FR3074293A1 - METHOD AND DEVICE FOR DETECTING AND DIAGNOSING VIBRATIONS OF AN AIRCRAFT RELATED TO A WEAR PHENOMENON OF MECHANICAL PARTS IN A GOVERNMENT. - Google Patents

Abstract

- Method and device for detecting and diagnosing vibration of an aircraft related to a phenomenon of wear of mechanical parts in a rudder. - The detection device (1) comprises a test construction module (6) configured to define a model of the steering position according to two test hypotheses including a first test hypothesis for which there is no vibration and a second test hypothesis for which there is a vibration, a sensor (7) for acquiring a position measurement of the rudder, a decision module (10) for deciding whether the first hypothesis must be rejected or retained with the help of a statistical test, an estimation module (14) for estimating an amplitude and a duration of the vibration, a calculation module (15) for calculating a comfort and controllability indicator, a comparison module (16) for comparing the comfort and controllability indicator on a scale of criticality values for obtaining a criticality indicator and a sending module (17) for sending the criticality indicator to a user device (18).

Description

The present invention relates to the field of electrical flight control systems for aircraft. More particularly, it relates to a method and a device enabling the detection and diagnosis of vibrations linked to mechanical play in a flight control system of an aircraft.

STATE OF THE ART

An electrical flight control system for an aircraft includes the flight controls, the control surfaces of the aircraft and all of the elements between the flight controls and the control surfaces. Flight control computers perform, among other things, the position control of the aircraft control surfaces by means of actuators. Generally, an actuator comprises a rod whose linear movement is controlled. At one end, the rod is linked to a control surface, via a ball joint. The other end is linked to a piston which separates two chambers containing hydraulic fluid. The principle used to move the rod is to inject hydraulic fluid into one chamber or the other depending on the desired movement such as extension or retraction of the rod. The enslavement of this rod mainly consists in determining a difference between a setpoint (or otherwise called "command"), resulting from calculations of the control laws, and a position of the actuator rod, measured by a dedicated sensor, and multiply this difference by a gain to obtain a servo current which is sent to the actuator. A particular component of the actuator, called a servovalve on a conventional actuator, makes it possible to convert this servo current into movement of hydraulic fluid between the two chambers of the actuator and therefore to move the rod which is linked to the control surface. On an actuator of another type, such as for example an electro-hydrostatic actuator (or EHA for "electro-hydrostatic actuator" in English), the servovalve is replaced by a motor and a pump. The motor controls the pump which injects hydraulic fluid into the chambers. At the end of the actuator rod, a ball joint articulates the control rod to the control surface to allow rotation of the control surface under the effect of the retraction or elongation of the rod. A hinge line, which includes a series of one or more small ball joints, articulates the control surface to the aircraft structure to allow the rotation of the control surface around its attachment to the structure. aircraft.

However, the coating inside the ball joint (s) can wear out and create play. These games lead in flight to limit cycle oscillations (LCO or “Limit Cycle Oscillation”) which are reflected in vibrations at the global level of the aircraft. These vibrations result in discomfort for passengers and crew and fatigue in the structure. If the vibration levels are too high, the pilot of the aircraft may decide to turn around or divert the aircraft to another airport. This also complicates maintenance tasks because it is very difficult to precisely locate the faulty actuator or to isolate the faulty small ball joint (s) from a hinge line. Patent DE 10 2006 058 903 relates to a static measuring device for automating the measurement of clearances in the ball joints of actuators. It requires the installation of the measuring device at the ball joints. The solution proposed by this patent is therefore not fully satisfactory.

STATEMENT OF THE INVENTION

The object of the present invention is to overcome these drawbacks by proposing a method and a device for detecting vibrations.

To this end, the invention relates to a method for detecting and diagnosing vibrations of an aircraft linked to a phenomenon of mechanical wear of mechanical parts, the mechanical parts being part of a corresponding joint:

- either to a ball joint articulating a control rod of at least one actuator to at least one control surface,

- Or to a hinge line articulating the control surface (s) to the aircraft, the actuator (s) being configured to move the positron of the control surface (s) according to a command given to the actuator (s).

According to the invention, the method comprises the following steps for each of the actuators:

- a test construction step, implemented by a test construction module, consisting in defining a model of the position of the control surface (s) according to two test hypotheses:

o a first test hypothesis for which there is no vibration linked to mechanical wear in the joint, o a second test hypothesis for which there is a vibration linked to mechanical wear in the joint;

- an acquisition step consisting in acquiring at least one position measurement of the control surface (s) by a sensor;

- a decision step, implemented by a decision module, consisting in deciding whether the first hypothesis should be rejected or retained, using a statistical test, from the position measurement (s) of the or control surfaces and the first test hypothesis;

if the first hypothesis is rejected, the method further comprises:

- an estimation step, implemented by an estimation module, consisting in estimating an amplitude and a duration of the vibration from the model according to the two test hypotheses;

- a step of calculating an indicator, implemented by a calculation module, consisting in calculating a comfort and maneuverability indicator from the estimation of the amplitude and duration of the vibration;

- a comparison step, implemented by a comparison module, consisting in comparing the comfort and controllability indicator to a scale of criticality values to obtain a criticality indicator;

- a sending step, implemented by a sending module, consisting in sending a signal representative of the criticality indicator to a user device.

Thus, thanks to the invention, it is not necessary to install specific sensors or gauges to detect the actuator (s) for which the articulation (ball joint or hinge line) has a wear phenomenon causing one or more vibrations.

According to a particular feature, the decision stage comprises the following sub-stages:

- a first sub-step for estimating unknown parameters, implemented by a first estimation sub-module, consisting in estimating unknown parameters of the model;

a substep for calculating a decision variable, implemented by a calculating sub-module, consisting in calculating, using the statistical test, a decision variable from the position measure (s) of the control surface (s) and the two test hypotheses;

a first comparison sub-step, implemented by a first comparison sub-module, consisting in comparing the decision variable with a first predetermined threshold value, the first assumption being retained if the decision variable is less than the first threshold value, the first hypothesis being rejected and the second hypothesis being retained if the decision variable is greater than the first threshold value.

According to one embodiment, the model of the position of the control surface (s) is defined as a function of the setpoint supplied to the actuator (s), the acquisition step consisting in addition of acquiring a signal representative of the setpoint supplied to the one or more actuators.

According to another embodiment, the acquisition step further consists in acquiring at least one vertical acceleration measurement and at least one lateral acceleration measurement by a vertical acceleration and lateral acceleration sensor;

the decision stage further comprising:

a first spectral analysis sub-step, implemented by a first spectral analysis sub-module, consisting in calculating a periodogram of the lateral acceleration measurement (s) and of the vertical acceleration measurement (s) and calculating a value of the periodogram at a frequency value estimated during the first substep for estimating unknown parameters;

a second comparison sub-step, implemented by a second comparison sub-module, consisting in comparing the value of the periodogram at the estimated frequency with a second predetermined threshold value, the first hypothesis being deemed to be adopted if the value of the periodogram at the estimated frequency is less than the second threshold value, the first hypothesis being deemed rejected and the second hypothesis being deemed to be adopted if the value of the periodogram at the estimated frequency is greater than the second threshold value;

a decision sub-step, implemented by a decision sub-module, consisting in retaining or rejecting the first hypothesis, the first hypothesis being retained if the first hypothesis is retained in the first comparison sub-step and if it is deemed to be retained in the second comparison sub-step, the first hypothesis being rejected and the second hypothesis being retained if it is rejected in the first comparison sub-step or if it is deemed to be rejected in the second comparison sub-step.

