained. This is calculated as the peak value. Here, s is a variable.

The comprehensive determination unit 50 comprehensively determines a plurality of feature amounts extracted for each data block by the feature amount calculation processing units 40-1 to 50-5 to determine whether the product to be inspected is normal or abnormal. .

In the comprehensive judgment unit 50, the plurality of feature amounts extracted for each of the data blocks are averaged or calculated in a plurality of N data blocks corresponding to the entire inspection target time, for example, 10 or more data blocks. By taking a value, the final feature amount is obtained, and the final feature amount becomes inference information for determining whether the product is normal or abnormal.

Here, the reason why the plurality of feature values take the maximum value in the plurality of data blocks corresponding to the entire inspection target time is that the abnormal state does not occur constantly but occurs in a cycle longer than the data block unit equivalent time. It is very effective in the case. Further, when an abnormal state occurs constantly within the data block unit equivalent time, by averaging over a plurality of data blocks corresponding to the entire inspection target time, for example, processing with poor repetition accuracy such as FFT processing is performed. The variation in the information obtained from the information can also be reduced to 1 / N, whereby the reliability of the information is greatly increased, so that an excellent inspection can be performed.

The comprehensive determination unit 50 compares the determination knowledge created by combining all or some of the plurality of feature amounts obtained by the feature amount calculation processing units 40-1 to 50-5 with a predetermined determination knowledge set in advance. Thereby, it is possible to obtain a comprehensive and stable determination result of normality and abnormality of the product to be inspected.

FIG. 2 is a block diagram showing a specific embodiment of the abnormality determination system shown in FIG.

In FIG. 2, the abnormality determination system includes an acceleration sensor 101 provided on the inspection object 100, and determines whether the inspection object 100 is normal or abnormal based on a measurement signal of the acceleration sensor 101.

A measurement signal of the test object 100 output from the acceleration sensor 101 disposed on the test object 100 is first amplified by the amplifier 200, and then digitally converted by the A / D conversion board (analog / digital conversion board) 300. The data is converted into measurement data and input to a personal computer (personal computer) 400.

The personal computer 400 determines whether the inspection object 100 is normal or abnormal by performing the processing described below in detail on the digital measurement data.

FIG. 3 is a block diagram showing an outline of processing in the personal computer (personal computer) 400 shown in FIG.

The measurement signal of the test object 100 output from the acceleration sensor 101 disposed on the test object 100 is converted into digital measurement data by the analog / digital conversion unit (A / D conversion unit) 310 of the A / D conversion board 300. It is converted and input to the personal computer 400.

The personal computer 400 is built in the personal computer 400 based on the memory 410 for storing digital measurement data (hereinafter simply referred to as measurement data) from the A / D conversion board 300 and the measurement data (original data) stored in the memory 410. The CPU (Central Processing Unit) has an internal processing unit 420 that executes predetermined processing by the CPU. The internal processing unit 420 by the CPU includes a pre-processing unit 421, a feature value calculation unit 422, and a FUZZY determination unit. 423.

Here, the processing of the pre-processing unit 421 and the feature value calculation unit 422 in the internal processing unit 420 by the CPU is performed based on the original data stored in the memory 410, as shown in FIG. 440, the intermittent component amplification processing unit 450, the low frequency amplification processing unit 460, and the high frequency amplification processing unit 470 are performed in parallel to perform the processing.

Hereinafter, the details of the processing of the FFT processing unit 430, the frequency processing unit 440, the intermittent component amplification processing unit 450, the low frequency amplification processing unit 460, and the high frequency amplification processing unit 470 will be described.

FIG. 5 is a block diagram showing details of the processing of the FFT processing unit 430.

In FIG. 5, the memory 410 is divided into fixed data blocks of, for example, 1024 points in units of conversion data of the analog / digital conversion unit (A / D conversion unit) 310 of the A / D conversion board 300, and the original data is stored therein. Is stored.

