CN110146156B

CN110146156B - Denoising method for fault vibration signal of aircraft engine rotor system

Info

Publication number: CN110146156B
Application number: CN201910566648.8A
Authority: CN
Inventors: 刘晓波; 梁春辉; 钟荣升
Original assignee: Nanchang Hangkong University
Current assignee: Nanchang Hangkong University
Priority date: 2019-06-27
Filing date: 2019-06-27
Publication date: 2021-02-09
Anticipated expiration: 2039-06-27
Also published as: CN110146156A

Abstract

The invention discloses a denoising method of a fault vibration signal of an aircraft engine rotor system, which comprises the steps of firstly collecting a rub-impact vibration signal of an aircraft engine rotor, carrying out repeated filtering processing and optimal threshold value on the signal, assigning a suboptimal solution found by iterative optimization of a particle swarm algorithm to an initial solution in a taboo search algorithm, then calculating an adaptability value, continuously searching the neighborhood of the initial value, flexibly using a taboo table and a privilege criterion, enabling the taboo search algorithm to accept some low-quality inferior solutions to a certain extent, preventing the algorithm from falling into a local extreme value, enhancing the global optimization capability of the algorithm, effectively solving the problem of slow late-stage collection speed of the particle swarm algorithm, and effectively avoiding premature falling into the local optimal solution although the convergence speed of the mixed particle swarm algorithm is reduced to a certain extent, the probability of finding the global optimal solution by the population is increased, and the optimization performance of the algorithm is improved on the whole.

Description

Denoising method for fault vibration signal of aircraft engine rotor system

Technical Field

The invention relates to a method for extracting signal characteristic information, in particular to a method for denoising a fault vibration signal of an aircraft engine rotor system.

Background

With the continuous development and progress of modern science and technology, the rotary machine is developed towards high speed and high efficiency on one hand, and is also challenged by more harsh working and operating environments, so that the potential risk of fault occurrence is correspondingly increased, and the rotor system of the core component is one of the parts with the highest fault occurrence rate. Particularly, an aircraft engine rotor system of an aircraft is a system with the highest failure rate, the most complex adjustment and the largest maintenance workload in aircraft mechanical equipment, and the safety, the reliability and the operation of the aircraft are directly influenced by the quality of the working state of the system; if the engine fails during operation, the airplane can not fly normally, personal safety is endangered, and great social and economic losses are caused.

It is statistical that the failure rate of an aircraft engine is about 30% of the failure rate of the whole aircraft, about 40% of the serious flight accidents of the aircraft caused by mechanical reasons are caused by the failure of the engine, the engine is different from the general machinery due to the characteristics of the engine, and even slight mechanical defects or damages can cause the vibration of a rotor system, including the vibration which is almost small and negligible to the vibration which is large enough to cause the damage of the engine. Due to the complexity of the rotor system of the aircraft engine, multiple faults are often represented by the same vibration, and the same fault is represented by different vibrations, so that the false alarm and the false alarm can be caused, and the reliability of fault diagnosis of the rotor system of the aircraft engine is seriously influenced.

At present, wavelet analysis can analyze abrupt change components and noise components of signals from the angles of time domain and frequency domain, so that background noise in the signals can be effectively eliminated; therefore, wavelet analysis is widely applied in the field of signal processing, but the wavelet basis function is difficult to select in wavelet denoising. The multi-wavelet transform is the further development of wavelet theory, and the multi-wavelet can make up for the defects of a single wavelet because the multi-wavelet can simultaneously have important characteristics required in signal processing such as orthogonality, compactness, symmetry, high approximation order and the like, the single wavelet only has one scale function, and the multi-wavelet simultaneously has a plurality of scale functions. Therefore, the multi-wavelet can have important characteristics of symmetry, orthogonality, compactness, high-order vanishing moment and the like which cannot be simultaneously possessed by a single wavelet, so that a better effect can be obtained in the field of signal processing.

