The invention relates to the field of monitoring actuators in aircraft, and specifically the monitoring of electromechanical actuators with an estimation of the game (in English "backlash") present in the actuator.

These actuators are, for example for mechanically actuating the flaps on aircraft wings.

S game described here is a mechanical play corresponding to the distance or the angle should follow a PI part of a mechanical system before transmitting motion or force to another part of said system P2 (see Figure 1). This phenomenon is common in mechanical coupling components, such as gearboxes, vis-a-ball, satellites roller bearings, etc. In general, the game is felt upon reversal of the actuator of the operating direction.

The knowledge of this game is needed to improve fault detection, and to establish diagnosis and prognosis, as used in methods of "Prognosis and Health Monitoring (PHM)."

In particular, it seeks to know the magnitude of the game, which is a constant and does not depend on the value at a given moment of the game from the actuator.

STATE OF THE ART

Several publications have proposed models and implementations of the game.

A common approach is to treat the game as a dead zone. Other approaches consider the part concerned as elastic, so that the game is included in this elasticity.

In 2007, Lagerberg [REF NL-EKF] developed a non-linear estimator for the amplitude and the game state using the Kalman filter formalism, wherein the estimator consider the system as alternating between a two linear modes, called "contact mode" and "game mode" in an absolute reference. But the process is costly in terms of resources.

There is thus no quick and effective method to evaluate this data.

PRESENTATION DE L'INVENTION

The invention provides a method for characterizing the amplitude of mechanical play of an electromechanical actuator, the electromechanical actuator comprising an electric motor and a movable member adapted to be moved by the motor, said clearance being defined as a distance or angle that must navigate through one of the motor and the movable member before transmitting motion or force to another,

the actuator comprising an own upstream sensor to measure a position of the motor and a clean downstream sensor to measure a position of the movable member, and an own effort sensor measuring a load applied to the movable member,

the method comprising the steps of:

- (El) Measurement of the motor position, the position of the movable member and the load applied to the movable member by means of three sensors,

- (E2) From computer modeling of the actuator and the measured position of the motor, the measured position of the movable member and the load applied to the movable member measured estimate of a value of backlash using a state observer, considering the game as a component of the state,

- (E3) Determination of the amplitude of the set from a set of dynamic variable the values ​​obtained in the previous step for different times.

The invention may include the following characteristics, taken alone or in combination:

the set S (t) is modeled as:

5(t) = 0 + ô(t)

where S is the game, b is a random white noise known spectral density, and t is time.

a component of the state of the state observer is a 5X gap (t) between the measured position of the movable member X2 (t) and the measured position of the motor Xl (t), so that the vector state x is expressed as follows:

x = [SX SX S] T

OÙ SX = X2 - X1,

and the measured output variable of the state observer is expressed in the following form:

y = [SX]

the state observer is modeled as follows:

x = Ax(t + Bu(t + Mw t)

y(t) = Cx(t) + v(t)

v and w are respectively disturbances and measurement noise, u (t) being the command, with u (t) being equal to the load applied to the movable member as measured by the force sensor,

with

1 T

0 — 0

C = [1 0 0]T

where K is a stiffness of the actuator, f a damping coefficient, m s a mass of the movable member,

- the state observer is an observer Kalman

- Kalman estimator is discretized,

- by selecting x (kT s ) = x (k) where T s is a sampling period, the observer is defined by the Kalman discrete state equations:

±(k + 1) = Adx(k) + Bdu(k) + Mdw(k)

y{k) = Cdx{k) + v{k)

or

Md = ln

c d = c

where I n is the identity matrix.

the amplitude of the game is calculated as a difference between an average of the maximum values ​​of estimated game and an average minimum of the set values ​​estimated from the set of values ​​of the set,

the amplitude of the AS game is calculated as follows

or

Srm = {s(k) , sm≥→sm +2 m,n lsm]

- the actuator is mounted on an aircraft and wherein the step of measuring (EI) is carried out with an operational flight profile.

The invention also relates to a monitoring unit for characterizing the amplitude of a set of an electro-mechanical actuator, electro-mechanical actuator comprising an electric motor and an own movable member to be moved by the motor , said clearance being defined as a distance or angle that must navigate through one of the motor and the movable member before transmitting motion or force to another,

the actuator comprising an own upstream sensor to measure a position of the motor and a clean downstream sensor to measure a position of the movable member, and an own effort sensor measuring a load applied to the movable member,

the unit being configured to:

- From computer modeling of the actuator and the measured position of the motor, the measured position of the movable member and the load applied to the movable member measured, estimating a value of the game using an observer considering the game state as a component of the system status,

- determining the amplitude of the set from a set of values ​​of the set obtained in the previous step for different times.