Original data composed of a time axis waveform stored in the memory 410 is first converted into frequency axis data by FFT (Fast Fourier Transform) in an FFT operation processing unit 431, and a feature amount 1 is extracted based on the frequency axis data. A specific frequency band a component is extracted by a unit 432-1, a specific frequency band b component is extracted by a feature amount 2 extraction unit 432-2, and a specific frequency band c component is extracted by a feature amount 3 extraction unit 432-3. .

Here, the specific frequency band a component is extracted by the characteristic amount 1 extraction unit 432-1, the specific frequency band b component is extracted by the characteristic amount 2 extraction unit 432-2, and the specific frequency band is extracted by the characteristic amount 3 extraction unit 432-3. The extraction of the c component is performed by extracting the peak value of each data block, and averaging the extracted peak values with a plurality of target data blocks. Variation can be suppressed.

特 徴 As for the feature value extracted by the process of the FFT processing unit 430, different feature values can be extracted by using a plurality of settings with different frequency bands.

FIG. 6 is a block diagram showing details of the processing of the frequency processing unit 440.

6, as in the case of FIG. 5, the memory 410 stores a fixed data of, for example, 1024 points in units of conversion data of the analog / digital conversion unit (A / D conversion unit) 310 of the A / D conversion board 300. The original data is stored in data blocks.

{Circle around (1)} The original data composed of the time axis waveform stored in the memory 410 is first subjected to pre-processing by the band-pass filter processing unit 441 to remove its noise component and background noise signal. Then, a feature value is extracted from the pre-processed data from which the noise component and the background noise signal have been removed.

The feature amount is extracted by extracting the feature amount 1 extractor 442-1, the feature amount 2 extractor 442-2, the feature amount 3 extractor 442-3, the feature amount 4 extractor 442-4, and the feature amount 5 extractor 442-. 5 is performed.

Here, the feature amount extraction by the feature amount 1 extraction unit 442-1 is to extract a feature amount from the pre-processed data by RMS (square root) operation, and the feature amount extraction by the feature amount 2 extraction unit 442-2. The quantity extraction is based on a feature value calculation function for obtaining a peak value, and the peak value is extracted from the data after the preprocessing in units of data blocks.

The feature value extraction by the feature value 3 extraction unit 442-3 is based on the above-described feature value calculation function for calculating the number of extreme values, and determines the extreme value of the waveform of the data array of the pre-processed data. The value obtained by counting the number of extreme values in the data array is the number of extreme values in the data array.

The feature quantity extraction by the feature quantity 4 extraction unit 442-4 is based on the above-described feature quantity calculation function for calculating the extreme value difference. For the value, the array of the absolute value of the difference between the extreme values before and after is obtained, and an array of the absolute value of the difference between the extreme values before and after is calculated. The value obtained by averaging is defined as an extreme value difference.

The feature value extraction by the feature value 5 extraction unit 442-5 is to calculate the average of the inclination of the data after the preprocessing.

Note that the feature extraction here can employ a method using a general vibration analysis index or a statistical function in addition to the above-described feature amount extraction method.

FIG. 7 is a block diagram showing details of the intermittent component amplification processing unit 450.

In FIG. 7, in the intermittent component amplification processing unit 450, first, the intermittent component amplification preprocessing unit 451 1) performs the Hilbert transform of the original data 2) converts the real data of the original data and the post-Hilbert transform data A complex process is performed as an imaginary part. 3) A smoothing process such as a moving average for calculating a square root array of the complex sequence is performed. 4) A process through a low-pass filter is performed.

According to the processing of the intermittent component amplification preprocessing unit 451, as shown in FIG. 8, a signal is obtained in which only the signal component having a strong gradient in the time axis waveform becomes a large signal, and the other signal becomes small. 8A shows the original waveform of the original data which is the time axis waveform stored in the memory 410, and FIG. 8B shows the original waveform after the pre-processing by the intermittent component amplification pre-processing unit 451. The waveform is shown.

Then, from the waveform pre-processed by the process of the intermittent component amplification pre-processing unit 451, that is, the waveform shown in FIG. The extraction of the extreme value difference by the 2 extractor 457-2 and the difference between the maximum value and the minimum value between the data blocks by the feature amount 3 extractor 457-3, that is, the extraction of the maximum-minimum between the data blocks are performed.

