FORM 2
THE PATENTS ACT, 1970
(39 of 1970) &
The Patents Rules, 2003
COMPLETE SPECIFICATION
(See section 10, rule 13)
A SYSTEM AND A METHOD FOR SUPPRESSING DISTURBANCES IN A FEEDBACK CONTROL SYSTEM
SAAB AB, a Swedish company of 581 88 Linkoping, Sweden
The following specification particularly describes the invention and the manner in which it is to be performed.

A system and a method for suppressing disturbances in a feedback control sys¬tem
Technical field
The present invention relates to a system and a method for suppressing frequency variant, narrow band disturbances in a feedback control system as defined in the preamble of claim 1 and 10, respectively.
Background of the invention
Interference or noise can be introduced in electronic control systems in many differ¬ent ways, for example, by having components or sensors susceptible to disturbances in their surroundings, such as variant magnetic fields, mains noise, a shaky and/or noisy environment etc. Various methods have been proposed in the prior art for es¬timating and alleviating such interferences.
For example, as a military tank is driving towards a destination, there is a need for stabilizing the gun barrel in order to be able to, for example, fire the gun while the vehicle is moving. A lot of things affect the elevation and the direction of the gun barrel while the tank is in motion, such as accelerating or using the brakes of the tank, the surrounding terrain, the road or off-road surface condition etc. Thus, for alleviating the effects of such factors a feedback control system, for ex¬ample a servo system, can be used to control and stabilize the gun barrel.
However, aside from the factors mentioned, another factor also affecting the stabil¬ity of the elevation and direction of the gun barrel is the shaking and thumping of the interconnected links of the driving bands of the tank. The driving bands consist of interconnected metal links rotating around driving wheels, where the intercon¬nected metal links sink into the ground and pushes the tank forward. Depending on

primarily the speed of the tank, the interconnected links will cause a frequency vari¬ant, narrow band disturbance to be introduced into the servo system. This frequency variant, narrow band disturbance will cause problems in servo systems according to prior art.
The reason for this is that the frequency variant, narrow band disturbance is located at frequencies that exceed the bandwidth of the feedback control system, that is, the servo system. It is therefore difficult to design the feedback control system so as to work satisfactorily in terms of all system design parameters, i.e. performance, stabil¬ity, robustness and accuracy etc., over the entire frequency spectrum of interest.
So depending on the characteristics of the interferences and how they are introduced in the electronic control system, there are situations, as described above, where none of the prior art solutions in a satisfying way can eliminate or suppress these distur¬bances.
Object of the invention
It is an object of the present invention to provide a stabile and robust system and method for suppressing frequency variant, narrow band disturbances in a feedback control system with an improved performance and accuracy.
Summary of the invention
The object of the present invention is achieved by a system for suppressing fre¬quency variant disturbances in a feedback control system, wherein the system is a inner feedback loop in the feedback control system comprising a real system con¬nected to a first filtering means, which is connected to a frequency estimation means and a system modelling means, wherein said system modelling means is connected to a second filtering means via a subtraction means, characterized in that the fre-

quency estimation means comprises means for estimating the main frequency of a disturbance, the first filtering means removes the frequency content in a first feed¬back signal around the estimated main frequency, the second filtering means re¬moves the frequency content in a second feedback signal around the estimated main frequency, the system modelling means comprises means arranged to use an second input signal from the first filtering means and the estimated main frequency for modelling and compensating for the damping and the phase loss of the real system, and the subtraction means is arranged to subtract an first output signal from the sec¬ond filtering means from an second output signal from the system modelling means to get an estimated disturbance signal added to the first input signal of the inner feedback loop.
The object of the present invention is further achieved by a method for suppressing frequency variant disturbances in a system as previously described, characterized in comprising the steps of estimating the main frequency of said disturbance, filtering out the frequency content in a first feedback signal and a second feedback signal around said estimated main frequency, modelling and compensating for the damping and the phase loss of said real system using said estimated main frequency and a second input signal, producing a damping and phase compensated second output signal, producing an estimated disturbance signal by subtracting a first output signal from said second filtering means from said second output signal, and adding said estimated disturbance signal to the first input signal of said inner feedback loop.
An advantage of the above described invention is that it can estimate frequency variant, narrow band disturbances without having to model/design the inverse of a real system. This is advantageous since designing the inverse of the real system so that the model correspond well regarding both added gain and phase shift across all desired frequencies, generally is very difficult. Another advantage is that the present invention can model a non-minimum phase system, whereas in the prior art the in¬verse of the real system becomes unstable.

