Patent Application
20230288482
The present disclosure relates to a diagnosis apparatus, a diagnosis method, and a diagnosis program for a rotary machine.
The present application claims priority based on Japanese Patent Application No. 2020-130180 filed Jul. 31, 2020, the entire content of which is incorporated herein by reference.
It has been proposed to detect an abnormality in a rotary machine on the basis of a current value measured during rotation of the rotary machine.
For example, Patent Document 1 discloses a diagnosis apparatus for diagnosing a machine including a rotary machine on the basis of current measured during rotation of the rotary machine. In this diagnosis apparatus, an abnormality in the machine is detected by comparing the distribution of current effective values acquired from the measured current with the distribution of current effective values acquired from current measured during normal operation of the rotary machine.
Depending on the properties of the rotary machine and the type of abnormality, even when an abnormality occurs in the rotary machine, there may not be much effect on the distribution of feature (e.g., current effective value) obtained from the measured current. Therefore, if abnormality detection of the rotary machine is based only on the distribution of one feature (current effective value in Patent Document 1) obtained from the measured current as described in Patent Document 1, it may not be possible to detect an abnormality appropriately depending on the properties of the rotary machine and the type of abnormality to be detected.
In view of the above, an object of at least one embodiment of the present invention is to provide a diagnosis apparatus, a diagnosis method, and a diagnosis program for a rotary machine whereby it is possible to detect an abnormality in the rotary machine appropriately.
A diagnosis apparatus for a rotary machine according to at least one embodiment of the present invention includes: a feature acquisition part configured to acquire, from a current waveform of a current measured during rotation of a rotary machine including a motor or a generator, a plurality of features each representing a characteristic of the current; and an abnormality determination part configured to determine whether there is an abnormality in the rotary machine on the basis of a divergence between a distribution of each of the plurality of features or a multi-dimensional distribution of the plurality of features and a reference distribution of each of the plurality of features or a reference multi-dimensional distribution during normal operation of the rotary machine.
Further, a diagnosis method for a rotary machine according to at least one embodiment of the present invention includes: a step of acquiring, from a current waveform of a current measured during rotation of a rotary machine including a motor or a generator, a plurality of features each representing a characteristic of the current; and a step of determining whether there is an abnormality in the rotary machine on the basis of a divergence between a distribution of each of the plurality of features or a multi-dimensional distribution of the plurality of features and a reference distribution of each of the plurality of features or a reference multi-dimensional distribution during normal operation of the rotary machine.
Further, a diagnosis program for a rotary machine according to at least one embodiment of the present invention is configured to cause a compute to execute: a process of acquiring, from a current waveform of a current measured during rotation of a rotary machine including a motor or a generator, a plurality of features each representing a characteristic of the current; and a process of determining whether there is an abnormality in the rotary machine on the basis of a divergence between a distribution of each of the plurality of features or a multi-dimensional distribution of the plurality of features and a reference distribution of each of the plurality of features or a reference multi-dimensional distribution during normal operation of the rotary machine.
At least one embodiment of the present invention provides a diagnosis apparatus, a diagnosis method, and a diagnosis program for a rotary machine whereby it is possible to detect an abnormality in the rotary machine appropriately.
Embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It is intended, however, that unless particularly identified, dimensions, materials, shapes, relative positions, and the like of components described in the embodiments shall be interpreted as illustrative only and not intended to limit the scope of the present invention.
(Configuration of Diagnosis Apparatus)
In some embodiments, the rotary machine to be diagnosed includes a motor. A rotary machine 1 shown in
The motor 4 may be configured to be driven by AC power. In the exemplary embodiment shown in
In some embodiments, the rotary machine to be diagnosed includes a generator. Such a rotary machine may include, for example, a turbine configured to be driven by a fluid and a generator configured to be driven by the turbine. The generator may be configured to generate AC power.
A diagnosis apparatus 20 is configured to diagnose the rotary machine 1 on the basis of a current measured by a current measurement part 10 during rotation of the rotary machine 1.
The current measurement part 10 is configured to measure a current supplied to the motor (for example, motor 4 in
The diagnosis apparatus 20 is configured to receive a signal indicating a current measurement value from the current measurement part 10. The diagnosis apparatus 20 may be configured to receive a signal indicating a current measurement value from the current measurement part 10 at a specified sampling period. Further, the diagnosis apparatus 20 is configured to process the signal received from the current measurement part 10 and determine whether there is an abnormality in the rotary machine 1. The diagnosis result by the diagnosis apparatus 20 may be displayed on a display part 40 (e.g., display; see
An abnormality in the rotary machine 1 to be diagnosed by the diagnosis apparatus is an abnormality in the rotary machine 1 that can affect the current measurement value from the current measurement part 10. Examples of such abnormalities include misalignment (center deviation), cavitation, belt loosening, and ground faults in the rotary machine 1.
As shown in
The diagnosis apparatus 20 includes a calculator equipped with a processor (e.g., CPU), a storage device (memory device; e.g., RAM), an auxiliary storage part, and an interface. The diagnosis apparatus 20 receives a signal indicating a current measurement value from the current measurement part 10 via the interface. The processor is configured to process the signal thus received. In addition, the processor is configured to process programs loaded into the storage device. Thereby, the function of each functional unit (current waveform acquisition part 22, etc.) is implemented.
