Multi-channel phase synchronization fusion spaceflight mechanism fault feature extraction method

Abstract

The invention discloses a multichannel phase synchronization fusion spaceflight mechanism fault feature extraction method, which comprises the following steps: carrying out vibration signal acquisition on a spaceflight structure, and further acquiring multichannel vibration signals of a fault mechanism; carrying out signal-noise separation on the multi-channel signal; analyzing the phase synchronization degree between the channels of the multi-channel low-noise signal through an improved MCMPC, and determining the phase compensation amount of each channel according to the phase synchronization degree; carrying out distributed estimation fusion on the low-noise signals after phase synchronization through local estimation noise cross covariance; performing adaptive decomposition on the fused signal fused in the step S4 through VMD, and outputting an intrinsic mode component; and according to expert experience knowledge and fault mechanism analysis, synchronously extracting spaceflight mechanism fault features from the output intrinsic mode components to realize spaceflight mechanism fault diagnosis. According to the method, the fault judgment reliability is improved, higher flexibility and accuracy are provided, sample collection and model training are not needed, and time and resources are saved.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical fault detection, and in particular to a method for extracting fault features of aerospace mechanisms by multi-channel phase synchronous fusion. Background Art

[0002] The continuous development of technologies such as lunar exploration, deep space exploration, space exposure, and space applications has put forward higher and higher requirements for the high reliability and long-life service capabilities of aerospace agencies. Factors such as harsh space environment and hardware aging make it inevitable for aerospace agencies to fail. If the failure characteristics of aerospace agencies can be detected in time and different types of failures can be identified, it will be of great significance to improve product quality and safety, ensure the success rate of aerospace missions, and avoid economic losses and catastrophic accidents.

[0003] Traditional fault feature extraction methods are based on Fourier transform and filtering technology, focusing on the spectrum analysis and processing of single-channel signals. Compared with single-channel signal processing, the fault feature extraction method of multi-channel phase synchronization fusion has significant advantages. Multi-channel improves the spatial resolution and spectrum utilization of signals by capturing richer information in the spatial, frequency or time domain. Phase synchronization fusion can enhance the strength of the target signal while suppressing noise interference, thereby significantly improving the signal-to-noise ratio and the robustness of signal processing. After searching the published patent applications, it was found that the multi-channel phase synchronization patent applications that are more relevant to the present invention include: (1) Invention patent CN202410979698.X provides a fault detection method for a multi-channel receiver. The invention constructs the amplitude characteristics and phase characteristics of a fault-free multi-channel receiver and a multi-channel receiver with multiple known fault types to realize sample set construction, and trains the fault detection model based on the sample set, so as to realize rapid fault detection of the multi-channel receiver. (2) Invention patent CN201910723964.1 provides a fault arc detection method based on current multi-channel time-frequency feature extraction. The method is as follows: continuously sample the current signal on the live wire, convert it through an ultra-high-speed ADC, and send it to the hardware digital signal processing unit. Bandpass filter the signal to filter out interference signals. Perform time-frequency analysis to extract arc feature vectors. The system performs segmented statistics on the time-frequency feature quantities extracted by the hardware module to form a time-frequency feature matrix, which is sent to the neural network for arc judgment. Patent (1) requires sample construction and model training, which not only consumes time and resources, but also depends on the quality and coverage of sample data. In contrast, the adaptive method proposed in the present invention can adjust the model parameters in real time during operation to maintain stable performance, especially in scenarios where data distribution changes dynamically. The time-frequency analysis proposed in patent (2) is usually based on a fixed framework and cannot adjust the analysis model according to real-time data, resulting in poor performance in non-stationary or complex environments. The adaptive method proposed in the present invention can dynamically optimize the time-frequency resolution or processing method according to the characteristics of the signal by adjusting the analysis strategy in real time, thereby providing higher flexibility and accuracy.

