GB2576159A - System for monitoring rotation of a shaft - Google Patents

Abstract

A system suitable for monitoring rotation of a shaft comprises and encoder disc and at least one sensor for detecting features on the encoder disc as the disc rotates with the shaft relative to the sensor. The disc is an assembled disc composed of a plurality of segments, each segment bearing a respective set of features and being separately securable to the shaft. A processor is provided for receiving output signals generated by the sensor in response to passing of the features and applying a correction to data derived from the output signals of the sensor to compensate for relative misalignment of the features of the assembled disc. The sensor may be a magnetic sensor and the disc segments may be made of sheet metal, the features detectable by the sensor being holes in the sheet metal. The disc may include an irregularity to indicate a reference radius. The processor may include a look-up table for storing an error value for each feature, indicative of the deviation of the angular position of the feature from a nominal reference position. The system may be used in a torque transmitting shaft.

Description

SYSTEM FOR MONITORING ROTATION OF A SHAFT

Field of the invention

The present invention relates to a system for monitoring rotation of a shaft and to a system for measuring the run out of a rotating shaft.

Background of the invention

It is often desirable to measure the run out of a rotating spindle. Where the spindle is the main shaft of a machine tool, run out measurement is useful for diagnostic, maintenance, and quality control purposes. When a spindle is running true, its axis of rotation should remain perfectly stationary. However, this condition is not even achieved in precision machines. In practice, even when new, some inaccuracy will exist and will cause its axis to wobble as the spindle rotates about its mean centre of rotation. The run out is a measurement of the maximum radial movement of the spindle axis relative to the mean centre of rotation.

Rotary shaft encoders are used in industry to measure the angular position and speed of rotation of a spindle. A shaft encoder comprises a wheel that is coupled to the spindle and has a number of features around its periphery. A stationary sensor, which is positioned to detect the features, generates a signal pulse as each feature passes it by. The number of pulses indicates the angular displacement of the spindle, while the frequency of the pulses is a measure of the rotational speed. As an example, if the entire disc has one hundred evenly spaced features, then fifty signal pulses would signify that the spindle had performed half a revolution.

There are several types of rotary shaft encoder. The wheel may be a flat disc or a drum and the features may be slots that are sensed optically, or magnetic stripes recorded on the cylindrical surface of a drum. Still other encoders use a toothed gear with optical or magnetic sensors to detect the passage of the individual teeth. There also exist high frequency/mm wave RF short range proximity sensors.

A system for measuring the lateral displacement of rotating shafts (such a system is hereinafter referred to as an Incremental Motion Encoder, or IME for short) is available from Gyrometric Systems Limited which uses a wheel of the same type as used in a shaft encoder. In this system, instead of sensing the features at only one location around the periphery of the disc, they are sensed at three or more locations using separate sensors. The signals from the sensors are processed to indicate radial displacement of the spindle.

In order to achieve great accuracy, timing values or time stamps for the passage of each feature by each encoder are collected and processed. This system is described in EP 0608234. The processing can be understood in principle by considering the effect of a lateral displacement of the spindle at one of the sensors. The displacement would produce a phase shift indicative of the displacement in a direction parallel to a tangent to the wheel at the position of the sensor concerned. By processing all the timing values measured at all the sensors a full plot of radial displacement can be produced. Effectively, run out will result in fluctuations in the relative phase of the signals from the different sensors and these phase differences provide an indication of the run out of the spindle.

Currently available IME system can measure run out with high accuracy on spindles rotating at speeds ranging from 0.5 rpm to 20,000 rpm. However, in order to achieve high accuracy, special attention needs to be paid to the manner in which the system is set up. In particular, if the run out of the spindle of a machine tool is to be measured, one needs first to mount the wheel so that its centre coincides with the axis of the spindle or else one needs to remove mathematically the error caused by the orbital motion of the wheel. The sensors must then be mounted so that they cannot move relative the machine bed but they too must be correctly centred on the disc while making contact with the neither the wheel nor the spindle.

An important application for IME’s is in off shore wind farms. The turbine blades of a wind turbine are connected to a rotor of an electrical generator by a spindle, usually constructed as a hollow shaft, that may have a diameter of up to 2m. The clearance between the rotor and stator of the generator may be only of the order of 6mm and contact between them would result in a catastrophic and very expensive failure of the wind turbine. Monitoring of the run out of the shaft is therefore of vital importance as it enables wear in the bearings supporting the shaft to be detected before contact occurs.