The invention also relates to an electro-mechanical actuator and associated monitoring unit, the monitoring unit being in conformity with what is described above.

PRESENTATION DES FIGURES

Other features, objects and advantages of the invention will become apparent from the following description, which is purely illustrative and not restrictive, and must be read with the accompanying drawings, wherein:

- Figure 1, already presented, represents a type of game in question in the present description,

- Figure 2 shows schematically an electromechanical actuator may be characterized in the context of the invention,

- Figure 3 schematically shows two actuators of Figure 2 installed in parallel, according to one embodiment not limited to,

- Figure 4 shows schematically a model of the actuator,

- Figure 5 represents the position (normalized) of different elements of the actuator as a function of time (normalized),

- Figure 6 shows a load applied to the actuator (normalized) as a function of time,

- Figure 7 shows the relative position (normalized) between the actuator elements as a function of time,

- Figure 8 shows the value of the estimated set (normalized) as a function of time,

- Figure 9 shows the estimated and set the speed of the actuator as a function of time,

- Figure 10 represents the estimated and set load of the actuator as a function of time.

DETAILED DESCRIPTION

A method for characterizing the AS amplitude of a set S (t) of an electro-mechanical actuator 10 will be described.

An example of electro-mechanical actuator is shown in Figure 2.

The illustrated actuator 10 is a linear displacement actuator. The actuator 10 includes a housing, a motor 11 and a movable part in the form of a movable rod 12. The motor 11 comprises a stator 14 fixedly mounted relative to the housing and a rotor 13 rotatable relative to the stator 14 . the movable rod 12 is movable in translation relative to the stator 14 of the motor 11 in a direction of displacement parallel to the axis of rotation of the rotor 13. the actuator 10 further includes a plurality of planetary rollers 16 arranged between the rotor 13 of the motor and the moving rod 12. rotation of the rotor 13 relative to stator 14 causes a displacement of the rod 12 in translation relative to the stator 14. for this purpose, the rotor 13 includes a threaded inner surface and the rod 12 includes a threaded outer surface.

The WO2010072932 discloses such an actuator.

Another example relates to actuator displacement rotary actuators, wherein the movable part is movable in translation, and the rotor of the motor causes the movable part to rotate relative to the stator.

3 shows an actuator assembly of a flight control rudder comprising two electromechanical actuators 10

and 20 identical to the actuator of FIG 2. In this case, the two electromechanical actuators 10, 20 are arranged in parallel between a frame 100 and a control surface 200. The advantage of such a configuration will be explained subsequently . In a particular embodiment, the frame 100 is an aircraft wing and the control surface 200 is a fin for controlling the rolling motion of the aircraft: the actuators used to rotate the fin relative to the wing of the aircraft as a function of the phase of flight (takeoff, landing, etc.). The connection between the frame 100 is each actuator 10, 20 is typically performed using a fixed ball joint and the connection between the rod 11 of the actuator 10, 20 and the fin 200 is typically carried out using a movable ball.

Each actuator comprises:

- an upstream sensor 41 capable of measuring a Xl position (t) of the engine 10, with respect to a fixed reference (e.g., motor stator), the position of XI engine consists more precisely in the position of the rotor 13 (in radian),

- a downstream sensor 42, capable of measuring a position X2 (t) of the movable piece 12 relative to a fixed reference (e.g., motor stator),

- a force sensor 43, capable of measuring a FI load applied to the movable member.

This information is collected by a processing unit. The processing unit comprises a computing unit and a memory for storing information. We also discuss surveillance unit to designate a processing unit configured to implement certain process steps.

The processing unit may comprise a computer on board the aircraft or a computer on the ground which has been transmitted the data.

The actuator 10 is modeled by computer using a model Mod. It is necessary to establish an equivalent dynamic model: The system rotor coupled to a screw can be considered as a transmission chain between two inertias or two masses. In the example of the actuator in Figure 2, the first inertia is the motor 11 and the second inertia is the transformation of the mass of the movable part 12 along the axis of the engine, using or not the ratio r reduction. It is thus possible to establish a linear position Xl (t) to the rotor in function of its angular position 91 (t):

Linear position is thus similar theoretical position calculated from the angular position.

C Ceettttee ééqquuiivvaalleennccee ppeeuutt êêttrree mmiissee eenn œœuuvvrree ssuurr tthhee ppiièèccee mmoobbiillee ddaannss llee ccaass dd''uunn aaccttiioonnnneeuurr AA mmoouuvveemmeenntt rroottaattiiff ..

