Sensor device for an electric machine, method for the operation of a sensor device

Abstract

A sensor device for an electric machine includes a rotor shaft mounted rotatably in a housing, with a signal generator that is or can be joined non-rotatably to the rotor shaft and is or can be arranged axially on the end face of the rotor shaft. A signal sensor is fixed to the housing opposite on the end face of the signal generator and at a distance from the signal generator. The signal sensor acquires an axial distance from the signal generator.

Claims

1. A method of operating a sensor device of an electric machine having a rotor shaft mounted rotatably in a housing, the sensor device comprising a signal generator joined non-rotatably to the rotor shaft and arranged axially on an end face of the rotor shaft, and a signal sensor fixed to the housing and arranged opposite to the end face of the signal generator at a distance from the signal generator, the method comprising: capturing a multi-part output signal of the signal sensor that depends on an angle of rotation; determining at least signal parameters, signal amplitude, signal phase and signal offset of a respective partial signal; calculating an intermediate variable; determining the amplitude of a specified harmonic; determining a current axial distance Ax between the signal generator and the signal sensor depending on the determined amplitude; and determining a current torque of the electric machine depending on the determined distance Δx.

2. The method according to claim 1, further comprising, if more than two partial signals are present, carrying out a Clarke transformation before calculating the intermediate variable.

3. The method according to claim 1, wherein the intermediate variable is determined in accordance with the following formula:
r(k)=√{square root over (c.sub.1(k).sup.2+c.sub.2(k).sup.2)}, where r(k) represents the multi-part output signal, c.sub.1(k) and c.sub.2(k) represent sine and cosine components of the input signal (d), and k represents the sampling time-point.

4. The method according to claim 1, wherein the amplitude of the specified harmonic is determined with the aid of the following formula: $a=\frac{1}{2n}{.Math.}_{k=1}^{2n}{\left(-1\right)}^{k}r\left(k\right),$ where a is the amplitude of the harmonic, n is the harmonic to be considered, k is the sampling time-point and r(k) is the input signal (e) at the sampling time-point k.

5. The method according to claim 1, further comprising: comparing the amplitude of the sensor signals determined from the signal parameters with the amplitude of the specified harmonic to verify the distance Δx.

6. The method according to claim 2, wherein an angle is determined as an output signal of the Clarke transformation, and the angle is corrected with an arctangent function.

7. The method according to claim 1, wherein the calculating of the intermediate variable includes calculating the intermediate variable from partial signals of the output signal.

8. The method according to claim 1, wherein the specified harmonic is the third harmonic.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) The disclosure is to be explained in more detail below with reference to the drawings, in which:

(2) FIG. 1 shows a simplified illustration of an electric machine with an advantageous sensor device,

(3) FIG. 2 shows a flow diagram for the explanation of an advantageous method for operating the sensor device,

(4) FIG. 3 shows a simplified illustration of sampling locations for carrying out the method,

(5) FIG. 4 shows a signal excursion of sensor signals,

(6) FIG. 5 shows a relationship between the signal excursion and an axial distance between a signal generator and a signal sensor of the sensor device,

(7) FIG. 6 shows a relationship between an amplitude of a specified harmonic in the sensor signal and the distance between signal generator and signal sensor,

(8) FIG. 7 shows a relationship between the axial distance and a torque of the electric machine, and

(9) FIG. 8 shows a further flow diagram for explanation of an advantageous development of the method.

DETAILED DESCRIPTION

(10) FIG. 1 shows a simplified illustration of an electric machine 1 with an advantageous sensor device 2. The electric machine 1 comprises a rotor 3 that is mounted rotatably on a rotor shaft 4 in a housing, not illustrated here. The rotor shaft 4 is formed in two segments, with a first shaft segment 5 that carries the rotor 3 and with a second shaft segment 6 which will be considered in more detail later. The rotor shaft 4 is formed as one piece according to a further exemplary embodiment. The sensor device 2 comprises a signal sensor 7 that is designed to interact with a signal generator 8. The signal generator 8 is arranged at the free end face of the rotor shaft 4, and the signal sensor 7 is arranged opposite the signal generator 8 at the end face, so that a distance Δx exists between signal generator 8 and signal sensor 7. A toothed wheel 9 is arranged on the shaft segment 6, being engaged with a toothed wheel 10 that is arranged non-rotatably on a support shaft 11 parallel to the rotor shaft 4. The toothed wheels 9, 10 are designed as helical gears, so that, depending on a torque generated by the electric machine 1, a force is output, acting axially on the rotor shaft 4, as a result of the helical gearing. The shaft segment 6 is mounted such that it can be displaced axially within a limited range, and is coupled non-rotatably to the shaft segment 4. As a result, the distance Δx between the signal generator 8 and the signal sensor 7 changes depending on the torque that is acting. The sensor device 2 furthermore comprises a control device 12 that is specially set up to carry out the method described in more detail below. The arrangement of the toothed wheels 10, 9 and the electric machine 1 can also be exchanged.

