GB2455801A - Monitoring the rotational speed of a shaft - Google Patents

Abstract

Monitoring the rotational speed of a shaft, in particular a shaft having a number of spaced-apart features on or rotating with the shaft, e.g. a gas turbine shaft. The arrangement comprises means whereby the "time of flight" of one feature between two sensors is determined, the period of time for successive features to pass a sensor is determined and averaged over a number of feature passages, a correction factor is calculated based upon the ratio of the successive passage period to the average, and the correction factor is applied to the "time of flight". This allows one to monitor with high accuracy and a fast response time, the rotational speed of a gas turbine shaft. The invention provides a method which allows one to accurately measure speed between two magnetic sensors (5, 6) and compensates for the effect of apparent changes in the magnetic distance between the sensors.

Description

Method and ADDaratus for Monitoring the Rotational Soeed of a Shaft The present invention is concerned with method and apparatus for monitoring the rotational speed of a shaft, in particular a shaft having a number of spaced features on or rotating with the shaft. Preferred embodiments of the invention are concerned with methods and apparatuses for processing signals in order to calculate, with high accuracy and a fast response time, the rotational speed of a gas turbine shaft.

Particular preferred embodiments of the invention that are described below in more detail may be used to generate a sequence of outputs or signals which represent the rotational speed of a gas turbine shaft. Preferred embodiments of the invention easily and effectively compensate for blade jitter and missing pulses or blades when used to process signals from a speed sensor in order to monitor the speed of a rotating turbine shaft.

The speed of a rotating gas turbine shaft is typically monitored by monitoring the movement of a magnetic toothed phonic or tone wheel which rotates with the gas turbine shaft. A magnetic speed probe monitors the changes in a magnetic field as a tooth passes through it. The passage of each tooth generates a probe signal pulse or signal peak and the probe signal train is used to calculate the rotational speed of the toothed wheel by measuring the time between successive pulses, or counting a number of pulses in a fixed time. The rotational speed of the gas turbine shaft is then derived from the speed of the phonic or tone wheel.

Magnetic variable reluctance sensors (including transformer probe sensors such as that disclosed in EP169,670) can be used to monitor the movement of a phonic wheel and therefore the rotational speed of a rotating shaft coupled thereto to the phonic wheel.

There is no easy access to the turbine shaft, so the toothed or phonic wheel is typically at a distance from the shaft and connected thereto via a long gear train.

A big disadvantage of such a system is that the gear train is expensive and heavy, and can only be replaced during a major engine overhaul. An alternative to the remote phonic wheel coupled to the turbine shaft by a gear train is to mount the phonic wheel directly on the shaft. However this requires additional space inside the engine for the wheel and probe fixture.

Eddy current sensors such as that disclosed in GB 2,265,221 can also be mounted on the outside of an engine and used to measure the rotational speed of a gas turbine shaft by directly monitoring movement of the blades mounted on the rotating shaft. If the separation between blades is known, then the rotational speed can be determined from the time between successive signal pulses where each signal pulse corresponds to passage of blade past the sensor.

Patent numbers GB 2,265,221, GB 2,223,103, US 3,984,713 and GB 1,386,035 each describe eddy current or inductive sensors which may be used to measure the rotational speed of a bladed shaft. The sensors described in these documents are speed or torque sensors, each comprising a magnet positioned so that the tips of the blades pass in close proximity to the magnet. When a blade is moving close to the sensor magnet, eddy currents are generated in the tip of the blade. These eddy currents generate their own magnetic fields which are detected by a coil located in the sensor. A rotating shaft with blades, such as that in a gas turbine, will therefore generate a series of pulses with the period between pulses representing the period between successive blades as they pass the sensor. The series of pulses can be used to determine the speed of the rotating shaft; the speed is calculated from the time measurements between the pulses.

It is possible to use other types of sensors, such as optical or RF (radio frequency) sensors, but these sensors cannot operate through the turbine casing and require direct access to the blades through a hole in the casing.