P Paarrttaanntt ,, llee mmooddèèllee eesstt iilllluussttrréé eenn ffiigguurree 44 ,, ooùù XX eesstt llaa ppoossiittiioonn ddaannss llee rrééfféérreennttiieell aabbssoolluu ((mm)) ,, ff eesstt llee ccooeeffffiicciieenntt dd''aammoorrttiisssseemmeenntt ((NN // ((mm // ss)) ),), ,, KK to eesstt llaa rraaiiddeeuurr, (, (NN // mm),), ,, to new SS, (, (tt),), eesstt llee jjeeuu ,, mm mm eesstt llaa mmaassssee eeqquuiivvaalleennttee dduu mmootteeuurr, (, (eenn ttrraannssllaattiioonn),),, (, (KKgg),), ,, mm ss eesstt llaa mmaassssee ddee llaa ppiieeccee mmoobbiillee, (, (kkgg),), ,, R and llaa Fi is eesstt cchhaarrggee aaeerrooddyynnaammiiqquuee, (, (NN),), ,, FFff eesstt lleess ffrrootttteemmeennttss, (, (NN),), ..

LL''iinneerrttiiee eenn ttrraannssllaattiioonn eesstt ddoonnnnééee ppaarr llaa ttrraannssffoorrmmaattiioonn ssuuiivvaannttee ::

Where Jm is the moment of inertia.

The force sensor 43 measures the load Fi, but all applied forces includes the load and friction:

F ext = I + T

A model given by Karam [REF-Karam] models the friction according to the following equation:

Ff = FVX + signe{x) [Fdry + lF ] ( 1)

Wherein X is the position in the absolute frame of reference, the overall friction Ff (N), the viscous friction Fv parameter (N / (m / s)), FDRY dry friction and η mechanical efficiency.

Friction include engine friction and those of the screw.

Considering the friction Ff negligible before the FI load, we get:

The method for characterizing the amplitude AS of the status of the actuator 10 comprises the steps El to E3 below.

In a first step E, the motor angular position 91 (t) of the motor is measured by the upstream sensor 41. As indicated above, at this angular position 91 (t) is given an equivalent linear position XI (t). In this same step, the position of the movable element X2 (t) is measured by the downstream sensor 42, and the applied load Fl (t) to the moving element is measured by the upstream sensor 41.

This data is then retrieved by the processing unit to be processed.

5 illustrates in solid lines XI and X2 in dotted lines (normalized values). The first graph in Figure 5, XI and X2 positions appear confused because of the scale: in fact, the relative difference in XI and X2 is low by absolute values. However, by enlarging the scale, as shown in the second graph of Figure 5, a difference is observed between XI and X2.

6 illustrates the load Fi (normalized values).

To implement the second stage E2, the variable considered is more defined in the absolute frame, but in a relative reference in relation to the motor inertia. We define:

Sx = x 2 - x 1

The fundamental principle of dynamics applied to the inertia of the movable member gives:

m X ' = - {K + ΑΚ) (δΧ - δΧ 0 - £) - (+ Af) Ox + F ext Where AK and Af are respectively uncertainties of stiffness and damping, δΧ 0 is the initial difference, which may be selected as zero.

The identification scheme of the game is based on a Kalman filter using the above equation. The main assumption is to consider that the dynamic variable S (t) is modeled as a full state contaminated with white noise which is in turn regarded as external disturbance, in the following form:

S(t) = 0 + b(t)

where s is the game, b is a random white noise known spectral density and unbiased, and t is time.

The setting equation of the Kalman filter is based on a state observer, whose condition is expressed as follows:

x = [tt d ' x s] T

and the measured output variable:

y = Sx

The observer is modeled by the following equations:

(x = Ax(t) + Bu(t) + w(t)

{ y(t) = Cxit) + v(t) ( )

v and w are respectively disturbances and measurement noise, u (t) being the command, with u (t) being equal to the applied load = Fext to the movable member as measured by the force sensor 43,

with

1

B = 0 — 0

C = [1 0 0]T

where K is a stiffness of the actuator, f a damping coefficient, m s a mass of the movable member.

Once established model and set equation, the Kalman filter must be implemented.

The estimator is discretized in a Ts time period. The measures are sampled at Ts.

Noting:

x(kTs) = x(k)

the discrete equations are given by

±(k + 1) = Adx(k) + Bdu(k) + Mdw(k)

y{k) = Cdx{k) + v{k)

These matrices are approximated on the basis of general solutions of the continuous system given by equation (5), and by integrating between times to = kT s and t = (k + l) T s ,

or

Md = /,n

c d = c

where L is the identity matrix.