A problem is encountered in the application of IME technology to wind turbines on account of the large size of the monitored shaft. Whereas small encoder discs can readily be purchased or manufactured with the features to be detected by the sensors positioned or formed accurately relative to one another, large diameter encoder discs cannot readily be manufactured to the desired tolerance, nor can they readily be fitted to the shaft to be monitored.

Summary of the invention

In accordance with the present invention, there is provided a system for monitoring rotation of a shaft which comprises an encoder disc and at least one sensor for detecting features on the encoder disc as the disc rotates with the shaft relative to the sensor, wherein the disc is an assembled disc composed of a plurality of segments, each segment bearing a respective set of features and being separately securable to the shaft, and a processor for receiving output signals generated by the sensor in response to passing of the features and applying a correction to data derived from the output signals of the sensor to compensate for relative misalignment of the features of the assembled disc.

The invention addresses the problem of producing a large-scale encoder disc by resorting to an assembled disc formed of initially separate segments that only form a complete encoder disc when fitted to the shaft. As correct alignment of the segments cannot be guaranteed to the necessary accuracy, on account of the manufacturing tolerance of both the segments themselves and of their fixings to the shaft, the invention further proposes analysing the signals of the sensor to determine relative misalignment and applying a correction to the signals when necessary.

In some embodiments, the sensor is a magnetic sensor and the disc segments are made of a sheet metal (which is preferably ferromagnetic), the features detectable by the sensor being holes in the sheet metal, two edges of each hole being radial with respect to the axis of rotation of the monitored shaft.

The joins between adjacent segments when mounted on the shaft preferably pass through holes so that any gaps between segments should not be mistaken by the sensors for additional features but may modify the width of the sensed features as measured in the circumferential direction of the shaft.

To form a circumferentially continuous encoder disc of sufficient rigidity to avoid distortion during operation, each segment may suitably be secured along its radially inner edge to a flange of the monitored shaft and bridge pieces may be provided to connect adjacent segments to one another adjacent their radially outer edges.

Because it is possible with a small disc to drive it so that it rotates at constant speed, the exact angular position of the features can be measured with little difficulty by analysing the output signals of the sensor while the disc is rotated. However, in the case of the rotor of a wind turbine, one cannot guarantee that the speed of rotation of the shaft will remain constant over one complete revolution.

To mitigate this problem and enable the precise angular position of each feature on the assembled encoder disc to be ascertained so as to enable subsequent compensation for relative misalignment of the features, in an embodiment of the invention the processor is programmed to derive an estimate of the speed of rotation of the shaft within a time window straddling or adjacent to the time of passage of two features past the sensor and the separation of the two features is determined from the measured time between the sensor signals resulting from the passage of the features and the estimated speed of the shaft during the time window.

Assuming for example that the exact angular separation between features numbered five and six is to be determined. During a first rotation of the encoder disc, all features are assumed to be correctly aligned and an approximate measurement of the speed of rotation of the shaft is made, for example by measuring the time between passage of the first and the tenth feature. The speed is best calculated over a window of features around the interval to be estimated and the window size should take into consideration possible fluctuations in speed over one revolution. Based on this speed measurement, one can arrive at a first better estimate of the angular distance between the featured numbered five and six by multiplying the time between the signals that they produce by the derived speed estimate. This process is repeated for all the features so that a first estimate of the distance between all the pairs of features can be ascertained.

This data, which takes into account possible misalignment of the features, will now be stored in a look-up table and the process repeated. The measurement of speed within the time window surrounding the passage of the features numbered five and six will now be known with greater accuracy and this will in turn be reflected in the measurement accuracy of the angular distance between these two features. By this iterative process an accurate calibration look-up table for the angular separation between all pairs of features can be built up.

From the table of angular intervals, a table of absolute angles from the once per revolution index can be made, by adding the angular intervals. If the interval angles do not add up to one revolution (360 degrees) due to cumulative errors, a correction may be made to scale the results, so that the sum is 360 degrees.

It is assumed above that there needs to be a reference radius relative to which the angular position of each feature is measured. This reference can be provided either by a separate sensor or by a discontinuity in the sequence of the features. For example, if the holes in the disc are on a nominally fixed pitch, one of the holes can be omitted or the separation between two adjacent holes can be omitted to create one hole that is wider than all the rest.

As an alternative to the calibration procedure described above, it is possible to provide two sensors that are spaced closer to one another than the angular distance between features. The same feature will now pass through both the first and the second sensor before the next feature reaches the first. From the time taken by the first feature to traverse the distance between the two sensors, one can arrive at a measurement of the speed of rotation of the shaft and that measurement can then be used to determine the angular separation between the first and the second feature. In this alternative method of estimating the current speed the iterative method to refine the speed estimation used in first method is not relevant. However, the “two-sensor” method could employ the use of a filtered window of readings about the interval being estimated (e.g. three). Further accuracy may be obtained by repeating the estimation and statistically calculating a best estimate.

