Method for reducing speed fluctuations of a permanent magnet excited electric motor

The method uses a learning device to record and compensate for synchronization fluctuations in permanent magnet motors, addressing sensor and gearbox-related issues, enhancing motor performance and reliability in high-precision applications.

DE102016122105B4Active Publication Date: 2024-11-21ABB (SCHWEIZ) AG
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
DE102016122105
Authority / Receiving Office
DE · DE
Patent Type
Patents
Current Assignee / Owner
Filing Date
2016-11-17
Publication Date
2024-11-21
Estimated Expiration
2036-11-17

AI Technical Summary

Technical Problem

Existing methods for reducing synchronization fluctuations in permanent magnet-excited electric motors, such as those used in robots and electric vehicles, are complex, require separate test measurements for each motor, and do not effectively compensate for fluctuations caused by systematic errors in sensors or gearboxes, leading to disruptive performance issues, especially at high precision and low speeds.

Method used

A method involving a learning device that performs test runs during motor operation to record measurement data, allowing a control unit to compensate for synchronization fluctuations by identifying and addressing the underlying causes, including sensor and gearbox-related disturbances, and is adaptable to changing magnetic properties without requiring separate test measurements for each motor.

Benefits of technology

This approach enables reliable compensation for all relevant disturbances, reduces torque ripple, extends motor availability, and prevents damage, while minimizing downtime and repair costs, making it suitable for series production and high-precision applications.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 00000000_0000_ABST
    Figure 00000000_0000_ABST
Patent Text Reader

Abstract

Method for reducing speed fluctuations of a permanent magnet excited electric motor (21) installed in a machine, characterized in that The following steps will be performed: - Using a learning device (10) that independently performs test runs of the motor (21) during operation of the machine and records measurement data that can be used to reduce synchronous fluctuations, - Adjusting a control unit (18) based on the measurement data such that it compensates for the speed fluctuation, and - Determining from the measurement data which component of the arrangement causes which speed fluctuation at which frequency, wherein a speed fluctuation caused by the motor (21) is decoupled from a speed fluctuation caused by a sensor by determining the motor-induced speed fluctuation and determining the sensor-induced speed fluctuation, wherein the speed fluctuation caused by the sensor is compensated first before a speed fluctuation caused by the motor (21) is compensated, the measurement data comprising a measured current (iq) or a correction current (iq, corr) and a motor angle (6) of the motor (21); wherein a speed fluctuation caused by the motor (21) is determined on the basis of the measured current (iq) or the correction current (iq, corr) by determining its amplitude and phase angle for each frequency of the measured current (iq) or the correction current (iq, corr), and a speed fluctuation caused by the sensor is determined using the sine and cosine of the motor angle (sin(6), cos(6)) of the motor (21) using a Fourier transform.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] The invention relates to a method for reducing speed fluctuations of a permanent magnet excited electric motor installed in a machine, according to the preamble of claim 1.

[0002] In many technical fields, electric motors are used in machines. These motors are essentially three-phase synchronous machines with permanent magnet excitation. Examples of such machines with electric motors include robots, especially powered robot joints, and electric vehicles.

[0003] In a permanent magnet three-phase synchronous motor, the rotating magnetic field of the three-phase winding causes the permanent magnet rotor to move. With the use of suitable controllers, such a motor exhibits the behavior of a DC motor. Such motors are used primarily in small drives. Using suitable gearboxes with a high gear ratio, high torques can be achieved even with small motors, which is particularly relevant in robotics.

[0004] Typically, the three-phase winding of such a motor is designed with three phases. Magnetization processes in the ferrous stator structure in the coils, as well as mechanical inaccuracies such as varying magnet strengths, differently shaped windings, or imprecisely placed magnets, can cause torque fluctuations, which manifest themselves as wow and flutter.

[0005] Malfunctions or errors, in particular systematic errors in the angle or current sensors as well as in the rotor bearing or in a connected gearbox, can also trigger periodic wow and flutter similar to those resulting from the torque fluctuations described above.

[0006] Such wow and flutter are disruptive when used in high-precision applications and at low speeds, as occurs in the case of robot joints, and are unsuitable when high precision is required.

[0007] Although motors with higher numbers of phases can be used to reduce torque fluctuations, this requires more effort and also causes significantly higher costs.

[0008] DE 10 2014 105 730 A1, which represents the closest prior art, shows a method in which the torque ripple of a motor is electronically controlled by applying correction data to the control current of the motor during its rotation.

[0009] The correction data are determined from mechanical data of the motor and stored in a memory so that they can be superimposed on the respective motor current depending on the position.

[0010] Several different data sets are stored in a function memory in the form of tables. The data sets are calculated taking into account influences resulting from the force and torque curve of the motor and, if applicable, any connected load.