FIG. 9 is a block diagram showing details of the intermittent component amplification pre-processing unit 451.

As shown in FIG. 9, the intermittent component amplification pre-processing unit 451 includes a Hilbert transform unit 452 that performs Hilbert transform of original data that is a time-axis waveform stored in the memory 410, and an original data that is a time-axis waveform stored in the memory 410. A complex number calculation unit 453 that forms a complex array using the data after the Hilbert transform by the Hilbert transform unit 452 as an imaginary part and a square root calculation unit that calculates an array of square roots of the complex array generated by the complex number calculation unit 453 454, a smoothing unit 455 for performing smoothing processing such as moving average on the array of square roots calculated by the square root calculating unit 454, and a low-pass filter 456 for removing noise components and the like from the waveform smoothed by the smoothing unit 455. It is composed.

製品 Generally, a product having a rotary drive unit such as a motor, that is, a representative state abnormality of the inspection object 100 has a bearing failure. A waveform including an intermittently generated high-frequency component typified by a bearing failure or the like (hereinafter referred to as an intermittent harmonic) usually has a higher amplitude than a component of a fundamental wave. In this case, the effective state and the maximum value of the waveform can be determined to determine the normal state.

However, if the intermittent harmonic has little difference compared to the amplitude component of the fundamental wave in a normal state or has a long generation cycle, it is very difficult to detect such an abnormal state.

Even if the original data is transformed to the frequency axis by FFT (Fast Fourier Transform), since the frequency of occurrence is small, the transformation result is calculated as the average frequency component of the target time, so that no effective difference appears.

Therefore, in the intermittent component amplification processing preprocessing unit 451, first, envelope processing, that is,
1) Hilbert transform the original data (Hilbert transform unit 452)
2) Create a complex array using the original data as the real part and the Hilbert-transformed data as the imaginary part (complex number calculator 453).
3) Calculate the square root array of the complex number array (square root calculator 454)
Execute the processing.

That is, first, a front envelope Ψ (t) is obtained. The front envelope Ψ (t) is a complex bandpass signal obtained by adding the Hilbert to the original data as an imaginary part, and is obtained by the following equation.

If the original data is χ (t) and the Hilbert transform data of χ (t) is ζ (t), the front envelope Ψ (t) is
Ψ (t) = χ (t) + ζ (t)
It becomes.

Further, since the envelope w (t) is defined as the amplitude of the previous envelope Ψ (t), the envelope w (t) is the sum of the square of χ (t) and the square of ζ (t). By taking the square root of, the following equation can be obtained.
w (t) = (χ2 (t) + ζ2 (t)) 1/2

処理 By this process, a waveform is obtained as if the wave of the original data was turned around at the center. This folding makes it easy to observe the generation cycle of the intermittent harmonic which is a problem.

FIG. 10 shows the processing result of the envelope processing as a waveform. 10A shows a waveform before the envelope processing, and FIG. 10B shows a waveform after the envelope processing.

Next, a smoothing process is performed. This smoothing process
1) Smoothing processing such as a moving average for calculating an array of square roots of a complex number array (smoothing processing unit 455)
2) Process of passing through a low-pass filter (low-pass filter 456)
Is performed by

The above-described smoothing process is preferably performed by a combination of both the process by the smoothing processing unit 455 and the process by the low-pass filter 456, but an approximate result can be obtained by either one of the processes.

That is, small waves are cut by the moving average processing, that is, the processing by the smoothing processing unit 455.

FIG. 11 shows a waveform of a moving average processing result by the smoothing processing unit 455 when the moving average coefficient of the smoothing processing unit 455 is “9”. In FIG. 11, (a) shows the waveform before the moving average processing, and (b) shows the waveform after the moving average processing.

