JPH1172379A - State judgment apparatus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a state judgment apparatus whose operating efficiency is enhanced when sensing information is judged to be abnormal by a method wherein the sensing information which is judged to be abnormal is prevented from being used to generate a judgment condition. SOLUTION: A state judgment apparatus is operated in a learning mode, a setting mode or a judgment mode. A judgment condition is generated on the basis of a signal which is input from a vibration sensor in the learning mode. The judgment condition is used in the setting mode and the judgment mode. Whether a mechanical apparatus as an object to be detected is normal or abnormal is judged on the basis of the signal which is input from the vibration sensor. At this time, a mode changeover switch 8 can take three states, i.e., a TEACH state, a SET state and a RUN state, and the states have a function which instructs a shift to the learning mode, the setting mode and the judgment mode. An ALARM-LED 19 is turned on when the sensor is disconnected or short-circuited. A PASS-LED 17 indicates that a judged result is 'normal'. In addition, an NG-LED 18 indicates that a judged result is 'abnormal'.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a state judging device for judging a state of a detection target based on a signal inputted from a sensor for detecting a vibration state of the detection target in various devices.

[0002]

2. Description of the Related Art As a state determination device of this type, a sensor for detecting vibration of equipment and the like, sensing information storage means for storing sensing information to be detected by the sensor,
2. Description of the Related Art There is known a vibration monitoring device including a determination condition generating unit that generates a condition for determining a state of a detection target and a determination unit that determines a state of a detection target (for example, Japanese Patent Application Laid-Open No.
No. 113351). Here, the generation of the determination condition is performed in the learning mode, and a function of notifying that the sensing information is abnormal when it is determined that the sensing information is abnormal is also disclosed. However, even in such a case, the sensing information determined to be abnormal is once adopted as a criterion for generating a determination condition, and it is left to the user to determine whether or not to remove the influence.

[0003]

However, when learning a reference signal in this type of state determination device, in the case of a vibration monitoring device, the mechanical device serving as the vibration source is stopped in the first place. Signal input may fail due to no vibration occurring, forgetting to set the vibration sensor, or improperly contacting the vibration sensor. The fact that the user is required to determine each time even when an abnormal signal is input is very troublesome because such a situation occurs relatively frequently. The present invention solves the above-described problem, and provides a state determination device capable of improving work efficiency when a signal input for generating a determination condition can be determined to be abnormal without waiting for a user's determination. The purpose is to:

[0004]

To achieve the above object, the present invention provides a sensing information storage means for storing a signal input from a sensor for detecting a state of a detection target or sensing information which is information extracted from the signal. And a determination condition generating means for generating a condition for determining the state of the detection target based on the sensing information stored in the sensing information storage means, and input from the sensor according to the condition generated by the determination condition generating means. In a state determination device including a determination unit that determines the state of the detection target from the signal, determine whether the sensing information is abnormal,
Abnormality monitoring that, when determined to be abnormal, prevents the sensing information determined to be abnormal from being used for the generation of the determination condition by the determination condition generation means and outputs a signal notifying that the abnormality has occurred. Means.

In the above configuration, when an abnormal input is made, the judgment condition is not updated based on the abnormal input, so that the judgment condition is not always updated automatically. Since the work of removing the influence of the abnormal input from the updated determination condition is not required, the processing after the abnormal input is performed can be performed efficiently.

[0006] The "sensing information" may be simply a digitized signal waveform detected by the sensor, or a so-called characteristic amount such as an average value or a maximum value extracted therefrom. The term “blocking” includes not storing the sensing information in the sensing information storage means and storing the sensing information in the sensing information storage means but not using the stored sensing information for generating the determination condition. The term “not stored” as used herein includes not only physically not storing, but also the case where it is physically stored but not logically stored. As for the “blocking” method, it is more preferable not to store the sensing information in the sensing information storage means. Then, when the signal from the sensor is input again, it is not necessary to delete the sensing information determined to be abnormal in advance or to overwrite the sensing information determined to be abnormal.

