METHOD AND SYSTEM FOR CONTROLLING THE PERFORMANCE OF HIGH-FREQUENCY, LOW-VARIABILITY HELICOPTER ENGINES

Abstract

The invention relates to a method for monitoring the health of a helicopter turbomachine, comprising: - a step (S45) of comparing flight data (T0, P0, N1, N2, T45, TQ, VCAS), this data being measured in flight and comprising data (N1, N2, T45, TQ, VCAS) for at least one turbomachine parameter, with data from a stabilized model of the turbomachine; - and a step (S48) of selecting, from said data for at least one parameter (N1, N2, T45, TQ) of the turbomachine (110), the data for which the difference, or residual, with the data from the stabilized model is less than or equal to a predefined value. Figure 4 (for the abstract)

Description

Title of the invention: METHOD AND SYSTEM FOR CONTROLLING THE PERFORMANCE OF HIGH-FREQUENCY, LOW-VARIABILITY HELICOPTER ENGINES

[0001] TECHNICAL FIELD AND PRIOR TECHNOLOGY

[0002] The invention relates to the field of data analysis, applied to monitoring the health of helicopter engines (or turbomachinery).

[0003] The performance parameters of a helicopter turboshaft engine are defined by the engine manufacturer to achieve specific power and fuel consumption at various speeds and within the flight envelope. These performance values, or margins, are referred to as "uninstalled" or "bare engine" and do not include intake and exhaust losses. They constitute the minimum reference performance guaranteed by the manufacturer. More precisely, for a given variable, the bare engine margin is defined as the difference between the reference value for that variable (reference variable) and the measured value.

[0004] However, once the engine is installed and running, it is affected by external "installation effects." These effects, which generally result in losses, reduce the turboshaft engine's performance level. These losses originate from a variety of mechanisms at the interface between the helicopter and the engine.

[0005] The Engine Power Check (EPC), which characterizes the performance margins of the installed engine, allows the helicopter pilot to verify the availability of the minimum guaranteed power of an engine in service once installed. These measurements represent the engine's performance once installed on the helicopter.

[0006] Currently, performance evaluations of engines installed on helicopters are based on operational flight procedures and on-board calculations carried out during stable phases, which can be rare and / or complex for a pilot to achieve.

[0007] In addition, these assessments may be subject to variability induced by the environment and the operation: effect of outside temperature, altitude, humidity, speed of advancement of the aircraft, etc.

[0008] Currently, two types of measurement are known and used.

[0009] First of all, there are the classic EPC measures.

[0010] They have the advantage of conforming to the flight manual or reference manual for engine performance.

[0011] They have the disadvantage of being carried out at a low frequency and of requiring the implementation, by the user, of an operational procedure which can sometimes be cumbersome and costly.

[0012] Methods based on mathematical conditioning of downloaded flight data are also known. Such a method is known, for example, from document FR3003032A1.

[0013] The advantage of this type of method is that it allows for a high frequency of measurements and calculations and is close to the calculation method in the flight manual. Furthermore, the high frequency of measurements makes it possible to improve the reliability of performance trend analysis.

[0014] But this type of method has the disadvantage of being dependent on phases considered to be stable (initially stable flight conditions and engine in stable thermomechanical conditions); however, such phases are not necessarily present for each mission or may be complex to achieve, which can be highly restrictive.

[0015] The problem therefore arises of the impossibility of performing an accurate and frequent evaluation of the performance of a helicopter engine in the absence of sufficiently long stable flight phases. Existing solutions rely on initially stable flight conditions, which can limit the frequency and accuracy of these evaluations and can also lead to a dispersion of measurements. Description of the invention

[0016] The invention aims to define a method and / or a device that overcomes the difficulties encountered with known methods of estimating or calculating engine performance, particularly for helicopter engines.

[0017] The invention relates in particular to a method, and / or respectively a system, for processing data measured in flight of a helicopter, referred to as flight data, comprising data of at least one parameter of a turbomachine of the helicopter, referred to as measured data of the turbomachine, for the purpose of monitoring the health of this turbomachine, this method, and / or this system comprising at least:

[0018] - a step, and / or respectively means (or programmed means or specially programmed), comparison of said data of said at least one parameter of the turbomachine with data from a stabilized model of the turbomachine;

[0019] - and a step, and / or respectively means (or programmed means or specially programmed), selection, from said data of said at least one parameter of the turbomachine, of the data for which the deviation, or residual, with the data of the stabilized model is less than or equal to a predefined value.