{Circle around (2)} FIG. 12 shows waveforms of processing results when a combination of both the moving average processing by the smoothing processing unit 455 and the processing by the low-pass filter 456 is performed. In FIG. 12, (a) shows a waveform before processing by the moving average processing and the low-pass filter processing, and (b) shows a waveform after processing by the moving average processing and the low-pass filter processing.

に よ り By the above processing, the processed waveform becomes a high-amplitude wave at the abnormal part, and as a result, it is possible to easily distinguish between normal and abnormal.

FIG. 13 shows a comparison of waveforms before and after processing when the product, that is, the inspection object 100 is normal, by the intermittent component amplification preprocessing unit 451. In FIG. 13, (a) shows a waveform before processing, and (b) shows a waveform after processing.

FIG. 14 shows a comparison of waveforms before and after processing when the product, that is, the inspection object 100 is abnormal, by the intermittent component amplification preprocessing unit 451. In FIG. 14, (a) shows the waveform before the processing, and (b) shows the waveform after the processing.

FIG. 15 is a block diagram showing details of the processing of the low-frequency amplification processing unit 460.

In FIG. 15, similarly to the case of FIG. 5, in the memory 410, for example, 1024 points are fixed in units of conversion data of the analog / digital conversion unit (A / D conversion unit) 310 of the A / D conversion board 300. The original data is stored in data blocks.

The original data composed of the time axis waveform stored in the memory 410 is first pre-processed by the low frequency processing unit 461. This preprocessing is
1) The original data is passed through a low-pass filter. 2) The original data passed through the low-pass filter is integrated in a short time.

{Circle around (4)} A feature value is extracted from the data preprocessed by the low-frequency processing unit 461.

特 徴 The extraction of the feature is performed by the feature 1 extractor 462-1, the feature 2 extractor 462-2, and the feature 3 extractor 462-3.

Here, the feature amount extraction by the feature amount 1 extraction unit 462-1 is based on a feature amount calculation function for obtaining a peak value, and the peak value is extracted from the data after the preprocessing in units of data blocks.

The feature value extraction by the feature value 2 extraction unit 462-2 is based on the feature value calculation function for finding the number of extreme values described above, and determines the extreme value of the waveform of the data array of the pre-processed data. The value obtained by counting the number of extreme values in the data array is the number of extreme values in the data array.

The feature value extraction by the feature value 3 extraction unit 462-4 is performed by calculating the difference between the maximum value and the minimum value between the data blocks, that is, the maximum-minimum value between the data blocks.

FIG. 16 is a block diagram showing details of the high-frequency amplification processing 470.

In FIG. 16, similarly to the case of FIG. 5, the memory 410 has a fixed data unit of, for example, 1024 points in units of conversion data of the analog / digital conversion unit (A / D conversion unit) 310 of the A / D conversion board 300. The original data is stored in data blocks.

The original data composed of the time axis waveform stored in the memory 410 is first pre-processed by the high frequency processing unit 471. This preprocessing is
3) The original data is passed through a high-pass filter. 4) The original data passed through the high-pass filter is differentiated in a short time.

{Circle around (4)} The feature amount is extracted from the data pre-processed by the high-frequency processing unit 471.

特 徴 The extraction of the feature amount is performed by the feature amount 1 extraction unit 472. That is, the feature value extraction by the feature value 1 extraction unit 472 is based on the above-described feature value calculation function for calculating the extreme value difference. For the extreme values in the data array obtained by the feature value calculation function for calculating the number of extreme values, Calculate the array of the absolute value of the difference between the extreme values before and after, and extract a predetermined number of data in descending order from the data in the array of the absolute value of the difference between the extreme values before and after, and calculate the average thereof. The resulting value is defined as the extreme value difference.

FIG. 17 shows an example of a waveform obtained by amplifying the respective characteristics of the original data in which the low-frequency component has an abnormality constantly and the high-frequency component has an abnormality intermittently by the low-frequency amplification processing and the high-frequency amplification processing. It is shown. In FIG. 17, (a) shows the waveform of the original data, (b) shows the waveform after the low-frequency amplification processing, and (c) shows the waveform after the high-frequency amplification processing.