[0007] In addition, when the value included in the sensing information is not within a predetermined value range, it can be determined that there is an abnormality. The “value included in the sensing information” is preferably a value related to the magnitude of the signal, such as an average value, a mean square value, and a maximum value of the input signal waveform. These values may be determined for specific frequency band components of the signal.

When the value contained in the sensing information is not within a predetermined range defined based on the value contained in other sensing information stored in the sensing information storage means before that time, it is determined that there is an abnormality. You can make it. As a result, a more accurate abnormality determination can be made based on non-abnormal sensing information. The value included in the other sensing information stored before may be the first stored value or the value stored immediately before,
It may be a value obtained by calculating from a plurality of values stored several times.

Further, the input of the signal from the sensor can be immediately stopped when it is determined that the sensing information is abnormal. Several tens of points to input the signal from the sensor
It can take seconds, but in such cases,
If the signal input is stopped immediately after the abnormality determination, the operation of redoing the signal input can be started without waiting for the completion of the input of the abnormal signal, and the operation can be performed efficiently.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be specifically described below with reference to the accompanying drawings. The state determination device described here inputs a signal from a vibration sensor that converts a vibration generated by a mechanical device having a built-in vibration source into an electric signal using a piezoelectric element, and the mechanical device is configured based on the input signal. This is a vibration judging device for judging whether it is normal or abnormal.
This vibration determination device operates in a learning mode, a setting mode or a determination mode, generates a determination condition based on a signal input from the vibration sensor in the learning mode, and uses the determination condition in the setting mode and the determination mode. It is determined whether the mechanical device to be detected is normal or abnormal based on the signal input from the vibration sensor. In the setting mode, the determination conditions can be manually adjusted.

The terms used hereinafter will be briefly described. The “PV value” is a value indicating the degree of separation of the signal to be determined. The degree of separation is a quantity that comprehensively represents how much an input signal differs from a signal serving as a criterion based on various feature amounts extracted from the input signal, which will be described in detail later. The “SV value” is a threshold value of the degree of separation for judging whether it is normal or abnormal. A value of 3 is set as a default value, and the value can be manually increased or decreased.

FIG. 1 shows a hardware configuration of the vibration judging device. This vibration judging device is provided with a one-chip microcomputer 1 (called a microcomputer) including a memory and an A / D converter. The microcomputer 1 has a transducer 4 including a sensor 2 for detecting vibration and an amplifier 3 for detecting a vibration to be detected. A signal is input from an A / D input port via an amplification degree variable amplifier 5 and a three-band filter 6. The microcomputer 1 has switches, which will be described in detail later.
A filter selection switch 7 for selecting a band filter 6;
Mode switch 8, START switch 9, STA
RT external input 10, TEACH switch 11, CLEA
R switch 12, SV-UP switch 13, and SV-
The DOWN switch 14 is connected. Further, the microcomputer 1 has a BUSY-LED1 as an LED indicator.
5, learning data LED16 (D1 to D5), PASS-
LED17, NG-LED18, ALARM-LED1
9, 7-segment LED with SV value 7 and 7-segment LED with PV value
22 is connected, and a PASS output 23 is provided as an output terminal.
NG output 24, ALARM output 25, PV value terminal output 2
6 are provided. The LEVEL-LED 29 is connected to the variable amplification amplifier 5 via the level meter IC 28. A power supply unit 30, a POWER-LED 31, and an EEPROM 32 (sensing information storage unit) for storing learning data are provided.

FIG. 2 is a front view of a panel of the vibration judging device, and FIG. 3 is a bottom view thereof. The switches and LED indicators that appear outside the apparatus main body will be described. The mode switch (MODE) 8 can take three states, TEACH, SET, and RUN, and has a function of instructing a transition to a learning mode, a setting mode, and a determination mode, respectively. The setting mode is a mode for manually adjusting the determination condition generated in the learning mode and performing the determination operation.

