The present invention provides accurate measurement of speed of an aircraft engine.

This is specifically to take into account the existence of noise in addition unpredictably to the useful signal, when measuring the speed of a motor shaft by sensor tone wheel.

PRESENTATION GENERAL DE INVENTION

Accurate detection of the rotational speed of an aircraft engine shaft is necessary, insofar as this information can control the cutoff or regulating the motor power supply, for example if the rotational speed or the acceleration exceed a predetermined safety value.

It is well known to measure a shaft rotation speed using a variable reluctance sensor associated with a phonic wheel. sensor invariably one can speak phonic wheel, or variable reluctance sensor. The phonic wheel is arranged to be secured to the shaft. Creates the sensor, with a winding, a magnetic field which closes either in the air to the space between two teeth of the phonic wheel or on a tooth of the phonic wheel. Thus, the measured magnetic flux varies depending on the passage of the teeth of the phonic wheel, proportional to the rotational speed of the tone wheel. The frequency of the alternating voltage generated is equal to the frequency of passage of the teeth of the phonic wheel, itself illustrative of the speed

rotation of the shaft, and the amplitude of this AC voltage depends on the gap and frequency of the signal. The Applicant has described in the French patent application published under number WO 2014/207369, a toothed nut which can be screwed around a shaft of an aircraft engine, for example for securing the shaft mechanically with a ball bearing, the teeth of which can cooperate with a rotational speed sensor operating by this principle. At least a portion of the teeth of the nut further comprises recesses, preserving the clamping function of the nut while providing a larger number of gap slots in the workpiece to allow an accurate measurement of rotational speed the tree.

Rotational speed measurement systems including phonic wheel can be used to measure the system says "N1" of rotation of the low pressure spool of a turbofan engine body. On the "direct drive" motors (direct drive blower by the low pressure shaft), this is also the N1 fan speed. A phonic wheel sensor can also be used to measure the N2 speed of rotation of the high pressure body of the twin-spool turbine engine, or the rotation speed of another motor rotating element.

The alternating sinusoidal acquired at the terminals of a variable reluctance sensor signal can be converted, by an electronic circuit placed upstream of the engine ECU, squarewave signal which the engine control unit can perform a frequency measurement. The transformation of the sinusoidal signal squarewave signal is particularly realized by a Schmidt trigger with a restart threshold and a predetermined reset threshold. The parallel signal output from Schmitt trigger switches to a high value when the input voltage exceeds the reset threshold, and switches to a low value when the input voltage drops below the restart threshold.

But the niche obtained signal to a phonic wheel speed sensor can be very disturbed due to noise. Parts of the sensor are subject to electromagnetic disturbances, due to mechanical vibrations of certain engine parts. One can observe occasional noise events, but also of the reciprocating type noise signals superimposed on the signal representative of the rotation of the tone wheel (useful signal). As such, Figure 1 illustrates, on a first axis, the voltage across the Sv sensor phonic wheel over time, in the case of a signal corrupted by an aliasing alternative type of higher frequency than the useful signal, and illustrates, on a second axis, the slot signal Sc obtained after aliasing of the voltage signal Sv, output trigger. The peak to peak amplitude of the noise may be locally higher than the voltage difference between the restart threshold and the UR AU reset threshold, the AC noise can cause noise switchings of the signal slot. A third line shows the variations of a counting signal CP successive periods of the pulse signal If over time. The timing illustrated by the curve of this third axis is restarted each time slot down switching signal, the high level to the low level. For example, during the period P shown in the figure, it does not occur spurious switching due to noise. However, the following period P2, the detected signal Sv AU crosses the reset threshold and then drops back below the restart threshold UR, due to the parasitic superimposed AC noise to the useful signal. The time between these two passages trigger thresholds is less than the period Pi, which causes successive fast switching of the pulse signal Se and detection, instead of a single period "complete" P2 which would was obtained without switching parasite, two successive periods PΣa and PΣb "fragmented" niche signal Se. The shaft rotation speed measurement is then disturbed. two successive periods PΣa and PΣb "fragmented" niche signal Se. The shaft rotation speed measurement is then disturbed. two successive periods PΣa and PΣb "fragmented" niche signal Se. The shaft rotation speed measurement is then disturbed.

