JPH05340799A - Vibration discriminating device - Google Patents

Abstract

PURPOSE:To more accurately discriminate the dangerousness of vibrations by extracting a plurality of feature amounts from a vibrational acceleration waveform based on a preset detection level. CONSTITUTION:The mobile electrode of a capacitance sensor 10 oscillates due to acceleration generated by an earthquake and the capacitance between the mobile electrode and a fixed electrode changes. A CR oscillation circuit 5 outputs a signal in corresponding to the capacitance change. In a microcomputer 2, the vibrational acceleration of the earthquake is first calculated from the output of a vibration detecting section 1 by means of a frequency measuring timer 17, arithmetic section 19, reference period storage section 18, etc., and a vibrational acceleration waveform is obtained by sampling the acceleration at regular time intervals. The microcomputer 2 extracts a plurality of feature amounts based on the acceleration waveform obtained from the output of the section 1 and discriminates whether or not the vibrations are caused by an earthquake by performing fuzzy inference based on the feature amounts.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a vibration discriminating apparatus which discriminates vibration caused by an earthquake or abnormal vibration of a machine.

[0002]

2. Description of the Related Art Conventionally, in an earthquake sensor that detects an earthquake,
The vibration acceleration is detected, and when the acceleration exceeds a preset level, processing such as issuing an alarm as an earthquake is performed.

By the way, various actuators, for example,
An air cylinder or the like is configured to detect whether or not the piston rod has reached a top dead center or a bottom dead center, and determine whether the air cylinder is operating normally.

[0004]

Since the conventional seismic sensor simply determines that there is an earthquake and issues an alarm when the magnitude of vibration acceleration exceeds a preset level. , It may not be possible to make an accurate determination.

Further, in the conventional air cylinder or the like, since it is determined whether or not the piston cylinder is operating normally by detecting the position of the piston rod, the abnormality cannot be detected until the operation of the air cylinder is stopped, and the operation is not performed. It was difficult to predict a failure before it stopped.

The present invention has been made in view of the above points, and it is possible to more accurately determine an earthquake and to predict a machine failure by determining an abnormal vibration of the machine. The purpose is to do so.

[0007]

In order to achieve the above object, the present invention is configured as follows.

That is, the vibration discriminating apparatus according to the first aspect of the present invention is based on a vibration detection section for detecting vibration and a detection level preset from a vibration acceleration waveform corresponding to the output of the vibration detection section. And a plurality of feature amounts are extracted, and a discriminating unit for discriminating an earthquake is provided based on these feature amounts.

According to a second aspect of the present invention, there is provided a vibration discriminating apparatus for discriminating abnormal vibration of a machine, the vibration detecting section detecting the vibration of the machine, and the vibration detecting section during normal operation of the machine. A learning unit that preliminarily learns a vibration acceleration waveform corresponding to the output of the vibration detection unit, and an abnormal vibration of the machine is determined based on the vibration acceleration waveform that is learned in advance and the vibration acceleration waveform corresponding to the output of the vibration detection unit. And a discriminating unit that operates.

[0010]

According to the first aspect of the present invention, a plurality of feature quantities are extracted from the vibration acceleration waveform, and based on these,
Since the earthquake is discriminated, the discrimination can be made more accurately than the conventional example in which the magnitude of the vibration acceleration is simply discriminated.

According to the second aspect of the present invention, the vibration acceleration waveform during normal operation of the machine is learned in advance, and the vibration acceleration waveform obtained from the vibration detection unit and the vibration acceleration waveform learned in advance are used. Since the abnormal vibration of the machine is determined, it is possible to detect the abnormality of the machine before the machine stops.

[0012]

Embodiments of the present invention will now be described in detail with reference to the drawings.

FIG. 1 is a block diagram of a vibration discriminating apparatus according to an embodiment of the present invention. This embodiment is an apparatus for detecting an earthquake and issuing an alarm.

