POWER CONVERSION DEVICE, ROTATING MACHINE SYSTEM, AND DIAGNOSIS METHOD

The power conversion device uses statistical analysis of current vector norms and zero-sequence currents to accurately detect intermittent bearing friction in rotating machines, enhancing failure detection accuracy and reducing the risk of overlooking mild deterioration.

JP7594434B2Active Publication Date: 2024-12-04HITACHI IND EQUIP SYST CO LTD
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
JP2020216259
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Priority Date
2020-01-16
Filing Date
2020-12-25
Publication Date
2024-12-04
Estimated Expiration
2040-12-25

AI Technical Summary

Technical Problem

Existing methods for diagnosing abnormalities in rotating machines, such as motors or generators, fail to detect intermittent issues like bearing friction fluctuations accurately, leading to potential overlooking of mild deterioration signs.

Method used

A power conversion device with a current measuring section, analysis target amount calculation, statistics sequential calculation unit, and abnormality degree calculation unit, which uses kurtosis and other statistical measures to detect abnormalities by analyzing current vector norms and zero-sequence currents, updating statistics sequentially to avoid rounding errors and improve accuracy.

Benefits of technology

Enables high-accuracy detection of intermittent bearing friction fluctuations and other abnormalities in rotating machines, even with low-performance microcontrollers, reducing the risk of overlooking early signs of failure.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007594434000007
    Figure 0007594434000007
  • Figure 0007594434000008
    Figure 0007594434000008
  • Figure 0007594434000009
    Figure 0007594434000009
Patent Text Reader

Abstract

To detect a sign of failure of a rotary machine with high accuracy.SOLUTION: A power conversion device includes a current measurement unit that measures a current flowing in a rotary machine, a to-be-analyzed amount calculation unit that calculates an amount to be analyzed from the current measured by the current measurement unit, a statistics sequential calculation unit that sequentially calculates statistics concerning a distribution of the amount to be analyzed calculated by the to-be-analyzed amount calculation unit, and anomaly calculation unit which calculates the degree of anomaly from the statistics obtained by the statistics sequential calculation unit.SELECTED DRAWING: Figure 1
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a power conversion device, a rotating machine system including a rotating machine such as a motor or a generator, and a diagnostic method. [Background technology]

[0002] When rotating machines such as motors and generators stop due to a sudden failure, it causes significant damage. In particular, the stoppage due to a sudden failure of rotating machines used in factory equipment has a large impact, such as reducing the operating rate of production equipment and forcing the review of production plans. Therefore, there is a growing need to prevent sudden failures of rotating machines by performing highly accurate failure prediction diagnosis while they are being used in the actual environment.

[0003] In response to such needs, there is known a technique for diagnosing various abnormalities by measuring the current during operation of a rotating machine and analyzing the frequency of the waveform. One such technique is a technique for diagnosing abnormalities by extracting the peak of the sideband wave of the power supply frequency through the frequency analysis of the current during operation.

[0004] Furthermore, Patent Document 1 is known as a technology for detecting an abnormality occurring in an observation target. In Patent Document 1, time series data of the values ​​of variables observed from the observation target is acquired, and a statistic that defines a probability density function of the variable at a certain time is calculated based on the value of the variable at the certain time and the statistic at a time prior to the certain time. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Patent Publication No. 2006-331300 Summary of the Invention [Problem to be solved by the invention]

[0006] The technology described above for diagnosing abnormalities by extracting the peaks of the sidebands of the power supply frequency through frequency analysis of the current during operation cannot detect abnormalities that do not have a characteristic frequency, such as intermittent increases and decreases in bearing friction, etc. Therefore, in the case of minor bearing deterioration, there is a problem that signs of abnormality may be overlooked.

[0007] Furthermore, Patent Document 1 describes that it is possible to appropriately select an observation value indicating an abnormality occurring in an observation target from among a plurality of observation values ​​obtained from the observation target. However, it does not sufficiently consider highly accurate detection of a fault sign of a rotating machine.

[0008] An object of the present invention is to detect a symptom of a failure in a rotating machine with high accuracy. [Means for solving the problem]

[0009] A preferred example of the present invention is a power conversion device including a current measuring unit that measures a current flowing through a rotating machine, an analysis target quantity calculation unit that calculates an analysis target quantity from the current measured by the current measuring unit, a statistics sequential calculation unit that sequentially calculates statistics related to the distribution of the analysis target quantity calculated by the analysis target quantity calculation unit, and an anomaly degree calculation unit that calculates an anomaly degree from the statistics obtained by the statistics sequential calculation unit. Effect of the Invention

[0010] According to the present invention, a sign of a fault in a rotating machine can be detected with high accuracy. [Brief description of the drawings]

