DE10247321B3 - Sensing rotary motion and torque from phase measurements and computerized linear transformation, adopts iterative approximation technique - Google Patents

Abstract

Phase offset is determined using iterative approximation, for at least one phase indicator trace, as follows. Commencing with a given offset (ao), this is added to the measured phase value (ak), followed by transformation in the time domain (Tk). The discontinuous portion of the transformed values (tk) is determined. A mean (tmean) is formed from the prior discontinuous portions (tk). Back calculation produces the offset value (DELTAa0). The offset value (ao,neu) is matched. Approximation is continued, adding the offset value (ao), until a given number of iterations is reached or a criterion for stopping is met.

Description

The invention relates to a method for detecting a movement or an angle of rotation with a sensor arrangement, in particular for detecting a movement or an angle of rotation or a torque on axes or shafts, according to the generic term of the main claim.

For example, to record the a steering wheel axis of a motor vehicle acting torque during the Rotation of the steering wheel very small changes in angle in both directions of rotation Steering wheel can be measured. Incremental angle encoders can be used here, an angular position based on the evaluation of optical, magnetic or otherwise as generated by the rotation and by suitable means evaluate detected impulses. To increase the measuring accuracy and especially for measuring the torque due to torsion on the rotating shaft a plurality of such incremental len measured values that occur periodically are evaluated so that Here several phase measured values occur, from which the one to be measured Size how e.g. the angle of rotation, an angle difference or the distance to one The goal is to determine.

The method is used, for example, in an optical angle sensor in which N parallel tracks are applied to a cylinder. Each of the N tracks (i = 1 ... N) contains n _i periods of phase information, which is represented, for example, in the optical case by n _i periods of light-dark transitions. Other sensor principles, for example magnetic or capacitive, are also possible here. The traces of the sensor can also be applied on one level instead of on a cylinder, for example in the case of a displacement sensor.

In the case of more than two phase signals, for example, one is used to evaluate such phase measurement values DE 101 42 449 A1 described method proposed. Here the measured phase values are mathematically transformed using a linear transformation method and evaluated with a predetermined weighting. It is also known from the DE 195 06 938 A1 that the phase signals can be evaluated by the simple or multiple use of a classic or modified vernier method.

Furthermore, to determine an angle difference from the DE 101 42 448 A1 It is also known that the phase measurement values are weighted and the integer and non-integer part is then determined. The non-integer part is proportional to the angular difference between two track groups of an incremental value sensor on a shaft, so that the torque on the shaft can be determined by multiplication with the spring rate of a torsion bar connected between the track groups.

From the DE 100 34 733 A1 it is also known per se that in an initialization phase a predetermined offset value is added to the phase measurement value and this in turn results in an adaptation of the offset value. An iterative approximation method for the offset adjustment of two orthogonal sensor signals is seen in itself from the DE 199 15 968 A1 known.

The generic method can, for example, with a corresponding sensor arrangement, as in the aforementioned DE 101 42 448 A1 described, are used on the steering shaft of a vehicle as a so-called torque angle sensor (TAS), in which the steering angle and the steering torque are to be output simultaneously. For this purpose, several optical or magnetic tracks with different phase information are attached on both sides of a torsion element. This phase information is recorded optically or magnetically and converted into corresponding, here electrical phase signals. If there are several tracks on each side of the torsion element, the phase signals are combined on one side, for example with the aid of a previously mentioned multidimensional vernier evaluation to form an ambiguous phase signal. The steering angle and the differential angle, which is proportional to the torque, are determined directly from two phase signals thus generated with the aid of the method known from the prior art.

In practice, the phase signals point however phase offsets due to, for example, mechanical installation tolerances, displacements of the individual phase tracks against each other or through a non-ideal Torsion element are caused. These phase offsets can be very grow up and thus reduce the robustness and accuracy of the evaluation or a reasonable one Evaluation even impossible do.

With the generic method mentioned at the beginning for detecting, for example, the angle of rotation and / or the torque on rotating mechanical components, phase measurements can be made by scanning by means of at least one phase generator on the rotating component of an assigned sensor can be evaluated. According to the invention in at least advantageously determining the phase offset of a phase generator by an iterative approximation method.

Here are called calculation steps an initialization with a predetermined offset value, an addition an offset value to the previous offset value, a transformation into a T range, a determination of the fractional part of the transformed values, averaging from the previous ones determined broken shares, a back calculation of the offset values, an adjustment of the offset values and a continuation in the addition of the Offset values up to a specified number of iterations or one predefined termination criterion has been reached.