The processing contents in the diagnosis apparatus 20 may be implemented as programs executed by the processor. The programs may be stored in the auxiliary memory. When executed, these programs are loaded into the storage device. The processor reads out the programs from the storage device to execute instructions included in the programs, respectively.
The current waveform acquisition part 22 is configured to acquire a current waveform 110 (see
The feature acquisition part 23 is configured to acquire a plurality of features (feature values) each representing a characteristic of the measured current from the current waveform 110 acquired by the current waveform acquisition part 22. The feature acquisition part 23 may be configured to acquire an effective value of the current for each of divided waveforms acquired by the divided waveform acquisition part 32, which will be described later.
The features of the current acquired by the feature acquisition part 23 may be, for example, a difference between maximum and minimum values of the current in the current waveform 110 (or in a divided waveform acquired from the current waveform) acquired by the current waveform acquisition part 22, an effective value of the current (the square root of the mean of the squares), an average value of the current (the mean of absolute values), a skewness of the current (the third-order moment around the mean normalized (divided) by the standard deviation cubed), or a crest factor (maximum value/effective value) of the current.
The feature acquisition part 23 may be configured to two or more of the above-described multiple types of features from the current waveform 110 as the plurality of features. In this case, the combination of two or more features may be for example, but not limited to, a combination of the effective value and the crest factor.
Alternatively, when the rotary machine 1 includes a three-phase motor or a three-phase generator, the feature acquisition part 23 may be configured to acquire, as the plurality of features, one or more features for each of the three-phase currents (winding currents) of the three-phase motor or the three-phase generator. In this case, the type of one or more features may be for example, but not limited to, the effective value.
The distribution acquisition part 25 is configured to calculate a distribution of each of the plurality of features acquired by the feature acquisition part 23 or a multi-dimensional distribution of the plurality of features. The multi-dimensional distribution of two features is a two-dimensional distribution.
The distribution of each of the plurality of features acquired by the distribution acquisition part 25 may be a probability distribution of each of the plurality of features. The multi-dimensional distribution of the plurality of features acquired by the distribution acquisition part 25 may be a multi-dimensional probability distribution of the plurality of features.
The reference distribution acquisition part 27 is configured to acquire a reference distribution of each of the plurality of features (the same features as those of the distributions acquired by the distribution acquisition part 25) during normal operation of the rotary machine 1 or a reference multi-dimensional distribution of the plurality of features. The reference distribution or the reference multi-dimensional distribution acquired by the reference distribution acquisition part 27 is acquired in advance during normal operation of the rotary machine 1 (when no abnormality has occurred). The reference distribution or the reference multi-dimensional distribution may be stored in the storage part 12 (see
The reference distribution of each of the plurality of features acquired by the reference distribution acquisition part 27 may be a probability distribution (reference probability distribution) of each of the plurality of features. Further, the reference multi-dimensional distribution of the plurality of features acquired by the reference distribution acquisition part 27 may be a multi-dimensional probability distribution (reference multi-dimensional probability distribution) of the plurality of features.
The divergence calculation part 29 is configured to acquire a divergence between each distribution or the multi-dimensional distribution calculated by the distribution acquisition part 25 and each reference distribution or the reference multi-dimensional distribution acquired by the reference distribution acquisition part 27.
The abnormality determination part 30 is configured to determine whether there is an abnormality in the rotary machine 1 (that is, determine the presence or absence of an abnormality in the rotary machine 1) on the basis of the divergence acquired by the divergence calculation part 29.
In some embodiments, the divergence calculation part 29 may be configured to calculate, as the above-described divergence, a distance between the probability distribution of each of the plurality of features calculated by the distribution acquisition part 25 and the reference probability distribution of each of the plurality of features during the normal operation acquired by the reference distribution acquisition part 27. Further, the abnormality determination part 30 may determine whether there is an abnormality in the rotary machine 1 on the basis of the plurality of distances thus calculated. The above-described distance is an index value that can quantify the difference between two probability distributions (probability density functions), and may be a Kullback-Leibler distance, a Pearson distance, a relative Pearson distance, or a L2 distance between the probability distribution of a certain feature and the reference probability distribution of the same feature.
In an embodiment, the abnormality determination part 30 may determine whether there is an abnormality in the rotary machine 1, using the largest one of the plurality of distances calculated (i.e., respective distances of the distributions of the plurality of features). For example, the abnormality determination part may be configured to determine that an abnormality has occurred in the rotary machine 1 when the largest one of the calculated distances is not less than a threshold, and may be determine that the rotary machine 1 is normal (no abnormality has occurred) when the largest one is less than the threshold.
In some embodiments, the divergence calculation part 29 may be configured to calculate, as the above-described divergence, a distance between the multi-dimensional probability distribution of the plurality of features calculated by the distribution acquisition part and the reference multi-dimensional probability distribution of the plurality of features during the normal operation acquired by the reference distribution acquisition part 27. Further, the abnormality determination part 30 may determine whether there is an abnormality in the rotary machine 1 on the basis of the distance thus calculated. For example, the abnormality determination part 30 may be configured to determine that an abnormality has occurred in the rotary machine 1 when the calculated distance is not less than a threshold, and may be determine that the rotary machine 1 is normal (no abnormality has occurred) when the distance is less than the threshold. The above-described distance is an index value that can quantify the difference between two probability distributions (probability density functions), and may be a Kullback-Leibler distance, a Pearson distance, a relative Pearson distance, or a L2 distance between the multi-dimensional probability distribution and the reference multi-dimensional probability distribution of the plurality of features.