[0011] [Figure 1] 1 is a basic configuration diagram of a power conversion device according to a first embodiment and a rotating machine system using the power conversion device; [Diagram 2] 4 is a flowchart for diagnosing an abnormality in a rotating machine system included in the power conversion device in the first embodiment. [Diagram 3] 3 shows typical phase current waveforms measured in the first embodiment. [Figure 4]4 is a typical current vector norm waveform calculated in the first embodiment. [Diagram 5] 4A to 4C are diagrams comparing typical instantaneous value histograms of the current vector norm obtained in the first embodiment in a normal state and an abnormal state. [Figure 6] 1 is a representative example of the kurtosis calculated sequentially in Example 1. [Figure 7] FIG. 11 is a basic configuration diagram of a power conversion device according to a second embodiment and a rotating machine system using the power conversion device. [Figure 8] 11 is a flowchart for diagnosing an abnormality in a rotating machine system included in a power conversion device according to a second embodiment. [Figure 9] FIG. 11 is a basic configuration diagram of a power conversion device according to a third embodiment and a rotating machine system using the power conversion device. [Figure 10] 11 is a flowchart for diagnosing an abnormality in a rotating machine system included in a power conversion device according to a third embodiment. [Figure 11] 10 is a flowchart for diagnosing an abnormality in a rotating machine system included in a power conversion device according to a fourth embodiment. [Figure 12] FIG. 13 is a basic configuration diagram of a power conversion device according to a fifth embodiment and a rotating machine system using the same. [Figure 13] 13 is a flowchart for diagnosing an abnormality in a rotating machine system included in a power conversion device according to a fifth embodiment. [Figure 14] FIG. 13 is a basic configuration diagram of a power conversion device according to a sixth embodiment and a rotating machine system using the same. [Figure 15] 13 is a flowchart for diagnosing an abnormality in a rotating machine system included in a power conversion device according to a sixth embodiment. [Figure 16] FIG. 13 is a basic configuration diagram of a power conversion device according to a seventh embodiment and a rotating machine system using the same. [Figure 17] 20 is a flowchart for diagnosing an abnormality in a rotating machine system included in a power conversion device in a seventh embodiment. [Figure 18] 13 is a schematic diagram of a phase current waveform measured in Example 7. FIG. [Figure 19] 13 is a comparison diagram of the distribution of the fundamental period of the phase current obtained in the seventh embodiment in a normal state and an abnormal state. FIG. [Figure 20]13 is a representative example of the standard deviation of the distribution of fundamental periods calculated sequentially in Example 7. [Figure 21] FIG. 13 is a basic configuration diagram of a power conversion device according to an eighth embodiment and a rotating machine system using the same. [Figure 22] 20 is a flowchart for diagnosing an abnormality in a rotating machine system included in a power conversion device according to an eighth embodiment. [Diagram 23] 13 is a flowchart for diagnosing an abnormality in a rotating machine system included in a power conversion device according to a ninth embodiment. [Figure 24] FIG. 23 is a basic configuration diagram of a power conversion device according to a tenth embodiment and a rotating machine system using the same. [Diagram 25] 23 is a flowchart for diagnosing an abnormality in a rotating machine system included in a power conversion device in a tenth embodiment. [Figure 26] FIG. 23 is a basic configuration diagram of a power conversion device according to an eleventh embodiment and a rotating machine system using the same. [Figure 27] 23 is a flowchart for diagnosing an abnormality in a rotating machine system included in a power conversion device in an eleventh embodiment. [Figure 28] FIG. 23 is a basic configuration diagram of a power conversion device according to a twelfth embodiment and a rotating machine system using the same. [Figure 29] 23 is a flowchart for diagnosing an abnormality in a rotating machine system included in a power conversion device in a twelfth embodiment. [Diagram 30] FIG. 23 is a basic configuration diagram of a power conversion device according to a thirteenth embodiment and a rotating machine system using the same. [Diagram 31] 23 is a flowchart for diagnosing an abnormality in a rotating machine system included in a power conversion device in a thirteenth embodiment. [Diagram 32] 1 is a schematic diagram of a power conversion device and a pump incorporating a rotating machine system using the same. [Diagram 33] 1 is a schematic diagram of an air compressor incorporating a power conversion device and a rotating machine system using the same. [Diagram 34] FIG. 1 is a schematic diagram of a power conversion device and a conveying table incorporating a rotating machine system using the power conversion device. [Diagram 35]1 is a schematic diagram of a machine tool incorporating a power conversion device and a rotating machine system using the same. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Hereinafter, embodiments of the present invention will be described with reference to the drawings. EXAMPLES

[0013] 1 is a basic configuration diagram of a power conversion device and a rotating machine system using the same in Example 1. This rotating machine system includes a power conversion device 10 having a main circuit 11 and current sensors 12a and 12b, a power source 20 connected to the power conversion device 10 to exchange electric power, and a rotating machine 30. The rotating machine 30 includes a bearing 31, and a load machine 40 is connected to the tip of the rotating shaft.

[0014] The power conversion device 10 further includes processing units such as a current measurement unit 13 that measures the phase currents detected by the current sensors 12a, 12b, a current vector calculation unit 14 that performs three-phase to two-phase conversion on the measured phase currents to calculate a current vector, an analysis target DC flow calculation unit 15 that calculates an analysis target quantity based on the calculated current vector, a statistical quantity sequential calculation unit 16 that sequentially calculates statistical quantities related to the distribution of the DC flow calculated by the analysis target DC flow calculation unit 15, an anomaly degree calculation unit 17 that calculates an anomaly degree from the statistical quantities obtained by the statistical quantity sequential calculation unit 16, and an output unit 18 that outputs a diagnosis result from the anomaly degree obtained by the anomaly degree calculation unit 17.

[0015] The current measurement unit 13, current vector calculation unit 14, analysis target DC flow calculation unit 15, statistical quantity sequential calculation unit 16, and anomaly degree calculation unit 17 are stored as programs in a recording unit such as a ROM, and a control unit (not shown) such as a microcomputer of the power conversion device 10 reads out the programs and executes the processing of each processing unit.

[0016] In Fig. 1, the current sensors 12a and 12b are installed between the rotating machine 30 and the output of the main circuit 11 of the power conversion device 10 so as to acquire two phase currents among the currents flowing through the rotating machine 30, but a current sensor may be installed for each of the three phase currents. Alternatively, a shunt resistor may be used. Also, a current sensor having diagnostic capabilities may be installed outside the power conversion device 10, such as the rotating machine 30, in addition to a current sensor for control installed inside the power conversion device 10.

[0017] A current sensor may be disposed on the positive or negative side of a DC bus (not shown) of the main circuit 11 to acquire the current flowing through the rotating machine 30. Alternatively, a current sensor may be disposed between the emitter or source of a switching element in the main circuit 11 and the DC bus (negative or positive side).

[0018] When some kind of lubrication problem occurs in the bearings of the rotating machine 30 and friction fluctuates intermittently, the amplitude of the motor current momentarily increases or decreases. When this frequency increases, the average spectrum level around the fundamental frequency of the phase current (e.g., ±20 Hz) rises, so in the past, frequency analysis such as FFT (Fast Fourier Transform) was carried out to detect the frequency components. However, in the case of minor bearing abnormalities that occur infrequently and do not show changes in the frequency components, there was a problem that the conventional method of focusing on frequency components would miss signs of abnormality.