A method of calculating shaft speed by measuring a time interval between consecutive blades passing a single sensor has been described in GB 2,414,300.

The inventors of the subject application are the first to realise that the method described in GB 2,414,300 is not suitable for shaft speed measurement when a fast response time is required at low shaft speeds. Effective operation of the predictor-limiter method described in GB 2,414,300 requires processing of time information from several time intervals and, especially when the rotational speed of the shaft is low, the response time of the system can be very stow. This is clearly problematic in applications where response time and accuracy are critical.

It is known that the speed of a moving object (e.g. a turbine blade or the tooth of a phonic wheel) can be determined by measuring the time taken for that object to pass from one sensor or monitoring station to another sensor or monitoring station. However, such speed measuring arrangements which use measurement of the utime of flight" between two points require one to know the distance between the two points. This has meant that such utime of flight" measurement systems have been considered unsuitable for measuring speed in environments or systems where the distance between the two sensors varies, such as in, for example, separate spaced magnetic sensors.

The distance between two spaced sensors depends on the mechanical spacing of the two sensors and can vary due to thermal expansion of the sensor(s) mounting fixture. Furthermore when the sensors are magnetic sensors such as eddy current sensors, this distance also depends on the magnetic field geometry of the sensors and blades. This magnetic geometry depends on several parameters such as the coil, magnet, blade dimensions, blade and magnet material, electrical and magnetic properties and their dependence on temperature. Since some of these properties are difficult to measure and all of them vary, time of flight measurements have been considered unsuitable for systems using sensors in environments such as gas turbines where the distance between the sensors will vary and accurate measurements are desired.

The present invention provides a method as defined in the independent claims to which reference should now be made. Some preferred features of the invention are set out in the dependent claims to which reference should now be made.

Preferred embodiments of the present invention will be described, by way of example only, with reference to the attached figures. The figures are only for the purposes of explaining and illustrating a preferred embodiment of the invention and are not to be construed as limiting the claims. The skilled man will readily and easily envisage alternative embodiments of the invention in its various aspects.

In the figures: Figures la and lb are a schematic illustration of a speed sensor set up to determine the time intervals between successive tips of the blades of a turbine as they move past the sensor, with figure 1 a illustrating the sensor arrangement and figure lb the output signal train from the sensor; Figures 2a and 2b are a schematic illustration of two speed sensors set up to determine the time of flight period of successive tips of the blades of a turbine as they move across the spacing between two sensors from the first sensor to the second sensor, in which figure 2a illustrates the sensor arrangement and figure 2b the output signal pulse trains from the sensors; Figure 3 is a block diagram illustrating a system embodying the present invention.

Figure 4 is a diagram illustrating the effects of blade jitter on monitoring systems such as that of figures 1 and 2; Figure 5 illustrates a variation of time intervals between consecutive blades when a shaft such as that of figure 1 is accelerating at a rate of about 2000 rpm/s from a speed of about 1000 rpm; Figure 6 is a diagram illustrating the effect of a missing pulse or blade and blade jitter on monitoring systems such as that of figures 1 and 2; Figure 7 illustrates a variation of time intervals between consecutive blades when a shaft such as that of figure 1, but with five missing blades, is accelerating at 2000 rpm/s from a speed of about 1000 rpm; Figure 8 is a diagram illustrating a pulse or signal train corresponding to the sensing of a position of a turbine blade having a number of missing pulses or signal peaks corresponding to, for example, missing, worn or damaged blades; Figure 9 is a graph illustrating, for a pulse or signal train such as that of figure 8, the dependence of the ratio of the latest blade period time to the average blade period time over the last AV periods on the total number of missing pulses, M, for different numbers of detected pulses, 0, in the latest blade period, as a function of blade jitter; Figure 10 is a diagram similar to that of figure 8, but showing the presence of M missing pulses in AV periods, with D missing pulses being present in the latest blade period and the AV periods including the latest blade period: and Figure 11 is a graph showing calculated values of the ratio of the latest blade period time to the average blade period time over the last AV periods for different numbers of detected missing pulses, 0, as a function of jitter using the predictor limiter method described in GB 2, 414, 300.