Advantageously, the model according to the first hypothesis has for expression: y (ri) = λχ (η -p) + w (ri),

- the model according to the second hypothesis has for expression: y (ri) = λχ (η -p ') + l ₁ cos (2nfri) + l ₂ sin (2nfn) + w (n);

in which :

- x represents the setpoint supplied to the actuator (s),

- y represents the position of the control surface (s),

- n corresponds to an iterative value between 0 and N - 1, where N represents the number of position measurements of the control surface (s) in an Observation window,

- λ represents an attenuation or an amplification of the position measurement of the control surface (s),

- p represents a delay on the setpoint introduced by the actuator (s),

- l _t and 1 ₂ represent parameters linked to an amplitude and to a phase φ of the anomaly considered as a sinusoid, the amplitude having for expression A - phase having for expression φ arctan (-ÿ,

- f represents the frequency of the vibration to be detected,

- w representing a Gaussian noise,

- the unknown parameters (α ₀ , a _1; β ₀ and ββ) have as expressions:

oa _Q = p, oa _x - Ç), ο β ₀ = λ,

In addition, the first substep of estimating unknown parameters consists of estimating your unknown parameters by estimating the maximum likelihood with a least squares type approach;

- · the statistical test implemented by the substep for calculating a decision variable corresponds to a generalized likelihood ratio test whose decision variable has the expression: _{GLRT =} y || ² ) or

GLRT = ^ (1111 ^ (^) 711 ² “111¼¾) 7 | Γ), in which:

ο Π _Η . _(αί) = HtÇai) (h? (αί) Ηι (αίΫ) H? (aj with i = 0 or 1, ° = / x (~ P) \ ο HoOo) = I h \ x (N -1 - p ) /

0 cos (2zr /) sin (2Tr /) cos (2nf (N - 1)) s \ n (2nf (N - 1))

- the first threshold value has the expression = Q ~ 2 (P _FA ), in which:

o Q ~ 2 corresponds to the inverse function of a complementary distribution function of the law of / ² centered at r degrees of freedom, o P _FA corresponds to a probability of false alarms.

In addition, the estimated amplitude of the vibration in the estimation step a, 02 -2 for expression A = + l ₂ ,

- the estimated duration of the vibration is calculated on at least one sliding observation window and has the expression T ~ W + (W - W _s ) (XB _s (n) - 1), in which:

o VK represents a width of the sliding window (s), o 10 represents a superposition between two sliding windows, o B _s is equal to a Boolean value taking the value 1, when the decision variable is greater than the first threshold value.

Furthermore, the comfort and controllability indicator calculated in the step of calculating an indicator has the expression / (t) = k _w ^{2 2} dt, in which k _w corresponds to a wear coefficient established empirically.

According to another embodiment, the decision step comprises the following sub-steps:

- a sub-step of estimation of unknown parameters, implemented by an estimation sub-module, consisting in estimating unknown parameters of the model;

a residue calculation sub-step, implemented by a residue calculation sub-module, consisting in calculating at least one residue of the position of the control surface (s);

- a second spectral analysis sub-step, implemented by a second spectral analysis sub-module, consisting in calculating a periodogram of the residue (s) of the position of the control surface (s),

- a third substep for determining a maximum, implemented by a maximum determination sub-module, consisting in determining a frequency for which the periodogram has a maximum value,

a third comparison sub-step, implemented by a third comparison sub-module, consisting in comparing the value of the periodogram calculated at the frequency for which the periodogram has a maximum value with a third threshold value, the first assumption being retained if the value of the periodogram at the frequency for which the periodogram has a maximum value is less than the third threshold value, the first hypothesis being rejected and the second hypothesis being retained if the value of the periodogram at the frequency for which the periodogram has a maximum value is greater than the third threshold value.

A residue corresponds to a difference between a measured parameter and an estimated parameter.

According to a variant, the model according to the first hypothesis has for expression: y (n) = λ ^ χ ^ η - 1) + Â ₂ y (n -1) + ιν (η),

- the model according to the second hypothesis has for expression: y (n) = λ _± χ (η - 1) 4- λ ₂ γ (η -1) + ^ cos (2rr / n) + l ₂ 5ΐη (2π / η ) + w (n), in which and λ ₂ respectively represent an attenuation or an amplification of the setpoint supplied to the actuator (s) and an attenuation or an amplification of the position measurement of the control surface (s).

The invention also relates to a device for detecting and diagnosing vibrations of an aircraft linked to a phenomenon of mechanical wear of mechanical parts forming part of a corresponding joint:

- either to a ball joint articulating a control rod of at least one actuator to at least one control surface,

- Or to a hinge line articulating the control surface (s) to the aircraft, the actuator (s) being configured to move the position of the control surface (s) according to a command given to the actuator (s).

According to the invention, the detection device comprises for each of the actuators:

- a test construction module configured to define a model of the position of the control surface (s) according to two test hypotheses:

o a first test hypothesis for which there is no vibration linked to mechanical wear in the joint, o a second test hypothesis for which there is a vibration linked to mechanical wear in the joint;

- a sensor configured to acquire at least one position measurement of the control surface (s);

- a decision module configured to decide whether the first hypothesis should be rejected or retained, using a statistical test, from the position measurement (s) of the control surface (s) and the first test hypothesis;

the system also includes the following modules which are implemented if the first hypothesis is rejected:

- an estimation module configured to estimate an amplitude and a duration of the vibration from the model according to the two test hypotheses;

- a calculation module configured to calculate a comfort and controllability indicator from the estimate of the amplitude and duration of the vibration;

- a comparison module configured to compare the comfort and controllability indicator to a scale of criticality values to obtain a criticality indicator;

- a sending module configured to send a signal representative of the criticality indicator to a user device.

the decision module includes the following sub-modules:

- a first sub-module for estimating unknown parameters configured to estimate parameters unknown to the model;

- a calculation sub-module configured to calculate, using the statistical test, the decision variable from the position measurement (s) of the control surface (s) and the two test hypotheses;

a first comparison sub-module configured to compare the decision variable with a first predetermined threshold value, the first hypothesis being retained if the decision variable is less than the first threshold value, the first hypothesis being rejected and the second hypothesis being retained if the decision variable is greater than the first threshold value.

In addition, the detection device further comprises a vertical acceleration and lateral acceleration sensor configured to acquire at least one measurement of vertical acceleration and at least one measurement of lateral acceleration;

- the decision module also includes:

a first spectral analysis sub-module configured to calculate a periodogram of the lateral acceleration measurement (s) and of the vertical acceleration measurement (s) and to calculate a value of the periodogram at a frequency value estimated by the first unknown parameter estimation sub-module;

a second comparison sub-module configured to compare the value of the periodogram at the estimated frequency with a second predetermined threshold value, the first assumption being deemed to be adopted if the value of the periodogram at the estimated frequency is less than the second threshold value, the the first hypothesis being deemed rejected and the second hypothesis being deemed to be accepted if the value of the periodogram at the estimated frequency is greater than the second threshold value;

o a decision sub-module configured to retain or reject the first hypothesis, the first hypothesis being retained if the first hypothesis is retained by the first comparison sub-module and if it is deemed to be retained by the second comparison sub-module, the first hypothesis being rejected and the second hypothesis being retained if it is rejected by the first comparison sub-module or if it is deemed rejected by the second comparison sub-module.

