Control system for a device equipped with at least one redundant sensor with fault detection and isolation of said sensors

Abstract

This control system for a device (1) equipped with at least one piece of equipment (1i) includes at least one sensor (3) capable of measuring one operating parameter of the device, and two measurement channels (3ia, 3ib) of the redundant sensor for each piece of equipment (1i). It also includes at least one control means (2a, 2b) configured to determine a command for at least one piece of equipment (1i) based on the measurement of at least one parameter and a selected measurement. The system includes a computing means (5) configured to determine the measurement selected by an observer (5a) comprising two Kalman filters using an operating model of the device (1). The computing means (5) is configured to detect a failure of one of the measurement channels (3ia, 3ib) and to recalibrate the operating model based on at least one individual characteristic of the device (1). (See Figure 1 for abbreviations.)

Description

Title of the invention: Control system for a device equipped with at least one redundant sensor for detecting and isolating failures of said sensors Technical field

[0001] The technical field of the invention is control systems, and more specifically observer control systems.

[0002] In a non-limiting application, the invention relates to cylinder control systems acting on an actuator of an aircraft engine, such as a cylinder controlling the injection of fuel into the engine, and the control of the operation of the cylinder. Previous techniques

[0003] The control systems of an aircraft engine according to the state of the art use various types of sensors which ensure the measurement of a monitored operating parameter. In the case of cylinders controlling the fuel injection, the sensors ensure the measurement of the position of the cylinders in order to monitor compliance with an injection instruction.

[0004] This measurement is redundant. Two measurement channels are therefore used and should ideally provide the same measurement value, within measurement uncertainties.

[0005] The control systems according to the state of the art do not allow a bias or a drift of one of the measurement channels of a redundant sensor to be located. Indeed, during a "cross check" failure, determined by a difference between measurement channels greater than a threshold, an arbitrary choice is then made of the channel presenting the minimum value, the channel presenting the maximum value or an average of the two measurement channels. There is therefore a risk of continuing to use an erroneous measurement or a value dependent on this erroneous measurement.

[0006] The following documents are known from the state of the prior art.

[0007] Document US20200201312 discloses a technology for isolating and detecting faults in aeronautical anemometric sensors (total pressure, static pressure, etc.). To achieve this, it is described to use a bank of KALMAN filters and its residuals (i.e. differences between the estimated measurement and the actual measurement) coupled with a truth table to detect and isolate a faulty sensor or to reconfigure it on its model in the event of a fault.

[0008] Document CN110118128 describes a similar teaching but based on a particular formatting of the residues (WSSR: Weighted Sum of Squares of Residuals) for isolation and detection.

[0009] Document US2009043447 also describes a similar teaching, but based on different indicators (delta on averages of the evolutions of the states and / or the residues) to detect and isolate faults. In addition, in the case where the faulty sensor cannot be isolated, a probabilistic approach (occurrence of faults) is used.

[0010] These different state-of-the-art documents make it possible to detect the fault of a sensor among a set of sensors measuring different physical quantities. Furthermore, these different documents are based on different Kalman filter architectures in the form of "banks": several different Kalman filters are used with the same sensors and the interpretation of the residuals (i.e. differences between the model outputs and the sensor measurements) makes it possible to detect the sensor of a physical quantity that is faulty.

[0011] The applicant proposed in the unpublished patent application FR 2 200 875 a method based on a Kalman filter using a model representative of the operation of the device for the detection of the healthy channel during a difference between measurement channels on a redundant sensor.

[0012] However, the use of such a method can reveal significant differences between the model predictions and the corresponding quantities measured in flight. These differences are due to several factors, in particular manufacturing variability, engine aging or variations in power and air sampling. Statement of the invention

[0013] In view of the above, the aim of the invention is to improve the accuracy of the model in order to reduce the deviations between the predictions and the corresponding quantities measured in flight.

[0014] The subject of the invention is therefore a system for controlling a device provided with at least one piece of equipment, the control system comprising at least one sensor capable of each measuring an operating quantity of the device, and two measurement channels of the redundant sensor for each piece of equipment, at least one control means configured to determine a command intended for the at least one piece of equipment as a function of the measurement of at least one operating quantity of the device and of a measurement selected for the controlled equipment.

[0015] The system further comprises a calculation means configured to determine the measurement selected for the at least one controlled equipment, by an observer comprising two Kalman filters each using the same operating model of the device, as a function of the at least one measurement of the operating quantity of the device and the measurements of the measurement channels of the redundant sensor for the controlled equipment.

