WO2019186909A1

WO2019186909A1 - Diagnosis device and diagnosis method

Info

Publication number: WO2019186909A1
Application number: PCT/JP2018/013287
Authority: WO
Inventors: 哲司加藤; 啓明小島; 牧　晃司; 岩路　善尚
Original assignee: 株式会社日立製作所
Priority date: 2018-03-29
Filing date: 2018-03-29
Publication date: 2019-10-03
Also published as: DE112018005613T5; JPWO2019186909A1; JP6801144B2

Abstract

This diagnosis device for diagnosing the state of a rotating machine that operates using a three-phase AC power supply comprises: a current measurement unit for acquiring simultaneous instantaneous values of physical data for a plurality of phases of the three-phase AC; a conversion unit for discretizing the instantaneous values acquired by the current measurement unit; an integration unit for determining, from the discretized values discretized by the conversion unit, the appearance densities of each combination of the discretized values for the physical data instantaneous values for each phase; an accumulation unit for accumulating appearance density data indicating the appearance densities of each combination determined by the integration unit; and a diagnosis unit for diagnosing the state of the rotating machine on the basis of the appearance density data accumulated by the accumulation unit.

Description

Diagnostic device and diagnostic method

The present invention relates to a technique for diagnosing equipment and facilities that operate with a three-phase AC power source.

If a rotating machine such as a motor (electric motor) or generator built into a production facility suddenly fails, unscheduled repair or replacement of the rotating machine is required, resulting in a reduction in the operating rate of the production facility or a review of the production plan. Is required. Similarly, when a power conversion device or cable connected to the rotating machine breaks down, unplanned repair work or replacement work is required, which requires a reduction in the operating rate of the production facility or a review of the production plan.

In order to prevent sudden failure of the rotating machine system (rotating machine and its associated equipment (cable, power converter)), stop the rotating machine system as appropriate and diagnose it offline to understand the deterioration. Sudden failure can be prevented to some extent. However, since the diagnosis is performed off-line, it is necessary to stop the rotating machine system, resulting in a reduction in the operating rate of the production facility. Some types of deterioration become apparent only when voltage is applied. Therefore, there is a need for diagnosing the state of a rotating machine based on current information of the rotating machine system.

There is Motor Current Signature Analysis (MCSA) as a method for diagnosing equipment and facilities based on current information of a rotating machine system. According to MCSA, specific frequency components corresponding to various factors such as rotor bar damage, rotor eccentricity, stator iron core damage, winding short-circuit, bearing deterioration, etc. on the current frequency spectrum. By detecting, failure or deterioration can be detected.

In Patent Document 1, especially in bearing diagnosis, a vibration sensor is installed at two locations, vibration sensor data is acquired from these vibration sensors, and an instantaneous value of each vibration sensor data is drawn on each axis. There is disclosed a method for determining an abnormality from a temporal change in the trajectory inclination or radius of a figure.

JP 2000-258305 A

However, the techniques disclosed in MCSA and Patent Document 1 have the following problems.

MCSA is required to detect specific frequency components with high accuracy. For this purpose, it is necessary to measure the current value for a long time at a high sampling rate. To perform long-time measurement at a high sampling rate, an expensive data logger is required, which increases diagnostic costs.

Also, there are motor failures that appear only for a short time. In order to detect the phenomenon without missing, it is necessary to measure the current over a long period of time at a high sampling rate and accumulate data. Therefore, the amount of data to be stored becomes enormous, and a terminal for storing data requires a large-capacity storage device, which increases costs. Also, if the amount of stored data becomes enormous, if the data is transferred via a communication path to send it to the cloud or a local terminal, it can occupy a large bandwidth of the communication path and hinder communication with other devices There is sex.

Further, in the technique disclosed in Patent Document 1, since the motor failure diagnosis is performed based on the data obtained by the vibration sensor, the position where the vibration change is sensitive to the motor failure, for example, the upper part of the motor, It is necessary to attach a vibration sensor, and the installation position of the vibration sensor is limited. In addition, the cost of diagnosis increases because a vibration sensor and an expensive data logger having a sampling rate sufficient to obtain the trajectory of the Lissajous figure are necessary.

An object of the present invention is to provide a technique for efficiently diagnosing an object to be diagnosed that operates with three-phase alternating current while suppressing the sampling rate and data amount of current measurement.

A diagnostic device according to one aspect of the present invention is a diagnostic device for diagnosing a state of a diagnosis target that operates with a three-phase AC power supply, and acquires an instantaneous value of physical data at the same time in a plurality of phases of the three-phase AC. And the conversion unit that discretizes the instantaneous value acquired by the acquisition unit, and the discrete density of the instantaneous value of each phase of the physical data of each phase from the discrete value obtained by discretization of the conversion unit, the appearance density of the combination Based on the accumulating unit to be obtained, the accumulating unit that accumulates the appearance density data indicating the appearance density for each combination obtained by the accumulating unit, and the diagnosis that diagnoses the state of the diagnosis target based on the appearance density data accumulated in the accumulating unit A section.