When pressed in the learning mode, the TEACH switch 11 starts learning and stores a set of sensing information (referred to as learning data in the present embodiment).
Up to five sets of learning data can be stored. CL
When the EAR switch 12 is pressed in the learning mode, the EAR switch 12 deletes one set of learning data in order from the newly stored learning data. When the START switch 9 is pressed in the setting mode, the determination starts. Hold down to make a continuous determination. This switch is effective only in the setting mode, and the instruction to start the determination in the determination mode is made by an external input. In the determination mode, erroneous operation is prevented by preventing input from switches on the panel including other switches. SV-UP switch 13
Increases the SV value setting by one when pressed in the setting mode. When the SV-DOWN switch 14 is pressed in the setting mode, the setting of the SV value is decreased by one.

The ALARM-LED 19 lights when the sensor is disconnected or short-circuited. The PASS-LED 17 indicates the determination result “normal”, and turns off at the start of the determination in the setting mode or the determination mode, and turns off the PV after the determination is completed.
The light turns on when the value is equal to or less than the SV value, and turns off when the PV value exceeds the SV value. However, in the second and subsequent determinations in the continuous determination, the previous lighting state is maintained even during the determination. NG-L
The ED 18 indicates the determination result “abnormal”, and turns off at the start of the determination in the setting mode or the determination mode, and turns on when the PV value exceeds the SV value after the determination is completed.
The light is turned off when the value is equal to or smaller than the SV value. However, in the second and subsequent determinations in the continuous determination, the previous lighting state is maintained even during the determination. These ALARM, PASS, and NG are also output as electric signals from the terminals of the terminal block 34 shown in FIG.

The BUSY-LED 15 is lit during the determination process in the setting mode or the determination mode. PV
-LED22, in the setting mode or the determination mode,
It turns off at the start of the judgment, and displays the PV value after the end of the judgment. However, in the second and subsequent determinations in the continuous determination, the previous display state is maintained even during the determination. PV value is the terminal block 3 shown in FIG.
4 is also output as an electric signal. SV-LE
D21 displays the SV value in the setting mode or the determination mode. LED 16 of D1 to D5 (learning data LE
D: storage state output means) sequentially lights up from D1 according to the number of sets of stored learning data. During the learning, the LED to be additionally lit by the learning blinks, and is continuously lit after the learning is completed. The filter selection switch 7 is provided in a filter selection switch storage 35 next to the terminal block 34 in FIG.
The switch is normally housed with the lid closed, and APF, LP in the setting mode or the determination mode.
It comprises three switches for determining whether signals of three bands, F and HPF, are input to the microcomputer 1, respectively.

FIG. 4 is a state transition diagram of the vibration judging device.
The operation of the device will be described. When the power is turned on, the microcomputer 1 is reset, and various states such as a stack and a register in the microcomputer 1 are initialized to predetermined states. (1) In the operation mode check A, the mode switch SW
When (switch) is TEACH, the mode is switched to the learning mode, and when SET or RUN, the mode is switched to the determination data calculation. In the initial state, the LEDs (D1 to D5) corresponding to the number of learning data stored in the EEPROM 32 are turned on.

(2) In the learning mode (corresponding to sensing information storage), when the mode switch SW is set to TEACH, the device operates as the learning mode. In this mode, learning processing / learning data deletion is executed. These functions are set by the operator through SW input, and cannot be controlled from outside. Available SWs are TEACH-SW and CLE
AR-SW. SV, PV, PASS,
The LED of NG is turned off, and the LEDs (D1 to D5) corresponding to the number of learning data stored in the EEPROM 32 are turned on.