(11) The sensor device 2 can be designed as a resolver, as a Hall-element-based or magnetoresistive (AMR, GMR or TMR) rotation speed encoder and/or as a phase encoder, as an inductive angle sensor or also as an eddy current effect sensor. All the sensor variants give rise to the disadvantage that measurement errors and inaccuracies can arise, resulting in particular from the inhomogeneity of the magnetic fields generated around them and the resulting magnetic fluxes through the coils. Angle errors result from this which, depending on the angle, repeat cyclically with each rotation and are therefore found in the frequency domain as harmonics, that is as multiples, and in particular integral multiples, of the rotation frequency. The second harmonic in the sensor signal in particular can make up a significant portion of the total, cyclically repeating, angle error.

(12) Because the inhomogeneity of the magnetic flux changes depending on the relative position, in particular the axial distance, between the signal generator 8 and the signal sensor 7, the strength of the harmonic interference is also a function of the relative position, in particular of the axial distance Δx. This distance Δx depends on mechanical tolerances and temperature influences as well as on the torque delivered by the electric machine 1, as explained above. In the case of sensors with three coils, an angle error that corresponds to a third harmonic results from the second harmonic in the sensor signal. This can be shown in the following calculation with three sensor signals s.sub.1, s.sub.2 and s.sub.3:
s.sub.1=cos(φ)+a.Math.cos(n.Math.φ)
s.sub.2=cos(φ−⅔π)+a.Math.cos(n.Math.(φ−⅔π))
s.sub.3=cos(φ−⅔π)+a.Math.cos(n.Math.(φ−⅔π))

(13) It can be seen that this relates to a three-phase system with an electrical phase shift of 120° between each of the signals. By means of the Clarke transformation, a two-phase signal c.sub.1, c.sub.2 is determined from the three-phase signal:

(14) $c_1=\frac{2}{3}{S}_1-\frac{1}{3}{S}_2-\frac{1}{3}{S}_3$$c_1=\cos \left(\phi \right)+\frac{1}{3}a\left(2\cos \left(n\phi \right)-\cos \left(n\left(\frac{2}{3}\pi +\phi \right)\right)-\cos \left(n\left(-\frac{2}{3}\pi +\phi \right)\right)\right)$$c_2=-\frac{1}{\surd 3}S2+\frac{1}{\surd 3}S3c_2=\sin \left(\phi \right)\frac{2a\sin \left(\frac{2\pi n}{3}\right)\sin \left(n\phi \right)}{\surd 3}$

(15) After this, an arctangent function can be used to estimate the angle φ.sub.est to be measured.
φ.sub.est=arctan 2(c.sub.1,c.sub.2)

(16) The error ε(φ) of this estimated angle is then given by

(17) $\mathrm{.Math.}\left(\phi \right)={\phi }_{\mathrm{est}}-\phi =\arctan 2\left(c_1,c_2\right)-\phi =\arctan \left(\frac{c_2}{c_1}\right)-\phi$

(18) By inserting different harmonics n, different harmonics result in the angle error which, to a first approximation, can be assumed to have the values in the following table, which were determined through a first-order Taylor expansion.

(19) TABLE-US-00001 Harmonic in the Approximated error in the signal domain (n) angle domain [degrees] 0 0 1 0 2 $-a.Math.\frac{180}{\pi }.Math.\sin \left(3\phi \right)$ 3 0 4 $a.Math.\frac{180}{\pi }.Math.\sin \left(3\phi \right)$ 5 $-a.Math.\frac{180}{\pi }.Math.\sin \left(6\phi \right)$ 6 0 7 $a.Math.\frac{180}{\pi }.Math.\sin \left(6\phi \right)$ 8 $-a.Math.\frac{180}{\pi }.Math.\sin \left(9\phi \right)$ 9 0 10 0$a.Math.\frac{180}{\pi }.Math.\sin \left(9\phi \right)$ 11 $-a.Math.\frac{180}{\pi }.Math.\sin \left(12\phi \right)$ 12 0 13 $a.Math.\frac{180}{\pi }.Math.\sin \left(12\phi \right)$ 14 $-a.Math.\frac{180}{\pi }.Math.\sin \left(15\phi \right)$ 15 0