According to one aspect of the present invention, it is possible to efficiently diagnose the state of a diagnosis target that operates with three-phase alternating current while suppressing the sampling speed and data amount of current measurement.

It is a block diagram of the comparative example of a diagnostic apparatus. 1 is a block diagram of a diagnostic device according to Embodiment 1. FIG. It is a figure which shows the data amount accumulate | stored in a storage part by the discretization in the conversion part shown in FIG. It is a figure which shows a current waveform in case a spectrum appears in the frequency 1Hz away from the fundamental frequency 50Hz of U phase current. FIG. 5 is a diagram showing a swell waveform generated by the current waveform shown in FIG. 4. It is a figure which shows the current waveform of the U phase of a normal state. It is a figure which shows the current waveform of the W phase of a normal state. It is a figure which shows the current waveform of the U phase in case a sideband wave generate | occur | produces. It is a figure which shows the current waveform of the W phase in case a sideband wave generate | occur | produces. FIG. 10 is a diagram illustrating a result of analyzing the degree of abnormality of the matrix A and the matrix B based on the current waveforms illustrated in FIGS. 6 to 9 with respect to the matrix C defined as a normal state. It is a block diagram of the diagnostic apparatus by Example 2. It is a figure which shows the abnormality degree of the matrix A obtained from the current waveform of the rotary machine of the normal state acquired by the control command A. It is a block diagram of the diagnostic apparatus by Example 3. FIG. It is a block diagram of the diagnostic apparatus by Example 4. It is a block diagram of the diagnostic apparatus by Example 5. FIG.

In the following, the present invention will be described with reference to the drawings by way of a plurality of examples. The following embodiments are merely illustrative examples, and the present invention is not limited to the following embodiments.

In the failure of a rotating machine system including a rotating machine such as an electric motor (motor) or a generator, a cable attached to the rotating machine, and a power conversion device, the location of the failure and the cause of the failure are diverse. For example, insulation deterioration, bearing deterioration, short circuit, disconnection, water immersion, etc. can be considered. Moreover, the electric motor is often installed for a long time in a harsh environment, and a diagnostic technique according to the installation condition is required.

FIG. 1 is a diagram showing a comparative example of a diagnostic device. In this comparative example, as shown in FIG. 1, the state of the rotating machine 3 connected to the power source 1 via the cable 2 is diagnosed. In the diagnostic device of this comparative example, the current measuring unit 4 acquires two-phase current data via the

current sensors

10a and 10b attached to the cable 2, and the acquired current data is obtained from the spectrum obtained by Fourier transform. After being stored in the storage unit 8 as a value of the specific frequency spectrum, the diagnosis unit 9 diagnoses the state of the rotating machine 3 based on the current data stored in the storage unit 8.

In the diagnostic apparatus configured as described above, since the change in the specific frequency spectrum is measured by Fourier transform, it is necessary to perform continuous measurement at a constant sampling rate. Therefore, it is necessary to increase the capacity of the memory for temporarily storing the measured data or increase the communication speed with the device for storing the data, and an expensive device is required.

In addition, since there is a phenomenon that appears only for a short time depending on the failure of the rotating machine 3, if a long time measurement is performed at a high sampling rate in order not to miss the state caused by the failure, the amount of accumulated data becomes enormous and expensive. Data storage terminal was necessary. In addition, when the amount of accumulated data becomes enormous, an expensive communication device that supports large-capacity communication is required to transmit data to the cloud or a local terminal.

The inventors discretize the intermittent sensor values for two phases among the load current values of the rotating machine with the number of measurement bits for each phase, and depending on the combination of instantaneous values of each phase and their appearance density Considered to make a diagnosis.

Since current sensor data acquired from a three-phase motor that is a three-phase rotating machine has a 120 degree deviation from each other, when two-phase current sensor data is combined, the Lissajous figure has an inclined elliptical shape. The inventors have noted that when the two-phase current is ideal perfect continuous sine wave data, the elliptical shapes of the first period and the second period completely overlap. The measured waveform deviates from the ideal sine wave, and the sampling interval is a finite value. Therefore, the probability of measuring a point with exactly the same phase is low, but it can be obtained at a predetermined time obtained intermittently. By accumulating the data for the period and accumulating the combination of instantaneous values and their appearance density, the data can be accumulated in a form that reduces the data amount compared to accumulating the two-phase waveform as time series data. In addition, by superimposing data for multiple cycles, even if the sampling interval is not constant or the sampling speed is slow, the combination of instantaneous values and the appearance density change the operating conditions of the diagnosis target. If there is no match, the equipment such as an expensive data logger and an expensive data storage device can be omitted. The important point here is the time synchronism of the first phase and the second phase. The first phase and the second phase need to be the same time or at an arbitrarily designed constant interval. The fluctuation of the measurement interval between the second phase directly leads to a decrease in diagnostic accuracy. In order to perform measurement at the same time or at regular intervals, it can be realized by using a device having excellent real-time characteristics such as a microcomputer.