As described above, by detecting a characteristic amount by executing a plurality of processes in parallel on the original data, it is possible to detect any abnormal state.

By the way, when the inspection object 100 is, for example, a motor, the types of abnormalities include a plurality of faults such as a bearing failure, a thread between a brush and a commutator, and an imbalance due to mechanical imbalance of a shaft or a rotor. Factors exist.

For example, if the types of defective products of a certain motor are three types, that is, a bearing defect, a thread defect, and a chatter defect, when these are collectively expressed, the entire set of rejected products to be inspected is X, and the bearing defect is X. If the set is A, the set of thread defects is B, and the set of chatter defects is C,
X = AUBUC (U indicates logical sum)
It becomes.

Therefore, it is not sufficient for the determination device for determining a rejected product to detect any one of the set A of the bearing failure, the set B of the thread failure, and the set C of the chatter failure. It must be able to detect and judge a passing level.

Now, in order to show how the inspector has determined one defect, a description will be given below with an example of a "noisy defect".

FIG. 18 shows the relationship between the noise level (decibel) measured by the sound level meter and the feeling that the inspector feels noisy by the crisp set and the fuzzy set. As will be apparent from FIG. 18, all judgments by the inspector are more appropriately represented by fuzzy sets than by crisp sets. In addition, all judgments made by the inspector vary depending on the physical condition of the same inspector if there are individual differences.

For this reason, in the case of the judgment by the inspector, there is no clear boundary, but the inspector has to judge the intermediate level as either pass or reject. ing. That part is the cause of the variation in passing and rejecting. As a result, it is extremely difficult to completely match the judgment in the system using the sound level meter with the judgment by the inspector.

FIG. 19 shows a distribution graph of a noise value obtained by a method specified in a method for measuring a noise level specified in JIS (Japanese Industrial Standards) as a feature amount of a good product and a defective product having a loud noise as a feature amount.

In FIG. 19, there is an intersecting portion between the peaks of the distribution of the non-defective product and the non-defective product. , And a defective product that falls below is overlooked, and as a result, the reliability of the determination device cannot be obtained.

Of course, when a certain feature amount and the judgment result of the inspector are represented by the distribution by the same means, the judgment can be made by the conventional binary logic judgment in which a judgment reference value is set between the mountains. Will be possible.

Therefore, in this embodiment, first, a certain number or more of samples of a clear non-defective product and a clear defective product are collected, and the distribution is examined as shown in FIG.

At this time, the distribution peaks of good and defective products must be separated, and if they intersect, the feature amount on the horizontal axis is not appropriate or the sample is wrong.

Next, two determination reference values SA and SB are obtained. The two determination reference values SA and SB can be obtained by the following equation, with μ1 as the average value of non-defective products, σ1 as the standard deviation, μ2 as the average value of defective products, and σ2 as the standard deviation.
SA = μ1 + 3σ1
SB = μ2−3σ2

At this time, SA <SB must be satisfied. If this is not the case, it means that the sample is not an effective feature amount for detecting the defect or the sample is strange.

Next, collect good samples and defective samples that are noisy by the sensory test of inspectors. Here, the samples of good quality and the noisy ones include those of intermediate grade. This is similarly distributed, and two determination reference values SA and SB are obtained for the sample.

FIG. 20 shows the result of obtaining the two judgment reference values SA and SB in the case of the sensory test by the inspector. At this time, if SA> SB holds, SA is set to SH and SB is set to SL. If SA ≦ SB holds, SA is set to SL and SB is set to SH.

Here, since there are a plurality of types of product defects, the above-mentioned effective characteristic amounts are determined for all the types of product defects, and the SL and SH are determined for each characteristic amount by a similar method. Here, two or more effective feature quantities may exist for one defect type.

Now, returning to FIG. 3, in the abnormality determination system according to the present embodiment, the FUZZY determination unit 423 includes a FUZZY inference unit, It is configured.

Here, the comprehensive determination unit determines from the inference result by the FUZZY inference unit that the abnormality of the inspection object 100 is “OK”, “GLAY”, or “NG”.

