This Application is a National Stage of International Application No. PCT/EP2017/059447 filed Apr. 20, 2017, claiming priority based on French Patent Application No. 1653497 filed Apr. 20, 2016, the entire contents of each of which are herein incorporated by reference in their entirety.
The invention relates to the field of monitoring actuators in aircraft, and more precisely the monitoring of electromechanical actuators by means of estimation of the play (backlash) present in the actuator.
These actuators are for example intended to mechanically actuate flaps on the wings of the aircraft.
The play S specified here is a mechanical play corresponding to the distance or the angle which a piece P1 of a mechanical system has to travel before it can transmit a movement or a stress to another piece P2 of said system (see
Knowing this play is necessary to improve detection of defect, and to set up diagnoses and prognoses, as used in methods of “Prognosis and Health Monitoring (PHM)”.
In particular, the aim is to know the amplitude of the play, which is a constant and does not depend on the value at a given instant of the play of the actuator.
Several publications have proposed modelling and implementations of this play.
A common approach is to consider the play as a dead zone. Other approaches consider the relevant piece as elastic, such that the play is included in this elasticity.
In 2007, Lagerberg [REF NL-EKF] developed a non-linear estimator for the amplitude and the state of the play by means of the formalism of the Kalman filter, wherein the estimator considers the system as alternating between two linear modes, called “mode of contact” and “mode of play” in an absolute frame of reference. But the method is costly in terms of resources.
There is therefore no rapid and effective method for evaluating this datum.
The invention proposes a method for characterisation of the amplitude of mechanical play of an electromechanical actuator, the electromechanical actuator comprising an electric motor and a mobile element capable of being set in motion by the motor, said play being defined as a distance or an angle which either of the motor and of the mobile element has to travel before it can transmit a movement or a stress to the other,
The invention can comprise the following characteristics, taken singly or in combination:
w and v being respectively perturbations and measuring noise,
u(t) being the command, with u(t) being equal to the load applied to the mobile element, measured by the force sensor,
with
where K is a rigidity of the actuator, f a damping coefficient, ms a mass of the mobile element,
The invention also relates to a monitoring unit for characterizing the amplitude of play of an electromechanical actuator, the electromechanical actuator comprising an electric motor and a mobile element suitable for being set in motion by the motor, said play being defined as a distance or an angle which either of the motor and of the mobile element has to travel before it can transmit a movement or a stress to the other,
The invention also relates to an electromechanical actuator and associated monitoring unit, the monitoring unit being according to what is described previously.
Other characteristics, aims and advantages of the invention will emerge from the following description which is purely illustrative and non-limiting, and which must be considered with respect to the appended drawings, in which:
A characterization method of the amplitude ΔS of a play S(t) of an electromechanical actuator 10 will be described.
An example of electromechanical actuator is illustrated in
The actuator 10 shown is a linear displacement actuator. The actuator 10 comprises a housing, a motor 11 and a mobile piece in the form of a mobile rod 12. The motor 11 comprises a stator 14 mounted fixed relative to the housing and a rotor 13 mobile in rotation relative to the stator 14. The mobile rod 12 is mobile in translation relative to the stator 14 of the motor 11, according to a direction of displacement parallel to the axis of rotation of the rotor 13. The actuator 10 also comprises a plurality of satellite rollers 16 arranged between the rotor 13 of the motor and the mobile rod 12. Rotation of the rotor 13 relative to the stator 14 causes displacement of the rod 12 in translation relative to the stator 14. To this effect, the rotor 13 comprises an internal threaded surface and the rod 12 comprises an external threaded surface. The rollers 16 present an external threaded surface with the same thread pitch as that of the rod 12, but this pitch is reversed relative to that of the rod 12. The external threaded surface of the rod 12 is suitable for cooperating with the internal threaded surface of the rotor 13 by means of the rollers 16 to convert a rotation movement of the rotor 13 into a translation movement of the rod 12.
Document WO2010072932 describes such an actuator.
Another example of an actuator relates to rotary displacement actuators, wherein the mobile piece is mobile in translation, and the rotor of the motor drives the mobile piece in rotation relative to the stator.
The link between the frame 100 is each actuator 10, 20 is typically made by means of a fixed ball and the link between the rod 11 of the actuator 10, 20 and the aileron 200 is typically made by means of a mobile ball.