The diagnostic device of the present embodiment is a diagnostic device for diagnosing the state of a diagnosis target that operates with a three-phase AC power source, and acquires an instantaneous value of physical data at the same time in a plurality of phases of a three-phase AC; A conversion unit that discretizes the instantaneous value acquired by the acquisition unit, and an integration unit that obtains the appearance density of the combination for each combination of discrete values of the instantaneous value of the physical data of each phase from the discrete value obtained by discretization of the conversion unit And an accumulating unit that accumulates appearance density data indicating the appearance density for each combination obtained by the integrating unit, and a diagnosis unit that diagnoses the state of the diagnosis target based on the appearance density data accumulated in the accumulating unit. It is to be prepared.

Also, the diagnostic device configured as described above can be incorporated into a rotating machine system. In particular, it is preferable to share a current sensor, a current measurement unit, and the like that are provided for controlling the rotating machine in the power converter, because the number of parts can be reduced. Further, a plurality of rotating machines may be connected to the power conversion device.

Furthermore, by classifying and evaluating the current data for each frequency condition of the current value measured by the current sensor or current measuring unit, even if the driving condition of the rotating machine changes, the rotation can be performed properly. It is possible to diagnose the state of the machine system.

Hereinafter, a more specific embodiment of the above-described diagnostic apparatus will be described.

FIG. 2 is a block diagram of the diagnostic apparatus according to the first embodiment.

As shown in FIG. 2, the diagnostic apparatus in the present embodiment diagnoses the state of the rotating machine 3 to be diagnosed that is electrically connected to the power source 1 via the cable 2. A conversion unit 5, a measurement bit number input unit 6, an integration unit 7, an accumulation unit 8, and a diagnosis unit 9. The diagnostic device may include a processor and a memory, and the processor may use the memory to execute a software program that defines the operation of each unit.

A three-phase AC voltage is output from the power source 1. There are two types of three-phase AC voltage output: when the inverter switching element is operated and controlled so that the motor speed and torque are at the desired values, and when the commercial power supply is connected directly Can be considered.

The current measurement unit 4 may be called an acquisition unit. The current measuring unit 4 obtains instantaneous values of current data as physical data at the same time in a plurality of three-phase alternating currents flowing through the cable 2 via the

current sensors

10 a and 10 b attached to the cable 2. At this time, the current measuring unit 4 acquires an instantaneous value of current data while maintaining time synchronism between a plurality of phases at an arbitrary interval for an arbitrary period.

The conversion unit 5 discretizes the instantaneous value acquired by the current measurement unit 4 with the set bit number A set by the measurement bit number input unit 6.

Integrating section 7, the discrete values by discretization of the conversion unit 5, (power of 2 A) 2 ^A row and column corresponding to a respective phase × 2 has a size of ^A (2 A squared), and line Each component of the column generates a matrix that represents the density of discrete combinations of instantaneous values of current data of each phase, thereby generating the density of combinations of combinations of discrete values of instantaneous values of current data of each phase. Ask for.

The accumulating unit 8 accumulates the appearance density data indicating the appearance density for each combination obtained by the accumulating unit 7.

The diagnosis unit 9 diagnoses the state of peripheral devices such as the rotating machine 3 and the power converter connected to the rotating machine 3 based on the appearance density data stored in the storage unit 8.

Hereinafter, a diagnostic method for diagnosing the state of the rotating machine 3 by the diagnostic device configured as described above will be described.

First, the current measurement unit 4 acquires current data in two phases of the three-phase alternating current flowing through the cable 2 via the

current sensors

10 a and 10 b attached to the cable 2. At that time, the current measuring unit 4 acquires an instantaneous value of current data while maintaining time synchronism between a plurality of phases at an arbitrary interval for an arbitrary period. The current sensor acquired by the current measuring unit 4 The

current data

10a and 10b do not necessarily have to be a constant sampling interval, and need not necessarily be continuous measurements. Further, it is preferable that the interval between the acquisition of the current data via the current sensor 10a and the acquisition of the current data via the current sensor 10b is constant allowing for some variation. By applying a measurement device that excels in real-time processing, a certain data acquisition interval that allows a certain variation can be obtained. As such a measuring device, for example, a measuring device using a microcomputer can be used.

Further, by making the interval between the acquisition of current data via the current sensor 10a and the acquisition of current data via the current sensor 10b constant, there is no need for continuous measurement. For example, the current measuring unit 4 can be designed such that after a certain amount of data is accumulated in the memory of the microcomputer, the accumulated data is stored in a storage device, and then the memory of the microcomputer is cleared and data accumulation is resumed. A system using a general-purpose current sensor or memory can be provided.