[0004] The present invention is related to the extraction of fault features of aerospace mechanisms. On a theoretical basis, the patents related to the present invention include: (1) Invention patent CN202310790923.0 provides a full life cycle adaptive fault diagnosis method for aerospace major product manufacturing equipment. First, the aerospace equipment fault data is used as a training set for training to determine the model parameters and establish the model. Then, a time series data anomaly detection system is established. The abnormal data screened by the time series data anomaly detection system is used to determine what kind of fault the abnormal data is, and the deep neural network parameters are updated for the data. (2) Invention patent CN202310451431.9 provides a launch vehicle servo mechanism stuck fault diagnosis method. The method determines the servo swing angle extension mark according to the swing angle deviation; determines whether the servo mechanism swing angle is abnormal according to the servo swing angle extension mark; if the servo mechanism swing angle is normal, it is considered that the servo mechanism has not been stuck; if the servo mechanism swing angle is abnormal, it is further determined whether the rocket body attitude angle deviation is abnormal; if the rocket body attitude angle deviation is normal, it is considered that the servo mechanism has not been stuck; if the rocket body attitude angle deviation is abnormal, it is considered that the servo mechanism has been stuck. (3) Invention patent CN202120596329.4 provides a fault diagnosis device for a remote sensing satellite receiving system, including a remote sensing satellite receiving end, a diagnosis unit, an information library and a manual operation terminal, wherein the remote sensing satellite receiving end is bidirectionally connected to the diagnosis unit, the diagnosis unit is bidirectionally connected to the information library, the information library is bidirectionally connected to the manual operation terminal, the manual operation terminal is connected to the remote sensing satellite receiving end through a switch, the remote sensing satellite receiving end is subjected to real-time data detection and fault diagnosis through the diagnosis unit, the diagnosis unit summarizes the diagnosis data and transmits it to the information library for storage, and the diagnosis unit can actively retrieve the diagnosis fault information in the information library at any time. The sample collection and model training of patent (1) is a time-consuming and resource-consuming process, and is also highly dependent on the quality of the sample data and its coverage. Summary of the invention

[0005] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a method for extracting fault features of aerospace mechanisms with multi-channel phase synchronous fusion, solve the possible problem of phase asynchronism of multi-channel signals, realize the fusion of fault feature information of multi-channel signals, use VMD to realize adaptive feature information separation and feature extraction, and provide sufficient criteria for aerospace mechanism fault identification. In order to achieve the above-mentioned purpose and other advantages according to the present invention, a method for extracting fault features of aerospace mechanisms with multi-channel phase synchronous fusion is provided, comprising:

[0006] S1. Collect vibration signals of aerospace structures and then collect multi-channel vibration signals of faulty mechanisms;

[0007] S2, performing signal-noise separation on the multi-channel signal to obtain a group of multi-channel low-noise signals and a group of multi-channel estimated noise signals;

[0008] S3, analyzing the phase synchronization degree between each channel of the multi-channel low-noise signal through the improved MCMPC and determining the phase compensation amount of each channel according to the phase synchronization degree;

[0009] S4, performing distributed estimation fusion on the phase-synchronized low-noise signal by locally estimating the noise cross covariance;

[0010] S5, adaptively decomposing the fused signal after fusion in step S4 through VMD, and outputting the eigenmode components;

[0011] S6. Based on expert experience and fault mechanism analysis, the aerospace mechanism fault characteristics are synchronously extracted from the output eigenmode components to achieve aerospace mechanism fault diagnosis.

[0012] The adaptive method introduced in the present invention directly decomposes the received signal without collecting samples or training models, thus saving time and resources.

[0013] The present invention uses multi-channel signal-to-noise separation, combines the analysis results of multiple features and the mechanism of mechanical failure, and extracts and identifies typical failure features. Multi-channel captures richer information in the space, frequency or time domain to improve the reliability of judgment.

[0014] The adaptive method proposed in the present invention can dynamically optimize the time-frequency resolution or processing method according to the characteristics of the signal by adjusting the analysis strategy in real time, thereby providing higher flexibility and accuracy.

[0015] The present invention also adopts an adaptive decomposition method to carry out aerospace mechanism fault feature recognition.

[0016] This application proposes a method for extracting fault features of aerospace mechanisms with multi-channel phase synchronization fusion, which is used to solve some difficulties in the current aerospace mechanism fault diagnosis methods. For example: (1) The diversity of compound faults, that is, aerospace mechanisms may have multiple different types of faults at the same time, such as gear tooth wear, gear tooth breakage, lubricant contamination, etc. The fault features of these faults vary and are difficult to be effectively distinguished and extracted by traditional feature extraction methods; (2) The amount of data is large, that is, the fault monitoring data of aerospace mechanisms that operate continuously for a long period of time usually has a large number of samples and high-dimensional characteristics. At the same time, spacecraft fault diagnosis has a high demand for real-time performance, which puts forward high requirements for the efficiency of fault extraction algorithms; (3) The noise interference is large, that is, due to the complex working conditions of aerospace mechanisms, the monitoring signal acquisition process is very susceptible to various noise interferences, which puts forward high requirements for the robustness of fault extraction algorithms. (4) Signal phase asynchrony, that is, due to the poor working conditions of aerospace mechanisms, there may be multi-channel acquisition asynchrony during the signal acquisition process, resulting in phase asynchrony between multi-channel signals. This will seriously affect the subsequent accuracy of aerospace mechanism fault feature extraction.