A known solution is to filter the unwanted noise in upstream electronics which produces the square wave signal. This is a low level solution implemented upstream of the engine ECU. It is common example of filtering by a low-pass filtering, the noise whose frequency is higher than the useful signal. However, the lowpass filtering is capable of shifting the measurement obtained. A phase-shifted extent may be troublesome for other functions that can perform the signal acquired at the terminals of the variable reluctance sensor, for example an engine balancing function, if the position of the shaft is taken as reference for the control other engine components. In addition, the low-pass filtering may reduce the amplitude of the signal. A thresholds restart and reset constant and reduced signal amplitude measuring rotational speed, the accuracy of the frequency measurement obtained is reduced. This problem is particularly relevant in low engine speed (eg engine start), where the ratio of peak to peak amplitude of a spurious noise alternative type on the amplitude of the useful signal may peak to peak be more important than for a higher engine speed.

Moreover, there are software solutions implemented at the application system of the digital motor interface (eg, a FADEC type system), realizing a post-processing of the acquired sinusoidal voltage to terminals a variable reluctance sensor. However, none of the existing solutions is specifically tailored to the problem of treatment of a noisy signal by alternating high frequency noise.

In particular, solutions of the prior art which are based on a simple averaging period measurements acquired in earlier times, to determine the period of the passage of the teeth of the phonic wheel at the current time, can be effective only on condition acquire a sufficient number of periods "complete" noiseless.

PRESENTATION GENERAL DE INVENTION

The present invention solves the problems mentioned above by a post-treatment period values ​​detected on the square wave signal, this processing can be implemented at the application software of the digital motor interface, to restore the "period useful "(that is to say the period of the useful signal which would be obtained without noise) from a set of time samples of the square wave signal stored in a ring buffer type memory. This post-treatment is based on the observation that the presence of parasites switches such as those highlighted in the signal Se in Figure 1 shown above, can be obtained by making the sum of several consecutive periods of the signal Se , a value very close to that of the period would be observed for the useful signal. The method of the invention allows to find the right are to be performed, by performing a plurality of sums of consecutive time samples and by comparing these sums to plausibility thresholds, calculated based on the period and speed measurements made previously .

The invention aims, in a first aspect, a speed measuring method of rotating a shaft of an aircraft engine, comprising steps of: acquiring an alternating signal of speed detection, the obtained terminals of a speed sensor comprising a phonic wheel rotated by the shaft,

transforming said AC signal into square wave signal, the switching wave signal to a low voltage level, respectively to a high level voltage in correspondence with the passage of the voltage of the AC signal below a restart threshold, respectively -Dessus a reset threshold,

recording in a ring buffer type memory, a plurality of samples of slot signal period, and comparing each of said samples to a lower limit of period and an upper limit of time, for determining said valid samples which value is between these two terminals;

if the number of valid samples is determined greater than a first threshold, determining from said samples a valid useful period of the pulse signal,

and at least on the condition that the number of valid samples is less than the first threshold, determination of the useful period of the signal by the following substeps:

calculating a plurality of sums of at least two samples,

calculating the average of a set comprising samples valid period and the amounts of the plurality of sums of at least two samples, the useful period of the pulse signal is taken equal to said average.

The method of the invention does not require to have samples of period corresponding to complete periods of the useful signal. Although it only has samples periods "fragmented" due to switching noise obtained on the niche signal, the method of the invention is able to find a period close to that which would be obtained for the useful signal.

The method of the invention has been developed starting from the observation that parasites switches such as those highlighted in the signal Se in Figure 1 shown above, can be obtained by making the sum of several consecutive periods of the signal Se, a value very close to that of the period would be observed for the useful signal.

Advantageously, this method may have the following additional features:

• the method comprises a further step of determining, among the plurality of sample sums calculated sums of valid samples which are between the lower and upper bounds period, the calculation of the useful period of the pulse signal being formed of averaging the group consisting samples valid period and are valid samples;

• if the sum of the number of valid samples and the number of valid samples are less than a second threshold, the effective period of the square wave signal is taken to be a useful period of niche signal obtained during a previous iteration of process ;

• the lower and upper bounds period are calculated from a consolidated period value, and a value of maximum variation period;

· The consolidated period is then taken as a useful period of slot obtained at an earlier time signal;

• the value of maximum change period is then obtained with a maximum gradient of Law speed where the maximum gradient at a given time is a function of speed;

• during a measurement initialization phase, phase during which the number of samples valid period does not exceed a third threshold or so during which period samples detected in the segment signal are dispersed them beyond d 'a

percentage predetermined threshold, a useful period of the pulse signal is estimated by a theoretical numerical model of the plan;

• the useful period of the pulse signal is determined during the initialization phase by a digital model of the plan based on thermodynamic parameters of the engine, preferably the value of the high pressure spool speed, atmospheric pressure, the compressor stator wedging .