According to one embodiment, the decision module comprises the following sub-modules:

- a second estimation sub-module configured to estimate parameters unknown to the model;

- a residue calculation sub-module configured to calculate at least one residue of the position of the control surface (s);

- a second spectral analysis sub-module configured to calculate a periodogram of the residue (s) of the position of the control surface (s),

- a maximum determination sub-module configured to determine a frequency for which the periodogram has a maximum value,

a third comparison sub-module, consisting in comparing the value of the periodogram calculated at the frequency for which the periodogram has a maximum value with a third threshold value, the first assumption being adopted if the value of the periodogram at the frequency for which the periodogram has a maximum value is less than the third threshold value, the first hypothesis being rejected and the second hypothesis being retained if the value of the periodogram at the frequency for which the periodogram has a maximum value is greater than the third threshold value.

The invention also relates to an aircraft, in particular a transport aircraft, comprising a device for detecting and diagnosing vibrations of an aircraft, as described above.

BRIEF DESCRIPTION OF THE FIGURES

The invention, with its characteristics and advantages, will emerge more clearly on reading the description made with reference to the appended drawings in which:

FIG. 1 represents an embodiment of the detection device,

FIG. 2 represents an embodiment of the detection method,

FIG. 3 represents an embodiment of an assembly comprising an actuator and a control surface,

- Figure 4 shows a graph representing a scale of criticality values.

DETAILED DESCRIPTION

FIG. 1 illustrates an embodiment of the device 1 for detecting vibrations of an aircraft linked to a phenomenon of mechanical wear of mechanical parts which are part of a joint.

The articulation may correspond either to a ball joint 2a articulating a control rod 3 of at least one actuator 4 to at least one control surface 5, or to a line of hinge 2b articulating the control surface (s) 5 to the aircraft. Figure 3 shows a ball joint 2a and a hinge line 2b seen in cross section.

The actuator (s) 4 are configured to move the position y of the control surface (s) 5 as a function of a set point x supplied to the actuator (s) 4 (FIG. 3). The detection device 1 is on board an aircraft.

A wear phenomenon leads to play in one or more joints 2a, 2b of one or more control surfaces 5 causing an LCO vibration.

The characteristics of the vibration to be detected are as follows.

This is a wear phenomenon that appears gradually. The older the aircraft, the more sensitive it is to play.

Once the phenomenon appears, it remains very localized in frequency. For example, on a given aircraft model and on a given control surface 5, the fluctuation for a given frequency is approximately 0.5 Hz.

As this is a wear phenomenon, its amplitude increases over time. There are generally three increasing levels in terms of acceleration felt by the pilot as a function of frequency: a first level from which the pilot detects the presence of vibrations, a second level from which the pilot reports the vibrations, for example, on a vibration report sheet and a third level from which the pilot chooses to divert the aircraft.

The approach chosen for vibration detection is based on the theory of hypothesis testing. In general, a hypothesis test has three phases.

The first phase consists in defining two hypotheses, a first hypothesis and a second hypothesis, corresponding to two different models of the functioning of a system.

The second phase consists of determining a statistical test allowing the first hypothesis to be rejected or retained based on a predetermined threshold taking account of observables.

The third phase consists in defining the predetermined threshold according to the performance desired for the statistical test.

According to one embodiment, the detection device 1 comprises, for each of the actuators 4, a test construction test module (TEST for "test construction module" 6 configured to define a model of the position y of the or control surfaces 5 according to two test hypotheses:

- a first test hypothesis for which there is no vibration linked to mechanical wear in the joint 2a, 2b,

- a second test hypothesis for which there is a vibration linked to mechanical wear in the joint 2a, 2b.

The modeling can correspond to a generic approach of the recursive identification type of a finite impulse response filter (or FIR filter for “Finite Impulse Response” in English) or an infinite impulse response filter (or HR filter for “Infinity Impulse” Response "in English).

The detection device 1 also comprises a SENSOR1 sensor 7 configured to acquire at least one measurement of the position y of the control surface (s) 5. The measurement of position y of the control surface (s) 5 corresponds to an observable.

There are generally two position measurements around the control surface: a linear displacement measurement of the control rod 3 acquired by a linear variation differential transducer or LVDT sensor 8 (LVDT for “Linear Variable Differential Transducer”) and a measurement of rotation of the control surface 5 acquired by a differential transducer with rotary variation or RVDT sensor 7 (RVDT for “Rotary Variable Differential Transducer”). Preferably, the measurement considered is that corresponding to the rotation measurement of the control surface 5 since the RVDT sensor 7 is the closest to the source of vibration to be detected. For control surfaces 5 which are not equipped with an RVDT sensor 7, it is possible to consider the measurements acquired by an LVDT sensor 8.

According to an alternative embodiment, the model of the position y of the control surface (s) 5 is defined as a function of the setpoint x supplied to the actuator (s) 4.

According to a preferred embodiment, the two test hypotheses modeled are as follows.

The model according to the first hypothesis has for expression:

y = y ° 4- seen.

The model according to the second hypothesis has for expression:

y = y ° + s + IV.

y ⁰ represents the theoretical position of the control surface 5.

y represents the measured position of the control surface 5.

w represents a measurement noise.

s represents the vibration caused by the wear phenomenon.

Depending on the expressions of y ⁰ and s, different solutions to the detection problem can be obtained. We suppose that the measurement noise wa the expression of a centered Gaussian noise of known variance. In the simpler case, we can consider the following model, discreetly:

For the first hypothesis:

y (n) - Ax (n - p) + w (n), with n = 0, ..., N - 1.

For the second hypothesis:

y (n) = A x (n - p) A- A sm (2iÎfn + <$>) + w (n), with n = 0, ..., N - 1.

Parameter A represents the attenuation or amplification of the position of the control surface 5 with respect to the setpoint x (for example, due to aerodynamic forces). The parameter p represents the delay introduced by the actuator 4. N represents the number of measurements of position y of the control surface (s) 5 in an observation window. Parameter x corresponds to the signal representative of the setpoint from the control laws sent to the actuator 4.

The vibration s caused by the phenomenon of wear can be assimilated to a sinusoid of amplitude A, of frequency f and of phase φ as in the preceding expression or, by removing the nonlinear dependence in φ, like: s (ri ) = cos & nfri) + l ₂ sin (2nfri).

The parameters l _t and 1 ₂ are linked to the amplitude A and the phase φ by the following expressions: A = - / ÏF + TF θί Φ = arctan (—p).

The test hypotheses can therefore be written:

For the first hypothesis:

y (n) ~ A x (n - p) + w (n).

For the second hypothesis:

y (n) - A x (n - p) + li cos (2nfri) + l ₂ si.n (2nfn) + w (n).

According to another embodiment, the test hypotheses can take the following expressions:

..... the model according to the first hypothesis has for expression: y (ri) = λ-ιχζτι - 1) + To ₂ y (n - 1) + w (n),

- the model according to the second hypothesis has for expression: y (n) = - 1) + λ ₂ γ (η - 1) + cos (2Tr / n) + l ₂ sin (2nfn) + w (n), in which Aj and  ₂ respectively represent an attenuation or an amplification of the setpoint x supplied to the actuator (s) 4 and an attenuation or an amplification of the position measurement y of the control surface (s) 5,

The detection device 1 also includes a DECISION decision module 10 configured to decide whether the first hypothesis should be rejected or retained, using a statistical test, from the position measurement (s) and of the control surface (s) 5 and the first test hypothesis. Preferably, the decision module 10 is configured to decide whether the first hypothesis should be rejected or not also from the setpoint x.