[0017] The present invention aims at a series of fault monitoring and diagnosis problems existing in aerospace institutions due to the particularity of the working environment. The improved multi-component mean phase coherence (MCMPC) is used to estimate the phase synchronization between channels and compensate for the phase imbalance of multi-channel signals, so as to solve the possible phase asynchrony problem of multi-channel signals. Then, the local estimation of the estimated noise cross covariance is used to perform distributed estimation fusion on the feature information in the multi-channel signal to form a new fusion signal, so as to realize the fusion of multi-channel signal fault feature information. Finally, VMD is used to realize adaptive feature information separation and feature extraction, so as to provide sufficient judgment criteria for aerospace institution fault identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flow chart of the method for extracting fault features of aerospace mechanisms by multi-channel phase synchronous fusion according to the present invention;

[0019] Figure 2 A multi-channel vibration signal of an aerospace mechanism in an embodiment of the method for extracting fault features of an aerospace mechanism by multi-channel phase synchronous fusion according to the present invention;

[0020] Figure 3 This is a graph of aerospace mechanism fault feature multi-channel synchronous extraction results in an embodiment of the aerospace mechanism fault feature extraction method using multi-channel phase synchronous fusion according to the present invention. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] Reference Figure 1 , a method for extracting fault features of aerospace mechanisms by multi-channel phase synchronous fusion, comprising:

[0023] Step S1, collecting vibration signals of aerospace mechanisms, and collecting multi-channel vibration signals y(t) of faulty mechanisms.

[0024] Step S2, using high-order singular value decomposition to separate the multi-channel signal of the aerospace agency from the noise, that is, decomposing it into a set of multi-channel low-noise signals y i (i=1,2,...,n) and a set of multi-channel estimated noise signals n i (i=1,2,..n)., n is the number of channels.

[0025] Step S3, using the improved MCMPC to analyze the phase synchronization degree r between each channel of the multi-channel low-noise signal obtained in step S2, can be expressed as:

[0026]

[0027] Where max(·) is the maximum value and K is the number of components. express The phase estimate of the Kth component of , where N is the length of the component. It can be expressed as:

[0028]

[0029] In the formula, Im(·) represents the imaginary part of the complex number, and Re(·) represents the real part of the complex number; represents the Hilbert transform; express The Kth component of . The phase compensation amount of each channel is determined according to the phase synchronization degree r, and the phase compensation between channels is completed. The low-noise signal after phase synchronization can be expressed as y i (i=1,2,...,n).

[0030] Step S4, using the local estimated noise cross covariance to obtain the phase-synchronized low-noise signal y in S3 i Distributed estimation fusion is performed. The vibration signals of channel i and channel j, i.e. y i (t) and y jThe fusion signal of (t) can be expressed as:

[0031]

[0032] where Σ i is the estimated mean square error matrix, Σ ij is the unknown cross-covariance matrix, which can be defined as:

[0033]

[0034] Where E(·) is the mathematical expectation operation, and the estimated noise It has been obtained in step S2. cor(·,·) is the correlation coefficient of two vectors, C i satisfy When n≥3, it is necessary to combine and superimpose the multi-channel signals in pairs to obtain a fused signal.

[0035] Step S5, using VMD to adaptively decompose the fused signal obtained in step S4, and outputting intrinsic mode functions (IMFs);

[0036] Step S6, based on expert experience and fault mechanism analysis, synchronously extract aerospace mechanism fault features from the output IMFs to achieve aerospace mechanism fault diagnosis.

[0037] 2. The component calculation characteristics of step S3 specifically include

[0038] Step S3-1, decomposing the multi-channel low-noise signal obtained in step S2, and using wavelet packet transform to decompose the signal into a wavelet tree.

[0039] Step S3-2, according to the minimum Shannon entropy principle of wavelet coefficients, select the best wavelet packet combination from the wavelet tree obtained in S3-1.

