Blind source separation method and system for helicopter main reducer planetary gear fault signal

Abstract

This invention discloses a blind source separation method and system for planetary gear fault signals in helicopter main gear reducers. It sets multiple frequency scales using calculated different fault characteristic frequencies as prior knowledge, and preliminarily locates the approximate position of fault frequency components and determines the number of fault types by detecting the location of the maximum value of the cross-correlation function. Then, by setting the bandwidth of a time-frequency bandpass filter, the original vibration signal is separated to obtain at least one fault vibration signal. Finally, each fault vibration signal is parametrically decomposed to reconstruct a new fault source signal. This method can accurately reconstruct fault source signals even when the number of source signals is unknown or the number of sensors is less than the number of source signals. Furthermore, through cross-correlation calculation and parametric decomposition, the complex blind source separation problem is transformed into an optimal estimation problem of model parameters, achieving accurate estimation of instantaneous frequency and instantaneous amplitude.

Description

Technical Field

[0001] This invention relates to the field of fault signal separation technology, and in particular to a blind source separation method and system for fault signals of planetary gears in helicopter main gear reducers, electronic equipment, and computer-readable storage medium. Background Technology

[0002] Blind Source Separation (BSS) is an important fault signal separation technique. It uses sensors to detect aliased signals during mechanical motion. Under conditions where the source signal and its corresponding transmission channel are unknown, it separates the source signal and its corresponding transmission channel characteristics from the acquired aliased signal, providing a new approach for monitoring equipment operating status and faults in noisy environments. Research focusing on helicopter transmission systems primarily concentrates on blind source separation of vibration signal faults, mainly including time-domain, frequency-domain, and time-frequency-domain blind source separation methods. Because helicopter transmission systems are inherently complex and operate in harsh environments, including constant and variable speeds, their vibration signals are often non-stationary. Existing blind source separation methods typically utilize time-frequency domain signal analysis to extract and separate fault characteristic components of the helicopter transmission system, then perform fault feature extraction and diagnosis. Commonly used time-frequency domain signal analysis methods include Short-Time Fourier Transform (STFT), Wigner-Ville Distribution (WVD), Wavelet Transform, and Hilbert-Huang Transform.

[0003] However, since blind source separation makes no specific assumptions about the mixing matrix and the source signals, the weak known conditions lead to two uncertainties in the separated source signals: amplitude uncertainty and order uncertainty. For signals belonging to the instantaneous mixing model, blind source separation methods can only recover the waveform of the source signal, but cannot guarantee that the absolute amplitude or position of the source signal remains unchanged. The vibration signal of the planetary gear in the helicopter main gear reducer belongs to a typical instantaneous mixing model. Therefore, existing blind source separation methods cannot guarantee that the amplitude and position of the separated source signals are accurate. In addition, blind source separation usually requires knowing the number of source signals and ensuring that the number of sensors is greater than the number of source signals. However, for rotating machinery, even a basic gearbox is composed of multiple bearings, gears, and other components, and the number of vibration sources is usually unknown. Moreover, the space available for installing sensors is limited, making the number of sensors less than the number of source signals. Existing blind source separation methods estimate the number of source signals by retaining the main components and ignoring the secondary components based on a preset threshold, or by treating the secondary components as noise. This method is difficult to accurately determine the threshold and is also difficult to define the boundary between secondary components and useless noise signals. Therefore, existing blind source separation methods are difficult to accurately separate source signals when the number of source signals is unknown or the number of sensors is less than the number of source signals. Summary of the Invention

[0004] This invention provides a blind source separation method and system for planetary gear fault signals in helicopter main reducers, as well as electronic devices and computer-readable storage media. It can accurately separate source signals when the number of source signals is unknown or the number of sensors is less than the number of source signals, and can ensure that the amplitude and position of the source signals are accurate.

[0005] According to one aspect of the present invention, a method for blind source separation of planetary gear fault signals in a helicopter main gear reducer is provided, comprising the following:

[0006] Obtain the structural parameters of the planetary gear train and calculate the fault characteristic frequencies of different fault types;

[0007] A frequency scale is generated based on the fault characteristic frequency of each fault type, with the fault characteristic frequency as the interval, the gear meshing frequency as the center frequency, and containing multiple pulses.

[0008] The original vibration signal is acquired, and the cross-correlation function between each frequency scale and the original vibration signal is calculated. The frequency position of each fault frequency component in the original vibration signal is initially located by detecting the position where the maximum value of the cross-correlation function occurs.

[0009] A time-frequency bandpass filter is set based on the frequency position of each fault frequency component in the original vibration signal, and the original vibration signal is separated by using at least one set time-frequency bandpass filter to obtain at least one fault vibration signal.

[0010] At least one fault vibration signal is parametrically decomposed and reconstructed to obtain the fault source signal.

[0011] Furthermore, for gear ring fault types, when it is a partial fault in the gear ring, the characteristic frequency of the gear ring fault is calculated based on the following formula:

[0012]

[0013] When the fault is a distributed fault in the gear ring, the fault characteristic frequency of the gear ring is calculated based on the following formula:

[0014]

[0015] For planetary gear failure types, the characteristic frequency of planetary gear failure is calculated based on the following formula:

[0016]

[0017] For the sun gear failure type, when it is a partial failure of the sun gear, the characteristic frequency of the sun gear failure is calculated based on the following formula:

[0018]

[0019] When the sun gear experiences a distributed failure, the characteristic frequency of the sun gear failure is calculated based on the following formula:

[0020]

[0021] Among them, f r f P and f S f represents the characteristic frequencies of faults in the ring gear, planet gears, and sun gear, respectively. m N represents the gear meshing frequency. r N represents the number of teeth on the gear ring. P N represents the number of teeth on a planetary gear. S NP represents the number of teeth on the sun gear and the number of planet gears in the planetary gear train.

[0022] Furthermore, the bandwidth of the time-frequency bandpass filter does not exceed the frequency difference between any two fault frequency components.

[0023] Furthermore, the process of parametrically decomposing at least one fault vibration signal includes the following:

[0024] One of the fault vibration signals is represented as an instantaneous frequency model. A parameterized time-frequency transformation is performed on the instantaneous frequency model, and the optimal transformation kernel parameters of the parameterized time-frequency transformation are estimated by using the ridge iterative fitting method. Based on the instantaneous frequency model and the optimal transformation kernel parameters, the instantaneous frequency of the signal component with the largest energy is estimated.

