Method and system for leak detection in a hydraulic circuit of an aircraft

Abstract

The invention relates to a method for detecting a slow leak in a hydraulic system of an aircraft, this system comprising a first hydraulic block (10), with redundancy due to a second hydraulic block (12), each block being powered by an accumulator (20, 22), this method comprising at least: - in-flight detection of the pressures upstream of each of the hydraulic blocks, during one or more time windows in which these blocks are not under stress; - then the following steps, implemented by a computer: * calculating a gradient of the pressure differences between the two hydraulic blocks; * detecting or identifying whether this gradient is greater than a first predefined threshold or whether a variation in this gradient on the last N (N > 1) flights is greater than a second predefined threshold.

Description

[0001] Description

[0002] Title of the invention: METHOD AND SYSTEM FOR DETECTING LEAKS IN AN AIRCRAFT HYDRAULIC CIRCUIT

[0003] Technical field and prior art

[0004] The invention relates to the field of leak detection in a redundant hydraulic circuit or block of an aircraft. In particular, it allows for the assessment of the condition and maintenance needs of such a circuit.

[0005] In particular it applies to the secondary braking circuit (used if the main circuit is malfunctioning) of an aircraft (for example an airplane or a helicopter), especially to the accumulator-powered hydraulic blocks of such a circuit.

[0006] Currently, no method or device exists to detect slow, progressive leaks in such a circuit. These leaks occur slowly and are observed over a long period, for example, several months or years.

[0007] However, a leak, for example from an accumulator, is critical both from a safety point of view (since there is then a lack of braking on the wheels concerned in the event of use of the secondary circuit) and from an operational point of view (the aircraft cannot take off once the leak is detected).

[0008] Indeed, during operation, a fault in such a circuit can be signaled by a fault message based on a low accumulator pressure threshold. This type of message prevents the aircraft from taking off.

[0009] However, a "slow leak" or "progressive leak" type fault cannot be identified before such a message is triggered. In other words, it cannot be anticipated and, moreover, does not always allow the source of the fault to be determined.

[0010] Furthermore, the signals and logic currently used for monitoring the operation of components in this type of circuit are not reliable enough to accurately track leakage, due to a strong dependence on exogenous environmental variables, particularly temperature, and on demands from internal components such as park, accumulator filling and braking demands.

[0011] The challenge, therefore, is to find a new method for detecting slow or progressive leaks of the type described above in a redundant hydraulic circuit or block of an aircraft. Such a method, and / or the development of an indicator for this type of leak, would make it possible to prevent more critical leaks that could cause operational interruptions.

[0012] In particular, we seek to find a process that allows us to overcome the influences, on such a circuit, of exogenous environmental variables, for example temperature, but also the influences resulting from the application of certain commands to the hydraulic circuit concerned, these commands having the effect of stressing or activating this circuit.

[0013] Description of the invention

[0014] The invention relates first to a method for detecting a slow leak in a hydraulic system of an aircraft (for example, an airplane), this system comprising a first hydraulic block redundant with a second hydraulic block, each block being supplied by an accumulator, this method comprising at least:

[0015] - in-flight detection of the pressures of each of the hydraulic blocks, for example upstream of each of these blocks, this detection being carried out during one or more time window(s) when these blocks are not under stress;

[0016] - then the following steps, implemented by computer:

[0017] * calculate a slope of the pressure differences, or a difference between the slopes of these pressures, between the 2 hydraulic blocks;

[0018] * detect or identify if this difference or slope is greater than a first predefined threshold and / or if this slope or a variation of this slope or difference over the last N (N> 1) flights is greater than a second predefined threshold.

[0019] N (N>1) is a configuration parameter. According to the invention, data from flight recordings are used to assess in a more precise and reliable way the presence of slow leaks in an accumulator-fed, instrumented and redundant hydraulic block.

[0020] For each flight, pressure data from redundant hydraulic blocks is acquired during at least part of the flight, for example upstream of these blocks, using pressure sensors, during one or more time windows during which the hydraulic blocks are not under stress.

[0021] A statistical processing of the pressure difference between the 2 parts of the circuit makes it possible, for example, to generate a leak indicator and to identify any exceedance of a static and / or dynamic threshold, the latter making it possible to characterize a slow evolution of this pressure difference.

[0022] The evolution of the pressure difference or of a leak indicator can be compared to one or more threshold values.