[0040] Step S3-3, calculate the spectrum of the best wavelet packet combination obtained in S3-2, then calculate the spectrum overlap area between all spectra, and form a kernel matrix W M×M , M is the number of wavelet packet components, the i-th wavelet packet component x i and the jth wavelet packet component x j The spectral overlap area w ij , that is, W M×M The i-th row and j-th column element in can be defined as:

[0041]

[0042] In the formula, min{·} is the minimum value, |·| is the absolute value, is the Fourier transform, ω is the frequency. According to WM×M Calculate the gradient matrix D M×M , the gradient matrix is ​​a diagonal matrix, and the elements of the main diagonal are W M×M The sum of all elements in the corresponding row, D M×M The i-th row and i-th column element d ii It can be expressed as:

[0043]

[0044] Step S3-4, D obtained in S3-3 M×M and W M×M Subtract and get the embedded Laplace matrix L M×M , which can be defined as:

[0045] L=DW

[0046] For L M×M Perform K clustering to obtain the clustering method corresponding to the wavelet packet components, cluster the wavelet packet components, and obtain K components. The value of K is generally n+1.

[0047] Example 1

[0048] A method for extracting faults of aerospace mechanisms by using multi-channel phase synchronous fusion is as follows:

[0049] Step S1, collecting vibration signals of aerospace mechanisms, and collecting multi-channel vibration signals y(t) of faulty mechanisms.

[0050] Figure 2 1 is a time domain waveform and spectrum diagram of a multi-channel vibration signal y(t) of an aerospace mechanism containing a harmonic reducer collected in an embodiment of the present invention. Figure 2 In the time domain of the vibration signal, there are several obvious pulses. However, due to the presence of noise, the period of the pulse cannot be determined. Figure 2 As shown, it can be seen that only combining time domain and frequency domain analysis cannot provide a reliable basis for aerospace mechanism fault diagnosis. Therefore, the present invention designs a new aerospace mechanism fault extraction method, and the specific method starts from step S2.

[0051] Step S2, using high-order singular value decomposition to separate the multi-channel signal of the aerospace agency into a group of multi-channel low-noise signals and a set of multi-channel estimated noise signals n is the number of channels.

[0052] Step S2-1, tensorize the multi-channel signal to form a new third-order trajectory tensor represents a real tensor, I1 and I2 represent the number of rows and columns of the tensor slice matrix, that is, is a real tensor of size I1×I1×n. The trajectory can be represented as a feature tensor and the noise tensor The superposition of

[0053]

[0054] Step S2-2, the trajectory tensor Decompose according to the mode, the decomposed expression is:

[0055]

[0056] In the formula is a tensor The left singular matrix in the j-expanded mode, V (j) is a right singular matrix, Σ (j) is a diagonal matrix of singular values.

[0057] Step S2-3, analyze the singular diagonal matrix of each mode and select the appropriate truncation parameter p n , truncate the left singular matrix of each expansion mode. Get the truncated left singular matrix:

[0058]

[0059] Step S2-4, construct a new core tensor using the truncated left singular matrix of each mode, the expression is:

[0060]

[0061] Where ×j represents the product in the j-expanded mode.

[0062] Approximately The low-noise feature tensor of The expression is:

[0063]

[0064] Step S2-5, according to the inverse process of tensor construction in step S2-1, Restore to multi-channel low-noise signal Similarly, the noise tensor can be extracted by taking the left singular matrix after taking the stage parameters in step S2-3 The multi-channel estimated noise can be restored That is, the signal-noise separation of multi-channel signals is completed.

[0065] Step S3, using the improved MCMPC to analyze the phase synchronization degree r between each channel of the multi-channel low-noise signal obtained in step S2, can be expressed as:

[0066]

[0067] Where max(·) is the maximum value and K is the number of components. express The phase estimate of the Kth component of , where N is the length of the component. It can be expressed as:

[0068]

[0069] In the formula, Im(·) represents the imaginary part of the complex number, and Re(·) represents the real part of the complex number. represents the Hilbert transform.

[0070] express The Kth component of . The phase compensation amount of each channel is determined according to the phase synchronization degree r, and the phase compensation between channels is completed. The low-noise signal after phase synchronization can be expressed as

[0071] Step S4, using the local estimated noise cross covariance to estimate the phase-synchronized low-noise signal obtained in step S3 Distributed estimation fusion is performed. The vibration signals of channel i and channel j, i.e. y i (t) and y j (t) fusion signal It can be expressed as:

[0072]

[0073] Among them, Σ i is the estimated mean square error matrix, Σ ij is the unknown cross-covariance matrix, which can be defined as:

[0074]

[0075] Where E(·) is the mathematical expectation operation. The estimated noise It has been obtained in step S2. cor(·,·) is the correlation coefficient of two vectors, C i satisfy When n≥3, it is necessary to combine and superimpose the multi-channel signals in pairs to obtain a fused signal.