[0011] The data sets can be retrieved position-dependently or time-dependently and linked to an input variable in order to be combined into position- and time-dependent instantaneous values.

[0012] The data sets that take the engine's force and torque curves into account are determined through measurement runs on a test bench to optimize or achieve the specified force and torque curves. Alternatively, the data sets can be calculated from a sample of measured parameters and / or characteristic curves.

[0013] In the process shown in DE 10 2014 105 730 A1, which is illustrated by the circuit diagram according to Fig. 1, is disadvantageous in that the data and / or parameters required to perform ripple compensation must first be determined in a test measurement on a suitable test bench. These are then stored in a memory.

[0014] Since the results of a test measurement on one engine cannot be transferred to another engine, even if the engine is of the same design, a separate test measurement must be carried out for each individual engine.

[0015] Another disadvantage is that the data obtained in the test can only be used to achieve the best possible reduction in torque ripple as long as the magnetic properties of the motor, which is excited by a permanent magnet (PM motor), do not change.

[0016] For example, as soon as a permanent magnet of the motor becomes demagnetized, even if only slightly, the data stored in the memory can no longer be used to achieve the best possible reduction of the torque ripple.

[0017] The data stored in the memory would then have to be adapted or updated to the changed magnetic properties of the motor using a new test measurement.

[0018] This procedure is extremely complex, as it would require the engine installed in a machine to be removed and then adapted to a test bench. Furthermore, the machine would be out of service for the duration of the test run due to the lack of an engine.

[0019] A further disadvantage of the prior art method is that only the torque fluctuations resulting from asymmetries in the electromagnetic structure of the PM motor and therefore periodically recurring are compensated, but not the speed fluctuations described above and caused by systematic errors in the angle or current sensors as well as in the rotor bearing or by a connected gearbox.

[0020] Simply reducing the periodically recurring torque fluctuations caused by the electromagnetic structure of the PM motor is not sufficient to achieve, for example, a jerk-free movement of a robot arm.

[0021] EP 2 235 818 B1 describes performing a calibration run to determine a phase difference between an external rotating field and an angle detected by a sensor. US 2016 / 0193941 A1 describes using Fourier analysis to obtain corrected position signals. Error correction of rotor position signals obtained from sensors is performed. EP 0 792 420 B1 is aimed at compensating for vibration forces. Against this background, it is proposed to determine pilot currents offline and store them as reference values ​​in table form. This table is then accessed during operation. DE 10 2010 062 273 A1 is based on a method in which the speed of a rotor is controlled. Although learning methods are also mentioned there, correction data are stored in a non-volatile memory.The invention is therefore based on the object of specifying a method which reduces as far as possible all of the speed fluctuations occurring in a permanent magnet excited electric motor installed in a machine, and which overcomes the disadvantages of the prior art and, in particular in the case of changed magnetic properties of a PM motor, reliably reduces the torque ripple of the motor as best as possible.

[0022] According to the invention, the above object is achieved by a method having the features of patent claim 1.

[0023] The method mentioned at the outset is characterized by the following steps being carried out: - using a learning device (10) which independently carries out test runs of the motor (21) during operation of the machine and thereby records measurement data which are used to reduce a speed fluctuation, and - Setting a control unit (18) on the basis of the measured data in such a way that it compensates for the wow and flutter.

[0024] In connection with the invention described here, the term ripple is used synonymously below instead of the term wow and flutter.

[0025] The method according to the invention enables all relevant disturbances to be compensated for with a solution approach that targets the respective cause. The learning device therefore first learns the individual influences of the disturbances before they are individually compensated for by the control unit.

[0026] Furthermore, the method according to the invention allows for use in series production. Such a method can also be integrated into standard testing steps during production and requires minimal time.

[0027] The method requires only a few test runs of short duration to determine the parameters required to compensate for ripple.

[0028] The test runs can be carried out both in load-free operation after the production of an individual component, for example after the assembly of a single joint, and on the completely assembled machine, e.g. a robot, in an area with little load change.

[0029] The process extends the availability and service life of all machines powered by PM motors. According to the invention, this is achieved by better preventing damage caused, in particular, by ripple torque in the PM motor, such as bearing damage. At the same time, unnecessary downtime and repair costs can be reduced.

[0030] It is advantageous to conduct the engine test runs in two directions of rotation. This makes it possible to identify and eliminate any time lag between a fault and its correction.

[0031] Another advantage is that several measurement signals are acquired and recorded as measurement data during at least one engine test run. This allows for a time-saving simultaneous recording of a large number of physical variables that may overlap and / or influence each other.

[0032] It is advantageous to determine at least one frequency of a ripple from the measurement data. This allows a periodically recurring disturbance to be detected or identified.