[0025] Solving the linear system based on the instantaneous frequency of the signal component with the highest energy, the instantaneous frequency and amplitude parameters of each signal component are estimated.

[0026] The corresponding signal component is reconstructed based on the instantaneous frequency and amplitude parameters of each signal component.

[0027] Furthermore, the process of solving the linear system based on the instantaneous frequency of the signal component with the highest energy, and estimating the instantaneous frequency and amplitude parameters of each signal component, includes the following:

[0028] First, the amplitude parameters of the fault vibration signal are estimated using the Tikhonov regularization method. Based on the instantaneous frequency and the estimated amplitude parameters, the corresponding signal components are reconstructed. Then, the reconstructed signal components are subtracted from the current fault vibration signal. The above steps are repeated for the remaining signal to obtain the instantaneous frequency and amplitude parameters of the remaining signal components.

[0029] Furthermore, after obtaining the instantaneous frequency and amplitude parameters of all signal components, the amplitude parameters of all signal components are jointly optimized and solved using the following formula:

[0030]

[0031] in, It represents the amplitude parameter vector matrix of all signal components, where K represents the number of signal components, and the superscript T indicates transpose. H represents the amplitude parameter vector matrix of all estimated signal components, argmin() represents the index function of the minimum value, s represents the fault vibration signal, and H = [H1...H2] is the amplitude parameter vector matrix of all estimated signal components. K ], which represents the instantaneous frequency matrix vector of all signal components, || ||2 represents the L2 norm, λ2 represents the regularization parameter, and λ2>0.

[0032] Furthermore, if the original vibration signal has only one type of fault, the reconstructed signal component is used as the new fault source signal; if the original vibration signal has at least two types of fault, the average value of the at least two reconstructed signal components is used as the new fault source signal.

[0033] In addition, the present invention also provides a blind source separation system for planetary gear fault signals of helicopter main gear reducers, employing the method described above, including:

[0034] The fault characteristic frequency calculation module is used to obtain the structural parameters of the planetary gear train and calculate the fault characteristic frequencies of different fault types.

[0035] The frequency scale generation module is used to generate a frequency scale based on the fault characteristic frequency of each fault type, with the fault characteristic frequency as the interval, the gear meshing frequency as the center frequency, and containing multiple pulses.

[0036] The cross-correlation calculation module is used to acquire the original vibration signal and calculate the cross-correlation function between each frequency scale and the original vibration signal. By detecting the location where the maximum value of the cross-correlation function occurs, the frequency position of each fault frequency component in the original vibration signal can be initially located.

[0037] The signal separation module is used to set a time-frequency bandpass filter based on the frequency position of each fault frequency component in the original vibration signal, and to use at least one set time-frequency bandpass filter to separate the original vibration signal to obtain at least one fault vibration signal.

[0038] The signal decomposition and reconstruction module is used to parametrically decompose at least one fault vibration signal and reconstruct the fault source signal.

[0039] In addition, the present invention also provides an electronic device, including a processor and a memory, wherein the memory stores a computer program, and the processor executes the steps of the method described above by calling the computer program stored in the memory.

[0040] In addition, the present invention provides a computer-readable storage medium for storing a computer program for blind source separation of fault signals of planetary gears of helicopter main gear reducers, wherein the computer program executes the steps of the method described above when running on a computer.

[0041] The present invention has the following beneficial effects:

[0042] The present invention provides a blind source separation method for planetary gear fault signals in helicopter main gear reducers. First, fault characteristic frequencies for different fault types are calculated based on the structural parameters of the planetary gear train. Then, using these calculated fault characteristic frequencies as prior knowledge, multiple frequency scales are set with the fault characteristic frequencies as intervals and the gear meshing frequency as the center frequency. By calculating the cross-correlation function between each frequency scale and the original vibration signal spectrum, the cross-correlation function always reaches its maximum value when the frequency components contained in the frequency scale approximately coincide with the fault frequency components of the original vibration signal. Therefore, by detecting the location of the maximum value of the cross-correlation function, the approximate frequency position of the fault frequency component in the original vibration signal spectrum and the number of fault types can be preliminarily located. Next, the bandwidth of a time-frequency bandpass filter is set according to the frequency position of each fault frequency component in the original vibration signal, and the original vibration signal is separated to obtain at least one fault vibration signal. Finally, each fault vibration signal is parametrically decomposed to reconstruct a new fault source signal, thereby achieving blind source separation of planetary gear fault signals in helicopter main gear reducers. The blind source separation method of the present invention can accurately reconstruct the fault source signal when the number of source signals is unknown and the number of sensors is less than the number of source signals. Furthermore, by cross-correlation calculation and parameterized decomposition, the complex blind source separation problem is transformed into the optimal estimation problem of model parameters, thereby achieving accurate estimation of instantaneous frequency and instantaneous amplitude and ensuring that the amplitude and location of the reconstructed fault source signal are accurate.

[0043] In addition, the blind source separation system for the planetary gear fault signal of the helicopter main reducer of the present invention also has the above-mentioned advantages.

[0044] In addition to the objectives, features, and advantages described above, the present invention has other objectives, features, and advantages. The invention will now be described in further detail with reference to the figures. Attached Figure Description

[0045] The accompanying drawings, which form part of this application, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0046] Figure 1 This is a flowchart illustrating a preferred embodiment of the method for separating blind sources of fault signals in the planetary gears of a helicopter main reducer according to the present invention.

[0047] Figure 2 yes Figure 1 A schematic diagram of the sub-process of step S5.

[0048] Figure 3 This is a schematic diagram of the spectrum of the observation channel selected during the simulation experiment in a preferred embodiment of the present invention.

[0049] Figure 4 This is a schematic diagram of the gear ring fault frequency detection scale generated during simulation experiments in a preferred embodiment of the present invention.

[0050] Figure 5 This is a schematic diagram of the cross-correlation results of gear ring faults calculated during simulation experiments in a preferred embodiment of the present invention.

[0051] Figure 6 This is a schematic diagram of the gear ring fault signal source separation result obtained during simulation experiments in a preferred embodiment of the present invention.

[0052] Figure 7 This is a schematic diagram of the planetary gear failure frequency detection scale generated during simulation experiments in a preferred embodiment of the present invention.