Preferably the system may also comprise first and second filtering means that are frequency variant, narrow band pass filters arranged to be adjusted in accordance with the estimated main frequency. This feature enables fast and accurate extraction of the main frequencies of the disturbances.
Preferably the system modelling means of the present invention may further com¬prise a third filtering means arranged to add a 90° phase shift to a second input sig¬nal at the estimated main frequency, a first gain multiplier arranged to remove the added gain of the third filtering means, a first and a second signal generator ar¬ranged to generate a third and fourth output signal respectively, a first and a second multiplier arranged to multiply the third output signal of the first signal generator with a fifth output signal of the first gain multiplier and the forth output signal of the second signal generator with the second input signal respectively, a summing node arranged to sum a sixth and seventh output signal of the first and second multiplier and a second gain multiplier arranged to remove the added gain of the real system. An advantage with this feature is that it uses simple trigonometric relationships to achieve an output signal that easily can be used to estimate the frequency variant, narrow band disturbance.
Preferably steps corresponding to this feature are also comprised in the method ac¬cording to the present invention.
Preferably the third filtering means in the system modelling means of the present in¬vention is a frequency variant, low pass filter arranged to be adjusted in accordance with the estimated main frequency. As previously mentioned, this feature enables fast and accurate extraction of the main frequencies of the frequency variant, narrow band disturbances.

Preferably the system may further comprise a first and a second signal generator ar¬ranged to generate the third and forth output signals in dependence of a table and also a second gain multiplier arranged to add a gain in dependence of the table, wherein the table comprises the phase shift and added gain of the real system at dif¬ferent frequencies. An advantage with this feature is that the table contains stored data that, for example, can be acquired by frequency sweeping, i.e. calculate a Bode diagram, the real system and can be easily retrieved by the first and second signal generator as well as the second gain multiplier.
In the method according to the present invention this feature may be enabled by fur¬ther comprising the step of generating the third and fourth output signals from a first and a second signal generator, respectively, and the step of removing the added gain of said real system by using a second gain multiplier is performed in dependence of a table comprising the phase shift and added gain of the real system at different fre¬quencies.
The frequency estimation means of the present invention may further comprises cal¬culation means arranged to use a predetermined, continuous function of a measur¬able system variable of the real system. This feature enables estimation of the main frequencies of the frequency variant, narrow band disturbances by inputting a meas¬urable system variable to the predetermined, continuous function.
In the method according to the present invention this feature may be enabled by hav¬ing the step of estimating the main frequency of said disturbance further comprise the step of using a predetermined, continuous function of a measurable system vari¬able of said real system.
Preferably the frequency estimation means of the present invention may further comprise calculation means arranged to use the discrete Fourier transform to calcu¬late the frequency spectrum of the feedback signal on a predetermined interval and

