Acquisition module for a system for monitoring a rotating machine, monitoring system and method

Abstract

The invention relates to an acquisition module for a monitoring system of a rotating machine, in particular of an aircraft engine, the acquisition module comprising at least one measurement sensor for measuring an analogue signal x(t) of a physical quantity of a member of the rotating machine, at least one sample-and-hold device configured to collect a sample of the analogue signal x(t) at sampling times t n and to maintain it constant between two sampling times t n, and at least one memory for storing samples, the sample-and-hold device being configured to collect a sample of the analogue signal x(t) at sampling times t n defined in a random manner, the sampling times t n being defined according to the following law t n=nΔ t+τ n, the samples and the sampling times t n being stored in the memory.

Claims

1. An acquisition module for a system for monitoring a rotating machine, the acquisition module comprising: at least one sensor for measuring an analog measurement signal x(t) of a physical quantity of a part of the rotating machine, at least one sample and hold device configured to collect a sample of the analog signal x(t) at sampling times t.sub.n and to maintain the sample constant between two sampling times t.sub.n, at least one memory for storing samples wherein the sample and hold device is configured to collect a sample of the analog signal x(t) at sampling times t.sub.n defined in a random manner, the sampling times t.sub.n being defined according to the following law:
t.sub.n=nΔ.sub.t+τ.sub.n in which, n is a natural integer, Δ.sub.t=1/f.sub.e where f.sub.e is the sampling frequency of the sample and hold device, τ.sub.n is a random variable having a uniform probability law over $\left[-\frac{{\Delta }_t}{2}\text{;}+\frac{{\Delta }_t}{2}\right],$ wherein the samples and the sampling times t.sub.n being stored in the at least one memory.

2. The acquisition module according to claim 1, comprising at least one analog bandpass filter having a cut-off frequency f.sub.cut, the analog bandpass filter being positioned between the measurement sensor and the sample and hold device.

3. The acquisition module according to claim 1, wherein the sampling frequency f.sub.e is below 20 kHz.

4. The acquisition module according to claim 1, comprising an analog-digital converter configured to transform the samples of the sample and hold device into digital samples x[nΔ.sub.t].

5. An aircraft comprising an aircraft engine and the acquisition module according to claim 1 of which the measurement sensor is positioned on a part of the aircraft engine.

6. A method for monitoring a rotating machine by use of the acquisition module according to claim 1, the method comprising: a step of measuring an analog measurement signal x(t) of a physical quantity of a part of the rotating machine, a step of sampling the analog signal x(t) at sampling times t.sub.n defined according to the following law:
t.sub.n=nΔ.sub.t+τ.sub.n in which, n is a natural integer, Δ.sub.t=1/f.sub.e where f.sub.e is the sampling frequency of the sample and hold device τ.sub.n is a random variable having a uniform probability law over $\left[-\frac{{\Delta }_t}{2}\text{;}+\frac{{\Delta }_t}{2}\right],$ a step of recording the samples and the sampling times t.sub.n.

7. The acquisition module according to claim 1, comprising a plurality of analog bandpass filters each having a cut-off frequency, the analog bandpass filters being positioned between the measurement sensor and the sample and hold device.

8. The acquisition module according to claim 7, wherein the analog bandpass filters are mounted in parallel with each other.

9. A monitoring system comprising the acquisition module according to claim 1 and at least one processing module configured to compute at least one health estimator from the samples and sampling times t.sub.n obtained from said acquisition module.

10. The monitoring system according to claim 9, wherein the processing module comprises a device for transforming the samples according to a Dirichlet transform.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) The invention will be better understood on reading the description that follows, given uniquely as an example, and by referring to the appended drawings among which:

(2) FIG. 1 is a schematic representation of an acquisition module according to the prior art,

(3) FIG. 2 is a schematic representation of a system for monitoring an aircraft engine according to the invention,

(4) FIG. 3 is a schematic representation of an acquisition module according to a first embodiment of the invention,

(5) FIG. 4 is a schematic representation of uniform sampling according to the prior art,

(6) FIG. 5 is a schematic representation of random sampling according to the invention and

(7) FIG. 6 is a schematic representation of an acquisition module according to a second embodiment of the invention.

(8) It should be noted that the figures set out the invention in a detailed manner for implementing the invention, said figures obviously being able to serve to better define the invention if need be.