During operation of the system, the look-up table is constantly updated to arrive at an accurate determination of the angular position of the features, which will even take into account possible variations that may occur during use, such as are a result of thermal expansion.

The monitoring system of the invention can be used in an IME to measure run out by the provision of a plurality of sensors around the circumference of the same encoder disc, as described in EP 0 608 234. If a system of the invention is used at each end of a shaft through which torque is transmitted, it is possible to measure the angular rotation of one end of the shaft relative to the other and thereby obtain an estimate of the torque loading on the shaft.

Brief description of the drawings

The invention will now be described further, by way of example, with reference to the accompanying drawings, in which:

Figure lisa section through a hollow shaft fitted with an assembled encoder disc associated with three pairs of sensors,

Figure 2 is a view of the segment of the encoder disc in Figure 1 that includes a missing feature,

Figure 3 is a waveform diagram of part of a signal produced by each of the sensors, and

Figure 4 is a flow chart indicating a process that can be adopted to compensate for positional inaccuracy of the features in the assembled encoder disc.

Detailed description of the drawings

Figure 1 shows a tube 10 fitted with an encoder disc 12 comprising eight segments 12a to 12h each extending over an arc of 45°. In practice, the number of segments would depend on the size of the tube 10 and the size to which each segment can be manufactured. Each segment 12a to 12h is made of a sheet of mild steel, with a typical thickness of 6mm, which is cut to the desired shape, such as by means of a laser or a waterjet.

The features in the disc 12 that are detectable by sensors are holes 14 or windows that are cut into the sheet metal at nominally regular angular intervals. The sides of the holes lie on radii of the tube 10 and the number and size of the holes 14 would again depend on the size of the tube 10. As an example, in the case where the tube 10 is a drive shaft connecting a generator to the blades of a wind turbine, the tube 10 may have a diameter of the order of 2m and the holes may be typically 1cm wide and spaced from one another by 1cm. This will result in the angular separation of the features being a little over 1°. The segments are secured a flange that is formed on of welded to the tube 10 by bolts passing through holes 16. To provide added stability, the radially outer sides of the adjacent individual segments are connected to one another by bridge pieces 22.

The segment 12a is shown to an enlarged scale in Figure 2, which shows holes 24 ion the segments and holes 26 in the bridge pieces for enabling the bridge pieces to be bolted in position. Instead, the bridge pieces may be spot welded to the segments or securing by crimping. The segment 12a differs from all the others in that one of the holes 14 is omitted, this being done intentionally to enable identifying of a reference radius of the encoder disc 12. As an alternative, this segment could have an additional feature, such as a projecting tab, or a hole on a different radius, that is detected by a separate sensor. As can be see from Figures 1 and 2, each of the segments has half a hole 14 at each end, so that the junction between each pair of adjacent segments passes through a hole 14. As a result, misalignment of the segments will only result in an error in the width of a hole, not in the creation of gap that could itself be mistaken for a feature.

The encoder disc 12 in Figure 1 is part of an IME system measuring run out, which for this purpose requires three sensors distributed around the circumference of the disc 12. In the illustrated embodiment, each sensor is formed as a sensor pair. Thus, there are two sensors 18a and 20a located at the 12 o’clock position and two further pairs 18b, 20b and 18c, 20c located near the 4 o’clock and 8 o’clock position, respectively.

Because the disc 12 is assembled from segments, the positional accuracy of the holes 14 around the circumference of the disc cannot be guaranteed. There is inevitably a tolerance in the position of the holes in the flange to which the segments are bolted and, to allow for this, the holes 16 in the segments must be enlarged, which adds to the positioning tolerance of the segments. The effect of possible misalignment and build-up of manufacturing tolerances is demonstrated by the two waveforms shown in Figure 3.

A mild steel disc with holes cut through it is intended for use a magnetic sensor. Such sensors are preferred as they can operate reliably in hostile and dirty environments, as demonstrated by the fact that they are used in motor vehicles to determine the angular position of the crankshaft. As the disc rotates and the holes pass through a sensor, an output signal of the sensor increases then decreases gradually. After suitable amplification, this signal is applied to a threshold detector and thus produces the square wave signals shown in Figure 3, wherein the rising and falling edges of the pulses coincide with the passages of the edges of the holes through the sensor. Such processing of the sensor output signals is conventional and need not be described in detail in the present context.