Advantageously, the decision module 10 comprises the following sub-modules:

an ESTIM1 estimation sub-module (ESTIM for 11) of unknown parameters configured to estimate unknown parameters at ₀ , at _1; /? _o and β _Α of the model;

a computation sub-module COMP1 (COMP for “computation module” in English) 12 configured to calculate, using the statistical test, a decision variable GLRT, GLRT from the position measurement (s) y of the or control surfaces 5 and two test hypotheses;

- a COMPAR1 comparison sub-module (COMPAR for “comparison module” in English) 13 configured to compare the decision variable GLRT, GLRT ' ¹ with a predetermined threshold value y _t , the first assumption being adopted if the decision variable GLRT , GLRT is less than the threshold value y _lt the first hypothesis being rejected and the second hypothesis being retained if the decision variable GLRT, GLRT is greater than the threshold value

Preferably, the calculation sub-module 12 is configured to calculate the decision variable GLRT, GLRT from the setpoint x also.

By writing the model according to the preferred embodiment in matrix form, we obtain:

y = Η ^ αββχ + w.

and in which a ₀ = p, tq -

x (-p) x (l - p), x (N - 1 - p)

'χ (-ρ) x (l - p), x (N - 1 - p) cos (2nf) cos (2ïïf (N - 1)) sin (2Kf) sin (2Kf (N - 1))

The unknown parameters are therefore a ₀ , cq, β ₀ and β _± for this model.

These unknown parameters a ₀ , cq, β ₀ and β <are estimated by a maximum likelihood estimate with a "least squares" approach.

Since the observables depend linearly on the parameters β ₀ and β _± , the estimates of the unknown parameters a _Ol a _lt β ₀ and β _± are as follows:

“O = argmax (y ^T Η (α0) (Η ^τ (a0W“ o)) ^ (¢ 0) 7) ”β _ΰ = (Η ^τ (α /) Η (“ S)) ” ¹ H ^T (ari ) y, “1 = argmax (y ^r H (a1) (H ^r (a1) H (a ₁ )) ¹ H ^T (a ₁ ) y), in which argmax represents the value of the variable for which the value of the function concerned reaches its maximum.

The calculation of unknown parameters can be the most time-consuming operation because it requires solving an optimization problem which is normally multidirectional. This operation can be simplified by using an approach called “alternating projection”. With this approach, it is possible to bring a multidimensional optimization problem to several monodimensionneis optimization problems, by reducing the computational cost at the price of being able to guarantee convergence towards the global maximum only under certain conditions.

The statistical test used corresponds to the generalized likelihood ratio test or GLRT test (for “generalized likelyhood ratio test” in English).

The expression for the decision variable GLRT, which is compared with the predetermined threshold value y _x , is given by the following expression, provided that the noise power is assumed to be known:

GLRT = ™ (|| y - H _o (¾) ^ 11 ² - || y - HAalMlQ, or by the following expression:

^{GLRT =} Ho ("o) ^y l | - | l ^{n ±} H ₁ (aï) y || )>

in which the matrices Π ¹ _Η are orthogonal projection matrices defined as follows:

Π ^ -πία) = / Π _Ηω = / - Η (α) (Η ⁷ ϊα) Η (α)) ¹ Η ^Γ (α).

It is important to observe that the objective of the algorithm is always to generate in real time a decision variable GLRT to compare with a threshold value γ _χ fixed a priori to detect a vibration caused by the phenomenon of mechanical wear. In addition, it is important to note that the unknown parameters α ₀ , α _{1 (} Po and Pi are useful either for constructing the GLRT test statistic or, in the event of detection, for estimating the amplitude A of the vibration.

From the expression of the decision variable GLRT, it is also possible to determine a value of the theoretical performances (in an asymptotic zone in the sense of a large signal to noise ratio), namely a probability of false alarms P _FA and a probability of detection P _D and theoretical threshold values associated with these probabilities. More precisely, for the modeling considered, we can show that:

P _FA = P (2GLRT> _Y1 | H „) = Ρ (χ ² _Γ > _Y1 ) = Q _xf ( _Y1 )

1¾ = P (2GLRT> yJHJ = P (x ² _r (k)> _Y1 ) = Q _{xf (k)} ( _Y1 ) in which Pîa corresponds to a probability of false alarms and P _D corresponds to a probability of detection, and in which Q _x z and Q _x Z (k) are the complementary distribution functions of laws of y ² centered at r degrees of freedom and of χ ² decentralized to r degrees of freedom with a parameter of decentrality k.

The degree of freedom r and the decentrality parameter k are expressed in asymptotic zones (large signal-to-noise ratio):

r = rg (4 - 4) k = -------- 5— σ ² in which:

- rg corresponds to the mathematical function “rank” (“rank” in English),

- Lf has the expression Lfin which η - Π ¹ < ^{) Η} ί ^{ · ^α ί) βί i 1 1Hî (æî)

Therefore the probability of false alarm P _FA can be calculated from the tail of the distribution of a law of% ² centered while the probability of detection PD can be calculated from the tail of distribution of a law of the type χ ² off center. The number of degrees of freedom r depends on the number of unknown parameters β0 and β _± . In our case, the number of degrees of freedom r is equal to three. The two probability laws P _FA and P _D can be easily calculated analytically or numerically.

To obtain the theoretical value of the optimal threshold value, it is sufficient to reverse the expression of the probability of false alarm ^p fa, that is:

Pi ⁼ Q _x / (Pfa).

It is also possible to demonstrate that the matrix expressions of the unknown parameters αο, α !, β ₀ and β _χ and of the GLRT test statistic, for the simple case considered, can be estimated explicitly and recursively without the use of matrices, which represents an interesting advantage from a computational cost point of view.

Thus, the estimates of the unknown parameters a ₀ , a _t , β ₀ and β _η estimated by the estimation sub-module 11, have the following expressions:

_Λ ____ i (x ^T (p) yŸ \ a ₀ = p ₀ = argmax ^^^^ J, yr _ ^ ^T (p ₀ ) yx ^T (p ^) x (p ^) '

argmax

'(x ^r (p) y) ² +2 (x ⁷ ' (p) x (p)) / y (f) -4 (x ^T (p) y) Ee [IXy (f)] - 4Im [lxy (f)] ^A 2 'x ^T (p) x (p) -2Ix (f) and

/ (* ^T (Pi) y) ~ 2Wxy (/)] \ x ^r (Pi) x (PÎ) -21 _x (f) '

2 / N (x ^T (pl) xXfi't)) c ^T (f ') (yr \ pi) xX.p ^)) 44 / NÇx' ⁱ '(p \) s (f') 'jlm [l _xy (f ')]

X ^T (fàx (pÎ) -2Ix (f) 2 / Al ^r (x ^T (^) x (^)) s ^T (/) (y-r '(^ ï) x (^ î)) - 4 / Jv (x ^r (^) c (/)) / m [rX3, (/)] i \ Y ^T C0Ï) ^ (? Ï) -2U (/) /

In the case of detection of a vibration signal, it is possible to directly use the decision variable GLRT, which is equivalent to the decision variable GLRT, obtained by using the expression Π ^{* 1} _H ( _a j = I - Π The GLRT decision variable, calculated by the calculation sub-module 10 (12), has the expression:

GLRT = m (|| Π H _{1 (S1)} y || ² - ΙΐΠ ^, νΙ | ² ) = i (x ⁷ '(^) y) ² +2 (x ^r (pî) x (pj)) / y (7) -4 (x ⁷ '(^) y) Re [/ Xy (f) j-4Im [7Xy (7)] ² _ (x ⁷ ' (p0) y) ² 2σ ² ^ x ^T (p '1) x (f1') - 21x (f ') x ^T ^ 0) x (PqJ J' in which r '(p) = x ^T (p) yx ^t (p') x (p) ^>

I _x (f) = ^ Kc ^r (f) x) ² + (s ^r (/) x) ² ],

Iy (f) = ^ [¢ ^ ¢ /) 7) ^ (^ (/) 7) ² ] and

4 //) = L (c ^T (/ ') x - Î5 ^r (/) x) (c ^T (/) y - is ^T (f') y) \.