extraction means arranged to extract the frequency which has the maximum energy peak value in the frequency spectrum and set the extracted frequency as the esti¬mated main frequency of the disturbance. This feature enables a more accurate and precise estimation of the main frequencies of the frequency variant, narrow band disturbances.
In the method according to the present invention this feature may be enabled by hav¬ing the step of estimating the main frequency of said disturbance further comprising the steps of using the discrete Fourier transform to calculate the frequency spectrum of said feedback signal on a predetermined interval, extracting the frequency which has the maximum energy peak value in said frequency spectrum and setting said ex¬tracted frequency as the estimated main frequency of said disturbance.
Preferably the predetermined interval may be determined by using the predeter¬mined, continuous function of the measurable system variable of the real system.
In the method according to the present invention this feature may be enabled by hav¬ing the step of estimating the main frequency of said disturbance further comprise the step of using said predetermined, continuous function of said measurable system variable of said real system to determine said predetermined interval.
Further, the measurable system variable of said real system can be the speed of the driving bands of a military tank.
Brief description of the drawings
The present invention will be described in more detail in the following, with refer¬ence to the appended drawings, in which:

Fig. 1 illustrates an example of how an out-of-band interference can be introduced into a servo system;
Fig. 2 shows a system according to the prior art;
Fig. 3 shows a first embodiment of a system according to the present invention;
Fig. 4 shows a second embodiment of system according to the present invention.
Description of the preferred embodiments
The description of the present invention will refer to embodiments as implemented for controlling the elevation and direction of the gun barrel of a military tank. This should, however, not be considered as limiting the present invention to these em¬bodiments, but instead be viewed as illustrative examples.
Fig. 1 schematically illustrates an example of a feedback control system 1, repre¬sented by the dotted lines, for controlling the output signal y of a real system G. In the feedback control system 1, the output signal y from the real system G is fed back to and subtracted from a reference signal r. The resulting signal z is used as an input into a controller A outputting the input signal u to the real system G. Also il¬lustrated in Fig. 1, is how a disturbance d is introduced and added to the input signal u before being inputted into the real system G which in turn outputs the signal y. Because of reasons mentioned in the background above, a disturbance can often be present in the input signal to the real system G, for example, as depicted in the feed¬back control system 1 in Fig. 1, and if the frequency or frequencies of the distur¬bance is below the bandwidth of the feedback control system 1, the feedback control system can be designed to eliminate or alleviate the effects of this disturbance on the output signal y. However, if the frequency range in which the disturbance is located is above the bandwidth of the feedback control system 1, the disturbance will re-

main and ultimately affect the output signal y, since it can not be attended to by the feedback control system 1. The disturbance d is such an out-of-band disturbance, that is, the frequency range of the disturbance is located outside of the bandwidth of the feedback control system 1, and is therefore depicted in the feedback control sys¬tem 1 as introduced before the real system G, and thereby also introduced in the feedback control system 1.
An example of the feedback control system 1 illustrated in Fig. l is the servo system for controlling the elevation and direction of the gun barrel on a military tank. The out-of-band disturbance signal d is in this case the previously mentioned frequency variant, narrow band disturbance caused by the interconnected links of the driving bands.
The feedback control system 1 could of course be any form of closed loop control system and is simply depicted for the benefit of providing a skilled person with a better understanding of the present invention. As an example, the reference signal r could be an output signal from a control unit, joystick etc., operated by a tank opera¬tor e.g. for aiming the gun barrel at a target, and the controller A could be could be a PID-controller designed and adapted for controlling the feedback control system 1.
Fig. 2 schematically illustrates an inner feedback loop for eliminating an out-of-band disturbance according to prior art.
An out-of-band disturbance signal d is introduced and added to a signal v, which comprises an input signal u and an estimate d' of the disturbance signal d. To obtain the estimated signal d', the output signal y is fed back into an estimated model G'"1 of the inverse of the real system G. From the output of the estimated model G'"1, the signal v is subtracted and the resulting signal sent to a frequency variant filter F(f). From the filter F(f), an estimated signal d' of the disturbance d is outputted.

In order to arrive at the model G'"1, a frequency analysis, e.g. a Bode-diagram, of the real system G is performed. By adapting a linear system to the measured data from the frequency analysis and inverting the linear system, an estimated model of the inverse system G'-1 is created.