In a preferred embodiment of the present invention, the speed of a gas turbine shaft 1 having, say, twenty-nine compressor blades mounted thereon is calculated based on measurements from two speed sensors such as the eddy current sensor type described in GB 2,265,221. The sensors measure the time interval between changes In the current induced therein and have their output connected to data processing apparatus. The data processing apparatus may be a digital engine control unit.

Referring to figure 1 showing a bladed gas turbine shaft 1, a shaft 4 has a disc 3 connected thereto. The disc has twenty-nine blades 2 (not all shown) fixed thereto at equal points spaced around the shaft's circumference. Each blade 2 is fixed to the disc by a pin (not shown). Each blade 2 can therefore rotate about its pin in a range of about +1-20 degrees. This rotation is known as jitter and can give rise to measurement errors.

Speed sensors 5, 6 (see Figures 2 and 3) are located near the path of the rotating blades and note the passage of each blade tip as it passes each of the sensors. Each sensor produces a pulse or a signal peak (see Figures 1 to 3) as a tip passes it.

In this text, reference is made to signal pulse or peak. For a magnetic sensor, what happens is that as a feature approaches and then moves away from the sensor, a signal similar to a sine wave results (i.e. having positive and negative peaks). The position of the pulse for the purposes of the described embodiments is usually taken to be the zero crossing point between the positive and negative peaks. Time periods are measured between respective zero crossings.

A data processor (not shown) is coupled to the speed sensors 5,6 which receives as an input the sequence of pulses generated by each of the sensors. There may be significant noise produced by the jitter effect, and missing pulses.

As illustrated in Figure 3 and discussed in more detail below, the system measures the time taken for a blade to pass from the first sensor 5 to the second sensor 6 by noting the time ToF between the pulses generated at the respective sensors by the blade. In order to then determine the rotational speed it is necessary to determine the distance between the two sensors 5, 6. A measure of distance between the sensors 5,6 is derived by dividing the time T0F between sensors 5,6 by the time between the passage of successive blades past a sensor 5 to obtain a correction or compensation coefficient or factor which is a measure of the relative proportion of the distance between the sensors to the distance between blades.

The data processor coupled to the speed sensors 5, 6 includes a counter which determines the time interval between successive signal pulses (t(1), t(2), (see Figure 1(b)) produced by one of the sensors e.g. sensor 5. The time interval between the successive signals at sensor 5 provides the blade period BPN (see Figure 3) for the NIh blade period. These time intervals may be stored in a memory.

The data processor also determines the time needed for a single blade to travel between sensors 5 and 6 spaced by distance S (Figure.2). This time interval provides the time of flight period ToFN for the Nth blade (see Figure 3). These time intervals are also stored in the memory.

The spacing, S, between the two sensors 5 and 6 should be large enough to achieve the required accuracy of time measurements for a given clock frequency.

In practice, S is limited by the available physical space in an application and is typically between 1 and 2 times the spacing between adjacent turbine blades. A preferred value of S is about 1.36 times the blade spacing where the blade tips pass sensor 5 and then pass sensor 6 after a normalised period of 1.36. If the spacing is 1 or 2, pulses from different sensors would come in at the same time to the processor for the Time of Flight calculation. A spacing of 1.5 would therefore be optimal from the point of view of reducing delay in processing the signals.

However as both sensors 5, 6, will typically be in a single sensor boxN, and it is desirable to keep the box as small as possible, a 1.36 spacing is a compromise.

As mentioned above, to monitor the shaft rotational speed it is necessary to know very accurately the distance, S, between the two sensors 5, 6. This distance depends on the mechanical spacing of the two sensors as discussed above and can vary due to the thermal expansion of the sensor mounting fixture. When the sensors are eddy current sensors, this distance also depends on the magnetic field geometry of the sensors and the blades. This magnetic geometry depends on several parameters such as the coil, magnet, blade dimensions, blade and magnet material, magnetic and electrical properties and theirs dependence on temperature.