According to a second aspect, the invention relates to a turbomachine engine calculator configured to implement the method steps described above.

The invention further relates in a third aspect a computer program product, for use in an application system for a turbomachine engine calculator, including code instructions allow the implementation of a method as above before.

PRESENTATION GENERALE DES FIGURES

Figure 1 has already been described above.

Other features, objects and advantages of the invention will become apparent from the following description, which is purely illustrative and not restrictive, accompanied by drawings additional annexes:

2 illustrates, on an example of the noisy signal in sensor output phonic wheel, false detections slot signal period.

Figure 3 shows a theoretical voltage signal resulting from the superposition of a useful signal and a noise alternating sinusoidal type, to illustrate the principle of the addition sample period.

4 shows schematically an embodiment of a processing algorithm sample period acquired in a circular memory to recalculate the review period.

5 shows one embodiment of an algorithm for determining likelihood interval within the algorithm of Figure 4.

Figure 6 illustrates the results obtained with the algorithm illustrated in Figure 4, on a shaft rotational speed signal from a measurement initialization phase and for a period comprising a restricted time range Z .

7 shows in reduced scale of more time, the Z range, the signals of Figure 6.

DETAILED DESCRIPTION OF EMBODIMENTS

Throughout the description below, it acquires an AC signal Sv to the terminals of a phonic wheel sensor, rotational speed detection of the shaft of the low-pressure (BP) of a two-spool turbojet. Downstream of the sensor and upstream of an Electronic Engine Controller (EEC) such as a FADEC3, the signal Sv is converted into a square wave signal Se. To do this, the signal Sv here is filtered by an RC filter and then clipped and filtered again, and finally by a crenelated Schmidt trigger UR restart threshold to 0 V and AU reset threshold to 0.232 V.

Periods of phonic wheel teeth to pass are measured in the segment signal Se, considering the elapsed times between two falling edges of the pulse signal. The noise-measuring wheel sensor here N1 of an engine. However, the same treatment modalities may also advantageously be used for measuring any other element revving the engine, such as for measuring the N2 speed.

However, as seen above in relation to Figure 1, the niche If signal can have switches that do not correspond to rising or falling edges to the useful part of the speed detection signal, but instead are caused by a voltage variation of a parasitic AC noise (non-useful part of the signal). This phenomenon is illustrated in Figure 2. In a first axis, a theoretical voltage Sv across a phonic wheel sensor is shown as a function of time t. Also shown on the same axis the useful part Sudu signal, theoretically obtained by subtracting the signal Sv the contribution of parasitic AC noise. On a second time axis, was superposed an associated window signal to signal Sv, and an associated slot signal theoretically useful signal Su, switching between a low level and a high level Ui U2. There switching these signals slots upwardly U2 level when the voltage of signal Sv passes beyond the AU threshold, and switching these signals slots down level U1 when the voltage of signal Sv passes below the threshold U R.

Is observed on the second axis, that the noisy signal Sv yield after aliasing a rising edge and a falling edge corresponding to the part 20 ', then a rising edge and a falling edge corresponding to the part 20', while the useful signal Su after aliasing gives one rising and one falling edge for the part 20. in addition, the rising and falling edges of the part 20 'are not

correspondence with the rising and falling edges of the part 20. Thus, the period detecting (e.g., the period between two successive fronts of slot signal descendants) is incorrect for the noisy signal Sv with respect to the period obtained for the useful signal Su. Now, it is the period of the wanted signal Su that is representative of useful N 1 diet.

In everything that follows, there is described a method for treating 1 0 applied to a plurality of values Ti, ..., T n period acquired in the square wave signal Se. 0 1 The method is implemented by an algorithm implemented in the application system of the engine control electronic control unit (EEC). It can reconstruct a consolidated period Pt from Ti values, ..., T n. As a reminder, a slot signal period is obtained here between two consecutive falling edges (U2 high level voltage to the low voltage level U1) of this segment signal. It is not necessary to assume in advance that such periods are "complete", that is to say associated with the theoretical useful signal detection system out unwanted noise, or "fragmented", that is, -dire calculated between signal interference If switching noise-related. 1 0 The method allows to reconstitute a t-th period Pt realizing a good approximation of the theoretical period of the wanted signal Su. Process 1 0 is here implemented digitally, the application program level, or application system, the EEC digital interface of an aircraft engine.

