Monitoring of an aircraft engine to anticipate the maintenance operations

1. A method for monitoring an aircraft engine, including the following steps of:

acquiring a time signal of the exhaust gas temperature residual margin of said aircraft engine ( 2 ),

smoothing said time signal to form a first curve (C 1 ) representing said temperature residual margin,

identifying decreasing pieces in said first curve,

constructing a second curve (C 2 ) by concatenation of said decreasing pieces, said second curve being continuous while being restricted to said decreasing pieces of said first curve, and

constructing a prediction model from said second curve to determine at least one failure forecast indicator (I 1 , I 2 ).

2. The method according to claim 1 , wherein identifying the decreasing pieces in said first curve includes the following steps of:

applying to the first curve (C 1 ) a statistical increase model decomposable into two independent parts formed of a first decreasing function representing a usual wear of the aircraft engine and of a second step function formed of randomly triggered hops representing ad hoc servicing operations on the aircraft engine,

looking for ascents corresponding to said hops, and

identifying the decreasing pieces by deleting the points from said ascents on the first curve.

3. The method according to claim 1 , wherein constructing by concatenation said second curve (C 2 ) includes a bonding of said decreasing pieces by displacing each previous piece, to make it join the following piece starting by the last piece and going back in time step by step.

4. The method according to claim 1 , wherein constructing said prediction model includes the following steps of:

constructing an autoregressive model modeling the evolution of the temperature residual margin using the record of said second curve, and

applying a dynamic filter to said autoregressive model to determine said at least one failure forecast indicator.

5. The method according to claim 4 , wherein the dynamic filter is selected among the set of the following particulate filters: a Bayesian filter, a Kalman filter, extended Kalman filters.

6. The method according to claim 1 , wherein constructing said prediction model includes the following steps:

constructing a linear model for the evolution of the temperature residual margin using the record of said second curve, and

applying a regression technique to said linear model to determine said at least one failure forecast indicator.

7. The method according to claim 1 , wherein said at least one failure forecast indicator is selected among a set of indicators comprising:

a first indicator (I 1 ) for estimating the probability of exceeding a failure threshold before a predetermined time horizon, and

a second indicator (I 2 ) for estimating the date of exceeding a failure threshold.

8. The method according to claim 1 , wherein acquiring said time signal of the temperature residual margin includes the following steps of :

acquiring over time measurements of the exhaust gas temperature of the aircraft engine,

normalizing said temperature measurements relative to a standard reference temperature thus forming normalized temperature measurements,

standardizing said normalized temperature measurements by taking into account context data thus forming standardized temperature measurements, and

computing the margins between said standardized temperature measurements and a predetermined maximum temperature value (as a function of the engine) to form said margin time signal.

9. A system for monitoring an aircraft engine, including:

acquisition and processing means ( 11 ) configured to collect a time signal of the exhaust gas temperature residual margin of said aircraft engine ( 2 ),

acquisition and processing means ( 11 ) configured to smooth said time signal thus forming a first curve (C 1 ) representing said temperature residual margin,

acquisition and processing means ( 11 ) configured to identify decreasing pieces in said first curve,

acquisition and processing means ( 11 ) configured to construct a second curve (C 2 ) by concatenation of said decreasing pieces, said second curve being continuous while being restricted to said decreasing pieces of said first curve,

acquisition and processing means ( 11 ) configured to construct a prediction model from said second curve to determine at least one failure forecast indicator (I 1 , I 2 ).

10. A non-transitory computer-readable medium having instructions stored thereon that, when executed by a computer, cause the computer to perform the monitoring method according to claim 1 .