The Kalman filter is defined by two steps: the prediction step and the update step.

The prediction step is described by the following equations:

x(k + 1 I k ) = Adx(k I k) + Bdu(k)

P {I k k + 1) = A d P {k I k) + M d W d M d T The update step is described by the following equations:

Kf (k + 1) = P(k + 1 I k). CdT{CdP{k + l \k)CdT + Va)-1 x(k + 1\ k + 1) = x{k + 1 \ k) + Kf (k + 1). ( (fc + 1) - Cdx{k + l \k) - Ddud(k + 1))

P (k + l \ k + l) = (s n - K f (k + l) C d ) P (k + l \ k) wherein Wd and Vd are respectively noise covariance matrices treatment and measurement, with known spectral densities, x (k + l \ k) and x (k + l \ k + 1) are predicted and estimated covariance state. P k + l \ k + l) and P k + l \ A :) are the prediction errors and covariance of the estimation error.

7 illustrates the estimated δΧ relative position, based on the data positions XI, X2 and load R shown in Figures 5 and 6.

Step E2 is typically implemented using the processing unit.

At the end of step E2, the dynamic variable forming the set S (t) is estimated. Each time value kTs, we therefore have a value of the set S (t).

The estimated set S (t) oscillates between a maximum value and a minimum value, as shown in Figure 8 (normalized values ​​between -1 and 1). The variations observed at the extreme values ​​are due to the load measured by the force sensor 43, which actually corresponds to the load R and Ff friction.

The aerodynamic load defines a positive charge, which tends to move the ball to the upper limit, while the dry friction,

are being added or subtracted depending on the sign (cf. equation (1)) oscillate the game around a given value. Figures 9 and 10 illustrate these situations by comparing the game with the estimated speed of a hand and the game with the other load. Note in Figure 9 that changes (around the maximum value) depends on the sign of the velocity and it is noted in Figure 10 that the value of the set S (t) is reversed with the load R.

It is found that fluctuations play S (t) is negligible compared to its values, which suggests that the approximation that the friction Ff are negligible compared to the R load is valid.

Finally, from estimates of the set S (t), in a third step E3, the amplitude of the AS set S (t) is determined from a set of values ​​estimated in step E2.

This assembly preferably includes a portion only of the estimated values.

For example, a method of determining the amplitude may be to take the average value of maximum values ​​of the estimated and set the average value of minimum values ​​of the estimated game, then subtract the two averages. It is then a kind of amplitude "peak to peak" averaged.

This method is formalized as follows:

OR

¾t) = {¾ / 5(t) < -wt)1 *lsMI)

This method has the advantage of being applicable regardless of the timing of the upstream and downstream position sensors Indeed, because of the offset error of these sensors, the 0 point of the mechanical play is never located middle of the range of the game.

The purpose of the step E3 is to achieve constant data regarding the set S (t) that characterizes the actuator. In practice, with the aging and wear of the actuator, the amplitude AS tends to increase. Nevertheless, the relative time scale for the implementation of the process, the amplitude can be considered a representative constant of the system status of this implementation period.

Step E3 is typically implemented by the processing unit. The obtained value of the amplitude AS is then stored in the memory.

The value of the AS amplitude obtained by the described method can then be used in the state of an actuator monitoring processes as part of a signature to the test actuator and is to be classified among a set of signatures from a database. A high value of the amplitude AS may be a sign of fatigue and / or aging of the actuator. By combining this data with other data characterizing the actuator, as coefficients of stiffness of the shaft, damping, or values ​​relating to travel of the stem, it is possible to establish classes of values ​​for actuator according to different states: in good condition, damaged state, fault state, for example.

The process described requires the acquisition of data (see step E). This acquisition can be performed during a flight with a profile

operational, that is to say during times when the actuator is requested by the aircraft to its flight (pitch, roll, take-off, etc.). The process can then be implemented without requiring an update specific provision of the aircraft, which optimizes costs and human time. When the acquisition is in flight, the method can be implemented with the use of a single actuator, as the load is applied directly by the physical effects of the flight.

Alternatively the tests may be made on the ground, on the tarmac or in a shed, during a check. In this case, the embodiment of the actuators as shown in Figure 3 is advantageous in so far as the actuator 20 can be biased to exert a load on the surface 200 while the actuator 10 is being tested.

In the case of an embedded processor, the steps E2 and E3 are performed either in flight or afterwards. In the case of a ground processing unit, steps E2 and E3 are performed after data recovery step El.

References

[NL-EKF] : Lagerberg, A. and Egardt, B. (2007). Backlash Estimation With Application to Automotive Powertrains. IEEE Transactions On Control System, Vol . 15, No. 3, May 2007.