[0053] Figure 8 This is a schematic diagram of the planetary gear fault signal source separation results obtained during simulation experiments in a preferred embodiment of the present invention.

[0054] Figure 9 This is a schematic diagram of the sun gear failure frequency detection scale generated during simulation experiments in a preferred embodiment of the present invention.

[0055] Figure 10 This is a schematic diagram of the sun gear fault signal source separation result obtained during simulation experiments in a preferred embodiment of the present invention.

[0056] Figure 11 This is a schematic diagram of the module structure of a blind source separation system for planetary gear fault signals of a helicopter main reducer, according to another embodiment of the present invention. Detailed Implementation

[0057] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0058] Reference Figure 1 A preferred embodiment of this application provides a method for blind source separation of planetary gear fault signals in a helicopter main gear reducer, including the following:

[0059] Step S1: Obtain the structural parameters of the planetary gear train and calculate the fault characteristic frequencies for different fault types;

[0060] Step S2: Generate a frequency scale based on the fault characteristic frequency of each fault type, with the fault characteristic frequency as the interval, the gear meshing frequency as the center frequency, and containing multiple pulses;

[0061] Step S3: Acquire the original vibration signal and calculate the cross-correlation function between each frequency scale and the original vibration signal. By detecting the location where the maximum value of the cross-correlation function occurs, the frequency position of each fault frequency component in the original vibration signal can be preliminarily located.

[0062] Step S4: Set a time-frequency bandpass filter based on the frequency position of each fault frequency component in the original vibration signal, and use the set time-frequency bandpass filter to separate the original vibration signal to obtain at least one fault vibration signal;

[0063] Step S5: Perform parametric decomposition on at least one fault vibration signal and reconstruct the fault source signal.

[0064] It is understood that the blind source separation method for the planetary gear fault signal of the helicopter main gear reducer in this embodiment first calculates the fault characteristic frequencies of different fault types based on the structural parameters of the planetary gear train. Then, using the calculated fault characteristic frequencies as prior knowledge, multiple frequency scales are set with the fault characteristic frequencies as intervals and the gear meshing frequency as the center frequency. By calculating the cross-correlation function between each frequency scale and the original vibration signal spectrum, the cross-correlation function will always reach its maximum value when the frequency components contained in the frequency scale approximately coincide with the fault frequency components of the original vibration signal. Therefore, by detecting the location of the maximum value of the cross-correlation function, the approximate frequency position of the fault frequency component in the original vibration signal spectrum and the number of fault types can be preliminarily located. Then, the bandwidth of the time-frequency bandpass filter is set according to the frequency position of each fault frequency component in the original vibration signal, and the original vibration signal is separated to obtain at least one fault vibration signal. Finally, each fault vibration signal is parametrically decomposed to reconstruct a new fault source signal, thereby realizing the blind source separation of the planetary gear fault signal of the helicopter main gear reducer. The blind source separation method of the present invention can accurately reconstruct the fault source signal when the number of source signals is unknown and the number of sensors is less than the number of source signals. Furthermore, by cross-correlation calculation and parameterized decomposition, the complex blind source separation problem is transformed into the optimal estimation problem of model parameters, thereby achieving accurate estimation of instantaneous frequency and instantaneous amplitude and ensuring that the amplitude and location of the reconstructed fault source signal are accurate.

[0065] It is understood that in step S1, for a given fault type, the vibration signal spectrum of the planetary gear train exhibits a periodic structure centered on the gear meshing frequency and interspersed with corresponding fault characteristic frequencies. Therefore, if the structural parameters of the planetary gear train are known, the fault characteristic frequencies corresponding to different fault types can be calculated, and the corresponding frequency components in the original vibration signal spectrum collected by the sensor can be located accordingly. The structural parameters of the planetary gear train include the number of teeth on each gear, the number of planet gears, the operating speed, and the gear meshing frequency. It is understood that for gear ring fault types, when it is a local fault on the gear ring, for example, damage to a single tooth surface of the gear ring, during one revolution of the planet carrier, the faulty tooth on the gear ring will mesh with each planet gear once. The gear ring fault characteristic frequency can then be calculated based on the following formula:

[0066]

[0067] When the fault is distributed across the gear ring, meaning each tooth surface of the gear ring is damaged, its fault characteristic frequency is equal to the rotational speed of the gear ring relative to the planetary carrier. In the case of a fixed-mount gear ring, the fault characteristic frequency is calculated based on the following formula:

[0068]

[0069] For planetary gear failure types, if it is a localized failure, such as pitting, during one revolution of the planetary gear relative to the planet carrier, the faulty tooth on the planetary gear will engage once. The component engaging with the faulty tooth could be the sun gear or the ring gear. This engagement will cause a sudden change in the vibration signal. Therefore, the characteristic frequency of planetary gear failure is calculated based on the following formula:

[0070]

[0071] If it is a distributed fault in the planetary gears, assuming the fault is located on each planetary gear tooth, as each faulty tooth engages in meshing, the distributed fault will produce periodic variations. The envelope of these periodic variations is replicated and fluctuates with the rotation frequency of the planetary gear relative to the planet carrier. The fault characteristic frequency of the distributed fault in the planetary gears is the same as the fault characteristic frequency of the local fault in the planetary gears.

[0072] For sun gear failure types, when it is a partial sun gear failure, the meshing of the faulty gear tooth with the planetary gears will introduce abrupt changes in the vibration signal. During one revolution relative to the planet carrier, the faulty area will mesh with each planetary gear. Therefore, the characteristic frequency of a partial sun gear failure is equal to the relative rotational frequency of the sun gear multiplied by the number of planetary gears. Specifically, the characteristic frequency of the sun gear failure is calculated based on the following formula:

[0073]

[0074] When the fault is a distributed fault in the sun gear, the principle is similar to that in the distributed fault of the gear ring, specifically calculated based on the following formula for the characteristic frequency of the sun gear fault:

[0075]

[0076] Among them, f r f P and f S f represents the characteristic frequencies of faults in the ring gear, planet gears, and sun gear, respectively. m f represents the gear meshing frequency. m =f c N r f c The frequency of planet carrier rotation, N r N represents the number of teeth on the gear ring. P N represents the number of teeth on a planetary gear. S NP represents the number of teeth on the sun gear and the number of planet gears in the planetary gear train.

[0077] Therefore, given the structural parameters of the planetary gear train, the fault characteristic frequencies corresponding to different fault types can be calculated using the above formula.