[0016] The calculation means is configured to detect a failure of one of the measurement channels. of the redundant sensor so as to exclude the measurement channel, the calculation means being further configured to recalibrate the operating model according to at least one individual characteristic of the device.

[0017] Such a control system allows the model to be recalibrated according to the individual characteristics of the device and therefore to improve the accuracy of the estimates made by the observer based on Kalman filters. The robustness of the accommodation of a failure is thus improved.

[0018] According to one characteristic, the calculation means comprises: - an observer comprising two Kalman filters each using the same operating model of the device and determining for one an estimate of a bias of a first measurement channel, and for the other an estimate of a bias of a second measurement channel, as a function of at least one measurement of the operating quantity of the device, of the measurement of the first measurement channel of the redundant sensor and of the measurement of the second measurement channel of the redundant sensor, - a comparison means configured to emit a selection signal as a function of a comparison between, on the one hand, the absolute value of the difference between the estimate of the bias of the first measurement channel and the estimate of the bias of the second measurement channel, and on the other hand, a predetermined verification threshold between the two measurement channels, - a selection device configured to transmit the selected measurement equal to the measurement of the first measurement channel, to the measurement of the second measurement channel or to the average of the measurements of the first measurement channel and of the second measurement channel according to said selection signal, and - a recalibration means configured to recalibrate the operating model used by said observer according to corrective factors representative of at least one individual characteristic of the device.

[0019] For example, the recalibration means comprises a memory intended for storing the corrective factors in the form of a map.

[0020] Advantageously, the recalibration means is configured to iteratively update the values ​​of the mapping when the device operates in a stabilized mode.

[0021] Such a recalibration means allows the recalibration of the model according to the individual characteristics of the device taking into account both the manufacturing variability and its evolution over time.

[0022] Preferably, the iterative updating is performed according to a predetermined gain factor.

[0023] Advantageously, the comparison means is configured to carry out the steps

[0024]

[0025]

[0026]

[0027]

[0028]

[0029]

[0030] following: the bias estimates are stored at the time when it is determined that the absolute value difference between the measurement of the first measurement channel and the measurement of the second measurement channel is greater than a predetermined threshold and a selection signal is issued indicating to the selection device to transmit a selected measurement equal to the measurement of the measurement channel of the redundant sensor for which the absolute value of its bias estimate is the lowest. Preferably, the bias estimates continue to be made even after storage. For example, the device is an aircraft engine and the measurement channels of the redundant sensor measure the position of the cylinders controlling the variable geometries of the compressors, the variable section of the nozzle or the fuel metering valve. According to another aspect, the invention relates to a method for controlling a device provided with at least one piece of equipment, at least one sensor capable of measuring an operating quantity of the device, and measurement channels of the redundant sensor for each piece of equipment, the control method comprising steps of: - determining a command intended for at least one piece of equipment as a function of the at least one measurement of the operating quantity of the device and a measurement selected for the controlled piece of equipment, - determining a measurement selected for the at least one controlled piece of equipment, by an observer comprising two Kalman filters each using the same operating model of the device adjusted by corrective factors representative of at least one individual characteristic of the device,based on at least one measurement of the operating quantity of the device and measurements of the measuring channels of the redundant sensor for the controlled equipment. Advantageously, the control method comprises a prior step of iteratively updating the corrective factors carried out when the device is operating in a stabilized mode. According to another aspect, the invention relates to an aircraft equipped with a control system or comprising a control method as described above. Brief description of the drawings Other aims, characteristics and advantages of the invention will appear on reading the following description, given solely by way of non-limiting example, and made with reference to the appended drawings in which: [Fig.l] illustrates the main elements of a control system based on two Kalman filters according to the invention; [Fig.2] illustrates the main elements of a calculation means provided on a control system according to the invention; and

[0031] [Fig.3] illustrates a flowchart of an update of the corrective factors according to the invention. Detailed description of at least one embodiment

[0032] The control system according to the invention is based on the use of two Kalman filters to detect and locate a simple failure of a redundant sensor in order to accommodate acquisition on the healthy channel instead of arbitrarily choosing a channel and potentially selecting the wrong channel. According to the invention, the two Kalman filters are identical, but each supplied by a different measurement channel, namely a first measurement channel called the local channel for one and a second measurement channel called the "cross" channel for the other.

[0033] In an application to an aircraft, and to the control of the engine of an aircraft, it is then possible to avoid an engine event, such as a loss of thrust control LOTC (acronym for “Loss of Thrust Control”).