As each blade is likely to have a different thickness and slightly different material properties such as resistivity which affect the response of magnetic sensors thereto, and the blade thicknesses of each blade may vary at different rates over time due to blade erosion, wear or damage, each blade has an associated unique sensor separation SN associated wit the Nth blade.

Since these blade properties vary with time as, for example, blade thickness changes due to wear and the magnetic properties vary with temperature, it is necessary to repeatedly calibrate (and then recalibrate) the distance SN for each blade.

The data processor and associated circuitry is used to calculate from the time of flight period measurements, the average blade speed over the distance between the sensors (i.e. T0F) 5,6, by: V= S/T (I) Where: V is the average blade speed; T is the time needed by one blade to travel between the two sensors, and SN is the distance between the two sensors for the Ntti blade.

In order to determine a measure for the spacing 5,6 between the sensors, a normalised distance, NDN, between the sensors is calculated for the Nth blade.

This can be defined as a ratio of the distance between the sensors, SN, to the nominal spacing between the turbine blades.

The normalised distance for the Nth blade, NDN is the ratio of the measured time of flight ToF to a nominal or average blade period NBPN for a number of determined blade periods up to that measured time of flight period for that Nth blade: = ! BP Where NBPN is the nominal blade period; N is the blade pulse identifier; i is the summation index; BP is the blade period; and K is the number of blades over which measurements are taken.

Preferably, the nominal blade period is determined over a single complete revolution (i.e. K corresponds to a single complete revolution of the shaft) of the shaft.

The normalised distance NDN for the Nth blade is then given by the ratio of the measured time of flight TOFN for that Nth blade to the nominal blade period NBP for that same blade period: NDN-ToF N i=N (2) 1= N-K where: N is the blade or blade pulse identifier; i is the summation index; TOFN is the time of flight BP is the blade period NDN is the normalised distance; and K is the selected number of blades.

Where there are missing or damaged blades resulting in a missing signal or signals; ToF (3)

ND

i= N -(K -M) where: N is the Blade Pulse identifier; i is the summation index; ToF is the time of flight BD is the blade period NDN is the normalised distance; K is the selected number of blades; and M is the number of missing pulses Preferably, the selected number K of time of flight periods correspond to a single complete revolution of a shaft with no missing pulses, so K-M corresponds to a single complete revolution where there are M missing blades.

The normalised distance for each blade can be determined continuously or simply repeatedly. The regularity of the calculation (and re-calculation) of the stored normalised distance depends on how much change there is believed to be between revolutions and how accurate the desired output. The more accurate the desired output and/or the more significant the rate of change, the more regular this re-calculation.

The normalised distance is stored in a memory, and is looked up for future measurements of time of flights for that blade and used to convert those time of flight measurements into an accurate rotational speed measurement.

When the normalised distance is known, the measured time of flight period for the latest blade to pass between the sensors is measured and these values are used to calculate the rotational speed of the bladed shaft, in revolutions per minute (rpm): 6OxNDN* EngineSpeed = (4)

TOFNXK

where:ToFN is the latest measured Time of Flight Period, and NDN is the normalised distance calculated from a previous revolution using equation 3 or 4.

As illustrated in Figure. 3, as each successive blade passes sensor 6, the value of the latest blade period is updated, and as each successive blade passes between sensor 5 and sensor 6, the latest time of flight period is measured. This allows the sums of the blade periods BP and time of flight periods TOFN to be updated if necessary. Hence, the normalised distance value can also be updated and the finally, the shaft rotational speed calculated and updated.

A result of the system described above is that the shaft rotational speed can be calculated from a single blade time of flight period measurement and is therefore updated as each successive blade period and time of flight period is recorded.

Shaft rotational speed can therefore be measured with a faster response time and a higher degree of accuracy than when the nominal distance N D is averaged over as revolution.