(20) It can be seen that, for example, the second harmonic in the sensor signal causes an unwanted third harmonic in the angle signal, whose amplitude is proportional to the amplitude of the second harmonics in the sensor signal. This cyclic error impairs the accuracy of the sensor device and can, for example, lead to a cyclically changing torque in the application with the electric machine 1.

(21) By means of the present sensor device 2 and the method described below for operation of same, an improved angle estimate with smaller amplitude angle errors is made available. In addition, the relative position, or its change, in particular of the axial distance Δx is estimated, so that with a known relationship between the axial distance and a further measured value, a conclusion can be drawn as to the further measured value, in particular of the torque of the electric machine. The advantage of a higher angular precision and a more robust angle signal of the sensor device 2 emerges from this. The method described below is simple to implement, since it can be implemented easily on the control device 12. The torque delivered is easy to determine through the method, and can therefore be captured without a separate sensor. Hardly any additional costs, if any, arise.

(22) The present method is based on the one hand on the observation of the amplitude of the harmonics, in particular of the second harmonics in the sensor signal during operation, and on the other hand on the observation of the signal amplitude/of the signal excursion when operating. Correction parameters for the angle error can be derived from these two values, and the distance between the signal generator 8 and the signal sensor 7 can be estimated. The estimation of the torque is then carried out depending on the estimated distance Δx. The process for the estimation of the second harmonics in the sensor signal when operating, the estimation of the distance between the signal generator 8 and signal sensor 7, as well as the estimation of the torque, which is carried out by the control device 12, is presented below.

(23) FIG. 2 shows the advantageous method with reference to a flow diagram. In a first step S1, an output signal a, which depends on the instantaneous angle of rotation y, is provided by means of the sensor device 2. The sensor device 2 is here in particular designed as an inductive sensor device. The output signal a can in particular be derived from the amplitude or the frequency of the coil signals of the sensor device 2. In the case of a three-phase design of the sensor device 2, the signal consists of three partial signals, and of two partial signals in the case of a two-phase design.

(24) The parameters of the partial signals are estimated from the output signal in a step S2, wherein the parameters are in particular respectively the signal amplitude (which corresponds to half of the signal excursion), the signal phase relative to the instantaneous angle of rotation φ and the signal offset (displacement from the zero position). These parameters represent the output signal B. For the estimation of the parameters it is necessary that the output signal a is processed at different angles of rotation φ.

(25) In an optional step S3, the partial signals of the output signal a are normalized with the aid of the previously determined parameter b, so that, for example, they have a low offset, have a predetermined amplitude and each have defined phases with respect to the instantaneous angle of rotation φ. A normalized output signal c results from this.

(26) In the case of a three-phase design of the sensor device 2, the three-phase signal is subsequently converted in a step S4 into a two-phase signal d. The Clarke transformation is used for this purpose. This step is omitted in the case of two-phase sensors.

(27) In a following step S5, a single intermediate variable e, which can be referred to as a vector length, is calculated from the two-phase output signal d. The following formula can be used as a basis for this:
r(k)=√{square root over (c.sub.1(k).sup.2+c.sub.2(k).sup.2)}
Here r(k) is the output signal e, while c.sub.1(k) and c.sub.2(k) are the two components (sine/cosine) of the input signal d. The variable k refers to the index of the sampling time-point.

(28) The determination of the amplitude of the harmonics to be examined, in particular of the third harmonics in the angle domain corresponds to the second harmonics in the signal domain cas is shown in the table for the case of n=2), takes place in processing step S6 using a plurality of values e. The following formula can be used for the third harmonic in the angle domain:

(29) $a=\frac{1}{6}\underset{k=1}{\overset{6}{.Math.}}{\left(-1\right)}^{k}r\left(k\right)$

(30) Here, a is the amplitude of the harmonics corresponding to the output signal f, k is the sampling time-point, and r(k) is the input signal e. It is also possible to take into account more than six sampling time-points in each rotation, wherein six sampling time-points at the respective minima and maxima are particularly expedient. These six sampling time-points are distributed evenly over a rotation, with a spacing of 60°, wherein the absolute position of the first sampling time-point r.sub.1 does not have to be at 0, but depends on the phase position of the harmonics of the error.