Figure 3 sheds light on a simplified theoretical example, the objectives and the principle of the method 1 0. a first axis, there is shown over time an example of voltage signal Sv, obtained by summing a useful signal Su theoretical sinusoidal low frequency, also shown in this first axis, and a high frequency noise signal. the restart threshold UR is also shown and the AU reset threshold of a Schmidt trigger used for the serrated signal Sv and the theoretical signal Su, these signals corresponding respectively

the slot signal Se, shown in parallel with a second axis, and the theoretical square wave signal SCT, shown in parallel to a third axis. It is noted that the period Pt which would be obtained for the SCT signal partially overlaps a first period Ti actually obtained for the signal Se, then completely overlaps two following periods T2 and T3, then partially overlaps a fourth unreferenced period in Figure 3.

However, by summing the consecutive periods T1, T2 and T3, there is obtained a very close period value of the value Pt, the latter constituting the appropriate period for the frequency measurement. One interpretation of this result is that both descendants Se marking fronts respectively the beginning of the period T1 and the end of the period T3 approximately correspond to consecutive falling edges that would be observed for the SCT signal, giving the period Pt. If is carried out an analogy between this theoretical example and a real case of detection measurement signal acquired at the terminals of a phonic wheel sensor, the actually observable signal across the sensor would be the signal Sv, giving the slot signal Se. An object of the method 10 is then of

slot is described below the steps of the method 10 of processing period values ​​of a signal acquired from a signal Sv rotational speed detecting taken across a phonic wheel sensor, the method 10 being here, as a reminder, implemented by the application software, or application system, the digital interface of the engine control electronic computer. The process steps 10 are hereinafter described in relation to Figure 4.

At step 100, the acquisition of the values Ti, ..., T n samples of slot signal period is performed in the following manner. A first counter detects the falling edges of a frequency sampling niche high signal with respect to that expected for the square wave signal. For example, the measuring board periods passage phonic wheel teeth may have a sampling period of the order of 0.25.10 "6s. A second counter detects the falling edges of the signal slot Se. Time values ​​elapsed between two consecutive falling edges of the pulse signal Sc can thus be stored in a circular memory circular buffer type B which can store a certain number of period measurements. Here it is preferable to implement the buffer B in the electronic circuit for measuring the period, instead of a digital implementation in the application system of the digital interface of the engine electronic control unit, since the calculation period of this calculator engine is typically of the order of 0.015 s and does not allow accurate measurement of tooth passing period of the phonic wheel (whose speed can reach 7000 rpm, at 60 teeth).n , and the dates of measurement of these samples, the application system.

The number, here denoted n, of samples taken above a minimum number of samples necessary for good measurement accuracy. The minimum number may correspond, for example, the number of teeth of the phonic wheel. By performing, in subsequent steps set out below of the method 10, an average of a number of consecutive samples exceeds the number of teeth of the phonic wheel, one gum possible discrepancies between measurements which would be linked to inhomogeneities geometry of the phonic wheel teeth. The expected frequency of the noise of which is to reduce the influence is also involved in determining the number of samples of the minimum

buffer B: more noise is high frequency compared to the wanted signal more periods "fragmented" may be necessary to reconstitute, on the principle described in relation to Figure 3, a useful passage period tooth of the phonic wheel .

In a step 1 10 samples Ti, ..., T n are sorted in order from oldest sample to the newest. This step 1 10 allows, in later stages, performing sums of consecutive periods in time. It is assumed, for convenience, the sorted samples retain their order of Ti to T n after sorting. The output of step 1 10, the application system thus has a vector consisting of n samples of consecutive periods.

After step 1 10, in this embodiment, there are two cases, depending on the value of a binary initialization signal that indicates whether the measurement period is in an initialized state or no.

Indeed, as will be described hereinafter, the method of reconstitution of consolidated period from step 120, a consolidated period value obtained in a previous calculation is not taken into input. These steps thus require at least satisfactory consolidated period value has been acquired.