Next, the conversion unit 5 discretizes the instantaneous value acquired by the current measurement unit 4 with the set bit number A set by the measurement bit number input unit 6, and the integration unit 7 determines the discrete value of the conversion unit 5. The appearance density of each combination of the discrete values of the instantaneous values of the two-phase current data is obtained from the discrete values obtained by the conversion, and the appearance density data indicating the appearance density for each combination is stored in the storage unit 8. Become. At this time, the conversion unit 5 discretizes the instantaneous value of the two-phase current data with a predetermined number of bits A, and the integration unit 7 has 2 ^A (2 to the power of 2) × 2 in which each row and column corresponds to a phase. ^{If the} application density is determined by generating a matrix having a size of ^A (2 to the power of A) and each component in the row and column representing the appearance density of a combination of discrete values of instantaneous values of current data of each phase The amount of current data for each phase can be compressed to a size of 2 ^A × 2 ^A.

Here, when the control pattern for the rotating machine 3 does not change and the rotating machine 3 operates at a constant fundamental frequency, the amount of data by discretization when an ideal sine wave current flows through the rotating machine 3 is obtained. Will be described. Note that the number of measurement bits of the data logger was 8 bits.

FIG. 3 is a diagram showing the amount of data stored in the storage unit 8 by discretization in the conversion unit 5 shown in FIG.

Suppose the data amount is 1 when measuring two-phase current for 200 seconds at a sampling rate of 200 Hz.

As technique 1, the waveform that is current data acquired by the current measurement unit 4 is discretized by the number of bits 8 set by the measurement bit number input unit 6 in the change unit 5, and the integration unit 7 sets 28 × 28 The appearance density of current values discretized by 8 bits, that is, 1 to 257, was obtained for each matrix element of the size matrix. For example, in a measurement time of 200 seconds, a value obtained by dividing the number of times that the discretized current value of the current sensor 10a is 257 and the discretized current value of the current sensor 10b is 257 by 40000 measurement points. Asked. The measurement for 200 seconds does not need to be continuous, and if the control pattern for the rotating machine 3 does not change and the rotating machine 3 operates at a constant fundamental frequency, the measurement should be divided several times. Can do. As a result, data necessary for diagnosis can be acquired even with a low-speed measuring terminal with a small memory capacity.

In this case, the data amount was 0.81, and the data amount was reduced by 19%. As a result, data necessary for diagnosis can be acquired even with a low-speed measuring terminal with a small memory capacity. In this example, the value divided by the number of measurement points is stored in each matrix element of the matrix, but the number of appearances is saved, and separately, the number of measurement points is saved in a form linked to the file that stores the number of appearances. Also good. In particular, when measuring is divided into several times, error accumulation due to integration can be reduced by storing the number of appearances and the number of measurement points separately.

Also, as a technique 2, a 28 × 28 size matrix is not prepared in advance, but (the current value 1 of the current sensor 10a, the current value 1 of the current sensor 10a, the number of appearances), ( Data was stored in the form of a discretized current value 1 of the current sensor 10a, a discretized current value 2 of the current sensor 10a, the number of appearances),. In Method 2, the data finally obtained is data excluding the component of Method 1 whose matrix element value is 0, and it is not necessary to prepare a blank cell. Only the value of the component in which a numerical value other than zero is stored is accumulated, and the amount of accumulated data can be further reduced. The timing for excluding the 0 component is not particularly limited.

As a data accumulation method, (discrete current value 1 of current sensor 10a, discretized current value 1 of current sensor 10a, number of appearances), (discrete current value 1 of current sensor 10a, current) The data may be stored in the form of the discretized current value 2 of the sensor 10a, the number of appearances),... Once the matrix of the method 1 is constructed, and then the discretization of the current sensor 10a is performed. Current value 1, discrete current value 1 of current sensor 10a, number of appearances), (discrete current value 1 of current sensor 10a, discrete current value 2 of current sensor 10a, number of appearances),・ You may convert to the form

Thereafter, the diagnosis unit 9 diagnoses the state of the rotating machine 3 based on the appearance density data accumulated in the accumulation unit 8.

Here, the diagnosis method of the diagnosis unit 9 when a spectrum is generated at a specific frequency for some reason when the control pattern for the rotating machine 3 does not change and the rotating machine 3 operates at a constant fundamental frequency will be described. To do.

4 is a diagram showing a current waveform when a spectrum appears at a frequency 1 Hz away from the fundamental frequency 50 Hz of the U-phase current, and FIG. 5 is a diagram showing a swell waveform generated by the current waveform shown in FIG. It is.

For example, when a sideband of the fundamental frequency is generated due to deterioration in bearing deterioration, and as a result, as shown in FIG. 4, a spectrum appears for some reason at a frequency 1 Hz away from the fundamental frequency of the U-phase current, 50 Hz, As shown in FIG. 5, the U-phase current appears as a waveform having a 1 Hz period undulation.

In the diagnostic apparatus shown in FIG. 1, in order to accurately separate components of 50 Hz and 51 Hz from this waveform, when the current flowing through the cable 2 is measured at a frequency of 200 Hz, data of 20000 points are continuously obtained for at least 200 seconds. It is necessary to measure. Furthermore, the frequency of appearance of sidebands increases with the progress of deterioration, and the frequency is low at the initial stage of deterioration, so that it is necessary to measure data for a long time. Therefore, it is necessary to apply a special and expensive measuring device in order to separate with high accuracy.