[0020] A method according to the invention can be implemented by computer. A system according to the invention may include a computer.

[0021] The data of said at least one parameter of the turbomachine may include data from at least one temperature probe, for example a probe which measures a temperature at the outlet of a high-pressure turbine of the turbomachine, and / or a torque sensor of the engine and / or a speed sensor of the shaft of the turbomachine which supplies power to the rotor and / or a speed sensor of a gas generator of the turbomachine.

[0022] Flight data may also include data on at least one environmental parameter, for example data on at least temperature and / or atmospheric pressure.

[0023] The stabilized model of the turbomachine can, for example, result from machine learning, and / or the equations of thermodynamics.

[0024] A method, and / or respectively a system, according to the invention may further comprise:

[0025] - a step or an additional step of comparison, respectively of the means (or programmed or specially programmed means) to perform a comparison of the residual, or of the rejected residual, to a threshold value;

[0026] - and an additional selection step or step, respectively of the means (or programmed or specially programmed means) to perform a selection, from said data, of at least one turbomachine parameter, of data or flight data for which the residual, or the recalibrated residual, is less than or equal to said threshold value,

[0027] and / or:

[0028] - a step or an additional step of comparison, respectively of the means (or programmed or specially programmed means), to perform a comparison, of the derivative of said turbomachine data for which the residual is calculated, with the derivative of said stabilized turbomachine model data;

[0029] - and a step or an additional step of selection from said data parameter of the turbomachine, respectively means for (or programmed or specially programmed means) to perform a selection, of data for which the difference between the derivative of this data and the derivative of the data of said stabilized model of the turbomachine is less than or equal to a second predefined value.

[0030] A method, and / or respectively a system, according to the invention, may comprise one or more of the following additional steps, and / or respectively means to, or programmed (or specially programmed) means to, implement one or more of the following steps or additional steps:

[0031] - a selection of data for which the NI rotational speed of the generator turbomachine gas is above a minimum threshold or is between a minimum value (X) and a maximum value (Y);

[0032] - and / or a calculation or estimation of at least one parameter of the turbomachine, a or several indicators, representative of the state of operation or health of the engine, for example from flight points retained following a treatment such as above, for example from flight points for which the residuals are low and their derivatives are also low;

[0033] - and / or, for example for at least one parameter of the turbomachine, a step estimation, of a margin by comparison between one or more measured value(s) of this parameter and one or more reference value(s), and possibly a calculation, of a non-linear regression of at least one margin and / or at least one margin filtering step.

[0034] The invention also relates to a computer program comprising instructions for implementing a method according to the invention, when this program is executed by a computer.

[0035] The invention also relates to a computer-readable recording medium on which a computer program according to the invention is recorded.

[0036] The invention also relates to a system for monitoring the health of a helicopter turbomachine, or a system for processing data measured in flight of a helicopter, referred to as flight data, comprising data on at least one parameter of a helicopter turbomachine, referred to as measured turbomachine data, this system comprising:

[0037] - means, or means programmed or specially programmed for, perform a comparison of these data from said at least one parameter of the turbomachine with data from a stabilized model of the turbomachine;

[0038] - means, or means programmed or specially programmed for, select data for at least one turbomachine parameter for which the deviation, or residual, with the stabilized model data is less than or equal to a predefined value.

[0039] The invention also relates to a health monitoring system for a helicopter turbomachine, comprising:

[0040] - sensors for measuring so-called flight data, this data comprising data for at least one parameter of the turbomachine;

[0041] - a system, according to the invention, for processing data measured in flight from a helicopter. Brief description of the drawings

[0042] Figures [Fig.1A] and [Fig.1B] represent the evolution over time of the variables TQ and T45 for 4 successive helicopter flights;

[0043] Figures [Fig.2A] and [Fig.2B] represent the evolution over time of the residuals of the parameters TQ and T45 for 4 successive helicopter flights;

[0044] Fig. 3 illustrates a method of filtering performance points, after their high-frequency evaluation by mathematical conditioning, and by filtering the residuals of the learned model;

[0045] The [Fig.4] represents steps of a process according to one embodiment of the invention;

[0046] Fig. 5 represents a system, comprising a helicopter, capable of implementing a method according to the invention;

[0047] Figure 6 represents a computer device capable of implementing a process according to the invention.

