JPH044534B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION Technical Field of the Invention The present invention relates to an abnormality inspection device that detects abnormalities in various devices based on mechanical vibration or sound.

Technical Background of the Invention Conventionally, in the manufacturing process of rotating equipment, a method has been developed that automatically performs comprehensive mechanical evaluation of the rotating equipment using vibration analysis. As a representative example, the first
The second method is to detect mechanical vibrations or sounds generated in the object to be inspected, analyze the power spectrum or 1/3 octave frequency of the detected signals, and compare and judge the patterns of normal products and abnormal products. Approximately 1KHz from the above detection signal by high pass filter
A method of extracting the above high frequency components and comparing and determining normal products and abnormal products based on their effective value, peak value, or the ratio of the two. Third, the signal in the high frequency region obtained in the same way is enveloped by an envelope detection circuit. There is a method of extracting the line signal and comparing and determining whether it is a normal product or an abnormal product.

Problems with the Background Art Since the first method above calculates the average power during measurement, it may be caused by ball flaws in the bearing of the object to be measured, abnormal rotation due to intervening dust, abnormal resonance of each component, etc. Detection of pulsed signals that occur only once in a while is difficult. On the other hand, with the second and third methods described above, it is possible to detect abnormal signals that cannot be detected with the first method, but on the other hand, it is possible to detect various abnormal signals that occur in vibration sources other than the object to be inspected during actual inspection. This method has the disadvantage that a normal product may be incorrectly determined to be an abnormal product due to the influence of external disturbances such as sound or mechanical vibration. In addition, it is difficult to detect subtle bearing defects.

Purpose of the Invention The present invention is a highly versatile device capable of accurately and quickly detecting abnormalities in a test object through statistical calculation processing by removing disturbances such as vibrations or sounds from vibration sources other than the test object. SUMMARY OF THE INVENTION To provide an abnormality testing device The abnormality testing device of the present invention has a detection signal obtained by a vibration detector such as a vibration pickup, which has a plurality of frequency bands that pass only frequency components in different frequency bands. The maximum peak value is detected once for each sampling period over an inspection time consisting of multiple sampling periods for each frequency component inputted to the analysis section, and the obtained maximum peak value is converted from analog to digital and then calculated. This is stored in the control unit, and the maximum peak value obtained by removing disturbance data from the stored maximum peak values is compared with a preset value to determine whether there is an abnormality in the object to be inspected. be.

Embodiment of the Invention FIG. 1 shows an electric circuit system diagram of an abnormality testing device according to this embodiment. For example, the output side of a vibration detector 1 such as a vibration pickup is connected via an amplifier 2 to frequency analyzers 301...30n that analyze frequency components of different frequency bands. however,
In this embodiment, n=28 (the same applies below).
These frequency analysis units 301...30n are connected to the amplifier 2
For example, if the center frequency is 20
Bandpass filters 401...40n for obtaining frequency components of 28 channels of frequency bands of 1/3 octave from Hz to 10000Hz and peak value detection circuits connected to the output sides of these bandpass filters 401...40n, respectively. It consists of 501...50n. The above peak value detection circuit 501...50
n is a sample-and-hold circuit 601...60n connected to the output side of the bandpass filter 401...40n, which is also the bandpass filter 40.
Comparators 701...70n each connected to the output side of 1...40n and the output side of the sample-and-hold circuit 601...60n, and the control input side of the sample-and-hold circuit 601...60n whose output side holds the positive half-wave peak value. It also consists of AND circuits 801...80n whose input sides are connected to the output sides of the comparators 701...70n, respectively.
The other input side of the AND circuits 801...80n is connected to the output side of a decoder 9, which will be described later.
Further, the frequency analysis units 301 . . . 30n are connected to the input side of the multiplexer 10. The output side of this multiplexer 10 is connected to the input side of an analog-to-digital converter 11. Analog-to-digital converter 11 is connected to microcomputer 13 by system bus 12.
This microcomputer 13 includes a central processing unit (hereinafter referred to as CPU) 14, an input/output interface 16 connected to this CPU 14 via a system bus 15, and a random access memory (hereinafter referred to as
It is written as RAM. ) 17, a read-only memory (hereinafter referred to as ROM) 18 in which a control program is stored, and a CRT interface 19. Further, as will be described later, the CPU 14 includes a first calculation means 14a that rearranges the maximum peak values obtained one by one for each sampling period within one inspection time in order of magnitude; a second calculation means for extracting a specific number of maximum peak values predetermined in accordance with the disturbance content for each frequency component as an effective maximum peak value among the maximum peak values rearranged in order of magnitude by the calculation means 14a; 14b and this second calculation means 1
The third calculation means 14c performs an abnormality determination by comparing the effective maximum peak value obtained in step 4b with a predetermined set value. A part of the RAM 17 is a non-volatile RAM for storing set values and normal patterns. In addition, although not shown, a microcomputer 13
is provided with an input section for externally storing the control program, setting values, etc. in the RAM 17 and ROM 18. The input/output interface 16 is connected to the analog-to-digital converter 11 via the system bus 12. moreover,
The input/output interface 16 is the system bus 2
0 to a decoder 9 and a multiplexer 10. Further, the input/output interface 16 is connected to the input side of a differentiating circuit 22 via a line 21. The output side of the differentiating circuit 22 is connected via a strobe line 23 to the input sides of the analog-to-digital converter 11 and the monostable multivibrator 24. The output side of the monostable multivibrator 24 is connected to the decoder 9 via a strobe line 25. the decoder 9;
Differentiator circuit 22 and monostable multivibrator 24
constitutes a timing circuit 26. moreover,
Timing circuit 26 and microcomputer 1
3 constitutes an arithmetic control section 27. Also, the above
A CRT (Cathode Ray Tube) 28 is connected to the CRT interface 19 for displaying test results.