[0033] Another advantage is that the measurement data can be used to determine which component of the system causes which ripple at which frequency. This allows individual sources of interference to be identified and compensated for individually.

[0034] With this in mind, measurement signals and / or data from a sensor are first compensated before measurement signals and / or data from the motor are used to determine and / or compensate for motor-related ripple. This allows a disturbance originating from a sensor to be masked out when compensating for motor disturbances. Sensor-related disturbances can be decoupled from motor-related disturbances.

[0035] It is advantageous to use a control unit that executes algorithms to compensate for ripples. This makes it possible to conduct iterative preliminary procedures during preliminary investigations and to adapt the procedure to different disturbances. The algorithms are learned by the learning device.

[0036] It is also advantageous to conduct test runs during a production process or at a production facility. This makes it possible to eliminate defects in an assembly or machine, especially a robot, or individual components during production.

[0037] A robot is advantageously used as the assembly. The process described here is applicable to the series production of robot joints. Test runs can be performed both in load-free operation during a standard production test after assembly of a single joint. Test runs can also be performed on a fully assembled robot.

[0038] The method described here applies to permanent magnet (PM)-excited electric motors and concerns measures for compensating their ripples. These measures reduce torque ripples and periodically recurring wow and flutter.

[0039] The method can also be used for electric vehicles if suitable test conditions can be achieved. Such test conditions could be created either by jacking up the electric vehicle or in a safe, uneven road area.

[0040] In the drawing show Fig. 1 a circuit diagram known from the prior art relating to a compensation of a ripple of a motor, Fig. 2 is a circuit diagram relating to a first, simplified embodiment of the invention, Fig. 3 a circuit diagram relating to a second embodiment of the invention, in which a motor control block performs a vector control of the switching block according to Fig. 2 takes over, Fig. 4 a circuit diagram relating to a third embodiment of the invention, in which the essential components of a learning device according to Fig. 3 are shown, Fig. 5 a flow chart for a preliminary investigation for performing ripple compensation, this method being used to optimally adapt a parameterizable control unit and a learning device to tasks, and Fig. 6 a robot with a motor-driven joint in which the method described here is used.

[0041] Fig. 1 shows a circuit diagram relating to a ripple compensation according to the prior art.

[0042] Fig. Figure 2 shows a circuit diagram relating to a method for compensating at least one torque ripple or wow and flutter occurring in an arrangement comprising a motor 21 having at least one permanent magnet.

[0043] This procedure involves the following steps: Using a learning device 10 which independently carries out test runs of the motor 21 and thereby records measurement data which are used to compensate for at least one ripple to be compensated.

[0044] Setting a control unit 18 based on the measured data such that it compensates for the at least one ripple.

[0045] Fig. 2 shows that in the method described here, a parameterizable control unit 18 is used instead of a memory module 1 according to Fig. 1 and a learning device 10 are used, with which the parameters required to compensate for the ripple can be learned.

[0046] The learning device 10 is capable of independently performing test runs during normal test steps of a production and learning the parameters required for successful compensation of ripple.

[0047] For this purpose, the switching unit S1 is in the position shown in Fig. 2, so that all specifications for vector control of motor 21 originate from learning device 10. In vector control, the measured motor currents are mapped to the rotor coordinates q,d using Clarke transformation 6 and Park transformation 7, which then become DC quantities and can thus be controlled without periodic AC components. The torque is controlled via PI controller 2, and any required field weakening is controlled via PI controller 3. The controller outputs are applied to the motor coils of motor 21 via an inverse Park transformation 4 and the pulse width modulation unit 5.

[0048] The parameterizable control unit 18 determines a compensation current based on a few parameters to suppress the ripple in the speed or in the torque or in the useful torque.

[0049] In addition, the parameterizable control unit 18 has various algorithms for suppressing disturbances originating from the angle of rotation or current sensors, the gearbox, the bearing of the rotor or the drive shaft of the motor 21.

[0050] In this respect, a control unit 18 is used which executes algorithms to compensate for ripples.

[0051] The learning device 10 finds the parameters with which the parameterizable control unit 18 can sufficiently suppress the ripple in the useful torque and synchronism disturbances.

[0052] Once the control unit 18 has been successfully trained and parameterized accordingly, the switching unit S1 is moved to a lower position so that any desired setpoint 9 can act on the motor 21 to control the motor 21. The setpoint 9 is a reference current i q,ref which generates a torque.

[0053] In Fig. 3 is the vector control of the switching block 15 from Fig. 2 are combined into a motor control block 20 and concentrated on the essential parts.

[0054] Ripple can occur in motor 21 due to asymmetries in the coils or permanent magnets of motor 21. Ripple can also result from systematic errors in a sensor, such as an angle encoder. Ripple can also be caused by signal conditioning. To successfully compensate for ripple, a suitable solution must be found and applied for the respective cause or fault.