The quantities 4 (/), Z //) correspond to the respective periodograms of the signal representative of the setpoint signal x from the control laws and of the signal representative of the position measurement y measured from the control surface 5, that is to say say to the signal power spectral density, obtained by calculating the square modulus of the Fourier transform.

A periodogram corresponds to the estimation of a spectral power of a signal. In general, the computation of the spectral power is carried out by discrete recursive Fourier transforms for one or more frequencies. The spectral power can be calculated by the Goertzel algorithm which has a low computational cost and the possibility of obtaining the spectral power by working directly on real scalar values recursively.

The decision variable GLRT ”is compared with the threshold value by the comparison sub-module 13. The comparison sub-module 13 can associate the result of the comparison with a boolean detection variable B _d . Thus, if the comparison indicates that the decision variable GLRT ”is greater than the threshold value y ₁ , the boolean decision variable takes the value 1 (or“ true ”).

The detection device 1 also comprises the following modules which are implemented if the first hypothesis is rejected:

- an ESTIM 14 estimation module configured to estimate a λ I ^λ 2 ^a 2 _Λ amplitude  = +1 ₂ and a duration T of the vibration from the model according to the two test hypotheses;

- a COMP 15 calculation module configured to calculate a comfort and controllability indicator / (t) from the estimate of the amplitude  and the duration T of the vibration;

- a COMPAR comparison module 16 configured to compare the comfort and controllability indicator / (t) with a scale of criticality values 19 to obtain a criticality indicator;

a SEND module (SEND) 17 configured to send a signal representative of the criticality indicator to a user device 18.

The fact of being able to estimate in real time the duration t and the amplitude de of the vibration therefore makes it possible to calculate, for each of the actuators 4, the comfort and controllability indicator l (t) which highlights the criticality of the vibration and its evolution in the life of the aircraft.

By way of example illustrated by FIG. 4, the comparison of the comfort and pilotability indicator 7 (t) can be made with a criticality value scale 19 comprising three steps 20, 21, 22 corresponding to three levels of criticality of the vibration. Each step 20, 21, 22 corresponds to a criticality indicator R (R for "red" or "rouge" in French), O (O for "orange), G (G for" green "or vert in French) to which i ! is possible to associate a maintenance recommendation over time. Thus, the first step 22 corresponds to a criticality indicator for which no maintenance action is necessary for the associated actuator. The second step 21 corresponds to a criticality indicator for which maintenance is necessary in the long term for the associated actuator. Finally, the third step 20 corresponds to a criticality indicator for which maintenance is necessary in the short term for the associated actuator 4.

In the example of FIG. 4, the curve representing the comfort and pilotability indicator / (t) reaches the level 20. Consequently, the criticality indicator corresponds to a criticality indicator R. Short-term maintenance term is therefore necessary for the associated actuator.

The user device 18 can translate the signal representative of the criticality by a maintenance recommendation message of the type:

"Schedule a maintenance task (to check clearances) in H hours of flight around the actuator X of the control surface S", with H corresponding to the number of hours of flight, X corresponding to an identifier of the actuator 4 and S corresponding to an identifier for the control surface 5.

Thus, according to the preferred embodiment, the detection device 1 has the following functionalities for an observation window and an actuator 4:

- Acquisition of at least one position measurement y of the control surface 5 by a sensor 7 and a signal representative of the setpoint x supplied to the actuator 4 resulting from the control laws.

- Estimation of unknown parameters at ₀ , cq. Discrete variations in the delay p and the frequency f of the vibration can be considered in the estimate.

- Calculation of the GLRT 'decision variable and generation of a Boolean detection variable B _d (which changes to the true value, or 1, when the GLRT decision variable is greater than the threshold value y _t ).

- Estimation of the amplitude A of the vibration from the estimation of the parameter β _χ

- By applying the algorithm to several sliding observation windows of width W, it is possible to estimate the duration T of the vibration. For example, from the Boolean detection variable B _d , the width of the window 14 / and the superposition between two windows W _s , the vibration time can be estimated by the following expression: T = W + ( W ~ W _s ) (£ B _d (n) - 1) = W _s + (W - W _s ) £ B _d (n).

- From the estimated amplitude  and the estimated duration T, it is possible to evaluate the criticality indicator of the phenomenon with the comfort and controllability indicator Z (t).

Since the increase in play in the joints 2a, 2b is due to a wear phenomenon, one way to assess the comfort and pilotability indicator / (t) is to calculate at the end of each flight a index as a criticality indicator and which has the expression:

kco = k _w J _T Â ² .dt, in which:

- T and  represent the estimated duration T and amplitude  of the anomaly

- k _w corresponds to a wear coefficient which makes it possible to calculate the amount of energy which is consumed by the wear processes after each flight (from the energy produced by equal vibration) to J _T Â ² , dt). The wear coefficient k _w is established empirically on the basis of the aircraft manufacturer's experience on the amplification of the play related to the wear phenomenon.

Recursively, for flight F _n we can consider that the comfort and pilotability indicator 7 (0 has the expression:

WFn) “iLCo (Fn-l) + k _w Âp _n . dt, in which the coefficient À _F represents a forgetting factor for previous flights F _n .

According to another embodiment, in the case where the unknown parameter Sj is substantially equal to p / and p, the decision module 10 comprises the following sub-modules:

- an ESTIM2 23 estimation sub-module configured to estimate unknown parameters a _Q , a _lt β ₀ and β ± of the model;

- a COMP2 residue calculating sub-module 24 configured to calculate at least one residue of the position y of the control surface (s) 5;

- a SPECTANA2 spectral analysis sub-module (SPECTANA for “spectral analysis module”) 25 configured to calculate a periodogram of the residue (s) of the position y of the control surface (s) 5,

- a MAXDET maximum determination sub-module (MAXDET) configured to determine a frequency for which the periodogram has a maximum value, ~ a COMPAR3 comparison sub-module 27, consisting of comparing the value of the periodogram calculated at the frequency for which the periodogram has a maximum value at a threshold value y ₃ , the first assumption being adopted if the value of the periodogram at the frequency for which the periodogram has a maximum value is less than the threshold value y ₃ , the first hypothesis being rejected and the second hypothesis being retained if the value of the periodogram at the frequency for which the periodogram has a maximum value is greater than the threshold value y ₃ .

The estimate of the amplitude  of the vibration can be determined using the following expression:

in which :

- N corresponds to the number of points contained in the observation window,

- P corresponds to the periodogram,

- F _c corresponds to a correction factor necessary to compensate for the phenomenon of spectral leaks (“spectral leakage” in English).

The local approach which only takes into account the position measurement y of the control surface by a sensor and / or the signal representative of the setpoint x supplied to the actuator 4 has the advantage of facilitating vibration isolation.

According to another embodiment, the use of a global approach in addition to the local approach can be envisaged to make the detection more robust. To do this, according to this embodiment, we consider the vertical acceleration measurements N _z or the lateral acceleration measurements N _y at the level of the inertial units (or 1RS for "Inertial Referenced System" in English) of the aircraft .

By analyzing in frequency the vertical and lateral accelerations N _zr N _y as a function of the control surface 5 considered, it is possible to confirm or not the presence of the vibration at the level of the aircraft. This additional analysis makes it possible to avoid confusing an LCO vibration with other types of local oscillations which may appear on the control surface 5 at frequencies close to the frequency of the LCO vibration.

In this embodiment, the detection device 1 also comprises a vertical acceleration and / or lateral acceleration sensor SENSOR2 28 configured to acquire at least one measurement of vertical acceleration N _z or at least one measurement of acceleration lateral N _y .