[0076] Step S5, using VMD to adaptively decompose the fused signal obtained in step S4, and output IMFs;

[0077] Step S6, based on expert experience and fault mechanism analysis, synchronously extract aerospace mechanism fault features from the output IMFs to achieve aerospace mechanism fault diagnosis.

[0078] Figure 3 This is a result diagram of multi-channel synchronous extraction of aerospace mechanism fault characteristics in an embodiment of the invention. Figure 3 As shown in the figure, the pulse signal has periodic impact characteristics. According to the expert experience and fault mechanism analysis, it is judged that it may be caused by the local wear of the flexible wheel of the harmonic reducer in the aerospace mechanism. In addition, the modulation characteristics are obtained, and this modulation component may be caused by the loosening of the bolts of the harmonic reducer in the mechanism.

[0079] The present invention can clearly and completely extract the fault characteristics of the aerospace mechanism.

[0080] The number of devices and processing scales described here are used to simplify the description of the present invention, and the application, modification and variation of the present invention will be obvious to those skilled in the art.

[0081] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and implementation modes. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the illustrations shown and described herein.

Claims

1. A method for extracting fault features of aerospace mechanisms by multi-channel phase synchronous fusion, characterized in that: The following steps are involved: S1. Collect vibration signals of aerospace structures and then collect multi-channel vibration signals of faulty mechanisms; S2, performing signal-noise separation on the multi-channel signal to obtain a group of multi-channel low-noise signals and a group of multi-channel estimated noise signals; S3, analyzing the phase synchronization degree between each channel of the multi-channel low-noise signal through the improved MCMPC and determining the phase compensation amount of each channel according to the phase synchronization degree; S4, performing distributed estimation fusion on the phase-synchronized low-noise signal by locally estimating the noise cross covariance; S5, adaptively decomposing the fused signal after fusion in step S4 through VMD, and outputting the eigenmode components; S6. Based on expert experience and fault mechanism analysis, the aerospace mechanism fault characteristics are synchronously extracted from the output eigenmode components to achieve aerospace mechanism fault diagnosis.

2. A method for extracting fault features of aerospace mechanisms by multi-channel phase synchronous fusion as claimed in claim 1, characterized in that: The phase synchronization degree in step S3 is expressed as follows: Among them, max(·) is the maximum value, K is the number of components; express The phase estimate of the Kth component of , where N is the length of the component; It can be expressed as: In the formula, I m (·) represents the imaginary part of a complex number, R e (·) represents the real part of a complex number; represents the Hilbert transform; express The Kth component of .

3. A method for extracting fault features of aerospace mechanisms by multi-channel phase synchronous fusion as claimed in claim 2, characterized in that: The step S3 also includes component calculation features, which are as follows: S31, decomposing the multi-channel low-noise signal obtained in step S2, and using wavelet packet transform to decompose the signal into a wavelet tree; S32, according to the minimum Shannon entropy principle of wavelet coefficients, select the best wavelet packet combination from the wavelet tree obtained in step S31; S33, calculating the frequency spectrum of the best wavelet packet combination obtained in step S32, and then calculating the spectrum overlap area between all the frequency spectra to form a kernel matrix; S34, calculating the gradient matrix according to the kernel matrix, subtracting the kernel matrix from the gradient matrix to obtain an embedded Laplace matrix; S35. Perform K clustering on the embedded Laplace matrix to obtain a clustering method corresponding to the wavelet packet components, cluster the wavelet packet components, and obtain K components.

4. The method for extracting fault features of aerospace mechanisms by multi-channel phase synchronous fusion according to claim 1, characterized in that: In step S4, the vibration signals of channel i and channel j, i.e., y i (t) and y j The fusion signal of (t) can be expressed as: where Σ i is the estimated mean square error matrix, Σ ij is the unknown cross-covariance matrix, which can be defined as: Where E(·) is the mathematical expectation operation, and the estimated noise It has been obtained in step S2; cor(·,·) is the correlation coefficient of two vectors, C i satisfy When n≥3, it is necessary to combine and superimpose the multi-channel signals in pairs to obtain a fused signal.

Images