[0034] [Fig.l] illustrates the main elements of an observer-based control system comprising two Kalman filters according to the invention.

[0035] As previously indicated, such a control system has applications in aeronautics and is used to control a device, such as an engine for an aircraft. For example, the control system can be used for controlling the operation of a fuel injection control actuator in an aircraft engine.

[0036] Thus, a device 1 is controlled by a control means 2a, 2b as a function of the difference between a setpoint CS of an operating parameter of the device 1 (for example, the rotation speed in the case of a motor) and a measurement Y by a sensor 3 of the operating parameter of the device 1 corresponding to the setpoint CS.

[0037] The difference between the CS setpoint and the Y measurement is determined by means of a subtractor 4.

[0038] The first control means 2a may be an open-loop or closed-loop control means and makes it possible to determine the command uc. The second control means 2b is a closed-loop control means.

[0039] More precisely, the first control means 2a makes it possible to determine a control setpoint uc as a function of the difference between the setpoint CS and the measurement Y when the first control means 2a is a closed-loop control means. The first control means 2a makes it possible to determine a control setpoint uc as a function of the measurement Y when the first control means 2a is an open-loop control means. This control setpoint uc is received by a second control means 2b in order to determine a command u of at least one device li taking into account a selected measurement us received from a computing means 5 comprising an observer 5a based on two Kalman filters 5ai and 5a2 ([Fig.2]).

[0040] It will be understood that the use of two control means 2a, 2b makes it possible to control several separate pieces of equipment li acting on the devices as a function of the measurement Y and / or the setpoint CS according to the structure of the first control means 2a. In the case of an aeronautical engine controlled in this way, the pieces of equipment li controlled may be the fuel meter or the variable geometry blades of said engine. These pieces of equipment contribute to the operation of the engine, and to the realization of the control, for example the rotation speed of the engine. It will thus be understood that to realize the setpoint CS through the measurement Y, the control u is of vector type, and comprises a measurement selected for each piece of equipment li of the device 1. It should be noted that the quantities uc, us, CS, Y, uL and ux may be vectors.

[0041] In order to ensure that each equipment li carries out the command determined by the second control means 2b, measurement channels 3ia, 3ib specific to each equipment li are used. Each pair of measurement channels 3ia, 3ib comes from the redundancy of a measurement sensor and its acquisition chain. They thus measure the same quantity so that the measurement is always available even in the event of failure of one of the measurement channels 3ia, 3ib of a specific redundant sensor. The measurements of the first measurement channel 3ia, or local channel, of several sensors (with i varying 1 to j) can be concatenated in the vector uL. The measurements of the second measurement channel 3ib, or cross channel, of several sensors can be concatenated in the vector ux. In [Fig.l], a single equipment li is illustrated for the sake of readability.It will be understood that an aircraft engine is generally equipped with multiple equipment li, each equipped with a pair of specific redundant sensors 3ia, 3ib.

[0042] Nevertheless, this redundancy then poses the problem of determining the measurement channels 3ia, 3ib to be taken into account in the calculation of the selected measurement us to be transmitted to the second control means 2b.

[0043] Indeed, simple cases such as the absence of a signal or identical signals can be treated with logical operators. However, more complex cases such as a drift or bias of a sensor measurement channel cannot be treated in this way.

[0044] The calculation means 5 receives as input the measurements uL,ux of the measurement channels 3ia,3ib of the redundant sensor and one or more measurements Y of the sensor 3. It outputs the selected measurement signal us to the second control means 2b. The calculation means 5 is provided with an observer 5a comprising two Kalman filters 5ai and 5a2( [Fig.2]), making it possible to determine the selected measurement us as a function of the signals from the sensors uL,ux,Y. The observer uses a model representative of the operation

[0045]

[0046]

[0047]

[0048]

[0049]

[0050]

[0051]

[0052]

[0053]