The vibration discriminating apparatus of this embodiment detects a vibration and outputs a corresponding frequency signal, and based on the output of the vibration detecting unit 1, whether or not there is an earthquake as described later. A microcomputer 2 for determining whether or not there is a detection unit, a setting unit 3 for externally setting a detection level, and the like; and an output unit 4 for outputting an alarm signal or the like to an alarm device (not shown) in response to an output of the microcomputer 2 when an abnormality occurs. ing.

The vibration detector 1 of this embodiment comprises a CR oscillation circuit 5 and a frequency divider 6 for dividing the output of the CR oscillation circuit 5.
It has and.

The CR oscillation circuit 5 includes a capacitance sensor 10 which will be described later.
A capacitance changing portion C composed of a resistor R, a stabilizing resistor Rf
And the inverters 7 to 9, and changes the capacitance of the capacitance sensor 10 with the frequency f _{0 according} to the relationship shown by the following equation.
It is converted into (Hz) and output.

F _out = 1 / (2.2CR) As shown in the exploded perspective view of FIG. 2 and the cross-sectional view of FIG. 3, the capacitance sensor 10 whose capacitance changes according to vibration is shown in FIG.
A printed wiring board 13 having a fixed electrode 12 formed therein and a movable electrode 14 having a vibration spring are housed in a case 11 via a spacer 15, and a cover 16 is attached. In this capacitance sensor 10,
The movable electrode 14 swings due to the acceleration generated by the vibration, and the electrostatic capacitance C (Farad) between the movable electrode 14 and the fixed electrode 12 is given by the following equation.

C = ε ₀ · (S / d) where S: the area where the electrodes face each other d: the distance between the electrodes ε ₀ : the permittivity, that is, the movable electrode 14 oscillates to generate between the electrodes. The capacity of the CR oscillation circuit 5 changes.
Outputs a signal having a frequency according to the change in capacitance.

The microcomputer 2 includes a frequency measurement timer 17, a reference cycle storage section 18 in which a reference cycle (cycle at rest) of the CR oscillation circuit output is stored in advance, and an output of the vibration detection section 1 for the vibration acceleration. An arithmetic unit 19 that performs arithmetic operations and the like, and a control unit 20 that performs fuzzy inference and learning to determine whether or not there is an earthquake as described below are provided.

In this microcomputer 2, first,
The vibration acceleration is calculated from the output of the vibration detection unit 1 as follows.

That is, the output of the vibration detector 1 is as shown in FIG.
When the electrode approaches the basic period T at rest (vibration to the + side) shown in (A), it becomes T ⁺ as shown in FIG. 4B, and when the electrode separates (-). Vibration to the side)
Becomes T ⁻ as shown in FIG. 4C, and the vibration acceleration is calculated by the following equation.

+ Side vibration acceleration (ΔT ⁺ ) = T ⁺ −T − side vibration acceleration (ΔT ⁻ ) = T ⁻ −T This calculation is performed by the frequency measurement timer 17, the calculation unit 19, the reference period storage unit 18, and the like. By sampling this acceleration at regular intervals, a vibration acceleration waveform can be obtained as shown in FIG.

In this embodiment, further, as shown in FIG. 6, moving average processing of sampled vibration acceleration data at four points is performed, and as shown in FIG. Has been cut. In FIG. 6, Xi is input, Yi is response output, T is time,
M = 4.

The microcomputer 2 of this embodiment is
As shown in the functional block diagram of FIG.
Based on the acceleration waveform obtained as described above from the output of, the feature amount described below is extracted, and fuzzy inference is performed based on this feature amount to determine whether or not there is an earthquake. The inference unit 22 and a function as a correction unit 23 that learns the acceleration waveform corresponding to the output of the vibration detection unit 1 in advance and corrects the determination in the inference unit 22 according to the installation situation are provided.

The correction unit 23 is given an acceleration waveform through the switching circuit 24 by an external operation at the time of initial setting or the like, and a learning unit 25 for learning the acceleration waveform, and based on the learning data of the learning unit 25. , And a gain adjusting unit 26 for correcting the determination output of the inference unit 22.

