JPH07122604B2 - Abnormality diagnosis method for rotating machinery - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a rotary machine that automatically determines the cause of an abnormality occurring in the rotary machine by using a mechanical or electrical signal indicating the state of the rotary machine. The present invention relates to an abnormality diagnosis method.

[Conventional technology]

In conventional abnormality diagnosis of rotating machinery, the deterioration index obtained from a plurality of features calculated from the analysis results of time series data of mechanical or electrical signals representing the state of rotating machinery is regarded as independent, and this deterioration is considered. The possibility of occurrence of each abnormality cause was estimated by combining only two linear factors, that is, the index and the weighting coefficient indicating the degree of causal relationship between the plurality of abnormality causes occurring in the rotating machine. For example, in the method disclosed in Japanese Patent Laid-Open No. 59-94018, a plurality of characteristic quantities are calculated from time series data of vibration signals obtained from a rotary machine to be subject to failure diagnosis and stored in advance in a storage unit. However, a plurality of failure causes are estimated based on a failure causal table that defines a causal relationship between the feature quantity and the failure cause, and a deterioration index indicating the degree of deterioration for each failure cause from the values of the plurality of feature quantities. In addition, the method is a method of diagnosing a failure of a rotating machine, which is converted into the following and linearly combines the deterioration index and a weighting coefficient indicating the degree of correlation between each failure cause to perform automatic diagnosis of the cause of the failure.

[Problems to be Solved by the Invention]

However, in the above-mentioned conventional method, it is possible to obtain a result that seems to have a large apparent possibility even for the cause of the abnormality that is unlikely to occur with the increase in the number of terms in the linear combination. It had the drawback of becoming difficult.

An object of the present invention is to provide a method for diagnosing an abnormality in a rotating machine that can clearly and automatically determine the cause of an abnormality occurring in the rotating machine.

[Means for Solving the Problems]

The present invention diagnoses an abnormality of a rotating machine based on a plurality of characteristic amounts calculated from various analysis results of time-series data of mechanical signals such as vibration indicating the state of the rotating machine or electric signals such as motor current. In the method described above, the matrix structure is used in which the cause of the abnormality that occurs in the rotating machine and the feature amount obtained from the analysis result of the time series data are taken as items,
An exchange coefficient group for converting the value of each feature quantity into a deterioration index indicating the degree of deterioration of the rotating machine, a weighting coefficient that defines the degree of causal relationship between each cause of abnormality and each feature quantity, and each feature quantity independently. A database having an index indicating the sex as an element is prepared in advance, and the values of a plurality of feature amounts calculated from the analysis result of the time series data of the detection signal indicating the state of the rotating machine and the database are used to A value obtained by combining a plurality of deterioration indexes, a weighting coefficient, and an index indicating the independence of the feature amount is obtained for each cause of abnormality, and the cause of abnormality is determined based on the magnitude of the value.

[Action and Example]

 The present invention will be described in detail below based on examples.

FIG. 1 is a block diagram showing a flow of abnormality diagnosis in the embodiment of the present invention. In FIG. 1, 1 is a starting point terminal, 2 is a diagnostic specification input section, 3 is a time series signal input section, 4 is a data analysis section, 5 is an abnormality cause estimation section, and 6 is a file for storing a database. , 7 are terminal terminals, respectively. In FIG. 1, the diagnostic specification input unit 2 indicates the configuration of a rotary machine such as a motor, a pump, and a speed reducer to be diagnosed, rotary element specifications such as the number of stages of the speed reducer, the number of gear teeth, and bearing specifications. Equipment specifications, vibration,
Input analysis conditions such as the type of detection signal such as current and detection position, data length, analysis frequency, etc. Time series signal input unit 3
If signal processing such as filtering is performed on a time-series detection signal input from a signal detection end (not shown), and if the input signal is an analog signal and the processing after 4 is performed digitally, , Time series data is captured by performing aliasing removal, analog / digital conversion, etc. FIG. 2 shows examples (a), (b) and (c) of vibration signals as examples of detection signals. The data analysis unit 4 performs frequency domain analysis and amplitude domain analysis of the captured time series data. As an example of the analysis, an example of the frequency analysis result of the vibration signal {signal (a) of Fig. 2} is shown in Fig. 3a, and an example of the amplitude probability density analysis result is shown in Fig. 3b. The abnormality cause estimation unit 5 calculates a plurality of feature amounts from the frequency domain analysis result and the amplitude domain analysis result of the time series data obtained by the data analysis unit 4. Next, with respect to a plurality of causes of abnormalities that are considered to occur in the rotating machine, at the time of diagnosis, the probability of occurrence of each is calculated. The database 6 stored in advance as a file is used for this calculation.

Next, a method of estimating the cause of abnormality will be described.

In the cause of anomaly estimation unit 5, the value S ₁ of the basic spectrum (d) in the frequency analysis result of the time series data illustrated in FIG.
Of the total sum of all spectrum values Ssum (S ₁ / Ssum),
The value S ₁ of the fundamental spectrum (d) and its harmonics (e),
The ratio {(S ₂ + S ₃ ) / S ₁ } to the sum of the spectral values S ₂ and S _{3 in} (f), or the maximum and effective values of the amplitude obtained from the amplitude probability density analysis results illustrated in Fig. 3b. A plurality of feature quantities such as the ratio of The database file 6 stores a database showing a causal relationship between these characteristic quantities and the cause of abnormality in the rotating machine. FIG. 5 shows an example of the structure of the database. This database is used to convert the above-mentioned plurality of characteristic amounts into a deterioration index Pij normalized to 0 to 1 for each of a plurality of possible causes of abnormality such as unbalance of rotating machinery and loosening of mounting bolts. Coefficients aij and bij (where i is the subscript corresponding to the cause of abnormality, j
Is a subscript corresponding to the feature amount), a weighting coefficient wij indicating how much each of the plurality of deterioration indices has correlation with each abnormality cause, and independence between each feature amount for each abnormality cause. It has a matrix structure with elements representing the index nij.
The coefficients aij and bij have the same units as the feature quantity xj (coefficient ai
The relationship between j and bij and the deterioration index Pij will be described later).