The divided waveform acquisition part 32 is configured to acquire a plurality of divided waveforms 112 by dividing the current waveform 110 acquired by the current waveform acquisition part 22 by a specified number of pulses (see
The filter 34 is a filter for reducing noise components (high frequency components) from the signal received from the current measurement part 10. The filter setting part 36 is configured to be able to change settings such as the time constant of the filter 34.
According to the findings of the present inventors, when an abnormality occurs in the rotary machine 1, the magnitude of the effect on each of the distributions of the plurality of features that can be acquired from the measured current varies depending on the properties of the rotary machine 1 and the type of abnormality. In this regard, the diagnosis apparatus 20 according to the above-described embodiments determines whether there is an abnormality in the rotary machine 1 on the basis of the divergence between the distribution of each of the plurality of features acquired from the current waveform 110 of the measured current and the reference distribution of each of the plurality of features or the divergence between the multi-dimensional distribution of the plurality of features and the reference multi-dimensional distribution of the plurality of features. Therefore, as compared to the abnormality determination based on the divergence between the distribution and the reference distribution of a single feature, it is possible to detect an abnormality more exhaustively for the characteristics of the rotary machine 1 and the types of abnormality. Thus, it is possible to detect an abnormality in the rotary machine 1 more appropriately.
(Diagnosis Flow of Rotary Machine)
Hereinafter, the diagnosis flow for a rotary machine according to an embodiment will be described more specifically. The following describes the case where the above-described diagnosis apparatus 20 is used to execute a diagnosis method for a rotary machine according to an embodiment, but in some embodiments, another apparatus may be used to execute the diagnosis method for a rotary machine.
In the embodiment shown in
Then, a current waveform 110 representing a change in measured current value over time is acquired by the current waveform acquisition part 22 on the basis of a signal received from the current measurement part 10 (signal indicating a current measurement value) (S4). Here,
Then, the current waveform 110 acquired in step S4 is divided by a specified number of pulses to acquire a plurality of divided waveforms 112 by the divided waveform acquisition part 32 (S6). In step S6, the plurality of divided waveforms 112 (divided waveforms with one pulse; see
The following describes the case where, in step S6, the current waveform 110 is divided by one pulse to acquire the plurality of divided waveforms 112. However, the following description can also be applied to the case where the current waveform 110 is divided by every two or more pulses to acquire divided waveforms.
Then, for each of the divided waveforms 112 obtained in step S6, a plurality of features each representing a characteristic of the measured current is acquired by the feature acquisition part 23 (S8). The feature acquisition part 23 may be configured to acquire a plurality of features for each of the divided waveforms acquired by the divided waveform acquisition part 32, which will be described later. Here, as an example, the effective value, which is the first feature, and the crest factor, which is the second feature, are acquired as the plurality of features.
Here, the effective value Irms of the current of each divided waveform 112 can be calculated as the square root of the mean (time mean) of the squares of current measurement values I of each divided waveform 112. If the current measurement value is obtained at a specified sampling period, the effective value Irms of the current of the divided waveform 112 can be expressed by the following equation (A), using the current values It at multiple measurement points in each divided waveform 112 and the time length T from the start point to the end point of each divided waveform 112.
Further, the crest factor Ief of the current of each divided waveform 112 can be calculated as a ratio of the maximum value Imax to the effective value Irms of the current measurement values I of each divided waveform 112. That is, the crest factor Ief can be expressed by the following equation.
I
ef
=I
max
/I
rms (B)
Then, a distribution of each of the plurality of features (effective value Irms and crest factor Ief) is acquired by the distribution acquisition part 25 for the plurality of divided waveforms 112 acquired in step S8. Here, a probability distribution of effective values Irms of the plurality of divided waveforms 112 obtained in step S8 and a probability distribution of crest factors Ief of the plurality of divided waveforms 112 obtained in step S8 are acquired.
In step S10, for example, a probability distribution shown by the curve 102 is obtained as the probability distribution of the effective value of the measured current. The curve 100 in
Although not depicted, a probability distribution of the crest factor of the measured current is also acquired in the same manner in step S10.
Then, a reference distribution, which is a distribution of each of the features of the measured current during normal operation of the rotary machine 1, is acquired by the reference distribution acquisition part 27. Here, a reference probability distribution of the effective value Irms and a reference probability distribution of the crest factor Ief are acquired. The reference distributions (e.g., reference probability distributions) of the effective value and crest factor may be, for example, acquired in advance and stored in the storage part 12. The reference distribution acquisition part 27 may acquire the reference distributions by reading the reference distribution of the effective value and the reference distribution of the crest factor stored in the storage part 12. The curve 100 in
Then, a distance between the probability distribution of each of the plurality of features calculated by the distribution acquisition part 25 and the reference probability distribution of each of the plurality of features during the normal operation acquired by the reference distribution acquisition part 27 is calculated by the divergence calculation part 29 (S14). Here, the relative Pearson distance is calculated as the distance. That is, the relative Pearson distance D1 between the probability distribution and the reference probability distribution regarding the effective value and the relative Pearson distance D2 between the probability distribution and the reference probability distribution regarding the crest factor are calculated.