The waveform A in Figure 3 shows the signal that should result from a perfectly formed disc rotating with constant speed. The pulses are of constant width and occur at precise regular intervals. Waveform B, on the other hand, illustrates the signal that one may obtain in practice when using the assembled disc of the present invention. It will be seen that there are differences dl to d4 in the time of arrival of the leading edge of the pulses from the idealised waveform A and that the pulse width is not uniform. These differences can be caused by a combination of factors, amongst them the errors in manufacture of the holes and in misalignment of the holes. These differences can importantly also be caused by variations in the speed of rotation of the shaft.

The assembled segmented disc can only be calibrated after it has been assembled on the shaft of a wind turbine and it needs to be noted that this large shaft is further connected to very large blades (possible 100m in length). On account of its vary large size and the variation of the wind forces acting on the blades, it is not possible to ensure that the tube 10 rotates at constant speed over a full revolution.

In one embodiment of the invention, this problem is overcome by adopting the iterative process represented by the flow chart in Figure 4. This process involves creating a look-up table in which an estimate of the positional error of each feature and assuming that, on account of the large moment of inertia, the rate of change of the speed of rotation can only be very small and the speed can be assumed to be constant if measured only over a small arc in the immediate vicinity of each feature.

In order to ascertain the identity of each feature, it is necessary to determine its angular position on the disc and for this purpose it is necessary to identify a reference position on the disc. This task is performed in step SI where position 0 is identified from the missing hole.

In step S2, as each feature Fi, F2, ... .F_n passes a sensor, its time or arrival ti... t_n is recorded. The system which processes the sensor output signals includes a look-up table. The look-up is a read/write memory having as many addresses as there are features and there is stored within each memory address a correction error to be added to, or a correction factor to be applied to, the measured time of arrival to compensate for the fact that the separation of the feature from the preceding feature is greater or less than the nominal separation between the features. In the first iteration of the process no correction is applied in step S3 and the actual times of arrival are used to compute positional errors of the holes.

In step S4, it is assumed that the shaft speed does not change significantly during passage of a small arc of the shaft past the sensor and that any irregularities in the times of arrival of the features within that arc are caused by misalignment of the features. On the basis of this assumption, errors in the angular positions of the features are determined and, in step S5, they stored in the look-up table.

Take for example the tenth feature Fio , one can make a prediction of its time arrival as being the time of arrival t9 of the previous feature F9 plus one sixth of the time between the previous six features F3 to F9, i.e. t9 + ((t9 -1=) / 6), or one sixth of the time between the three preceding and the three following features, i.e. t9 + ((tn -1?) / 6). Such a prediction is compared with the time tlO, the actual time of arrival of the feature F10, and a value indicative of the error in the angular position of the feature F10 is stored in the look-up table. If the value to be stored in the look-up table is to be a measure of angular displacement, the time error would be multiplied by the speed of rotation of the shaft at the time of the measurement.

In the first iteration, it was assumed that the time t9 could be used as a starting point for predicting the time tio and that the time difference between t9 and t3 or between ti3 and t?, as the case may be, could be relied upon as a correct indication of speed measurement. In the second iteration, the errors in t3 and t9 will have been reduced to some extent in step S3 allowing a more accurate determination of the positional errors to be stored in the look-up table and as the process is repeated the compensation accuracy increases.

Once the look-up table has been fully populated and time readings taken over several cycles of rotation of the shaft, a point will be reached when the stored values will cease to require updating and those values are then used during operation of the IME to correct measurements made by the sensors to compensate for inaccuracies in the angular position of the features.

The routine shown in Figure 4 is not merely performed once to calibrate the encoder disc but is run at all times that the IME is in operation to take into account possible movement or distortion of the individual segments.

The above description is not intended as a complete description of the system for processing the sensor signals and only relates to the routine for compensation for random errors in the angular positioning of the features. The system may addition perform other routines, for example to correct for systematic errors by checking for example that the sum of the errors measured over the entire circumference of the encoder is zero.

The processing described above is performed at each of the sensors 18a to 18c and 20a to 20c. If the sensors use different look-up tables, cross referencing of the tables provides an additional safeguard against errors resulting from inaccuracy in the angular positioning of the features. As an alternative, however, they may all share the same lookup table allowing the calibration of the encoder disc to be completed more rapidly.

The illustrated embodiment allows an alternative calibration routine to be adopted because the provision of two sensors at each sensing location, e.g. 18a and 20a, allows a measurement of the instantaneous speed of rotation of the shaft to be made from the time taken for the same feature to pass between the two sensors. That speed measurement can be used to predict the time of arrival of the immediately following feature and a positional error can be determined and stored in the look-up table in the same manner a described previously. In this case, an error in the relative position of the sensor 18a and 20a would give rise to an incorrect estimation of the speed but such a systematic error would be detected as the sum of the errors in the look-up table would not be equal to zero.

