KR101808461B1 - Method and apparatus for predicting remaining life of a machine - Google Patents

Abstract

The present invention relates to a method of predicting residual life of a machine and, more specifically, comprises: a step of receiving a signal sensed by a rotating machine; a step of determining a state of the rotating machine by learning a deterioration index from the signal; a step of calculating a deterioration value of the rotating machine from the learned deterioration index; a step of comparing the calculated deterioration value with a threshold value to determine a failure of the rotating machine; a step of comparing the deterioration value with a previous deterioration value, and correcting the deterioration value; and a step of re-learning the corrected deterioration value and the deterioration index.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for predicting remaining life of a machine,

The present invention relates to a method and apparatus for predicting residual service life of a machine.

Sudden failure or malfunction of the equipment can seriously reduce the operating efficiency of the equipment and the yield of the product. Therefore, it is necessary to diagnose the abnormality beforehand and predict the remaining service life of the machine before occurrence of the failure of such equipment. In general, in order to predict the service life of a machine, the degree of machine deterioration is monitored by monitoring the vibration state of the machine, and the remaining service life is predicted based on the monitoring data. However, the above method has a problem in that accurate prediction of the remaining service life is difficult according to the monitoring data. As shown in FIG. 2, there is no change in deterioration value at a median fault, and it is difficult to predict the remaining service life. Also, in a severe fault, the deterioration value is unstable and it is difficult to predict the remaining life of the machine

Korean Patent Publication No. 10-2015-0104459

SUMMARY OF THE INVENTION It is an object of the present invention to provide a method for estimating the remaining life of a machine by extracting a deterioration index from an input signal of a machine and correcting the deterioration value.

An object of the present invention is to provide an apparatus for predicting the remaining life of a machine by extracting a deterioration index from an input signal of a machine and correcting the deterioration value.

According to another aspect of the present invention, there is provided a method for predicting residual service life of a machine,

Receiving a signal that senses a rotating machine, determining a state of the rotating machine by learning a deterioration index from the signal, calculating a deterioration value of the rotating machine from the learned deterioration index, Comparing the numerical value with a threshold value to determine a failure of the rotating machine; comparing the deterioration value with a previous deterioration value and correcting the deterioration value; And re-learning the corrected deterioration value and the deterioration index.

The deterioration index may include RMS information, crest factor information, skewness information, peak-to-peak value, impulse value and mean value information, Or the like.

In the step of determining the state of the rotating machine, the deterioration index may be input to a classifier previously learned based on the deterioration index of the rotating machine to diagnose whether the machine is defective.

The classifier may include a support vector machine (SVM) classification model or a naive Bayes classifier.

In the step of calculating the deterioration value,

It is possible to calculate the deterioration value by analyzing the correlation of past history data of the rotating machine and selecting a deterioration index in which the deterioration index of the deterioration index increases proportionally.

The step of determining the failure of the rotating machine may include comparing the deterioration value with a threshold value to predict a remaining life of the machine when the deterioration value is greater than a threshold value, Can be corrected.

When the deterioration value is smaller than the previous deterioration value, it can be corrected to a predetermined error range or to be equal to the previous deterioration value.

The step of re-learning the corrected deterioration value and the deterioration index may include inputting the corrected deterioration value to a regression model based on the artificial neural network that has been learned in advance, to estimate the next deterioration value.

According to another aspect of the present invention, there is provided an apparatus for predicting remaining life of a machine, the apparatus comprising: an input unit for receiving a signal sensed by a rotating machine; A state determiner for determining a state of the rotating machine by learning a deterioration index from the signal; A calculation unit for calculating a deterioration value of the rotating machine from the learned deterioration index; A failure determination unit for comparing the calculated deterioration value with a threshold value to determine a failure of the rotating machine; A correction unit for comparing the deterioration value with a previous deterioration value and correcting the deterioration value; And a re-learning unit for re-learning the corrected deterioration value and the deterioration index.

The deterioration index may include RMS information, crest factor information, skewness information, peak-to-peak value, impulse value and mean value information, Or the like.

The state determiner may diagnose whether the machine is defective by inputting the deterioration index into a classifier previously learned based on the deterioration index of the rotating machine.

The pre-learned classifier may be a support vector machine (SVM) classifier or a naive Bayes classifier.