[0078] It is understood that in step S2, when different fault sources exist in the planetary gears of the helicopter main reducer, i.e., when one or more of the following are present: sun gear fault, planet gear fault, and ring gear fault, the signal sequence model obtained by the vibration sensor can be represented as:

[0079]

[0080] in, x represents the vibration signal sequence collected by the vibration sensor. A (n) represents the fault-free vibration signal sequence, RTIV(n) represents the fault vibration signal introduced by the gear ring fault, PTIV(n) represents the fault vibration signal introduced by the planetary gear fault, and STIV(n) represents the fault vibration signal introduced by the sun gear fault.

[0081] Since there is no mutual interference between the signal components of different fault sources, the above signal sequence model can actually be represented as a superposition of periodic spectral structures centered on the gear meshing frequency and with corresponding fault characteristic frequencies as intervals, that is:

[0082]

[0083] Among them, Freq(f m ) indicates that in f m A spectral sequence containing frequency components exists at f, and so on for others. mf represents the gear meshing frequency. r f represents the characteristic frequency of gear ring faults. p f represents the characteristic frequency of planetary gear failure. S Let k ∈ N*, which is a constant representing the characteristic frequency of the sun gear fault. Therefore, according to formula (7), regardless of whether there are single or multiple fault components, the vibration signal of the planetary gear train contains periodic signal components corresponding to the fault type. Thus, given the known structural parameters of the planetary gear train, the corresponding fault characteristic frequency can be calculated as prior knowledge. Furthermore, by setting a frequency scale with the fault characteristic frequency as the interval and the gear tooth meshing frequency as the center frequency, and then calculating the cross-correlation function between the frequency scale and the original vibration signal spectrum, the detection of the fault-related vibration components can be achieved. Therefore, a frequency scale with the corresponding fault characteristic frequency as the interval and the gear tooth fitting frequency f is generated. m A frequency scale Freq(m) centered at the x-axis and containing K pulses is defined. Furthermore, each Freq(m) has the same frequency resolution as the original vibration signal spectrum x(m) of the planetary gear train. Taking a sun gear failure as an example, the frequency f in Freq(m) is... m ±kf S A pulse exists at a certain point. The amplitude of the pulse in the frequency scale Freq(m) can be set according to the actual situation. Usually, the amplitude at the center frequency is set to 1.5, the amplitude at the fault characteristic frequency is set to 1, and the amplitude at other frequencies is set to 0.

[0084] It can be understood that in step S3, the spectral sequence of the original vibration signal is acquired, and the cross-correlation function between each frequency scale Freq(m) and the spectral sequence of the original vibration signal is calculated. Then, the frequency position of each fault frequency component in the spectral sequence of the original vibration signal is initially located by detecting the position where the maximum value of the cross-correlation function occurs. Specifically, the cross-correlation function is a concept in signal analysis, which represents the degree of correlation between two time series, that is, it describes the degree of correlation between the values ​​of signals x(t) and y(t) at any two different times t1 and t2. When describing the correlation between two different signals, these two signals can be random signals or deterministic signals. Let f1(t) and f2(t) be energy signals, which are generally complex functions of time, called: Let f1(t) and f2(t) be the cross-correlation function. The cross-correlation function has the following properties: (1) R 12 (t)=R 21 (t); (2) For f1(t) and f2(t) which are functions with the same period, the cross-correlation function has the same periodicity. In digital signal processing, the cross-correlation formula for continuous signals can be expressed as: The cross-correlation formula for discrete signals can be expressed as: Specifically, if the discrete signal is a binary signal, the cross-correlation function should be expressed as: Where A and D represent the number of identical and different symbols between the x sequence and the y sequence after cyclic shifting n bits, respectively. Therefore, in step S3, the cross-correlation function between the frequency scale Freq(m) and the original vibration signal spectrum x(m) is specifically calculated based on the following formula: And detect the maximum value R(n) of the cross-correlation function. max The location of the fault frequency component is determined to pinpoint its position within the original vibration signal spectrum. For planetary gear trains, due to fluctuations in rotational speed, the actual frequency components deviate slightly from the theoretical values. However, since the vibration signal contains corresponding periodic vibration frequency components, the cross-correlation function will always reach its maximum value when the frequency components on the frequency scale approximately coincide with the fault frequency components of the detected signal. By detecting the location of the maximum cross-correlation function value, the approximate frequency position of the fault frequency component can be preliminarily located.

[0085] It is understood that the original vibration signal of the planetary gears in the helicopter main reducer is a non-stationary signal. The instantaneous amplitude of a non-stationary signal can reflect the instantaneous bandwidth of the signal in the time-frequency domain. This invention represents the signal amplitude function as a Fourier series, and the order L of the Fourier series directly determines the amplitude bandwidth, and also the instantaneous bandwidth of the signal. Therefore, the signal decomposition method of this invention has similar properties to time-frequency domain bandpass filtering. This method can extract signal components within a specific frequency range on the time-frequency plane. This property is very effective for decomposing the vibration signal of a planetary gear train where instantaneous frequency fluctuations are caused by speed fluctuations. To illustrate this property, the original vibration signal can be processed to obtain:

[0086]

[0087] in, cd represents UV, or s(t) represents the original vibration signal, K represents the number of signal components, L represents the Fourier order, and r(t) represents the noise signal. The above equation shows that each signal component consists of a series of harmonic components, and the instantaneous frequency range of these harmonic components is [f...]. i (t)-LF0,f i (t)+LF0]. Therefore, the signal decomposition process can be regarded as having a center frequency of f. i The bandwidth of the time-frequency bandpass filter (t) is: BW = LF S / N, where N represents the number of sampling points, F S BW represents the sampling frequency, BW represents the bandwidth, and L represents the Fourier order, which takes a value between 2 and 1. 7 ~2 8Therefore, in step S4, the center frequency f of the time-frequency bandpass filter is determined based on the frequency position of each fault frequency component in the original vibration signal. i (t), then set the bandwidth of the time-frequency bandpass filter, and use at least one set time-frequency bandpass filter to separate the original vibration signal to obtain at least one fault vibration signal.