As information to be input to the FUZZY inference unit, the effective feature determined as described above is a variable of the antecedent part.

FIG. 21 shows a membership function of each feature amount for fuzzy inference by the FUZZY inference unit in the FUZZY determination unit 423.

In FIG. 21, (a) shows the membership function when SA> SB. Here, the label of this membership function is composed of three labels of SML, MDL, and LGL, and the horizontal axis coordinate of the trapezoid is, as shown in FIG. And the intersection of the MDL and LRG is set to be the SH determined by the above-described method.

に お い て In FIG. 21, (b) shows the membership function when SA ≦ SB. In this case, as shown in FIG. 21B, two membership functions of SML and LRG are set.

FIG. 22 shows a consequent part for fuzzy inference by the FUZZY inference unit in the FUZZY determination unit 423. Here, the fuzzy inference by the FUZZY inference unit is characterized in that it is performed with respect to two conclusions of “bad” and “bad”.

Further, as shown in FIG. 22A, the consequent variable of “bad” has two singletons “OK” and “NG”, and the consequent variable of “bad” is that shown in FIG. As shown in (b), it has two singletons, “OK” and “GRAY”.

FIG. 23 shows an example of a fuzzy determination rule for performing fuzzy inference in the FUZZY inference unit in the FUZZY determination unit 423. Here, the number of features for fuzzy inference may be as many as necessary, but in the fuzzy determination rule shown in FIG. 23, five types of features are used as antecedents for fuzzy inference. The case where it is used as a unit is shown.

In FIG. 1 indicates a rule that is “OK” in two conclusions. When there are a plurality of defective types A, B, and C for the entire set, the set of non-defective products is
Ac * Bc * Cc (* indicates logical product)
(Here, Ac, Bc, and Cc are complements of A, B, and C, respectively.)
Therefore, this rule No. 1 is a rule that a product is clearly non-defective only when the condition of all effective feature values = SML is satisfied by an AND condition.

Rule No. 2 to Rule No. No. 5 is the rule for conclusion 2, that is, "funny". That is, the rule No. 2 to Rule No. According to 5, if any of the feature amounts is MDL, a “GRAY” degree of “funny” is generated.

Rule No. 7 to Rule No. 7. No. 11 is the rule for conclusion 1, that is, "bad". That is, the rule No. 7 to Rule No. 7. According to 11, if any of the feature amounts is LRG, an “NG” degree of “bad” occurs.

Rule No. 12, Rule No. Reference numeral 13 denotes a rule that generates a “bad” degree when a plurality of feature values have an MDL grade. This is the result of examining the data obtained from the sample, and if a corresponding rule is found, it is added in that combination.Especially, even if a new label is created for membership determination of each feature quantity for combination determination Good.

In addition, in addition to the above, when there is a feature amount candidate in which a combination produces a “funny” degree, a rule that generates a “gray” degree of “funny” with a label that can be identified as the AND condition is added. Is also good.

FIG. 24 shows that the degrees of conformity α and β are obtained from each rule. When a mini-max operation is performed on the fitness of each feature, the fitness of two singletons of each conclusion as a composite of each output is obtained.

{For example, the rule No. in FIG. In step 1, the SML suitability α1, α2, α3, α4, α5 and the LRG suitability β1, β2, β3, β4, β5 are obtained for all feature amounts. As the degree of conformity of all the condition parts, the minimum value obtained by λ = min (α1, α2, α3, α4, α5) is obtained by mini-calculation. I do. Similarly, the grade μ of “GRAY” μ = min (β1, β2, β3, β4, β5) is also obtained from β.

適合 Similarly, for the “bad” of the conclusion 1, the fitness α and β are obtained.

Next, as a Max operation, for each label of conclusion 1 and conclusion 2, the maximum value of each rule that outputs the label is obtained.

Next, one output value (grade; y) is obtained from the composite fuzzy output for each conclusion by the centroid method.