The calculation unit may analyze the correlation of the past history data of the rotating machine and calculate the deterioration index by selecting a deterioration index in which the deterioration index of the deterioration index increases proportionally.

The failure judging unit compares the deterioration value with a threshold value, judges that the deterioration value is a failure when the deterioration value is larger than the threshold value, predicts the remaining service life of the machine, and corrects the deterioration value when the deterioration value is smaller than the threshold value.

If the deterioration value is smaller than the previous deterioration value, the correction unit may correct the deterioration value to a value within a preset error range or to be equal to the previous deterioration value.

The re-learning unit may estimate the next deterioration value by inputting the corrected deterioration value into a regression model based on the artificial neural network that has been learned in advance.

According to the method and apparatus for predicting the remaining service life of a machine according to the present invention, it is possible to accurately predict the remaining service life of a machine even in a situation where it is difficult to accurately predict the degree of defect or a degree of defect.

FIG. 1 is a flowchart illustrating a method of predicting residual service life of a machine according to an embodiment of the present invention.
2 is a diagram showing a change in deterioration value according to the degree of defect according to an embodiment of the present invention.
3 is a diagram illustrating an example of degradation index features according to an embodiment of the present invention.
4 is a diagram showing an example of selecting a deterioration index showing a degree of deterioration according to an embodiment of the present invention.
5 is a diagram showing a result of the deterioration value correction according to the embodiment of the present invention.
6 is a block diagram of an apparatus for predicting remaining life of a machine according to another embodiment of the present invention.
7 is a diagram showing an example of estimating the remaining service life according to the embodiment of the present invention.

Hereinafter, some embodiments of the present invention will be described in detail with reference to exemplary drawings. In the drawings, like reference numerals are used to denote like elements throughout the drawings, even if they are shown on different drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

It is also to be understood that the terms such as " comprises, "" comprising," or "having ", as used herein, mean that a component can be implanted unless specifically stated to the contrary. But should be construed as including other elements. All terms, including technical and scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs, unless otherwise defined. Commonly used terms, such as predefined terms, should be interpreted to be consistent with the contextual meanings of the related art, and are not to be construed as ideal or overly formal, unless expressly defined to the contrary.

Hereinafter, the present invention will be described in more detail with reference to the accompanying drawings.

FIG. 1 is a flowchart illustrating a method of predicting residual service life of a machine according to an embodiment of the present invention.

As shown in FIG. 1, a method for predicting residual service life of a machine includes receiving a signal sensed by a rotating machine (S10), learning a deterioration index from the signal (S20), determining a state of the rotating machine A step S30 of determining whether the current state of the rotating machine is normal, a step S40 of calculating the deterioration value of the rotating machine, and a step of determining whether the deterioration value is larger than the threshold value A step S60 of comparing the deterioration value with a previous deterioration value, a step S90 of re-learning the corrected deterioration value and the defect information, As shown in FIG.

First, in step S10, in which a signal sensing the rotating machine is input, a vibration sensor may be provided to sense a signal of the rotating machine. At this time, any sensor capable of detecting a mechanical signal is possible, and is not limited to a vibration sensor. The vibration sensor can sense both time domain data and frequency domain data.

The time domain data refers to vibration data according to a change in time, and the frequency domain data refers to vibration data according to a frequency change.

Accordingly, when the vibration sensor senses the rotating machine, a deterioration index can be generated based on the sensed signal. In this case, the deterioration index may be determined by, for example, a mean value, a root mean square (RMS), a crest factor, a skewness, a kurtosis or a probability density function (PDF ). The crest factor is the peak value of the data divided by its rms value within a given time interval, and is sometimes referred to as the ridge factor. The skewness can indicate the direction and degree of asymmetry about the mean value in the frequency distribution of the data. Kurtosis can indicate the degree of sharpness of the data distribution.

The degradation index associated with the frequency domain may also include a peak level or an overall peak level. The frequency spectrum may also be included. This can be obtained by performing fast Fourier transformation on a vibration amount over time for a predetermined time. In addition, the frequency spectrum may be subjected to a filtering process to improve the reliability of the frequency domain data. By using band pass filtering, only data of a specific frequency range can be obtained.

When the deterioration index is generated, a step of inputting the deterioration index to a classifier previously learned in the classifier and learning is performed. (S20) At this time, the classifier is classified into a support vector machine (SVM) A naive Bayes classifier, but is not limited thereto.

