JPS5863831A - Monitoring method for periodically moving body - Google Patents

Abstract

PURPOSE:To detect abnormality by separating time serial data to periodically moving components and non-periodically moving components by Fourier transform, and further calculating effective values for the former and normalized values for the latter by inverse Fourier transform, and making overall evaluation. CONSTITUTION:An oscillation detector 2 is mounted to an oscillation generating body 1 such as a bearing, a gear or the like, and the output of the device 2 is stored via a sampling device 3 into a storage device 4. An arithmetic device 5 separates time serial data to periodically moving components and non-periodically moving components by Fourier transform, and further separates the same to periodically moving components and non-periodically moving components in time regions by inverse Fourier transform. Effective values are calculated from the former, and the K value normalized by dividing the higher order moments of propability density functions by variance. Both values are inputted to a comparator 7 together with the data relating to the normal levels and life levels stored in a reference data storage device 6. and when the body 1 is decided overall to be abnormal, an alarm is emitted by an alarm device 8.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to objects that perform periodic motion such as bearings, gears, etc., that is, periodic moving bodies, and a method for monitoring devices equipped with these objects.

In general, if abnormalities such as scratches on parts, eccentricity of the rotating shaft, or poor lubrication occur in periodic moving bodies such as bearings and gears, and equipment equipped with these, if left untreated, the bearings, gears, and other parts may become damaged. This causes not only damage but also failure and destruction of the entire equipment equipped with them. Therefore, accurately understanding such abnormalities is an extremely important issue in the maintenance and management of equipment including bearings, gears, and the like. Conventionally, in order to accurately grasp such anomalies, a medium sensor is installed in a periodic moving body, and from the time series data obtained from the output signal, its effective value, that is, RMS (R
The diagnosis and monitoring of objects to be monitored, including periodic moving objects such as bearings and gears, as well as equipment equipped with these objects, has been performed by determining the value of oot Mean 5 square) and analyzing it.

However, x([): Time series data (1=1,2...
・N)N: When evaluating based on the number of data and this RMS value, it is suitable for understanding the time series data as a whole, but this RMS value depends on the size of the monitored object, the size of the load, the rotation speed, etc. Since it is not possible to establish new standards, it is cumbersome to accumulate individual normal data, and the sensitivity is low when determining subtle abnormalities. There were also some difficulties.

The present invention has been made in view of the above circumstances, and when monitoring a periodic moving body, in addition to the effective value such as the RMS value described above, the probability density function (Probability Density) of time series data obtained from the monitored body is used. F
(hereinafter abbreviated as P, D, F) and normalize the disturbance of its distribution, that is, P, D, F
, which is the normalized value of the higher-order moments (select even-numbered moments that indicate the spread of the distribution) divided by its variance (
We aim to provide a highly sensitive monitoring method for periodic moving bodies that can comprehensively grasp time-series data and determine abnormalities using universal criteria by incorporating the K value).
target

Therefore, in general, if the thread is normal, its P, D, F
.

is considered to have a normal distribution, and in this case the value is 3.0. However, since periodic motion components that are the object of monitoring in the present invention are always mixed with periodic motion components, P, D, and F do not necessarily have a normal distribution when the system is normal, and K
The value is lowered (K=1.5 in the case of only a sine wave as an example of a periodic motion signal), and the evaluation of the K value becomes inaccurate. Consider separating the obtained signal into periodic motion components and non-periodic motion components, and evaluating each component in an appropriate manner.

In other words, the latter has a normal distribution when the system is normal, but when the system deteriorates, its P, D, and F are thought to be disturbed, so if you calculate the 1st shift, you can make an accurate judgment based on the above principle. It becomes possible. Regarding the former, since the kinetic energy increases as the system deteriorates, it may be evaluated using, for example, an effective value.

The details are detailed below.

The method for monitoring one periodic motion body according to the present invention is to perform Fourier transformation on time series data obtained by sampling the vibrations of a periodic motion body at a constant period, thereby converting the periodic motion component and the non-periodic motion component. By performing inverse Fourier transform on at least one of these, the time series data in ntl is separated into a periodic motion component and a non-periodic motion component in the time @ domain, and for the periodic motion component, its effective value is Calculate,
In addition, for non-periodic motion components, the higher-order moments of P, D, and F formed by them are divided by their variance to calculate a normalized value, and the two calculated values are combined for evaluation. Therefore, it is characterized by detecting abnormalities in periodic moving bodies.

First, the principle of the method of the present invention will be explained. In the method of the present invention, in order to understand the time series data as a whole, the RMS value shown in equation (1) is used, and in order to determine abnormality using a universal criterion, P and D of the time series data are used. , F
A value obtained by dividing the higher-order moment of , by its variance, and normalizing it, for example, the value shown in equation (2) below, is used.

...(2) However, P(x): P, D, F of x(t)
.

E: Average operation/'4'' Fourth-order moment σ2: The variance of x(t), that is, the time series data, is separated into a periodic motion component and an aperiodic motion component by performing Fourier transformation, and at least one of these is calculated. By applying inverse Fourier transform, the time series data is separated into periodic motion components and aperiodic motion components in the time domain, and RM is applied to the periodic motion components.
The S value is calculated, and the value is calculated for the non-periodic motion component, and both calculated values are combined for a comprehensive evaluation.

The present invention will be explained in detail below based on the drawings. FIG. 1 is a schematic block diagram of a device used to implement the method, and a vibration generator 1 houses a vibration detection device 2 that detects the vibration and converts it into an electric signal. The output of this vibration detection device 2 is input to a sampling device 3, where it is sampled at a constant period, converted from analog data to digital data, and stored in a storage device 4. The time-series data x(t) stored in the five arithmetic units and the storage device 1'i4 is stored and the arithmetic processing described below is performed.