The transfer function of the inner feedback loop in Fig. 2, from the input signal u and the disturbance signal d to the output signal y, is shown in equation 2:
The estimated model G'-1 is also arranged to be causal according to the following equation 1 in the Laplace domain:

However, a problem in the prior art is that it is very difficult to find a linear system which can, after being adapted to the measured data of the real system G and in¬verted and made causal, accurately correspond to the behaviour of the real system G across the whole desired frequency range.
Another problem in the prior art is when the real system G is a system with a non-minimum phase behaviour. This implies that the linear system adapted to the meas¬ured data of the real system G has a transfer function in the Laplace domain which has zeros in the right half plane of the s-plane. Consequently, the inverse of the lin¬ear system adapted to the measured data of the real system G'"1 has a transfer func¬tion in the Laplace domain which has poles in the right half plane, which means that is G'"1 is unstable.
These problems render the prior art solutions difficult to design in view of all sys¬tem design parameters, i.e. performance, stability, robustness and accuracy etc., over

the entire frequency spectrum of interest and are therefore not able to eliminate out-of-band disturbances in a satisfying way.
Fig. 3 illustrates a first embodiment of the present invention comprising an inner feedback loop for eliminating an out-of-band disturbance.
As in Fig. 1, an out-of-band disturbance signal d is added to a signal v, which com¬prises an input signal u and an estimate d' of the disturbance signal d. To get the es¬timated signal d', the output signal y is first fed back into to an adjustable, narrow band pass filter Bi which is arranged to output a signal S.
For the adjustable, narrow band pass filter Bl, the main frequency f0 of the distur¬bance signal d can be approximately estimated by the frequency estimation means 3A according to the following equation 3:

The known, continuous function F(x) of the measurable system variable x, for ex¬ample the speed of the driving bands, can be achieved empirically by testing and ex¬perimenting on the real system G. Therefore, the narrow band pass filter B1, can be designed as time-variant band pass filter Bi(f), wherein the band pass frequency is set to be the estimated main frequency f, which means that the variation of the time-variant band pass filter B1(f’) is adapted to the variation of the main frequency f0 of the frequency variant, narrow band disturbance.
Then, in the inner feedback loop, the signal S can be written according to either one of the following equations 4-5:



wherein the equation 5 the real system G has shifted the phase of the input signal v
+ d at the frequency f with the angle cpG and multiplied it with a gain This, of
course, since the adjustable, band pass filter Bi is a narrow band pass filter with zero phase shift and gain 1 (0 dB) at the frequency f, which removes most of the fre¬quencies outside the main frequency f0 of the frequency variant, narrow band distur¬bance d.
The signal S is used as input into a low-pass filter A(F) and also into a multiplier M2. The low-pass filter A(f) is designed to add a phase shift of 90° to its input sig¬nal S by also utilizing the same known, continuous function F(x) as the adjustable,
narrow band pass filter Bi(P). Unfortunately, it also adds a gain of -=to the signal
S. The resulting signal is therefore multiplied by a gain - V2 by using a gain multi¬plier KA. This means that the output signal S1 from the gain multiplier KA in com¬parison to the signal S has changed sign and has been shifted 90° before being in¬putted into the multiplier M1.

It should be understood that the main purpose of the low-pass filter A(f) is to add a phase shift of 90° to the input signal S and for which there are several different filter designs applicable. In this case, the consequential added gain of the low-pass filter A(P) is just a bi-product of the filter and could be removed in any suitable way. The signal S1 can therefore be written according to the following equation 6

The signals S1 and S, wherein S is the output signal from the adjustable, narrow
band pass filter B1(f), are then multiplied with the signals S2 and S3 respectively in

me multipliers Mi and M2. The signals S2 and S3 are created in signal generators SG2 and SG3, respectively, and according to the following equations 7 and 8:

where the angle (pG is the phase shift of the real system G at the frequency f which is obtained from a table of system parameters T that has been achieved by a fre¬quency analysis of the real system G.
The outputs from the multipliers M1 and M2 are then added to form the signal S0. Because of the trigonometric relationships between the output signals from the mul¬tipliers Mi and M2, the signal S0 becomes identical to the output from the filter B1(f) although including a phase shift of the angle q>G. The trigonometric relation¬ships in the signal S0 can be written according to the following equations 9-11:

Therefore the summed output signal S0 can be written according to the following equation 12:


Then the summed output signal S0 is, by using a gain multiplier KG, multiplied with a gain KG, also obtained from the previously mentioned table of system parameters T. This results in that the attenuation, or damping, and phase loss which the real sys¬tem G has enforced on the control signal v and the disturbance signal d at the fre¬quency f resulting in the output signal y have been compensated for without having to invert an adapted linear system model of the real system G.
The signal v, for which most of the frequencies outside the main frequency f0 of the frequency variant, narrow band disturbance d has been filtered out by being passed through an adjustable, narrow band pass filter B2(f), identical to the adjustable, nar¬row band pass filter B1(f), is then subtracted from the signal S' leaving the resulting signal d' comprising an estimate of the disturbance signal d. As can be seen from Fig. 3, this results in that the estimated disturbance signal d' is subtracted from the input signal u and, since d' approximates d, thus eliminates or at least suppresses the effects of added the disturbance signal d on the output signal y from the real sys¬tem G.
This means that the present invention can eliminate or suppress the effects of a dis¬turbance signal introduced to the real system G, even when the real system G is a non-minimum phase system, without resulting in an unstable control loop. It also means that the present invention can eliminate or at least suppress out-of-band, fre¬quency variant noise in a feedback control system, for example, the feedback control system 1 in Fig. 1.
Fig. 4 illustrates a second embodiment according to the present invention. The inner feedback loop depicted in Fig.4 is identical to the inner feedback loop in Fig. 3, ex¬cept that this second embodiment also comprises frequency estimation means 4A instead of 3A, which includes calculation means DFT and extraction means max, for improving the estimation of the main frequency of the out-of-band, frequency variant disturbance d.

So, in addition to the output signal y being fed back into to the adjustable, narrow band pass filter B1(f), the output signal y is also fed back into the calculation means DFT. The calculation means DFT uses the discrete Fourier transform to cal¬culate the frequency spectrum of the output signal y on a predetermined interval I, defined according to the equation 13:

where f0 is estimated by the predetermined known continuous function F(x), accord¬ing to equation 3 described in reference to Fig. 2, and a is a suitable parameter de¬pending on, and is preferably calibrated for, each specific implementation of the present invention.
The extraction means max then searches across the frequencies of the frequency spectrum of the interval I for a peak energy value. Once a maximum energy value for the frequency spectrum has been found, the frequency f* where this maximum energy value occurs is extracted and set as the main frequency of f0 of the distur¬bance signal d. The frequency value f* is then used for adjusting the band pass fre¬quencies of the filters B1(f*) and B2(f*), and also for adjusting the filter A(f*). Also, it should be noted the calculation means DFT and the extraction means max can easily be implemented integrally and are separated herein only for the sake of clarity.
The calculation means DFT enables the inner feedback loop to automatically and more accurately estimate the main frequency f0 of the disturbance signal d.