[KARAM] Karam W. generators of static and dynamic forces high power electromagnetic technology. Ph. D. dissertation, University of Toulouse, 2007.

claims

1. A method for characterizing the amplitude (AS) of a set (S (t) of an electro-mechanical actuator (10), the electromechanical actuator comprising an electric motor (11) and a movable member ( 1) adapted to be set in motion by the motor, said clearance (S) being defined as a distance or an angle (m, rad) must travel one of the motor (11) and the movable member (12) before transmitting motion or an effort to another,

the actuator (10) comprising an upstream sensor (41) adapted to measure the motor position (XI (t)) and a downstream sensor (42) capable of measuring a position of the movable member (X2 (t)), and a force sensor (43) adapted to measure a load applied to the movable member (Fl (t)),

the method comprising the steps of:

- (El) Measurement of motor position (XI (t)) of the position of the movable member (X2 (t)) and the applied load (Fl (t)) to the movable member to the using three sensors,

- (E2) From computer modeling of the actuator and the measured position of the motor (Xl (t)) of the measured position of the movable member (X2 (t)) and the applied load ( fl (t)) to the movable member measured estimate of a mechanical set value (S (t)) using a state observer considering the play (S (t)) as a component state, the set S (t) is modeled as:

s(t) = 0 + b(t)

where S is the game, b is a random white noise known spectral density, and t is time,

a component of the state of the state observer is a 5X gap (t) between the measured position of the movable member X2 (t) and the measured position of the motor Xl (t), so that the vector state x is expressed as follows:

x = [tt d ' x s] T

OÙ SX = X2 - X1,

and the measured output variable of the state observer is expressed in the following form:

y = [5xF

- (E3) Determination of the amount of play (AS) from a set of dynamic variable the values ​​obtained in the previous step for different times (t).

2. The method of claim 1 wherein the state observer is modeled as follows:

x = Ax(t + Bu(t + Mw t)

y(t) = Cx(t) + v(t)

v and w are respectively disturbances and measurement noise, u (t) being the command, with u (t) being equal to the load applied to the movable member (12) measured by the force sensor (43 )

with

0 1 0

-K -f K

A =

m s m s m s

0 0 0

1

B = 0

ms

C = [1 0 0]T

where K is a stiffness of the actuator, f a damping coefficient, m s a mass of the movable member.

3. A method according to one of claims 1 to 2, wherein the state observer is a Kalman observer.

4. The method of claim 3, wherein the estimator of Kalman is discretized.

5. Proceed in the Bible a quelconque des revendications précédentes, the amplitude est du jeu (AS) is recalculated at a talented entre une moyenne des valeurs du jeu maximales Estime (S) et une moyenne des valeurs du jeu minimales Estime (S ) Models ensemble de valeurs du jeu.

6. Process according to claims 4 and 5 in combination, wherein the amount of play (AS) is calculated as follows:

OR

max[S(/c)] + min[S(/c)]

Sp(fc) = 5(fc) / 5(fc) >

7. A method according to any preceding claim, wherein the actuator is mounted on an aircraft and wherein the step of measuring (EI) is carried out with an operational flight profile.

8. A surveillance unit to characterize the amplitude (AS) of a set (S (t)) of an electromechanical actuator, the electromechanical actuator comprising an electric motor and an own movable member to be moved by the engine, said set being defined as a distance or an angle (m, rad) must travel one of the motor and the movable member before transmitting motion or force to another,

the actuator (10) comprising an upstream sensor (41) adapted to measure the motor position (XI (t)) and a downstream sensor (42) capable of measuring a position of the movable member (X2 (t)), and a force sensor (43) adapted to measure a load applied to the movable member (Fl (t)),

the unit being configured to:

- From computer modeling of the actuator and the measured position of the motor (Xl (t)) of the measured position of the movable member (X2 (t)) and the applied load (Fl (t )) to the movable member measured, estimating a value of the clearance (S (t)) using a state observer considering the play (S (t)) as a component of the system status , the set S (t) is modeled as:

5(t) = 0 + ô(t)

where S is the game, b is a random white noise known spectral density, and t is time,

a component of the state of the state observer is a 5X gap (t) between the measured position of the movable member X2 (t) and the measured position of the motor Xl (t), so that the vector state x is expressed as follows:

x = [sx sx

OÙ SX = X2 - X1,

and the measured output variable of the state observer is expressed in the following form:

and

- determining the amount of play (AS) from a set of values ​​of the set obtained in the previous step for different times (t).

9. electro-mechanical actuator and associated monitoring unit, the monitoring unit being in accordance with claim 8.