Advantageously, the iteration performed for so long be until there is a distribution of the fractional part of the transformation around the zero point of the transformation range. In the case, that the average determined in the first iteration is less than is a predetermined value (e.g. 1/16), can also be a fixed value (e.g. 0.25) are added to the mean. To determine a absolute offset values then only need location information of the mechanical component for calculating a known angle or phase value can be used.

With the method according to the invention is thus advantageously achieved that one based on the next embodiment described below explained Determination of the first unknown phase offset is carried out with the aim that a sensor arrangement, for example for measuring the angle of rotation or torque on a shaft, the correct ones after adding the offset values to the input signals Values for provides the angle and the angle difference. Thus, the greatest possible robustness opposite the sensor arrangement Phase errors can be achieved. The invention is particularly related to this suitable for the offset values of the phase signals of an optical torque angle sensor (TAS) to determine in an optimal way.

drawing

An embodiment of a sensor arrangement to carry out of the method according to the invention is explained using the drawing. Show it:

1 1 shows a schematic view of a sensor arrangement for detecting the angle of rotation or an angle difference of an axis or shaft,

2 a signal flow graph for an N-dimensional phase evaluation,

3 a representation of the distribution of the results of a transformation of phase values and

4 and 5 one way of taking into account the offset values determined.

description of the embodiment

In 1 is an axis in a schematic view 1 Shown as a rotating component, whose angle of rotation Φ and whose angle difference Φ - δ can be determined with a sensor arrangement, which in principle from the state of the art mentioned in the introduction to the description DE 101 42 448 A1 is known.

The method according to the invention is described here using a four-dimensional phase evaluation (N = 4) of the signals from optical signal transmitters 2 to 5 and 6 to 9 , with those in sensor modules 10 and 11 Here, for example, optical phase detectors (optical sampling and phase determination), the phase signals α ₁ to α ₄ and α ₅ to α ₈ , are determined. From these signals, taking into account a respective phase offset, α ₀ can then be used in computing modules 12 and 13 Here, for example, 4-dimensional 3Noni us computing blocks (4-dimensional vernier processing), a new phase signal β ₁ and β _{2 can be} calculated. This can then be done using a further computing module 14 , for example a modified vernier calculation module (modif. nonius processing), the angular position Φ _m and the torque δ _{m of} the shaft or axis 1 be determined.

However, the method can be transferred in a simple manner to more or to fewer dimensions N, wherein in 2 the signal flow graph is generally shown for an N-dimensional phase evaluation of the phase signal α to determine the value for the new phase signal β. The mathematical operations performed for this are also referred to below 3 to 5 shown.

In practice, normalized angles and phase signals are generally considered, ie the signals have a value range from 0 to 1, corresponding to an angle or phase value from 0 to 2π. By a quantization unit 20 after 2 Using mathematical methods known per se, an integer grid point is assigned to the real-value phase signals after a transformation in T-space, an exemplary grid point using the number 30 in the 3 is shown in principle. The greatest possible robustness of the sensor arrangement is achieved here when the average of the measuring points 31 lies on a grid point in T-space before quantization.

The starting point for a computational consideration here is to be a set of P vectors α _k , each with four phase measurements. These measured values are transformed into a 3 × 4 matrix using the transformation M ₁ previously mentioned T region transformed.

T  = M 1 * α  (1)

Now the fractional part t becomes t = T - round ( T 1 ) (2) considered, from which the average is formed in each dimension:

Now the offset values are determined on the basis of these mean values so that when the fractional fraction t _{mean is recalculated} , taking the offsets into account, the mean value is 0.

berechnet. Die dabei, verwendete Matrix M ~ ₁ besteht aus den ersten drei Spalten der Matrix M ₁. Der vierte Offsetwert wird als 0 angenommen. In gleicher Weise könnten auch drei andere Offsetwerte berechnet werden, die dann zu verwendende Matrix M ~ ₁ enthält dann genau die zu den Offsetwerten gehörenden Spalten.Since the transformation with the matrix M _{1 represents} a projection, when calculating back from the values t _mean to the offset values, not all offset values, but only three values can be determined here. Mathematically, this is due to the property Rank( M 1 ) = 3 (4) expressed. Thus, for example, the offset values α ₀₁ , α ₀₂ , α _{03 are} generated using the relationship

calculated. The matrix M ~ _{1 used here} consists of the first three columns of the matrix M ₁ . The fourth offset value is assumed to be 0. Three other offset values could also be calculated in the same way, the matrix M ~ ₁ then to be used then contains exactly the columns belonging to the offset values.