When the reference probability distribution is p(x), and the probability distribution is p′(x), the relative Pearson distance between the probability distribution and the reference probability distribution can be calculated, for example, by ∫qα(x)[{p(x)/qα(x)}−1]2dx, where qα =αp+(1−α)p′ (0≤α<1).
Then, using the plurality of distances (i.e., the relative Pearson distance D1 regarding the effective value and relative Pearson distance D2 regarding the crest factor) calculated in step S14, it is determined by the abnormality determination part 30 whether there is an abnormality in the rotary machine 1 (S16). In step S16, the largest one of the plurality of distances may be used for abnormality determination of the rotary machine 1.
For example, of the two distances D1 and D2, if the relative Pearson distance D1 regarding the effective value is the larger, the relative Pearson distance D1 regarding the effective value is used to determine whether there is an abnormality in the rotary machine 1. If the relative Pearson distance D1 is not less than a preset threshold (Yes in S16), it is determined that an abnormality has occurred in the rotary machine 1 (S18). Conversely, if the relative Pearson distance D1 is less than the threshold (No in S16), it is determined that the rotary machine 1 is normal (no abnormality has occurred) (S20).
The determination results in steps S18 and S20 may be displayed on the display part (S22).
As already described, when an abnormality occurs in the rotary machine 1, the magnitude of the effect on each of the distributions of the plurality of features that can be acquired from the measured current varies depending on the properties of the rotary machine 1 and the type of abnormality. Further, the larger the distance between the probability distribution of each of the plurality of features and the reference probability distribution of each of the plurality of features, the higher the possibility that an abnormality occurs in the rotary machine (the degree of abnormality of the rotary machine 1). In this regard, according to the above-described embodiment, an abnormality in the rotary machine 1 can be detected appropriately on the basis of the largest one (e.g., relative Pearson distance D1 regarding the effective value) of the plurality of distances acquired (the above-described relative Pearson distances D1 and D2).
Next, an embodiment shown in
In the embodiment shown in
In step S40, for example, a multi-dimensional probability distribution shown in
Then, a reference multi-dimensional distribution, which is a distribution of the features of the measured current during normal operation of the rotary machine 1, is acquired by the reference distribution acquisition part 27. Here, a reference multi-dimensional probability distribution of the effective value Irms and the crest factor Ief is acquired. The reference multi-dimensional distribution (e.g., reference multi-dimensional probability distribution) of the effective value and the crest factor may be, for example, acquired in advance and stored in the storage part 12. The reference distribution acquisition part 27 may acquire the reference multi-dimensional distribution by reading the reference multi-dimensional distribution of the effective value and the crest factor stored in the storage part 12.
Then, a distance between the multi-dimensional probability distribution of the plurality of features calculated by the distribution acquisition part 25 and the reference multi-dimensional probability distribution of the plurality of features during the normal operation acquired by the reference distribution acquisition part 27 is calculated by the divergence calculation part 29 (S44). Here, the relative Pearson distance is calculated as the distance.
That is, the relative Pearson distance Dm between the multi-dimensional probability distribution and the reference multi-dimensional probability distribution regarding the effective value and the crest factor is calculated.
Then, using the distance (i.e., the relative Pearson distance Dm regarding the effective value and the crest factor) calculated in step S44, it is determined by the abnormality determination part 30 whether there is an abnormality in the rotary machine 1 (S46). If the relative Pearson distance Dm is not less than a preset threshold (Yes in S46), it is determined that an abnormality has occurred in the rotary machine 1 (S48). Conversely, if the relative Pearson distance Dm is less than the threshold (No in S46), it is determined that the rotary machine 1 is normal (no abnormality has occurred) (S50).
In the above-described embodiment, one value (e.g., the relative Pearson distance Dm) indicating the divergence is calculated for the plurality of features (e.g., the effective value and the crest factor). Therefore, the single index thus calculated is used to determine whether the rotary machine 1 is normal or abnormal, which facilitates the abnormality determination of the rotary machine 1.
Here,
For simplicity of explanation, in
In the multi-dimensional probability distribution when the rotary machine 1 is normal shown in
On the other hand, it is possible that, although there is a difference in the multi-dimensional probability distribution regarding a plurality of features between normal and abnormal operations of the rotary machine 1, there is no difference in the probability distribution regarding a single feature. For example, there is no difference in the probability distribution regarding the effective value (single feature) shown in
Thus, even if an abnormality occurs in the rotary machine 1, the distance between the distribution (e.g., probability distribution) and the reference distribution (e.g., reference probability distribution) can be zero when focusing on only a single feature. The use of the distance calculated in this way may not be sufficient to appropriately detect an abnormality in the rotary machine 1.