The inference unit 22 calculates, for example, the vibration acceleration from the acceleration waveform shown in FIG. 9, calculates the vibration cycle from the zero-cross point, and further extracts the following four types of feature quantities 1 to 4.

That is, the feature quantity 1 is shown in FIG.
ESU of energy exceeding the detection level shown in
M, the feature amount 2 is a total energy SESUM of energy within a predetermined time, and the feature amount 3 is a frequency TS in which the time (half cycle) between points where the acceleration waveform crosses zero is 0.3 s or less. The characteristic amount 4 is the frequency TB in which the time (half cycle) between the points at which the acceleration waveform crosses zero exceeds 0.3 s.

Now, the process for calculating energy from the acceleration waveform will be described.

The velocity v can be calculated by integrating the acceleration a, that is, by obtaining the area of the acceleration waveform. The velocity v and the energy E have a relation of E = (1/2) · mv ² , that is, the energy E has a proportional relation with the square of the velocity v.

Therefore, a speed / energy conversion table is prepared in advance, and the energy is calculated from the conversion table.

In this embodiment, the determination by the inference unit 22 is performed by the membership functions shown in FIGS. 10 and 11 and the fuzzy rules shown in Tables 1 to 3 below.

That is, FIG. 10A shows that the feature quantity 1 (ES
UM) membership function, "SLM", "M
It has three labels, "DL" and "BIG".
0 (B) shows the membership function of the feature amount 2 (SESUM), which has three labels of “SML”, “MDL”, and “BIG”, and FIG. 10C shows the feature amount 3 ( T
S) shows the membership function of "SML", "BI
10D has two labels, “G”, and FIG. 10D shows the membership function of the feature amount 4 (TB).
It has two labels, "L" and "BIG".

FIG. 11 shows the membership function of the judgment output which is the consequent part, and has two labels, "unout (alarm non-output)" and "out (alarm output)".

Fuzzy rules are shown in Tables 1 to 3 below.
In this embodiment, fuzzy inference is performed according to the fuzzy rules in Table 1 based on the feature quantities 1 and 2 corresponding to energy, and "un ou"
When it is determined to be “t”, the fuzzy rule in Table 2 is used. When it is determined to be “out”, the fuzzy rule is used in Table 3 to determine the feature amounts 3 and 4 corresponding to the periodic frequencies. Finally, fuzzy inference is performed to make a judgment.

That is, Table 2 shows the fuzzy rule when the first fuzzy inference determines "un out".
Table 3 shows the fuzzy rules when the first fuzzy inference determines “out”.

[0036]

[Table 1]

[0037]

[Table 2]

[0038]

[Table 3]

In the fuzzy inference by the inference unit 22, as shown in FIG. 12, first, data is sampled (step 1), the above-mentioned four types of feature quantities 1 to 4 are extracted (step 2), and then (Step 3).

That is, the degree of conformity is obtained in the membership function corresponding to each fuzzy rule in Table 1 by the characteristic amount 1 (ESUM) and the characteristic amount 2 (SESUM). Then, for each fuzzy rule, the smallest value of the conformance of each antecedent part is selected as the antecedent conformity (M
IN operation). In this way, the membership functions related to the consequent part of each fuzzy rule are cut by the conformance of the antecedent part obtained in each fuzzy rule, and the membership functions related to all the cut fuzzy rules are overlapped (MAX Calculation), the output value of the position corresponding to the position of the center of gravity of the superimposed figure is obtained, and it is determined whether or not there is an earthquake (“out (alarm output)”) based on this output value.