Conclusion 2 "Fun"
y = (0 × λ + 1 × μ) / (λ + μ)
Similarly, the grade z is determined for “bad” in the conclusion 1.

グ レ ー ド Thus, in the FUZZY inference unit of the FUZZY determination unit 423, a grade indicating the “bad” degree and the “bad” degree is selected.

In the comprehensive judgment unit of the FUZZY judgment unit 423, a threshold value m for "bad" and a threshold value n for "bad" are set. Each grade is compared with the threshold value, and classified according to the following conditions. You.
if y <man and z <n then "OK"
if y ≧ m and z <n then “GRAY”
if z ≧ n then “NG”
This result is the final determination result and is output to the outside.

According to the above-mentioned method, as a test result, a reliable “OK”, a reliable “NG” and a non-reliable one can be determined. Further, by analyzing data with the test object 100 determined as “GRAY”, It is possible to narrow the width of “GRAY”.

In the present embodiment, the acceleration sensor 101 attached to the inspection object 100 is used as the measuring means. However, non-contact measurement using a sound wave or a laser beam or measurement of the driving current at hand is performed. There is also a method of measuring.

FIG. 1 is a functional block diagram showing an embodiment of an abnormality determination method in an abnormality determination system configured by applying the abnormality determination method and device according to the present invention. FIG. 2 is a block diagram showing a specific embodiment of the abnormality determination system shown in FIG. 1. FIG. 3 is an exemplary block diagram showing an outline of processing in the personal computer (personal computer) shown in FIG. 2. FIG. 4 is an exemplary block diagram for explaining processing of a preprocessing unit and a feature amount calculation unit in an internal processing unit by the CPU shown in FIG. FIG. 5 is a block diagram showing details of the processing of the FFT processing unit shown in FIG. 4. FIG. 5 is a block diagram showing details of the processing of the frequency processing unit shown in FIG. 4. FIG. 5 is a block diagram illustrating details of an intermittent component amplification processing unit illustrated in FIG. 4. FIG. 5 is a waveform chart for explaining the processing of the intermittent component amplification processing preprocessing section shown in FIG. 4. FIG. 9 is a block diagram showing details of an intermittent component amplification processing preprocessing unit 451 shown in FIG. 8. FIG. 10 is a waveform diagram illustrating a result of the envelope processing performed by the Hilbert transform unit, the complex number calculation unit, and the square root calculation unit illustrated in FIG. 9. FIG. 10 is a waveform chart illustrating a moving average processing result by the smoothing processing unit illustrated in FIG. 9. FIG. 10 is a waveform diagram illustrating a processing result when a combination of both the moving average processing by the smoothing processing unit and the processing by the low-pass filter illustrated in FIG. 9 is performed. FIG. 10 is a waveform diagram showing a comparison between a waveform before processing and a waveform after processing when the product, that is, the inspection target is normal, by the intermittent component amplification processing preprocessing unit shown in FIG. FIG. 10 is a waveform chart showing a comparison between a waveform before processing and a waveform after processing when a product, that is, an inspection target is abnormal, by the intermittent component amplification processing preprocessing unit shown in FIG. FIG. 5 is a block diagram illustrating details of processing of a low-frequency amplification processing unit illustrated in FIG. 4. FIG. 5 is a block diagram showing details of the high-frequency amplification processing shown in FIG. 4. The figure which showed an example of the waveform which each characteristic was amplified by the low-frequency amplification processing and the high-frequency amplification processing from the original data in which abnormality normally occurs in a low-frequency component and intermittently shows an abnormality in a high-frequency component. The figure which expressed the relationship between the noise level (decibel) measured by the sound level meter and the inspector feeling noisy by the crisp set and the fuzzy set. FIG. 5 is a distribution graph showing, as feature values, noise values obtained by a method specified in a method of measuring a noise level specified in JIS (Japanese Industrial Standards) for non-defective products and defective products having noisy sounds. The figure which showed the result of having calculated | required two determination reference values SA and SB in the case of the sensory test of an inspector. FIG. 4 is a diagram showing membership functions of respective feature amounts for fuzzy inference by a FUZZY inference unit in a FUZZY determination unit shown in FIG. 3. FIG. 3 is a diagram showing a consequent part for fuzzy inference by a FUZZY inference unit in the FUZZY determination unit shown in FIG. 2. FIG. 3 is a diagram illustrating an example of a fuzzy determination rule for performing fuzzy inference in a FUZZY inference unit in the FUZZY determination unit illustrated in FIG. 2. FIG. 7 is a view for explaining that the degrees of conformity α and β are obtained from each rule. FIG. 9 is a block diagram showing a processing method by a conventional computer system configured by incorporating an FFT analyzer. FIG. 27 is a view for explaining processing of a metamorphosis unit shown in FIG. 26. The figure which shows an example of the waveform generated from the test object of normal and abnormal. The figure explaining the problem by the frequency analysis using FFT. The figure explaining the conventional method using the filter method.