[0023] Such a treatment is more refined and reliable than known techniques because:

[0024] - the use of the pressure difference makes it possible to neutralize effects due for example to exogenous environmental variables, in particular temperature; in other words, the invention makes it possible to decontextualize the measurement of the leak in order to be more precise and thus make it possible to detect slow leaks, for example from a fluid accumulator;

[0025] - the data collected comes from phases during which the valves and various components of the hydraulic blocks concerned are not under stress.

[0026] This allows us to neutralize effects that could mask slow changes due to pressure leaks.

[0027] Therefore, a method according to the invention makes it possible to compare pressure level trends between two redundant hydraulic circuits during phases where the circuits are not under stress.

[0028] The statistical processing may involve calculating a confidence interval, for example at 97.5%, for example also based on, or from, a linear regression of the gaps or pressure differences between the 2 circuits, a confidence interval from which an indicator can be calculated.

[0029] Such an indicator, or consolidated indicator, makes it possible to reflect a significant difference between the gaps or differences in pressure or between the slopes of these gaps or differences.

[0030] Preferably, pressure measurements are taken only when the blocks are not under stress.

[0031] A method according to the invention may further include triggering an alert signal if at least one slope of the pressure differences or a difference between the slopes of these pressures is greater than the first predefined threshold or if the variation of this difference or this slope over the last N flights is greater than the second predefined threshold.

[0032] The hydraulic system or each hydraulic block may include a portion or at least a portion of the aircraft's secondary braking circuit.

[0033] The invention also relates to a data processing system, or respectively a computer-readable recording medium for implementing data processing, said data being measured in flight from an aircraft (for example, an airplane) and comprising pressure data from a first hydraulic block and a second hydraulic block with which the first hydraulic block is redundant, this system comprising means specifically programmed or configured for, or respectively this medium comprising instructions for, when read by a computer:

[0034] - calculate a slope of the pressure differences, or a difference between the slopes of these pressures, between the two hydraulic blocks;

[0035] - calculate or detect if this slope or difference is greater than a first predefined threshold or if this slope or a variation of this slope or difference over the last N flights is greater than the second predefined threshold.

[0036] The invention also relates to a data processing system, or respectively a computer-readable recording medium for implementing data processing, said data being measured in flight from an aircraft (for example, an airplane), comprising pressure data from a first hydraulic block and a second hydraulic block with which the first hydraulic block is redundant, each block being able to be powered by an accumulator, this pressure data being measured during one or more time window(s) when these blocks are not under load, this system comprising means specifically programmed or configured for, or respectively this medium comprising instructions for, when read by a computer:

[0037] - calculate a slope of the pressure differences, or a difference between the slopes of these pressures, between the 2 hydraulic blocks;

[0038] - calculate or detect if this slope or difference is greater than a first predefined threshold or if this slope or a variation of this slope or difference over the last N flights is greater than a second predefined threshold.

[0039] N (N>1) is a configuration parameter.

[0040] According to embodiments of the invention, said means can be specifically programmed or configured to, or respectively said recording medium can include instructions to, when read by a computer:

[0041] - calculate a confidence interval, for example at 97.5%, for example still based on a linear regression or on the slope of a linear regression of pressure differences;

[0042] - and / or calculate a consolidated indicator that reflects a difference or significant difference between pressure slopes or pressure deviations;

[0043] - and / or trigger an alert signal if at least one pressure slope or pressure deviation or pressure deviation is greater than the first predefined threshold or if the variation of this slope or the slope of pressure deviations over the last N flights is greater than the second predefined threshold.

[0044] In a method or device according to the invention:

[0045] - each hydraulic unit may include part or at least part of the aircraft's secondary braking circuit; - and / or each hydraulic unit may be arranged in parallel with main supply units, each main supply unit being equipped with a main pressure generator;

[0046] - and / or at the output of each of the hydraulic blocks there is a set, or plate, of valve(s) which will allow the fluid from the corresponding hydraulic block to be distributed to wheels.

[0047] The invention therefore makes it possible to detect slow or progressive leaks early enough and to anticipate situations for which the only solution would be to immobilize the aircraft, due to the failure to detect the leaks early enough.

[0048] The invention also relates to a predictive maintenance monitoring system, comprising a data processing system according to the invention and / or respectively a recording medium according to the invention, readable by computer to implement data processing according to the invention.