[0054] of the device 1 and of at least one piece of equipment 11 and makes it possible to isolate a faulty measurement channel coming from one of the measurement channels 3ia, 3ib of the redundant sensor measuring each piece of equipment 11. The selected measurement us is obtained by comparing the estimated additive biases on the first measurement channel 3ia with the estimated additive biases on the second measurement channel 3ib. The Kalman filters then allow, by varying in real time the additive biases on each of the model inputs, to converge the model outputs with the Y measurements of the motor. Convergence allows to minimize the residuals. The biases obtained then directly represent the difference between the real measurement of the controlled equipment (determined by observation via its effect on the device) and the chosen measurement uL, ux from one of the measurement channels 3ia,3ib of the sensor. The adjustment of the filter performance (gains) is optimally achieved through the resolution of a standard problem (e.g., RICCATI algebraic equation). More precisely, the calculation means 5 carries out the estimation of the biases associated with each measurement channel 3ia or 3ib of the redundant sensor. [Fig.2] illustrates the main elements of the calculation means 5 according to the invention. The calculation means 5 comprises an observer 5a using two Kalman filters referenced 5ai and 5a2, a comparison means 5b, and a selection device referenced 5c. As indicated previously, the observer 5a uses a model representative of the operation of the device 1. In the preferred embodiment, the model used is a tangent linearized model and the observer 5a comprises two Kalman filters using this same model, but each fed by a different measurement channel. It remains, however, possible to use any disturbance estimator which uses a tangent linearized model. It is also possible to replace the tangent linearized model with a physical and non-linear model. In the latter case, the Kalman filters used to estimate the disturbances on the actuator signals are extended Kalman filters which make it possible to locally linearize the problem. The representative model of the operation of the device is based on a state representation of the system represented by the device 1. As is known, a state representation of a linear system can be written: xk+1 = xk + B uk ( 1 ) . yk = Cxk+Duk where A, B, C, D are matrices that allow us to represent the contributions of the states xk and the commands uk on the value of the states at the following instant xk+[ and the

[0055]

[0056]

[0057]

[0058]

[0059] yk measurements. However, real systems are generally non-linear and their state representation then takes the following general form: (-^+1= / ½. «J <2) I where f and g are any functions which allow the system to be linearized by Taylor series developments around any operating point of the system. We then obtain in the vicinity of a point {Xq, Wq) the following linearized tangent state representation: df I / \ ùf | (3)

[0060]

[0061] in which, ' A = B = C = D- egg | df | dh i ck ! \ dh! 3« (4) In the context of accommodating measurement deviation failure on a redundant sensor on an aircraft engine, such as an LVDT (acronym for "Linear Variable Differential Transformer") sensor measuring the linear position of a cylinder on one of the engine actuators, the following state representation can be chosen: XN2R XN25R Or: XN2R = XN2 jT2lTsld XN25R = WF32R2 A8 ARDE ASTA XN2R XN25R y “ DPQ23 PS32Q2 (5) XN25 fr25lTsul WF32R2 =--— Z 72 y - / P2 A PS32Q2 = PS32 / P2

[0062] With: - XN2 the rotation speed of the low pressure (LP) shaft, - XN25 the rotation speed of the high pressure (HP) shaft, - T2 the temperature in the thermodynamic plane of the LP compressor, - T25 the temperature in the thermodynamic plane of the HP compressor, - P2 the pressure upstream of the LP compressor, - WF32 the fuel flow into the combustion chamber, - XN2R the reduced speed of the LP shaft, - XN25R reduced HP shaft speed, - IVF327?2 reduced fuel flow in the combustion chamber, - A8 the variable section of the nozzle, - ARDE the angle of the LP compressor inlet steering wheel, - ASTA the angle of the variable stators of the HP compressor, - DPQ23 the ratio between static pressure and total pressure in an intermediate plane between the LP and HP compressors, - PS32Q2 reduced static pressure in the combustion chamber, - ?std a standard reference temperature, - ?std a standard reference pressure.

[0063] Note that the XN2 and XN25 regimes are both in the x states, since they are governed by a differential equation that we consider through the state representation, and in the measurements since we measure them.

[0064] We then have:

[0065] / XN2R \ k+l / XN2R[j \XN25Rq ) = A ' k \ / XN2R \ / kl ■ / XN2R XN2R0 / ^^32X2] iWF32R20 XN25R (7) 0 1, and Or

[0066] ran «12 “ [«21 a22. A41 C23 C'24

[0067] ' 0 0 0 0 ■ 0 0 0 0 D = 41 41 4s 44 .41 42 4 3 44.

[0068] The coefficients Clij, b;k Cij, dÿ present in the matrices A, B, CD as the elements of *o (XN2RG and XN25R^ (WF32R2^ A80, ARDE0 and ASTA^ and >0 (XN2Rq, XN25R^ DPQ23o and PS32Q2q) are obtained by a classical identification procedure.

[0069] It should be noted that the first two lines of C and D are trivial since we have XN2Rk = XN2Rk and XN25Rk = XN25Rk.