[0088] Optionally, the Fourier order is adjusted so that the bandwidth of the time-frequency bandpass filter does not exceed the frequency difference between any two fault frequency components. It is understood that increasing the filter bandwidth helps compensate for instantaneous frequency estimation errors in the subsequent parametric decomposition process. Even if the instantaneous frequency estimation error is large, the target signal component can still be successfully extracted as long as a sufficiently large filter bandwidth is used. However, if the bandwidth is too large, more noise components or other irrelevant signal components may be extracted. Therefore, in this invention, the filter bandwidth is set to not exceed the frequency difference between any two fault frequency components. This ensures that the subsequent parametric decomposition process can successfully extract the target signal component while avoiding the inclusion of unnecessary noise components or irrelevant signal components in the decomposed target component.

[0089] Understandable, such as Figure 2 As shown, in step S5, the process of parametrically decomposing at least one fault vibration signal includes the following:

[0090] Step S51: Represent one of the fault vibration signals as an instantaneous frequency model, perform parameterized time-frequency transformation on the instantaneous frequency model, and use the ridge iterative fitting method to estimate the optimal transformation kernel parameters of the parameterized time-frequency transformation. Based on the instantaneous frequency model and the optimal transformation kernel parameters, estimate the instantaneous frequency of the signal component with the largest energy.

[0091] Step S52: Solve the linear system based on the instantaneous frequency of the signal component with the highest energy, and estimate the instantaneous frequency and amplitude parameters of each signal component;

[0092] Step S53: Reconstruct the corresponding signal component based on the instantaneous frequency and amplitude parameters of each signal component.

[0093] It is understandable that the vibration signal of the planetary gear train of the main reducer has a frequency characteristic that fluctuates with time, meaning the signal frequency is a function of time. This type of signal can be represented by a frequency-modulated signal model, but we will now use a cosine signal model to represent it: Where s(t) represents the vibration signal, and a(t) is the amplitude function of the vibration signal. It is the phase function of the vibration signal. The analytic signal of signal s(t) is defined as: z(t) represents the analytic signal, and j represents the imaginary unit. Represents the Hilbert transform of the signal. * indicates convolution. Therefore, it can be seen that the analytic signal z(t) is a complex signal, whose real part is equal to the original signal, i.e., R[z(t)] = s(t), where R[] represents the real part of the signal. When a(t) and Under certain conditions, it can be obtained That is, the analytic signal can be represented as: The signal represented by this formula contains only one component, hence it is called a single-component signal.

[0094] The vibration signal of a planetary gear train typically contains multiple signal components, namely the rotational frequency and its higher-order harmonics. This type of signal is called a multi-component signal, which can be represented as: K represents the number of signal components, s i (t) represents the i-th signal component, a i (t) and Let represent the amplitude function and phase function of the i-th signal component, respectively. Similarly, the analytic signal of the multi-component signal can be obtained through the Hilbert transform as follows:

[0095]

[0096] Let Z(f) represent the Fourier transform of the analytic signal z(t), then we have:

[0097]

[0098] For a planetary gear train, its vibration signal s(t) consists of several physically meaningful signal components s i Composed of r(t) and noise r(t), it can be expressed as:

[0099] If f(t) represents the instantaneous frequency that fluctuates with time, then each signal component can be represented in the form of a frequency-modulated component, specifically: θ i0 Let represent the initial phase of the i-th signal component. To eliminate the nonlinear effect of the initial phase, the above equation can be rewritten as:

[0100]

[0101] Among them, u i (t)=a i (t)cosθ i0 v i (t)=-a i (t)sinθ i0 These two represent two newly defined signal component amplitude functions.

[0102] Since the instantaneous amplitude of a wave signal is directly related to its instantaneous bandwidth in the time-frequency domain, a Fourier series model is used to characterize the signal's amplitude function in order to accurately describe the signal's bandwidth properties. The expression is as follows:

[0103]

[0104]

[0105] Where L represents the order of the Fourier model, and Denotes the Fourier coefficients to be estimated, and the frequency resolution of the Fourier model is . F is a constant. S Let Q represent the sampling frequency of the signal, and N represent the number of sampling points. When Q = 1, the above formula is a standard Fourier model, which can only characterize periodic functions. When Q > 1, a redundant Fourier model is obtained, which can characterize more complex functions. Therefore, this invention sets Q = 2. The redundant Fourier model essentially extends the existing N-point signal into a QN-point periodic signal. Therefore, the N-point signal before extension can be an aperiodic signal. The Fourier order L controls the rate of amplitude change, i.e., the degree of amplitude modulation, and also determines the instantaneous bandwidth of the signal model. Therefore, by adjusting the Fourier order L, vibration signals with different degrees of modulation can be analyzed.

[0106] As can be seen from the above frequency modulation component model, the amplitude function u i (t) and v i (t) Related terms to instantaneous frequency and The relationship between them is linear. Therefore, the instantaneous frequency parameters can be estimated first, and then the amplitude parameters can be estimated by solving the linear system. Finally, the frequency modulation components can be reconstructed, that is, the signal components can be reconstructed, thereby realizing signal decomposition.

[0107] Therefore, in step S51, a parameterized time-frequency analysis method is used to estimate the instantaneous frequency, taking one of the signal components s as an example. i Taking the instantaneous frequency estimation of (t) as an example, the signal component (i.e., a fault vibration signal obtained by decomposition) is expressed in analytical form as an instantaneous frequency model, and the expression is:

[0108]

[0109] Among them, z i (t) represents the analytical form of the instantaneous frequency model of the i-th signal component. P represents the frequency constant. i Let κ() represent the instantaneous frequency model parameters of the i-th signal component, and let κ() represent the kernel function. The above analytic signal z... i (t) can be obtained by analyzing the real signal si (t) is obtained by performing a Hilbert transform.