As shown in figure 4, blade jitter is caused by blades (or blade tips) not being in their nominal equally spaced positions. This offset from the nominal position results from the fact that blades are mounted on a pin and the forces acting thereon can cause the blades to pivot around the pin. This jitter causes the blade period times measured from blade to blade to differ from one blade to the next, causing it to appear as if the rotational speed is constantly changing. The speed change is not smooth and continuous, but appears as noise. Taking measurements over one revolution reduces the effect of blade jitter -for every increase in blade separation caused by jitter of a particular blade, there is a corresponding decrease because as one blade, for example, moves closer to the preceding blade, it must move further away by the same distance from its following blade.

A further limitation of using the time of flight principle described above is that it can be only used where the blade jitter movement velocity during the time period, T, required for the blade to cover the distance, S, is several orders of magnitude smaller than the blade rotational velocity.

The calibration technique described above involves a comparison of the time of flight period measurements for each successive blade that passes between the sensors 5,6, with the shaft speed measurements obtained from the cumulative total of the blade to blade period measurements recorded during a previous complete revolution of the shaft. The shaft speed over one complete revolution is calculated from the time period measurements for all of the blades fitted to the compressor disc to pass one of the sensors. The number of blades must therefore be counted to secure correct measurements.

In order to accurately calibrate the distance S, we have to know the number of blades. In operation, the speed sensors may not detect the presence of one or more blades and as a result, they will not generate a pulse signal. This may be caused by one or more blades being damaged (e.g. having a bent or broken tip), or the pulse amplitude generated by the sensors might be too small to be detected by the processing circuitry connected to the sensors. This condition shall be referred to as the presence of missing pulses'.

If one or more missing pulses are present, the apparent time between successive blades passing one of the sensors (e.g. sensor 6) -the blade period -appears much longer than the condition where there are no missing pulses (see figure 6).

This results in a low shaft rotational speed calculation compared to when there no missing pulses. It is therefore important to compensate for missing pulses when calculating shaft rotational speed.

To simplify the following description of the preferred embodiment, normalised times will be used, where 1' shall represent the nominal time period between perfectly positioned blades passing the sensors.

For example and with reference to figures 4 and 6, if there is assumed to be � 7.5% blade jitter on each blade, this will cause the blade period to vary between a minimum blade period of: BPlfl= 1-2 xO.075= 1-0.15=0.85 (5) and a maximum blade period of: BP = 1 + 2 x 0.075 = I + 0.15 = 1.15 (6) However, when one missing pulse is present, as shown in Figure. 6, the normalised minimum blade period is calculated by: BPmin= 1 + I -(2x0.075)= 1.85 (7) and a maximum blade period is calculated by: BPmax = 1+1 +2 X 0.075 = 2.15 Therefore, the presence of a single missing pulse gives a nominal normalised blade period of 2, two missing pulses give a nominal Blade Period BP of 3, and so on.

Various methods for detecting the number of missing pulses present in a number of measured blade periods are envisaged and shall now be considered in turn. In the following discussions of the methods, AV' is used for the number of blade periods over which an average blade period is taken, and M' is the total number of missing pulses present In that sample taken over AV blade periods (see Figure 8).

I

One method of detecting missing pulses is based on finding the ratio, R, of the latest blade period (LBP) to the average blade period over the last AV periods of measurements (BPAV) (see Figure. 8). In this method, the AV number of periods does not include the latest blade period.

As each successive blade tip passes one of the sensors (say sensor 5 in Figures 2, 3), a pulse is generated and the attached circuitry measures the latest blade period LBP) (the elapsed time between successive blades passing the same sensor) and these values are stored in the memory.

When a predetermined number AV of blades have been detected by the sensor, an average blade period over the AV periods is calculated.

If there are M' missing pulses in a compressor disc having K' blades and in the latest blade period, there are D' missing pulses (see Figure 8), the latest blade period (LBP) is given by: LBP= D+1� 2xj (8) where: j is the maximum value of blade jitter The total of the blade periods to be averaged is: T = AV+M�2xj.(9) The average blade period over the last AV periods is then: BPAV = T/AV (10) The values of LBP and BPAV are then used to calculate the Raflo, R, of the latest blade period (LBP) to the average blade period over the last AV periods of measurements (BPAV) is then: Ratio = LBP/BPAV=(LBP/T)XAV (11) This ratio is at a maximum when LBP has its largest value and T has its smallest value: Ratio=AVxA2J -(12) This ratio is at a minimum when LBP has its smallest value and I has its largest value:.