Next, the operation of the abnormality inspection device of this embodiment will be described in detail.

First, a calculation program to be described later is stored in the ROM 18, and the test time (3 seconds in this embodiment), sampling period Δt (0.1 seconds in this embodiment), and peak value are input to the RAM 17 via the input section. Number of invalid peak values to exclude N _IP
(two in this embodiment) are stored (step 29 in FIG. 2). Note that the length of the sampling period △t is determined by the disturbance (for example, shock vibration when the conveyor stops).
It is preferable to set the time so that it is approximately equal to the influence time. The length of the inspection time is determined depending on the nature of the abnormality of the object to be detected (in this embodiment, a rotational abnormality of a rotating device). Further, the number of invalid peak values _NIP is determined based on the empirically determined number of disturbances that occur within the above-mentioned inspection time. Then, the vibration detector 1 is brought into contact with a rotating device (not shown), which is an object to be inspected, and the rotating device is started. Mechanical vibration or sound from the above rotating equipment is detected by vibration detector 1.
An electrical signal with a waveform corresponding to the vibration waveform at
Converted to SA. This signal SA is amplified by the amplifier 2, and then passed through each bandpass filter 401...
40n, each bandpass filter 401...
40n, the signals are converted into signals SC1...SCn having frequency components in predetermined different frequency bands. Hereinafter, the signal processing method of the signal SCn mainly in the case of n=28, that is, 28 channels (center frequency is 10000 Hz) will be described in detail by way of example based on FIGS. 1 and 3. The signal SCn has a sinusoidal waveform as shown in Fig. 3 (other signals SC1...SC(n-1) also have different wavelengths, but the waveforms are almost sinusoidal. This signal SCn is a sample The sample and hold circuit 60n outputs a signal SDn (see FIG. 3) indicating the peak value to the comparator 70n.The comparator 70n sequentially inputs The signal SDn is set to the set value, and the signal
When the signal SDn is smaller than the signal SCn, a signal SEn indicating "0" is output to the AND circuit 80n, and when the signal SDn is larger than the signal SCn, a symbol SEn indicating "1" is output to the AND circuit 80n. Signal SCn shown in Figure 3
is monotonically increasing in the section 30, so a signal SEn indicating "0" is output to the AND circuit 80n (see FIG. 3). On the other hand, AND circuit 80n
Since the signal SFn indicating "1" is always outputted to the other input terminal of
SGn (see FIG. 3) is output to a sample and hold circuit 60n. This sample-and-hold circuit 60n samples the peak value when the signal SFn from the AND circuit 80n is "0", and holds the peak value when the signal SFn is "1". Therefore, in section 30, the sample-and-hold circuit 60n is in the peak value sampling state, so the signal SDn also increases as shown in FIG. However, the signal
Since the signal SEn becomes "1" across the inflection point 31 of SCn, it is "1" from the AND circuit 80n.
The signal SGn indicating the sample and hold circuit 6
0n and the peak value at the inflection point 31
Pn ₁ is held, and a signal SDn indicating this peak value Pn ₁ is output to the multiplexer 10. Third
Although not shown in the figure, other frequency analysis units 301...3
Similarly, from 0(n-1), the peak value P ₁₁ ...
Signals SD1...SD(n-1) indicating P _(o-1)1 are simultaneously output to the multiplexer 10. On the other hand, CPU
14 outputs a signal specifying a desired frequency analysis section 30n via the input/output interface 16 to the decoder 9 and multiplexer 10 via the system bus 20. With this signal SH, only the signal SDn indicating the peak value Pn ₁ of the frequency component that has passed through the bandpass filter 40n is outputted to the analog-digital converter 11 as the signal SIn (see FIG. 3). Time △t ₁ (several tens of μs) from the time when the signal SIn is input to the analog-to-digital converter 11
After the elapsed time, a pulse signal SJ is sent from the CPU 14 via the input/output interface 16 for a time △t ₂ (approximately 30
ms or less) to the differentiating circuit 22 (see FIG. 3). The differentiating circuit 22 outputs a signal SK obtained by differentiating the signal SJ to the monostable multivibrator 24 and the analog-to-digital converter 11. The signal SIn at the time when this signal SK is input is converted into a digital value data signal SLn by the analog-digital converter 11, and this data signal