The patent with the publication number of CN101968379B and the patent application with the publication number of CN107506709A respectively disclose two methods for extracting the characteristic information of the fault signal of the rotor system of the aircraft engine, wherein the two methods measure the vibration signal by using an eddy current displacement sensor, and respectively analyze by adopting wavelet clusters with different concepts, and mine implicit characteristic information from a large amount of vibration displacement signals to realize the extraction of the working condition characteristic, the former method still has certain defects in extraction precision and speed, the latter extraction method is improved in a breakthrough manner on the basis of the former method, but the former method and the latter method still have a common problem, namely, a noise background frequency spectrum in an extraction result cannot be effectively filtered, the frequency spectrum impact amplitude of weak fault characteristic information is close to that of the frequency spectrum of the fault characteristic information, thereby influencing the reasonable extraction of the rotor fault characteristic to a certain extent, therefore, in order to remove noise interference and more reasonably diagnose faults of the rotor system of the aircraft engine and extract signal characteristics, further improvement is needed through other directions; therefore, the noise reduction research of the rotor system fault vibration signal is of great significance for preventing sudden major accidents.

Disclosure of Invention

(1) Technical problem to be solved

Aiming at the defects of the prior art, the invention aims to provide a denoising method for a fault vibration signal of an aircraft engine rotor system

(2) The technical proposal is that the method comprises the following steps,

in order to solve the technical problem, the invention provides a denoising method of a fault vibration signal of an aircraft engine rotor system, which is specifically a multi-wavelet adjacent coefficient denoising method based on hybrid particle swarm optimization, and the approximate steps of denoising the fault signal of the aircraft engine rotor system by the method are as follows: firstly, performing multi-wavelet transformation on a preprocessed signal, then calculating parameters required by a hybrid particle swarm optimization algorithm through a decomposition coefficient, then solving an optimal threshold value through the hybrid particle swarm optimization algorithm according to a minimization principle of a fitness function, performing corresponding threshold value processing on the decomposition coefficient, and then performing multi-wavelet reconstruction and post-processing to obtain a de-noising signal.

A mixed particle swarm optimization algorithm is also called a PSO algorithm, firstly, the dimension of the algorithm is determined according to the number of independent variables in an optimization problem to be solved, then, an initial position and an initial speed are randomly given to all particles in a swarm, through iterative search of the algorithm, in each iterative optimization process, the particles mainly guide the flight position and the flight speed of the particles by tracking two parameters, wherein one parameter refers to the best position experienced by the particle individuals in the iterative process and is called as an individual extreme value, and the other parameter refers to the best position experienced by the whole swarm and is also called as a global extreme value.

The particle swarm optimization algorithm is a new bionic intelligent optimization technology. The particle swarm optimization algorithm can converge on a global optimal solution with a high probability, and experimental results show that the particle swarm optimization algorithm is suitable for optimizing in environments such as single-target optimization, constraint target optimization and the like.

The taboo search algorithm is a simulation of human intelligence process, namely, when a person searches an unknown area, the searched area can automatically avoid secondary search, further the whole unknown space can be searched in the shortest time, if the unknown space is not searched, the searched area is searched again, the taboo search algorithm is based on the thought, a plurality of specific search directions (or called movements) are determined as heuristics from an initial solution, then the direction which reduces the fitness function value the most is selected for searching, in order to prevent the algorithm from falling into a local extreme value, the taboo search algorithm adopts a flexible memory technology to record improved movements so as to guide the next search direction of the algorithm, namely the taboo table, the realized movements in the iterative process are saved through the taboo table, and the algorithm is prevented from falling into a loop, resulting in a reduction in search efficiency. While employing the "privileged criteria" allows certain states to be disabled, so that the algorithm has the possibility of accepting some poor quality solutions, preventing the algorithm from falling into local extrema.

The taboo search algorithm is based on neighborhood iterative search, avoids some repeated searches through a single-in and single-out taboo table, and utilizes special privilege criteria (scofflaw criteria) to privilege some good states previously taboo by the taboo table, thereby ensuring the diversity of search. The selection of the neighborhood structure, the acquisition of the initial solution, the determination of the taboo length, the scofflaw criterion and the establishment of the termination criterion all have critical influence on the performance of the taboo search algorithm. The neighborhood function inherits the characteristics of local neighborhood search and is used for performing neighborhood search on the current solution; the flexible use of the tabu table and privilege criteria not only enables the algorithm to avoid repeated searches, but also ensures the efficient search of algorithm diversification, enabling the algorithm to finally realize global optimization.