In this embodiment, a test is performed on samples of period Ti, ..., T n sorted in the buffer B. Alternatively, it could also perform a test for initialization on samples of unsorted period . If testing a number of samples among samples Ti, ..., T n (eg, ten samples), or alternatively all the samples are sufficiently consistent with each other, that is to say, if their values are poorly dispersed with respect to a threshold percentage of predetermined dispersion D.

For example, if the initialization test requires that three samples period are somewhat scattered them, and for a value of threshold

PI-P2 dispersion D equal to 5%, it is checked whether the logic equation 100 * <5%

AND 100 *

P2 <5% is checked.

Alternatively, the initialization test would be to verify that a number of time samples from Ti, ..., T n included in a plausibility interval [Pmin, Pmax], where P min and P max values are example calculated using a theoretical model depending on the operational conditions of the engine, exceeds a predetermined threshold A3.

As the initialization test does not give a positive result for an iteration of the method 1 0, the value of the binary initialization signal remains at zero, and the measurement is considered uninitialized. During this initialization phase where the measurement is not initialized yet, instead of determining a consolidated period Pt from an earlier period consolidated PM, Pt is calculated by a theoretical numerical model Pth at a step 500D ( D mode of step 500 determination period Pt). The theoretical numerical model Pth may correspond to a theoretical model value N1, based on thermodynamic parameters of the engine, as the value of the high pressure spool speed, atmospheric pressure, the compressor stator wedging. Results obtained during a phase

If the initialization test is negative for an extended period (relative to a predetermined time), which implies that the signal is very noisy, it is possible to increase the value of the binary initialization signal 1 and initialize the algorithm with a consolidated period value Pt after the theoretical numerical model giving the Pth period.

It is possible to take into account in establishing the probability interval [min Pt, Pt max] the following iterations after initialization, an error in the determination of this theoretical period, due to uncertainty measurement of engine speed used in the theoretical model.

If the initialization test gives a positive result, the binary initialization signal is incremented to the value 1. The initialization signal then remains equal to 1 until the end of the measurement period passage phonic wheel teeth, and in subsequent iterations of the method, one thus operates consolidated period values ​​obtained previously. During the iteration of the method where the value of the binary initialization signal is incremented to 1, one can calculate an initial period consolidated Pto from the samples determined consistent. For example, Pto value can be taken as the arithmetic average of the samples considered consistent. The next step 120 may then be implemented in the next iteration,

Provided that the measurement period is initialized after step 1 10 samples sorting, the method 10 continues with a step 120 of calculating a lower bound and an upper bound of useful period. Alternatively, when the measurement period is initialized, step 120 may be implemented prior to step 1 10, or in parallel with step 1 10. The step 120 consists in establishing a probability interval that will then determine which of the sorted samples Ti, ..., T n , who may correspond to periods "complete", ie samples of valid period for frequency measurement phonic wheel teeth pass.

Description of the algorithm 120 for calculating the interval of likelihood for the t-th period Pt passage phonic wheel teeth delimited by a Pt min lower bound and an upper bound Pt max, is described in relation to the Figure 5.

At step 121 are acquired Pt-i values ​​and Vt-i, respectively the (tl) th value of phonic wheel teeth passage period previously determined using the algorithm 100, and ( tl) th corresponding rotation speed value. Periods are, for example in milliseconds, and the rotational speeds are, for example expressed in rotations per minute (rpm).

Next, at step 122, and with a predetermined law of maximum possible gradient rotational speed of the tone wheel (maximum gradient N1) is the maximum gradient calculating speed for the period Pt. The max i Gisi-law may be derived from engines or prior calculations trials. It is here expressed as a function of the shaft speed, but may depend on other parameters relating to operating conditions of the engine. The rotational speed gradient can be expressed in rpm per second.

At step 123, the GN-i max value (Vt - i) obtained is used to obtain the maximum variation speed in the period Pt is multiplied by the value of maximum gradient GN-i max (VT. -i) by the Pt-i period to obtain an estimate of this maximum variation. This maximum variation value is here in absolute value, and may lead to either an acceleration or a deceleration of the rotation of the tone wheel. This calculates the speed Vt max obtained by adding to Vt-i this maximum variation, and the speed Vt min obtained by subtracting from this maximum variation VM.

Alternatively, the maximum variation along the lines of a deceleration (minimum relative value) of the rotational speed may not be equal to the opposite of the maximum variation range in the direction of acceleration.

