JPS58100717A - Abnormality detector - Google Patents

Abstract

PURPOSE:To detect the occurrence of an abnormality of a machine precisely by detecting the difference between an oscillation value at an optional point of time and an estimated oscillation value. CONSTITUTION:A signal SW from a pickup 1 is applied to an addition point 3 and a linear forecasting filter 4 through an amplifier 2. The filter 4 samples and inputs the signal SW at a period tau to calculate the estimated value S'n of a sampled value Sn on the basis of a set forecasting coefficient ai and a sample value Si, and then inputs a samples value Sn-1 to apply it to the addition point 3 time tau later. The addition results Sn-S'n is passed through an absolute value circuit 5 as a forecasting error E(z) and compared by a comparator 6 with a signal Vr from a reference amplitude setter 7 to generate a fault signal Sc regarding an engine when¦E(z)¦>Vr. The signal Sc triggers a one-shot circuit 8 and a pulse signal Pt is counted by a counter 9, whose counted value DC is displayed on a display device 13.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an abnormality detection device for detecting an abnormality in a machine (for example, an engine) having a moving body that performs synchronous motion based on the characteristics of vibrations generated in the machine.

In general, a machine (for example, an engine) that has a moving body that performs synchronous motion (e.g., an engine) generates inherent vibrations due to the periodic motion of the moving body, but the bearings of the moving body may suffer damage such as cracks or other abnormalities in their parts. and abnormal vibration occurs in the machine. Therefore, by detecting this abnormal vibration, it is possible to detect a machine failure or abnormality.

Conventionally, the East method measures engine vibration using a general-purpose measuring instrument such as a vibration meter or waveform analyzer, and then detects abnormal vibration by comparing the vibration characteristics of a normal engine with the measurement results. However, in machines with complex configurations such as engines, the vibration transmission path from the source of abnormal vibration to the vibration detection point is complex, so noise components generated along this path may include There were many cases in which abnormal signals that should have been detected were buried, and as a result, abnormalities could not be detected.

The present invention has been made to solve the above-mentioned problems, and an object of the present invention is to provide an abnormality detection device that can accurately detect abnormalities in machines such as engines that have moving bodies that perform crocodile motion. shall be.

According to the present invention, focusing on the fact that there is a strong correlation between vibrations occurring in a machine such as an engine at any point in time and signals before and after the vibrations, a linear prediction method is used to form an analytical model of vibrations. There is. However, based on the analysis model of the lever, does it vibrate when it is attached to a predetermined position on the machine? Predictively calculate the detection signal of the pickup to be detected after a predetermined time has elapsed from the detection signal of the pickup to be detected at an arbitrary time, and calculate the detection signal of the pickup after a predetermined time has elapsed from the arbitrary time and the result of the above predictive calculation. The above objective is achieved by detecting machine abnormalities from the difference between the prediction error and the amplitude value of the prediction error.

Hereinafter, the present invention will be described in detail based on embodiments of the accompanying drawings.

FIG. 1 shows an example of an output signal from a vibration detection pickup attached to a bearing of an engine. In addition,
The bearing section supports a moving body that moves synchronously, and the detection signal indicates a signal when there is no abnormality in the bearing section or the like.

It is well known that when the output signal 8wg of the pickup is sampled at a predetermined sampling interval in seconds, the sample value +3n of the signal Sv at any time tn has a strong correlation with the sample value at each sample time adjacent to the time tn. It is being

Therefore, the sample violet S1 of the signal Elv every period t
If the current value of is an, this sample value Sn is a range of M sample values 8n −t that are correlated with itself,
It can be predicted using 5n-2,...an-H.

Now, assuming that the predicted value 13n of the sample value Sn is expressed by K as shown in the following equation (j).

Here, a±(i = 1, 2...M) is a prediction coefficient.