[0048] DETAILED DESCRIPTION OF SPECIFIC EMBODIMENTS

[0049] During a helicopter flight 100 ([Fig.5]), numerous parameters (or variables) can be recorded using different sensors (or probes) 140i, 1402 ..., 140n (Figures 5, 6). For example, one or more of the parameters of the helicopter's turbomachine (or engine) 110 ([Fig.5]) can be measured, for example NI (speed of the gas generator), and / or N2 (speed of the shaft that supplies power to the rotor), and / or T45 (temperature at the outlet of the high-pressure turbine), and / or TQ (engine torque), and / or VCAS (speed of the helicopter relative to the atmosphere).

[0050] One or more data points can also be recorded for one or more environmental parameters (or variables), for example: atmospheric temperature T0, and / or atmospheric pressure PO, ....

[0051] Within the scope of the present invention, a stabilized engine model (step S42, [Fig. 4]), referred to as the stabilized model or reference model of the turbomachine, produced by machine learning or according to the thermodynamic equations appropriate for this helicopter turbomachine, hereinafter referred to as the "defining equations," can be used. The reference engine model is constructed, for example:

[0052] - from flight data recorded and selected as "stationary" in the case of the machine learning approach;

[0053] - or from the stationary thermodynamic equations - without component temporal - for the thermodynamic approach.

[0054] The thermodynamic model is presented, for example, in the form of a calculation program that combines the maps describing the thermodynamic operation of the main components of the engine (compressors, combustion chamber and turbines), the description of the secondary air system, and other assumptions enabling the solution of the system of thermodynamic equations describing the turbomachine's physics. These equations describe the evolution of the air's thermodynamic properties (temperature, pressure, flow rate, etc.) as it passes through the various components (compressor, turbine, etc.) of the turbomachine, as well as the balance between the work done by the high-pressure turbine and the work done by the compressor. This modeling allows us to solve the system of equations and arrive at a balanced thermodynamic cycle. This, in turn, makes it possible to calculate the flow rate, pressure, and temperature at each engine plane, as well as the power delivered to the helicopter.

[0055] The model built by machine learning also takes the form of a computational program, which, for example, takes the same inputs as the thermodynamic model and calculates the same variables. The difference with the thermodynamic model lies in the equations used for the calculations. In the case of the learned model, the equations are determined using an experimental design that provides the results obtained for each variable in phases labeled as stable (for example, by operators) as a function of the input variables. The equations are the results of mathematical regressions (linear regression, random forest, support vector machine, neural network, or Gaussian mixture model) between the results and the inputs.

[0056] This model may or may not incorporate the installation effects related to the helicopter.

[0057] This model will allow, from the data relating to the parameters F of the turbomachine recorded during a flight, to establish a prediction F for each of these parameters (step S43 of [Fig.4]).

[0058] The variable TQ (torque) is represented in [Fig.1A] and, in [Fig.1B], the variable T45 (temperature at the outlet of the high-pressure turbine) for 4 successive flights of a helicopter.

[0059] In these figures, curve I (respectively II) represents the data measured by the helicopter's data collection unit, while curve F (respectively IF) represents the data simulated with a stabilized model.

[0060] Preferably, this data is filtered (step S44 of [Fig. 4]) to correspond to a sufficiently high engine speed to be representative of the operating conditions of the engine delivering power in steady-state mode at the top of its power range. For example, data are selected for which the rotational speed NI of the gas generator is above a minimum threshold X or is between a minimum value X and a maximum value Y (step 44 of [Fig. 4]).

[0061] Let F be a variable, and F' the estimate of this same variable with the stabilized model selected.

[0062] The term “residual” will be used to designate the difference between the measured variable F and the estimated variable F;

[0063] Residue = F - F.

[0064] These residuals are calculated (step S45 of [Fig.4]): the difference between the measured value of each parameter and its value predicted by the stabilized model is thus established.

[0065] We seek the minimum of the calculated residual:

[0066] Residumin = min (F - F)

[0067] An acceptable minimum is preferably less than or equal to the measurement errors of the corresponding probe, for example, the temperature, torque, or rotational speed probe (or, for example, equal to these cumulative errors, since engine performance depends on all these probes), but can be reduced even further compared to these measurement errors to limit the variability of the results. For example, a residual on the order of Kelvin (for example, between 1K and 5K) is used for the temperature residual, and a residual on the order of 0.2% of torque is used for the torque residual (for example, between 0.1% and 1%). The limit is therefore set by the user, who can choose it to be less than or equal to the accuracy of the corresponding probe.