We Claim:
1. A system for suppressing frequency variant disturbances in a feedback con¬trol system (1), wherein said system is a inner feedback loop in the feedback control system (1) comprising a real system (G) connected to a first filtering means (B1(f); B1(f*)), which is connected to a frequency estimation means (3A; 4A) and a system modelling means (3B; 4B), wherein said system mod¬elling means (3B; 4B) is connected to a second filtering means (B2(f); B2(f*)) via a subtraction means (5),
characterized in that
said frequency estimation means (3A; 4A) comprises means for estimating the main frequency (f; f*) of a disturbance (d);
said first filtering means (B1f); B1f*)) is arranged to remove the frequency content in a first feedback signal (y) around said estimated main frequency
(f; f*);
said second filtering means (B2(f); B2(f*)) is arranged to remove the fre¬quency content in a second feedback signal (v) around said estimated main frequency (f; f*);
said system modelling means (3B; 4B) comprises means arranged to use an second input signal (S) from said first filtering means (B1(f); B1f*)) and said estimated main frequency (f; f*) for modelling and compensating for the damping and the phase loss of said real system (G); and said subtraction means (5) is arranged to subtract an first output signal from said second filtering means (B2(f); B2(f*)) from an second output signal (S') from said system modelling means (3B; 4B) to get an estimated disturbance signal (d') added to an first input signal (u) of said inner feedback loop.

2. A system according to claim 1, wherein said first and second filtering means (B1(f); B1(f*); B2(f); B2(f*)) are frequency variant, narrow band pass filters arranged to be adjusted in accordance with said estimated main frequency (f; f*).
3. A system according to any of the claims 1 to 2, wherein said system model¬ling means (3B; 4B) comprises
a third filtering means (A(f); A(f*)) arranged to add a 90° phase shift to a second input signal (S) at said estimated main frequency (f ;f*); a first gain multiplier (KA) arranged to remove the added gain of said third filtering means (A(f); A(f*));
a first and a second signal generator (SG2, SG3) arranged to generate a third and fourth output signal (S2, S3) respectively;
a first and a second multiplier (M1, M2) arranged to multiply said third output signal (S2) of said first signal generator (SG2) with a fifth output signal (S1) of said first gain multiplier (KA) and said fourth output signal (S3) of said second signal generator (SG3) with said second input signal (S) respectively; a summing node arranged to sum a sixth and seventh output signal of said first and second multiplier (M1, M2) and a second gain multiplier (KG) ar¬ranged to remove the added gain of said real system (G).
4. A system according to claim 3, wherein said third filtering means (A(f); A(f*)) is a frequency variant, low pass filter arranged to be adjusted in accor¬dance with said estimated main frequency (f; f*).
5. A system according to claim 3 or 4, wherein said first and second signal gen¬erator (SG2, SG3) are arranged to generate said third and fourth output signals (S2, S3) in dependence of a table (T) and also said second gain multiplier (KG) is arranged to add a gain in dependence of said table (T), wherein said

table (T) comprises the phase shift and added gain of the real system (G) at different frequencies.
6. A system according to any of the claims 1 to 5, wherein said frequency esti¬mation means (3A; 4A) comprises calculation means arranged to use a prede¬termined, continuous function (F(x)) of a measurable system variable (x) of said real system (G).
7. A system according to any of the claims 1 to 5, wherein said frequency esti¬mation means (3A; 4A) comprises
calculation means (DFT) arranged to use the discrete Fourier transform to calculate the frequency spectrum of said feedback signal (y) on a predeter¬mined interval (I) and
extraction means (max) arranged to extract the frequency which has the maximum energy peak value in said frequency spectrum and set said ex¬tracted frequency as the estimated main frequency (f; f*) of said disturbance (d).
8. A system according to claim 7, wherein said predetermined interval (I) is de¬termined by using said predetermined, continuous function (F(x)) of said measurable system variable (x) of said real system (G).
9. A system according to any of the claims 6 or 8, wherein said measurable sys¬tem variable (x) of said real system (G) is the speed of the driving bands of a military tank.
10. A method for suppressing frequency variant disturbances (d) in a system for suppressing frequency variant disturbances in a feedback control system (1), wherein said system is a inner feedback loop in the feedback control system (1) comprising a real system (G) connected to a first filtering means (B1(f);

