ADAPTIVE FILTERING METHOD

Abstract

The invention relates to a method for filtering an input signal (3b, 4b, 5b) relative to a physical variable of a turbine engine (9), the input signal being digitised, the method implementing frequency filtering of said signal in a computer (6) of a control system (7) of said turbine engine (9), said signal being provided at the input of the computer, a digital derivative of said signal being intended for being used by the control system (7), characterised in that it involves: —detecting an amplitude variation of said variable on said input signal, by a step of generating a second derivative signal (S) of the input signal and a step of comparing a value of the second derivative value of the input signal with at least one predetermined threshold (S.sub.1 . . . S.sub.n); and —adapting the frequency filtering of said input signal as a function of the detected amplitude variation of said variable, by a step of controlling a controlled filter (PB.sub.11) capable of applying frequency filtering to the input signal, so that the controlled filter applies or does not apply the frequency filtering as a function of a result of the comparison step.

Claims

1. A method for filtering an input signal, the method implementing, in a computer of a control system of the turbine engine, frequency filtering of the input signal, the input signal measuring a physical quantity of a turbine engine, the input signal being digitized, the input signal being supplied at the input of the computer, a numerical derivative of the input signal being intended to be used by the control system, the method comprising the following steps: detecting, in the input signal, a variation of amplitude of the physical quantity, by a step of generating a second derivative signal of the input signal and a step of comparing a value of the second derivative signal with a predetermined threshold, adjusting the frequency filtering of the input signal depending on the detected variation of amplitude of the physical quantity, by a step of controlling a controlled filter capable of applying frequency filtering to the input signal, so that the controlled filter applies, or does not apply, the frequency filtering depending on a result of the comparison step.

2. The method according to claim 1, wherein the input signal passes through a second filter distinct from the controlled filter so that a frequency filtering of the input signal is accomplished by the second filter regardless of the result of the comparison step, the second filter being arranged either upstream or downstream of the controlled filter.

3. The method according to claim 1, wherein: the step detecting, in the input signal, a variation of amplitude of the physical quantity further comprises comparing the value of the second derivative signal with predetermined thresholds, and the step of adjusting the frequency filtering of the input signal depending on the detected variation of amplitude of the physical quantity further comprises controlling controlled filters, each capable of applying frequency filtering to the input signal, so that each controlled filter applies, or does not apply the frequency filtering depending on a result of the comparison to one of the predetermined thresholds.

4. The method according to claim 3, wherein: each controlled filter is capable of applying frequency filtering of the low-pass type, each controlled filter being associated with a predetermined threshold; the control of a controlled filter is configured so that the controlled filter does not apply the frequency filtering if a value of the second derivative signal of the input signal is greater than its associated predetermined threshold.

5. The method according to claim 3, comprising the steps of: receiving by a comparator the second derivative signal of the input signal at a first input of the comparator and a signal corresponding to a predetermined threshold at a second input of the comparator, generating, by the comparator, a comparison signal corresponding to the result of the comparison between the signals received at the first input and the second input of the comparator, receiving by a channel selector the comparison signal at a control input of the channel selector, each channel selector being associated with a controlled filter capable of applying frequency filtering, receiving by the channel selector at a first input channels of the channel selector a first signal being a signal received by the controlled filter associated with the channel selector, receiving by the channel selector at a second input channel of the channel selector a second signal being the output signal of the controlled filter associated with the channel selector, the output signal being the signal received by the controlled filter to which the controlled filter has applied frequency filtering, and transmitting to an output of the channel selector one of the first signal and second signal depending on the value of the comparison signal.

6. The method according to claim 5, wherein the step of generating a second derivative signal of the input signal comprises the following steps: generating a first derivative signal of the input signal by application of a differentiating filter to the input signal, generating a first derivative signal filtered by application of at least one frequency filter of the “low-pass” type to the first derivative signal, and generating a second derivative signal by application of a differentiating filter to the first derivative signal filtered.

7. The method according to claim 5, wherein at least one frequency filter of the “low-pass” type is configured to apply an average over several time steps.

8. The method according to claim 1, wherein the physical quantity of the turbine engine is the angular speed of an engine shaft of the turbine engine.