SLn is transmitted to the RAM 17 via the input/output interface 16 CPU 14 and stored as the peak value Pn ₁ at a predetermined address in the RAM 17 (step 32 in FIG. 2). In addition, in FIG. 3, the abnormality inspection is temporarily interrupted at the time Δt ₃ between the leading edge of the part showing "1" of the signal SK and the trailing edge of the part showing "0" of the signal SF1. In order to collect data accurately in real time, settings are made to satisfy Δt ₃ <1/2F _nax (where F _nax is the highest frequency) according to Shannon's sampling theorem. On the other hand, the monostable multivibrator 24 to which the signal SK is input
From there, the signal SM is output to the decoder 9 as a strobe signal using the trailing edge of the signal SK as a trigger.
From the decoder 9 which inputs this signal SM, the signal
A signal SFn indicating "0" for reset is output to the AND circuit 80n designated by SH. As shown in FIG. 3, a signal indicating "0" is output from the AND circuit 80n which inputs the signal SFn indicating "0" to the sample-and-hold circuit 60n, and the peak value is at point 33 of the signal SC in FIG. Then, as described above, the next sampling period Δt (where Δt>Δt ₂
It is set so that Here, n is the number of frequency analysis sections. ) is _held and output to the multiplexer 10. On the other hand, when the peak value Pn ₁ is stored in the RAM 17, the signal SH specifying the frequency analysis section 301 is sent to the decoder 9 and the multiplexer 10, and the signal SJ is sent to the differentiating circuit after {△t-(△t ₂ ×n)} time. 2
Similarly to the frequency analysis section 30n, the signal SI1 indicating the peak value _P11 is converted into a digital value data signal SL1 by the analog-digital converter 11, and this data signal SL1 is
It is transmitted to the RAM 17 and stored at a predetermined address in the RAM 17 as the peak value _P11 . Thereafter, a signal SF1 indicating "0" for resetting is output to the AND circuit 801, and the sample-and-hold circuit 601 returns to the sampling state again. In this way, the maximum peak value P ₂₁ ...P _(o-1)1 held in each of the other frequency analysis sections 302 ... 30 (n-1) within the sampling period Δt is sequentially calculated.
Store it at a predetermined address in RAM 17. Note that the pulse interval of the signal SJ output from the microcomputer 13 is △t ₂ during the period of sampling SL1...SLn corresponding to the signals SL1...SLn in FIG.
and the next sampling period △t from signal SLn
The pulse interval until sampling signal SL1 at is {△t−(△t ₂ ×
n)}. Note that the program to obtain the pulse interval of signal SJ is stored in ROM in advance.
It is stored in 18. Furthermore, the time Δt ₂ is adjusted such that Δt ₂ ×n is equal to the sampling period Δt.
may be set. Each sampling period △t above
The storage of peak values in the RAM 17, which is performed sequentially for each frequency analysis section 301...30n, is repeated m times as shown in FIG.
In the RAM 17, the peak value is stored for each frequency analysis section 301...30n. Therefore, m pieces of data are stored for each frequency analysis section 301...30n. When the number of data stored in the RAM 17 reaches m×n (where m is the number of sampling times and n is the number of frequency analyzers) (step 34 in FIG. 2), the first calculation of the CPU 14 is performed. In the means 14a, each frequency analysis section 301...30n
A computation is performed to rearrange the peak values in ascending order (step 35 in FIG. 2), and the computation results are stored in the RAM 17 as shown in FIG. However, the CPU
In order to exclude the influence of disturbances from vibration sources other than the object to be inspected, such as shock vibration when the conveyor is stopped and shock vibration when the air cylinder is operated, the second calculation means 14b of No. 14 calculates the frequency component for each frequency component. RAM17
Exclude N _IP data (for example, 2 data) from the larger side of the peak values arranged in descending order of the values stored in , and extract the N _IP +1st data as the effective maximum peak value P _EM1 ...P _EMo . Store it in the RAM 17 (step 36 in Figure 2). However, CPU14
The third calculation means 14c calculates, for each frequency component, the setting values P _T1 ...P _To set in advance in the RAM 17 and the effective maximum peak value P corresponding to each of these setting values P _T1 ...P _To . _EM1 ...P _EMo are compared, and if even one of the effective maximum peak values P _EM1 ...P _EMo exceeds the set value, it is determined that there is an abnormality in the object to be inspected (Step 37 in Figure 2). The above setting values can be calculated by setting the peak value in the RAM in the same way as described above for, for example, about N = 100 (pieces) of normal objects to be inspected before the actual inspection.
The CPU 14 calculates the average value i and the standard deviation σ _i of the effective maximum peak value P _EMki for each frequency component of all test objects using the following formula stored in the ROM 18.
Store in 7.