The vertical acceleration and / or lateral acceleration sensor 28 may include a vertical acceleration sensor and / or a lateral acceleration sensor.

The decision module 10 further comprises:

a spectral analysis sub-module SPECTANA1 29 configured to calculate a periodogram of the lateral acceleration measurement (s) N _y and of the vertical acceleration measurement (s) N _z and to calculate a value of the periodogram at a value of frequency estimated by the estimation sub-module 11 of unknown parameters;

a COMPAR2 comparison sub-module 30 configured to compare the value of the periodogram at the estimated frequency with a predetermined threshold value y ₂ , the first assumption being deemed to be adopted if the value of the periodogram at the estimated frequency is less than the threshold value y _z , the first hypothesis being deemed to be rejected and the second hypothesis being deemed to be accepted if the value of the periodogram at the estimated frequency is greater than the threshold value y ₂ ;

a DECISION1 decision sub-module 31 configured to retain or reject the first hypothesis, the first hypothesis being retained if the first hypothesis is retained by the comparisons sub-module 13 and if it is deemed to be retained by the comparison sub-module 30, the first hypothesis being rejected and the second hypothesis being retained if it is rejected by the comparison sub-module 13 or if it is deemed to be rejected by the comparison sub-module 30.

In this embodiment, the threshold value has the expression:

^{Υ2 = σ2} '° ⁸ ά ·

According to another embodiment, it is assumed that the frequency of the LCO vibration is not superimposed with the bandwidth of the control laws. In this case, it is possible to limit ourselves to analyzing in frequency the measurement of position y of the control surface 5 with possibly a confirmation according to the global approach described in the previous embodiment with the measurements of vertical acceleration JV _Z or the lateral acceleration measurements N _y at inertial units.

The frequency of the LCO vibration is estimated by determining the maximum of the periodogram of the position y measurement of the control surface 5.

According to another embodiment, to improve the quality of the estimation of the spectral power, it is possible to envisage filtering around the LCO vibration frequency. Thus, the detection device 1 can comprise a bandpass filter around the frequency of the LCO vibration before the estimation of the spectral power or of the periodogram.

According to another embodiment, if certain parameters are already known, it is possible to inject the value of the known parameter in the calculations before determining the GLRT decision variable. For example, if the frequency of the vibration is already known, it suffices to limit the calculation of the periodogram for a single point in frequency.

Without limitation, the modules and the sub-modules are integrated into a central unit or a computer.

By way of example, the modules and sub-modules of the detection device 1 can correspond to algorithms implemented in software in the central unit.

In particular, the modules can be stored in at least one memory area of the central unit.

The invention also relates to a method for detecting and diagnosing vibrations of an aircraft (FIG. 2).

The detection method includes the following steps for each of the actuators:

a step E1 of test construction, implemented by the test construction module 6, consisting in defining a model of the position y of the control surface (s) 5 according to two test hypotheses:

o a first test hypothesis for which there is no vibration linked to mechanical wear in the joint 2a, 2b, o a second test hypothesis for which there is a vibration linked to mechanical wear in the joint 2a , 2b;

- An acquisition step E2 consisting in acquiring at least one position measurement y of the control surface (s) 5 by the sensor 7;

a decision step E3, implemented by the decision module 10, consisting in deciding whether the first hypothesis should be rejected or retained, by means of a statistical test, from the position measurement (s) y of ia or control surfaces 5 and the first test hypothesis;

if the first hypothesis is rejected, the method further comprises:

- an estimation step E4, implemented by the estimation module 14, consisting in estimating an amplitude et and a duration T of the vibration from the model according to the two test hypotheses;

a step E5 of calculating an indicator, implemented by the calculation module 15, consisting in calculating a comfort and piotability indicator 7 (t) from the estimation of the amplitude  and of the duration t vibration;

a comparison step E6, implemented by the comparison module 16, consisting in comparing the comfort and piotability indicator / (t) with a scale of criticality values 19 to obtain a criticality indicator;

a sending step E7, implemented by the sending module 17, consisting in sending a signal representative of the criticality indicator to a user device 18.

Advantageously, the decision step E3 comprises the following sub-steps:

a sub-step E31 of estimation of unknown parameters, implemented by the estimation sub-module 11, consisting in estimating unknown parameters α ₀ , α _χ , βο ^and βι of the model;

a sub-step E32 of calculating a decision variable, implemented by the calculation sub-module 12, consisting in calculating, using the statistical test, the decision variable GLRT, GLRT ”from of the position measurement (s) y of the control surface (s) 5 and of the two test hypotheses;

a comparison sub-step E33, implemented by the comparison sub-module 13, consisting in comparing the decision variable GLRT, GLRT ¹ ′ with a predetermined threshold value γ _χ , the first hypothesis being retained if the decision variable GLRT , GLRT ¹ 'is less than the threshold value the first hypothesis being rejected and the second hypothesis being retained if the decision variable GLRT, GLRT ”is greater than the threshold value Yi

According to one embodiment, the acquisition step E2 consists in further acquiring a signal representative of the setpoint x supplied to the actuator (s) 4.

According to another embodiment, the acquisition step E2 also consists in acquiring at least one vertical acceleration measurement 1V _Z or at least one lateral acceleration measurement N _y by the vertical acceleration sensor and / or lateral acceleration 28.

Thus, the decision step E3 further comprises:

a spectral analysis sub-step E39, implemented by the spectral analysis sub-module 29, consisting in calculating a periodogram of the lateral acceleration measurement (s) N _y and of the acceleration measurement (s) vertical N _z and calculating a value of the periodogram at a frequency value estimated during the sub-step E31 of estimation of unknown parameters;

a comparison sub-step E310, implemented by the comparison sub-module 30, consisting in comparing the value of the periodogram at the frequency estimated at a predetermined threshold value γ ₂ , the first hypothesis being deemed to be adopted if the value of the periodogram at the estimated frequency is less than the threshold value y ₂ , the first hypothesis being deemed to be rejected and the second hypothesis being deemed to be adopted if the value of the periodogram at the estimated frequency is greater than the threshold value y ₂ ;

o a decision sub-step E311, implemented by the decision sub-module 31, consisting in retaining or rejecting the first hypothesis, the first hypothesis being retained if the first hypothesis is retained in the comparison sub-step E33 and if it is deemed to be retained in the comparison sub-step E310, the first hypothesis being rejected and the second hypothesis being retained if it is rejected in the comparison sub-step E33 or if it is deemed to be rejected in the comparison sub-step E310.

Advantageously, the substep E31 of estimating unknown parameters consists in estimating the unknown parameters a ₀ , a _it β _ΰ , & by an estimation of the maximum likelihood with a least squares type approach.

The statistical test implemented by sub-step E32 for calculating a GLRT decision variable corresponds to a generalized likelihood ratio test described above.

According to one embodiment, the decision step E3 comprises the following sub-steps:

a sub-step E34 for estimating unknown parameters, implemented by the estimation sub-module 23, consisting in estimating unknown parameters “ ₀ < ^α ι> βο θί βι model;

a sub-step E35 of residue calculation, implemented by the residue calculation sub-module 24, consisting in calculating at least one residue of the position y of the control surface (s) 5;

a spectral analysis sub-step E36, implemented by the spectral analysis submodule 25, consisting in calculating a periodogram of the residue (s) of the position y of the control surface (s) 5,

- a sub-step E37 of determining a maximum, implemented by the maximum determination sub-module 26, consisting in determining a frequency for which the periodogram has a maximum value,

a comparison sub-step E38, implemented by the comparison sub-module 27, consisting in comparing the value of the periodogram calculated at the frequency for which the periodogram has a maximum value with a threshold value y ₃ , the first hypothesis being retained if the value of the periodogram at the frequency for which the periodogram has a maximum value is less than the threshold value γ ₃ , the first hypothesis being rejected and the second hypothesis being retained if the value of the 5 periodogram at the frequency for which the periodogram has a maximum value is greater than the threshold value y ₃ .