By this first fuzzy inference, "un
If it is determined to be “out”, the goodness of fit is obtained in the membership function corresponding to each fuzzy rule in Table 2 by the feature amount 3 (TS) and the feature amount 4 (TB). Then, for each fuzzy rule, the smallest value of the matching degree of each antecedent part is selected as the antecedent part matching degree (MI.
N operation). In this way, the membership functions related to the consequent part of each fuzzy rule are cut by the conformance of the antecedent part obtained in each fuzzy rule, and the membership functions related to all the cut fuzzy rules are overlapped (MAX Calculation), the output value at the position corresponding to the position of the center of gravity of the superimposed figure is obtained, and the final determination is made based on this output value.

In addition, the first fuzzy inference causes "ou
If it is determined to be “t”, the goodness of fit is obtained in the membership function corresponding to each fuzzy rule in Table 3 from the feature quantity 3 (TS) and the feature quantity 4 (TB).
Then, for each fuzzy rule, the smallest value of the matching degree of each antecedent part is selected as the antecedent part matching degree (MIN calculation). In this way, the membership functions related to the consequent part of each fuzzy rule are cut by the conformance of the antecedent part obtained in each fuzzy rule, and the membership functions related to all the cut fuzzy rules are overlapped (MAX Calculation), an output value at a position corresponding to the position of the center of gravity of the superimposed figure is obtained, and it is finally determined whether or not there is an earthquake (“out”) based on this output value.

Next, the operation of the correction unit 23 for correcting the determination output value by the gain adjustment unit 26 so that the determination output of the inference unit 22 will be considered in consideration of the installation situation will be described.

The learning section 25 is given an acceleration waveform through the switching circuit 24 by an operation from the outside at the time of initial setting or the like, and based on this, an energy level of ordinary vibration (normal vibration in normal time). Is calculated as shown in FIG.
And the inference unit 2 based on the relationship shown in Table 4 below.
The gain adjustment amount, that is, the correction coefficient, for the determination output from 2 is obtained.

[0045]

[Table 4]

In this learning unit 25, as shown in FIG. 14, vibration acceleration is sampled from the acceleration waveform every 5 milliseconds (step 1), and a cycle at which the acceleration waveform crosses zero is obtained (step 2). The energy level of microvibration is calculated in the same manner as 22 (step 3), and the gain adjustment amount is obtained by using this energy level as an input,
It is determined whether 10 seconds have passed (step 5), and when 10 seconds have passed, the process ends. That is, in this embodiment, learning is performed for 10 seconds.

In this way, the judgment output of the inference unit 22 is corrected by the gain adjustment amount obtained by the learning unit 25.

Therefore, it is possible to make a determination according to the installation situation.

Furthermore, in this embodiment, since the drift of the sensor output can be considered due to the influence of the surrounding environment and the like,
This drift correction is performed as follows.

That is, as shown in FIG. 15, the detected acceleration exceeds the detection level, for example, +5 gal and is 10
When it does not cross the origin value of the sensor within a second, it is determined that drift has occurred, and the measured value at that time is used as the drift correction value.

As described above, since the feature quantity is extracted from the vibration acceleration waveform and the earthquake is discriminated and the alarm is output, it is more accurate than the conventional example in which the earthquake is discriminated only by the magnitude of the acceleration. It can be distinguished.

Since this vibration discriminating apparatus can discriminate an earthquake according to a detection level preset from the outside, it is possible to detect a vibration that affects a machine tool requiring high accuracy, for example. Therefore, it is possible to reduce machining defects due to the machine tool.

Further, since the capacitance type acceleration sensor is used for detecting the vibration, the cost can be reduced as compared with the conventional seismic sensor.

FIG. 16 is a block diagram of a vibration discriminating apparatus 27 according to another embodiment of the present invention. The vibration discriminating apparatus 27 is attached to an air cylinder 28 as shown in FIG. The abnormal vibrations of No. 28 are detected, and the portions corresponding to the above-described embodiment are designated by the same reference numerals.

The vibration discriminating device 27 of this embodiment detects the vibration and outputs the corresponding frequency signal to the vibration detecting unit 1.
And a microcomputer 29 that determines whether or not the vibration is abnormal based on the output of the vibration detection unit 1, a setting unit 3 that sets the detection level and the like from the outside, and a response to the output of the microcomputer 29. In addition, the output unit 4 outputs an alarm signal or the like to an alarm device (not shown) when an abnormality occurs.