Here, i indicates the frequency component, i=1
...n. Also, k is the number of objects to be inspected and k=...
1N, and P _EMki is the peak value for the frequency component i of the k-th test object. Setting values P _{T1 .} . . P _To are determined for each frequency component by the following formula stored in the ROM 18 based on the average value i and standard deviation σ _i determined by the above formula.

P _Ti =i+Zσ _i (where i=1...n) Here, Z is a variable that can be arbitrarily set according to the inspection environment and inspection criteria. The above setting value P _T1 ...P _To ,
The effective maximum peak values _PEM1 ... _PEMo and the determination results regarding the presence or absence of an abnormality are displayed on the CRT 28 for each frequency component as shown in FIGS. 6 and 7 (Step 38 in FIG. 2). In these figures, the number n of frequency components is 28. In addition, CRT28
A program for displaying is stored in the ROM 18 in advance.

FIG. 6 shows a CRT display of an inspection result when the waveform is caused by a so-called "blink abnormality" when a rotating part of the object to be inspected comes into contact with a fixed part as shown in FIG. In FIG. 6, the shaded part of the bar graph indicates an abnormal part, and when the effective maximum peak value is larger than the set value, it is determined to be "abnormal". The band representing the abnormality is displayed by a rectangle 39 directly above the bar graph representing the abnormality. Similarly, Fig. 7 shows a CRT display of an inspection result when the waveform is due to an abnormality in the bearing, such as a ball flaw, as shown in Fig. 9, and the effective maximum peak value is larger than the set value. Sometimes it is judged as "abnormal".