[0110] Based on the principle of parametric time-frequency analysis, a parametric time-frequency transform with a kernel function matching the signal model can be defined, as shown in the following expression:

[0111]

[0112] Where TF() represents the parameterized time-frequency transform function, t represents time, and f represents frequency. This represents the transformation kernel parameters, and σ represents the window width of the Gaussian window function. It represents the frequency demodulation operator, used to reduce the signal z. i (v) frequency modulation degree, It represents the frequency compensation operator, used to compensate for instantaneous frequency errors caused by frequency demodulation. This represents a Gaussian window function with a window width of σ. This invention assumes that κ(τ; 0) ≡ 0, meaning that when the kernel parameter is zero, the transform kernel function is always zero. The above-described kernel function model for parameterized time-frequency transform... This is consistent with the instantaneous frequency model of the signal. It can be seen that when the kernel function is zero, then... The parameterized time-frequency transform degenerates into a short-time Fourier transform. In the most ideal case, if the zero-transform kernel parameters are equal to the instantaneous frequency model parameters of the signal, i.e. Then you can get That is, the signal is demodulated into a stationary signal with a constant frequency. In addition, signal The instantaneous frequency value at v = t and the original signal z i (v) Consistency, meaning that the parameterized time-frequency transform does not change the actual instantaneous frequency of the signal. Therefore, the parameterized time-frequency transform is essentially a transformation of the signal... Unlike existing short-time Fourier transforms, this one performs a frequency demodulation operator. The frequency demodulation effect significantly reduces the frequency modulation degree of the signal within each analysis window function. Therefore, the parameterized time-frequency transformation method of this invention can be used not only for signals with instantaneous frequency fluctuations but also for analyzing strongly time-varying frequency-modulated signals. Therefore, the optimal transformation kernel parameter P for the parameterized time-frequency transformation... i The instantaneous frequency model parameter P i Therefore, the instantaneous frequency of a signal can be accurately extracted by estimating the optimal transform kernel parameters. In this invention, the ridge iterative fitting method is preferably used to estimate the optimal transform kernel parameters. The process of estimating the optimal transform kernel parameters using the ridge iterative fitting method is prior art and will not be described in detail here. Then, the estimated optimal transform kernel parameters are substituted into the analytical form of the instantaneous frequency model to estimate the instantaneous frequency of the signal component.

[0113] In step S52, the process of solving the linear system based on the instantaneous frequency of the signal component with the highest energy, and estimating the instantaneous frequency and amplitude parameters of each signal component, includes the following:

[0114] First, the amplitude parameters of the fault vibration signal are estimated using the Tikhonov regularization method. Based on the instantaneous frequency and the estimated amplitude parameters, the corresponding signal components are reconstructed. Then, the reconstructed signal components are subtracted from the current fault vibration signal. The above steps are repeated for the remaining signal to obtain the instantaneous frequency and amplitude parameters of the remaining signal components.

[0115] It is understandable that the instantaneous frequency model parameter P is estimated in step S51. i Next, the model parameters of the amplitude function need to be estimated to reconstruct the signal components. As mentioned earlier, there is a linear relationship between the amplitude function and the instantaneous frequency correlation term in the signal model. Therefore, given the instantaneous frequency, the amplitude function can be estimated by solving a linear system. For ease of explanation, the signal model is rewritten in matrix form as follows: Where s=[s(t0)...s(t) N-1 )] T , r = [r(t0)...r(t)] N-1 )] T ,s(t N-1 ) and r(t N-1 ) represent the t-th N-1 The signal and noise values ​​at time t0, s(t0) and r(t0) represent the signal and noise values ​​at time t0, respectively, with the superscript T indicating transpose, H i and y i Let these represent the instantaneous frequency matrix and amplitude parameter column vector of the i-th signal component, respectively. For ease of representation, let diag[] denotes a diagonal matrix, and F denotes an N×(2L+1) Fourier model matrix, which can be represented as:

[0116] The above equation shows that the amplitude parameter vector y of the i-th signal component i With the instantaneous frequency matrix H i The relationship is linear, while the instantaneous frequency matrix H i From instantaneous frequency f i (t) is determined. Therefore, in the instantaneous frequency matrix H i Given the conditions, the magnitude parameter y can be estimated by solving the linear system. iHowever, such mathematical inverse problems are usually ill-posed. Therefore, this invention employs the Tikhonov regularization method to solve this problem, that is, to estimate the amplitude parameters of the fault vibration signal. The amplitude parameter vector can be estimated using the following formula:

[0117]

[0118] in, Let ||i||2 represent the amplitude parameter vector matrix of the estimated i-th signal component, ||i||2 represent the L2 norm, and λ1 represent the regularization parameter, λ1 > 0, which aims to improve the numerical stability of solving ill-posed problems. The analytical solution of the above equation can be given by the following expression: Where I1 represents the identity matrix. The corresponding signal components can then be reconstructed using the obtained amplitude parameters, specifically based on the following formula: This represents the i-th reconstructed signal component.

[0119] It is understandable that, since the instantaneous frequency estimation method in step S51 can only obtain the instantaneous frequency of the signal component with the highest energy in the current signal, it can be used... and Reconstruct the signal component with the largest energy Then reconstruct the signal components Subtract the current fault vibration signal s to eliminate the influence of this signal component on the instantaneous frequency estimation of other signal components. Then repeat the above steps on the remaining signal to estimate the instantaneous frequency of other signal components with lower energy and reconstruct the corresponding signal components.

[0120] It is understood that the instantaneous frequency estimation method and amplitude parameter estimation method described above can initially extract all signal components. However, the amplitude parameter estimation method described above will have a large error when dealing with adjacent or intersecting signal components. Therefore, this invention performs joint optimization on all extracted signal components to improve the amplitude estimation accuracy. Optionally, after obtaining the instantaneous frequency and amplitude parameters of all signal components, the amplitude parameters of all signal components are jointly optimized and solved using the following formula:

[0121]

[0122] in, It represents the amplitude parameter vector matrix of all signal components, where K represents the number of signal components, and the superscript T indicates transpose. H represents the amplitude parameter vector matrix of all estimated signal components, argmin() represents the index function of the minimum value, s represents the fault vibration signal, and H = [H1...H2] is the amplitude parameter vector matrix of all estimated signal components. K], which represents the instantaneous frequency matrix vector of all signal components, || ||2 represents the L2 norm, λ2 represents the regularization parameter, and λ2>0.

[0123] Specifically, in step S52 above, the instantaneous frequencies of all signal components can be estimated and their corresponding instantaneous frequency matrix H can be solved. i If i = 1, 2, ..., K, then This can be rewritten as: s = Hy + r, H = [H1...H K ], Then, the amplitude parameters of all signal components are jointly optimized using the following formula: Similarly, the analytical solution can be expressed as: I2 indicates that it is related to H T H is an identity matrix of the same scale. Each signal component can be reconstructed using the joint amplitude optimization results, as expressed by: for The magnitude function of the subvectors can be estimated using the following two equations: in, and They represent u respectively i (t) and v i The estimated value of (t), where F represents the Fourier model matrix, and for Two sub-vectors. Finally, the instantaneous amplitude a. i (t) can be estimated by the following formula:

[0124] It can be understood that in step S53, after obtaining the instantaneous frequency and amplitude parameters of each signal component, the formula can be used to... The corresponding signal components are obtained through reconstruction.