On the other hand, in the diagnostic apparatus shown in FIG. 2, the current measurement unit 4 acquires current values of two or more phases while maintaining time synchronism between phases at an arbitrary interval, and at the conversion unit 5, The current data of two or more phases is discretized for each phase with the number of bits set in the measurement bit number input unit 6, and the accumulating unit 7 sets the number of appearances or the appearance density of the discretized current data in each matrix element. Then, the diagnosis unit 9 compares the data accumulated as application density data in the storage unit 8, and the rotation unit 3 or the power conversion device connected to the rotary unit 3 or the like from the change. By diagnosing the state of the peripheral devices, it is possible to diagnose the rotating machine 3 without increasing the sampling speed and the data amount.

Also, in order to acquire high frequency information, it is desirable to measure asynchronously with the switching timing of the power converter.

When the sampling rate is decreased, it is desirable that the current measurement unit 4 measures the current flowing through the cable 2 asynchronously with the fundamental frequency via the

current sensors

10a and 10b. Specifically, at a sampling rate having a period that is an integral multiple of the fundamental wave, a value of only a specific phase is always obtained, and there is a risk of a failure in the case of deterioration in which a change appears only in a specific phase. Therefore, it is desirable that the sampling rate is a frequency different from an integral multiple of the fundamental wave period. As a result, it is possible to acquire the same matrix as when data is acquired at a high sampling rate.

FIG. 6 is a diagram illustrating a U-phase current waveform in a normal state, and FIG. 7 is a diagram illustrating a W-phase current waveform in a normal state. FIG. 8 is a diagram showing a U-phase current waveform when a sideband wave is generated, and FIG. 9 is a diagram showing a W-phase current waveform when a sideband wave is generated. The sampling rate is 200 Hz, and the data for 1 second is enlarged in the data measurement time of 100 seconds.

In a normal 50 Hz sine wave without undulation, the U-phase current waveform is as shown in FIG. 6, the W-phase current waveform is as shown in FIG. 7, and the U-phase and W-phase currents are: They are 120 degrees out of phase with each other.

When the waveforms shown in FIGS. 6, 7, 8, and 9 are measured and acquired by the current measuring unit 4, the current measured by the converting unit 5 as shown in the description of the method 1 described above. The waveform is discretized with the number of bits 8 set by the measurement bit number input unit 6, and the accumulating unit 7 converts the current value discretized by 8 bits, that is, 1 to 257 into each matrix element of a 28 × 28 size matrix. Appearance densities were stored in matrix A and matrix B, respectively. The matrix A is based on the waveforms shown in FIGS. 6 and 7, and corresponds to the normal state matrix, and the matrix B is based on the waveforms shown in FIGS. Corresponding to

In the diagnosis unit 9, the matrix A and the matrix B are compared with the matrix C set in advance as a normal state by learning, and the normal state is calculated by calculating how close the matrix A and the matrix B are to the matrix C. The degree of deviation, that is, the degree of abnormality can be examined. As the deterioration progresses, the variation of the current value increases, and accordingly, the appearance density of the region having a high appearance density is lowered and distributed to the region having a low appearance density. As a result, the value of each matrix element in the matrix changes, the number of matrix elements having a value of 0 decreases, and changes from the matrix defined as the normal state.

The diagnosis unit 9 outputs the diagnosis result after performing the above diagnosis. Means for transmitting the diagnosis result to the user can be selected as appropriate, and methods for transmitting to the user include display on the display, lighting of the lamp, notification by e-mail, and the like. The contents are also (1) a method for displaying a matrix to be diagnosed on the screen and allowing the user to determine whether or not it is compatible, (2) a method for quantifying the difference in the matrix to be diagnosed in some way and telling the user, (3) A method of notifying the user when a predetermined threshold is exceeded can be considered.

Here, application of machine learning can be considered as a method of quantifying the difference of the matrix to be diagnosed in (2) above. The machine learning algorithm may be selected so that the difference in the matrix to be diagnosed becomes clear. For example, the Bhattacharyya coefficient (Bhattacharyya distance) Z of the matrix A to be diagnosed with respect to the matrix C defined as a normal state is the matrix A Where A _{i, j} is the matrix element and C _{i, j} is the matrix element of the matrix C, the following equation is defined.

In the above equation, if the Bhattacharyya coefficient Z is large, it can be diagnosed that it is far from the normal state, that is, the deterioration is progressing, and if the coefficient Z is small, it can be diagnosed that the normal state. It can be said that the Bhattacharyya coefficient Z represents the degree of abnormality. In this way, the matrix data is held in the storage unit 8, and the diagnosis unit 9 compares the matrix held in the storage unit with the reference matrix that is predetermined as indicating the normal state. By diagnosing the normality of the state, it is possible to easily know how much the state is different from the normal state and to diagnose the normality.