Rati=AVx D+1-2xj --(13) AV+M+2xi By way of an example of the detection of the total number of missing pulses in AV blade periods, Equations 12 and 13 set out above have been used to calculate the Ratio, and Ratiomin for various values of blade jitter and various values of D (number of missing pulses in last blade period) and M (number of missing pulses in a single revolution of compressor disc). Figure. 9 is a graph showing the relationship between the Ratio, R, and the jitter value, calculated for: AV = 20, 0 = 0 to 5 and M = 0 to 5.

For each value of D, the relevant lines' are the outermost lines (e.g. lines 7, 8 for 0=1 where 7 is the line corresponding to D=1, M=5, Mm, and 8 is the line corresponding to D=1, M=1, max) plotted for that value of 0, (see Figure 9).

As each successive blade passes sensor 6, the value of R (see equation 11) for the latest blade period is calculated and then plotted on the graph of Figure. 9. If the plotted value lies in the region before the relevant lines for one D value cross those for another D value -i.e. where the value of R lies within one of hatched regions A, B, C, D, E or F (Figure. 9), the value of 0 can be ascertained from the graph as there is only one possible region or outcome.

For example (see figure 9), if when AV=20 and M is between 0 and 5, the value of jitter is 0.12, and the calculated value of the ratio, R, is 1.9, R falls within hatched region B of Figure 9, the graph shows that the value for D (missing pulses in latest blade period) is therefore 1. However, if when AV=20 and M is between 0 and 5 and the value of jitter is say 0.2, the value of the ratio R will fall to the right of hatched region B, beyond the maximum allowable jitter limit. As a result, D cannot be ascertained from the graph of Figure 9. Experiments suggest that jitter values are not that great so the graph can be used as such situations are unlikely to arise.

For a turbine compressor disc having, say, twenty-nine blades, AV must be less than 29-M, but the larger it is the better resolution it has, so 20 is a compromise.

The M and 0 values are selected as 0 to 5 because, in practice, when 5 out of 29 blades are damaged, turbine vibrations are so large that the turbine must be shut down. M and/or D could however be equal to 6, 7, 8 or 9.

To simplify the process described above of determining the value of 0, the value of 0 that corresponds to a range of values of R for the latest blade period can be ascertained from Table 1.

Table 1 -Calculation of 0 for given ranges of R No. of missing Range of values of ratio Max. jitter value pulses, D 0< R<1.348 0.156 0 1.34< R < 2.18 0.128 1 2.18 < R < 2.98 0.12 2 2.98 < R < 3.75 0.12 3 3.75 < R < 4.49 0.131 4 4.49<R 5

S 0.159

As described above, it is only possible to determine D for a value of R where the jitter value does not exceed the given maximum jitter value given in Table 1, which corresponds to the jitter limit of the hatched regions A to F (Figure. 9). If the value of blade jitter exceeds this maximum permissible jitter value, the calculations cannot be performed.

The maximum value of jitter for a given turbine is established during a separate test. In practice, the blade jitter experienced by the turbine blades during shaft rotation is much smaller than the limits given in Table 1.

To establish the number of missing pulses, M, in a full revolution of the bladed shaft, the calculations have to be performed K - D times, where: K is the number of blades attached to the compressor disc, and D is the sum of detected missing pulses in the Latest Blade Period. D is measured during every measurement,so the sum of Os should give M, but the calculations are performed K -ZD times to try and avoid counting the same Ds twice In a second method of detecting the presence of missing pulses, the number of blade periods to be averaged, AV, includes the Latest Blade Period as shown in Figure. 10. Therefore, the total AV time period may include more missing pulses than in the previously described method as, in the subject method, AV includes both any missing pulses M and any missing pulses D.