Each actuator comprises:
This information is collected by a processing unit. The processing unit comprises a calculation unit and a memory to store the information. It is also referred to a monitoring unit to designate a processing unit configured to conduct some steps of the method.
The processing unit can comprise a computer on board the aircraft, or a computer on the ground to which data have been transmitted.
The actuator 10 is modelled by computer by means of a Mod model. it is necessary to set up an equivalent dynamic model: The rotor system coupled to a screw can be considered as a transmission chain between two inertias or two masses. In the example of the actuator in
The linear position is therefore an equivalent theoretical position calculated from the angular position.
This equivalence can be executed on the mobile piece in the case of a rotary movement actuator.
Initially, the model is illustrated in
The inertia in translation is given by the following transformation:
Where Jm is the moment of inertia.
The force sensor 43 measures the load Fl, but all the applied stress comprise the load and the friction:
Fext=Fl+Ff
A model given by Karam [REF-KARAM] models the friction according to the following equation:
Ff=Fv{dot over (X)}+sign e({dot over (X)})[Fdry+η|Fl|] (1)
where X is the position in the absolute frame of reference, Ff the overall friction (N), Fv the viscous friction parameter (N/(m/s)), Fdry the dry friction and η the mechanical output.
Frictions comprise the friction of the motor and those of the screw.
Given the friction Ff as negligible before the load Fl, there is:
Fext=Fl
The characterisation method of the amplitude ΔS of the state of the actuator 10 comprises the following steps E1 to E3.
In a first step E1, the angular position of the motor θ1(t) of the motor is measured by the upstream sensor 41. As indicated previously, attributed to this angular position θ1(t) is an equivalent linear position X1(t). In this same step, the position of the mobile element X2(t) is measured by the downstream sensor 42, and the load applied Fl(t) to the mobile element is measured by the upstream sensor 41.
These data are then retrieved by the processing unit to be processed.
To perform the second step E2, the variable considered is no longer defined in the absolute frame, but in a relative frame relative to the inertia of the motor, defined as:
δX=X2−X1
The fundamental principle of the dynamic applied to the inertia of the mobile piece gives:
ms{umlaut over (δ)}X=−(K+ΔK)(X−δX0−S)−(f+Δf){dot over (δ)}X+Fext
where ΔK and Δf are respectively rigidity and damping uncertainties, δX0 is the initial deviation, which can be selected as zero.
The identification scheme of the play is based on a Kalman filter using the preceding equation. The main hypothesis consists of considering that the dynamic variable S(t) is modelled as an integral state polluted with white noise which in turn is considered an external perturbation, in the following form:
{dot over (S)}(t)=0+b(t)
where S is the play, b is random white noise of known spectral density and without bias, and t is the time.
Equating of the Kalman filter is done on the basis of a state observer, whereof the state is expressed as follows:
x=[δX{dot over (δ)}XδS]T
and the magnitude of measured output is:
y=δX
The observer is modelled according to the following equations:
w and v being respectively perturbations and a measuring noise,
u(t) being the command, with u(t) being equal to the load applied Fext=Fl to the mobile element, measured by the force sensor 43,
with
where K is a rigidity of the actuator, f a damping coefficient, ms a mass of the mobile element.
Once the model is set up and equated, the Kalman filter must be implemented.
The estimator is discretized according to a time period Ts. The measurements are therefore sampled at Ts.
Noting:
x(kTs)=x(k)
the discrete equations are given by
{dot over (x)}(k+1)=Adx(k)+Bdu(k)+Mdw(k)
y(k)=Cdx(k)+v(k)
These matrices are approximated starting out form general solutions of the continuous system given by the equation (5), and by integrating between the instants t0=kTs and t=(k+1)Ts,
where
Ad=eAT
Bd=∫0TseApBdp
Md=In
Cd=C
where In is the identity matrix.
The Kalman filter is defined by two steps: the prediction step and the updating step.
The prediction step is described by the following equations:
{circumflex over (x)}(k+1|k)=Ad{circumflex over (x)}(k|k)+Bdu(k)
P(k+1|k)=AdP(k|k)AdT+MdWdMdT
The updating step is described by the equations following:
Kf(k+1)=P(k+1|k)·CdT(CdP(k+1|k)CdT+Vd)−1
{circumflex over (x)}(k+1|k+1)={circumflex over (x)}(k+1|k)+Kf(k+1)·(y(k+1)−Cd{circumflex over (x)}(k+1|k)−Ddud(k+1))
P(k+1|k+1)=(In−Kf(k+1)Cd)P(k+1|k)
in which Wd and Vd are respectively noise covariance matrices of processing and measuring, with known spectral densities, {circumflex over (x)}(k+1|k) and {circumflex over (x)}(k+1|k+1) are the predictions and the estimations of covariance state. P(k+1| k+1) and P(k+1| k) are the prediction and covariance errors of the estimation error.