(Step S30). In this case, the state of the machine means that the deteriorated state (normal or defective) is determined do. If it is determined that the current state is normal (S40), step S10, which is a step of receiving a signal sensing the rotating machine, is performed again. However, if it is determined that the current state is not normal, a step (S50) of calculating the deterioration value of the rotating machine is performed.

At this time, in order to calculate the deterioration value, the correlation of the past history data of the rotating machine is analyzed, and a deterioration index capable of indicating the degree of deterioration of the machine in the deterioration index is selected to calculate the deterioration value. S50)

Thereafter, it is determined whether or not the calculated deterioration value is larger than the threshold value (S60).

The threshold relates to the machine failure threshold and can be preset by the user. If the deterioration value is greater than the threshold value, step S70 of estimating the remaining life of the machine is performed. If the deterioration value is smaller than the threshold value, the deterioration value is compared with the previous deterioration value and corrected (S80)

The proposed technique corrects the deterioration value by taking advantage of the fact that the deterioration phenomenon can not be weakened. If the current deterioration value is smaller than the previous deterioration value, the current deterioration value is corrected to the same value as the previous deterioration value and corrected. In another embodiment of the present invention, when the present deterioration value is smaller than the previous deterioration value, the present deterioration value can be corrected within a predetermined error range.

(S90) This is done by learning the regression model based on the artificial neural network through the past history data, setting the newly corrected deterioration index and deterioration index It means learning. Then, the next regression value is estimated, the next deterioration index is estimated, and step S60 is performed again.

As described above, the present invention has an effect of accurately predicting the remaining life of the machine by correcting the deterioration value.

2 is a diagram showing a change in deterioration value according to the degree of defect.

As shown in FIG. 2, there is no change in deterioration value at a median fault, and it is difficult to predict the remaining service life. Also, in the case of a severe fault, the deterioration value is unstable, which makes it difficult to predict the remaining life of the machine. Therefore, by correcting the deterioration value, which is a method proposed by the proposed technique, the remaining service life of the machine can be accurately predicted even in the above case.

3 is a diagram illustrating an example of degradation index features according to an embodiment of the present invention.

The deterioration index can be calculated by averaging the root mean square (RMS), crest factor, skewness, kurtosis or probability density function (PDF) An impulse factor, and the like.

In addition, as shown in FIG. 3, the distribution of the deterioration indexes can be determined as a healthy state, an incipient fault, a progreesing fault, and a severe fault Do.

4 is a diagram showing an example of selecting a deterioration index showing a degree of deterioration according to an embodiment of the present invention.

In FIG. 4, a root mean square (RMS), a crest factor, a mean value, a skewness, and a crest factor are shown. However, the deterioration index is only an example, and is not limited to the deterioration index shown above.

As shown in FIG. 4, in the case of the root mean square (RMS) among the four types of deterioration indexes, the deterioration index increases proportionally as the deterioration progresses. Therefore, it can be seen that the root mean square (RMS) is an index that can best represent the deterioration state of the machine, and the deterioration index is selected in step S50.

5 is a diagram showing a result of the deterioration value correction according to the embodiment of the present invention.

5A is a diagram showing a case where the deterioration of the machine is severe and the deterioration value is unstable. FIG. 5B shows a case where the current deterioration value is smaller than the previous deterioration value, the current deterioration value is corrected by changing it to the same value as the previous deterioration value, or if the current deterioration value is smaller than the previous deterioration value, Is a graph obtained by correcting the present deterioration index.

6 is a block diagram of an apparatus for predicting remaining life of a machine according to another embodiment of the present invention.

6, the apparatus for predicting residual service life of a machine may include a sensing unit 100, a control unit 200, and an output unit 300. The control unit 200 may include a state determination unit 210, A calculation unit 220, a failure determination unit 230, a correction unit 240, and a re-learning unit 250.

First, the sensing unit 100 senses a signal of the rotating machine. The sensor used at this time is provided with a vibration sensor, but is not limited thereto. For example, the vibration sensor may be a displacement type sensor or a speed meter sensor.

In addition, the machine residual life predicting apparatus of the present invention may further include (1) an input unit (not shown). At this time, an input unit (not shown) may be provided for data transmission between the user and the control unit 200. For example, an input (not shown) may be used to input the information and diagnostic conditions of the rotating machine.