[0019] In this embodiment, in order to capture the instantaneous increase and decrease of the amplitude of the motor current at a time when it occurs less frequently, the motor current is measured by the current measuring unit 13, and then converted to a DC amount by the analysis target DC amount calculating unit 15. For example, it may be the norm of the current vector obtained by converting the phase current into three-phase to two-phase, or it may be the torque current (q-axis current), or it may be the zero-phase current. When an instantaneous value histogram of the DC amount is created, an abnormality sign appears in the form of an increase in events that deviate significantly from the average value. In the world of statistics, kurtosis is often used as an index of the number of outliers in a probability density distribution. Kurtosis is a dimensionless quantity defined as the ratio of the fourth moment around the average value to the square of the variance, and therefore has the advantage of not being directly affected by the magnitude of the motor load.

[0020] However, power conversion devices are usually not equipped with a memory that can temporarily store all the data to be statistically processed. Therefore, the statistical calculation unit sequentially updates the statistical calculation value each time a piece of data is input.

[0021] FIG. 2 is a flowchart for diagnosing an abnormality in a rotating machine system provided in the power conversion device of this embodiment. First, in step S100, the diagnostic mode is started. In addition to a method of selecting from the setting items of the power conversion device, a mechanical button for starting the diagnostic mode may be pressed, or a "diagnosis mode" button displayed on a display may be touched. Alternatively, the diagnostic mode may be set to start automatically at a specific date and time, or when the amplitude of the fundamental frequency of the phase current continues for a specific period of time and falls within a specific range. Alternatively, the diagnostic mode may be set to start automatically before or after a specific rotating machine control operation.

[0022] Next, the current measuring unit 13 receives the currents detected by the current sensors 12a and 12b and measures the phase currents of at least two phases (step S101). Alternatively, the phase currents of all phases may be measured. Alternatively, the current of the DC portion may be measured and regarded as the phase current of each phase according to the time. The current measurement is performed in synchronization with the voltage pulses generated by the power conversion device 10 to drive the rotating machine 30. That is, the signals from the current sensors 12a and 12b are sampled in synchronization with the voltage pulses.

[0023] Sampling may be performed once or twice per voltage pulse. Sampling may also be performed once per multiple voltage pulses. For example, an example of two-phase phase current waveforms obtained by sampling twice per voltage pulse is shown in FIG. 3. When the power conversion device 10 performs current measurement synchronized with the voltage pulse for control purposes, the current measurement can also be used for the failure sign detection of this embodiment.

[0024] Next, the current vector calculation unit 14 performs three-phase to two-phase conversion on the measured phase currents according to the following formula (1) to calculate a current vector (step S102). However, if only two phase currents are measured, a value derived on the assumption that formula (2) is satisfied is used as the phase current of the remaining phase.

[0025]

number

[0026] where i U、 i V、 i W indicates the measured current of each phase, U, V, and W, and i α , i β indicates the current components on the two-phase fixed axes α-axis and β-axis after three-phase to two-phase conversion.

[0027]

number

[0028] Next, the analysis target DC amount calculation unit 15 calculates the current vector norm expressed by the formula (3) for the obtained current vector (step S103). Fig. 4 shows an example of the calculated current vector norm.

[0029]

number

[0030] Alternatively, the current vector may be further transformed into a rotating coordinate system using equation (4) to calculate the q-axis current (torque current), which may be used in the subsequent analysis instead of the current vector norm. The rotor position θ appearing in equation (4) may be obtained from a signal from a rotational position sensor such as a resolver or encoder installed in the rotating machine 30, or may be estimated from information contained in the speed electromotive force or current of the rotating machine 30.

[0031]

number

[0032] where i d , i q indicate the d-axis current and the q-axis current, respectively, and θ indicates the rotor position.

[0033] When bearing deterioration causes intermittent increases in friction, there will be an increasing number of instances where the instantaneous value of the current vector norm or q-axis current significantly deviates from the average value. For example, the histogram of the instantaneous value of the current vector norm is thought to change as shown in Fig. 5 as bearing deterioration progresses.

[0034] Such changes can occur not only in the current vector norm and q-axis current, but also in the zero-phase current. To perform such statistical processing with high precision, a minimum of 10,000 samples are required. Normally, memory is required to store and expand them all at once, and the computing power of the microcomputer that performs the calculations is also required.

[0035] In particular, when calculating quantities with large orders such as kurtosis, it is necessary to first calculate the mean and variance, which results in a very large amount of calculations. Although the performance of microcomputers installed in power conversion devices is improving year by year, most of the resources are allocated to motor control, and there is a limit to the surplus memory and surplus computing power that can be used for anomaly detection.

[0036] Therefore, in this embodiment, the statistical sequential calculation unit 16 derives an equation for updating various statistical quantities when the (n+1)th data is received while holding various statistical quantities calculated from n data, and proceeds with the calculation sequentially (step S104). The DC amount to be analyzed is x i When (i=1,2,…,n), the average value μ n , variance σ n 2 , skewness S n , kurtosis K n is calculated as shown in equation (5).

[0037]

number

[0038] If we expand and organize the contents of the n sums that appear on the right-hand side of each equation in equation (5), we can derive equations that update various statistics when the (n+1)th data item is received while retaining the various statistics calculated from n data items. Using these, we can sequentially calculate the mean, variance, skewness, and kurtosis. Figure 6 shows a representative example of sequentially calculated kurtosis. We can see that kurtosis is larger in an abnormal state compared to a normal state.

[0039] By calculating a statistical quantity called kurtosis for the DC amount calculated by the analysis target DC amount calculation unit 15, intermittent abnormalities such as bearing friction can be distinguished from normality and detected with high accuracy.

[0040] Next, the anomaly degree calculation unit 17 calculates the degree of anomaly from the various statistics obtained by the statistics sequential calculation unit 16 (step S105). For example, the obtained kurtosis is used as the degree of anomaly. Alternatively, the moving average of the kurtosis is calculated and used as the degree of anomaly.