Explanation of reference numerals

10. Analog / digital conversion unit (A / D conversion unit)
Reference Signs List 20 memory 30-1 preprocessing section 1 (filtering) section 30-2 preprocessing 2 (intermittent component amplification processing) section 30-4 preprocessing 4 (high frequency amplification processing) section 30-5 frequency conversion section by FFT 40-1 Quantity calculation processing unit 40-2 Feature calculation processing unit 40-3 Feature calculation processing unit 40-4 Feature calculation processing unit 40-5 Feature calculation processing unit 100 Inspection object 101 Acceleration sensor 200 Amplifier 300 A / D conversion Board (analog / digital conversion board)
310 analog / digital converter (A / D converter)
400 PC (personal computer)
410 Memory 420 Internal processing unit by CPU 421 Preprocessing unit 422 Feature calculation unit 423 FUZZY determination unit 430 FFT processing unit 431 FFT calculation processing unit 432-1 Feature amount 1 extraction unit 432-2 Feature amount 2 extraction unit 432-2
432-3 Feature 3 Extraction Unit 432-3
440 Frequency processing unit 441 Bandpass filter processing unit 442-1 Feature amount 1 extraction unit 442-2 Feature amount 2 extraction unit 442-3 Feature amount 3 extraction unit 442-4 Feature amount 4 extraction unit 442-5 Feature amount 5 extraction unit 450 Intermittent component amplification processing unit 451 Intermittent component amplification preprocessing unit 452 Hilbert transform unit 453 Complex number calculation unit 454 Square root calculation unit 455 Smoothing processing unit 456 Low-pass filter 457-1 Feature amount 1 extraction unit 457-2 Feature amount 2 extraction unit 457 -3 feature amount 3 extraction unit 460 low frequency amplification processing unit 461 low frequency processing unit 462-1 feature amount 1 extraction unit 462-2 feature amount 2 extraction unit 462-3 feature amount 3 extraction unit 470 high frequency amplification processing unit 471 high frequency processing Unit 472 feature amount 1 extraction unit

Claims (2)

In an abnormality determination device that determines an abnormality of the product based on measurement data measured by a measurement unit,
Having intermittent component analysis means for analyzing the intermittent component of the measurement data,
The above-mentioned intermittent component analysis means,
Intermittent component analysis preprocessing means for preprocessing the intermittent component,
An intermittent component analysis feature amount extracting unit that analyzes the intermittent component preprocessed by the intermittent component analysis preprocessing unit and extracts a feature amount;
An abnormality determination device comprising: The intermittent component analysis pre-processing means,
Hilbert transform means for Hilbert transforming the measurement data,
A real number part of the data before the conversion of the Hilbert conversion means, a complex number array generating means for generating a complex number array with the converted data as an imaginary part,
A root mean square array calculating means for calculating a root mean square array of the complex number array generated by the complex number array generating means,
Smoothing processing means for performing a smoothing process such as moving average of the array of the root mean square calculated by the root mean square array calculating means,
Low-pass filter means for extracting a low-frequency component from the processing result of the smoothing processing means,
The abnormality determination device according to claim 1, further comprising:

Images