The denoising method of the fault vibration signal of the aeroengine rotor system comprises the following specific steps:

firstly, measuring rub-impact vibration signals of a certain number of aeroengine rotors in a certain sampling period according to a determined time interval or sampling frequency through an eddy current displacement sensor;

step two, performing repeated filtering processing on the acquired impact vibration signal of the aircraft engine rotor, performing multi-wavelet decomposition and reconstruction on the signal subjected to repeated filtering processing to obtain a multi-wavelet coefficient, wherein the multi-wavelet decomposition and reconstruction are performed according to the following formula,

in the formula, c_j,k＝(c_1,j,k,···,c_r,j,k)^TLow frequency component of dimension r, d_j,k＝(d_1,j,k,···,d_r,j,k)^THigh-frequency components of r dimension are conjugate transpose operation;

step three, setting mixed particlesConstant factor in subgroup optimization algorithm, wherein learning factor c is set₁＝c₂1.4995, the number m of particle groups is 40, the maximum number of iterations T is 200, and the maximum value ω of the weighting coefficients_max0.9, minimum value ω of weight coefficient_minMaximum velocity v of the particle equal to 0.4_max＝0.2·λ_maxMaximum value λ of particle group position parameter λ_maxAnd a minimum value λ_minThe method is carried out by the following formula,

λ_j＝(2·logn_j)/q

in the formula, n_jWhen q is 0.1 and 1 respectively for the length of multiple wavelet coefficients, lambda can be obtained respectively_maxAnd λ_min；

Step four, obtaining an optimal threshold value through a function based on the GCV criterion fitness to obtain the fitness value of each particle in the population, wherein the function based on the GCV criterion fitness is the following formula,

in the formula, D and D_λRespectively a noise-containing multi-wavelet coefficient and a multi-wavelet coefficient processed by a threshold lambda rule, wherein N is the total number of the multi-wavelet coefficients, and N is₀The number of the multi-wavelet coefficient set to 0 in the threshold processing process;

step five, for each particle, the fitness value of each particle and the best position p which the particle has undergone_iIf so, x is compared_iAs the current best position p_iThen, the fitness value is compared with the fitness value of the best position that all particles have undergone; if it is better, it is taken as the best position p of all the particles currently_g；

Step six, after the individual extreme value and the global extreme value are obtained through the step five, the speed and the position of the particles in the population are respectively updated through the following formula,

v_ij(t+1)＝w·v_ij(t)+c₁r₁(t)[p_ij(t)-x_ij(t)]+c₂r₂(t)[p_gj(t)-x_ij(t)]

x_ij(t+1)＝x_ij(t)+v_ij(t+1)

wherein w represents an inertial weight coefficient, c₁And c₂Represents a learning factor (also called acceleration constant), i ═ 1, 2, ·, D, v_ijDenotes the velocity, v, of the particle_ij∈[-v_max，v_max]，v_maxIs a constant; limiting the particle velocity by manual setting, parameter r₁And r₂Is a uniform random number ranging from 0 to 1 to enhance the randomness of the particle search;

step seven, judging whether the algorithm in the step six reaches the maximum iteration times, returning to the step four to continue operation if the algorithm does not reach the maximum iteration times, and outputting an optimal value lambda if the algorithm reaches the maximum iteration times_best；

Step eight, starting a tabu search algorithm, setting the maximum iteration times, setting a null tabu table, and performing the optimal solution lambda generated by the hybrid particle swarm optimization algorithm_bestAs an initial solution of the tabu search algorithm, and assigning the initial solution to the current solution;

generating a plurality of neighborhood solutions by using a current solution neighborhood function, and determining a plurality of candidate solutions from the neighborhood solutions;

tenth, obtaining an optimal threshold of the algorithm based on a GCV criterion fitness function, obtaining a fitness value of each particle in a group, judging whether a current candidate solution meets a privilege criterion, if so, replacing the current solution with the optimal candidate solution meeting the privilege rule, replacing an object entering a privilege table earliest by taking the optimal candidate solution as a privilege object, replacing the current optimal solution with the optimal candidate solution meeting the privilege rule, and then directly skipping to the twelfth operation; if not, then executing step eleven downwards; the function based on the fitness of the GCV criteria is given by,

in the formula, D and D_λAre respectively containing noiseThe multi-wavelet coefficient and the multi-wavelet coefficient after threshold lambda rule processing, wherein N is the total number of the multi-wavelet coefficients, and N is the total number of the multi-wavelet coefficients₀The number of the multi-wavelet coefficient set to 0 in the threshold processing process;