The step E2 is typically conducted by means of the processing unit.
On completion of the step E2, the dynamic variable forming the play S(t) is estimated. For each time value kTs there is therefore a value of the play S(t).
The estimated play S(t) oscillates between a maximal value and a minimal value, as illustrated in
The aerodynamic load defines a positive load which tends to displace the play towards the upper limit, while the dry friction, being added or subtracted as a function of the sign (see equation (1)) has the play oscillate around a given value.
It is clear that the fluctuations of the play S(t) are negligible relative to its values, which tends to demonstrate that the approximation according to which frictions Ff are negligible relative to the load Fl is valid.
Finally, from estimations of the play S(t), in a third step E3 the amplitude ΔS of the play S(t) is determined from a set of values estimated at the step E2.
This set preferably includes only some of the estimated values.
For example, a method for determination of amplitude can consist of taking the average value of the maximal values of the estimated play and the average value of the minimal values of the estimated play, then subtracting the two averages. This is about an amplitude of averaged “crest to crest” type.
This method is formalized as follows:
This method has the advantage of being applicable irrespective of the setting of the upstream and downstream position sensors. In fact, because of the error in offset of these sensors, the point 0 of the mechanical play is never located at the centre of the range of play.
The aim of the step E3 is to obtain a constant data relative to the play S(t), which allows characterizing the actuator. In practice, with ageing and wear on the actuator, the amplitude ΔS tends to increase. Yet on a time scale relative to executing the method, the amplitude can be considered as a constant value representative of the state of the system over this execution period.
The step E3 is typically conducted by the processing unit. The value obtained from the amplitude ΔS is then stored in the memory.
The value of the amplitude ΔS obtained by the described method can then be used in monitoring methods of the state of an actuator, as component of a signature suitable for the actuator tested and which must be classified from among a set of signatures coming from a database. A high value of the amplitude ΔS can be a sign of fatigue and/or ageing of the actuator. By coupling this datum to other data characterizing the actuator, such as rigidity coefficients of the rod, damping, or even values linked to displacements of the rod, it is possible to establish classes of values relative to actuators according to different states: good state, damaged states, defect state, for example.
The described method needs acquisition of data (see step E1). This acquisition can be made during a flight with an operational profile, that is, during instants when a strain is put on the actuator by the aircraft for its flight (pitch, roll, take-off, etc.). The method can be implemented without the need for particular provision of the aircraft, which optimizes costs and man hours. When acquisition is done in flight, the method can be implemented with the use of a single actuator, since the load is applied directly by the physical effects of flight.
Alternatively, the tests can be conducted on the ground, on the tarmac or in a hangar, during a control. In this case, the embodiment of the actuators such as presented in
In the case of an embedded processing unit, the steps E2 and E3 are conducted either in flight, or a posteriori. In the case of a processing unit on the ground, the steps E2 and E3 are conducted after retrieval of data from step E1.
Number | Date | Country | Kind |
---|---|---|---|
16 53497 | Apr 2016 | FR | national |
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/EP2017/059447 | 4/20/2017 | WO | 00 |
Publishing Document | Publishing Date | Country | Kind |
---|---|---|---|
WO2017/182593 | 10/26/2017 | WO | A |
Number | Date | Country |
---|---|---|
2 940 430 | Jun 2010 | FR |
Entry |
---|
Joonas Sainio, “Backlash compensation in electric vehicle powertrain”, Master's Thesis Espoo, Apr. 18, 2016, 121 pages. |
Adam Lagerbrg et al., “Backlash Estimation With Application to Automotive Powertrains”, IEEE Transactions on Control Systems Technology, IEEE Service Center, May 2007, pp. 483-493, vol. 15, No. 3. |
French Preliminary Search Report for FR 1653497, dated Sep. 7, 2016. |
International Search Report for PCT/EP2017/059447, dated Jul. 11, 2017. |
Number | Date | Country | |
---|---|---|---|
20190144135 A1 | May 2019 | US |