B1(f*)), which is connected to a frequency estimation means (3A; 4A) and a system modelling means (3B; 4B), wherein said system modelling means (3B; 4B) is connected to a second filtering means (B2(f); B2(f*)) via a sub¬traction means (5),
characterized in
comprising the steps of:
estimating the main frequency (f; f*) of a disturbance (d);
filtering out the frequency content in a first feedback signal (y) and a
second feedback signal (v) around said estimated main frequency (f;
modelling and compensating for the damping and the phase loss of said real system (G) using said estimated main frequency (f; f*) and a second input signal (S); producing a damping and phase compensated second output signal
(S');
producing an estimated disturbance signal (d') by subtracting a first
output signal from said second filtering means (B2(f); B2(f*)) from
said second output signal (S'); and
adding said estimated disturbance signal (d') to a first input signal (u)
of said inner feedback loop.
11. A method according to claim 10, wherein the step of modelling and compen¬sating for the damping and the phase loss of said real system (G) using said estimated main frequency (f; f*) and the filtered first feedback signal (y) fur¬ther comprises the steps of:
adding a 90° phase shift to an second input signal (S) at said esti¬mated main frequency (f ;f*) by using a third filtering means (A(f); A(f*));

removing the added gain of said third filtering means (A(f); A(f*))
by using a first gain multiplier (KA);
generating third and fourth output signal (S2, S3) from a first and a
second signal generator (SG2, SG3), respectively;
multiplying said third output signal (S2) of said first signal generator
(SG2) with a fifth output signal (S1) of said first gain multiplier (KA)
and said fourth output signal (S3) of said second signal generator
(SG3) with said second input signal (S) by using a first and a second
multiplier (M1, M2), respectively;
summing a sixth and seventh output signal of said first and second
multiplier (M1, M2), respectively, by using a summing node;
removing the added gain of said real system (G) by using a second
gain multiplier (KG).
12. A method according to claim 11, wherein the step of generating said third and fourth output signal (S2, S3) from a first and a second signal generator (SG2, SG3), respectively, and the step of removing the added gain of said real system (G) by using a second gain multiplier (KG) is performed in depend¬ence of a table (T) comprising the phase shift and added gain of the real sys¬tem (G) at different frequencies.
13. A method according to any of the claims 10 to 12, wherein the step of esti¬mating the main frequency (f; f*) of said disturbance (d) further comprises the step of:
using a predetermined, continuous function (F(x)) of a measurable system variable (x) of said real system (G).
14. A method according to any of the claims 10 to 13, wherein the step of esti¬
mating the main frequency (f; f*) of said disturbance (d) further comprises
the steps of:


using the discrete Fourier transform to calculate the frequency spec¬trum of said feedback signal (y) on a predetermined interval (I); extracting the frequency which has the maximum energy peak value in said frequency spectrum; and
setting said extracted frequency as the estimated main frequency (f; f*) of said disturbance (d).
15. A method according to claim 14, wherein the step of estimating the main fre¬quency (f; f*) of said disturbance (d) further comprises the step of:
using said predetermined, continuous function (F(x)) of said measur¬able system variable (x) of said real system (G) to determine said pre¬determined interval (I).

ABSTRACT
A system for suppressing frequency variant disturbances in a feedback control sys¬tem, wherein the system is a inner feedback loop in the feedback control system comprising a real system connected to a first filtering means, which is connected to a frequency estimation means and a system modelling means, wherein said system modelling means is connected to a second filtering means via a subtraction means, characterized in that the frequency estimation means comprises means for estimat¬ing the main frequency of a disturbance, the first filtering means removes the fre¬quency content in a first feedback signal around the estimated main frequency, the second filtering means removes the frequency content in a second feedback signal around the estimated main frequency, the system modelling means comprises means arranged to use an second input signal from the first filtering means and the esti¬mated main frequency for modelling and compensating for the damping and the phase loss of the real system, and the subtraction means is arranged to subtract an first output signal from the second filtering means from an second output signal from the system modelling means to get an estimated disturbance signal added to the first input signal of the inner feedback loop.
Fig. 4