In the above embodiment, the peak value detection circuits 501...50n are positive half-wave peak value circuits, but as shown in FIG.
Absolute value circuits 4001 . . . 400n may be interposed between 0n and 0n. By doing so, it becomes possible to detect negative half-wave peak values as well, thereby improving detection accuracy. In addition, as shown in FIG. 11, for each bandpass filter 401...40n, a sample-and-hold circuit for detecting a negative half-wave peak value is provided in parallel with a peak value detection circuit 501...50n, which is a positive half-wave peak value circuit. Circuit 601'...60n', comparator 70
1'...70n', and a peak value detection circuit 501'...50n' consisting of an AND circuit 801'...80n' is connected, and the output difference between the two is inputted to a subtraction circuit 4101...410n to obtain a peak-peak value (positive/negative peak value). You may also ask for This improves detection accuracy. In addition, in the above example
The peak values stored in the RAM 17 are sorted in descending order, and then the N _IP + 1st valid maximum peak value is extracted. However, due to a software program change, the peak values are sorted in ascending order, but due to a change in the software program, the peak values are sorted in descending order, but due to a change in the software program, the peak values are sorted in ascending order. The effective maximum peak value may be extracted from among the peak values obtained.
Furthermore, the abnormality inspection device of the present invention is not limited to rotating equipment as an object to be inspected, and can also be used to detect abnormalities in, for example, nuclear reactors, manufacturing equipment, and the like. At this time, it is necessary to appropriately change the sampling period Δt and the number of invalid peak values N _IP depending on the length of the influence time of the disturbance. Furthermore, if the detected content is the point in time when an abnormality occurs such as tool breakage of a cutting tool, the inspection time may be set to 10 seconds, for example, and the presence or absence of an abnormality may be determined every 10 seconds. In addition, in the above embodiment, the set value P _T1 ...P _{To (} normal pattern) is set in advance before the inspection, but each time the number of objects to be inspected accumulates a certain number, the above formula is calculated. You may also update _To . By doing so, corrections due to changes over time during manufacturing can also be automatically performed. Furthermore, although a CRT is used as the display unit in the above embodiment, it may be a printing device or the like.

Effects of the Invention The abnormality inspection device of the present invention detects the maximum peak value for each sampling period within the inspection time consisting of a plurality of sampling periods, and sequentially stores the detected peak values in the arithmetic control section. It is now possible to reliably capture not only abnormal signals that occur continuously, but also pulse-like abnormal signals that occur intermittently, and the calculation control unit can now exclude maximum peak values caused by disturbances. This dramatically improves the accuracy of abnormality determination. Therefore,
Ball flaws on the bearing, abnormal rotation due to dirt,
It is possible to detect all kinds of abnormalities, including resonance abnormalities, tool damage, equipment failures, etc.

[Brief explanation of drawings]

FIG. 1 is an electric circuit system diagram of the abnormality testing device of the present invention, FIGS. 2 and 3 are a flowchart and timing chart for explaining the operation of the abnormality testing device of the present invention, respectively, and FIG. 4 is a diagram of the present invention. Figure 5 is a diagram schematically showing the maximum peak values rearranged in descending order in the RAM of the arithmetic and control unit; Figures 6 and FIG. 7 is a diagram showing the CRT of the abnormality inspection device of the present invention that displays the inspection results, FIGS. 8 and 9 are diagrams showing signal waveforms when there is a hit abnormality and a bearing abnormality, respectively, and FIGS. 10 and 11. The figure is an electrical circuit diagram of a frequency analysis section in another embodiment of the present invention. 1: Vibration detector, 401...40n: Band pass filter, 501...50n: Peak detection circuit, 1
4a: first calculation means, 14b: second calculation means, 14c: third calculation means.

Claims (1)

[Scope of Claims] 1. An abnormality inspection device characterized by having the following configuration. (a) A vibration detector that detects vibrations generated from the object being inspected and converts them into electrical signals. (b) A plurality of filters that input the electrical signal output from the vibration detector and decompose the electrical signal into frequency components of a plurality of frequency bands. (c) A plurality of peak values which are connected to the output side of the above filter separately, and which input electrical signals of each of the above frequency components and detect the maximum peak value within each of the above sampling periods within an inspection time consisting of a plurality of sampling periods. detection circuit. (d) Input the electric signals indicating the maximum peak values outputted from the plurality of peak value detection circuits separately, and calculate the maximum peak value detected one by one for each sampling period within the above inspection time to each of the above. A first calculation means for rearranging each frequency component in order of magnitude. (E) Regarding the maximum peak values rearranged in the order of magnitude in the first calculation means, the specific maximum peak value predetermined in accordance with the disturbance content is used as the effective maximum peak value for abnormality detection. A second calculation means for extracting each frequency component. (F) Comparing the effective maximum peak value for each frequency component obtained by the second calculation means with a predetermined set value and determining abnormality of the object to be inspected based on the comparison result. 3 calculation means.