9. A computer of a control system of the angular speed of an engine shaft of the turbine engine, wherein the computer is configured to implement a method according to claim 1, to receive an input signal that measures a physical quantity of the turbine engine and to generate as output a base signal used in the control of the angular speed of the engine shaft.

10. An assembly including a turbine engine comprising at least one engine shaft and a control system of the angular speed of an engine shaft of the turbine engine, wherein the control system includes a computer according to claim 9 configured to receive an input signal measuring a physical quantity of the turbine engine and to generate as output a base signal used in the control of the angular speed of the shaft, the control system being configured to control the turbine engine on the basis of the base signal.

11. The assembly according to claim 10, wherein a numerical derivative of the angular speed of the engine shaft is used by the control system to control the turbine engine during a regime acceleration of the engine, and wherein the input signal is digitized using a quantification step and a sampling step, the control system is designed to supply angular acceleration setpoints which can have values smaller than a ratio of the quantification step to the sampling step, the input signal being a measurement of the angular speed.

12. A computer program comprising instructions configured to implement each of the steps of the method according to claim 1 when the program is executed on a computer.

Description

PRESENTATION OF THE FIGURES

[0058] Other features and advantages of the invention will still be revealed by the description that follows, which is purely illustrative and not limiting, and must be read with reference to the appended drawings in which:

[0059] FIG. 1 shows schematically a twin spool dual flow turbojet in which the speeds of rotation N1 and N2 are controlled by a control system;

[0060] FIG. 2 shows a functional diagram of an adaptive digital filter according to a particular example of implementation of the invention;

[0061] FIG. 3 shows a functional diagram of an embodiment of an adaptive digital filter according to a particular example of implementation of the invention, and comprising two thresholds;

[0062] FIG. 4 shows the dN2/dt signal over time in a stable regime, within a controlled system in which the control signal makes use of the derivative signal dN2/dt.

[0063] FIG. 5 shows the N2 signal over time in a rapid variation regime, within a controlled system in which the control signal makes use of the derivative signal dN2/dt.

DETAILED DESCRIPTION

[0064] FIG. 1 shows schematically a twin spool dual flow turbojet 9 controlled by the control system 7. At least one of the physical quantities of the turbojet is measured thanks to a sensor: the rotation speed N1 of the lowpressure spool 1 can be measured by the speed sensor 3a which produces the input signal 3b depending on N1, the rotation speed N2 of the high-pressure spool 2 can be measured by the speed sensor 4a which produces the input signal 4b depending on N2.

Other physical quantities can be measured such as for example the pressure in a zone of the high-pressure spool 2 measured by a pressure probe 5a which produces the input signal 5b.
The input signals 3b,4b and 5b are sent to a computer 6 which is a part of the control system 7.
The computer 6 delivers a base signal 8 relating to one of the physical quantities of which it has received the input signal, for example an angular speed signal.
This base signal 8 is used in the control of the turbojet, and in the production of the control signal 10 originating in the control system delivered to a control block of the turbine engine.

[0065] FIG. 2 shows a functional diagram of the adaptive digital filter according to a particular implementation of the invention.

[0066] The input signal in(t) is applied to the inputs of the differentiating filters D.sub.11 and D.sub.21, the function of which consists of replacing the digitized value received in(t) with the value

[00001]$\frac{\mathrm{in}\left(t\right)-\mathrm{in}\left(t-\Delta t\right)}{\Delta t}$

where Δt is the time sampling step.

[0067] These digital filters use calculation programs to determine the numerical derivative of the digitized input signal.