The learning process is executed when the TEACH-SW is pressed (turned on). When the TEACH-SW is continuously pressed, the learning process is continuously performed. That is, while the TEACH-SW is ON, a maximum of five learning processes are continuously performed up to the learning data D5. The data is D1
Learning is performed in the order from to D5. In the initial state of the learning process, P
Turn off the V-LED. In order to perform the automatic gain processing (automatic adjustment of the amplification degree of the amplification degree variable amplifier 5), the automatic gain processing is performed when learning D1, the gain level set thereafter is set as the amplifier magnification, and the value of the sensor signal value is set. The amplifier magnification is set so that the range is within a predetermined range.
During learning, the corresponding learning data LED blinks. The learning data calculation will be described later. After the learning is completed, the blinking learning data LED is turned on.

The procedure of the automatic gain processing will be described. An auto gain process is performed during learning of D1, and the gain level set thereafter is set as the amplifier magnification. next,
The amplifier magnification is set so that the half amplitude D of the sensor signal falls within a predetermined range. Regarding the gain setting method, the amplifier is switched in 32 steps to adjust the gain so that the amplitude of the APF input waveform becomes an optimum level. In order to prevent the gain setting from being affected by spontaneous vibration or noise,
Adjust the gain by averaging the absolute value of the waveform. Except for single oscillation and noise, the half amplitude of the vibration waveform is set to be in the range of 0.8 ≦ D ≦ 1.2. The average absolute value (AVE) is represented by the following equation.

(Equation 1)

The deletion process is executed when the CLEAR-SW is pressed (turned on). Until the CLEAR-SW is once turned off, the next learning processing and deletion processing are not performed. One of the currently stored learning data is deleted in the order of D5 to D1. PV-LE in the initial state of the deletion process
Turn off D. Learning L corresponding to the deleted learning data
The ED (any one of D1 to D5) is turned off. When the D1 data is deleted (that is, when all data is deleted), the setting data (SV value, gain level, etc.) is set to a default value.

(3) In the judgment data calculation (corresponding to the judgment condition generation), the mode switch SW is set to SET or RUN.
When the learning data is stored, the calculation of the determination data is executed. In the initial state, PV, PASS, N
Turn off each LED of G. If there is no learning data, the learning mode wait state is reached, and no input other than the mode switching SW is accepted. If there is learning data stored in the EEPROM 32, a determination data calculation is performed, and the SV-LED
Displays the SV value.

(4) In the operation mode confirmation B, when the mode switching SW is TEACH, the operation mode confirmations A and S
The setting mode is switched to the ET mode, and the determination mode is switched to the RUN mode. In the initial state, the PV, PASS, and NG LEDs are turned off. The LEDs (D1 to D5) corresponding to the stored learning data are turned on.

(5) If the learning data is stored and the mode switch SW is set, it operates as the setting mode.
In the setting mode, SV value setting / filter selection / judgment processing is executed. These functions are set by the operator by operating a switch, and cannot be controlled by an external signal line. Available switches are START-SW,
SV-UP-SW and SV-DOWN-SW. In the initial state, the PV, PASS, and NG LEDs are turned off. The LEDs (D1 to D5) corresponding to the learning data stored in the EEPROM 32 are turned on. SV-LED
Indicates the SV value. In the SV value setting process, the SV-UP
SV value UP processing using -SW, SV-DOWN-S
SV value DOWN processing is performed using W (corresponding to determination condition generation).

The determination process in the setting mode is performed in the STAR mode.
Executes when the T-SW is pressed. When the START-SW is continuously pressed, the determination process is continuously performed. At the start of the determination process, the LEDs of PV, PASS, and NG are turned off. In the continuous determination, the PV, PASS, and NG LEDs are not turned off at the start of the second and subsequent determination processes.
During the determination process, BUSY is turned on. After the end of the PV value calculation and the judgment calculation, the judgment result is output. The degree of separation PV is displayed on the PV-LED in 11 levels. PASS is set to ON when PV is greater than SV, and OF is set when PV exceeds SV.
F. NG is set to OFF when PV is lower than SV, PV
Turns on when SV exceeds SV. Regarding the filter selection process, in the setting mode, the contents of the FILTER-SW are always read (except during the determination and during the continuous determination), and a combination of filters to be used is selected according to the result.