The process from detecting vibration to obtaining the acceleration waveform and performing the moving average process is the same as in the above-described embodiment.

In this embodiment, after the moving average processing, a secondary filter (software LPF) is used to classify the vibration frequency into bands.

That is, as shown in FIG. 18, the elapsed time t until the amplitude outputs Yi and Yi of the acceleration waveform subjected to the moving average processing separate from the origin and cross the origin next time.
(1/2 cycle) is input, a range is provided by this cycle t, and stratification is performed. From the result, the input waveform Yi is multiplied by the filter coefficient kn to obtain the output of the secondary filter. Learning and inference to be described later are performed based on the output of this secondary filter.

Microcomputer 29 of this embodiment
19 corresponds to the learning unit 30 that preliminarily learns the vibration acceleration waveform during normal operation of the air cylinder 28, the vibration acceleration waveform that has been preliminarily learned, and the output of the vibration detection unit 1, as shown in the functional block diagram of FIG. It has a function as an inference unit as a discriminating unit for discriminating abnormal vibration of the air cylinder 28 based on the vibration acceleration waveform.

In this embodiment, when the equipment is installed or the maintenance is completed, the vibration acceleration waveform during the normal operation of the air cylinder 28 is learned in advance, as shown in the flowchart of FIG.

That is, the external switch enters the waveform learning mode (step 1), the operation is started (goes) from the bottom dead center to the top dead center (step 2), and the acceleration waveform data at that time is sampled (step 3). Further, the operation is started from the top dead center to the bottom dead center (return) (step 4), the acceleration waveform data at that time is sampled (step 5), and it is judged whether or not 4 cycles are completed (step 6). ) When the four cycles are completed, the learning data is stored (step 7) and the process is completed.

In this learning mode, the following basic data is collected from the waveform data.

Going cycle time TSF Returning cycle time TSR Going first peak acceleration FGF Going second peak acceleration SGF Returning first peak acceleration FGR Returning second peak acceleration SGR Setting (threshold) acceleration or more Number of times acceleration has occurred GPLS Next, processing during normal operation after the learning is completed will be described with reference to the flowchart in FIG.

First, the acceleration waveform data is sampled (step 1), the going and returning cycle times are sampled (step 2), and the feature state is extracted and fuzzy inference is performed as described later to determine the operating state. It is analyzed (step 3) and it is judged whether or not it is normal (step 4), and when it is abnormal vibration, an alarm is output (step 5).

In this normal operation, the same data as in the initial learning is sampled and the next data is used as the feature amount.

Initial GPLS-Current GPLS = GP
LSS Initial TSF-Current TSF = TSFS Initial TSR-Current TSR = TSRS Initial FGF-Current FGF = FGFS Initial FGR-Current FGR = FGRS Fuzzy inference is performed using these five types of feature values as inputs. Determine whether the vibration is abnormal.

This fuzzy reasoning is shown in FIG. 22 and FIG.
The membership function shown in Table 1 and the fuzzy rules in Table 5 below are used.

That is, FIG. 22A shows that the feature quantity 1 (GP
LSS) membership function, "SML",
5 of "MSL", "MDL", "MBG", "BIG"
22B, the feature quantity 2 (T
SFS) membership function, "SML",
5 of "MSL", "MDL", "MBG", "BIG"
22C, the feature quantity 3 (T
SRS) membership function, "SML",
5 of "MSL", "MDL", "MBG", "BIG"
22 (D) has 4 labels.
GFS) membership function, "SML",
5 of "MSL", "MDL", "MBG", "BIG"
22E, the feature quantity 5 (F
GRS) membership function, "SML",
5 of "MSL", "MDL", "MBG", "BIG"
It has individual labels.