[0041] Then, the anomaly degree calculation unit 17 judges whether the calculated degree of anomaly exceeds a preset threshold value (step S106). If the threshold value is exceeded, it instructs the output unit 18 to diagnose that there is a sign of a failure and issue a warning. The threshold value may be determined based on the results of a separate study using a test machine or the like. Alternatively, the anomaly degree calculation unit 17 may determine the threshold value by learning in a real environment. For example, the degree of anomaly is calculated multiple times during a learning period, the average value and the fluctuation range are found, and the threshold value is set to a level that does not cause false alarms.

[0042] The calculated degree of anomaly may be output together with other parameters that define the operating state of the rotating machine 30 so that precise diagnosis can be performed later in machine learning or the like. If such multivariate data is recorded using a logger or the like and analyzed using a machine learning algorithm such as vector quantization clustering, it may be possible to detect even weak anomalies that cannot be detected by threshold judgment.

[0043] Finally, the output unit 18 outputs the diagnosis results such as the degree of abnormality and the failure sign judgment result obtained by the abnormality degree calculation unit 17 (step S107) to complete the diagnosis. The display method may be a display, a lamp, a buzzer, or the like that appeals to the five human senses, or may be recorded on paper or an electronic file. Alternatively, the results may be transmitted via a communication network.

[0044] As described above, according to the power conversion device of this embodiment and the rotating machine system using the same, diagnosis is performed based on statistics, so that intermittent bearing friction fluctuations that cannot be detected by frequency analysis can be detected with high sensitivity.

[0045] According to the first embodiment, it is possible to realize a power conversion device that detects a symptom of a failure in a rotating machine system with high accuracy or at an early stage, and a rotating machine system using the power conversion device. EXAMPLES

[0046] 7 is a basic configuration diagram of a power conversion device and a rotating machine system using the same in a second embodiment. Explanation of the same parts as in the first embodiment will be omitted. The difference from the first embodiment is that an addition residual storage unit 19 is provided for storing an addition residual generated when the statistics sequential calculation unit 16 sequentially calculates various statistics.

[0047] A low-performance microcomputer that is normally installed in the power conversion device 10 cannot handle double-precision floating-point numbers at a practical speed. Therefore, the number of significant digits of the variables used in sequential calculation of various statistics must be reduced, and rounding errors become a problem when the number of samples is increased.

[0048] Specifically, if you try to calculate the four types of sums shown in equation (6) sequentially, the larger n becomes, the more likely it is that a rounding error will occur because a very small amount will be added to a very large partial sum.

[0049]

number

[0050] Therefore, in the second embodiment, when the statistics sequential calculation unit 16 performs sequential calculations, the generated addition residual (addition remainder) is stored in the addition residual storage unit 19, and the addition residual is updated each time, thereby avoiding rounding errors. That is, when adding a very large partial sum to a small value, the statistics sequential calculation unit 16 first adds the small value to the addition residual stored in the addition residual storage unit 19, and adds it to the partial sum. Then, the statistics sequential calculation unit 16 calculates the increase in the partial sum (the amount actually added) and updates the addition residual (step S204).

[0051] In this way, it is possible to prevent the addition residual from increasing like a snowball and suppress the rounding error below a certain amount. The anomaly calculation unit 17 calculates the anomaly degree from the statistics obtained by the statistics sequential calculation unit 16.

[0052] 8 is a flowchart for diagnosing an abnormality in a rotating machine system provided in a power conversion device according to embodiment 2. The difference from the flowchart in embodiment 1 is that in step S204, various statistics are sequentially calculated while storing and updating the additive residuals.

[0053] In Patent Document 1, since a process of adding the variable value at each time to a very large value of the same order as the number of samples occurs, when the number of samples approaches the number of significant digits of the variable, the effect of rounding error cannot be ignored and the calculation fails. Therefore, there is a problem that a low-performance microcomputer normally used in the power conversion device 10 cannot handle a sufficient number of samples.

[0054] In the second embodiment, since the statistics are calculated sequentially while avoiding rounding errors, the diagnosis can be performed even with a low-performance microcomputer and a small-capacity memory mounted for control in the power conversion device 10. Therefore, even with a low-performance microcomputer, the diagnosis can be performed with high accuracy at a practical speed.

[0055] According to the second embodiment, a sign of a failure in a rotating machine system can be detected at low cost and with high accuracy. EXAMPLES

[0056] 9 is a basic configuration diagram of a power conversion device and a rotating machine system using the same in embodiment 3. Description of the same parts as in embodiment 1 or embodiment 2 will be omitted.

[0057] In the third embodiment, there is no current vector calculation unit 14, and a zero-phase current calculation unit calculates a zero-phase current from a current measured by a current measurement unit 13 as an analyzed DC amount calculation unit 15. A statistics sequential calculation unit 16 calculates statistics from the calculated zero-phase current. In the third embodiment, current sensors 12a, 12b, and 12c are arranged to sense three-phase currents.

[0058] 10 is a flowchart for diagnosing an abnormality in a rotating machine system provided in a power conversion device in Example 3. The difference from Example 1 or Example 2 is that an analysis target DC amount calculation unit (zero-phase current calculation unit) 15 simultaneously measures three phase currents to calculate a zero-phase current, or directly measures the zero-phase current and sets it as the analysis target DC amount (step S301).

[0059] According to the third embodiment, the calculation in the analysis target DC amount calculation unit 15 is simplified, and diagnosis is possible even with a microcomputer having lower performance. EXAMPLES

[0060] Fig. 11 is a flowchart for diagnosing an abnormality in a rotating machine system provided in a power conversion device according to the fourth embodiment. The basic configuration diagram of the power conversion device and the rotating machine system using the same in the fourth embodiment is the same as Fig. 7 in the second embodiment. Descriptions of the same parts as those in the first to third embodiments will be omitted.

[0061] The difference between the first to third embodiments is that the statistics sequential calculation unit 16 further derives a second statistic regarding the distribution of the first statistic obtained by calculation (step S405), and the anomaly degree calculation unit 17 further calculates the degree of anomaly from the second statistic (step S406). For example, the kurtosis and average value of an instantaneous value histogram of the DC amount are sequentially calculated for one second, and a histogram of the final values ​​is created. Then, the statistics (average value and variance) of the histogram are further sequentially calculated.