[0055] Fig. Figure 4 shows the main blocks of the learning facility 10.

[0056] The learning device 10 comprises or contains a measuring device 11, a measurement data memory 12, an evaluation device 13, a parameter optimization module 14 and a well-adjusted speed controller 16.

[0057] The speed controller 16 compensates for speed changes caused by torque ripple by compensating torque changes at a constant speed setting.

[0058] However, since the speed controller 16 can only regulate once a disturbance has already occurred, the compensation current generated by the speed controller 16 has a time offset to the torque ripple or to the disturbance that triggers the ripple.

[0059] Therefore, test runs of the motor 21 are carried out at a low constant speed in two directions of rotation in order to keep the time offset small and to be able to eliminate the time offset by comparing the two directions of rotation.

[0060] Excessively high test speeds can lead to the speed controller 16 no longer being able to follow or even showing resonance effects that falsify the measurement data.

[0061] Therefore, the transfer function of the speed controller 16 should be known in order to be able to adjust a suitable test speed to the frequency of a ripple to be measured and to keep the time required for the test run short.

[0062] During at least one test run of the engine 21, several measurement signals are acquired and recorded as measurement data.

[0063] The effects of disturbances are visible in several measurement signals with varying degrees of intensity. Therefore, in the method described here, several measurement signals are recorded during an initial test run for a few seconds or engine revolutions, once the speed controller 16 has set the constant specified speed.

[0064] When enough measurement data or data points have been recorded for subsequent evaluation, the recording of the first test run is stopped and the second test run is prepared by inverting the direction of rotation.

[0065] As soon as a specified speed or target speed in the reverse direction is reached, a second data recording is started. After the same measurement time, the second data recording is terminated and the motor 21 is stopped again.

[0066] The measurement data or data points recorded during both test runs are temporarily stored in the measurement data memory 12 separately according to test run and then evaluated in the evaluation device 13.

[0067] The recorded measurement data is evaluated in the form of a multi-channel Fourier transform analysis.

[0068] The evaluation procedure becomes more robust if the Fourier transformation is performed over a cumulative motor angle θ for a fixed number of revolutions and not over time.

[0069] Especially with larger speed fluctuations, a time-based Fourier transformation would produce blurred and inaccurate analysis results. Fourier analysis is designed to provide an amplitude and phase angle for each relevant disturbance.

[0070] If one analyses in this way the reference current i specified by the speed controller 16 q,ref , the delay caused by the speed controller 16 leads to a shift of the phase angle in the direction of rotation.

[0071] By averaging the phase angles from both test runs in different directions of rotation at the same speed, the phase shift caused by the speed controller 16 can be compensated.

[0072] In this way, the following measurement signals can be analyzed: - The reference current i specified by the speed controller 16 q,ref - The measured current i q - The correction current i q,corr - The speed ω - sin θ s and cos θ s from the angle sensor 17 - The sum of squares (sin 2 θ s + cos 2 θ s ) or the root of it - The total current i a + i b + i c

[0073] The following relationships apply:

[0074] The Fourier coefficients for a frequency n are determined according to the following formulas (1) and (2): αn=1k∑θi=0m⋅2πƒ(θi)⋅cos(n⋅θi) bn=1k∑θi=0m⋅2πƒ(θi)⋅cos(n⋅θi)

[0075] The following applies: n=Frequency=Periods / Revolution θi=motor angle at i-th measurement m=number of engine revolutions f(θi)=function value of the analyzed quantity at the i-th measurement

[0076] For an amplitude A at a frequency n, the following applies: An=an2+bn2

[0077] The phase angle φ at a frequency n is: tan(ϕn)=αnbn

[0078] For the mean value of the phase angle ϕ¯n The following applies to the two directions of rotation forward (“forward”) and backward (“back”). ϕ¯n=ϕn,forward+ϕn,back2

[0079] Formulas (1) to (5) explain the Fourier analysis.

[0080] For each frequency, namely each disturbance frequency, the amplitude values ​​according to formula (3) as well as the average of the phase angles across both directions of rotation from formula (5) are relevant. Which function values ​​f(θi) are used depends on the source of the disturbance. For disturbances in motor 21, the measured current f(θi)=i q and the correction current f(θi)=i q,corr . For the angle sensor 17, the sine or cosine function as well as the sum of squares are analyzed in this way.

[0081] The choice of the appropriate analysis channel depends on which influencing factor is to be compensated.

[0082] While all disturbances are mixed in the speed measurement signal and consequently also in the currents, the trigonometric signals and their square sum or the root thereof essentially show the errors of the angle sensor 17.

[0083] For correctly functioning current sensors, the total current is constantly zero according to Kirchhoff's law. However, if the current gain of the current sensors differs, the total current exhibits a fault-related ripple with a pole reversal frequency.