Here, when using a matrix in which the number of appearances is stored in each matrix element, it is necessary to match the number of measurement points of the normal state and the matrix to be diagnosed. In this case, a local subspace method can be cited as a diagnostic technique. In the local subspace method, for each matrix element of the matrix to be diagnosed, two closest points are selected from the two-dimensional space defined by the matrix row and column defined as normal, and the two points are connected. In this method, the degree of deterioration is defined by the distance between the straight line and the point to be diagnosed. In addition, it may be effective to weight the appearance density of normal values, especially when the data can be recognized as an abnormal state, the discriminant is adjusted so that the matrix of the normal state and the abnormal state can be easily distinguished. It is desirable.

In addition, when using the local subspace method, in addition to the method of calculating the distance of all matrix elements to be diagnosed and quantifying the change in the matrix, the average value of the distances of all points to be diagnosed and specifying An arbitrary evaluation method can be selected in accordance with the variation error of the current waveform, such as quantification based on the distance of only the phase point. When priority is given to the calculation speed, a clustering method such as vector quantization clustering or K-means clustering can be used. In addition, a technique called a deep neural network, which is a method for automatically finding feature quantities based on a large amount of data, can be applied.

FIG. 10 is a diagram illustrating a result of analyzing the degree of abnormality of the matrix A and the matrix B based on the current waveforms illustrated in FIGS. 6, 7, 8, and 9 with respect to the matrix C defined as a normal state.

The matrix A based on the current waveform shown in FIGS. 6 and 7 and the matrix B based on the current waveform shown in FIGS. 8 and 9 can analyze the degree of abnormality from the Bhattacharyya distance Z according to the above formula. Since the Bhattacharyya distance Z of the matrix A does not exceed a preset threshold as shown in FIG. 10, the diagnosis unit 9 determines that it is normal. On the other hand, the Bhattacharyya distance Z of the matrix B exceeds the preset threshold value as shown in FIG.

As described above, it is possible to detect an increase in the degree of abnormality before the rotating machine 3 fails depending on whether or not the Bhattacharyya distance Z exceeds a preset threshold value.

As described above, since the appearance density of combinations of instantaneous values of current data of each phase is accumulated and accumulated, the amount of data to be stored can be suppressed. Further, since the state of the rotating machine 3 to be diagnosed is diagnosed based on the accumulated data obtained by integrating the appearance density of the combination of instantaneous values of the current data of each phase, it is not necessary to measure the current data at a high sampling rate.

In the present embodiment, the case where a spectrum is generated at a specific frequency has been described. However, even when the type of deterioration does not generate a spectrum at a specific frequency, the deterioration causes the current to be generated by the load of the rotating machine 3 or the impedance change. Since some change appears, the abnormality can be detected by the diagnostic apparatus shown in FIG. 2 and the diagnostic method described above. Specifically, the deterioration that does not appear as a change in the spectrum of the specific frequency is specifically a deterioration other than the deterioration in which the change appears as a peak of the specific frequency that can be detected by the MCSA, that is, grease deterioration, thermal deterioration of the insulating material, and moisture absorption. Things are assumed.

In addition, according to the diagnostic device of the present embodiment, it is possible to diagnose a motor system including equipment that is electrically or mechanically connected to the rotating machine 3, such as a cable, a power converter, and a load, in addition to the rotating machine 3. is there. In a motor system including peripheral devices connected to the rotating machine 3, even if a peripheral device other than the rotating machine 3 fails or deteriorates, the current flowing through the rotating machine 3 changes due to the impedance and load of those devices changing. Therefore, it is possible to detect deterioration by this method.

Next, the case where the control pattern for the rotating machine 3 changes will be described. When the control pattern for the rotating machine 3 changes, the fundamental frequency, the carrier frequency, and the like of the rotating machine 3 change, so the diagnosis apparatus shown in FIG. 1 sometimes diagnoses this as abnormal. Further, when learning is performed as a normal state including a state in which the control pattern is changed, a change due to deterioration may be overlooked and reported. Therefore, it is desirable to diagnose normality for each control pattern.

FIG. 11 is a block diagram of a diagnostic apparatus according to the second embodiment.

As shown in FIG. 11, the diagnostic apparatus according to the present embodiment is different from that shown in FIG.

The command unit 13 gives a control command indicating control pattern information to the integrating unit 7. The control command can be any command value that can be output by the power source 1 such as a voltage command value, current command value, excitation current command value, torque current command value, speed command value, frequency command value, or any equivalent value You can choose to.

The accumulating unit 7 obtains the appearance density of each combination for each control command to the rotating machine 3, and the accumulating unit 8 accumulates the appearance density data for each control command.

Hereinafter, a case where the state of the rotating machine 3 is diagnosed under the two conditions of the control command A and the control command B having different voltage command values and frequency command values as control commands will be described. The control command A corresponds to a fundamental current frequency of 50 Hz, and the control command B corresponds to a fundamental current frequency of 100 Hz.