[0062] According to the fourth embodiment, by performing diagnosis using the derived statistics regarding the distribution of statistics, it is possible to detect a more subtle abnormality symptom. EXAMPLES

[0063] Fig. 12 is a basic configuration diagram of a power conversion device and a rotating machine system using the same in the fifth embodiment. Fig. 13 is a flowchart for diagnosing an abnormality in the rotating machine system provided in the power conversion device of the fifth embodiment. Descriptions of the same parts as in the first to fourth embodiments will be omitted.

[0064] The difference between the first to fourth embodiments is that the present embodiment includes a rotational frequency estimation unit 21 that estimates the rotational frequency of the rotating machine 30, and a feature waveform extraction unit 22 that extracts a waveform in an arbitrary frequency range between zero and the rotational frequency, and an anomaly degree calculation unit 17 calculates and diagnoses the degree of anomaly using both the degree of anomaly calculated from the statistics obtained by the statistical sequential calculation unit 16 and the degree of anomaly calculated from the waveform extracted by the feature waveform extraction unit 22. Next, the processes of the rotational frequency estimating unit 21, the feature amount waveform extracting unit 22, and the abnormality degree calculating unit 17, which are different from those in the second embodiment, will be described.

[0065] In parallel with the analysis target DC amount calculation unit 15, the rotation frequency estimation unit 21 estimates the rotation frequency of the rotating machine 30 (step S504). When the analysis target DC amount calculation unit 15 performs rotor position estimation in step S503, it is possible to estimate the rotation frequency at the same time. When a rotation position sensor is not installed in the rotating machine 30, the rotation frequency may be estimated from the output frequency of the power conversion device 10 and the number of poles of the rotating machine 30. The output frequency of the power conversion device 10 is estimated from phase current information or extracted from control command information.

[0066] If we ignore slippage that occurs in induction motors at a maximum of a few percent, the rotational frequency can be estimated from the output frequency of the power conversion device 10 and the number of poles of the rotating machine 30 as follows. Since rotational frequency = output frequency of power conversion device / number of pole pairs, for example, if the output frequency is 50 Hz and there are four poles (two pole pairs), the rotational frequency is estimated to be 25 Hz. If the output frequency is 30 Hz and there are six poles (three pole pairs), the rotational frequency is estimated to be 10 Hz. If the output frequency is 100 Hz and there are two poles (one pole pair), the rotational frequency is estimated to be 100 Hz.

[0067] In response to the results of the current vector norm or torque current calculated by the analysis target DC amount calculation unit 15 and the rotational frequency estimated by the rotational frequency estimation unit 21, the feature waveform extraction unit 22 extracts a waveform in the frequency range between zero, which is the specified frequency range, and the rotational frequency, for the current vector norm or torque current (step S506).

[0068] The feature waveform extraction unit 22 can extract a specific frequency range by passing the current vector norm or the torque current through a filter.

[0069] Next, the anomaly degree calculation unit 17 calculates a second degree of anomaly from the waveform obtained by the feature waveform extraction unit 22 (step S507). For example, the degree of anomaly may be calculated by adding up the absolute value or square value of the obtained waveform for a certain period of time. Alternatively, the calculated degree of anomaly may be normalized by the amplitude of the fundamental frequency component of the phase current, or the current vector norm or a DC component extracted from the torque current.

[0070] This makes it possible to perform diagnosis even when the DC component changes over time due to load fluctuation. Also, when communicating the degree of anomaly to the outside, the communication load can be reduced. As described in the second embodiment, the anomaly degree calculation unit 17 of this embodiment performs an anomaly diagnosis based on the first degree of anomaly and the second degree of anomaly calculated from the statistics obtained by the statistics sequential calculation unit 16.

[0071] Instead of calculating the anomaly degrees individually, a single anomaly degree may be calculated by clustering from the beginning. By doing as in this embodiment, a wider range of abnormal phenomena in the rotating machine system can be detected. EXAMPLES

[0072] Fig. 14 is a basic configuration diagram of a power conversion device and a rotating machine system using the same in the sixth embodiment. Fig. 15 is a flowchart for diagnosing an abnormality in the rotating machine system provided in the power conversion device of the sixth embodiment.

[0073] Description of the same parts as in the first to fifth embodiments will be omitted. The difference from the first to fifth embodiments is that a characteristic frequency component extracting unit 23 is provided to extract the magnitude of the rotation frequency and its multiple frequency components.

[0074] The characteristic frequency component extraction unit 23 obtains the rotation frequency from the rotation frequency estimation unit 21, and obtains the current vector norm or the torque current from the analysis target DC amount calculation unit 15. The characteristic frequency component extraction unit 23 extracts the magnitude of the rotation frequency and its multiple frequency components for the current vector norm or the torque current (step S606). In this embodiment, the anomaly degree calculation unit 17 calculates and diagnoses the degree of anomaly (step S607) using both the first degree of anomaly calculated from the statistics obtained by the statistical sequential calculation unit 16 and the second degree of anomaly calculated from the specific frequency component value extracted by the feature frequency component extraction unit 23.

[0075] The characteristic frequency component extraction unit 23 may perform a Fast Fourier Transform (FFT) on the current vector norm or the torque current to select the characteristic frequency component to be acquired, or may extract the characteristic frequency component to be acquired by using a filter without performing an FFT.

[0076] In the sixth embodiment, the characteristic frequency component extraction unit 23 extracts the magnitude of the rotation frequency and its multiple frequency components for the current vector norm or the torque current. However, as in the third embodiment, the characteristic frequency component extraction unit 23 may extract the magnitude of the rotation frequency and its multiple frequency components for the zero-phase current.

[0077] Note that instead of calculating the anomaly degrees individually, a single anomaly degree may be calculated by clustering from the beginning. By doing as in this embodiment, a wider range of abnormal phenomena in the rotating machine system can be detected.