(6) If the learning data is stored and the mode switch SW is RUN, the apparatus operates as a determination mode.
In the determination mode, a determination process is performed. In order to make the setting by the operator impossible, control by SW is not performed, and only control by an external signal line is performed. The determination process is almost the same as the determination process in the setting mode (both correspond to the determination).

Next, the algorithm (feature calculation and learning) used in the learning process will be described. Feature amount calculation: A feature amount is calculated from sampling data (256 continuous input waveforms), and the features of the input waveform are grasped. Four types of feature amounts are prepared by the following arithmetic expressions. By calculating the characteristic amounts for the signals from the three different filters, a total of 12 types of characteristic amounts are obtained. FIG. 5 shows the contents of the feature amount.

## EQU2 ## Range = maximum value−minimum value Average absolute value = (1 / N) Σ | x _i | Maximum PP = peak TO peak maximum value Slope average = (1 / N) Σ | x _{i + 1} -x _i | proviso X _i is the sampled data values, N is the number of sampling data.

Learning: Sampling for each filter and calculation of each feature are performed 16 times, and 16 values are obtained for each combination of the filter and the feature. The average and the deviation of the 16 values are calculated for each combination of the filter and the characteristic amount, and stored in the EEPROM 32. At this time, EEPRO
One set of data stored in M32 is learning data.
The deviation σ is calculated by the following equation.

(Equation 3)

FIG. 6 shows a memory map of the EEPROM 32. The new learning data is stored in an area next to an area in which valid learning data previously learned is stored. At the time of the deletion processing, the number of learning data stored effectively is managed. For example, if one data is deleted after learning up to data 4, the number of learning data becomes three. ~ 3 will be used. The data 4 still remains in the EEPROM, and if it is learned again, it will be overwritten on the data 4.

The algorithm used in the determination processing in the setting mode and the determination mode will be described. Judgment data calculation (corresponding to judgment condition generation): With respect to learning data (up to a maximum of five) stored effectively, the average μ calculated by the above learning for each combination of a filter and a feature amount for each learning data. _{From n} and deviation σ _n (n is from 1 to the number P of learning data stored effectively), for each combination of a filter and a feature amount, a composite average M and a composite deviation, which are the average and deviation of all learning data, S is calculated by Expression 3.

(Equation 4)

Judgment: A judgment operation is started by a START-SW (setting mode) or a START external input (judgment mode). Sampling and feature value calculation are performed four times for each filter, and four values are obtained for each combination of filter and feature value. Based on the composite average and composite deviation calculated above, the degree of separation is calculated for each sampling and for each combination of filter and feature value. Select the largest degree of separation in a set of samplings (a set of samplings consists of three samplings using each filter once each). The average value over the four sets of samplings at the maximum degree of separation is displayed as a PV value, and PASS or NG is output based on the threshold value determination with the SV value. The calculation formula and procedure of the PV value are shown below. (A) waveform sampling (3 filters), (b) calculation of each feature (4 features × 3 filters), and (c) (b) the degree of separation | (Y−M) / S | The maximum distance R is calculated.

(Equation 5)

(D) (a) to (c) are calculated four times corresponding to four sets of samplings, and the average of the obtained four Rs is calculated as P
V value.

PV = (1 / Q) (R ₁ + R ₂ +... + R _Q ) PV: Judgment result Q: Average number of times (4 times)

Next, abnormal processing will be described. Auto gain abnormality process: If the sensor signal does not fall within the predetermined range in the auto gain process in the learning mode, the learning is stopped, the blinking D1 to D5 are turned off, and the result is displayed using the PV-LED. To display a predetermined symbol.