FIG. 23 shows the membership function values of the judgment output, which is the consequent part, and is "NB (good)" and "NM".
It has five labels, “Good”, “ZR (normal)”, “Slightly stopped probability (maintenance preparation required)”, and “Stop probability large (maintenance required)”.

The fuzzy rules are as shown in Table 5 below.

[0071]

[Table 5]

As described above, the feature amount is extracted from the data obtained from the acceleration waveform during normal operation and the data obtained from the acceleration waveform during operation, and fuzzy inference is performed to determine abnormal vibration. Before the 28 stops abnormally, the abnormality can be detected, that is, the failure can be predicted. Note that FIG. 24 (A) shows an acceleration waveform during normal operation in 1/2 cycle, and FIG.
Shows an example of acceleration waveform during abnormal vibration.

In the above-mentioned embodiment, the abnormal vibration of the air cylinder is discriminated, but it goes without saying that the present invention can be applied to the discrimination of the abnormal vibration of other actuators and other machines.

[0074]

As described above, according to the present invention, a plurality of characteristic quantities are extracted from a vibration acceleration waveform based on a preset detection level to discriminate an earthquake, so that an accurate discrimination can be performed. ..

Further, the vibration acceleration waveform during normal operation of the machine is learned in advance, and the abnormal vibration of the machine is discriminated based on the learned vibration acceleration waveform and the vibration acceleration waveform detected by the vibration detecting section. So it is possible to detect it before the machine completely fails.

[Brief description of drawings]

FIG. 1 is a block diagram showing an overall configuration of an embodiment to which the present invention is applied.

FIG. 2 is an exploded perspective view of a capacitance sensor.

FIG. 3 is a sectional view of a capacitance sensor.

FIG. 4 is a waveform diagram showing a change in frequency due to vibration.

FIG. 5 is a waveform diagram of vibration acceleration.

FIG. 6 is a block diagram of moving average processing.

FIG. 7 is an acceleration waveform diagram for explaining moving average processing.

FIG. 8 is a functional block diagram of the embodiment shown in FIG.

FIG. 9 is an acceleration waveform diagram for explaining extraction of a feature amount.

FIG. 10 is a diagram showing a membership function of the antecedent part.

FIG. 11 is a diagram showing a membership function of a consequent part.

FIG. 12 is a flowchart for explaining the operation.

FIG. 13 is a diagram showing a relationship between gain and energy level.

FIG. 14 is a flowchart for explaining the operation.

FIG. 15 is a diagram for explaining drift correction.

FIG. 16 is a block diagram of another embodiment of the present invention.

FIG. 17 is a perspective view of an air cylinder to which the embodiment of FIG. 16 is applied.

FIG. 18 is an explanatory diagram of a secondary filter.

FIG. 19 is a functional block diagram of the embodiment shown in FIG.

FIG. 20 is a flowchart for explaining the operation.

FIG. 21 is a flowchart for explaining the operation.

FIG. 22 is a diagram showing a membership function of the antecedent part.

FIG. 23 is a diagram showing a membership function of the consequent part.

FIG. 24 is a diagram showing acceleration waveforms in a normal state and an abnormal state.

[Explanation of symbols]

 1 Vibration detection part 2,29 Microcomputer 22,31 Inference part 25.30 Learning part

Claims (2)

[Claims] 1. A vibration detecting section for detecting vibration, and a plurality of characteristic quantities are extracted from a vibration acceleration waveform corresponding to the output of the vibration detecting section based on a preset detection level, and based on these characteristic quantities. A vibration discriminating device comprising: a discriminating unit for discriminating an earthquake. 2. A vibration discriminating apparatus for discriminating abnormal vibration of a machine, wherein a vibration detecting section for detecting vibration of the machine, and a vibration acceleration waveform corresponding to an output of the vibration detecting section during normal operation of the machine. And a discriminating unit for discriminating abnormal vibration of the machine based on the vibration acceleration waveform learned in advance and the vibration acceleration waveform corresponding to the output of the vibration detecting unit. Vibration discriminating device.