[0068] The output of the differentiating filter D.sub.11 is connected to the set of low-pass filters denoted PB.sub.10, PB.sub.11, PB.sub.12 . . . PB.sub.1n, which are arranged in series. Each low-pass filter PB.sub.11, PB.sub.12 . . . PB.sub.1n is a controlled filter associated with a channel selector denoted SW.sub.1, SW.sub.2 . . . SW.sub.n, and each low-pass filter input is connected to the first input channel E.sub.1, E.sub.2 . . . E.sub.n of its associated control means. The second input channel F.sub.1, F.sub.2 . . . F.sub.n−1 of the channel selector SW.sub.1, SW.sub.2 . . . SW.sub.n−1 is connected to the output L.sub.2, L.sub.3 . . . L.sub.n of the channel selector associated with the following filter in the series. The input F.sub.n of the channel selector SW.sub.n is connected to the output of its associated filter PB.sub.1n, and the output L.sub.1 of the channel selector SW.sub.1 is connected to the output of the adaptive digital filter and corresponds to the output signal out(t). Each channel selector SW.sub.1, SW.sub.2 . . . SW.sub.n, receives a binary control signal at its control input J.sub.1, J.sub.2 . . . J.sub.n, and, depending on the value of the control signal, it delivers at the output L.sub.1, L.sub.2 . . . L.sub.n either the signal received at the first input channel E.sub.1, E.sub.2 . . . E.sub.n, or the signal received at the second input channel F.sub.1, F.sub.2 . . . F.sub.n. The control input J.sub.1, J.sub.2 . . . J.sub.n, of each channel selector is connected to the output of a comparator element described below.

[0069] The differentiating filter D.sub.21 is connected in series with, in this order, two low-pass filters PB.sub.21 and PB.sub.22, a differentiating filter D.sub.22 and a block V which transforms the signal received at its input into its absolute value. The output of the block V is connected to the inputs A.sub.1, A.sub.2 . . . An of the comparator elements C.sub.1, C.sub.2 . . . C.sub.n. Each input B.sub.1, B.sub.2 . . . B.sub.n of the comparators C.sub.1, C.sub.2 . . . C.sub.n receives a constant and predetermined signal S.sub.1, S.sub.2 . . . S.sub.n. The signal generated at the output of the comparator corresponds to the two possible results of the comparison between its two inputs. The output O.sub.1 (respectively O.sub.2, O.sub.3 . . . O.sub.n) of the comparator C.sub.1 (respectively C.sub.2, C.sub.3 . . . C.sub.n) is connected to the control input J.sub.1 (respectively J.sub.2, J.sub.3 . . . J.sub.n) of the channel selectors.

[0070] When the system is in operation, the digitized input signal of the adaptive filter in(t) is injected into the differentiating filter D.sub.21 which produces the numerical first derivative of the signal, then PB.sub.12 and PB.sub.22 filter this derivative to avoid the quantification noise interfering excessively with the signal. D.sub.22 and V produce a signal S which gives the variation level of the real signal and the regime or amplitude swept by the measured physical quantity. This value is then distributed over the n parallel lines each of which leads to a comparator associated with a particular threshold. All the comparators operate identically: if A.sub.k≥B.sub.k the output generated at Ok equals 1, if A.sub.k<B.sub.k the output generated at O.sub.k equals 0.

The different thresholds predetermined before the filtering operation are selected so that S.sub.1>S.sub.2>S.sub.3> . . . >Sn.
Two consecutive thresholds S.sub.k>S.sub.k+1 bound the value V so that: S.sub.k>S≥S.sub.k+1
At the input of the comparator C.sub.k+1, A.sub.k+1=S≥B.sub.k+1=S.sub.k+1 and the signal O.sub.k+1 delivered at the output equals 1.
This is also the case for all the comparators for which the threshold is less than S.sub.k+1, i.e. the thresholds S.sub.k+2, S.sub.k+3 until S.sub.n.
At the input of the comparator C.sub.k, A.sub.k=S<B.sub.k=S.sub.k and the signal Ok delivered at the output equals 0.
This is also the case for all the comparators for which the threshold is greater than S.sub.k, i.e. the thresholds S.sub.k−1, S.sub.k−2 until S.sub.1.

[0071] The comparators C.sub.1 to C.sub.n deliver the signal 0 or 1 at the control input of the channel selectors SW.sub.1, SW.sub.2 . . . SW.sub.n.

All the channel selectors operate identically, the signal delivered at the output L is equal either to the signal received at the first input channel E if the control signal equals 1, or the signal received at the second input channel F if the control signal equals 0.
The comparators C.sub.1 to C.sub.k deliver signal 0 to the control input of the channel selectors SW.sub.1, SW.sub.2 . . . SW.sub.k. The signals received at the second input channels F.sub.1 (respectively F.sub.2 . . . F.sub.k) are therefore delivered to the outputs L.sub.1 (respectively L.sub.2 . . . L.sub.k).
The comparators C.sub.k+1 to C.sub.n deliver signal 1 to the control input of the channel selectors SW.sub.k+1, Sw.sub.k+2 . . . SW.sub.n. The signals received at the first input channels E.sub.k+1 (respectively E.sub.k+2 . . . E.sub.n) are therefore delivered to the outputs L.sub.k+1 (respectively L.sub.k+2 . . . L.sub.n).