Here, the index nij representing independence will be described. When the existing feature quantity xk does not have any meaning for the cause of abnormality for the first time when the other feature quantity xl is significant, for example, for the looseness of the mounting bolt, the feature obtained in the example of FIG. 3a. The quantity {(S ₂ + S ₃ ) / S ₁ } has meaning only when another feature quantity (S ₁ / Ssum) is large enough [(S ₁ / Ssu
When m) is small, {(S ₂ + S ₃ ) / S ₁ } is meaningless even if it is evaluated]. In such a case, the feature value xk
[For example, as an index nik indicating the independence of {(S ₂ + S ₃ ) / S ₁ }], another feature quantity xl having the strongest correlation with xk [eg, (S ₁ / Ssu
m)] and the deterioration index Pil for the same abnormal cause i
Set the value of to be the index value. Alternatively, when another feature amount xm is considered to be independent of all other feature amounts, “1” is set as the value of the index nim.

These coefficients aij, bij, wij and index nij are simple knowledge bases determined based on the physical properties of abnormal phenomena of rotating machinery and many years of experience. Can be changed. The abnormality cause estimation unit 5 converts the plurality of feature quantities xj as described above into the deterioration index Pij by using the conversion coefficients aij and bij described above, for example, using the following relational expression (1).

If xj <aij, Pij = 0, if aij ≦ xj <bij, Pij = (xj−aij) / (bij−aij) If xj ≧ bij, Pij = 1 (1) Equation (1) The relationship is shown in FIG.

Fi that indicates the possibility of occurrence of one abnormal cause
Is calculated by synthesizing a plurality of deterioration indexes Pij, weighting factors wij corresponding to the respective Pij, and an index nij that also indicates the independence of the feature amount corresponding to the respective Pij. For example, in the case of synthesis by linear combination, it is calculated by the following equation (2).

Fi = Wi ₁ ni ₁ Pi ₁ ＋ Wi ₂ ni ₂ Pi ₂ ＋… ＋ WitnitPit (2) Here, the subscript t represents the total number of feature quantities xj, and the sum of Wi ₁ , Wi ₂ …, Wit at each i. Is 1.

Fi takes a value from 0 to 1, and the closer this value is to 1, the higher the possibility that the cause i of abnormality is occurring in the rotating machine. The closer it is to 0, the lower the possibility of occurrence. means. If the calculation of the equation (2) is performed for all the abnormal causes that are considered to occur in the rotating machine, it is easy to know which abnormal cause is likely to occur at the time of diagnosis.

〔The invention's effect〕

As described above, according to the present invention, with respect to a plurality of feature amounts calculated from the analysis result of the time series data of the detection signal representing the state of the rotating machine, each feature amount has a plurality of other causes with respect to each abnormality cause. Different index values are given depending on whether they have a correlation with the quantity (whether they are independent or not), and in order to calculate the value that indicates the possibility of occurrence of an abnormal cause, the deterioration index converted from the characteristic amount, the characteristic amount and the abnormal cause Since the weighting coefficient indicating the degree of correlation and the index indicating the independence of the feature amount are used, it is possible to accurately determine the cause of abnormality of the rotating machine.

[Brief description of drawings]

FIG. 1 is a block diagram showing a flow of abnormality diagnosis in an embodiment of the present invention, FIG. 2 is a waveform diagram showing an example of a detection signal taken from a rotating machine, and FIGS. 3a and 3b are time series data. FIG. 4 is a plan view showing an example of the analysis result of FIG. 4, FIG. 4 is a graph showing an example of the relationship between the characteristic amount and the deterioration index, and FIG. 5 is a plan view showing a database for determining an abnormality cause from the characteristic amount. 1: Start point terminal, 2: Diagnostic specification input section 3: Time series signal data input section, 4: Data analysis section 5: Abnormality cause estimation section, 6: Database file 7: End point terminal

Claims (1)

[Claims] 1. A method for diagnosing an abnormality of a rotating machine based on a plurality of feature amounts calculated from analysis results of time series data of detection signals representing the state of the rotating machine, and the cause of the abnormality occurring in the rotating machine. Takes a matrix structure with the feature amount obtained from the analysis result of the time series data as an item,
A group of conversion coefficients for converting the value of each feature quantity into a deterioration index that indicates the degree of deterioration of the rotating machine, a weighting coefficient that defines the degree of causal relationship between each cause of abnormality and each feature quantity, and independent of each feature quantity. A database having as an element an index indicating the sex is prepared in advance, and a plurality of feature value values calculated from the analysis result of the time series data of the detection signal indicating the state of the rotating machine and the database are used, respectively. It is characterized in that a value obtained by combining a plurality of deterioration indexes, a weighting coefficient, and an index indicating the independence of the feature amount is obtained for each of the abnormality causes, and the cause of the abnormality is determined by the magnitude of the value. Abnormality diagnosis method for rotating machinery.