DETAILED DESCRIPTION

(9) A monitoring system according to an embodiment of the invention for monitoring an engine of an aircraft such as an airplane or a helicopter will henceforth be described. However, it goes without saying that the invention applies to any rotating machine, in particular, a wind turbine, a terrestrial vehicle engine, etc.

(10) With reference to FIG. 2, an aircraft P in flight and a control center CC on the ground for monitoring the aircraft P are represented. The aircraft P comprises several engines of which at least one engine is monitored continuously in order to track its state of health. To this end, the monitoring system comprises an acquisition module 10 connected to one or more engines of the aircraft and a processing module 20 situated on the ground in the control center CC. In this example, the acquisition module 10 can communicate with the processing module 20 via a satellite link S.

(11) A first embodiment of an acquisition module 10 is shown in FIG. 3. The acquisition module 10 comprises a sensor for measuring 11 an analog signal representative of a physical quantity of a part of the aircraft engine.

(12) The acquisition module 10 comprises a measurement sensor 11 configured to measure a physical quantity, for example an acceleration, and to convert it into an analog signal x(t), notably, a voltage signal. The measurement sensor 11 is placed on a part of the aircraft engine, for example, in the vicinity of a fan, a bearing, etc. Such a measurement sensor 11 is known to those skilled in the art and may have various forms as a function of the physical quantity to measure.

(13) Moreover, the acquisition module 10 comprises a sample and hold device 13, configured to collect a sample of the analog signal x(t) at sampling times t.sub.n and to maintain it constant between two sampling times t.sub.n. The sample and hold device 13 has a sampling frequency f.sub.e which is chosen to respect the Shannon theorem, that is to say, be at least equal to double the maximum frequency contained in the analog signal x(t) to sample.

(14) According to the invention, the sample and hold device 13 is configured to collect a sample of the analog voltage signal x(t) at sampling times t.sub.n defined according to the following law
t.sub.n=nΔ.sub.t+τ.sub.n in which n is a natural integer Δ.sub.t=1/f.sub.e where f.sub.e is the sampling frequency of the sample and hold device (13) τ.sub.n is a random variable having a uniform probability law over

(15) $\left[-\frac{{\Delta }_t}{2}\text{;}+\frac{{\Delta }_t}{2}\right].$

(16) In this example, with reference to FIG. 3, the acquisition module comprises a timer 14 to determine the sampling times t.sub.n as indicated above.

(17) Thus, thanks to the invention, for monitoring an aircraft engine, the sampling times t.sub.n are determined in a random manner and no longer in a uniform manner as in the prior art. In other words, unlike the prior art in which the time between two sampling times t.sub.n was fixed (FIG. 4), the time between two sampling times t.sub.n is henceforth variable (FIG. 5).

(18) The use of a uniform probability law over

(19) $\left[-\frac{{\Delta }_t}{2}\text{;}+\frac{{\Delta }_t}{2}\right]$
for the random variable τ.sub.n is advantageous because it allows optimal spectral statistical properties, the quadratic estimation error being minimal. Furthermore, it makes it possible to obtain unbiased estimators and a minimum variance of the sampling noise. After having performed theoretical computations, it appears that the optimal statistical law for sampling a signal with a number N of samples acquired randomly over an acquisition time T is the uniform statistical law centered on nT/M=nΔ.sub.t and of width T/M=Δ.sub.t.

(20) Similarly, theoretical computations have shown that the frequency resolution within the context of random sampling is equal to the inverse of the acquisition time T. As a reminder, the acquisition time T is equal to the frequency resolution for a uniform sampling. This property is particularly advantageous for analyzing mechanical signals of a rotating machine because the frequency resolution must be chosen smaller than the spacing between 2 frequency harmonics in order to be able to identify it.

(21) Unlike the prior art, a sampling frequency f.sub.e lower than in the prior art may be chosen throughout the acquisition time T without limiting the pass band of the spectrum of the signal, that is to say, without reducing the Nyquist frequency. Such a sampling frequency f.sub.e is advantageous given that it makes it possible, on the one hand, to limit the number of samples and thus the number of data to transmit to the processing module 20 and, on the other hand, to obtain an optimal frequency resolution, which is advantageous during processing of the samples as will be described hereafter to obtain relevant spectral estimators.

(22) To enable a qualitative estimation, the number of samples must nevertheless remain sufficient in order to obtain a variance of the sampling noise of low value, the variance of the sampling noise being inversely proportional to the numbers of samples. Also, the number of samples may be reduced but within the limit of the desired precision.

