Method and system for determining one or more faults in a rotating machine of an aircraft
By measuring and analyzing the acoustic, electrical, and vibration signal spectra of rotating machines, a fault spectrum library is formed, which solves the problem of difficulty in identifying complex faults in existing technologies and enables early and accurate identification and diagnosis of faults.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SAFRAN VENTILATION SYST
- Filing Date
- 2021-12-09
- Publication Date
- 2026-07-14
AI Technical Summary
Existing technologies struggle to accurately identify complex or distributed faults in rotating machines, especially those with multiple manifestations, leading to inaccurate and imprecise fault diagnosis.
By measuring the acoustic, electrical, and vibration signals of rotating machines and converting them into a spectrum, a fault spectrum library is formed using weighting coefficients. The frequencies of interest are then precisely determined, and multiple spectra are compared to identify faults.
It enables early and reliable identification of faults in rotating machines, accurately distinguishes between single and complex faults, and improves the accuracy and reliability of fault diagnosis.
Smart Images
Figure CN116601475B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of fault monitoring of rotating machinery in aircraft. The invention is particularly applicable to rotating machinery that includes at least one electrical device, especially an electric motor. Background Technology
[0002] Rotating machines are known to be equipped with various sensors to identify known faults in the machine. For example, vibration sensors can be used to detect imbalances, acoustic sensors can be used to detect rolling bearing faults, or electrical sensors can be used to detect electrical faults in the generator of the rotating machine.
[0003] The use of such sensors is robust for identifying “simple” faults, i.e., those with a single cause and a single effect. In reality, faults in rotating machinery can affect many components of the rotating machinery to a lesser extent (several effects). These “complex” or “distributed” faults are more difficult to identify because they are harder to detect by a single sensor. The same applies to combinations of “pure” or “complex” faults (multiple causes). One object of the present invention is to enable the detection of all faults that may occur simultaneously or may have multiple manifestations, i.e., vibrational, auditory, or electrical manifestations.
[0004] For example, frequency analysis can reveal a periodic fault with a first frequency and another electrical fault with a second frequency close to the first frequency, which may correspond to an imbalance in the shaft of a rotating machine. It can be difficult for operators to determine whether they are facing a single fault or two distinct faults.
[0005] In practice, to process sensor measurements, it is known to determine the spectrum, particularly through Fourier transform, in order to check for the presence of possible features at the machine's characteristic frequencies. Given the spectral dispersion at high frequencies, it is difficult to accurately isolate features from faults; this inaccuracy leads to inaccurate fault diagnosis. This determination is difficult with strong signals (very obvious faults), but even more difficult with latent signals (slightly obvious faults), which are then confused with broadband noise or very close spectral lines. Traditionally, the presence / severity of a fault is not precisely determined but estimated over a frequency range, making it impossible to separately determine two faults with overlapping frequency ranges.
[0006] Early and reliable identification of faults in rotating machines is required, whether they are pure faults, complex faults, independent faults, or combinations of faults.
[0007] In the prior art, methods for determining faults in rotating machines are known according to patent applications WO2018 / 222341A1, EP2693176A1 and FR2952177A1. Summary of the Invention
[0008] This invention relates to a method for determining one or more faults in the rotating machinery of an aircraft, the method comprising:
[0009] The step of measuring at least one acoustic signal and at least one vibration signal during the same operating phase of a rotating machine;
[0010] The steps to convert each acoustic signal and each vibration signal into an acoustic spectrum and a vibration spectrum respectively to obtain a spectrum set;
[0011] The step of determining the set of frequencies of interest within the spectrum set, where each frequency of interest belongs to a predetermined frequency range;
[0012] The steps for forming a fault spectrum library include each fault spectrum comprising at least one frequency line, each frequency line being obtained by a linear combination of the frequency of interest and a predetermined weighting coefficient.
[0013] The steps to compare each spectrum with each fault spectrum to obtain a score, and
[0014] The steps to determine the faults of rotating machines are based on the analysis of the obtained scores.
[0015] According to the present invention, the set of frequencies of interest is determined very precisely based on sound and vibration signals. Due to the different properties of the signals, it is advantageous to avoid bias in a single measurement. An accuracy of 0.1 Hz is achieved, making it possible to accurately determine the frequencies of interest, rather than within a frequency range as in the prior art. Therefore, a fault spectrum library can be formed through linear combinations of relevant frequencies of interest. In fact, due to the accuracy of the frequencies of interest, the fault spectrum can include frequencies (lines) that are very close to each other and characteristic of a specific fault. Since each target frequency is very precise, harmonics (multiples of the target frequency) can also be accurately determined.