[0125] Furthermore, in step S5, the fault source signal can be reconstructed based on at least one signal component obtained after parametric decomposition and reconstruction. If the original vibration signal has only one fault type, the reconstructed signal component is directly used as the new fault source signal. If the original vibration signal has at least two fault types, the noise caused by different fault types accumulates and superimposes, resulting in errors in the reconstructed fault source signal components for each vibration signal component. Assuming that the noise in the reconstructed signal components satisfies the white noise assumption, the average value of the at least two reconstructed signal components can be taken as the new fault source signal. in, F represents the fault source signal that is being reconstructed. t This indicates the number of fault sources, i.e., the number of fault types.

[0126] It is understood that, in order to verify the effectiveness of the blind source separation method of the present invention, the inventors of this application used the planetary gear train structure parameters shown in Table 1 for simulation and set the sampling frequency F. s =2000Hz, sampling time is 80s, number of sampling points is N=160000, then the frequency resolution should be 0.0125Hz. The fault types are set as distributed fault of the gear ring, partial fault of the planetary gear, and distributed fault of the sun gear, as shown in Table 2. Taking one observation channel as an example, its spectrum is as follows... Figure 3 As shown.

[0127] Table 1. Structural Parameters of Planetary Gear Train

[0128] <![CDATA[Planetary carrier rotation frequency f c > 2.5Hz <![CDATA[Number of teeth N of the gear ring r > 96 <![CDATA[Number of planet gear teeth N P > 34 <![CDATA[Number of teeth of the sun gear N S > 28 <![CDATA[Mesh frequency f of gear teeth m > 240Hz

[0129] Table 2. Fault characteristic frequencies of different fault types

[0130] <![CDATA[Characteristic frequency f of local planetary gear fault P > 7.0625Hz <![CDATA[Characteristic frequency f of sun gear distribution fault S > 8.5875Hz

[0131] The blind source separation method of this invention is then used to sequentially separate the gear ring fault, planetary gear fault, and sun gear fault. Taking the gear ring fault as an example, the characteristic frequency of the gear ring fault is calculated and a corresponding frequency detection scale is generated, as shown below. Figure 4 As shown, the further calculated cross-correlation results are as follows: Figure 5 As shown. The maximum value is taken from the cross-correlation results calculated above, and the bandwidth BW of the time-frequency bandpass filter is set, where the order L of the Fourier model is 2. 7 =96, resulting in a bandwidth BW of 1.2Hz. The fault frequency components were extracted using a parametric signal decomposition method, and the results are as follows. Figure 6 As shown, the characteristic frequency components of the gear ring fault were effectively separated. Similarly, the frequency scale and separation results for the planetary gear fault are shown below. Figure 7 and Figure 8 As shown, the frequency scale and separation results of the sun gear failure are as follows: Figure 9 and Figure 10 As shown, this demonstrates that the blind source separation method of the present invention can successfully separate the three fault characteristic frequency components.

[0132] In addition, such as Figure 11 As shown, another embodiment of the present invention also provides a blind source separation system for planetary gear fault signals in a helicopter main gear reducer, preferably employing the method described above, including:

[0133] The fault characteristic frequency calculation module is used to obtain the structural parameters of the planetary gear train and calculate the fault characteristic frequencies of different fault types.

[0134] The frequency scale generation module is used to generate a frequency scale based on the fault characteristic frequency of each fault type, with the fault characteristic frequency as the interval, the gear meshing frequency as the center frequency, and containing multiple pulses.

[0135] The cross-correlation calculation module is used to acquire the original vibration signal and calculate the cross-correlation function between each frequency scale and the original vibration signal. By detecting the location where the maximum value of the cross-correlation function occurs, the frequency position of each fault frequency component in the original vibration signal can be initially located.

[0136] The signal separation module is used to set a time-frequency bandpass filter based on the frequency position of each fault frequency component in the original vibration signal, and to use at least one set time-frequency bandpass filter to separate the original vibration signal to obtain at least one fault vibration signal.

[0137] The signal decomposition and reconstruction module is used to parametrically decompose at least one fault vibration signal and reconstruct the fault source signal.

[0138] It is understood that the blind source separation system for the planetary gear fault signal of the helicopter main gear reducer in this embodiment first calculates the fault characteristic frequencies of different fault types based on the structural parameters of the planetary gear train. Then, using the calculated fault characteristic frequencies as prior knowledge, multiple frequency scales are set with the fault characteristic frequencies as intervals and the gear meshing frequency as the center frequency. By calculating the cross-correlation function between each frequency scale and the original vibration signal spectrum, the cross-correlation function always reaches its maximum value when the frequency components contained in the frequency scale approximately coincide with the fault frequency components of the original vibration signal. Therefore, by detecting the location where the maximum value of the cross-correlation function occurs, the approximate frequency position of the fault frequency component in the original vibration signal spectrum and the number of fault types can be preliminarily located. Then, the bandwidth of the time-frequency bandpass filter is set according to the frequency position of each fault frequency component in the original vibration signal, and the original vibration signal is separated to obtain at least one fault vibration signal. Finally, each fault vibration signal is parametrically decomposed to reconstruct a new fault source signal, thereby realizing the blind source separation of the planetary gear fault signal of the helicopter main gear reducer. The blind source separation system of the present invention can accurately reconstruct the fault source signal when the number of source signals is unknown and the number of sensors is less than the number of source signals. Furthermore, by cross-correlation calculation and parameterized decomposition, the complex blind source separation problem is transformed into the optimal estimation problem of model parameters, thereby achieving accurate estimation of instantaneous frequency and instantaneous amplitude, and ensuring that the amplitude and location of the reconstructed fault source signal are accurate.

[0139] In addition, another embodiment of the present invention provides an electronic device including a processor and a memory, wherein the memory stores a computer program, and the processor executes the steps of the method described above by calling the computer program stored in the memory.

[0140] In addition, another embodiment of the present invention provides a computer-readable storage medium for storing a computer program for blind source separation of fault signals of planetary gears of helicopter main gear reducers, wherein the computer program executes the steps of the method described above when running on a computer.