Finally, at step 124, we deduce from these values ​​maximum and minimum speed Vt max Vt min respectively the minimum values ​​min Pt and Pt max maximum period. is first obtained, from Vt Vt max and min, maximum frequency values ​​and minimum respectively ft max and min ft passage of the phonic wheel tooth. For example, if the rate is expressed in rpm, we first performs a division by 60 (for values ​​of speed in revolutions per second), then the obtained values ​​are multiplied after division by the number of teeth of the phonic wheel to obtain the maximum and minimum frequencies teeth passage. Finally, for the periods corresponding Pt Pt min and max are respectively considered the inverse of the frequency ft ft max and min,

Returning here to the diagram of Figure 4, at step 200, it is determined, from Ti, ..., T n and from the range of likelihood delimited by periods

Pt and Pt max min, samples valid Ti. . . , Tki whose value is between periods Pt Pt min and max, and the number ki of the valid samples. It is noted here Vi all samples period buffer B designated valid at this stage.

It is then considered a first threshold value Ai of the number of valid samples. Comparing the number of valid samples ki (not necessarily consecutive) than the threshold value Ai.

If k is greater than or equal to Ai, it is estimated that one has enough valid samples acquired for, by averaging (arithmetic mean by here, that could be an alternative geometric) the valid samples, a t- Pt th period for the passage of the teeth of the phonic wheel which is meaningful to the extent N1. In other words, among the samples in the buffer B, there is sufficient sample that is estimated they will come no parasites and commutations correspond to periods "complete". Thus if k is greater than or equal to Ai, the application system of the computer

motor control electronics determines a step 500A (procedure A of step 500) that Pt is equal to the valid samples ki arithmetic averaging period contained in B.

Choosing a value of Ai threshold which is sufficient to obtain consolidated periods have meaning for the frequency measurement, but not too high to avoid complicating exaggeratedly calculations and not to overload the memory of the application system. For example, it is interesting to consider a threshold Ai being equal to the number of teeth of the phonic wheel or a multiple of the number of teeth: for instance, if the teeth are compared to other teeth irregularities geometry which would warrant period inequalities in the useful signal (not related to the presence of noise), the algorithm would reduce the influence of these irregularities by averaging over a number of periods that cover the entire perimeter of the tone wheel.

If, however, k is less than Ai, we move to the next step 300 of calculating sample sums.

As seen above in relation to Figure 3, a principle of addition of several samples invalid period is to find sums of samples corresponding to periods "complete" the useful signal, by divesting the influence noise due to noise switching. In practice, are the search algorithm of consecutive samples, the sum value is within the previously determined likelihood interval at step 120, namely between Pt min

Thus, in the embodiment described here, the algorithm calculates, for an index i varying from 1 to n-1 (n being the number of buffer elements B) and for strictly superscript j to i and variant i + 1 to n, the sum denoted S samples of buffer B period from Ti to Tj, being recalled that the samples Ti, ..., T n have been sorted by date order in step October 1.

In other words, here you realize all possible sums of consecutive samples of the buffer B.

Next, at step 400, it is determined, among Si2, ..., Sm, S23, ..., S (ni) n and from the range delimited by the likelihood Pt and Pt max min periods, the are valid samples Si, ..., Sk2 whose value is between the Pt and Pt max min periods and the kΣ number of these are valid samples. It is noted here V2 all samples are considered valid. Alternatively, one could consider the upper and lower limits of values ​​of probability interval that would be different from the values ​​determined in step 200, or alternatively, it is possible to apply other criteria that membership interval [Ptmin, Emax] to select are valid, or can not do are valid selection.

Similarly to step 400, there are two cases according to the order number between the W2 and a second number of A2 threshold are valid periods relationship. A2 threshold is taken as Ai. Preferably, the A2 threshold is taken greater than Ai, for the calculation of the consolidated period Pt average are valid samples requires a larger number of are valid. This is justified because the fact of admitting the use of sample sums (and not only sample period) in the calculation of the consolidated period decreases the accuracy of the measurement. If the number of kΣ are valid samples is greater than or equal to the threshold A2, the algorithm determines in step 500B that the consolidated period Pt is equal to the mean, here arithmetic, the

This latter method of consolidated accounting period is very advantageous, because contrary to the calculation of the step 500A, it does not require to have a predetermined number of samples taken only valid in

buffer B. In other words, the calculation of step 500B can be achieved even if all the samples acquired in the period buffer B, or many of them are "fragmented" and result from unwanted commutations . The algorithm reconstructs from inconsistent periods taken alone (given the speed gradient law used), a useful period for calculating system, which not only provides better measurement accuracy in filtering out, but also allows the algorithm to operate even if no period gained on the embattled signal Se is "complete".