(31) FIG. 3 shows in this connection a simplified illustration of the sampling locations or sampling time-points for the determinations of the amplitude of the third harmonics of the angle error.

(32) The amplitudes of the sensor signals, which can thus be extracted from the signal parameters b, are determined in step S7. These amplitudes represent the output signal g. FIG. 4 shows in this connection in a diagram of a demodulated output signal over the angle of rotation φ with a signal excursion g.

(33) In processing step S8, relationships known from reference measurements between the signal excursion g and the distance Δx between the amplitudes of harmonics in the angle error, in particular of the third harmonics in the angle domain, are used to deduce the distance Δx from the known amplitudes of the harmonics of the error f and the signal excursion g. This distance represents the output signal h. The relationship between the signal excursion g and the axial distance Δx can be given, within a defined range of, for example, 0.2 mm to 4 mm, in the form of a monotonic function that results from the weaker electromagnetic field at greater distances, as is shown by way of example in FIG. 5.

(34) In this connection, FIG. 5 shows in a diagram for example the signal excursion g against the distance Δx.

(35) FIG. 6 shows in a simplified form the relationship of the amplitude of the harmonics A.sub.H to the axial distance Δx. If this amplitude is understood to be related to a fixed phase position, it can also be negative in certain distance ranges. Typical values for this amplitude lie in the range between 0% up to 20% in relation to the mean value of the vector length e.

(36) In step S9 the torque is estimated from the previously estimated distance Δx with the aid of the relationship, determined previously in reference measurements, between the distance Δx and the torque of the electric machine. This relationship between the distance Δx and the torque M.sub.d can typically be represented as a function with three distances, as shown in FIG. 7.

(37) FIG. 7 shows the distance Δx between the signal generator 8 and the signal sensor 7 plotted against the torque M.sub.d. The relationship is divided into three sections (1), (2) and (3). Section (1) results with a torque MD in the one direction, and section (3) with a torque in the other direction, wherein these two sections can typically be described as linear functions that depend on the mechanical construction in the electric machine 1 and the sensor device 2 in particular. Section (2) lies in the transition region between sections (1) and (2), and results from the bearing play of the mechanical construction which typically leads to a higher gradient (distance divided by torque) in this region.

(38) The process for an improved correction, which is carried out on the control device 12, is presented below with reference to FIG. 8. The steps S1, S2, S3 and S4 are already known from the exemplary embodiment of FIG. 2. In a step S10 an angle is calculated as the output signal (i) from the input signals (f), which represent sine/cosine signals over one rotation of the signal generator 8 with respect to the signal sensor 7, making use of the arctangent function. A correction of the angle error that arises a result of harmonics in the signal domain, in particular as a result of the second harmonics in the signal domain, is advantageously integrated into the method on at least one of three points A, B and C as a method step. Point A occurs here as step S11, which follows step S3 and thus takes place before the Clarke transformation in step S4, point B occurs as step S12 following the Clarke transformation in step S4, and point C occurs as step S13 after the arctangent function. Any desired combinations of multiple correction blocks A, B, C are possible here in order to obtain a corrected angle φ.sub.korr in step S14.

(39) In the case of the correction A, the output signal d is calculated for the three-phase signal (C) with the help of a previously determined, uncorrected angle φ.sub.unkorrigient and the previous determined amplitude a and phase Δ of the harmonics n that is to be corrected, in that for each signal phase, and for every harmonic to be corrected, the value of the respective cosine function is determined and subtracted from the signal (c):
a.Math.cos(n.Math.φ.sub.unkorr+θ)

(40) In the case of correction B, the same procedure is applied to the two-phase signal (e) in order to generate the output signal (f).

(41) In the case of correction C, the correction is applied in the angle domain to the previously determined angle (g), whereby the conversion of the amplitude and the harmonic number n in the signal domain into the error amplitude and the harmonic number in the angle domain takes place in accordance with the table discussed previously.

(42) It is particularly advantageous with correction C that it is only to be applied to a single signal, and therefore requires the least computing effort. It is advantageous for correction A that the achievable correction precision is particularly high, since the correction can be applied to all three sensor signals s.sub.1, s.sub.2, s.sub.3 separately.