The controller 200 receives the signal sensed by the sensing unit 100 and appropriately processes the signal to predict the remaining service life of the rotating machine.

The output unit 300 may be used to receive and output the machine remaining life prediction result from the control unit 200. [

Further, a storage unit (not shown) may be further included. The storage unit (not shown) may be used to store signals sensed by the sensor, cumulative defect data, and diagnostic results.

At this time, the controller 200 may sense both the time domain data and the frequency domain data.

The time domain data refers to vibration data according to a change in time, and the frequency domain data refers to vibration data according to a frequency change.

The state determination unit 210 may generate a deterioration index by receiving a signal that the vibration sensor senses the rotating machine. In this case, the deterioration index may be determined by, for example, a mean value, a root mean square (RMS), a crest factor, a skewness, a kurtosis or a probability density function (PDF ). The crest factor is the peak value of the data divided by its rms value within a given time interval, and is sometimes referred to as the ridge factor. The skewness can indicate the direction and degree of asymmetry about the mean value in the frequency distribution of the data. Kurtosis can indicate the degree of sharpness of the data distribution.

The frequency domain data are those relating to the periodicity of the data, and may include, for example, the primary peak level or the entire peak level in the frequency spectrum. The frequency spectrum can be obtained by performing fast Fourier transformation on a vibration amount over time for a predetermined time. In addition, the frequency spectrum may be subjected to a filtering process to improve the reliability of the frequency domain data. By using band pass filtering, only data of a specific frequency range can be obtained.

After generating the deterioration index, the control unit 200 controls the state deciding unit 210 to input the deterioration index into the learned classifier and learn it. At this time, the classifier is a support vector machine (SVM) classifier or a naive Bayes classifier, but is not limited thereto.

The controller 200 inputs the deterioration index to the classifier learned by the state determiner 210 to determine the current state of the machine. At this time, the current state of the machine is to judge the deterioration state, that is, to judge whether the machine is normal or defective. Thereafter, the controller 200 controls the input unit 100 to detect the signal of the rotating machine when it is determined that the current state is normal. If it is determined that the current state is not normal, the control unit 200 controls the calculation unit 220 to calculate the deterioration value of the rotating machine.

At this time, in order to calculate the deterioration value, the controller 200 analyzes the correlation of past history data of the rotating machine, selects a deterioration index that can indicate the degree of deterioration of the machine among the deterioration indexes, .

Then, the failure determination unit 230 determines whether the calculated deterioration value is greater than a threshold value.

The threshold relates to the machine failure threshold and can be preset by the user. If the deterioration value is greater than the threshold value, the controller 200 predicts the remaining service life of the machine. If the deterioration value is less than the threshold value, the controller 200 compares the deterioration value with the previous deterioration value, (240).

At this time, the correcting unit 240 corrects the deterioration value using the fact that the deterioration phenomenon can not be weakened. If the current deterioration value is smaller than the previous deterioration value, the current deterioration value is corrected to the same value as the previous deterioration value and corrected. According to another embodiment of the present invention, when the present deterioration index value is smaller than the previous deterioration value, the correction unit 240 can correct the current deterioration value within a predetermined error range.

Then, the control unit 200 controls the re-learning unit 250 to re-learn the corrected deterioration value and the deterioration index. At this time, the re-learning unit 250 learns the regression model based on the artificial neural network through the past history data, and re-learns the newly corrected deterioration value and deterioration index. Then, the next regression value is estimated.

The control unit 200 compares the newly corrected deterioration value and deterioration index with the threshold value again in the re-learning unit 250.

As described above, the present invention has an effect of accurately predicting the remaining life of the machine by correcting the deterioration value.

7 is a diagram showing an example of estimating the remaining service life according to the embodiment of the present invention.

7 is a graph showing the degree of deterioration of the machine, the y-axis of the graph represents the deterioration value, and the x-axis represents time. In the present invention, in order to estimate the remaining service life, the remaining service life of the machine is estimated by calculating the time interval between the initial deterioration value and the final deterioration value.

While the present invention has been described in connection with what is presently considered to be practical exemplary embodiments, it is to be understood that the invention is not limited to the disclosed embodiments, but, on the contrary, It will be understood. It is therefore to be understood that the embodiments described above are illustrative in all aspects and not restrictive.