[0084] In order to optimally adapt the learning device 10 to the type of motor 21, the type of angle sensor 17 used and any gearbox used, a preliminary investigation with suitable test runs is carried out in accordance with Fig. 5 carried out.

[0085] The preliminary investigation can determine which frequencies the observed speed fluctuations show and whether these originate from the angle sensor 17, the motor 21 or the gearbox 24.

[0086] At least one frequency of a ripple is determined from the measurement data. In particular, the measurement data determines which component of the arrangement causes which ripple and at which frequency.

[0087] There is a fundamental difference between the compensation of ripples due to disturbances in the motor 21 and those due to those in the angle sensor 17.

[0088] While the speed fluctuations resulting from the motor 21 generally depend on the motor current or on a load torque, this dependence does not exist for disturbances resulting from the angle of rotation sensor 17.

[0089] The distinction between the origin of the speed fluctuations from the motor 21 or from the angle sensor 17 is particularly crucial when the ripples due to these disturbances show the same frequency.

[0090] If this is the case, it is essential to first eliminate the fault in the angle sensor 17 before performing the compensation of the ripple generated by a fault in the motor 21.

[0091] In this respect, measurement signals and / or measurement data coming from a sensor are first compensated before measurement signals and / or measurement data coming from the motor 21 are used to determine and / or compensate for motor-related ripple.

[0092] The preliminary investigation according to Fig. 5 is an iterative process. First, a test run is performed with a constant speed in one direction, and measurement data is recorded for later evaluation.

[0093] A second test run is then carried out in the opposite direction of rotation at the same speed and a second set of measurement data is recorded.

[0094] The two test runs are carried out in Fig. 5 is identified as a test run pair TP by step 31, which occurs after a start step 30.

[0095] The respective recording of the measurement data is marked as step 32.

[0096] Once the two test runs and the recording of the measurement data are completed, a multi-channel Fourier transform analysis is performed, which is marked as step 33.

[0097] This is followed by step 34 to determine the frequencies of the ripples and their origin.

[0098] Step 34 is explained in more detail below.

[0099] In order to distinguish whether the observed disturbances originate from the motor 21 or from the angle sensor 17, the selected measurement signal and the detected frequency of a ripple due to a disturbance are taken into account.

[0100] For each type of disturbance there is a characteristic fundamental frequency and associated harmonics.

[0101] Asymmetries in rotor magnets show the pole change frequency as the fundamental frequency.

[0102] Asymmetries in the magnetic fields of a stator show a fundamental frequency which, according to the formula f s = 1 / number of teeth.

[0103] For the conduct of the preliminary investigation, the Fig. 4 shown measurement data memory 12 and the evaluation device 13 can be used.

[0104] The recording of the measurement data and the subsequent evaluation can also be carried out externally with suitable data transmission.

[0105] The advantages of performing the analysis externally are generally greater flexibility, the larger available memory, and the better graphical representation of the Fourier transform analysis.

[0106] Once the frequency of a ripple and the origin of the underlying disturbance have been identified, the evaluation unit 13 can be specifically aligned to the frequency to be addressed, thus simplifying the process.

[0107] In particular, the two Fourier coefficients then only have to be determined for a few frequencies in the evaluation device 13, for example in order to obtain amplitude and angle.

[0108] In the preliminary investigation, it is first determined which wow and flutters are caused by disturbances in the rotation angle sensor 17.

[0109] An overview of sensor-related disturbances can be obtained after just two test runs, for example in the low-load case, in particular by carrying out steps 31 and 32.

[0110] Since disturbances in the rotation angle sensor 17 also influence the motor angle θ, these disturbances also have an effect in the form of simulated small changes in speed.

[0111] The speed controller 16 also reacts to simulated speed changes by compensating setpoint current specifications.

[0112] However, these overlap with the corrections made by the speed controller 16 as a result of detected torque fluctuations.

[0113] Therefore, specifying a reference current i q,ref primarily for assessing the remaining interference and provides information about which frequencies have actually led to wow and flutter.

[0114] In order to detect the disturbances originating from the angle sensor 17, the Fourier transformation of the measurement signals sin θ is used. s and cos θ s from the angle sensor 17 and the Fourier transformation of the sum of squares (sin 2 θ s + cos 2 θ s ) or its square root. This can be done in step 33.

[0115] First, errors in signal processing must be expected. Such errors can be offset errors or scaling errors.

[0116] These errors can easily be determined from the Fourier coefficients of the measured data or measurement signals sin θ s and cos θ s from the rotation angle sensor 17 if the Fourier coefficients are determined exactly over an integer multiple of revolutions.

[0117] A zeroth order, namely a Fourier coefficient a0 for a frequency of 0, provides the offset.