[0078] According to the sixth embodiment, by performing diagnosis using both the degree of abnormality calculated from statistics and the degree of abnormality calculated from specific frequency component values ​​extracted by the characteristic frequency component extraction unit 23, it is possible to increase the number of abnormal phenomena that can be detected. EXAMPLES

[0079] FIG. 16 is a basic configuration diagram of a power conversion device and a rotating machine system using the same in a seventh embodiment. Explanation of the same parts as in the first embodiment will be omitted. The difference from the first embodiment is that at least one phase of phase current needs to be measured, and that a phase current fundamental period calculation unit 24 that calculates the fundamental period of the measured phase current for each period is provided as an analysis object quantity calculation unit, instead of the current vector calculation unit 14 and the analysis object DC amount calculation unit 15. Note that, although a current sensor is installed to measure the phase current of one phase in FIG. 16, it may be installed in two or more phases. Alternatively, a shunt resistor may be used.

[0080] In this embodiment, in order to capture instantaneous increases and decreases in the fundamental period of the motor current at a time point when they occur less frequently, the motor current is measured by a current measuring unit, and the value is recorded for each fundamental period of vibration. The value of the fundamental period of the phase current as the analysis target quantity may be derived from the interval of time when the phase current value becomes zero. The timing of becoming zero occurs twice per period, but only the timing when the phase current value changes from negative to positive may be used, or only the timing when the phase current value changes from positive to negative may be used. As another example of calculating the fundamental period, it may be calculated from the interval of time when a specific amplitude value is satisfied. For example, the fundamental period may be calculated from the interval of time when the phase current reaches its maximum peak.

[0081] When the distribution of the fundamental period is calculated, an abnormal sign appears in the form of an increase in events that deviate from the average value. As an index of the amount of variation in the probability density distribution, kurtosis may be used as in the first embodiment, or standard deviation may be used. The standard deviation is the square root of the variance, which is the second moment around the average value, and has the same dimensions as the fundamental period.

[0082] FIG. 17 is a flowchart for diagnosing an abnormality in a rotating machine system provided in a power conversion device in the seventh embodiment. First, in step S700, a diagnostic mode is started. In addition to a method of selecting from the setting items of the power conversion device, a mechanical button for starting the diagnostic mode may be pressed, or a "diagnosis mode" button displayed on a display may be touched. Alternatively, the diagnostic mode may be set to start automatically at a specific date and time, or when the amplitude of the fundamental frequency of the phase current continues for a specific period of time and falls within a specific range. Alternatively, the diagnostic mode may be set to start automatically before or after a specific rotating machine control operation.

[0083] Next, in step S701, the phase current of at least one phase is measured. Alternatively, the phase current of two or more phases may be measured. Alternatively, the current of the DC part may be measured and regarded as the phase current of each phase according to the time. A schematic diagram of a phase current waveform is shown in FIG. 18. The current measurement is performed in synchronization with a voltage pulse generated by the power conversion device to drive the rotating machine. That is, the signal from the current sensor is sampled in synchronization with the voltage pulse. Sampling may be performed once or twice per voltage pulse. Sampling may also be performed once per multiple voltage pulses.

[0084] Next, in step S702, the value of the fundamental period is calculated for each period for the measured phase current. When the bearings deteriorate and friction increases intermittently, the fundamental period of the phase current will deviate from the average value more and more. For example, the distribution of the fundamental period is thought to change as shown in Figure 19 as bearing deterioration progresses.

[0085] Next, in step S703, the mean value, variance, skewness, and kurtosis of the DC quantity to be analyzed (here, the fundamental period of the phase current) are sequentially calculated using the same formulas as in the first embodiment. After the sequential calculation has been continued for a certain period of time, the mean value, variance, skewness, and kurtosis are recorded, and each value is returned to zero before the sequential calculation is resumed. The standard deviation value is obtained by taking the square root of the recorded variance value. A schematic diagram of the standard deviation of the distribution of the fundamental period thus obtained is shown in FIG. 20. It can be seen that the standard deviation is larger in an abnormal state compared to a normal state.

[0086] Next, in step S704, the degree of abnormality is calculated from the various statistics obtained. For example, the obtained standard deviation or kurtosis is used as the degree of abnormality. Alternatively, the moving average of the standard deviation or kurtosis is calculated and used as the degree of abnormality.

[0087] In step S705, it is determined whether the calculated degree of abnormality exceeds a preset threshold, and if it does, it is diagnosed that there is a sign of failure and a warning is issued. The threshold may be determined based on the results of a separate study using a test machine or the like, or may be determined by learning in a real environment. For example, the degree of abnormality is calculated multiple times during the learning period, the average value and the fluctuation range are obtained, and the threshold is set to a level that does not cause false alarms.

[0088] Furthermore, the calculated degree of anomaly may be output together with other parameters that define the operating state of the rotating machine so that a precise diagnosis can be made later by machine learning, etc. By recording such multivariate data using a logger or the like and analyzing it using a machine learning algorithm such as vector quantization clustering, it may be possible to detect even weak anomalies that cannot be detected by threshold judgment.

[0089] Finally, in step S706, the diagnosis result is displayed and the diagnosis is completed. The display method may be a display, a lamp, a buzzer, or the like that appeals to the five human senses, or may be recorded on paper or in an electronic file, or may be transmitted via a communication network.

[0090] As described above, according to the seventh embodiment, it is possible to realize a power conversion device and a rotating machine system using the same that can detect signs of a failure in a rotating machine system with high accuracy or at an early stage, even when abnormal symptoms appear in the time axis direction rather than the amplitude direction of the motor current. EXAMPLES

[0091] FIG. 21 is a basic configuration diagram of a power conversion device and a rotating machine system using the same in the eighth embodiment. Explanation of the same parts as in the seventh embodiment will be omitted. The difference from the seventh embodiment is that an addition residual storage unit 19 is provided for storing an addition residual generated when various statistics are sequentially calculated by the statistics sequential calculation unit 16. The method of updating the addition residual is the same as in the second embodiment. This makes it possible to prevent the addition residual from snowballing and suppress the rounding error to a certain amount or less.

[0092] 22 is a flowchart for diagnosing an abnormality in a rotating machine system provided in a power conversion device according to the eighth embodiment. The difference from the seventh embodiment is that in step S803, various statistics are calculated sequentially while storing and updating the additive residual. This allows even a low-performance microcomputer to perform diagnosis at a practical speed with high accuracy.