S

In this method, the cumulative total of the blade periods to be averaged is: T AV+(M-D)�2xj (14) The average blade period is calculated by: BPAV = T/AV (15) The ratio of the latest blade period to the average blade period over the last AV periods of measurements is then calculated by: Ratio = LBP I BPAV = LBP/T x AV Again, this ratio is at a maximum when LBP has its largest value and I has its smallest value (Equation 9) and the ratio is at a minimum when LBP has its smallest value and T has its largest value (Equation 10).

RatiQux=AVxAV'2'J (16) RatilN=AVxA,2 (17) The same process as that described for the above method is then used to ascertain the values of D and M for use in calculating the shaft rotational speed.

When the blade jitter values are larger than the maximum permissible values listed in Table 1, or it is possible that more than 5 missing pulses may be present in K periods, it may be more reliable to use a method based on the measurements of blade to blade times using the predictor-limiter method of GB 2,414..

The predictor limiter-method removes missing pulses blade periods and therefore blade to blade time measurements give us a Reference Period.

Using this method, the number of missing pulses present in one period is obtained by calculation of the ratio: Latest Blade Period R= (18) Reference Period The latest blade period is the elapsed time between pulses produced by successive blades passing a single sensor and is calculated by: LBP =D+1�2xj (19) where: D is the number of missing pulses, and j is the jitter value.

The Reference Period is obtained using predictor-limiter calculations as described in GB 2.414,300. In normalised notation, the reference period is equal to 1, hence the ratio of Equation 18 is simply equal to the latest blade period.

The calculations must be performed K -ID times to establish M as for method 1 described above.

The results of calculations of the ratios of equation 18 are shown in Figure. 11

and Table 2.

Table 2 -Calculation of 0 for ranges of the ratio, R No. of missing Range of Values of Ratio Max. Jitter range pulses 0 0<R<1.5 0.25 0 1.5< R <2.5 0.25 1 2.5<R<3.5 0.25 2 3.5 < R < 4. 5 0.25 3 4.5 < R < 5.5 0.25 4 5.5 < R<6.5 0.25 5 6.5<R 6 0. 25 As with the previous methods of determining the number of missing pulses, the value of D is only admissible where the maximum blade jitter value is not exceeded. In practice, blade jitter of 0.25 is never observed, hence the value of O should always be ascertainable from Table 2.

Whichever method of determining the number of missing pulses present is employed, the result is the total number of missing pulses, M, that result from a complete revolution of the bladed shaft. As shown in the flow diagram of Figure.

3, when the value of M has been determined, the cumulative total blade period over a total revolution of the bladed shaft can be calculated, i.e. the sum of the individual blade periods for K-M blades, where K is the total number of blades on the shaft and can be counted.

Claims (26)

1. A system according to any preceding claim including: (I) first and second sensors for sensing movement of said features; (ii) means for measuring the time of flight between a first signal generated as a selected feature passes the first sensor and a second signal generated as the selected feature passes the second sensor; (iii) means for measuring the time between a third signal 9enerated by a feature as it passes one of the sensors and a fourth signal generated as the next feature passes the same one of the sensors; (iv) means for determining average time for a selected number of features to pass the respective sensor of step (iii); (v) means for determining a correction factor for the selected feature of step (ii) consisting of the ratio of the time of step (ii) to the average time of step (iv); and (vi) memory to store the correction factor; and (vii) means to apply the correction factor to subsequent measured times of flight of that selected feature.

2. A system according to claim I wherein a correction factor is determined for each feature on the shaft.

3. A system according to any preceding claim wherein the average time of step (iii) for the selected number of features to pass the sensor is taken over a single complete revolution of the shaft.