Further, the difference between the predicted value an and the sample value an, that is, the prediction error En, is calculated by the following equation 11). εn
= sn-Σa4 B'n -1= n)1ba1 The prediction coefficient athl is determined to minimize the root mean square value of this prediction error εn, and is determined by the following simultaneous equations '+ii). Ru.

By converting the equations i) and n) Y by 2, we get
The following equations 'tv) and v) are obtained.

Sn(g)=H(g)8(z) -----
--・-'1v)z(g) = +3(g)(x-a(
g)) ----------v) Here, H(M) is an operator that minimizes the prediction error v yg (g) yt5, and is calculated by the following equation v1).

Note that FIG. 2 shows the relationship between these equations.

The linear prediction filter 4 shown in FIG. 3 uses this operator H(
-) is a conventionally known filter means.

Hereinafter, the structure and operation of this filter 4 will be briefly explained. This filter 4 includes <CM-1) 2- arithmetic units 4a, an example of which is shown in the figure.
(M-1) addition points 4C for calculating the sum of the output signals of the multipliers 4C and each multiplication @4C in which the prediction coefficients a and ~aM shown in the formula 1) are set as multipliers. It is composed of.

However, this linear prediction filter 4 sequentially operates the z-1 calculator 4 every time a periodic sampling pulse Pg is added.
It acts to shift the input of a to the right with a unit delay. The predicted value 8n is output after the period T of the sampling pulse P8 has elapsed since the sample value 5R-s was input.

FIG. 4 shows an embodiment of an abnormality detection device according to the present invention to which the linear prediction filter 4Y is applied.

In FIG. 4, the acceleration pickup 1 is attached to a bearing part of the engine, etc., and the vibration at the attachment point is detected by an electric signal of 8W.
The signal 8W is applied to the summing point 3 and the linear prediction filter 4 via the amplification rI2.

The edge shape prediction filter 4 receives the output signal 8 of the pickup 1.
W is sampled and input at the sampling pulse P8 of period i, and the preset prediction coefficient az (1=n-1, n-zt-n-r4) and the sampling value B i of the input signal 8W are input. (i = n-1.
n-i...n-M) Sample value 8n based on K
The predicted value of 8n9 is calculated and output. This predicted value 13n is calculated at the addition point 3 after time τ after inputting the sample value an-11.
added to. Then, the addition result of this addition point 3 (8n-
an) is the absolute value circuit 5 as the prediction error 1 (g).
is output to.

By the way, when the bearing part, which is the mounting position of the pickup l, is normal, the output signal E1w of the pickup is as shown in Fig. 5 (
As shown in a), in this case there is a prediction error! (IB> is the minimum, that is, it becomes white noise-like as shown in FIG. 5(b)K).

Further, if some damage occurs to the bearing portion, impulse-like vibrations are periodically generated due to the damage. In this case, the vibration signal due to the impulse-like vibration leaks a noise component Kll, and therefore the signal 131y detected by the pickup becomes as shown in FIG. 6e)K. Note that the arrow in the same figure indicates a portion that includes a signal due to the above-mentioned impulse-like vibration, and the prediction error IC(g) in this portion is equal to the peak signal 1 shown in FIG. 6(b).
Hail as p.

FIG. 7 shows the amplitude occurrence probability of a prediction error of 1 (g), with the amplitude mu on the horizontal axis and the probability P (mu) on the vertical axis. Fig. 5(b) Prediction error B when the engine shown in K is normal. Also, the 6th
The probability density distribution of the prediction error l (-) when an abnormality occurs in the engine shown in Figure 0) is the peak signal 1
Because it includes p, the song 4151L has a shape in which the length of 5 km is extended to the larger part K.

Therefore, if the engine is normal,
1 An abnormality in the engine can be determined by detecting a state in which the amplitude of the prediction error IC(z) is larger than the maximum amplitude unevenness r of the prediction error It (=), and the device shown in FIG. Make a decision.