[0118] A first order, namely a Fourier coefficient a p , b p for a pole change frequency p of the rotation angle sensor 17, provides the amplitude or any crosstalk from the other signal.

[0119] Offset errors lead to a fluctuation of the sum of squares with the fundamental frequency, namely the pole change frequency, of the angle sensor 17.

[0120] Errors in the amplitude ratio or crosstalk errors lead to fluctuations in the sum of squares at twice the fundamental frequency, namely the pole change frequency, of the angle of rotation sensor 17.

[0121] The sum of squares is used more for control than for extracting errors because it represents a combined signal from two sources.

[0122] In a special embodiment in which a 10-pole resolver is used as the rotation angle sensor 17, a further error can be detected in this way, which is reflected in the sum of squares with four times the pole reversal frequency and in the individual signals with three times the pole reversal frequency, namely the third harmonic.

[0123] Therefore, a compensation algorithm is installed in the parameterizable control unit 18, which uses two parameters to calculate the measurement signals sin θ s and cos θ s from the angle sensor 17 so that the square sum of the corrected measuring signals (sin 2 θ* s + cos 2 θ* s ) no longer show the fluctuation in the fourfold pole change frequency.

[0124] The parameters for correcting an offset, a scaling, a crosstalk and, if necessary, a third harmonic are in Fig. 5 is summarized in step 35. Step 35 focuses on the parameters that affect sensor-related ripples.

[0125] The preliminary investigation is carried out with the corrected measurement signals sin θ* s and cos θ* s from the angle sensor 17 as soon as the sum of squares (sin 2 θ* s + cos 2 θ* s ) shows a smooth course without significant interference frequencies.

[0126] In Fig. 5 shows the decision on the success of the sensor signal correction by a decision node in step 38.

[0127] As soon as the sensor-related disturbances or errors are sufficiently well compensated so as not to interfere with further investigations, the measurement data obtained from the test run pairs are processed according to step 31.

[0128] After performing the multi-channel Fourier transform analysis according to step 33, these measurement data are used to determine the parameters relating to motor-related ripple in step 36.

[0129] This is done using the measured current i q or based on the specified reference current i q,ref of the speed controller 16 added correction current i q,corr the parameterizable control unit 18.

[0130] An iterative procedure is applied for each frequency detected in step 34. Step 34 aims to determine the frequencies of the ripple and their origin.

[0131] At each iteration step, for each frequency or interference frequency to be tested, the amplitudes and phase angles are determined from the Fourier transformation of the two measurement signals measured current i q and correction current i q,corrfor each direction of rotation and their mean values ​​were calculated based on both directions of rotation and compared with each other.

[0132] As far as possible, a match in amplitude and phase angle should be achieved because this means that the speed controller 16 cannot detect any deviation at this frequency and therefore does not have to supply a compensating reference current iq,ref for this interference frequency.

[0133] It is advisable to start the first iteration step with a very small amplitude so that the phase shift and amplitude increase caused by the true disturbance become visible.

[0134] Once the phase angle has been found to be stable by means of a difference calculation, the amplitude for this interference frequency can also be adjusted by multiplying it by the reciprocal amplitude ratio.

[0135] This functionality is realized via the weight factor X introduced in the following formula (6).

[0136] It is kept small at least for the first iteration step and only increased to values ​​of 0.9 to 0.95 when the phase angle has been found and it is ensured that the detected amplitude A n,l-1 (i q ) in the measuring signal of the measured current i q is also greater than the amplitude A n,l-1 (i q,corr ) in the impressed compensation signal i q,corr .

[0137] For the amplitude parameter pA n‚l , in the I-th iteration step: pAn,ι=pAn,ι−1⋅pAn,ι−1(iq)pAn,ι−1(iq,corr)⋅X with: X = weight factor < 1

[0138] For the phase parameter pϕ n,l in the I-th iteration step: pϕn,l=pϕn,l−1+ϕn¯(iq)−ϕn¯(iq,corr)

[0139] Formulas (6) and (7) apply to the optimization of the parameters concerning the motor in iteration step I.

[0140] The weighting factor X is initially small until the phase angle is close to the target value and it is certain that the measured amplitude of the measured current i q is greater than that of i q,corr .

[0141] This is the only way to ensure that a fault actually exists and is not just simulated by the previous correction value.

[0142] In principle, the iterative procedure used later in the self-learning process is a simplification of the procedure used in the preliminary investigation according to Fig. 5.

[0143] Since the frequencies of the ripples and their origin have already been determined in step 34 during the preliminary investigation, this step is omitted in the self-learning process.

[0144] The evaluation device 13 also only has to calculate the Fourier coefficients amplitude and phase for each known frequency or interference frequency according to the formulas (1) to (5) for the measurement signals suitable for detecting the interference.

