Method and a system for diagnosing an aircraft from measurements performed on the aircraft

• US 7,949,439 B2
• Filed: 08/02/2007
• Issued: 05/24/2011
• Est. Priority Date: 09/11/2006
• Status: Expired due to Fees

• Current Assignee: Airbus Helicopters
• Original Assignee: Eurocopter
• Inventors: Tran, Bang
• Primary Examiner(s): NGUYEN, TAN QUANG
• Application Number:
  US11/832,764
• Publication Number:
  US 20080065281A1
• Time in Patent Office: 1,391 Days
• Field of Search: 701/3, 701/29, 701/120, 702/35, 702/182, 706/16, 706/52, 244/1.N, 455/450
• US Class Current: 701/3
• CPC Class Codes: G01H 1/00 Measuring characteristics of vibrations in solids by using direct conduction to the detector G01H9/00, G01H11/00 take precedence

Abstract

A method of diagnosing an aircraft by analyzing at least one spectrum obtained from measurements taken on the aircraft, in particular from vibration measurements, in which the spectrum (spectra) to be analyzed is/are input to a calculation module (23) together with at least one first reference spectrum, the calculation module determining a diagnosis value (Diag) as a function of the spectrum (spectra) to be analyzed, of the reference spectrum (spectra), and of parameters determined from at least two reference spectra including the first reference spectrum.

Claims

1. A method of diagnosing an aircraft by analyzing at least one spectrum (V{right arrow over (te)}st, V{right arrow over (te)}st1, V{right arrow over (te)}st2) obtained from measurements made on the aircraft, in particular from vibration measurements, the method comprising the following successive operations:

synchronizing a plurality of measurement time data series;

calculating a mean series of measurement time data from a plurality of synchronized measurement time data series;

transforming from the time domain to the frequency domain in order to obtain the measurement frequency data vector from the mean series of time data;

imputing the frequency data vector(s) corresponding to the spectrum (spectra) for analysis to a calculation module (23) together with at least one first reference spectrum data vector ({right arrow over (VR)}1, {right arrow over (VR)}2, {right arrow over (VR)}3, {right arrow over (VR)}4), the calculation module determining, for each input vector, a plurality of degree-of-membership vectors for the components of the input vector, specifying their respective degrees of membership to a plurality of classes, the calculation module determining a diagnosis value (Diag) as a function of the degree-of-membership vectors, and as a function of the spectrum (spectra) to be analyzed, of the reference spectrum (spectra), and of parameters (a_m, b_m, c_m, α

_i,j, β

_i,j, δ

_i,j) determined from at least two reference spectra including said first reference spectrum.

2. A method according to claim 1, in which:

in a preparatory stage, at least two reference spectra are input to the parameterized calculation module (23) which outputs a diagnosis value, and then the parameters of the calculation module are modified by iterations in order to obtain a reference value (Valref) at the output; and

then

in a diagnosis stage, one or more spectra can be analyzed together with some of the reference spectra are input to the parameterized calculation module which outputs the diagnosis value, and a fault is diagnosed when said diagnosis value exceeds the reference value.

3. A method according to claim 1, in which a calculation module is used having a plurality of operators structured in layers and in a network, including parametric membership operators (A_i, B_j), and parametric weighting operators (P_i,j) applied to the spectra presented at its inputs.

4. A method according to claim 1, in which, in the diagnosis stage, at least one spectrum data, or frequency data vector for diagnosis is applied together with at least one reference spectrum data vector to the inputs of the calculation module which determines, for each input vector, a plurality of degree-of-membership vectors for the components of the input vector, specifying their respective degrees of membership to a plurality of classes, and calculates a (single) diagnosis value as a function of the degree-of-membership vectors, and in which a fault of the aircraft is indicated when the diagnosis value exceeds a threshold or reference value.

5. A program comprising code applied to a medium, such as a memory, or embodied as a signal, the code being readable and/or executable by at least one data processor unit, to diagnose the (good or bad) operation of an aircraft, the code having code segments for performing the operations of a method according to claim 1.

6. A system for diagnosing an aircraft, the system comprising:

a read member for reading a data medium and arranged to read data relating to measurements made on the aircraft;

a database containing reference spectrum data for the aircraft;

transformation apparatus for transforming the measurement data from the time domain to the frequency domain, which apparatus is connected to the read member to receive the measurement data and to output spectrum data to be analyzed; and

a calculation member connected to the database and to the transformation apparatus, and programmed to perform the operations of a method according to claim 1.

7. A method according to claim 1, in which the calculation module (23) has a layer structure including a layer (CO1) of membership operators, and in which a vector is applied to the input of the membership operators of the module (23) in order to obtain at the output of each membership operator a degree-of-membership vector that presents the same number of dimensions as the input vector.

8. A method according to claim 1, in which each of the spectrum data vectors is the result of applying a Fourier transform to measurement data obtained by sampling signals delivered by microphones or accelerometers at a frequency of not less than about 1 kHz (in particular a frequency of about 5 kHz, 10 kHz, 50 kHz, or 250 kHz).

9. A method according to claim 1, in which each of the spectrum data vectors comprises about one to several thousand components.

10. A method according to claim 9, in which the calculation module (23) includes a layer (CO2) of multiplication operators (□

), and in which each pair of degree-of-membership vectors is applied to the inputs of the multiplication operators of the module (23) in order to obtain at the output from each multiplication operator a value for the product of the corresponding pair of vectors, said product representing the degree of activation of a rule associated with the pair of membership classes corresponding to the pair of vectors.

11. A method according to claim 10, in which the calculation module (23) includes a layer (CO3) of normalization operators (N), and in which each input vector product value is applied to the input of the normalization operators of the module (23) in order to obtain at the output from each normalization operator a normalized value for the product of the corresponding pair of vectors.

12. A method according to claim 11, in which the calculation module (23) includes a layer (CO4) of weighting operators (P_i,j) for the input vectors, and in which the normalized vector product values and the input vector are applied as inputs to the weighting operators to obtain at the output from each weighting operator a weighted vector.

13. A method according to claim 12, in which the calculation module (23) includes a layer (CO5) comprising a summing operator (□

) for summing the weighted vectors, and in which each weighted vector is input to the summing operator to obtain at the output from the summing operator a overall vector, and the diagnosis value is calculated as a function of (is equal to) the norm of the overall vector.

14. A method according to claim 9, in which the membership operators of the first layer (CO1) present continuous and differentiable membership curves (CC1-CC5) that are generally bell-shaped or Gaussian.