A condition for the calculation of consolidated period 500B step is however to succeed, from samples invalid content in the buffer B, rebuild sums of consecutive time samples that are determined valid, as included in the likelihood min interval between Pt and Pt max. Preferably the consolidated period determination algorithm provides, if this condition is not performed (that is to say, if the period of samples acquired in the buffer B are too noisy), assign all the same Pt value, so as to continue the implementation of the process 10 on subsequent iterations by having a consolidated period value Pt. Here, if there is not enough are valid sample period, Pt period is taken equal to the PM consolidated period determined in the previous iteration, at a step 500C. As a reminder, during the initialization phase where one does not have a first value of consolidated satisfactory Pt period, one can apply a theoretical model dependent on engine operating conditions to obtain Pt.

The results obtained by applying the process 10 of a voltage signal Sv acquired across a phonic wheel sensor are illustrated from an initialization of the measurement stage, in Figures 6 and 7. These figures show on a first axis Sv voltage over time, identifying restart thresholds UR and UA resetting of the Schmidt trigger producing the aliasing signal Sv. on a second axis, are shown superimposed consolidated Pt period obtained by applying the method 1 0, on the same time scale, and another Pti period value that would be obtained from the same signal Sv by applying a simple averaging algorithm sample periods. Finally, the third axis shown on the same time scale the value of the binary signal

PTI value corresponds to a period value obtained without distinguishing between valid samples and samples not disabled (on the criterion explained in relation to step 200), and without performing sums of consecutive time samples measured on a window signal obtained . from the signal Sv Pti the period calculation algorithm (not described in detail herein) includes the same initialization criterion than that which occurs in the initialization test to manage the signal Ui: calculating Pti can be initialized if a number of time samples are not dispersed them beyond the dispersion threshold D. Thus, in Figure 6, the time abscissa corresponding to a value 0 of the signal Ui (phase initialization) are not associated with a value of Pti. On the other hand,

Once the measurement correctly initialized and the signal Ui incremented to the value 1, the value Pti is obtained by making an arithmetic average of all the period samples in the buffer B. The value Pt, it is obtained according to method 1 0, so by not means that the valid samples if the algorithm determines a number greater than the first threshold A1, and conducting otherwise sums of consecutive samples as described above. On a

Most of the time range shown in FIG 6, for example the time range where the measurement is initialized and the above referenced range Z, all samples in the buffer B are determined as valid in step 200 of the method 10, and the period Pt corresponds to the arithmetical mean of all these samples. In this time range, Pt period is equal to the period Pti.

In contrast, the Z range, is observed on the first axis of Figure 6 that the maximum amplitude of the signal Sv is greater than for the values ​​of the signal Sv obtained previously. This is indicative of a here aliasing reciprocating type, which comes to be superimposed to the useful signal during the measurement of the voltage across the sensor phonic wheel. On this beach, many period samples among the samples acquired and stored in the circular buffer B are "fragmented" and correspond to spurious switching of the associated slot signal to signal Sv. 7 gives to see, on a time scale less, the curves of the first and second axes of Figure 6, the Z range of time value Pti is always obtained by the arithmetical mean of the same number of period of samples, although many of these samples are "fragmented" and correspond to much lower periods to periods of useful signal. Thus, when the signal Sv is very noisy, the value of Pti can reach 1, 0.10"5 s, while it is in the order of 2.0x10 " 4 s on the measurement time ranges where the signal Sv is little or no noisy. However, thanks to the consolidated period algorithm that finds consistent values of useful signal period from the time "fragmented" recorded in the ring buffer B, the Pt value is almost not decreased on the beach Z, not descending below 1 7.10 " 4 s. the Pt measurement period is thus hardly affected by the presence of parasitic AC noise on the beach Z.