[0070] As the engine behavior evolves with its load, we can identify matrices A, B, C, D and vectors xo> for operating points progressively distributed between idle and full throttle. We then obtain a state representation conforming to equation 1 for each operating point, and between two identified operating points the system model is given by a term-by-term interpolation of the elements of these matrices.

[0071] The operating quantity representing the position along the operating line may typically be PS32Q2, i.e. the reduced static pressure in the combustion chamber. However, it remains possible to use any other quantity evolving bijectively with the engine load. For example, it is possible to use operating quantities such as the pressure P32, the speeds XN2, XN25, the reduced speeds XN2R, XN25R, the fuel flow WF32 or the reduced fuel flow WF32R2.

[0072] This representation is just on the ground, but the behavior of the engine also varies according to the ambient air conditions, themselves a function of the altitude (noted ALT) and the Mach number (noted XM) of the aircraft. The values ​​of the coefficients of A, B, C, D, Xq, m0, y are therefore not only derived from a one-dimensional interpolation along the operating line, but also from an interpolation in a cube whose three dimensions are typically PS32Q2, ALT and XM.

[0073] We can then express the state representation in the following form: = A(seq) • (x^C^) ) +B(seq) • (uk-u0(seq) ) (8) . >'k'>'o(se^) = C(se^ ' (xk-*o())+D (seq) • (uk-uQ(seq) ) Where (PS32Q2 (9) seq= ALT . XM

[0074] It should be noted that although the determination of the values ​​of the elements x& has been described with reference to an interpolation typically carried out from a table of values, it is possible to use instead, mathematical functions representing the evolution of each quantity of the model along an operating line.

[0075] Also, although the description refers to the altitude and the Mach number to take into account the ambient air conditions, it remains possible to replace these sequencings with other quantities representative of the air conditions at the engine terminals.

[0076] The Kalman filters 5ai and 5a2 of the observer 5a make it possible, taking into account one or more measurements Y, to determine, respectively, an estimate of the bias of the first measurement channel ûL corresponding to the measurement uL of the first measurement channel 3ia of the redundant sensor, and an estimate of the bias of the second measurement channel ûx corresponding to the measurement ux of the second measurement channel 3ib of the redundant sensor.

[0077] The comparison means 5b is configured to determine a selection signal, denoted sel, as a function of the comparison between, on the one hand, the absolute value of the difference between the estimation of the bias ûL of the first measurement channel and the estimation of the bias ûx of the second measurement channel, and on the other hand, a predetermined verification threshold between the two measurement channels, a so-called “cross-check” threshold.

[0078] The choice of the selected measurement is made according to the cross-check threshold. If the cross-check threshold is not exceeded, the arithmetic mean of the two measurement channels is selected.

[0079] If the cross-check threshold is exceeded, there are two possible cases. According to a first case, if the estimated bias on the first measurement channel of the actuator concerned by the cross-check failure is greater in absolute value than the estimated bias on the second measurement channel, the measurement of the second channel is selected. According to a second case, if the estimated bias on the first channel of the actuator concerned by the cross-check failure is less than or equal in absolute value to the estimated bias on the second measurement channel, the measurement of the first channel is selected.

[0080] Depending on the result of the comparisons, the selection signal sel is then transmitted to the selection device 5c also receiving as input the measurement uL of the first measurement channel 3ia and the measurement ux of the second measurement channel 3ib.

[0081] The selection device 5c comprises a calculation means 5d configured to carry out the average of the measurement uL of the first measurement channel 3ia and of the measurement ux of the second measurement channel 3ib.

[0082] The selection device 5c also comprises a logic gate 5e receiving as input on the one hand the selection signal sel and on the other hand the measurement uL of the first measurement channel 3ia, the measurement ux of the second measurement channel 3ib and the average of the uL measurement of the first measurement channel 3ia and the ux measurement of the second measurement channel 3ib.

[0083] Depending on the selection signal sel received, the selection device 5c transmits the measurement uL of the first measurement channel 3ia, the measurement ux of the second measurement channel 3ib or the average of the measurement uL of the first measurement channel 3ia and the measurement ux of the second measurement channel 3ib in the form of the selected measurement signal us.

[0084] The following [Table 1] illustrates the truth table resulting from the actions of the comparison means 5b and the selection device 5c.

[0085] [Tables 1] 1 uL - ux 1 > P ÛL > Ûx salt 0 N / A (uL+ ux) / 2 1 0 uL 1 1 ux

[0086] The calculation means 5 makes it possible to detect that a measurement channel 3ia, 3ib is broken or that it is faulty. A measurement channel is considered to be faulty if the difference measured between the two channels is greater than the cross-check threshold and if the estimate of the bias associated with the other channel is lower in absolute value than that of the channel in question. The observer then uses the measurements provided by the healthy measurement channel of the redundant sensor. This arbitration can be done in the form of a truth table.