[0016] Furthermore, comparing each spectrum with the spectrum of each fault provides an overall understanding of the importance of each fault. The tonal noise of the spectrum can therefore be easily processed to identify faults.
[0017] The heterogeneity of faults listed in the fault spectrum library takes into account high reliability and robustness. The method according to the invention makes it possible to identify strong faults as well as hidden or potential faults. This allows operators to determine the causes of failures in rotating machinery, assuming they have information on both the acoustic and vibrational components of each fault.
[0018] Preferably, each frequency of interest corresponds to the fundamental frequency.
[0019] Preferably, during the step of comparing the spectrum with the determined fault spectrum to obtain a score, a basic score in the spectrum is calculated for each weighted frequency of interest in the determined fault spectrum, and a score is determined from the set of basic scores.
[0020] Because of the accuracy in determining each frequency of interest, it is possible to check whether each frequency line of the fault spectrum exists in the spectrum. Such a comparison is impossible in the prior art because the frequencies are not accurately known. According to the invention, if the fault spectrum includes two frequency lines close to a few Hz, two independent fundamental fractions are measured, which improves the correlation for determining the fault.
[0021] Preferably, the aircraft rotating machine includes at least one electrical device, and the method includes:
[0022] The steps of measuring at least one acoustic signal, at least one electrical signal, and at least one vibration signal during a single operating phase of a rotating machine, and
[0023] The steps of converting each acoustic signal, each electrical signal, and each vibration signal into an acoustic spectrum, an electrical spectrum, and a vibration spectrum, respectively, to obtain a spectrum set.
[0024] The combined use of electrical signals with vibration and acoustic signals is relevant because it allows for the visualization of distributed or combined faults. Obtaining more of the spectrum improves the accuracy of identifying frequencies of interest and helps pinpoint different types of faults.
[0025] Preferably, the electrical signal is measured in an electrical device, and more preferably in a motor.
[0026] Preferably, during the step of determining the set of frequencies of interest, at least two spectra are functions of at least one frequency of interest:
[0027] The first estimate of the frequencies of interest is determined in the first spectrum.
[0028] A second estimate of the frequency of interest is determined in the second spectrum, and
[0029] The frequency of attention should be determined based at least on the first estimate and the second estimate of the frequency of attention.
[0030] According to the present invention, at least two spectra are used to obtain independent estimates of the frequency of interest. This reduces the inaccuracy of its determination, particularly by eliminating biases or traps in interpreting the frequency of interest. Preferably, the two spectra have different properties.
[0031] Preferably, the step of determining the set of frequencies of interest is performed from a link library that associates at least each frequency of interest with one or more spectra. This link library makes it possible to determine in which spectrum the frequencies of interest should be searched. This linking basis is preferably obtained through analytical calculations or feedback.
[0032] Preferably, the linked library associates at least each frequency of interest with one or more spectra with weighting factors. The weighting factors enable the calibration of the correlation of the frequencies of interest within the spectrum. Therefore, if the first spectrum has a higher weighting factor than the second spectrum, the first estimate of the frequencies of interest will be given more significant consideration when determining them. Consequently, the spectrum that is more representative of the frequencies of interest is highlighted.
[0033] Preferably, each frequency of interest is identified within each spectrum, and the obtained frequencies of interest are weighted according to their spectral correlation. Therefore, all spectra are used to limit bias. The more spectra available, the higher the accuracy.
[0034] According to one aspect of the invention, in the step of determining the set of frequencies of interest within a predetermined frequency range,
[0035] The moving average is calculated over the spectrum by limiting a window with a fixed predetermined value;
[0036] Remove the fixed moving average of the spectrum to determine at least one peak corresponding to the frequency of interest.
[0037] When the peak value of the spectrum is very narrow (precisely defined), the calculation of the fixed moving average is fast and convenient, especially for low and mid frequencies.