FIG. 2 is an explanatory diagram showing this processing procedure.

First, in FIG. 2(a), time series data x(t) is subjected to Fourier transformation as shown in equation (3) below.

Flash −j27Lft X(f) = / x(t)e dt −
(3) fl', t: Ordinal number representing the sampling point f: Frequency j: Imaginary unit This means that the time series data This is an operation to separate the two. The data X(f) thus obtained
For example, if the motion period of the periodic moving body of the monitored object is 30Hz
In this case, the frequency f forms a peak at 30 Hz. Therefore, if the frequency f is 30Hz and its harmonics (60Hz
, 90Hz...) as shown in Figure 2 (b), the periodic luck! IIIJjt minute A
, (f) is shown in equation (4) below! /11 <frequency ti
It can be obtained by separation in the l'f region.

A(f) -X(f)・P(f)...
(4) IL1, P(f): 1 if f is IP 1 frequency and its harmonic frequency from the monitored object, 0 otherwise. ) As shown in the formula, Figure 2C
’), the periodic luck in the time domain is 10.
J1 把minute a(t) is calculated.

a(t) −/A(f) e d f
・-(5) And non-periodic luck in the time 1ili range! lνJ
The component b(t) is derived from the periodic motion component a(t) from the time series data x(t) as shown in equation (6) below (see Figure 2).
Calculated by subtracting.

b(t) −x(t) = a(t) ・(6)
At this time and in the same region, the aperiodic motion component b(t) is obtained by Fourier transforming the time X column data x(t). Growth
After separating B(f) as shown in equation (7) below, B(f) -X(f) -A(f)
...(7) The non-periodic luck: It may be calculated by inverse Fourier transforming the IiI component B(f), and the component a(t) of the periodic luck 1 in the time domain can also be calculated by Instead of using the method shown in Figure 2, it is of course possible to obtain the aperiodic motion component b(t) in the time domain using the method described above and then subtract it from the time series data x(t). be.

Therefore, for Zhou J (Jl Unpongmu component a (t)), the following (
8) Take the RMS value in Figure 2 (e) as shown in formula 2, and for the non-periodic motion component b(t), take the value in Figure 2 (f) as shown in formula (9) below. Take.

However, i-nee! -da a(t) (average value of a(t))
N t-+ / (b(t))'P(x)dx K-□...(9) [/(b(t))2P(x)dx,)2However, P
(x): P, D, F of b(t).

The data regarding the RMS value and the value thus obtained are as follows:
The information is input to the comparison device 7 together with the data regarding the normal level and the lifespan level stored in the reference data storage device 6, and as a result of comparing both data there, if it is comprehensively determined that the eyelid is abnormal, the prefectural notification device d8 An alarm signal has been sent to the area, and an alarm is being issued. In order to make a comprehensive judgment, it is possible to evaluate the data related to the RMS value and the value, for example, at 51 points, and then use the average to make the overall evaluation.
Since the seven rolls due to ll1i11a's decision are large, it is only necessary to create a simple judgment logic as appropriate.

When monitoring a periodic moving body using a certain device 1q, for example, when monitoring a reducer that decelerates and transmits assistance using a reversing gear supported on a bearing, the time series data obtained is a periodic motion component. is mixed with a sine wave (
Theoretically, the value is 1.5 when the original series data is a sine wave.
becomes. ), and if we take the value of in normal conditions using the conventional method, it will be approximately 2.7, whereas it is theoretically 30. In contrast, the above-described apparatus for carrying out the method of the present invention is used to separate the time series data into a periodic motion component and a non-periodic motion component. I'd
If we take the value for the J component under normal conditions, it will be approximately 3.0, and accurate determination will be possible. In addition, in the method of the present invention, valuable information on the periodic motion component is also evaluated by taking the RMS value for the periodic motion component.
It is also possible to understand time series data as a whole.

As described in detail below, in the case of the present invention, time series data is subjected to Fourier fm to be separated into a periodic motion component and a non-periodic motion component, and at least one of these is subjected to inverse Fourier transform. Therefore, between the periodic motion component and the non-periodic motion component in the time domain, the effective value (RMS value) is calculated for the former, and the higher-order moment of the probability density function formed by it is calculated for the latter. A normalized value (K value) is calculated by dividing the value by its variance, and both calculated values are combined for a comprehensive evaluation, so it is possible to understand the time series data as a whole, and also to identify abnormalities using universal judgment criteria. A highly sensitive method for monitoring periodic moving objects that can be used for judgment becomes possible.

Therefore, the present invention makes a significant contribution to the improvement of abnormality detection technology for periodic moving bodies.

[Brief explanation of the drawing]

FIG. 1 is a schematic block diagram of an apparatus used to implement the present invention, and FIG. vJ2 is an explanatory diagram showing the arithmetic processing procedure according to the present invention. 1... Shaking off generator 2... Vibration detection device 5...
Arithmetic device 7...Comparison circuit patent applicant: Sumitomo Metal Industries, Ltd. agent Patent attorney: Noboru Kono

Claims (1)

[Claims] 1. Apply Fourier transform to the time series data obtained by sampling the vibration of a periodic body at a constant period. For example, separate the periodic motion component and the non-periodic motion component, and perform inverse Fourier transform on at least one of them. The time series data is separated into a periodic motion component and a non-periodic motion component in the time domain by applying It is characterized by detecting abnormalities in periodic moving bodies by calculating the normalized value by dividing the higher-order moment of the probability density function composed of by its variance, and by evaluating the combination of values from both sides. A method for monitoring periodic moving objects.