[0049] Brief description of the drawings

[0050] [Fig. 1] represents an example of an aircraft hydraulic circuit to which the invention can be applied;

[0051] [Fig. 2] represent curves of evolution over time of the pressure of two hydraulic blocks and the identification of an acquisition window;

[0052] Figure 3 illustrates a method for calculating or estimating the slope of pressure differences between two hydraulic blocks;

[0053] Figure 4 represents an example of tracking, from one flight to another, of the consolidated indicator, and a comparison with respect to thresholds, respectively static and dynamic.

[0054] Figure 5 represents a computer device capable of implementing a process according to the invention.

[0055] Detailed description of embodiments of the invention

[0056] Figure 1 represents an example of an aircraft hydraulic system (for example, an airplane) to which the invention can be applied. This system comprises two hydraulic blocks 10, 12 (in fact: a redundant hydraulic block), each supplied by an accumulator 20, 22 (containing pressurized oil) and equipped with a pressure sensor 30, 32. One or the other of these hydraulic blocks 10, 12 may have a leak, in particular a slow leak which must be detected.

[0057] These hydraulic blocks 10, 12 are arranged in parallel with main power supply blocks 14, 16, each equipped with a main pressure generator 24, 26.

[0058] At the outlet of each of the hydraulic blocks 10, 12 there is an assembly, or plate, 40, 42 of valve(s) which will allow the fluid from the block corresponding to wheels 50i-50s to be distributed.

[0059] Whether through the main system or the secondary system, the hydraulic flow reaches each wheel of the aircraft (for example, an airplane) to brake it.

[0060] Sensors 30, 32 will allow the acquisition or measurement of accumulator pressures during time windows in which these blocks are not under stress, and preferably only when the blocks are not under stress.

[0061] Pressure sensors 30 and 32 measure the pressure in the accumulators, which supply pressure to the hydraulic blocks 10 and 12. It is at these pressures that the impact of a leak will be most noticeable when the hydraulic system is not under load. The hydraulic system is not under load when there is no demand from the parking garage, no demand to refill the accumulators, and no braking.

[0062] For example, Figure 2 shows the evolution of the pressures of the two accumulators 20 and 22 (curve I in thick line, curve II in thin line), pressures acquired by sensors 30 and 32, during an aircraft takeoff phase, when these blocks are under load. The x-axis is graduated in time, in this example from approximately 10:30 to approximately 10:45, and the y-axis is in psi. This takeoff phase is followed by a window 60 during which these blocks and the accumulators are not under load; that is, no de-loading command is sent to them by the aircraft's control system, and during which these blocks do not supply fluid.

[0063] Within this window 60, the pressure of each of these hydraulic blocks can be measured, for example at a frequency of a few Hz (e.g., between 4 and 8 Hz), or continuously. This allows us to obtain the time evolution of the pressure for each of the hydraulic blocks.

[0064] The measured data is collected in a data storage device located on board the aircraft.

[0065] During a flight, several 60 windows (which may be narrow) can be identified and exploited within the framework of the present invention.

[0066] There is no minimum duration requirement for the 60-day windows. In fact, this aspect can be managed statistically using a confidence interval, as explained below. Calculating an indicator using this confidence interval allows for the elimination or filtering of excessively narrow 60-day windows.

[0067] The 60 time windows during which the blocks and accumulators are not used are not of fixed duration. They can sometimes be long, making the estimation of the slope difference by linear regression and the calculation of the confidence interval statistically representative, and sometimes short, in which case these calculations may be less statistically representative. In all cases, whether long or short, these 60 windows never cover the entire flight, since there is always at least one use of the blocks and accumulators during a flight.

[0068] Rather than separately measuring the pressure slopes of the two accumulators 20 and 22 (whose pressures are captured respectively by sensors 32 and 30), we study the difference in these slopes between the two circuits. Indeed, exogenous environmental variables (particularly temperature) have a significant impact on pressure variations. Since the two accumulators operate in a similar environment, comparing the slopes helps to limit the effect of exogenous variables on these slopes. For each flight, it is therefore possible to acquire or measure the pressures upstream of the hydraulic blocks 10 and 12 using the pressure sensors 30 and 32 during these time windows 60.

[0069] One of these windows 60 is shown in Figure 3, which includes part of curves I and II from Figure 2. A linear regression allows us to identify the evolution of the pressure of one of the blocks and the pressure of the other block.

[0070] The pressure slope is calculated using linear regression. Since the window sizes can be small, the effect of random noise on the regression coefficient can be significant, leading to insignificant values. To mitigate this problem, a confidence interval is constructed for the difference in slopes.