[0093] According to the eighth embodiment, a sign of a failure in a rotating machine system can be detected at low cost and with high accuracy. EXAMPLES

[0094] Fig. 23 is a flowchart for diagnosing an abnormality in a rotating machine system provided in a power conversion device according to the 9th embodiment. The basic configuration diagram of the power conversion device and the rotating machine system using the same in the 9th embodiment is the same as Fig. 21 in the 8th embodiment. Descriptions of the same parts as those in the 7th and 8th embodiments will be omitted.

[0095] The difference from the seventh and eighth embodiments is that the statistics sequential calculation unit 16 further derives a second statistic from the distribution of the first statistic obtained by calculation (step S904), and the anomaly degree calculation unit 17 further calculates the degree of anomaly from the second statistic (step S905). For example, the standard deviation, kurtosis, and mean value of the distribution of the fundamental period are sequentially calculated for one second, and a histogram of the final values ​​is created. Then, the statistics (mean value and variance) of the histogram are further sequentially calculated.

[0096] According to the ninth embodiment, by performing diagnosis using the statistics derived regarding the distribution of the statistics, it is possible to detect a more subtle abnormal symptom. EXAMPLES

[0097] 24 is a basic configuration diagram of a power conversion device and a rotating machine system using the same in the tenth embodiment. The system includes a current vector calculation unit 14 and an analysis target DC amount calculation unit 15 similar to those in the first and second embodiments, and a phase current fundamental period calculation unit 24 similar to those in the seventh and eighth embodiments.

[0098] Fig. 25 is a flowchart for diagnosing an abnormality in a rotating machine system provided in a power conversion device in the tenth embodiment. This is a combination of the flowchart in the second embodiment (Fig. 8) and the flowchart in the eighth embodiment (Fig. 22). Description of the same parts as those in Fig. 8 or Fig. 22 will be omitted.

[0099] In step S1005, statistics are sequentially calculated for each of the histograms of the fundamental period of the phase current calculated in step S1002 and the current vector norm and torque current calculated in step S1004.

[0100] Then, in step S1006, a first degree of anomaly regarding the fundamental period of the phase current and a second degree of anomaly regarding the current vector norm or torque current are calculated, and anomaly diagnosis is performed using both of them. Alternatively, instead of calculating the degrees of anomaly individually, a single degree of anomaly may be calculated from the beginning by unifying them through clustering.

[0101] According to the tenth embodiment, a wider range of abnormal phenomena in the rotating machine system can be detected. EXAMPLES

[0102] 26 is a basic configuration diagram of a power conversion device and a rotating machine system using the same in the eleventh embodiment. The system includes an analysis target DC amount calculation unit (zero-phase current calculation unit) 15 similar to that in the third embodiment, and a phase current fundamental period calculation unit 24 similar to that in the seventh and eighth embodiments.

[0103] Fig. 27 is a flowchart for diagnosing an abnormality in a rotating machine system provided in a power conversion device in the eleventh embodiment. This is a combination of the flowchart in the third embodiment (Fig. 10) and the flowchart in the eighth embodiment (Fig. 22). The same parts as those in Fig. 10 or 22 will not be described.

[0104] In step S1104, statistics are sequentially calculated for each of the histograms of the fundamental period of the phase current calculated in step S1102 and the zero-phase current calculated in step S1103.

[0105] Then, in step S1105, a first anomaly degree related to the fundamental period of the phase current and a second anomaly degree related to the zero-phase current are calculated, and an anomaly diagnosis is performed using both of them. Alternatively, instead of calculating the anomaly degrees individually, a single anomaly degree integrated by clustering may be calculated from the beginning.

[0106] According to the eleventh embodiment, the calculation in the analysis target DC amount calculation unit 15 is simplified, and diagnosis is possible even with a microcomputer having lower performance, while a wider range of abnormal phenomena in the rotating machine system can be detected. EXAMPLES

[0107] Fig. 28 is a basic configuration diagram of a power conversion device and a rotating machine system using the same in Example 12. As in Example 5, a rotational frequency estimation unit 21 and a feature waveform extraction unit 22 are added to the basic configuration diagram of Example 10 (Fig. 24). Then, an anomaly degree calculation unit 17 performs an anomaly diagnosis using both the anomaly degree calculated from the statistics obtained by the statistics sequential calculation unit 16 and the anomaly degree calculated from the waveform extracted by the feature waveform extraction unit 22.

[0108] Fig. 29 is a flowchart for diagnosing an abnormality in a rotating machine system provided in a power conversion device of the twelfth embodiment. Explanation of the same parts as those in the flowchart of the fifth embodiment (Fig. 13) and the flowchart of the tenth embodiment (Fig. 25) will be omitted. The difference from the tenth embodiment is that it includes a rotational frequency estimation unit 21 for estimating the rotational frequency of the rotating machine 30, and a feature waveform extraction unit 22 for extracting a waveform in an arbitrary frequency range between zero and the rotational frequency, and an anomaly degree calculation unit 17 calculates and diagnoses the degree of anomaly using both the degree of anomaly calculated from the statistics obtained by the statistics sequential calculation unit 16 and the degree of anomaly calculated from the waveform extracted by the feature waveform extraction unit 22.

[0109] The processing of the rotational frequency estimation unit 21 and the feature waveform extraction unit 22 is the same as in the fifth embodiment, and therefore a description thereof will be omitted. The processing of the anomaly degree calculation unit 17 calculates a first anomaly degree related to the fundamental period of the phase current, a second anomaly degree related to the current vector norm or torque current, and a third anomaly degree related to the waveform obtained by the feature waveform extraction unit 22 (step S1208). Then, an anomaly diagnosis is performed using each anomaly degree. Instead of calculating the anomaly degrees individually, a single anomaly degree integrated by clustering may be calculated from the beginning. By performing as in this embodiment, a wider range of abnormal phenomena in the rotating machine system can be detected. EXAMPLES

[0110] Fig. 30 is a basic configuration diagram of a power conversion device and a rotating machine system using the same in Example 13. As in Example 6, a rotational frequency estimation unit 21 and a feature frequency component extraction unit 23 are added to the basic configuration diagram of Example 10 (Fig. 24). Then, an anomaly degree calculation unit 17 performs an anomaly diagnosis using both the anomaly degree calculated from the statistics obtained by the statistics sequential calculation unit 16 and the anomaly degree calculated from the frequency components extracted by the feature frequency component extraction unit 23.