[0038] According to another aspect of the invention, in the step of determining the set of frequencies of interest within a predetermined frequency range,
[0039] The most frequent peak value detected in the spectrum;
[0040] For each peak, its width is measured at its frequency to form a sample;
[0041] The process of performing linear regression on a sample to determine the peak width as a function of frequency;
[0042] The moving average over the spectrum is calculated by limiting a window with variable values determined by linear regression;
[0043] Remove the variable moving average of the spectrum to determine at least one peak corresponding to the frequency of interest.
[0044] The calculation of the variable moving average is a function of the peak width process as a function of frequency, which allows for the efficient removal of all noise to highlight the peaks in the spectrum, especially for high frequencies.
[0045] The present invention also relates to a system for determining one or more faults in the rotating machinery of an aircraft, the system comprising:
[0046] At least one acoustic sensor and at least one vibration sensor are configured to measure at least one acoustic signal and at least one vibration signal, respectively, during the same operating phase of the rotating machine.
[0047] At least one calculator, configured as follows:
[0048] Each acoustic signal and each vibration signal are converted into an acoustic spectrum and a vibration spectrum, respectively, to obtain a spectrum set;
[0049] Determine the set of frequencies of interest within the spectrum set, where each frequency of interest belongs to a predetermined frequency range;
[0050] A fault spectrum library is formed, each fault spectrum includes at least one frequency line, and each frequency line is obtained by a linear combination of the frequency of interest and a predetermined weighting coefficient;
[0051] Each spectrum is compared with each fault spectrum to obtain a score;
[0052] One or more faults in the rotating machine are determined by analyzing the obtained scores.
[0053] The present invention also relates to an aircraft rotating machine comprising the aforementioned system. In other words, the fault determination method can be implemented on a ship. Attached Figure Description
[0054] The invention will be better understood by reading the following description given as an example and by referring to the following drawings given as a non-limiting example, wherein the same reference numerals denote similar objects.
[0055] Figure 1 : Figure 1 It is a schematic diagram of an aircraft that includes rotating machines equipped with sensors of different types.
[0056] Figure 2 : Figure 2 This is a schematic diagram of the steps of a method for determining a fault according to an exemplary embodiment of the present invention.
[0057] Figure 3 : Figure 3 This is another schematic diagram of the three fault spectra and their lines.
[0058] Figure 4A : Figure 4A This is a schematic diagram comparing the lines of the first sound spectrum and the first fault spectrum.
[0059] Figure 4B: Figure 4B This is a schematic diagram comparing the lines of the first sound spectrum and the second fault spectrum.
[0060] Figure 5 : Figure 5 This is another schematic diagram of the steps of a method for determining a fault according to an exemplary embodiment of the present invention.
[0061] It should be noted that the accompanying drawings illustrate the invention in detail to facilitate its implementation, and the drawings can, of course, be used to better define the invention if necessary. Detailed Implementation
[0062] This invention relates to a rotating machine for aircraft, particularly airplanes or helicopters. The invention will be described in particular with respect to rotating machines including electrical equipment, especially an electric motor that extracts mechanical energy from at least one shaft of the rotating machine. It goes without saying that the invention is also applicable to rotating machines without electrical equipment.
[0063] Rotating machines refer to any of the following: turbines, rotating electric machines, generators, pumps, reduction gears, agitators, reactors, compressors, gearboxes, fans, turbine units, turbochargers, etc.
[0064] refer to Figure 1 The illustration shows an aircraft A equipped with two rotating machines T, specifically turbine engines. In this example, each rotating machine T includes multiple measuring sensors, specifically at least one electrical sensor CE, at least one acoustic sensor CA, and at least one vibration sensor CV. The rotating machine T includes a generator GE, which comprises at least one stator component and at least one rotor component rotatably mounted relative to the stator component. The rotor component has a fixed number of magnetic poles.
[0065] exist Figure 1 In the examples, for clarity and brevity, only one sensor of each type is shown, but it goes without saying that there can be more. See then, for reference... Figure 3 Sensors CE, CA, and CV are configured to acquire measured values, specifically acoustic signals Ai of quantity i (i is a natural number), electrical signals Ej of quantity j (j is a natural number), and vibration signals Vk of quantity k (k is a natural number).
[0066] Preferably, each acoustic sensor CAi is configured to measure a single acoustic signal Ai, and the same applies to each electrical sensor CEj and each vibration sensor CVk, to measure a single electrical signal Ej and a single vibration signal Vk, respectively.