The above description relates only to the processing of the signals of one sensor to enable the precise angular position of the shaft as measured at that sensor to be measured at all times. If two such shaft encoders are mounted at locations that are spaced from one another axially along a shaft, a comparison of the precise measurements of the different angular positions of the shaft at the two locations can enable the torsion in the shaft to be measured and thereby allow the torque transmitted through the shaft to be estimated.

The precise measurements at three angular spaced positions allow run out to be measured, as is known from EP 0 608 234.

When used in a wind turbine, it is important that the IME equipment should have a very long MTBF (mean time between failure) as access to the equipment for repair or maintenance would be difficult and costly. It is therefore vital for the equipment to be robust. As the assembled segmented disc is attached at numerous locations to the torque transmitting tubular shaft, it cannot move relative to the shaft nor separate from it. Being make of sheet metal it is unlikely to break.

While the provision of two sensors at each location increases the probability that a sensor might fail, as the sensors in each pair do not work in conjunction with one another but either one can perform the task of the other, the duplication halves the probability of failure.

Magnetic sensors have no moving parts and can operate in hostile environments. It is for this reason that they are preferred to alternative sensors such as optical sensors.

Claims (12)

1. A system for monitoring rotation of a shaft, comprising an encoder disc, and at least one sensor for detecting features on the encoder disc as the disc rotates with the shaft relative to the sensor, wherein the disc is an assembled disc composed of a plurality of segments, each segment bearing a respective set of features and being separately securable to the shaft, and wherein a processor is provided for receiving output signals generated by the sensor in response to passing of the features and applying a correction to data derived from the output signals of the sensor to compensate for relative misalignment of the features of the assembled disc.

2. A system as claimed in claim 1, wherein the sensor is a magnetic sensor and the disc segments are made of a sheet metal, the features detectable by the sensor being holes in the sheet metal, two edges of each hole being radial with respect to the axis of rotation of the monitored shaft.

3. A system as claimed in claim 2, wherein the joins between adjacent segments when mounted on the preferably pass through holes.

4. A system as claimed in any preceding claim, wherein each segment is secured along its radially inner edge to a flange of the monitored shaft and bridge pieces are provided to connect adjacent segments to one another adjacent their radially outer edges.

5. A system as claimed in any preceding claim, wherein the disc includes an irregularity to indicate a reference radius relative to which the angular positions of all the features are measured.

6. A system as claimed in claim 5, wherein the features are disposed around the assembled encoder disc with a nominally constant pitch and one of the features is omitted to indicate the reference radius.

7. A system as claimed in any preceding claim, wherein the processor is programmed to derive an estimate of the speed of rotation of the shaft within a time window straddling or adjacent to the time of passage of two features, to predict the time of passage of the second feature through the sensor based on the estimated speed of rotation of the shaft, and to measure the difference in time between the predicted and actual times of passage of the second feature past the sensor.

8. A system as claimed in claim 7, wherein the processor includes a look-up table for storing for each feature an error value indicative of the deviation of the angular position of the feature from a nominal reference position, the error value being calculated from the measurement of the difference in time between the predicted and actual passage of the feature past the sensor.

9. A system as claimed in claim 8, wherein the storage of values in the lookup table is performed during consecutive cycles of rotation of the shaft and wherein the recorded times of passage of features that are used to predict the time of passage of each feature are modified using data stored in the look-up table from a previous cycle.

10. A system as claimed in any one of claims 1 to 6, wherein at least at one sensing location the system is provided with two sensors that are spaced closer to one another than the angular distance between adjacent features and wherein the processor is operative (i) to estimate the speed of rotation of the shaft from the time of passage of a feature past the two sensors, (ii) to predict the expected time of passage of each feature from the time of the passage of the preceding feature and the estimated speed of the shaft, and (iii) to store in a look-up table an error value indicative of the deviation of the angular position of each feature from a reference position based on the difference between the predicted time and the measured time of passing of each feature.

11. A system as claimed in any preceding claim, wherein multiple sensing locations are distributed about the circumference of the encoder disc and the processor is programmed to determine run out of the shaft based on comparison of phase measurements at the different sensing locations.

12. A torque transmitting shaft fitted with two systems as claimed in any preceding claim spaced from one another along the length of the shaft, wherein a processor is operative to compare phase measurements made by the two system to provide an indication of the magnitude of the torque transmitted through the shaft.