[0145] There is then also a suitable algorithm for compensation in the parameterizable control unit 18 in order to be able to efficiently suppress the detected disturbance with the parameters determined from the measurement.

[0146] The method just described for optimizing the parameters relating to the motor 21 is integrated into the learning device 10 as an algorithm of the parameter optimization module 14.

[0147] To compensate for the motor-related torque ripple with frequency n, the parameterizable control unit 18 generates the correction current i q,corr = pA n * cos (n*θ) - pΦ n ) and the reference current i q,ref added.

[0148] The algorithm of the parameter optimization module 14 is also supplemented by the previously described methods for optimizing the sensor-related parameters (cf. Fig. 4).

[0149] Therefore, it is possible to test whether the learning device 10 and the parameterizable control unit 18 are functioning correctly during the preliminary test. Preferably, the learning device 10 is installed in a microcontroller or implemented as such.

[0150] It can also be checked whether after a few iteration steps through the learning facility 10 necessary parameters for correction can be found.

[0151] Another aim of the preliminary investigation is to find a simple, parameterizable formula for the current-dependent speed fluctuations in order to be able to draw conclusions about the behavior at higher loads from the few test runs in the low-load case during later tests during series production.

[0152] Basically, one looks for a suitable simple amplitude function for all current-dependent frequencies or interference frequencies.

[0153] In most cases, however, it is sufficient to install the amplitude functions only for the fundamental frequencies of the rotor and stator of motor 21, because the harmonics are significantly weaker and do not cause as much interference.

[0154] However, it is also necessary to determine a phase function because the correction currents i q,corr are only set with a delay by the motor control block 20 and therefore lead to a speed-dependent shift.

[0155] Therefore, this delay must be compensated for by an adequate lead time. However, the phase function has only a slight current dependence, which can usually be neglected.

[0156] The test runs described here are carried out during a production process or at a production site.

[0157] According to one embodiment, a robot is used as the assembly. According to this embodiment, the preliminary examination is carried out on an assembled robot 22. This is shown in Fig. 6 is shown schematically.

[0158] The robot 22 comprises a base 28 to which an arm 25 is rotatably connected via a joint 23. The motor 21 and a gear 24 are installed in the joint 23 to drive the rotational movement of the arm 25.

[0159] The angle of rotation θ can be detected by means of the angle of rotation sensor 17. At the upper end of the arm 25, a further joint 26 is arranged, which can be constructed analogously to the joint 23 in order to move an end effector 27 mounted on the further joint 26. The end effector 27 can be any tool and, in the present embodiment, is designed as a gripper.

[0160] To shorten the measurement time during the preliminary examination, a weight is attached to the arm 25 or the end effector 27. The weight is moved at a slow speed through a semicircle from a horizontal position to a vertical position to an opposite horizontal position and back again.

[0161] Multiple measurement data sets are recorded externally. For example, twelve measurement data sets with different loads are collected in each direction of movement, depending on the angular position of the load arm, namely arm 25.

[0162] The respective fundamental frequencies are determined by the number of teeth and permanent magnets of the motor. Since in this example the stator of motor 21 is constructed with 18 teeth and the rotor consists of 20 permanent magnets, the Fourier coefficients for the fundamental frequencies, namely frequency = 18 / revolution and frequency = 20 / revolution, as well as the frequencies of the harmonics = 36 / revolution, = 54 / revolution, and = 40 / revolution, are calculated from the measured current i. q determined and plotted against the average load current during the respective measurement data recording.

[0163] It turns out that only for the fundamental waves an amplitude function with a slight shift at the zero crossing of the frequency = 20 / revolution or at the frequency = 18 / revolution has to be installed.

[0164] The current dependence is significantly lower for harmonics and overall they cause less interference.

[0165] The amplitude function found for a 21-tooth motor can also be used for other motor sizes with otherwise the same design, namely a stator with 18 teeth and a rotor with 20 permanent magnets.

[0166] The phases change barely with the current, but significantly when the direction is reversed. The phase functions are therefore determined for a freely rotating single joint under low-load conditions.

[0167] The amplitude functions and the phase functions found in the preliminary investigation are integrated into the parameterizable control unit 18 in addition to the algorithms required for correcting the sensor errors.

[0168] The evaluation unit 13 is designed to detect the frequencies = 18 / revolution, = 20 / revolution, = 36 / revolution, = 54 / revolution and = 40 / revolution, which are due to a fault in the motor 21, as well as the frequencies expected from the rotation angle sensor 17.

[0169] The algorithm of the parameter optimization module 14 is tuned to first compensate for the sensor-related disturbances or errors and then, in further iteration steps, the motor-related disturbances.