• 1. Initialisierung der Offsets α ₀ mit 0 α 0 = 0 (6)
• 2. Addition der Offsets α' k = α k + α 0, k = 1, ..., P (7)
• 3. Transformation in den T-Bereich und Bestimmung des gebrochenen Anteils und Mittelwertbildung
  
• 4. Rückrechnung auf die Offsetwerte
  
• 5. Anpassung der Offsetwerte α 0,neu = α 0,alt + Δα 0 (10)
• 6. Fortführung der Iteration durch Sprung auf den Schritt 2. bis die gewünschte Anzahl an Iterationen bzw. ein Abbruchkriterium erreicht ist.

Because of the nonlinearity in the relationship ( 2 ) the offset calculation must be done iteratively. The starting point is the set of P vectors α _k with corresponding measured values. P denotes the number of measurements carried out. The iteration proceeds in the following steps:

• 1. Initialization of the offsets α ₀ with 0 α 0 = 0 (6)
• 2. Addition of the offsets α ' k = α k + α 0 , k = 1, ..., P (7)
• 3. Transformation into the T-range and determination of the fractional part and averaging
  
• 4. Back calculation on the offset values
  
• 5. Adjustment of the offset values α 0, new = α 0, old + Δ α 0 (10)
• 6. Continue the iteration by jumping to step 2. until the desired number of iterations or an abort criterion is reached.

If the offset values determined in this way are used, the t values, as fractional parts of T, are around the zero point after the 3 arranged. The 3 shows an example distribution of the t-values before the offset adjustment. Due to the proposed offset determination and the previously described addition of the offset values, the in the 3 shown cloud 31 in the origin around the grid point 30 , The offset values determined in this way guarantee the greatest possible robustness of the sensor arrangement against phase errors.

To completely determine the offset of the sensor arrangement, position information, for example the installation position of the sensor arrangement, is also required. With the previously determined offset values α ₀₁ , α ₀₂ and α ₀₃ , the angle or phase β _m can be calculated for a known angle or phase value β. The difference β 0 = β - β m (11) is the angular or phase offset, which must be added at the end of the evaluation. In order to increase the accuracy when determining β ₀ , however, it is advantageous if this value is determined over several measurements.

den Offset β_o auf den Eingang der Anordnung zurückzurechnen. Die Werte für n_i sind dabei die bei der mehrdimensionalen Auswertung verwendeten Periodenzahlen, so dass sich eine Anordnung nach der 5 ergibt.With the determined offset values α ₀₁ , α ₀₂ and α ₀₃ as well as βo, the entire arrangement is compared, as shown in the block diagram after 4 is shown. It is possible from here with the help of the relationship

to calculate the offset β _o back to the input of the arrangement. The values for n _i are the period numbers used in the multi-dimensional evaluation, so that an arrangement according to the 5 results.

In the case of a multi-stage arrangement, as in principle in the 1 is shown, the method can also be used several times. Only a single angle information, for example the zero point position, is then required for the adjustment of the absolute angle, although in this application it also makes sense to use all offset values analogously to the relationship ( 12 ) to count back to the entrance.

Stagnation in the iteration following the steps 1 to 6 described in an unstable point to prevent it, you can extend the iteration, for example, that for the case that the determined mean value in the first iteration is less than a specified value (e.g. 1/16), also a fixed one Value (e.g. 0.25) can be added to the mean. alternatives this is about given the determination of the variance.

Claims (4)

Method for detecting the movement, the angle of rotation and / or the torque on moving mechanical components, in which - phase measurement values are evaluated which are generated by scanning at least one phase encoder track on the linearly or rotationally moving component by means of a respectively assigned sensor and in which - the phase measured values are mathematically transformed into a new area by means of a linear transformation, characterized in that - the phase offset of at least one phase encoder track is determined by an iterative approximation method in which the following method steps are carried out: - an initialization with a predetermined offset value ( α ₀ ), - addition of the offset value ( α _o ) to the phase measurement ( α _k ), - transformation into a T range ( T _k ), - determination of the fractional part of the transformed values ( t _k ), - averaging ( t _mean ) from the previously determined broken en shares ( t _k ), - Recalculation to the offset value (Δ α ₀ ), - adjustment of the offset value ( α ₀ , _new ) and - a continuation of the approximation process in the addition of the offset value ( α ₀ ) until a predetermined number of iterations or a predetermined termination criterion has been reached. Procedure according to. Claim 1, characterized in that - the iteration is carried out until there is a distribution of the fractional portions of the transformed values ( t _k ) around the zero point. Method according to claim 1 or 2, characterized in that - in the event that the determined mean ( t _mean ) in the first iteration is less than a predetermined value, a fixed value is added to the mean ( t _mean ). Method according to one of the preceding claims, characterized characterized that - to Determination of an absolute offset value a position information of the mechanical component for calculating a known angle or Phase value is used.