Gain tolerance range abnormality processing: Second and subsequent times (D
2 to D5), it is confirmed that the vibration level is within the allowable range. The threshold value (permissible range) calculation method is as follows: (a) One-third of the lower limit value of the threshold value calculated in the above-described auto gain processing is set as the threshold value of the lower limit of the permissible range, and the difference between these lower limit values is calculated. W. (B)
The difference W calculated in (a) is also added to the upper limit threshold value of the auto gain process, and is set as the upper limit threshold value of the allowable range. The gain allowable range is as shown in FIG. If the sensor signal does not fall within the allowable range, the learning is stopped and D1 to D5
Are turned off, and the result is displayed in a predetermined symbol using PV-LED.

Sensor abnormality processing: Monitors sensor disconnection / short circuit. If an error is detected, TEACH-SW, ST
ART-SW and START external inputs are not accepted. If a sensor abnormality is detected during the learning process, the learning data is discarded, the learning process is stopped, and the blinking D1 to D5 are turned off. If a sensor abnormality is detected during the determination processing, the data being determined is discarded, and the determination processing is stopped. In case of sensor abnormality, ALARM is turned on, PASS, NG,
Turn off each LED of the PV. ALA if sensor is normal
Turn off the RM.

In the above-described embodiment, the function in the learning mode and the manual adjustment function of the SV value in the setting mode correspond to the "sensing information storage means" and "determination condition generating means", and the determination function in the setting mode and the determination mode corresponds to " The sensor abnormality processing and the auto gain abnormality processing correspond to the “abnormality monitoring means”, respectively.

The present invention can be variously modified without being limited to the above embodiment. For example, it is possible to determine whether or not the sensing information is abnormal in the learning process based on an arbitrary feature amount or the like.

[Brief description of the drawings]

FIG. 1 is a hardware configuration diagram of a vibration determination device according to an embodiment of the present invention.

FIG. 2 is a front view of a panel of the vibration determination device.

FIG. 3 is a bottom view thereof.

FIG. 4 is a state transition diagram of the vibration determination device.

FIG. 5 is a diagram showing the contents of a feature amount.

FIG. 6 is a memory map diagram of the EEPROM 32.

FIG. 7 is a diagram showing an allowable gain range.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 One chip microcomputer 2 Sensor 9 START switch (judgment means) 10 START external input (judgment means) 11 TEACH switch (sensing information storage means) 13 SV-UP switch (judgment condition generation means) 14 SV-DOWN switch (judgment condition) Generation means) 22 7-segment LED of PV value (output of abnormality monitoring means) 32 EEPROM (sensing information storage means)

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Akiyuki Nakayama 10 Ohanaron Dodocho, Ukyo-ku, Kyoto-shi, Kyoto

Claims (4)

[Claims] 1. A sensing information storage unit that stores a signal input from a sensor that detects a state of a detection target or sensing information that is information extracted from the signal, based on the sensing information stored in the sensing information storage unit. A determination condition generating unit that generates a condition for determining the state of the detection target; and a determination unit that determines the state of the detection target from a signal input from the sensor according to the condition generated by the determination condition generation unit. In the state determination device comprising: determining whether the sensing information is abnormal, when it is determined to be abnormal, the sensing information determined to be abnormal, the determination condition of the determination conditions by the determination condition generating means Abnormality monitoring means for preventing use for generation and outputting a signal to notify that an abnormality has occurred. And a state determination device. 2. The state determination device according to claim 1, wherein the abnormality monitoring unit determines that the abnormality is abnormal when a value included in the sensing information is not within a predetermined value range. 3. The abnormality monitoring unit according to claim 1, wherein the value included in the sensing information is a predetermined range defined based on a value included in other sensing information stored in the sensing information storage unit before then. 2. The state judging device according to claim 1, wherein when there is no such condition, it is judged that there is an abnormality. 4. The state judging device according to claim 1, wherein the abnormality monitoring unit further stops inputting a signal from the sensor as soon as it determines that the sensing information is abnormal.