The foregoing description is merely illustrative of the technical idea of the present invention, and various changes and modifications may be made by those skilled in the art without departing from the essential characteristics of the present invention. Therefore, the embodiments disclosed in the present invention are intended to illustrate rather than limit the scope of the present invention, and the scope of the technical idea of the present invention is not limited by these embodiments. The scope of protection of the present invention should be construed according to the following claims, and all technical ideas within the scope of equivalents should be construed as falling within the scope of the present invention.

100: sensing part
200:
210:
220:
230: Fault determination unit
240:
250:
300:

Claims (16)

Receiving a signal sensing the rotating machine;
Determining a state of the rotating machine by learning a deterioration index from the signal;
Calculating a deterioration value of the rotating machine from the learned deterioration index;
Comparing the calculated deterioration value with a threshold value to determine a failure of the rotating machine;
Comparing the deterioration value with a previous deterioration value and correcting the deterioration value; And
Re-learning the corrected deterioration value and the deterioration index,
Wherein the correcting step corrects the deterioration value within a predetermined error range when the deterioration value is smaller than the previous deterioration value or predicts the remaining service life of the machine that corrects the deterioration value by changing the deterioration value to the same value as the previous deterioration value Method for predicting machine residual life.
The method according to claim 1,
The deterioration index,
Peak value, an impulse value, and an average value of the RMS information, the crest factor information, the skewness information, the peak-to-peak value, the impulse value, The remaining life of the machine is predicted.
The method according to claim 1,
The step of determining the state of the rotating machine includes:
Wherein the deterioration index is input to a classifier previously learned based on a deterioration index of the rotating machine to diagnose whether or not the machine is defective.
The method of claim 3,
Wherein the pre-learned classifier is a support vector machine (SVM) classifier or a naive Bayes classifier.
The method according to claim 1,
In the step of calculating the deterioration value,
Wherein the deterioration index is calculated by analyzing a correlation between past history data of the rotating machine and a deterioration index in which the deterioration index of the deterioration index increases proportionally to calculate the deterioration value.
The method according to claim 1,
The step of determining the failure of the rotating machine includes:
Comparing the deterioration value with a threshold value; predicting a remaining life of the machine by determining that the deterioration value is a failure when the deterioration value is greater than a threshold value; and correcting the deterioration value when the deterioration value is less than the threshold value.
delete The method according to claim 1,
Wherein the step of re-learning the corrected deterioration value and the deterioration index comprises:
And inputting the corrected deterioration value to a regression model based on a learned neural network to estimate a next deterioration value.
An input unit for receiving a signal sensed by the rotating machine;
A state determiner for determining a state of the rotating machine by learning a deterioration index from the signal;
A calculation unit for calculating a deterioration value of the rotating machine from the learned deterioration index;
A failure determination unit for comparing the calculated deterioration value with a threshold value to determine a failure of the rotating machine;
A correction unit for comparing the deterioration value with a previous deterioration value and correcting the deterioration value;
And a re-learning unit for re-learning the corrected deterioration value and the deterioration index,
Wherein the correcting unit corrects the deterioration value within a predetermined error range when the deterioration value is smaller than the previous deterioration value or corrects the deterioration value by changing the deterioration value to the same value as the previous deterioration value.
10. The method of claim 9,
The deterioration index,
Peak value, an impulse value, and an average value of the RMS information, the crest factor information, the skewness information, the peak-to-peak value, the impulse value, And the remaining life of the machine is predicted.
10. The method of claim 9,
The state determination unit may determine,
Wherein the degradation index is input to a classifier previously learned based on a deterioration index of the rotating machine to diagnose whether the machine is defective.
12. The method of claim 11,
Wherein the pre-
An apparatus for predicting residual life of a machine that is a support vector machine (SVM) classifier or a naive Bayes classifier.
10. The method of claim 9,
The calculation unit may calculate,
Wherein the deterioration index is calculated by analyzing a correlation between past history data of the rotating machine and a deterioration index in which the deterioration index of the deterioration index increases proportionally.
10. The method of claim 9,
Wherein the failure determination unit
Comparing the deterioration value with a threshold value and predicting a remaining service life of the machine when the deterioration value is greater than a threshold value and determining the failure and correcting the deterioration value when the deterioration value is less than the threshold value.
delete 10. The method of claim 9,
The re-
And inputting the corrected deterioration value to a regression model based on a previously learned artificial neural network to estimate a next deterioration value.

Images