[0087] The calculation means 5 comprises a recalibration means referenced 5f configured to recalibrate the operating model of the device 1 used by the observer 5a as a function of corrective factors representative of at least one individual characteristic of the device.

[0088] The recalibration consists of correcting the linearized tangent state representation described by equations 8 which becomes: xk+l -xa)rr(seq) - A(seq) ■ (xk-xcm.(seq')) +B(seq) ■ (uk-u^,.(seq)) (10) . = C(seq) ■ (xk-xcorr(seq) ) +D(seq) • (uk-ucorr(seq) )

[0089] where XN'lR^y XN25Rcorr DPQ23corr PS32Q2mrr WF32R2corr AQ F^corr ARDEcorr ASTAcorr XN2R0-km(PS32Q2) (11) XN25R„kXAi25l!(PS32Q2) DPQ230-kDFaa(PS32Q2) PS32Q20- kPsi2&(PS32Ql) WF32R20-kw32lt2(PS32Q2) A80-kAi(PS32Q2) ARDE(,ktila,E(PS32Q2) ASTAiykASTA(PS32Q2)

[0090] The reference point xo )'o no longer appears directly in the state representation, but each element of equation 11 is corrected by multiplying each operating quantity by a corrective factor noted k, specific to the corrected quantity and which depends on the position along the operating line, typically as a function of PS32Q2.

[0091] Such a recalibration by corrective factors k makes it possible to correct the operating model of the device by taking into account the individual characteristics of the device.

[0092] To ensure the effectiveness of the Kalman filters in designating the healthy channel, and to prevent a recalibration from updating corrective factors when a difference between measurement channels has appeared, the calculation means 5 is configured according to a predefined difference threshold. If the difference measured between the measurement channels is below the predefined difference threshold, the calculation means is configured to perform the recalibration and to deactivate the Kalman filters. Conversely, if the difference measured between the measurement channels is equal to or above the predefined difference threshold, the calculation means is configured to deactivate the recalibration and activate the Kalman filters. For example, if the measurement differences between the local channel and the cross channel are less than 50% of the cross-check threshold for each of the four actuators considered, the recalibration is activated and the Kalman filters estimating the disturbances on the control signals are deactivated.Otherwise, the Kalman filters are enabled and the recalibration is disabled. Alternatively, it is still possible to keep the 5ai and 5a2 Kalman filters enabled and disable the recalibration when the predefined deviation threshold is exceeded.

[0093] The corrective factors are stored in a memory 5g of the recalibration means 5f in the form of a map. The map may correspond to a set of tables, each table linking a corrective factor to an operating parameter of the device.

[0094] The resetting means 5f is configured to regularly update this car tography in order to take into account the evolution of the individual characteristics of the device 1 over time, particularly in relation to aging.

[0095] When first used, the mapping or tables are initialized with unit values. In other words, before the first flight, the elements of each table containing the values ​​of the corrective factors are all equal to 1.

[0096] The updating of the corrective factors is carried out in a similar manner at each iteration and for each element to be recalibrated.

[0097] The updating of the corrective factors is carried out when the device is operating in a stabilized regime. The operation in a stabilized regime is determined with respect to a predefined threshold. For example, a controlled device is considered to be operating in a stabilized regime if the average variation of parameters representative of its operation considered over at least one predetermined period is less than a predetermined threshold. For example, an aircraft engine can be considered to be operating in a stabilized regime if the following indicator I(t) has a value less than 1. rHHOQl / / / ) - 1 / 1-^25 / 5(^-^5^(^-) I 1^25 / 5(:)^25 / 5(^ ) 1^25 / 5(:)-^25 / 5( / -5.0 I 1^32252(:)-^32 / 52( / -1^) [ j WF32R2( / )-WF32R2(!-3.ï) I \WF32iait)-WF32R2(l-5x) |\ [vvyoj - ôtl XN25R[()-Seu^x^ | + | XA'25A(? ) 5^^ |+| XN2SR(i) | + | WF32R2(àSeitiiKryZI,z | + j W32R2(l]Sei<'lKnzRZ | + | WF32R2(F)Xc!^ |) (12)

[0099] With: ThresholdXN25R —ThresholdWF32R2 — 1% (13)

[0100] In other words, if I(t) = 1 then the average variation of XN25R and WF32R2, taken with respect to the instants t - kv, t - 35 and t - 5s, is worth 1%.