[0067] Preferably, signals Ai, Ej, and Vk are stored in a computer memory during flight for later processing, particularly on the ground. Optionally, the aircraft includes an onboard calculator for real-time processing of signals Ai, Ej, and Vk.
[0068] refer to Figure 1 The rotating machine T includes at least one calculator ECU, which is configured to perform several calculation functions when implementing the method, specifically:
[0069] Each acoustic signal, each electrical signal, and each vibration signal is converted into an acoustic spectrum, an electrical spectrum, and a vibration spectrum, respectively, to obtain a spectrum set;
[0070] Determine the set of frequencies of interest within the spectrum set, where each frequency of interest belongs to a predetermined frequency range;
[0071] A fault spectrum library is formed, each fault spectrum includes at least one frequency line, and each frequency line is obtained by a linear combination of the frequency of interest and a predetermined weighting coefficient;
[0072] Each spectrum is compared with each fault spectrum to obtain a score;
[0073] One or more faults of the rotating machine T are determined by analyzing the obtained scores.
[0074] These calculation steps will be explained in detail later.
[0075] Now refer to Figure 2 Describe the method used to determine faults in rotating machine T. Figure 2 Steps ET1-ET6 of the method are illustrated schematically.
[0076] In this example, using sensors CA, CE, and CV, the method includes measuring the acoustic signal Ai, the electrical signal Ej, and the vibration signal Vk during the same operating phase of the rotating machine T. Signals Ai, Ej, and Vk are time signals known to a person skilled in the art.
[0077] During the operational phases of a rotating machine T, particularly taxiing, takeoff, cruise, or landing, if some faults are highlighted based on the mechanical, thermal, or thermodynamic conditions of the rotating machine T, data collection during the same phases makes it possible to characterize these faults in a robust manner.
[0078] Preferably, signals Ai, Ej, and Vk are measured simultaneously and correlated with each other. This makes it easier to determine the frequency of interest, as will be explained later. After acquiring signals Ai, Ej, and Vk, they are noisy (broadband noise) due to the operating environment of the rotating motor T, making this method difficult to use. Furthermore, the tone noise of each signal (fundamental frequency and harmonics) is difficult to determine individually.
[0079] refer to Figure 2 The method includes the step of transforming the ET2 signals Ai, Ej, and Vk into spectra SAi, SEj, and SVk. In other words, the method involves converting each acoustic signal Ai, each electrical signal Ej, and each vibration signal Vk into an acoustic spectrum SAi, an electrical spectrum SEj, and a vibration spectrum SVk, respectively, to obtain a set of spectra. This transformation in the frequency domain is preferably performed via a Fourier transform, but it is self-evident that it can be performed in different ways.
[0080] In a known manner, the spectrum comprises raw data (tone noise) corresponding to the fundamental frequency and harmonic frequencies, as well as secondary data (broadband noise, related equipment) that generate noise in the raw data.
[0081] According to the present invention, reference Figure 2 The method includes the step of determining the frequency set JFI of interest to ET3 from the spectrum sets SAi, SEj, and SVk.
[0082] Preferably, the set of frequencies of interest JFI includes frequencies of interest f1, f2, and fg (g is a natural number), which are predetermined for each rotating machine T and belong to a predetermined frequency range PFg.
[0083] In this example, the rotating machine T is a two-shaft machine and includes a low-pressure shaft, a high-pressure shaft, a low-pressure compressor, a high-pressure compressor, a combustion chamber, a high-pressure turbine, and a low-pressure turbine.
[0084] At least one spectrum is a function of at least one frequency of interest, preferably several frequencies of interest. Advantageously, referring to this figure, there exists a linked library DB_FI associated with each frequency of interest fg, one or more spectra SAi, SEj, SVk.
[0085] For example, the first acoustic spectrum SA1 includes a first frequency of interest f1 corresponding to the rotational frequency of the low-voltage shaft, while the first electrical spectrum SE1 includes the first frequency of interest f1 and a second frequency of interest f2 corresponding to the frequency of the generator's rotor components. The first vibration spectrum SV1 also includes the second frequency of interest f2. This example is for illustrative purposes only, and needless to say, the combination between the spectrum and the frequency of interest can be diverse.
[0086] In other words, each frequency of interest is associated with one or more frequencies in the spectrum. One or more frequencies of interest, fg, may appear clearly in the spectrum (frequencies with significant amplitude and / or far from other frequencies). Conversely, other frequencies of interest may be more difficult to identify (frequencies with low amplitude, those with noise, and / or those close to other frequencies).