(23) Unlike the prior art, the sampling times t.sub.n are required by the processing module 20 in order to compute the estimators and must thus be saved as will be described hereafter. The conservation of the sampling times t.sub.n is not penalizing given that the number of samples is considerably reduced compared to the prior art.

(24) In this example, the acquisition module 10 further comprises an analog bandpass filter having a cut-off frequency f.sub.cut. The analog bandpass filter 12 is positioned between the measurement sensor 11 and the sample and hold device 13 as illustrated in FIG. 3.

(25) Although having a structure analogous to an anti-aliasing filter according to the prior art, the analog bandpass filter 12 according to the invention makes it possible to reject components of the analog signal x(t) above the cut-off frequency f.sub.cut in order to minimize the variance of the spectral estimators computed by the processing module 20. The cut-off frequency f.sub.cut thus defines the useful frequency range to explore. Also, the cut-off frequency f.sub.cut is chosen to be above the maximum value of the desired spectral range. In a preferred manner, the ratio f.sub.cut/f.sub.e is comprised between 3 and 6.

(26) Random sampling has the advantage of not being restricted by the Nyquist limit like uniform sampling. However random sampling has an incidence on the random noise present in the spectrum. Theoretical computations have shown that the variance of the random noise is proportional to the energy of the signal measured. For this reason, it is advantageous to filter, by an analog bandpass filter 12, undesirable frequencies which are above the cut-off frequency f.sub.cut.

(27) The acquisition module 10 also comprises an analog-digital converter 15 configured to transform the samples maintained constant by the sample and hold device into digital samples x[nΔ.sub.t] quantified on a limited number of bits which depends on the desired precision. Such an analog-digital converter 15 is analogous to the prior art and known to those skilled in the art.

(28) As illustrated in FIG. 3, the acquisition module 10 further comprises a storage memory 16 for storing the digital samples x[nΔ.sub.t] as well as the sampling times t.sub.n.

(29) In this example, the acquisition module 10 further comprises a communication module 17, for example by satellite means, so as to transmit data from the memory 16 to the processing module 20 situated on the ground in the control center CC.

(30) Unlike the prior art, the sampling times t.sub.n must be stored in the memory 16 in order to allow the processing module 20 to compute estimators of the state of health of the aircraft engine. However, since the number of digital samples x[nΔ.sub.t] is greatly reduced compared to the prior art, the amount of data having to be transmitted between the acquisition module 10 and the processing module 20 is smaller than in the prior art. This advantageously decreases the transmission cost and reduces the dimensioning of the memory 16.

(31) Thanks to the invention, an acquisition module 10 is obtained which makes it possible to collect a restricted number of digital samples x[nΔ.sub.t] and sampling times t.sub.n while enabling the computation of relevant estimators by the processing module 20 as will now be described.

(32) The processing module 20 comprises a computation unit, in particular a microprocessor, for carrying out mathematical operations from the digital samples x[nΔ.sub.t] obtained randomly as well as the sampling times t.sub.n.

(33) In a preferred manner, the processing module 20 implements a Dirichlet transform defined in the following manner:

(34) $X\left(f\right)={\mathrm{DIR}}_{n.fwdarw.m}^N\left\{x\left[n{\Delta }_t\right]\right\}=\frac{1}{N}\underset{n=1}{\overset{N}{.Math.}}x\left[t_n\right]e^{-j2{\pi \mathrm{ft}}_n}.$

(35) Where f designates the continuous frequency variable theoretically varying up to infinity and N designates the number of samples.

(36) In order to be able to apply the Dirichlet transform in an optimal manner, knowledge of the frequency resolution Δ.sub.f is desired. As a reminder, the frequency resolution Δ.sub.f is equal to the inverse of the acquisition time T: Δ.sub.f=1/T.

(37) In practice, it is not useful to compute frequencies above the cut-off frequency f.sub.cut of the analog bandpass filter 12 given that the Fourier coefficients are practically zero. Also, in order to reduce the computation cost, m must take the following values: m=0,1, . . . M−1 with M=f.sub.cut/Δ.sub.f. Unlike the DFT, the maximal frequency index may exceed the number of samples: M>N, random sampling not being limited by the Shannon theorem.

(38) An example of computation of estimators is described hereafter.