[0101] The following describes the flowchart of [Fig.3] which summarizes the architecture of the calculation code associated with the updating of the corrective factors and implemented by the recalibration means 5f within the framework of the control system according to the invention.

[0102] The description is given by way of example and with reference to WF32R2, i.e. the reduced fuel flow rate in the combustion chamber. The calculation code can of course be transposed in a similar manner to any other operating quantity to be recalibrated.

[0103] As indicated previously, the corrective factors kWF32R2 are stored in a memory 5g of the recalibration module 5f in the form of a map or tables.

[0104] Updating the corrective factors begins with a step 10 of determining the input values.

[0105] The input values ​​consist of atmospheric parameters measured by probes, mechanical parameters measured by sensors present on the engine and calculated parameters specified below.

[0106] Preferably, the input values ​​include the current values ​​of PS32Q2, ALT, XM and the fuel flow rate, denoted WF32R2measured. The input values ​​also include the current value of WF32R20 obtained by interpolation as a function of PS32Q2, ALT and XM from a table obtained by a procedure conventional identification. The input values ​​further include the value of the correction factor kWF32R2 obtained by linear interpolation in the corresponding table as a function of PS32Q2, as a function of the coordinate xPS32Q2 within a given mesh and the corresponding tabular values ​​at the nodes of the mesh.

[0107] During the following calculation step 11, the recalibration module 5f first calculates the corrected value WF327?2C(9rr of the current reference point by multiplying the current value of the quantity to be corrected by the corresponding corrective factor:

[0108] WF32R2c„rr WF32R2O • kWF32R2 (14)

[0109] During the same calculation step 11, the recalibration module 5f calculates the relative difference A WF32R2 between the expected fuel flow rate and the measured fuel flow rate:

[0110] a ïyF'jop'? WF32R2cwFWF32R2„,e,urea (15) WF32R2r:orr

[0111] During the same calculation step 11, the recalibration module 5f calculates the filtered relative deviation A WF32R2f between by multiplying the relative deviation by a predetermined gain factor gainWF32R2 which makes it possible to adjust the speed of convergence:

[0112] A WF32R2fihre A W32Æ2 • gainWF32R2 (16)

[0113] During the following update step 12, the recalibration module 5f assigns the filtered relative deviation A IVF32R2f between the two adjacent nodes to the current value of PS32Q2, in the table of corrective factors kWF32R2, in proportion to the proximity of the current PS32Q2 with each of these two nodes. By denoting IpsxiQZ the index of the current mesh of this table, we then have:

[0114] K (T \ k (t \ . A Ii / mm P^2Q2(1 ps-iiQ^-xps^Qi kWF32R2 ( hs32Q2 ) kwF32R2 ( 1PS32& ) + AH r 3^2 / / / f,, • ^2^(^ (17)

[0115] and

[0116] kWF32K2 (IPSi2& + 1) *- kWF32K2 (IPS32Q2 + 1) + A W327?2 / (7fre - (18)

[0117] Updating the corrective factors of the quantities XN2Rq, XN25Rq, DPQ23^ PS32Q2iy, ARDE0 and ASTA^ are identical to that of WF327?20, replacing the kWF32ia table with the corresponding table.

[0118] The control system has been described in relation to a single quantity measured by two specific measurement channels 3ia, 3ib of a redundant sensor. Nevertheless, the control system can be extended to at least two quantities each measured by two specific measurement channels of redundant sensors. In this case, the same observer 5a based on two Kalman filters 5ai and 5a2 is used in order to determine the estimation of the bias of the measurement channels for each quantity taken into account.

[0119] For each quantity, we then have two bias estimates, a means of comparison, as well as a means of selection.

[0120] The control system allows: • to isolate, for each redundant sensor, the faulty channel from the healthy channel and thus be able to reconfigure itself, • to use the same observer based on two Kalman filters and fed by different measurement channels. Thus, unlike the state of the art which uses a bank of different Kalman filters, this makes it possible to optimize the control system (reducing the impact on memory or CPU load) • to obtain an estimate of the bias directly dependent on the measurements and easily interpretable subsequently (unlike the residuals used in the state of the art) • to improve the accuracy of disturbance estimates and therefore the robustness of the designation of the healthy channel by recalibrating the model used by the observer based on the Kalman filter.