[0087] Preferably, the library DB_FI associates each frequency of interest fg with one or more spectra having a weighting factor δ. In this example, the library DB_FI includes at least one spectrum SA1, SE1, SV1 associated with the weighting factor δ for each frequency of interest f1, f2. Preferably, the library DB_FI includes theoretical values for each frequency of interest f1, f2, which helps to determine the theoretical value in the spectrum by limiting the search window (a defined frequency range).
[0088] In this example, the linked library DB_FI includes the following associations.
[0089] f1: [SA1; δ1]; [SE1; δ2] of frequency range PF1
[0090] f2: [SE1; δ3]; [SE1; δ2] of frequency range PF2
[0091] For each predetermined frequency of interest f1, f2, the frequency is searched in the relevant spectrum. In this example, on the one hand, the first predetermined frequency of interest f1 is searched in the first acoustic spectrum SA1 to obtain the first estimated frequency f1e1 associated with the first weighting factor δ1; on the other hand, the first predetermined frequency of interest f1 is searched in the first electrical spectrum SE1 to obtain the second estimated frequency f1e2 associated with the second weighting factor δ2.
[0092] To determine the first predetermined frequency of interest, f1, a weighted calculation is performed based on estimated frequencies f1e1 and f1e2 and weighting factors δ1 and δ2. In this example, the first weighting factor δ1 is greater than the second weighting factor δ2; therefore, the first estimated frequency f1e1 is considered more than the second estimated frequency f1e2 in determining the predetermined frequency of interest, f1. Any mathematical weighting function can be implemented, particularly a weighted average.
[0093] Advantageously, searching for frequencies of interest across several spectra limits inaccuracy. This inaccuracy is low because the spectra have different properties. Therefore, the spectra do not benefit from the same measurement bias, which is advantageous. Preferably, the frequencies of interest are associated with each spectrum via a weighting factor.
[0094] Preferably, each frequency of interest corresponds to the fundamental frequency of the tone noise.
[0095] Preferably, to determine predetermined frequencies of interest in the low and mid frequencies, a moving average is calculated over the spectrum by defining a window with a predetermined fixed value. This moving average is removed from the spectrum to highlight peaks. This method is effective when the peak width is nearly constant across the entire spectrum.
[0096] At high frequencies, as the peak width increases, the most prominent peak in the spectrum is detected, and then the width of each peak at its frequency is measured to form a sample.
[0097] Then, linear regression is performed on the samples to determine the peak width variation process as a function of frequency. Advantageously, a polynomial is obtained, preferably a linear polynomial.
[0098] Therefore, the moving average over the spectrum can be calculated by defining a window with variable values, determined from linear regression, specifically from the obtained polynomial linear regression. The moving average is no longer fixed but varies proportionally to the peak width over the frequency range. The moving average is then removed from the spectrum to highlight the peaks, thereby emphasizing the frequencies of interest.
[0099] After determining step ET3, it is advantageous to obtain the set of frequencies of interest JFI, which includes multiple frequencies of interest fg that have been precisely determined.
[0100] refer to Figure 2 and 3 The method includes the step of forming a fault spectrum library DB_DEF, where each fault spectrum S_DEFm is obtained by a linear combination of frequencies fg of interest having predetermined weighting coefficients αg and βg. Preferably, the weighting coefficients αg and βg are determined based on feedback.
[0101] Each fault spectrum is in the form of a set of frequencies, i.e., a frequency distribution. Preferably, each fault spectrum is in the form of a weighted Dirac comb.
[0102] According to the present invention, each fault spectrum S_DEFm includes at least one frequency line, each frequency line corresponding to a linear combination of frequencies of interest fg with predetermined weighting coefficients. This method is relevant when the frequencies of interest fg are precisely determined. The linear combination makes it possible to highlight the interaction phenomena (slippage) between the various frequencies of interest in a rotating machine.