Claims (15)

1. Method for detecting and diagnosing vibrations of an aircraft linked to a phenomenon of mechanical wear of mechanical parts, the mechanical parts being part of a corresponding joint: - either to a ball joint (2a) articulating a control rod (3) of at least one actuator (4) to at least one control surface (5), - Or to a hinge line (2b) articulating the control surface (s) (5) to the aircraft, the actuator (s) (4) being configured to move the position (y) of the control surface (s) (5) in function an instruction (x) supplied to the actuator (s) (4), characterized in that it comprises the following steps for each of the actuators (4): - a test construction step (E1), implemented by a test construction module (6), consisting in defining a model of the position (y) of the control surface (s) (5) according to two test hypotheses: o a first test hypothesis for which there is no vibration linked to mechanical wear in the joint (2a, 2b), o a second test hypothesis for which there is a vibration linked to mechanical wear in the articulation (2a, 2b); - an acquisition step (E2) consisting in acquiring at least one position measurement (y) of the control surface (s) (5) by a sensor (7); - a decision step (E3), implemented by a decision module (10), consisting in deciding whether the first hypothesis should be rejected or retained, using a statistical test, from the measurement (s) position (y) of the control surface (s) (5) and the first test hypothesis; if the first hypothesis is rejected, the method further comprises: - an estimation step (E4), implemented by an estimation module (14), consisting in estimating an amplitude (A) and a duration (T) of the vibration from the model according to the two test hypotheses ; a step (E5) of calculating an indicator, implemented by a calculation module (15), consisting in calculating an indicator of comfort and piotability (f (t)) from the estimation of the amplitude (Â) and duration (T) of the vibration; a comparison step (E6), implemented by a comparison module (16), consisting in comparing the comfort and piability indicator (7 (t)) with a scale of criticality values (19) to obtain a criticality indicator; - a sending step (E7), implemented by a sending module (17), consisting in sending a signal representative of the criticality indicator to a user device (18). 2. Method according to claim 1, characterized in that the decision step (E3) comprises the following substeps: a first sub-step (E31) for estimating unknown parameters, implemented by a first estimation sub-module (11), consisting in estimating unknown parameters (“0,“ !, β ₀ and / JJ of the model; a sub-step (E32) for calculating a decision variable, implemented by a calculation sub-module (12), consisting in calculating, using the statistical test, a decision variable (GLRT, GLRT ”) from the position measurement (s) (y) of the control surface (s) (5) and the two test hypotheses; a first comparison sub-step (E33), implemented by a first comparison sub-module (13), consisting in comparing the decision variable (GLRT, GLRT ”) with a first threshold value (yj predetermined, the first hypothesis being retained if the decision variable (GLRT, GLRT ”) is less than the first threshold value (yj), the first hypothesis being rejected and the second hypothesis being retained if the decision variable (GLRT, GLRT”) is greater at the first threshold value (^). 3. Method according to any one of claims 1 or 2, characterized in that the model of the position (y) of the control surface (s) (5) is defined as a function of the instruction (x) supplied to the actuator (s) ( 4), the acquisition step (E2) consisting in further acquiring a signal representative of the setpoint (x) supplied to the actuator (s) (4). 4. Method according to any one of claims 1 to 3, characterized in that: the acquisition step (E2) furthermore consists in acquiring at least one vertical acceleration measurement (N _z ) and at least one lateral acceleration measurement (N _y ) by a vertical acceleration sensor and lateral acceleration (28); - the decision step (E3) further includes: a first (E39) spectral analysis sub-step, implemented by a first spectral analysis sub-module (29), consisting in calculating a periodogram of the lateral acceleration measurement (s) (N _y ) and the vertical acceleration measurement (s) (N _z ) and calculating a value of the periodogram at a frequency value estimated during the first sub-step (E31) of estimation of unknown parameters; a second comparison sub-step (E310), implemented by a second comparison sub-module (30), consisting in comparing the value of the periodogram at the estimated frequency with a second predetermined threshold value (y ₂ ). the first hypothesis being deemed to be adopted if the value of the periodogram at the estimated frequency is less than the second threshold value (y ₂ ), the first hypothesis being deemed to be rejected and the second hypothesis being deemed to be adopted if the value of the periodogram at the estimated frequency is greater than the second threshold value (y ₂ ); o a decision sub-step (E311), implemented by a decision sub-module (31), consisting in retaining or rejecting the first hypothesis, the first hypothesis being retained if the first hypothesis is retained in the first sub-step ( E33) of comparisons and if it is deemed to be retained in the second substep (E310) of comparison, the first hypothesis being rejected and the second hypothesis being retained if it is rejected in the first substep (E33) of comparison or if it is deemed rejected in the second substep (E310) of comparison. 5. Method according to any one of claims 1 to 4, characterized in that: - the model according to the first hypothesis has for expression: y (n) = λχ (η - p) + w (n), - the model according to the second hypothesis has for expression: y (n) - λχ (η - p) + cos (2TT / n) + l ₂ sïn (2nfri) + w (ri); in which : - x represents the setpoint supplied to the actuator (s) (4), - y represents the position measurement of the control surface (s) (5), - n corresponds to an iterative value between 0 and N -1, where N represents the number of position measurements (y) of the control surface (s) (5) in an observation window, - λ represents an attenuation or an amplification of the position measurements (y) of the control surface (s) (5), - p represents a delay on the setpoint (x) introduced by the actuator (s) (4), - and 1 ₂ represent parameters related to an amplitude (ri) and a phase (φ) of the anomaly considered as a sinusoid, the amplitude having for expression A - ^ Ιβ + ΐββ the phase having for expression φ = ^arctan H) · - f represents the frequency of the vibration to be detected, - w representing a Gaussian noise, - the unknown parameters (cr ₀ , a _lt β ₀ and ββ) have as expressions: o "o = P, °", = £ ο β < _} = λ, ί ^λ ο βι - | 1-L \ ί ₂ 6. Method according to any one of claims 1 to 5, characterized in that: - the first sub-step (E31) of estimation of unknown parameters (cr ₀ , β ₀ , ββ) consists in estimating the unknown parameters (a _Oî a _lt β ₀ , ββ) by an estimation of the maximum likelihood with an approach least squares type; - the statistical test implemented by the sub-step (E32) for calculating a decision variable (GLRT, GLRT ”) corresponds to a generalized likelihood ratio test including ia decision variable (GLRT, GLRT”) a for expression: GLKT = ^ (lln ^ yir - lln ^ yll ² ) or GLRT ”= ^ (ΙΙΠη ^ ΗΓ - ΙΙΠΗ ₀ (ά3) yll ² ). in which : ο Π _{Η / (αί)} = Ηί (αβ) (h? (α ^ Η ^ αβή Hf (αβ) with i = 0 or 1, ° = 1 - n ^, x (-p) \ x (l - p )) x (N - 1 - p) / 'χ (-ρ) χ (1 - ρ), χ (Ν - 1 - ρ) 1 ¢ 05 (2 /) cos (2 / (ZV - 1)) sin (2 /) sin (2 / (N - 1)) - the first threshold value (/ J has the expression y _± - Q _v ï (JP _FA )> in To which: o Q ~ è corresponds to the inverse function of a complementary distribution function of law of / centered at r degrees of freedom, °? FA corresponds to a probability of false alarms. 7. Method according to any one of claims 1 to 6, characterized in that: the estimated amplitude (ri) of the vibration in the estimation step (E4) has the expression ri = + l ₂ , - the estimated duration (T) of the vibration is calculated on at least one sliding observation window and has the expression f = W + (WW _s ) ŒB _s (n) ..... 