[0071]

[0072] > A- in the following manner.

[0073] We assume that the model is defined, for example, by:

[0074]

[0075] the pressures of the 2 circuits (pressures captured respectively by sensors 30 and 32), and where

[0076]

[0077] are centered, independent, and identically distributed Gaussian variables (with potentially different variances for the two circuits). We then have:

[0078]

[0079] AP = Pg — J? y

[0080] Or e s is a Gaussian noise, independent and identically distributed. We then have the pivotal random variable:

[0081]

[0082] where ■ ; 'Î denotes a Student's t-variable ¥ degrees of freedom,

[0083]

[0084] is the estimated standard deviation of A P. The 97.5% confidence interval for:

[0085]

[0086] where feys is the 97.5% quantile of the Student's t-distribution with N-2 degrees of freedom. In practice, one can, for example, use the "statsmodel" package in Python via its statsmodel.api.ols function to calculate the slope of the linear regression on gt-yt as well as the associated confidence interval.

[0087] The quantile of the Student's law is, for example, calculated with scipy via its scipy.stats.t.isf function.

[0088] Specifically, for a 60-pixel window, we can calculate, using the statsmodel.api.OLS function from the Python statsmodel package:

[0089] 1) ÂP as the linear regression slope on St J't;

[0090] 2) -' p as the standard deviation of Ap;

[0091] 3) ÇB.975 as the quantile of the Student's t-distribution (which is calculated by the scipy.stats.t.isf function of Python scipy).

[0092] With this, we obtain the confidence interval (CI) on

[0093]

[0094] >

[0095]

[0096] Within each confidence interval, to obtain a consolidated indicator that reflects a significant difference between the slopes (as explained above, the window sizes can be small), we can choose, as the Ind indicator, the true value with the minimum absolute value within that interval. Specifically, when the confidence interval contains the value 0, the difference between the slopes is not statistically significant, and the indicator is 0.

[0097] For example :

[0098] ■ If IC = [-0.1, 0.3], then the indicator Ind is 0;

[0099] ■ If IC = [-0.2, -0.15], then the indicator Ind is -0.15 (real value of minimum absolute value over the interval);

[0100] ■ If IC = [0.2, 0.3], then the Ind indicator is 0.2 (again, the actual value has the minimum absolute value over the interval). Figure 4 shows the evolution of a system as described above over several flights, for a period that can be quite long, here 2.5 years. There is no minimum duration for the period covered: it depends on the intensity of the leak. Each point in Figure 4 is associated with, or calculated from, a window. The Ind indicator (for example, as explained above) shown is taken from the confidence interval of the slope

[0101]

[0102] Therefore, this is a statistical estimate of this slope. Figure 4 represents the evolution of the Ind indicator, and thus the evolution of the estimate of this slope. We can see in this figure 4 that a point M exceeds a predefined statistical threshold and that a dynamic evolution begins from a certain date.

[0103] The point beyond the static threshold may trigger a maintenance-type intervention.

[0104] The evolution of the dynamic threshold indicates a general trend in the system, a trend that reflects the evolution of a slow leak. The stronger this leak, the more peaks are observed, and the more quickly the dynamic threshold is crossed.

[0105] Therefore, data processing, or statistical processing, according to the invention can implement:

[0106] - the determination of a confidence interval (CI);

[0107] - the calculation of an indicator (or consolidated indicator) Ind over this confidence interval;

[0108] - the identification of the evolution of this indicator over one or more flights of the aircraft;

[0109] - possibly the comparison of the indicator, and / or its evolution, during this flight(s) with a static threshold and / or a dynamic threshold (which or which may have been predetermined by a user).

[0110] A warning or alert signal may be generated if:

[0111] - the consolidated indicator is above a static threshold SEUIL_STATIC which can be configured;

[0112] - and / or the trend tracking (the slope) over the last N flights is greater than a dynamic threshold SEUIL_DYNAMIQ.UE which can be configured.

[0113] N is a configuration parameter.

[0114] The value of N can be determined by the indicator designer. The aim is to calculate the indicator's slope, particularly through linear regression on the last N flights. Choosing an N that is too large risks masking or smoothing out a potential trend change, which could lead to insufficient dynamic detection of a slow leak. Conversely, choosing an N that is too small would make dynamic detection highly sensitive to noise and seasonal fluctuations, which could cause premature or even incorrect dynamic detection of a slow leak.