[0103] For example, for two frequencies of interest, f1 and f2, refer to Figure 3 Three fault spectra S_DEF1, S_DEF2, and S_DEF3 with predetermined weighting coefficients α and β are determined:
[0104] S_DEF1=R1(α11xf1+β11xf2)+R2(α12xf1+β12xf2)
[0105] S_DEF2=R1(α21xf1+β21xf2)+R2(α22xf1+β22xf2)
[0106] S_DEF3=R1(α31xf1+β31xf2)+R2(α32xf1+β32xf2)
[0107] For clarity, each fault spectrum is represented by only two lines, but needless to say, their number can vary. Due to the accuracy of the frequencies of interest f1 and f2, each ray is precisely determined. For example... Figure 3 As shown, lines R1 and R2 are far from fault spectra S_DEF1 and S_DEF3, but close to fault spectrum S_DEF2.
[0108] refer to Figure 2 and 3 The method includes the step of comparing each spectrum SAi, SEj, SVk of ET5 with each fault spectrum S_DEFm to obtain a score SCORE.
[0109] During the step of comparing the spectra SAi, SEj, SVk with the determined fault spectrum S_DEFm to obtain the fractional SCORE, the basic fractional SCORE_E in the spectra SAi, SEj, SVk is calculated for each frequency line of the determined fault spectrum S_DEFm, and the fractional SCORE is determined from the set of basic fractional SCORE_E. In particular, the hue bandwidth height of the readout line is determined.
[0110] For example, refer to Figure 4A The spectrum SA1 is compared with the first fault spectrum S_DEF1, and the energy associated with each frequency line R1, R2 in the spectrum SA1 is measured to obtain two basic scores SCORE_E1, SCORE_E2. In this example, the spectrum SA1 is represented as a continuous curve, but it can also be in the form of a straight line. In this example, to form the score SCORE of the first fault spectrum S_DEF1, the basic scores SCORE_E1, SCORE_E2 are simply summed, but it goes without saying that other mathematical calculations can be performed. In this example, the basic scores SCORE_E1, SCORE_E2 are low (or even empty).
[0111] Similarly, refer to Figure 4BThe same spectrum SA1 is compared with a second fault spectrum S_DEF2, which includes frequency lines R1 and R2 that are close to each other. Advantageously, a high score is obtained because each frequency line is considered independently, which is impossible with prior art methods, in which energy is measured only over a frequency range. Figure 4B In the study, there are two basic scores SCORE_E1, a significant value SCORE_E2, and a moderate population score SCORE.
[0112] For example, the fault database DB_DEF includes four identified faults: DEF1, DE2, DEF3, and DEF4. In this example, the base score DB_SCORE is obtained for each fault, which is low, medium, or high.
[0113] Table 1
[0114] DEF1 DEF2 DEF3 DEF4 DEF5 SA1 Low middle Low Low high SE1 Low Low Low Low Low SV1 Low middle Low Low Low
[0115] This method includes steps to determine faults in the ET6 rotating machine by analyzing scores associated with faults in the scoring library DB_SCORE, in order to identify the most probable fault. In this example, fault DEF5 is a loud acoustic fault because it is only visible on the first acoustic spectrum SA1. Fault DEF2 is a potential fault that could be interpreted as noise or a replica of fault DEF5. Its characterization on the first acoustic spectrum SA1 and the first vibration spectrum SV1 allows it to be determined in a practical manner.
[0116] This invention provides a global view, making it possible to detect low-level or hidden faults. Furthermore, multi-standard analysis allows for a better understanding of the root cause of the fault. In this case, fault DEF5 corresponds to a bearing failure, and a simple analysis would lead to the conclusion that the bearing itself is faulty. By identifying fault DEF2, which is associated with adjacent components, it can be concluded that there is a bearing assembly or installation fault, even though the bearing does not actually have such a fault.
[0117] Preferably, all identified faults associated with their spectrum are stored in a knowledge base for later use in predictive maintenance.
Claims
1. A method for determining one or more faults in the rotating machinery (T) of an aircraft, the method comprising: The step of measuring at least one acoustic signal (Ai) and at least one vibration signal (Vk) during the same operating phase of the rotating machine (T) (ET1). The steps involve converting each acoustic signal (Ai) and each vibration signal (Vk) into an acoustic spectrum (SAi) and a vibration spectrum (SVk) respectively (ET2) to obtain the spectrum set (SAi, SVk). The steps for determining the set of frequencies of interest (JFI) within the (ET3) spectrum set (SAi, SVk), where each frequency of interest (fg) belongs to a predetermined frequency range. The step of forming a fault spectrum library (DB_DEF) involves each fault spectrum (S_DEFm) including at least one frequency line (R1, R2), which is obtained by a linear combination of the frequency of interest (fg) and predetermined weighting coefficients (α, β). The steps of comparing each acoustic spectrum (SAi), each vibration spectrum (SVk), and each fault spectrum (S_DEFm) (ET5) to obtain a score (SCORE) are as follows. The steps to determine the faults of the rotating machine (T) by analyzing the obtained score (SCORE).