[0170] Thus, in a learning process, for example during the first commissioning of an assembled motor 21, the necessary parameters for suppressing speed fluctuations due to sensor-related disturbances and motor-related torque fluctuations can be determined from measurements in the low-load case.

[0171] The determination of the parameters can also be repeated on an assembled robot 22 if necessary by repeating the measurements required for the learning process in the low-load case.

[0172] Suitable parameters for a simple linear amplitude function can also be determined from the amplitudes and average currents measured in the “forward” and “backward” directions of movement.

[0173] If transmission-related disturbances, for example from gearbox 24, or disturbances from the bearings are also to be compensated, a further optimization step for the characteristic frequencies of these disturbances is required.

[0174] Typically, these frequencies are much lower and are in the range of =1 / revolution or =2 / revolution.

[0175] However, the described method can only compensate for periodically recurring disturbances and thus foreseeable disturbances.

[0176] In the case of randomly occurring faults in the bearings or effects due to gear play, predictability is generally not possible and therefore no correction can be planned.

[0177] As an alternative to the described method using the amplitudes according to formula (3) and the phase mean values ​​according to formula (5), an optimization method can also be constructed directly using the Fourier coefficients according to formula (1) and formula (2).

[0178] However, the described method has proven to be effective for the fundamental frequencies of the quantities originating from the rotor or stator, because speed-dependent delays can be more easily mapped to the phase shift.

[0179] It is also conceivable to use a combination of both methods for the harmonics by using the phase shift to be applied for the fundamental wave as the shift angle in the calculation of the Fourier coefficients according to formulas (1) and (2) and thus compensating the speed-dependent component in the same way as for the fundamental wave. List of reference symbols 1 memory module 2 PI controllers 3 PI controllers 4 Inverse Park transformation (from rotating rotor system to stationary stator system) 5 Pulse width modulation unit 6 Clarke transformation (from 3-phase system to Cartesian coordinate system) 7 Park transformation (from stationary stator system to rotating rotor system) 9 Target specification 10 Learning facility 11 Measuring device 12 measurement data memories 13 Evaluation device 14 Parameter optimization module 15 Switch block 16 speed controllers 17 Angle sensor 18 Control unit (parameterizable µC unit) 20 Engine control block 22 robots 21 Engine 23 Joint 24 gearboxes 25 arms 26 additional joint 27 End effector 28 Base S1 switching unit

Claims

[1] Method for reducing wow and flutter of a permanent magnet excited electric motor (21) installed in a machine, characterized by , that the following steps must be carried out: - using a learning device (10) which independently carries out test runs of the motor (21) during operation of the machine and thereby records measurement data which are used to reduce a speed fluctuation, - setting a control unit (18) on the basis of the measured data in such a way that it compensates for the wow and flutter, and - Determining from the measured data which component of the arrangement causes which wow and flutter at which frequency, wherein a wow and flutter caused by the motor (21) is decoupled from a wow and flutter caused by a sensor by determining the motor-induced wow and flutter and determining the sensor-induced wow and flutter, wherein the sensor-induced wow and flutter is first compensated before a wow and flutter caused by the motor (21) is compensated, the measured data comprising a measured current (iq) or a correction current (iq, corr) and a motor angle (θ) of the motor (21); wherein a wow and flutter caused by the motor (21) is determined on the basis of the measured current (iq) or the correction current (iq, corr) by determining its amplitude and phase angle for each frequency of the measured current (iq) or the correction current (iq, corr), and a sensor-induced wow and flutter is determined from the sine and cosine of the motor angle (sin(6), cos(6)) of the motor (21) using a Fourier transformation. [2] Method according to claim 1, characterized by that the test runs of the motor (21) are carried out in two directions of rotation of the motor (21). [3] Method according to claim 1 or 2, characterized by that the test runs are carried out at a constant engine speed of the engine (21). [4] Method according to one of the preceding claims, characterized by that during at least one test run of the engine (21) several measurement signals are detected and recorded as measurement data. [5] Method according to one of the preceding claims, characterized by that at least one frequency of a wow and flutter is determined from the measured data. [6] Method according to one of the preceding claims, characterized bythat the sensor is a rotation angle sensor (17) and determines a motor angle (6) of the motor (21). [7] Method according to one of the preceding claims, characterized by that a control unit (18) is used which executes algorithms to compensate for wow and flutter.

Citation Information

Patent Citations

  • Method for determining the speed of a synchronous machine

    DE102010062273A1

  • Method and arrangement for reducing the torque ripple of a DC motor

    DE102014105730A1

  • Method of compensating for periodic vibration forces in an electrical induction machine

    EP0792420B1

  • Method for ascertaining a correction value for the angle of the rotor of an electrically commuted reversible synchronous motor

    EP2235818B1

  • System and method for error correction in angular position sensors

    US20160193941A1