In this regard, in the embodiment described with reference to
In some embodiments, the diagnosis method according to the flowchart shown in
That is, in this case, in step S32, current of each of the three phases of the three-phase motor or three-phase generator is measured. As a result, current measurement values for three phases are obtained. In steps S34 and S36, current waveforms and divided waveforms are acquired for each of the three-phase currents. In step S38, one or more features (e.g., effective values) are acquired for each of the three-phase currents as the plurality of features. In step S40, a multi-dimensional distribution of the features (e.g., effective values) of the three-phase currents is acquired. When one feature is used, the multi-dimensional distribution is a three-dimensional distribution. In step S42, a reference multi-dimensional distribution of the features (e.g., effective values) of the three-phase currents is acquired. In step S44, the divergence (distance) between the multi-dimensional distribution and the reference multi-dimensional distribution is calculated. Then, in steps S46 to S50, it is determined whether there is an abnormality in the rotary machine 1 on the basis of the divergence.
According to the above-described embodiment, since features corresponding to each of the three-phase currents of the three-phase motor or three-phase generator are acquired as the plurality of features and the distance between the multi-dimensional probability distribution and the reference multi-dimensional probability distribution of these features is acquired, it is possible to detect an abnormality in the rotary machine 1 including the three-phase motor or three-phase generator appropriately on the basis of the distance acquired.
In some embodiments, in steps S6, S36, the divided waveform acquisition part 32 may acquire a plurality of divided waveforms 112 by dividing the current waveform 110 acquired in step S4 at a plurality of zero-crossing points ZP (e.g., ZP0 to ZP3 in
In the case of the current waveform 110 shown in
When dividing the current waveform 110, it is conceivable to divide the current waveform by a specified frequency (such as frequency associated with the rotation speed of the rotary machine), but in this case, the number of samples per period may not be stable, depending on the sampling interval of the measurement device or the like. In this regard, according to the above-described embodiment, the current waveform 110 is divided at the zero-crossing points. Thus, it is possible to obtain a plurality of divided waveforms 112 whose current values are zero at the start point (zero-crossing point) and the end point (i.e., zero-crossing point). Therefore, for each of the plurality of divided waveforms 112 thus obtained, it is possible to acquire the plurality of features appropriately in steps S8, S38.
In the example shown in
As described above, current measurement values may be acquired as discrete measurement values at each predetermined sampling period. In this regard, in the above-described embodiment, the zero-crossing points ZP can be identified by linear interpolation of two measurement values with different signs (e.g., PA and PB) among the plurality of current measurement values acquired at the specified sampling period Ts. Thus, even if the plurality of discrete current measurement values does not include a measurement point with zero current value, the current waveform 110 can be divided into the divided waveforms 112 appropriately.
In some embodiments, in steps S4, S34, the current waveform acquisition part 22 may reduce noise components (high frequency components) from the signal received from the current measurement part 10 (signal indicating a current measurement value) with a filter 34 to acquire the current waveform 110. In an embodiment, in steps S6, S36, the divided waveform acquisition part 32 may identify the zero-crossing points ZP from the current waveform 110 obtained on the basis of the signal processed by the filter 34.
In a current waveform obtained from a signal containing noise, points with zero current value may randomly appear in addition to the inherent (i.e., noise-free) zero-crossing points ZP due to waveform disturbance caused by noise. In this regard, in the above-described embodiment, since the zero-crossing points ZP are identified on the basis of the signal from which noise components have been reduced by the filter 34, the divided waveforms 112 can be obtained by dividing the current waveform 110 more appropriately on the basis of the zero-crossing points ZP thus identified.
As shown in
Then, the number of current measurement points (number of samples) included between each zero-crossing point of the plurality of zero-crossing points ZP is acquired (S106). Further, the maximum and minimum values of the number of current measurement points included between each zero-crossing point are acquired (S108).
Then, it is determined whether the difference between the maximum and minimum values obtained in step S108 is within an allowable range (S110). If the difference is outside the allowable range (No in S110), the filter setting part 36 increases the time constant of the filter 34 (S112) and returns to step S102. Then, steps S102 to S108 are repeated using the filter 34 with the new time constant set.
On the other hand, if the difference is within the allowable range in step S110 (Yes in S110), a plurality of divided waveforms are obtained on the basis of the current waveform 110 and the zero-crossing points ZP obtained in the last steps S102 and S104 (S114, S6 in
Here,
The current waveform shown in
In this case, there is a large variation in the time length between zero-crossing points (time length of waveforms 1 to 5 in
Here,
As described above, when the signal contains noise, points with zero current value appear randomly in addition to the inherent zero-crossing points ZP. For this reason, the divided waveforms obtained on the basis of such apparent zero-crossing points zp may have large variations in the length from the start point to the end point (period of divided waveform) and the number of samples.
In this regard, according to the above-described embodiment, the filter setting part 36 increases the time constant of the filter 34 until the difference between the maximum and minimum values of the number of sampling measurement values of the current included in each of the divided waveforms (or between a pair of zero-crossing points in the current waveform 110) falls within the allowable range. Thus, it is possible to reduce variation in the number of sampling current measurement values included in the divided waveforms 112 obtained on the basis of the zero-crossing points ZP from the signal processed by the filter 34. Thus, it is possible to obtain the divided waveforms by dividing the current waveform 110 more appropriately.
In an embodiment, the filter setting part 36 may be configured to repeatedly increase the time constant by a predetermined amount until the difference in the number of sampling measurement values of the current included in each of the divided waveforms (or between a pair of zero-crossing points in the current waveform 110) falls within the allowable range. That is, in an embodiment, in step S112, the time constant of the filter 34 may be increased by a predetermined amount. In this case, the time constant of the filter 34 increases in proportion to the number of loops in steps S102 to S110.