[0141] Common computer-readable storage media include: floppy disks, flexible disks, hard disks, magnetic tapes, any other magnetic media, CD-ROMs, any other optical media, punch cards, paper tape, any other physical media with perforated patterns, random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), flash erasable programmable read-only memory (FLASH-EPROM), any other memory chips or cartridges, or any other media readable by a computer. Instructions may further be transmitted or received by a transmission medium. The term transmission medium can include any tangible or intangible medium used to store, encode, or carry instructions for machine execution, and includes digital or analog communication signals or intangible media that facilitate communication of such instructions. Transmission media include coaxial cables, copper wires, and optical fibers, which contain conductors for transmitting a bus of computer data signals.

[0142] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0143] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0144] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0145] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0146] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0147] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

[0148] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for blind source separation of helicopter main gearbox planetary gear failure signals, characterized in that, Includes the following: Obtain the structural parameters of the planetary gear train and calculate the fault characteristic frequencies of different fault types; A frequency scale is generated based on the fault characteristic frequency of each fault type, with the fault characteristic frequency as the interval, the gear meshing frequency as the center frequency, and containing multiple pulses. The original vibration signal is acquired, and the cross-correlation function between each frequency scale and the original vibration signal is calculated. The frequency position of each fault frequency component in the original vibration signal is initially located by detecting the position where the maximum value of the cross-correlation function occurs. A time-frequency bandpass filter is set based on the frequency position of each fault frequency component in the original vibration signal, and the original vibration signal is separated by using at least one set time-frequency bandpass filter to obtain at least one fault vibration signal. At least one fault vibration signal is parametrically decomposed and reconstructed to obtain the fault source signal; The process of parametrically decomposing at least one fault vibration signal includes the following: One of the fault vibration signals is represented as an instantaneous frequency model. A parameterized time-frequency transformation is performed on the instantaneous frequency model, and the optimal transformation kernel parameters of the parameterized time-frequency transformation are estimated by using the ridge iterative fitting method. Based on the instantaneous frequency model and the optimal transformation kernel parameters, the instantaneous frequency of the signal component with the largest energy is estimated. Solving the linear system based on the instantaneous frequency of the signal component with the highest energy, the instantaneous frequency and amplitude parameters of each signal component are estimated. The corresponding signal component is reconstructed based on the instantaneous frequency and amplitude parameters of each signal component; The process of solving a linear system based on the instantaneous frequency of the signal component with the highest energy, and estimating the instantaneous frequency and amplitude parameters of each signal component, includes the following: First, the amplitude parameters of the fault vibration signal are estimated using the Tikhonov regularization method. Based on the instantaneous frequency and the estimated amplitude parameters, the corresponding signal components are reconstructed. Then, the reconstructed signal components are subtracted from the current fault vibration signal. The above steps are repeated for the remaining signal to obtain the instantaneous frequency and amplitude parameters of the remaining signal components. After obtaining the instantaneous frequency and amplitude parameters of all signal components, the amplitude parameters of all signal components are jointly optimized and solved using the following formula: ； in, It represents the amplitude parameter vector matrix of all signal components, K represents the number of signal components, and the superscript T indicates transpose. This represents the amplitude parameter vector matrix of all estimated signal components, argmin() represents the index function of the minimum value, and s represents the fault vibration signal. It represents the instantaneous frequency matrix vector of all signal components. Represents the L2 norm. Represents the regularization parameter. .

2. The blind source separation method for planetary gear fault signals in helicopter main gearboxes as described in claim 1, characterized in that, For gear ring fault types, when it is a partial fault, the characteristic frequency of the gear ring fault is calculated based on the following formula: ； When the fault is a distributed fault in the gear ring, the fault characteristic frequency of the gear ring is calculated based on the following formula: ； For planetary gear failure types, the characteristic frequency of planetary gear failure is calculated based on the following formula: ； For the sun gear failure type, when it is a partial failure of the sun gear, the characteristic frequency of the sun gear failure is calculated based on the following formula: ； When the sun gear experiences a distributed failure, the characteristic frequency of the sun gear failure is calculated based on the following formula: ； in, , and These represent the fault characteristic frequencies of the ring gear, planet gears, and sun gear, respectively. N represents the gear meshing frequency. r N represents the number of teeth on the gear ring. P N represents the number of teeth on a planetary gear. S NP represents the number of teeth on the sun gear and the number of planet gears in the planetary gear train.

3. The blind source separation method for planetary gear fault signals in helicopter main gear reducers as described in claim 1, characterized in that, The bandwidth of the time-frequency bandpass filter does not exceed the frequency difference between any two fault frequency components.

4. The blind source separation method for planetary gear fault signals in helicopter main gearboxes as described in claim 1, characterized in that, If the original vibration signal has only one fault type, the reconstructed signal component is taken as the new fault source signal; if the original vibration signal has at least two fault types, the average value of the at least two reconstructed signal components is taken as the new fault source signal.

5. A blind source separation system for planetary gear fault signals in a helicopter main gear reducer, comprising the method described in any one of claims 1 to 4, characterized in that, include: The fault characteristic frequency calculation module is used to obtain the structural parameters of the planetary gear train and calculate the fault characteristic frequencies of different fault types. The frequency scale generation module is used to generate a frequency scale based on the fault characteristic frequency of each fault type, with the fault characteristic frequency as the interval, the gear meshing frequency as the center frequency, and containing multiple pulses. The cross-correlation calculation module is used to acquire the original vibration signal and calculate the cross-correlation function between each frequency scale and the original vibration signal. By detecting the location where the maximum value of the cross-correlation function occurs, the frequency position of each fault frequency component in the original vibration signal can be initially located. The signal separation module is used to set a time-frequency bandpass filter based on the frequency position of each fault frequency component in the original vibration signal, and to use at least one set time-frequency bandpass filter to separate the original vibration signal to obtain at least one fault vibration signal. The signal decomposition and reconstruction module is used to parametrically decompose at least one fault vibration signal and reconstruct the fault source signal.

6. An electronic device, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program, and the processor executes the steps of the method as described in any one of claims 1 to 4 by calling the computer program stored in the memory.

7. A computer-readable storage medium for storing a computer program for blind source separation of fault signals from planetary gears in a helicopter main gear reducer, characterized in that, The computer program, when run on a computer, performs the steps of the method as described in any one of claims 1 to 4.

Images