2. The method of claim 1, for an aircraft rotating machine (T) comprising at least one electrical device (GE), the method comprising: The step of measuring (ET1) at least one acoustic signal (Ai), at least one electrical signal (Ej), and at least one vibration signal (Vk) during the same operating phase of the rotating machine (T). The steps are to convert each acoustic signal (Ai), each electrical signal (Ej), and each vibration signal (Vk) into an acoustic spectrum (SAi), an electrical spectrum (SEj), and a vibration spectrum (SVk) respectively to obtain a spectrum set (SAi, SEj, SVk).
3. The method according to claim 2, wherein, During the step of comparing the acoustic spectrum (SAi), electrical spectrum (SEj), and vibration spectrum (SVk) with the determined fault spectrum (S_DEFm) (ET5) to obtain a score (SCORE), a basic score (SCORE_E) is calculated for each weighted frequency of interest in the determined fault spectrum (S_DEFm) in the acoustic spectrum (SAi), electrical spectrum (SEj), and vibration spectrum (SVk), and a score (SCORE) is determined from the set of basic scores (SCORE_E).
4. The method according to claim 1, wherein, During the step of determining the set of frequencies of interest (JFI) for (ET3), at least two spectra are functions of at least one predetermined frequency of interest (fg): A first estimate of the frequencies of predetermined interest is determined in the first spectrum. A second estimate of the predetermined frequency of interest is determined in the second spectrum, and The frequency of the planned attention (fg) is determined at least based on a first estimate of the frequency of the planned attention and a second estimate of the frequency of the planned attention.
5. The method of claim 1, wherein the step of determining (ET3) the set of frequencies of interest (JFI) is performed by a link library (DB_FI) that associates each frequency of interest with one or more spectra.
6. The method according to claim 5, wherein, The link library (DB_FI) associates at least each frequency of interest with one or more spectra with weighting factors (δ).
7. The method according to claim 1, wherein, During the step of determining the set of frequencies of interest (JFI) within a predetermined frequency range (ET3), By defining a window with fixed predetermined values, a moving average is calculated over the spectrum. Remove the fixed moving average of the spectrum to determine at least one peak corresponding to the frequency of interest.
8. The method according to claim 1, wherein, During the step of determining the set of frequencies of interest (JFI) within a predetermined frequency range (ET3), The most frequent peak value detected in the spectrum. For each peak, its width is measured at its frequency to form a sample. The process of performing linear regression on a sample to determine the peak width as a function of frequency. The moving average over the spectrum is calculated by limiting the window to values of variables determined by linear regression. Remove the variable moving average of the spectrum to determine at least one peak corresponding to the frequency of interest.
9. The method according to claim 2 or 3, wherein, The electrical signal (Ej) is measured in the electrical equipment (GE).
10. A system for determining one or more faults in the rotating machinery (T) of an aircraft, the system comprising: At least one acoustic sensor (CAi) and at least one vibration sensor (CVk) are configured to measure at least one acoustic signal (Ai) and at least one vibration signal (Vk) respectively during the same operating phase of the rotating machine (T). At least one calculator (ECU) configured as follows: Each acoustic signal (Ai) and each vibration signal (Vk) is converted into an acoustic spectrum (SAi) and a vibration spectrum (SVk) respectively, in order to obtain a spectrum set (SAi, SVk). Determine the set of frequencies of interest (JFI) within the spectrum set (SAi, SVk), where each frequency of interest (fg) belongs to a predetermined frequency range. A fault spectrum library (DB_DEF) is formed, and each fault spectrum (S_DEFm) includes at least one frequency line (R1, R2). Each frequency line (R1, R2) is obtained by a linear combination of the frequency of interest (fg) and predetermined weighting coefficients (α, β). Each acoustic spectrum (SAi), each vibration spectrum (SVk), and each fault spectrum (S_DEFm) are compared to obtain a score (SCORE). One or more faults in the rotating machine (T) are determined by analyzing the obtained scores (SCORE).
11. An aircraft rotating machine (T) comprising the system according to claim 10.