According to the above-described embodiment, since the time constant is repeatedly increased by the predetermined amount until the difference between maximum and minimum values of the number of sampling measurement values of the current included in the divided waveforms (or between a pair of zero-crossing points in the current waveform) falls within the allowable range, it is possible to reliably reduce variation in the number of sampling current measurement values included in the divided waveforms 112 obtained on the basis of the zero-crossing points ZP from the signal processed by the filter 34. Thus, it is possible to obtain the divided waveforms 112 by dividing the current waveform 110 more appropriately.
The contents described in the above embodiments would be understood as follows, for instance.
(1) A diagnosis device (20) for a rotary machine (1) according to at least one embodiment of the present invention includes: a feature acquisition part (23) configured to acquire, from a current waveform of a current measured during rotation of a rotary machine including a motor (4) or a generator, a plurality of features each representing a characteristic of the current; and an abnormality determination part (30) configured to determine whether there is an abnormality in the rotary machine on the basis of a divergence between a distribution of each of the plurality of features or a multi-dimensional distribution of the plurality of features and a reference distribution of each of the plurality of features or a reference multi-dimensional distribution during normal operation of the rotary machine.
According to the findings of the present inventors, when an abnormality occurs in the rotary machine, the magnitude of the effect on each of the distributions of the plurality of features that can be acquired from the measured current varies depending on the properties of the rotary machine and the type of abnormality. In this regard, with the above configuration (1), it is determined whether there is an abnormality in the rotary machine on the basis of the divergence between the distribution of each of the plurality of features acquired from the current waveform of the measured current and the reference distribution of each of the plurality of features or the divergence between the multi-dimensional distribution of the plurality of features and the reference multi-dimensional distribution of the plurality of features. Therefore, as compared to the abnormality determination based on the divergence between the distribution and the reference distribution of a single feature, it is possible to detect an abnormality more exhaustively for the characteristics of the rotary machine and the types of abnormality. Thus, it is possible to detect an abnormality in the rotary machine more appropriately.
Additionally, in the above configuration (1), when using the multi-dimensional distribution and the reference multi-dimensional distribution of the plurality of features, one value indicating the divergence is calculated for the plurality of features. Therefore, the single index thus calculated is used to determine whether the rotary machine is normal or abnormal, which facilitates the abnormality determination of the rotary machine. Additionally, when using the multi-dimensional distribution and the reference multi-dimensional distribution of the plurality of features, since the divergence is calculated in relation to the plurality of features, changes in the distribution of the plurality of features when an abnormality occurs in the rotary machine can be grasped in more detail, compared to the case where the divergence is calculated for the distribution of a single feature. Therefore, it is possible to improve the abnormality detection performance of the rotary machine.
(2) In some embodiments, in the above configuration (1), the abnormality determination part is configured to acquire a distance between a probability distribution of each of the plurality of features and a reference probability distribution of each of the plurality of features during the normal operation, and determine whether there is an abnormality in the rotary machine on the basis of the plurality of distances acquired.
With the above configuration (2), since the distance between the probability distribution of each of the plurality of features and the reference probability distribution of each of the plurality of features is acquired as indicating the divergence between the distribution of each of the plurality of features and the reference distribution of each of the plurality of features during the normal operation of the rotary machine, it is possible to detect an abnormality in the rotary machine appropriately on the basis of the plurality of distances acquired.
(3) In some embodiments, in the above configuration (2), the abnormality determination part is configured to determine whether there is an abnormality in the rotary machine, using the largest one of the plurality of distances.
The larger the distance between the probability distribution of each of the plurality of features and the reference probability distribution of each of the plurality of features, the higher the possibility that an abnormality occurs in the rotary machine (hereinafter, also referred to as the degree of abnormality of the rotary machine). In this regard, with the above configuration (3), an abnormality in the rotary machine can be detected appropriately on the basis of the largest one of the plurality of distances acquired.
(4) In some embodiments, in the above configuration (1), the abnormality determination part is configured to acquire a distance between a multi-dimensional probability distribution of the plurality of features and a reference multi-dimensional probability distribution of the plurality of features during the normal operation, and determine whether there is an abnormality in the rotary machine on the basis of the distance acquired.
With the above configuration (4), since the distance between the multi-dimensional probability distribution of the plurality of features and the reference multi-dimensional probability distribution of the plurality of features is acquired as indicating the divergence between the multi-dimensional distribution of the plurality of features and the reference multi-dimensional distribution of the plurality of features during the normal operation of the rotary machine, it is possible to detect an abnormality in the rotary machine appropriately on the basis of the distance acquired.
(5) In some embodiments, in the above configuration (4), the rotary machine includes a three-phase motor or a three-phase generator, and the feature acquisition part is configured to acquire, as the plurality of features, one or more features corresponding to each of the three-phase currents of the three-phase motor or the three-phase generator.
With the above configuration (5), since features corresponding to each of the three-phase currents of the three-phase motor or three-phase generator are acquired as the plurality of features and the distance between the multi-dimensional probability distribution and the reference multi-dimensional probability distribution of these features is acquired, it is possible to detect an abnormality in the rotary machine including the three-phase motor or three-phase generator appropriately on the basis of the distance acquired.