[0111] Fig. 31 is a flowchart for diagnosing an abnormality in a rotating machine system provided in a power conversion device of the thirteenth embodiment. Explanation of the same parts as those in the flowchart of the sixth embodiment (Fig. 15) and the flowchart of the tenth embodiment (Fig. 25) will be omitted. The difference from the tenth embodiment is that it includes a rotational frequency estimation unit 21 for estimating the rotational frequency of the rotating machine 30, and a characteristic frequency component extraction unit 23 for extracting the magnitude of the rotational frequency and its multiple frequency components, and an anomaly degree calculation unit 17 calculates and diagnoses the degree of anomaly using both the degree of anomaly calculated from the statistics obtained by the statistical quantity sequential calculation unit 16 and the degree of anomaly calculated from the waveform extracted by the characteristic frequency component extraction unit 23.

[0112] The processing of the rotational frequency estimation unit 21 and the feature frequency component extraction unit 23 is the same as in the sixth embodiment, and therefore a description thereof will be omitted. The processing of the anomaly degree calculation unit 17 calculates a first anomaly degree related to the fundamental period of the phase current, a second anomaly degree related to the current vector norm or torque current, and a third anomaly degree related to the frequency component obtained by the feature frequency component extraction unit 23 (step S1308). Then, an anomaly diagnosis is performed using each anomaly degree. Instead of calculating the anomaly degrees individually, a single anomaly degree integrated by clustering from the beginning may be calculated. By performing as in this embodiment, it is possible to increase the number of abnormal phenomena that can be detected.

[0113] In the seventh to thirteenth embodiments, the fundamental period of the phase current as the analysis target quantity is calculated as an example. However, the present invention is not limited to this, and the same effect can be obtained by replacing the phase current fundamental period calculation unit with a phase current fundamental frequency calculation unit that calculates a fundamental frequency that is the reciprocal of the fundamental period. EXAMPLES

[0114] 32 is a schematic diagram of a pump incorporating a power conversion device according to any one of the first to thirteenth embodiments and a rotating machine system using the same. A motor connected to the power conversion device is equipped with an impeller and sucks and pumps liquid such as water. It is preferable to perform an abnormality diagnosis during a time period when the motor is rotating at a constant speed and load, so the diagnosis is performed, for example, at night when the amount of water pumped is relatively stable during the day. EXAMPLES

[0115] 33 is a schematic diagram of an air compressor incorporating a power conversion device according to any one of the first to thirteenth embodiments and a rotating machine system using the same. In an air compressor, there is usually a time period immediately after startup during which the motor is in substantially no-load operation, and it is therefore desirable to perform diagnosis at that time. EXAMPLES

[0116] 34 is a schematic diagram of a conveying table incorporating the power conversion device according to the first to thirteenth embodiments and a rotating machine system using the same. In the conveying table, there is usually a time period during which the motor operates at a constant speed and load between workpiece conveying processes, so it is desirable to perform diagnosis at that timing. EXAMPLES

[0117] 35 is a schematic diagram of a machine tool, particularly a servo press or an injection molding machine, incorporating the power conversion device according to the first to thirteenth embodiments and a rotating machine system using the same. In these machine tools, there is usually a time period during which the motor is operating at almost no load before and after applying pressure to the workpiece, so it is desirable to perform the diagnosis at that timing.

[0118] The power conversion device 10 in the above embodiment can be applied to power conversion devices such as general-purpose inverters, servo amplifiers, DCBL controllers, etc. In addition to being applicable to rotating machine systems incorporating the above-mentioned power conversion device, the above embodiment can also be applied to wind power generation systems, hybrid automobiles, electric automobiles, fuel cell vehicles, railway cars, etc. [Explanation of symbols]

[0119] 10 Power conversion device 11 Main circuit 12a, 12b, 12c Current Sensors 13 Current measurement section 14 Current vector calculation section 15 Analysis target DC amount calculation section 16 Statistics sequential calculation part 17 Abnormality calculation part 18 Output section 19 Addition residual storage section 20 Power supply 21 Rotational frequency estimation unit 22 Feature waveform extraction unit 23 Feature frequency component extraction unit 24 phase current basic cycle calculation section 30 Rotating Machine 31 Bearings 40 load machine

Claims

1. a current measuring unit that measures a current flowing through a rotating machine; an analyte amount calculation unit that calculates an analyte amount from the current measured by the current measurement unit; A statistical quantity for sequentially calculating a statistical quantity relating to the distribution of the analysis target quantity calculated by the analysis target quantity calculation unit A sequential calculation unit; an anomaly calculation unit that calculates an anomaly degree from the statistics obtained by the statistics sequential calculation unit, The statistical value sequential calculation unit The residual error generated in the calculation is stored in the residual error storage unit. The statistical sequential calculation unit stores the additive residual generated during sequential calculation in the additive residual storage unit and updates it each time, and the power conversion device first adds a tiny value to the additive residual stored in the additive residual storage unit, adds it to a partial sum, calculates the increment of the partial sum, and updates the additive residual.

2. 2. The power conversion device according to claim 1, The statistical value sequential calculation unit deriving a second statistic regarding the distribution of the statistic; The abnormality degree calculation unit The power conversion device calculates an abnormality degree from the second statistical amount.

3. 2. The power conversion device according to claim 1, The analysis target amount calculation unit A power conversion device that calculates a fundamental period of a phase current as the analysis quantity.

4. The power conversion device according to claim 3, A power conversion device that calculates the fundamental period of the phase current from an interval between times when the phase current becomes zero.

Citation Information

Patent Citations

  • Abnormality determination apparatus, control method, automobile, and program

    JP2006331300A

  • Controller of rotary machine

    JP2012257360A

  • Diagnostic equipment of electric motor

    JP2016195524A

  • Electric motor diagnosis device and switching device

    WO2014156386A1