[0068] A step of smoothing the residuals obtained (step S46 of [Fig.4]), for example by moving average and / or by Kalman filtering, can then be applied.

[0069] The residuals of each variable T45 and TQ can then be estimated per flight session for both engines and mathematically recalibrated (step S47 of Figure 4) to be comparable in absolute terms. The recalibration can be performed using a mathematical optimization method, by finding the "offset" e that minimizes the difference between the two signals. To obtain e such that the difference between pfy and p(t) is minimal, an optimization formula can be used, which can be solved with differential calculus techniques, for example, the least squares method: e = argmiu / (FU) — F(t) — f )2dt

[0070] Moreover, the dynamic phases of flight produce significant and highly noisy residuals.

[0071] Thus, Figures 2A and 2B represent the evolution over time of the residuals of the parameters TQ and T45 for 4 successive helicopter flights; more precisely:

[0072] - in [Fig.2A] the curves Iqd (respectively Iqg) represent the evolution, in function of time, residual data of TQ for the right (respectively left) motor;

[0073] - in [Fig.2B] the curves I45d (respectively I45g) represent the evolution, in function of time, residual data of T45 (high pressure turbine outlet temperature) for the right engine (respectively left).

[0074] In these figures, phases A, B, C, D, and E, which are the most dynamic phases, have also been identified and cannot be used for measurements. The stabilized zones, located between two phases adjacent to phases A-E, can be short, for example, on the order of 3 seconds (or of duration between 3 seconds and 30 seconds, 2 minutes, or 3 minutes), but can be exploited thanks to the present invention. One of the important advantages of the present invention is the ability to use extremely short phases, much more so than with known techniques.

[0075] It can be seen from these figures that, for phases A - E, there are significant noisy residuals (in other words, the difference between the measured and estimated values ​​is substantial). These significant noisy residuals can be identified by their value, which exceeds a threshold S (or first predefined value), which can be set by an operator. They can therefore be removed from the dataset, retaining only the sufficiently small residuals; for example, under engine bench test conditions, the dispersion in TQ is chosen to be between 0.1% and 1% and in T45 between 1K and 5K. Therefore, only the residuals F - F can be retained for which:

[0076] IF-Fe|^S

[0077] This inequality allows us to select only the flight points that have a calibrated residual F - F - e with a small absolute value. We calibrate F and F (F - F being the residual) with e to be able to determine the points of interest when F - F - e is small for convenience (rather than having very varied S values ​​between the variables); but we can, alternatively, use the residual F - F instead of the calibrated residual; this residual will then be compared to a threshold value Sr.

[0078] To avoid points of overlap, where the measurement and model curves intersect (the residual is zero there, but the variability is potentially significant), this type of processing can also be applied to the time derivative of the residual: the aim is to ensure that the time derivative of the difference between the measured and estimated values ​​is low or less than a threshold S' (or a second predefined value). The threshold S' used is, for example, between 0 and 25 K / s for temperature and between 0 and 2% / s for torque (S' may depend on the data acquisition frequency since these are discrete time series):

[0079] I d(F - F) / dt I < S'

[0080] The set of measurement points (“flight points”) obtained by a method as described above is reliable and corresponds to stable flight phases, which may be short (for example, a few seconds as indicated above) but which are not not imposed on the pilot; a significant advantage of the invention is the ability to utilize stable zones, even if these are very brief, and which can be much more frequent, even if the piloting is dynamic. In other words, the invention allows the use of data from these flight phases considered stable, even if these phases are short and in the absence of sufficiently long stable phases. The invention even allows phases that were previously not considered stable to be classified as stable.

[0081] For each flight, the amount of usable data can thus be increased. The invention therefore allows for more frequent evaluations, without dependence on stable flight conditions.

[0082] For the flight points retained following the above processing, for example the flight points for which the residuals are low and their derivatives are also low, one or more indicators, representative of the operating state or health of the engine, can be developed or calculated.

[0083] For example, an estimate of the margin(s) (e.g., in % for torque and in °C for T45) from the performance margin definition equations can be made (step S31, Figures 3 and 4). These equations can be derived from the initially chosen reference model.

[0084] A performance margin is defined as the difference between a measured value and a reference value (for example a limit value), to which can be added an installation loss model that depends on the helicopter.

[0085] Such a margin is an indicator of the health of the helicopter's turbomachine.