The integration unit 7 integrates the appearance density obtained by the conversion unit 5 for each current value distributed to the control command A and the control command B based on the fundamental frequency input to the command unit 13. Then, after data of an arbitrarily determined period or score is accumulated, the data is accumulated in the accumulation unit 8 as appearance density data.

Thereafter, the diagnosis unit 9 compares the appearance density data stored in the storage unit 8 with the data defined as normal acquired in the same state of the control command information to determine the degree of abnormality of the target data, and the normality of the rotating machine 3 Diagnose sex.

FIG. 12 is a diagram showing the degree of abnormality of the matrix A obtained from the current waveform of the rotating machine 3 in the normal state acquired by the control command A. The matrices α and β are matrices defined by the current waveform of the rotating machine 3 in the normal state measured by the control commands A and B, respectively.

As shown in FIG. 12, when the rotating machine 3 in the normal state is diagnosed using the matrix α based on the current data of the rotating machine 3 in the normal state using the matrix α, the Bhattacharyya distance Z of the matrix A is shown in FIG. Thus, since it does not exceed the preset threshold value, the diagnosis unit 9 determines that it is normal.

On the other hand, when the normal state rotating machine 3 is diagnosed using the matrix β of the matrix A based on the current data of the normal state rotating machine 3, the Bhattacharyya distance Z of the matrix A is set in advance as shown in FIG. Therefore, the diagnosis unit 9 determines that the rotating machine 3 is abnormal even though the rotating machine 3 is in a normal state.

As described above, in this embodiment, the control input to the command unit 13 is performed in order to diagnose the state of the rotating machine 3 based on the matrix composed of the U-phase and W-phase currents classified as having the same control command. By classifying the current waveform based on the command and accumulating the appearance density for each classified data, the control pattern and the current information are combined to improve the diagnostic accuracy. Thereby, a suitable diagnosis can be performed for each control state.

In addition, in the integrating | accumulating part 7, it is not necessary to use all the command values which the power supply 1 can output, and only the command value with high sensitivity with respect to degradation of a detection target can be used. The command value having high sensitivity may be selected so that the degree of similarity between the matrix in the deteriorated state and the matrix in the normal state is compared and the difference between the deteriorated state and the normal state is easily visible. In addition, if the number of conditions is increased, the amount of data increases by the amount classified according to the conditions even if the current data has the same measurement time. Therefore, the number of command values to be used should be limited by the trade-off between diagnostic accuracy and data volume. Can do.

If the information of the command unit 13 cannot be referred to as in the diagnostic device shown in FIG. 11, the characteristic frequency of the current waveform of the rotating machine 3 can be substituted.

FIG. 13 is a block diagram of a diagnostic apparatus according to the third embodiment.

As shown in FIG. 14, the diagnostic apparatus in the present embodiment is different from that shown in FIG. 2 in that it has a frequency extraction unit 12.

When recognizing the control pattern of the rotating machine 3 based on the characteristic frequency of the current waveform of the rotating machine 3, the output of the current measuring unit 4 is branched and input to the frequency extracting unit 12, as shown in FIG. The frequency extraction unit 12 extracts a characteristic frequency such as at least one of the fundamental frequency and the carrier frequency, and then the integration unit 7 obtains an appearance density of the combination for each combination for each measurement condition with the same characteristic frequency. In the storage unit 8, appearance density data is stored for each characteristic frequency.

As described above, in this embodiment, the output of the current measuring unit 4 is branched in order to diagnose the state of the rotating machine 3 based on the matrix composed of the U-phase and W-phase currents classified as the same control command. Then, the characteristic frequency is extracted by the frequency extraction unit 12, the current waveform is classified based on the characteristic frequency, and the appearance density is integrated for each classified data, thereby combining the control pattern and the current information to obtain diagnosis accuracy. Improved. Thereby, a suitable diagnosis can be performed for each control state.

FIG. 14 is a block diagram of a diagnostic apparatus according to the fourth embodiment.

As shown in FIG. 14, the diagnostic apparatus in this embodiment is different from that shown in FIG. 2 in that a plurality of rotating machines 3-1 to 3-n are connected to one power source 1. Is.

In this embodiment, the load currents of all the rotating machines 3-1 to 3-n were measured by the

current sensors

10a and 10b attached to the cable 2 connecting the power source 1 and the rotating machines 3-1 to 3-n. Diagnose using the results. The diagnosis method is equivalent to the case where one rotating machine is connected, and the diagnosis is performed by using the measured command data as the distribution of the Lissajous figure together with the classification using the control command value and the reflection of the command value information.

In addition, since the current data of the plurality of rotating machines 3-1 to 3-n is used for diagnosis as one load current, the change in the current waveform of the deteriorated rotating machine is diluted by the current waveforms of the other rotating machines. Will end up. Therefore, it is desirable to evaluate the degree of abnormality by machine learning, particularly the local subspace method.

FIG. 15 is a block diagram of a diagnostic apparatus according to Embodiment 56.