[0072] Under these conditions, the output signal of the system out(t) corresponds to the digitized input signal of the adaptive filter in(t) which has passed through the differentiating filter D.sub.11, then the low-pass filter PB.sub.10, then the controlled low-pass filters PB.sub.11, PB.sub.12 . . . PB.sub.1k+1 and finally the channel selectors SW.sub.1, SW.sub.2 . . . SW.sub.k+1.

The system therefore behaves in the following manner: if S.sub.k>S≥S.sub.k+1 then the output signal of the system out(t) corresponds to the digitized input signal in(t) of the adaptive filter which has passed through k low as filters.

[0073] Under these conditions, if S increases the number k is reduced and the filtering is reduced, and conversely if S is reduced the number k increases and filtering increases.

[0074] The system therefore allows adjusting the number of controlled low pass filters which the numerical first derivative passes through depending on the value S, i.e. adjusting the intensity of filtering depending on the level of variations of the real signal, i.e. depending on the amplitude swept by the signal. In other words, in the case where the measured physical quantity represents an engine regime, the intensity of the filtering is adjusted depending on the level of variations in the regime.

[0075] In this system, the output signal of the system out(t) corresponds to the digitized input signal in(t) of the adaptive filter which has passed at least through the filter PB.sub.10, distinct from the controlled filters PB.sub.11, PB.sub.12 . . . PB.sub.1n so that a frequency filtering of the input signal is accomplished by the filter PB.sub.10 regardless of the result of the comparisons within the comparators C.sub.1, C.sub.2 . . . C.sub.n.

[0076] In other words, the input signal (3b, 4b, 5b) passes through filter PB.sub.10 regardless of the value of the signal S which gives the level of variation of the real signal and the regime or the amplitude swept by the measured physical quantity.

[0077] The filter PB.sub.10 is arranged upstream of the controlled filters PB.sub.11, PB.sub.12 . . . PB.sub.1n, but the filter PB.sub.10 can also be arranged downstream of these controlled filters, and be positioned after the output L.sub.1 of the channel selector SW.sub.1.

[0078] This particular implementation example of the invention requires low-pass filters, which are elements known to a person skilled in the art, like filters with finite or infinite impulse response, weighted moving averages, low pass filters of order 1, order 2, order 4, etc. A 2-step moving average receiving an input signal U(t) delivers the signal

[00002]$\frac{U\left(t\right)+U\left(t-\Delta t\right)}{2}$

where Δt is the time sampling step. A 4-step moving average corresponds to the formula

[00003]$\frac{U\left(t\right)+U\left(t-\Delta t\right)+U\left(t-2\Delta t\right)+U\left(t-3\Delta t\right)}{4}.$

FIG. 3 shows an adaptive digital filter functional diagram according to a particular implementation example of the invention, corresponding to the system shown in FIG. 2 in the case where n=2, with a single difference: the low-pass filter PB.sub.10 is placed after the output of the last channel selector instead of being placed right after the differentiating filter D.sub.11.

[0079] Once again the output signal of the system out(t) corresponds to the digitized input signal in(t) of the adaptive filter which has passed at least through the filter PB.sub.10, distinct from the controlled filters PB.sub.11 and PB.sub.12. The input signal (3b, 4b, 5b) passes through the filter PB.sub.10 regardless of the value of the signal S which gives the level of variation of the real signal and the regime or the amplitude swept by the measured physical quantity.

[0080] This time, the filter PB.sub.10 is arranged downstream of the controlled filters PB.sub.11, and PB.sub.12, but the filter PB.sub.10 can also be arranged upstream of these controlled filters, and be located between the differentiating filter Di 1 and the controlled filter PB.sub.11.

The behavior of this system is exactly that of the system described by FIG. 2 in the case where n=2 and where: [0081] PB.sub.11, PB.sub.12 and PB.sub.21 are averages over 4 steps [0082] PB.sub.10 and PB.sub.22 are averages over 2 steps.
The system therefore allows adjusting the number of filters passed through (between 1 and 3) by the first numerical derivative depending on the value S compared to two thresholds, i.e. adjusting the intensity of filtering depending on the level of variations of the real signal, i.e. on the amplitude swept by the signal representing the measured physical quantity.