(39) Let x=[x[t.sub.1], x[t.sub.2] . . . x[t.sub.N]].sup.T (.sup.T designates the transpose operator) a size vector N sampled randomly at the times t.sub.1 . . . t.sub.N et X=[X[Δf], X[2.Math.Δf] . . . X[M.Math.Δf]].sup.T its spectrum estimated by the Dirichlet transform, the relationship between x and X is expressed in matrix form as:
X=Φ.Math.x
where Φ designates the matrix of size (M, N) of the Dirichlet transform such that:
Φ((m,n)=e.sup.j2π(m-1)Δft.sup.n,1<m≤M et 1<n≤N

(40) Similarly, the square envelope spectrum (SES) could also be expressed as:
P=Φ.Math.x.sup.2
where x.sup.2=[x.sup.2[t.sub.1],x.sup.2[t.sub.2] . . . x.sup.2[t.sub.N]].sup.T.

(41) Such estimators are advantageous because they make it possible to reveal in a relevant and reliable manner defects of an aircraft engine and, more generally, a rotating machine.

(42) An exemplary embodiment of a method for monitoring an aircraft engine will henceforth be described by means of the monitoring system described previously. With reference to FIG. 2, the monitoring system comprises an acquisition module 10 connected to one or more engines of the aircraft P and a processing module 20 situated on the ground in the control center CC. In this example, the acquisition module 10 communicates with the processing module 20 by a satellite link S.

(43) The monitoring method comprises a step of measuring, by the measurement sensor 11 of the acquisition module 10, an analog signal x(t) representative of a physical quantity of a part of the rotating machine, for example, an acceleration.

(44) Then, the method comprises a step of filtering, by the bandpass filter 12, frequencies above the cut-off frequency f.sub.cut of said bandpass filter 12. Next, the method comprises a step of sampling the analog signal x(t) at sampling times t.sub.n defined in a random manner according to the law defined previously. In this example, the sampling frequency f.sub.e is below 20 kHz, preferably, comprised between 100 Hz and 1000 Hz, further preferably, comprised between 300 Hz and 500 Hz so as to limit the number of samples.

(45) Then, the monitoring method comprises a step of digital conversion of the samples by the analog-digital converter 15 and a step of recording the digital samples x[nΔ.sub.t] with the sampling times t.sub.n in the memory 16.

(46) Advantageously, the amount of data recorded in the memory 16 is limited compared to the prior art, which enables a less expensive transmission by the communication module 17.

(47) The data of the memory 16, that is to say the digital samples x[nΔ.sub.t] as well as the sampling times t.sub.n are received by the processing module 20 which can thus compute relevant estimators mathematically. Advantageously, the processing module 20 implements a Dirichlet transform from the digital samples x[nΔ.sub.t] and the sampling times t.sub.n in order to form relevant indicators, notably, the square envelope spectrum (SES).

(48) Thus, thanks to the invention, relevant estimators may be obtained with a lesser amount of data while conserving relevant data characteristic of rotating machines and, in particular, aircraft engines.

(49) A second embodiment of an acquisition module 10 is represented with reference to FIG. 6. For reasons of brevity and clarity, only structural and functional differences with the first embodiment of FIG. 3 will be described.

(50) In this embodiment, as illustrated in FIG. 6, the bandpass filter 12 is replaced by a plurality of analog bandpass filters 18.sub.1, 18.sub.k which are mounted in parallel with each other.

(51) The bandpass filters 18.sub.1, 18.sub.K make it possible to cut the frequency range into a plurality of elementary frequency ranges which are each sampled in an independent manner by the sample and hold device 13. Each elementary frequency range is defined by its central frequency (f.sub.c.sup.K) and its pass band (B.sub.p) as illustrated in FIG. 6. The use of bandpass filters 18.sub.1, 18.sub.K is interesting in the case of a planetary or straight gearing in which the signature of defects appears around the tooth-meshing frequency, which is known. In this case, an analog bandpass filter may be applied with a central frequency f.sub.c equal to the tooth-meshing frequency f.sub.eng and a band width B.sub.p equal to f.sub.eng/2.

(52) Advantageously, it is possible to parameterize each elementary frequency range to highlight in a relevant manner a potential defect of the part of the aircraft engine over the elementary frequency range. An elementary frequency range makes it possible to increase the signal to noise ratio, which facilitates the detection of a potential malfunction.