10 The process described above thus allows, without spot among the switching signal niche ones that fit a parasite noise and those that correspond to the useful measurement signal N1, greatly reduce the influence of parasitic noise AC type over the measurement period of passage phonic wheel teeth. The speed measurement sensor wheel and thus the extent of N1, are thereby made reliable. This solution, simple implementation in the application system of an engine digital interface, is much more effective than a simple averaging of a constant number of samples period, as seen above. Moreover, this solution is advantageous with respect to known solutions of applying a type of low-pass filtering,

WE CLAIMS

1. A method for rotational speed measurement of a shaft of an aircraft engine from a square wave signal (Se), said slot signal being obtained from an alternating signal (Sv) of rotation speed detecting acquired across a speed sensor comprising a phonic wheel rotated by the shaft,

said window signal (Se) switching to a low level voltage (Ui), respectively to a high level voltage (U2), in correspondence with the passage of the voltage of the alternating signal (Sv) below a threshold restarting (UR), respectively above a reset threshold (UA),

the method comprising the steps of:

- recording, in a circular buffer type memory (B), a plurality of samples (T, ..., T n ) slot signal period

(I know),

- comparing each of said samples to a lower terminal period (Pt min) and an upper terminal period (Pt max) for determining said valid samples whose value is between the two terminals;

- if the number of valid samples is determined greater than a first threshold (Ai), determining from said samples a valid useful period (Pt) of the pulse signal (Se),

the method being characterized in that at least on the condition that the number of valid samples is less than the first threshold (Ai), the determination of the useful period (Pt) of the pulse signal (Se) implements the following steps :

- calculating a plurality of sums of at least two samples

(Sl2, .. . ,Sl n,S23, .. . ,S(n-1)n),

- calculating the average of a set comprising samples valid period and the amounts of the plurality of sums of the

least two samples, the useful period (Pt) of the pulse signal is taken equal to said average.

2. A method for measuring velocity according to claim 1, comprising a further step of determining, among the plurality of sample sums calculated sums of valid samples which are between the lower and upper bounds period (Pt min, Pt max), the calculation of the effective period (Pt) of the pulse signal being formed by calculating the average of the group consisting of valid samples and the amounts determined valid samples.

3. A method for measuring velocity according to one of claims 1 or 2, wherein if the sum of the number of valid samples and the number of valid samples are less than a second threshold (A2), the effective period (Pt) of the pulse signal is taken as a useful period of the pulse signal (Pt-1) obtained during a previous iteration of the method.

4. A method for measuring velocity according to one of claims 1 to 3, wherein the lower and upper limits (min Pt, Pt max) period are calculated from a consolidated period value, and a value maximum variation period.

5. A measuring method according to claim 4, wherein the consolidated period is taken to be a useful period of slot obtained at an earlier time signal (PM).

6. A measuring method according to one of Claims 4 or 5, wherein the maximum change period value is obtained using a maximum gradient law rotational speed (NG-i max), where the maximum gradient at a given instant is a function of the rotational speed.

7. The measurement method according to one of claims 1 to 6, wherein during a measurement initialization phase, phase during which the number of samples valid period does not exceed a third threshold (A3), a period review of the square wave signal is estimated by a theoretical numerical model period (Pth).

8. Measuring Method according to one of claims 1 to 7, wherein during a measurement initialization phase, during which period samples detected in the segment signal (Se) are dispersed between them beyond of a predetermined threshold (D), a useful period of the pulse signal is estimated by a theoretical numerical model period (Pth).

9. The measurement method according to claim 7 or 8, wherein the useful period of the pulse signal is determined during the initialization phase by a digital model based on thermodynamic parameters of the engine, preferably taken from the value of the high body slimming pressure or the value of speed of another rotating member of the engine, atmospheric pressure, the compressor stator wedging.

10. aircraft engine calculator configured to implement the steps of, taking into a square wave signal (Se):

recording, in a circular buffer type memory (B), a plurality of samples (T, ..., T n ) period of the pulse signal (Se), comparing each of said samples to a lower limit of period (Ptmin) and an upper terminal period (Emax) for determining said valid samples whose value is between these two terminals,

if the number of valid samples is determined greater than a first threshold (Ai), determining from said samples a valid useful period (Pt) of the pulse signal (Se),

and if the number of valid samples is less than the first threshold (Ai), determination of the useful period (Pt) of the pulse signal (Se) as follows:

calculating a plurality of sums of at least two samples

(Sl2, . . . , Sl n,S23, . . . , S(n-1 )n),

calculating the average of a set comprising samples valid period and the amounts of the plurality of sums of at least two samples, the useful period (Pt) of the pulse signal is taken equal to said average.

January 1. A computer program product, for use in an application system for a turbomachine engine calculator, including code instructions allow the implementation of a method according to any one of claims 1 to 9.