[0086] For example, a margin for temperature T45 indicates whether the turbomachine is operating at a temperature sufficiently close to, for example, the temperature recommended by the manufacturer. A margin relative to the torque TQ of the drive shaft indicates whether the turbomachine is delivering sufficient power, which is the case if the measured torque is not too far from the torque recommended by the manufacturer.

[0087] Margin filtering can be performed by applying mathematical conditioning according to industry rules (engine or aircraft operating phase (e.g., helicopter), external environment (pressure, temperature, etc.)). For example, an operator may choose to condition the margins on external pressure or altitude to ensure that the margins are only evaluated within a flight envelope where system losses have been well defined.

[0088] Furthermore, since a margin depends on engine variables (T0, PO, NI, N2, T45, TQ, VCAS), both environmental and helicopter-dependent, a nonlinear regression can be applied to the previously calculated margins; regression tools such as random tree forests or support machine regression can be used. nonlinear vectors, deep neural networks, or Gaussian mixture models. For example, for the temperature T45 at the outlet of the high-pressure turbine: M-45 - 7'0, FO, ,V2. TQ, VCAS,...)

[0089] And, for the TQ couple:

[0090] Mtq = f'TQ (T0, PO, NI, N2, T45, TQ, VCAS...)

[0091] Where:

[0092] - NI is the rotation speed of the gas generator;

[0093] - T0 is the ambient temperature;

[0094] - PO is the ambient pressure;

[0095] - N2 is the speed of the shaft that supplies power to the rotor;

[0096] - VCAS is the speed of forward movement of the helicopter relative to the atmosphere;

[0097] - TQ is the torque of the motor.

[0098] This new regression, by design, yields high scores (R² coefficient greater than 0.95, with a mean absolute error less than 0.1 for the temperature margin, for example). Indeed, this new regression replicates the equations used to calculate the margins. The difference in modeling lies in the underlying equations used, which depend on the chosen regression model. On the other hand, this new regression can take into account new flight parameters that a person skilled in the art deems appropriate. With this new regression, we can calculate a new residual. For large residuals, the points are considered unexplained by the model—which, it should be noted, has excellent validation scores—and these points can be removed from the potential performance estimates.

[0099] A filtering of the margins in accordance with this new nonlinear regression can then be performed. For example, temperature margin points estimated with this new model with an error greater than 5K are removed. More generally, the criteria used are again configurable by the expert, who typically chooses to retain only the points whose residual with respect to the new learned model is, for example, on the order of Kelvin (between 1K and 5K) for the temperature residual, and 0.2% of torque for the torque residual (between 0.1% and 1%). In other words, only the margin estimates that can be explained by this new learned model can be kept in the end. What is then explained by the model are the points that the model recalculates with a small error. Points that are recalculated with a large error are considered unexplained by the model.

[0100] Figure 3 represents the calculation of margins (step S31), then their filtering (step S32), and the non-linear regression steps (step S33), and the filtering of conformal margins. to non-linear regression (step S34), which leads to filtered margin data (step S35). In a process according to the invention, steps S32-S35 are optional.

[0101] Figure 4 represents an example of implementing a method according to the invention comprising the steps described above. As already explained, some of these steps may be optional.

[0102] A helicopter 100 to which the invention can be applied is schematically represented in [Fig.5].

[0103] It includes a turboshaft engine 110 and is equipped with a number of sensors 140i, 1402 ..., 140n which allow, during a flight, the measurement of one or more of the parameters NI, T0, PO, N2, VCAS, TQ already mentioned above.

[0104] The data measured using sensors 140i, 1402 ..., 140n can be transmitted to a computer 120 or an on-board computer. This computer comprises, for example ([Fig. 6]), a central processing unit, which itself includes a microprocessor 56, a set of non-volatile memories and RAM 57, peripheral circuits, all these elements being coupled to a bus 55. Data can be stored in the memory areas, in particular data for implementing a method according to the present invention: these areas form a computer-readable recording medium for implementing a method according to the invention or containing instructions for, when read by a computer, implementing a method according to the invention; other types of media may include a USB key or any other type of data storage medium used in computing and which, when read by a computer, allows implementing a method according to the invention.The onboard computer or computer is therefore programmed or configured to implement a process according to the invention. Means 59 will allow the management of the input (and in particular from sensors 140i, 1402 ..., 140n) and output data flow, and towards the various components of the helicopter. These means for receiving and / or transmitting wireless data may also be provided.