That is, the absolute value circuit 5 shown in the same figure has the above prediction error 1.
The absolute value lx (z) of (z) is outputted and added to the + side input of the comparator 6. Therefore, the comparator 6 compares the signal corresponding to the amplitude r set by the reference amplitude setter 7 with the pressure signal lFi(g)lt-4-, and
When K(g)l>Vr, that is, when the engine is abnormal, a signal SC is generated. The signal sC triggers the one-sit circuit 8 to generate a predetermined pulse signal Pt (see FIG. 3(0))'.
KK added.

The count value DC of this counter 9 is the curve Lt in FIG.
(This corresponds to the enclosed part DK.

In this embodiment, the presence or absence of an abnormality is detected by displaying a count value DC11 on the display device 13 for each unit time (for example, one second). Note that the clock signal generator 10 outputs a one-second clock, and its output signal pa is input to the reset terminal of the counter 9 via the delay circuit 11, and a latch that latches the count value of the counter 9. The signal is input to the circuit 12 as a latch signal.

In this way, according to the device of the above embodiment, the display device 1
An abnormality in the engine can be detected from the display contents of 3. Of course, when the engine is normal, the counter 9 does not perform counting operations, so the displayed value on the display device 13 becomes zero.

In the above-described embodiment, one amplitude value vr is used as a comparison standard for the amplitude of the predicted error E (=s), but the present invention is not limited to this. In other words, since the magnitude of the prediction error amplitude and the degree of damage are generally proportional, the fourth It is possible to more accurately grasp the shape of the amplitude probability density distribution of the prediction error as shown by the curve in the figure, and thereby the degree of damage in each part of the engine can be determined more precisely. Of course, in this case, the comparison circuit 6tJL: reference amplitude value Vr
You need as many as

Further, in the above-described embodiments, the apparatus was constructed using analog circuits, but it may also be constructed using digital circuits.

Furthermore, if it is constructed using digital circuits, it goes without saying that a microcomputer can be used as a component.

As explained above, the abnormality detection device according to the present invention detects an abnormality in the machine based on the difference between the vibration value picked up from the machine at any point in time and the predicted vibration value at the above-mentioned arbitrary point predicted before that point in time. Since the presence or absence of the problem is detected, it is possible to eliminate the deficiencies of the conventional device described above and accurately detect an abnormality in the machine.

Further, the device of the present invention can be widely applied to abnormality detection of machines other than engines by appropriately changing the prediction coefficients set in the linear prediction filter shown in the embodiment.

[Brief explanation of the drawing]

Fig. 1 is a waveform diagram showing an example of the output signal of a vibration detection pickup attached to the bearing of the engine, and Fig.
Block diagram for explaining the action of the linear prediction filter,
FIG. 3 is a block diagram showing an example of a linear prediction filter,
1g4 is a block diagram showing an embodiment of the abnormality detection device according to the present invention, FIG. 5(a) is a waveform diagram showing a signal 8W when the engine is normal, and FIG. The prediction error ml when the engine is normal is the waveform diagram shown in (g). Prediction error when an abnormality occurs 1! The waveform diagram shown in FIG. 7 showing i (g) is a graph showing the amplitude probability density distribution of the prediction error E (Z). 1... Acceleration pickup, 3-... Addition point, 4...
・Linear prediction filter, 5...Absolute value circuit, 6...Comparator, 7...Reference amplitude setter, 8-...One-shot circuit, 9...Counter, 10...Clock Signal generator, 13...display device.

Claims (1)

[Claims] An abnormality detection device that detects an abnormality in a machine from vibrations of the machine, which includes a moving body that performs synchronous motion, includes a pickup attached to the machine to detect vibrations, and vibration characteristics when the machine is normal. calculation means for predicting and calculating the detection signal of the pickup at S after a predetermined period of time has elapsed from the detection signal of the pickup at an arbitrary point in time; An abnormality detection device comprising means for determining an abnormality in the machine based on the difference between the detection signal and the output signal of the calculation means.