(6) In some embodiments, in any one of the above configurations (2) to (5), the distance includes a Kullback-Leibler distance, a Pearson distance, a relative Pearson distance, or a L2 distance.
With the above configuration (6), since the Kullback-Leibler distance, the Pearson distance, the relative Pearson distance, or the L2 distance is acquired as the distance between the probability distribution of each of the plurality of features and the reference probability distribution of each of the plurality of features or the distance between the multi-dimensional probability distribution of the plurality of features and the reference multi-dimensional probability distribution of the plurality of features, it is possible to detect an abnormality in the rotary machine appropriately on the basis of the distance acquired.
(7) In some embodiments, in any one of the above configurations (1) to (6), the plurality of features includes a difference between a maximum value and a minimum value, an effective value, an average value, a skewness, or a crest factor of the current in the current waveform.
With the above configuration (7), since the difference between maximum and minimum values, the effective value, the average value, the skewness, or the crest factor of the current in the current waveform of the measured current is used as the plurality of features, by acquiring the divergence between the distribution and the reference distribution of each of these features or the divergence between the multi-dimensional distribution and the reference multi-dimensional distribution of these features, it is possible to detect an abnormality in the rotary machine appropriately on the basis of the divergence.
(8) In some embodiments, in any one of the above configurations (1) to (7), the diagnosis apparatus for a rotary machine includes a divided waveform acquisition part (32) configured to acquire a divided waveform with a specified number of pulses from the current waveform. The feature acquisition part is configured to acquire the plurality of features for each divided waveform.
With the above configuration (8), since the divided waveform with a specified number of pulses is acquired from the current waveform obtained by current measurement, by acquiring the plurality of features for each divided waveform thus obtained, it is possible to acquire the distributions or multi-dimensional distribution of the plurality of features and the reference distributions or reference multi-dimensional distribution of the plurality of features appropriately. Therefore, it is possible to acquire the divergence between the distribution and the reference distribution of each of the plurality of features or the divergence between the multi-dimensional distribution and the reference multi-dimensional distribution of the plurality of features appropriately on the basis of the distributions thus acquired, and it is possible to detect an abnormality in the rotary machine appropriately on the basis of the divergence.
(9) In some embodiments, in the above configuration (8), the divided waveform acquisition part is configured to acquire a plurality of the divided waveforms by dividing the current waveform at a plurality of zero-crossing points (ZP) of the current waveform where the current passes through zero and a sign of the current changes in the same direction.
When dividing the current waveform, it is conceivable to divide the current waveform by a specified frequency (such as frequency associated with the rotation speed of the rotary machine), but in this case, the number of samples per period may not be stable, depending on the sampling interval of the measurement device or the like. In this regard, with the above configuration (9), the current waveform is divided at the zero-crossing points of the current waveform where the current passes through zero and the sign of the current changes in the same direction (from negative to positive, or from positive to negative). As a result, it is possible to obtain a plurality of divided waveforms whose current values are zero at the start point and the end point, and for each of the plurality of divided waveforms thus obtained, it is possible to acquire the plurality of features appropriately.
(10) In some embodiments, in the above configuration (9), the current waveform is represented as a curve connecting measurement values of the current acquired at a specified sampling period. The divided waveform acquisition part is configured to identify the zero-crossing points by linear interpolation of two of the measurement values with different signs.
Current measurement values may be acquired as discrete measurement values at each predetermined sampling period. With the above configuration (10), since the zero-crossing points are identified by linear interpolation of two measurement values with different signs among the plurality of current measurement values acquired at a specified sampling period, even if the plurality of discrete current measurement values does not include a measurement point with zero current value, the current waveform can be divided into the divided waveforms appropriately.
(11) In some embodiments, in the above configuration (10), the diagnosis apparatus for a rotary machine includes a filter (34) configured to reduce or remove noise components from a signal indicating the current. The divided waveform acquisition part is configured to identify the zero-crossing points on the basis of the signal processed by the filter.
In a signal containing noise, points with zero current value may randomly appear in addition to the inherent (i.e., noise-free) zero-crossing points due to waveform disturbance caused by noise. In this regard, with the above configuration (11), since the zero-crossing points are identified on the basis of the signal from which noise components have been reduced by the filter, the divided waveforms can be obtained by dividing the current waveform more appropriately on the basis of the zero-crossing points thus identified.
(12) In some embodiments, in the above configuration (11), the diagnosis apparatus for a rotary machine includes a filter setting part (36) configured to increase a time constant of the filter so that a difference between a maximum value and a minimum value of the number of sampling measurement values of the current included in each of the plurality of divided waveforms falls within an allowable range.
As described above, when the signal contains noise, points with zero current value appear randomly in addition to the inherent zero-crossing points. For this reason, the divided waveforms obtained on the basis of such apparent zero-crossing points may have large variations in the length from the start point to the end point (period of divided waveform) and the number of samples. In this regard, with the above configuration (12), since the time constant is increased so that the difference between maximum and minimum values of the number of sampling measurement values of the current included in the plurality of divided waveforms falls within the allowable range, it is possible to reduce variation in the number of sampling current measurement values included in the divided waveforms obtained on the basis of the zero-crossing points from the signal processed by the filter. Thus, it is possible to obtain the divided waveforms by dividing the current waveform more appropriately.