1), in which: o W represents a width of the sliding window ( _s) , o W _s represents a superposition between two sliding windows, o B _s is equal to a Boolean value taking the value 1, when the decision variable is greater than the first threshold value ( xJ. 8. Method according to any one of claims 1 to 7, characterized in that the comfort and pilotability indicator (Z (t)) calculated in step (E5) of calculation of an indicator has the expression / (t) = ri ² · dt, in which k _w corresponds to an empirically established wear coefficient. 9. Method according to any one of claims 1 to 3, characterized in that the decision step (E3) comprises the following substeps: - a sub-step (E34) of estimation of unknown parameters, implemented by an estimation sub-module (23), consisting in estimating unknown parameters (α ₀ / "ι, βο ^and βι) of the model; - a sub-step (E35) of residue calculation, implemented by a residue calculation sub-module (24), consisting in calculating at least one residue of the position (y) of the control surface (s) (5); - a second sub-step (E36) of spectral analysis, implemented by a second sub-module of spectral analysis (25), consisting in calculating a periodogram of the residue or residues of the position (y) of the one or more control surfaces (5), - a third sub-step (E37) of determining a maximum, implemented by a maximum determination sub-module (26), consisting in determining a frequency for which the periodogram has a maximum value, a third comparison sub-step (E38), implemented by a third comparison sub-module (27), consisting in comparing the value of the periodogram calculated at the frequency for which the periodogram has a maximum value with a third value threshold (y ₃ ), the first hypothesis being retained if the value of the periodogram at the frequency for which the periodogram has a maximum value is less than the third threshold value (y ₃ ), the first hypothesis being rejected and the second hypothesis being retained if the value of the periodogram at the frequency for which the periodogram has a maximum value is greater than the third threshold value (y ₃ ). 10. Method according to claim 9, characterized in that: - the model according to the first hypothesis has for expression: y (n) = Àyxfjt - 1) + À ₂ y (n - 1) + w (n), - the model according to the second hypothesis has for expression: y (n) Λ ·] Χ (η - 1) + λ ₂ γ (η - 1) + cos (2nfri) + l ₂ sm (2nfri) + w (n) , in which Aj and A ₂ respectively represent an attenuation or an amplification of the setpoint (x) supplied to the actuator (s) (4) and an attenuation or amplification of the position measurement (y) of the control surface (s) (5) . 11. Device for detecting and diagnosing vibrations of an aircraft linked to a phenomenon of mechanical wear of mechanical parts forming part of a corresponding joint: - either to a ball joint (2a) articulating a control rod (3) of at least one actuator (4) to at least one control surface (5), - Or to a hinge line (2b) articulating the control surface (s) (5) to the aircraft, the actuator (s) (4) being configured to move the position of the control surface (s) (5) as a function of instruction (x) supplied to the actuator (s) (4), characterized in that it includes, for each of the actuators (4): - a test construction module (6) configured to define a model of the position (y) of the control surface (s) (5) according to two test hypotheses: o a first test hypothesis for which there is no vibration linked to mechanical wear in the joint (2a, 2b), o a second test hypothesis for which there is a vibration linked to mechanical wear in the articulation (2a, 2b); - a sensor (7) configured to acquire at least one position measurement (y) of the control surface (s) (5); - a decision module (10) configured to decide whether the first hypothesis should be rejected or retained, using a statistical test, from the position measurement (s) (y) of the control surface (s) (5) and the first test hypothesis; the device (1) further comprises the following modules which are implemented if the first hypothesis is rejected: - an estimation module (14) configured to estimate an amplitude (4) and a duration (T) of the vibration from the model according to the two test hypotheses; - a calculation module (15) configured to calculate a comfort and controllability indicator (7 (t)) from the estimate of the amplitude (ri) and the duration (T) of the vibration; - a comparison module (16) configured to compare the comfort and controllability indicator (Z (t)) with a scale of criticality values (19) to obtain a criticality indicator; - a sending module (17) configured to send a signal representative of the criticality indicator to a user device (18). 12. Device according to claim 11, characterized in that the decision module (10) comprises the following sub-modules: - a first estimation sub-module (11) of unknown parameters configured to estimate unknown parameters (tr ₀ , α _{1 (} β ₍₎ and ββ) of the model; - a calculation sub-module (12) configured to calculate, using the statistical test, the decision variable (GLRT, GLRT) from the position measurement (s) (y) of the control surface (s) (5) and two test hypotheses; a first comparison sub-module (13) configured to compare the decision variable (GLRT, GLRT) with a first threshold value (predetermined yj, the first hypothesis being retained if the decision variable (GLRT, GLRT) is less than the first threshold value (xJ, the first hypothesis being rejected and the second hypothesis being retained if the decision variable (GLRT, GLRT) is greater than the first threshold value (yJ. 13. Device according to any one of claims 11 or 12, characterized in that: the detection device (1) further comprises a vertical acceleration and lateral acceleration sensor (28) configured to acquire at least one vertical acceleration measurement (N _z ) and at least one lateral acceleration measurement ( W _y ); - the decision module (10) further comprises: a first spectral analysis sub-module (29) configured to calculate a periodogram of the lateral acceleration measurement (s) (N _y ) and the vertical acceleration measurement (s) (N _z ) and to calculate a value from the periodogram to a frequency value estimated by the first sub-module for estimating unknown parameters (11); a second comparison sub-module (30) configured to compare the value of the periodogram at the estimated frequency with a second predetermined threshold value (γ ₂ ), the first assumption being deemed to be adopted if the value of the periodogram at the estimated frequency is lower at the second threshold value (γ ₂ ), the first hypothesis being deemed to be rejected and the second hypothesis being deemed to have been adopted if the value of the periodogram at the estimated frequency is greater than the second threshold value (Y2); a decision sub-module (31) configured to retain or reject the first hypothesis, the first hypothesis being retained if the first hypothesis is retained by the first comparisons sub-module (13) and if it is deemed to be retained by the second comparison sub-module (30), the first hypothesis being rejected and the second hypothesis being retained if it is rejected by the first comparison sub-module (13) or if it is deemed to be rejected by the second comparison sub-module ( 30). 14. Device according to any one of claims 11 or 12, characterized in that the decision module (10) comprises the following sub-modules: - a second estimation sub-module (23) configured to estimate unknown parameters (a ₀ , a _lt β ₍₎ and ββ) of the model; - a residue calculation sub-module (24) configured to calculate at least one residue of the position (y) of the control surface (s) (5); - a second spectral analysis sub-module (25) configured to calculate a periodogram of the position residue (s) (y) of the control surface (s) (5), - a maximum determination sub-module (26) configured to determine a frequency for which the periodogram has a maximum value, a third comparison sub-module (27), consisting in comparing the value of the periodogram calculated at the frequency for which the periodogram has a maximum value with a third threshold value (y ₃ ), the first assumption being retained if the value of periodogram at the frequency for which the periodogram has a maximum value is less than the third threshold value (y ₃ ), the first hypothesis being rejected and the second hypothesis being retained if the value of the periodogram at the frequency for which the periodogram has a value maximum is greater than the third threshold value (y ₃ ). 15. Aircraft, characterized in that it comprises a device for detecting and diagnosing (1) vibrations of an aircraft such as that specified under any one of claims 11 to 14.