[0115] The static threshold and / or the dynamic threshold can be predetermined, or selected by a user.

[0116] The data from sensors 30, 32 can be stored in a memory area, for example of a computer or calculator which may be located on board the aircraft; then the data is transmitted to the ground, to a computer 120 (figure 5) of an analysis or maintenance center for processing and statistical analysis according to the invention.

[0117] The data measured using the pressure sensors 30, 32 can therefore be transmitted to this computer 120 (or calculator or microcomputer). This includes, for example (Figure 5), 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 memory areas, including data for implementing a process according to the present invention. These areas form a computer-readable storage medium for implementing a statistical data processing method according to the invention, or they contain instructions that, when read by a computer, enable the implementation of the statistical data processing portion of a process according to the invention. Other types of storage media may include a USB key or any other type of data storage medium used in computing, which, when read by a computer, enables the implementation of a process according to the invention. The computer or calculator is therefore programmed or configured to implement statistical data processing of a process according to the invention. Means 59 will manage the flow of input data (particularly from sensors 30, 32) and, optionally, output data.These means of receiving and / or transmitting wireless data may also be provided for.

[0118] The data can be analyzed and processed by statistical data processing of a method according to the invention. This analysis and / or processing can be carried out by this computer 120 or calculator, for example, in a maintenance center to which the data is transmitted.

[0119] A method or device according to the invention can implement a leak indicator calculated according to the time evolution of the pressure upstream of a hydraulic block.

[0120] This indicator can be compared to a leak detection threshold, then tracked in trend for comparison to a leak progression detection threshold in order to make predictive maintenance recommendations.

[0121] The invention therefore makes it possible to detect progressive leaks and to anticipate situations for which the only solution would be to immobilize the aircraft if, according to the invention, it were not implemented.

[0122] The invention has been implemented in a predictive maintenance monitoring system applied to a fleet of A350 (-900 and -1000) aircraft. Several cases of leaks were detected, leading to recommendations for equipment removal.

Claims

DEMANDS 1. A method for detecting a slow leak in an aircraft hydraulic system, this system comprising a first hydraulic block (10) redundant with a second hydraulic block (12), each block being supplied by an accumulator (20, 22), this method comprising at least: - detection in flight of the pressures of each of the hydraulic blocks (10, 12), during one or more time window(s) when these blocks are not under stress; - then the following steps, implemented by computer (120): * calculate a slope of the pressure differences between the 2 hydraulic blocks; * detect or identify if this slope is greater than a first predefined threshold and / or if a variation of this slope over the last N (N> 1) flights is greater than a second predefined threshold.

2. A method according to claim 1, comprising calculating a confidence interval based on a linear regression of pressure deviations.

3. Method according to claim 2, comprising the calculation of a consolidated indicator (Ind) which reflects a difference between the slopes of the pressure deviations.

4. A method according to any one of claims 1 to 3, further comprising triggering an alert signal if at least one slope of the pressure differences is greater than the first predefined threshold or if the variation of this slope over the last N flights is greater than the second predefined threshold.

5. A method according to any one of claims 1 to 4, wherein each hydraulic block (10, 12) comprises at least a portion of the aircraft's secondary braking circuit.

6. A system (120) for processing in-flight data measured from an aircraft, referred to as flight data, comprising pressure data from a first hydraulic block (10) and a second hydraulic block (12), with which the first hydraulic block (12) is redundant, each hydraulic block (10, 12) being able to be powered by an accumulator (20, 22), said data being measured during one or more time windows when these blocks are not in use, this system comprising means (12) specifically programmed or configured to: - calculate a slope of the pressure differences between the 2 hydraulic blocks; - calculate or detect if this slope is greater than a first predefined threshold and / or if a variation of this slope over the last N (N> 1) flights is greater than a second predefined threshold.

7. System according to claim 6, said means (12) being specifically programmed or configured to calculate a confidence interval (CI) based on a linear regression of pressure deviations.

8. System according to claim 7, said means (12) being specifically programmed or configured to calculate a consolidated indicator (Ind) which reflects a difference between the slopes of the pressure deviations.

9. System according to any one of claims 6 to 8, said means (12) being specifically programmed or configured to trigger an alert signal if at least one slope of the pressure deviations is greater than the first predefined threshold or if the variation of this slope over the last N flights is greater than a predefined threshold.

10. System according to any one of claims 6 to 9, each hydraulic block comprising at least a portion of the aircraft's secondary braking circuit.

Images