As shown in FIG. 15, the diagnostic apparatus in the present embodiment measures each of the three-phase load currents supplied from the power source 1 to the rotating machine 3 by the current sensors 10a to 10c, as shown in FIG. The point to do is different.

In the present embodiment, the current measurement unit 4 obtains instantaneous values of current data at the same time in the three phases of the three-phase alternating current, and the integration unit 7 obtains the three-phase from the discrete values obtained by the discretization of the conversion unit 5. The appearance density of each combination of discrete values of instantaneous values of current data is obtained. In this case, when comparing the distributions, it is possible to obtain and evaluate the distribution of the Lissajous figure with a plurality of combinations of two selected from three or more current sensors. Further, a multidimensional matrix (tensor) can be defined by three or more current sensors, and motor deterioration can be evaluated from changes in the tensor.

In the case of the three-phase AC rotating machine 3, the current value measured by the current sensors 10a to 10c is zero measured by clamping the three-phase components in addition to the U-phase, V-phase, and W-phase load currents. Diagnosis of phase current, two-phase current measured by clamping two arbitrarily selected phases, leakage current measured by clamping both the winding start and end of winding of the motor, and current flowing from the motor to the ground A current that increases sensitivity is arbitrarily selected, and a current sensor may be installed at that position.

Thus, since the state of the rotating machine 3 is diagnosed based on the three-phase AC three-phase current data, a more accurate diagnosis is possible.

DESCRIPTION OF SYMBOLS 1 ... Power supply, 2 ... Cable, 3 ... Rotating machine, 4 ... Current measurement part, 5 ... Conversion part, 6 ... Measurement bit number input part, 7 ... Accumulation part, 8 ... Accumulation part, 9 ... Diagnosis part, 10a, 10b , 10c ... current sensor, 12 ... frequency extraction unit, 13 ... command unit

Claims

A diagnostic device for diagnosing the state of a diagnostic target that operates with a three-phase AC power source,
An acquisition unit that acquires instantaneous values of physical data at the same time in a plurality of phases of the three-phase alternating current;
A conversion unit for discretizing the instantaneous value acquired by the acquisition unit;
From a discrete value obtained by discretization of the conversion unit, an integration unit for obtaining an appearance density of the combination for each combination of discrete values of instantaneous values of physical data of each phase;
An accumulating unit that accumulates appearance density data indicating the appearance density for each combination obtained by the integration unit;
Based on the appearance density data stored in the storage unit, a diagnosis unit that diagnoses the state of the diagnosis target;
A diagnostic device comprising:
The conversion unit discretizes an instantaneous value of the physical data of two phases with a predetermined number of bits A,
The accumulating unit has a size of 2 A (2 to the power of A) × 2 A (2 to the power of A) corresponding to each phase of the row and column, and each component of the row and column is physical data of each phase. Generates a matrix representing the density of occurrences of a combination of discrete values of instantaneous values of
The diagnostic device according to claim 1.
The storage unit holds the data of the matrix,
The diagnosis unit diagnoses the normality of the state of the diagnosis target by comparing a matrix held in the storage unit and a reference matrix that is predetermined as indicating a normal state,
The diagnostic device according to claim 2.
The diagnosis unit diagnoses the normality of the state of the diagnosis target based on a difference between the matrix accumulated in the accumulation unit and the reference matrix,
The diagnostic device according to claim 3.
The accumulation unit accumulates only the values of components in which numerical values other than zero are stored in the matrix,
The diagnostic device according to claim 2.
The diagnostic object includes a controllable rotating machine,
The integration unit obtains the appearance density of the combination for each combination for each control command to the rotating machine,
The storage unit stores the appearance density data for each control command.
The diagnostic device according to claim 1.
The diagnostic object includes a controllable rotating machine,
The integrating unit obtains the appearance density of the combination for each combination for each characteristic frequency that is changed by the control to the rotating machine,
The storage unit stores the appearance density data for each feature frequency.
The diagnostic device according to claim 1.
The characteristic frequency is at least one of a fundamental frequency and a carrier frequency of a current input to each phase of the rotating machine.
The diagnostic device according to claim 7.
The acquisition unit acquires an instantaneous value of physical data at the same time in the three phases of the three-phase alternating current,
The integration unit obtains an appearance density of the combination for each combination of discrete values of instantaneous values of three-phase physical data from discrete values obtained by discretization of the conversion unit.
The diagnostic device according to claim 1.
A diagnostic method for diagnosing the state of a diagnosis target that operates with a three-phase AC power source,
Obtain the instantaneous value of physical data at the same time in the multiple phases of the three-phase AC,
Discretizing the acquired instantaneous value,
From the discrete values by the discretization, obtain the appearance density of the combination for each combination of discrete values of the instantaneous value of the physical data of each phase,
Accumulating appearance density data indicating the appearance density for each of the obtained combinations,
Diagnosing the state of the diagnosis object based on the accumulated appearance density data;
Diagnosis method.