[0083] FIG. 4 shows the signal dN2/dt over time in the stable regime, within a controlled system, the control signal of which makes use of the derivative signal dN2/dt. The four curves correspond to four different filtering situations of the derivative signal: dN2/dt is either filtered by a low-pass filter of order 4 (curve 31) or filtered by an average over 4 steps (curve 32) or filtered by an adaptive digital filter (curve 33) according to one implementation example of the invention corresponding to the functional diagram of FIG. 2, or not filtered (curve 34).

All these curves are centered on the value zero because, the angular speed N2 being constant, the derivative dN2/dt is zero on average. The curve 34 shows the highest signal-to-noise ratio; it is due essentially to the sampling noise, which is not filtered.
Curve 31 shows the lowest signal-to-noise ratio; it corresponds the highest filtering.
Curve 32 has a signal-to-noise ratio located between those of curves 31 and 34, because it corresponds to less filtering than the low-pass filter of order 4.
Finally, curve 33 has a signal-to-noise ratio located between those of curves 31 and 32. This shows that the performance of the adaptive digital filter in a stable regime are better than for a 4-step moving average but poorer than those of a low-pass filter of order 4.

[0084] FIG. 5 shows the signal N2 over time in a regime of rapid variations, within a controlled system with a loop that makes use of the derivative signal dN2/dt. The four curves correspond to the four filtering situations previously described for FIG. 4: a low-pass filter of order 4 (curve 41) or filtered by average over 4 steps (curve 42), adaptive digital filter (curve 43) according to a particular implementation example of the invention corresponding to the functional diagram of FIG. 2, not filtered (curve 44).

Curve 44 again shows the highest signal-to-noise ratio, still due to the unfiltered sampling noise. The signal varies considerably in passing from approximately 1200 rpm to 600 rpm, then it increases again until the value 1000 rpm.

[0085] The curve 41 shows the greatest delay relative to the unfiltered signal: this signal does decrease after curve 44 has reached its minimum and the signal even continues to decrease when curve 44 has already increased.

Furthermore, the amplitude of the variation of the unfiltered signal is not restored, the signal 41 passes from 1200 rpm to 0 rpm, then increases again toward 700 rpm.
Curve 42 has a smaller delay relative to the unfiltered signal and better restores the amplitude of the variation of the unfiltered signal.
Finally, curve 43 shows an even smaller delay and restores even better the amplitude of the variation of the unfiltered signal.
The adaptive digital filter therefore has better performance than the moving average over 4 points, both in a stable regime and in a regime with rapid variations. It has a favorable signal to noise ratio for control and little delay relative to the unfiltered signal in the regime with rapid variations.

[0086] This filter allows sufficiently small delay relative to the real signal and has a sufficiently high signal-to-noise ratio for controlling the system in the stable regime as in the regime with rapid variations.

[0087] The invention applies to the generation of a base signal used in the control of a turbine engine, in particular the control of angular speed of an engine shaft of the lowpressure spool or of the high-pressure spool of a twin spool dual flow turbine engine.

[0088] In particular, the adaptive digital filter is able to be used to control the turbine engine in acceleration of the engine regime. In the case where the speed of the shaft is digitized with a sampling step and a quantification step and it has quantification noise, the adaptive digital filter allows managing the control of angular acceleration setpoints which can have values less than the ratio of a quantification step to a sampling step of the digitized input signal. A numerical derivative of the angular speed is then used for control; this derivative can be generated by an adaptive filter as presented above.

[0089] The base signal can be generated based on an input signal relating to the angular speed N1 of the lowpressure spool, or N2 of the high-pressure spool, or of any engine shaft of a turbine engine in general. It can be generated from a measurement signal relating to other physical quantities of a turbine engine such as for example a pressure, the measurement of which is used in the control of the turbine engine.

[0090] The invention is not limited to the embodiments described and shown in the appended Figures. Modifications remain possible, particularly from the standpoint of the constitution of the various elements or the substation of technical equivalents, without however departing from the scope of protection of the invention.