4. A system according to any preceding claim wherein the correction factor is periodically updated.

5. A system according to any preceding claim wherein the correction factor is a normalised distance NDN for the Nth feature on the shaft given by the equation:

TOFN

NDN i-N

i=N.K i where N Is the feature or feature pulse identifier; i is the summation index; TOFN is the measured time of flight for the Nth feature; BP is the time between the third and fourth signals for each respective pair of features; and K is the selected number of features

6. A system according to any preceding claims, wherein the features are evenly spaced around the shaft.

7. A system according to any of claims 1 to 5, wherein the features are irregularly spaced around the shaft and the selected number of features corresponds to a complete revolution of the shaft.

8. A system according to any preceding claim for measuring the rotational speed of a gas turbine shaft.

9. A system according to claim 8 wherein the spaced features are the teeth of a phonic wheel coupled to and rotating with the gas turbine shaft.

10. A system according to claim 8 wherein the spaced features are either compressor or turbine blades mounted on the gas turbine shaft.

11. A system according to claim 14 or 15 wherein the correction factor is a normalised distance and NDN is given by the equation:

TOFN NDN= i=N i

1= N -(K -M) where N is the Blade or tooth Pulse identifier; i is the summation index; TOFN is the measured time of flight for the Nth blade or tooth; BP is the time between successive blades or pulses passing one of the sensors; K is the selected number of blades or teeth; and M is the number of missing pulses or signal peaks from a signal train representing the shaft blades or phonic wheel teeth.

12. A system according to any of the preceding claims wherein a correction factor representing a normalised distance between the sensors is applied to the time taken for one of said features to move between the two sensors.

13. A method for monitoring the rotational speed of a shaft having a number of spaced features on and/or rotating with the shaft, including the steps of: (I) measuring the time of flight between a first signal generated as a first feature passes a first sensor and a second signal generated as the first feature passes a second sensor; (ii) measuring the time between a third signal generated by a feature as it passes one of the sensors and a fourth signal generated as the next feature passes the same one of the sensors; (iii) determining the average time of step (ii) for a selected number of features to pass the respective sensor of step (ii); (iv) determining a correction factor for the selected feature consisting of the ratio of the time of step (ii) to the average time of step (iii); and (vi) applying the correction factor to measured times of flight.

14. A method according to claim 13 wherein the correction factor is determined for each feature on the shaft.

15. A method according to claim 13 or claim 14 wherein the average time of step (iii) for the selected number of features to pass the sensor is taken over a single complete revolution of the shaft.

16. A method according to any of claims 13 to 15 wherein the correction factor is periodically updated.

17. A method according to any of claims 13 to 16 wherein the correction factor NDN for the Nth feature on the shaft given by the equation:

TOFN NON =

i=N-K-where N is the feature or feature of pulse identifer; i is the summation index; TOFN is the measured time of flight for the Nth feature; BP is the time between the third and fourth signals for each respective pair of features; and K is the selected number of features.

18. A method according to any of claims 13 to 17, wherein the features are evenly spaced around the shaft.

19. A method according to any of claims 13 to 17, wherein the features are irregularly spaced around the shaft and the selected number of features corresponds to a complete revolution of the shaft.

20. A method according to any of claims 13 to 14 for measuring the rotational speed of a gas turbine shaft.

21. A method according to claim 20 wherein the spaced features are the teeth of a phonic wheel coupled to and rotating with the gas turbine shaft.

22. A method according to claim 20 wherein the spaced features are either compressor or turbine blades mounted on the gas turbine shaft.

23. A method according to claim 21 or 22 wherein the correction factor is a normalised distance and NDN is given by the equation:

TOFN NDN= BPi

i=N-(K -M) where N is the Blade or tooth Pulse identifier; i is the summation index; TOFN is the measured time of flight for the Nth blade or tooth; BP is the time between successive blades or pulses passing one of the sensors; K is the selected number of blades or teeth; and M is the number of missing pulses or signal peaks from a signal train representing the shaft blades or phonic wheel teeth.

24. A system according to any of claims 13 to 23 wherein a correction factor representing a normalised distance between the sensors is applied to the time taken for one of said features to move between the two sensors.

25. A system substantially as hereinbefore described with reference to the attached figures.

I

26. A method substantially as hereinbefore described with reference to the attached figures.