(13) In some embodiments, in the above configuration (12), the filter setting part is configured to repeatedly increase the time constant by a predetermined amount until the difference falls within the allowable range.
With the above configuration (13), since the time constant is repeatedly increased by the predetermined amount until the difference between maximum and minimum values of the number of sampling measurement values of the current included in the divided waveforms falls within the allowable range, it is possible to reliably reduce variation in the number of sampling current measurement values included in the divided waveforms obtained on the basis of the zero-crossing points from the signal processed by the filter. Thus, it is possible to obtain the divided waveforms by dividing the current waveform more appropriately.
(14) A diagnosis method for a rotary machine according to at least one embodiment of the present invention includes: a step (S8, S38) of acquiring, from a current waveform of a current measured during rotation of a rotary machine including a motor or a generator, a plurality of features each representing a characteristic of the current; and a step (S16 to S20, S46 to S50) of determining whether there is an abnormality in the rotary machine on the basis of a divergence between a distribution of each of the plurality of features or a multi-dimensional distribution of the plurality of features and a reference distribution of each of the plurality of features or a reference multi-dimensional distribution during normal operation of the rotary machine.
With the above method (14), it is determined whether there is an abnormality in the rotary machine on the basis of the divergence between the distribution of each of the plurality of features acquired from the current waveform of the measured current and the reference distribution of each of the plurality of features or the divergence between the multi-dimensional distribution of the plurality of features and the reference multi-dimensional distribution of the plurality of features. Therefore, as compared to the abnormality determination based on the divergence between the distribution and the reference distribution of a single feature, it is possible to detect an abnormality more exhaustively for the characteristics of the rotary machine and the types of abnormality. Thus, it is possible to detect an abnormality in the rotary machine more appropriately.
Additionally, in the above method (14), when using the multi-dimensional distribution and the reference multi-dimensional distribution of the plurality of features, one value indicating the divergence is calculated for the plurality of features. Therefore, the single value thus calculated is used to determine whether the rotary machine is normal or abnormal, which facilitates the abnormality determination of the rotary machine. Additionally, when using the multi-dimensional distribution and the reference multi-dimensional distribution of the plurality of features, since the divergence is calculated in relation to the plurality of features, changes in the distribution of the plurality of features when an abnormality occurs in the rotary machine can be grasped in more detail, compared to the case where the divergence is calculated for the distribution of a single feature. Therefore, it is possible to improve the abnormality detection performance of the rotary machine.
(15) A diagnosis program for a rotary machine according to at least one embodiment of the present invention is configured to cause a compute to execute: a process of acquiring, from a current waveform of a current measured during rotation of a rotary machine including a motor or a generator, a plurality of features each representing a characteristic of the current; and a process of determining whether there is an abnormality in the rotary machine on the basis of a divergence between a distribution of each of the plurality of features or a multi-dimensional distribution of the plurality of features and a reference distribution of each of the plurality of features or a reference multi-dimensional distribution during normal operation of the rotary machine.
With the above program (15), it is determined whether there is an abnormality in the rotary machine on the basis of the divergence between the distribution of each of the plurality of features acquired from the current waveform of the measured current and the reference distribution of each of the plurality of features or the divergence between the multi-dimensional distribution of the plurality of features and the reference multi-dimensional distribution of the plurality of features. Therefore, as compared to the abnormality determination based on the divergence between the distribution and the reference distribution of a single feature, it is possible to detect an abnormality more exhaustively for the characteristics of the rotary machine and the types of abnormality. Thus, it is possible to detect an abnormality in the rotary machine more appropriately.
Additionally, in the above program (15), when using the multi-dimensional distribution and the reference multi-dimensional distribution of the plurality of features, one value indicating the divergence is calculated for the plurality of features. Therefore, the single value thus calculated is used to determine whether the rotary machine is normal or abnormal, which facilitates the abnormality determination of the rotary machine. Additionally, when using the multi-dimensional distribution and the reference multi-dimensional distribution of the plurality of features, since the divergence is calculated in relation to the plurality of features, changes in the distribution of the plurality of features when an abnormality occurs in the rotary machine can be grasped in more detail, compared to the case where the divergence is calculated for the distribution of a single feature. Therefore, it is possible to improve the abnormality detection performance of the rotary machine.
Embodiments of the present invention were described in detail above, but the present invention is not limited thereto, and various amendments and modifications may be implemented.
Further, in the present specification, an expression of relative or absolute arrangement such as “in a direction”, “along a direction”, “parallel”, “orthogonal”, “centered”, “concentric” and “coaxial” shall not be construed as indicating only the arrangement in a strict literal sense, but also includes a state where the arrangement is relatively displaced by a tolerance, or by an angle or a distance whereby it is possible to achieve the same function.
For instance, an expression of an equal state such as “same” “equal” and “uniform” shall not be construed as indicating only the state in which the feature is strictly equal, but also includes a state in which there is a tolerance or a difference that can still achieve the same function.
Further, an expression of a shape such as a rectangular shape or a cylindrical shape shall not be construed as only the geometrically strict shape, but also includes a shape with unevenness or chamfered corners within the range in which the same effect can be achieved.
On the other hand, an expression such as “comprise”, “include”, and “have” are not intended to be exclusive of other components.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2020-130180 | Jul 2020 | JP | national |
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/JP2021/004490 | 2/8/2021 | WO |