[0121] Applied to an aircraft, and in particular to the control of an aircraft engine, the control system makes it possible to be robust to failure during flight and to provide a diagnosis for replacement at the end of the mission.

Claims

Claims

1. System for controlling a device (1) provided with at least one piece of equipment (li), the control system comprising at least one sensor (3) each capable of measuring an operating quantity of the device, and two measurement channels (3ia, 3ib) of the redundant sensor for each piece of equipment (li), at least one control means (2a, 2b) configured to determine a command to the at least one piece of equipment (li) as a function of the measurement of at least one operating quantity of the device (1) and of a measurement selected for the controlled piece of equipment (li), the system further comprising a calculation means (5) configured to determine the measurement selected for the at least one controlled piece of equipment (li), by an observer (5a) comprising two Kalman filters (5ab 5a2) each using the same operating model of the device (1), as a function of the at least one measurement of the operating quantity of the device and of the measurements of the measurement channels (3ia,3ib) of the redundant sensor for the controlled equipment (li), characterized in that the calculation means (5) is configured to detect a failure of one of the measurement channels (3ia, 3ib) of the redundant sensor so as to exclude the measurement channel, the calculation means (5) being further configured to recalibrate the operating model as a function of at least one individual characteristic of the device (1).,

2. A control system according to claim 1, wherein the computing means (5) comprises: - an observer (5a) comprising two Kalman filters (5ab 5a2) each using the same operating model of the device and determining for one (5a0) an estimate of a bias (ûL) of a first measurement channel (3ia), and for the other (5a2) an estimate of a bias (ûx) of a second measurement channel (3ib), as a function of the at least one measurement of the operating quantity of the device, of the measurement of the first measurement channel of the redundant sensor and of the measurement of the second measurement channel of the redundant sensor, - a comparison means (5b) configured to emit a selection signal as a function of a comparison between, on the one hand, the absolute value of the difference between the estimate of the bias (ûL) of the first measurement channel (3ia) and the estimate of the bias (ûx) of the second measurement channel (3ib), and on the other hand, a predetermined verification threshold between the two measurement channels, - a selection device (5c) configured to transmit the selected measurement equal to the measurement of the first measurement channel, to the measurement of the second measurement channel or to the average of the measurements of the first measurement channel and of the second measurement channel as a function of said selection signal, and - a recalibration means (5f) configured to recalibrate the operating model used by said observer (5a) as a function of corrective factors representative of at least one individual characteristic of the device.

3. System according to claim 2, in which the recalibration means (5f) comprises a memory (5g) intended for storing the corrective factors in the form of a map.

4. System according to claim 3, in which the recalibration means is configured to iteratively update the values ​​of the mapping when the device operates in a stabilized mode.

5. The system of claim 4, wherein the iterative updating is performed based on a predetermined gain factor.

6. Control system according to any one of claims 2 to 5, in which the comparison means (5b) is configured to carry out the following steps: the bias estimates are stored at the moment when it is determined that the difference in absolute value between the measurement of the first measurement channel and the measurement of the second measurement channel is greater than a predetermined threshold and a selection signal is emitted indicating to the selection device (5c) to transmit a selected measurement equal to the measurement of the measurement channel of the redundant sensor for which the absolute value of its bias estimate is the lowest.

7. Control system according to any one of claims 1 to 6, in which the device is an aircraft engine and the measurement channels of the redundant sensor measure the position of the cylinders controlling the variable geometries of the compressors, the variable section of the nozzle or the fuel metering valve.

8. Method for controlling a device (1) provided with at least one piece of equipment (li), at least one sensor (3) capable of measuring an operating quantity of the device, and measurement channels (3ia, 3ib) of the redundant sensor for each piece of equipment (li), the control method comprising steps of: - determining a command intended for at least one piece of equipment (li) as a function of the at least one measurement of the operating quantity of the device and a measurement selected for the controlled piece of equipment (li), - determining a measurement selected for the controlled piece of equipment (li), by an observer (5a) comprising two Kalman filters (5ab 5a2) each using the same operating model of the device adjusted by corrective factors representative of at least one individual characteristic of the device (1),depending on at least one measurement of the operating quantity of the device (1) and the measurements of the measuring channels (3ia, 3ib) of the redundant sensor for the controlled equipment (li).,

9. Method according to claim 8 comprising a prior step of iteratively updating said corrective factors carried out when the device (1) operates in a stabilized regime.

10. Aircraft provided with a control system according to any one of claims 1 to 7 or provided with a system implementing a control method according to claim 8 or 9.

Images