[0105] The data can be analyzed and processed by a method according to the invention. This analysis and / or processing can be carried out by this computer 120 or calculator, and / or by a computer or calculator in a maintenance center to which the data is transmitted. The latter may have the same type of elements as described above in connection with [Fig. 6].

[0106] The invention makes it possible to significantly increase the number of possible evaluation points and improves accuracy by constantly comparing flight data to a static model, even in the absence of stable flight phases in the sense of operational flight phases. In terms of evaluation frequency, between 2 and 100 times more performance measurement points are obtained for each flight performed and analyzed.

[0107] Consequently, engine performance can be calculated on missions for which known methods cannot be applied due to insufficiently long stable phases. This avoids requiring the helicopter crew to perform maneuvers or even a dedicated flight to evaluate engine performance.

Claims

Demands

1. A method for processing data (TO, PO, NI, N2, T45, TQ, VCAS) measured in flight of a helicopter (100), referred to as flight data, comprising data of at least one parameter (NI, N2, T45, TQ) of a turbomachine (110) of the helicopter (100), referred to as measured data of the turbomachine (110), for the purpose of monitoring the health of this turbomachine (110), this method being implemented by computer and comprising at least: - a step (S45) of comparing said data of said at least one parameter (NI, N2, T45, TQ) of the turbomachine (110) with data of a stabilized model of the turbomachine; - and a selection step (S48) of said data of said at least one parameter (NI, N2, T45, TQ) of the turbomachine (110), of the data for which the deviation, or residual, with the data of the stabilized model is less than or equal to a predefined value.

2. Method according to claim 1, the data (NI, N2, T45, TQ) of said at least one parameter of the turbomachine (110) comprising data from at least one temperature probe (T45) of the turbomachine and / or a torque sensor (TQ) of the turbomachine and / or a rotational speed sensor (N2) of a shaft of the turbomachine and / or a speed sensor (NI) of a gas generator of the turbomachine.

3. Method according to claim 1 or 2, the flight data further comprising data (T0, PO) of at least one environmental parameter, the latter data preferably comprising data of at least atmospheric temperature (T0) and / or atmospheric pressure (PO).

4. Method according to any one of claims 1 to 3, the stabilized turbomachine model resulting from machine learning and / or the equations of thermodynamics.

5. A method according to any one of claims 1 to 4, further comprising: - a step of comparing the residual, or the rejected residual, to a threshold value (S, Sr); - and an additional step of selecting, from said data of said at least one parameter (NI, N2, T45, TQ) of the turbomachine (110), the data for which the residual, or the rejected residual, is less than or equal to said threshold value (S, Sr).

6. A method according to any one of claims 1 to 5, further comprising: - a step (S48) of comparing the derivative of said turbomachine (110) data for which the residual is calculated, with the derivative of said stabilized model of the turbomachine; - and an additional step of selecting, from said data of at least one parameter (NI, N2, T45, TQ) of the turbomachine (110), the data for which the difference between the derivative of this data and the derivative of said stabilized model of the turbomachine is less than or equal to a second predefined value (S').

7. A method according to any one of claims 1 to 6, comprising an additional step (S44) of selecting the data for which the rotational speed NI of a gas generator of the turbomachine (110) is above a minimum threshold or between a minimum value (X) and a maximum value (Y).

8. A method according to any one of claims 1 to 7, comprising, for at least one parameter of the turbomachine (110), a margin estimation step (S31) by comparison between one or more measured value(s) of this parameter and one or more reference value(s).

9. A system (120) for processing data measured in flight of a helicopter (100), referred to as flight data, comprising data of at least one parameter (NI, N2, T45, TQ) of a turbomachine (110) of the helicopter (100), referred to as measured data of the turbomachine (110), this system comprising: - means (53, 54, 56, 57) for comparing this data (NI, N2, T45, TQ, VCAS) of said at least one parameter of the turbomachine with data of a stabilized model of the turbomachine; - means (53, 54, 56, 57) for selecting data of said at least one parameter of the turbomachine for which the deviation, or residual, with the data of the stabilized model is less than or equal to a predefined value.

10. A turbomachine (110) health control system of a helicopter (100), comprising: - sensors (140i, 1402 ..., 140n) for measuring flight data (T0, PO, NI, N2, T45, TQ, VCAS), this data comprising data (NI, N2, T45, TQ, VCAS) of at least one parameter of the turbomachine; - a system, according to claim 9, for processing data measured in flight from a helicopter.

Images