step eleven, when the candidate solution is superior to the current solution and inferior to the current optimal solution, determining the taboo attribute of the candidate solution, if the candidate solution is not in the taboo table, assigning the candidate solution to the current solution of the next iteration, and replacing the object which enters the taboo table earliest by the candidate solution; when the candidate solution is inferior to the current solution and is inferior to the current optimal solution, a new neighborhood solution is generated again by using the current solution;

step twelve, judging whether the algorithm reaches an end condition, if so, outputting an optimal solution and terminating the algorithm; if not, jumping to the step nine to continue operation;

and thirteen, processing the multi-wavelet coefficients according to a threshold rule, and performing multi-wavelet reconstruction on the processed coefficients to obtain a denoising signal of the fault vibration of the rotor system of the aircraft engine.

Through comparative experiments, the results show that: compared with the traditional multi-wavelet adjacent coefficient denoising method, the multi-wavelet adjacent coefficient denoising method based on the hybrid particle swarm optimization has the advantages of obviously improving the background noise denoising effect, being more advantageous than the Db2 single wavelet and GHM multi-wavelet methods in denoising effect, and providing technical support for feature extraction and fault diagnosis in the later period.

Common threshold selection methods in the prior art include a DJ uniform threshold, a confidence interval threshold based on zero-mean normal distribution, an ideal threshold and a GCV threshold. Most of the wavelet threshold denoising algorithms utilize the statistical properties of noise to achieve denoising. In many practical applications, however, a priori knowledge about the noise is not known and needs to be estimated. Instead, the threshold is determined by a GCV function, which relies only on the input and output data, independent of the noise energy and its real data. Moreover, the current experiments prove that the threshold value obtained by utilizing Generalized Cross Validation (GCV) is an asymptotic optimal solution in the sense of least mean square error. Therefore, the threshold is obtained by utilizing the GCV principle, any information of noise does not need to be acquired in advance, and the detail characteristics of the original image can be well maintained while denoising.

Compared with an actual de-drying method, and combined with signal-to-noise ratio (SNR) and Mean Square Error (MSE), the best de-noising effect is the de-noising method provided by the invention, and then the multi-wavelet adjacent coefficient de-noising method is adopted, and the single wavelet threshold value Db2 has the worst de-noising effect; moreover, in terms of the denoising effect obtained by adopting different threshold functions for the same denoising method, the soft threshold denoising effect in each denoising method is obviously superior to that of the hard threshold denoising. From the above, it can be seen that the noise reduction method provided by the present invention has the maximum signal-to-noise ratio (SNR) and the minimum Mean Square Error (MSE) value, which indicates that the method of the present invention has the highest SNR, the low signal distortion degree and the best noise reduction effect.

The comparison shows that the hybrid particle swarm optimization algorithm is simple to operate and has strong global swarm capacity, but the convergence speed and precision are gradually reduced in the later stage of the algorithm, and the algorithm is easy to fall into precocity, so that the algorithm can only obtain suboptimal solution; the taboo search algorithm has stronger dependence on the initial solution in the initial stage, and the quality of the initial solution often directly influences that the algorithm can not converge to the global maximum solution, so the invention combines the taboo search and the particle swarm optimization algorithm in a breakthrough manner, namely, the mixed particle swarm algorithm, the method of the invention assigns the sub-optimal solution found by the particle swarm algorithm through iterative optimization to the initial solution in the taboo search algorithm, then calculates the fitness value, enables the taboo search algorithm to accept some inferior quality solutions with low quality to a certain extent through continuous search of the neighborhood of the initial value and flexible use of the taboo table and the special criterion, prevents the algorithm from falling into a local extreme value, enhances the global optimization capability of the algorithm, effectively solves the problem of slow collection speed in the later stage of the particle swarm algorithm, and comparatively speaking, the convergence speed of the mixed particle swarm algorithm is reduced to a certain extent, but the method can effectively avoid the premature trapping of the population into the local optimal solution, increase the possibility of finding the global optimal solution by the population and improve the optimization performance of the algorithm on the whole.

(3) Advantageous effects

Compared with the prior art, the invention has the beneficial effects that: the denoising method combines the hybrid particle swarm optimization algorithm and the multi-wavelet adjacent coefficient denoising method, carries out threshold optimization on vectorized multi-wavelet decomposition coefficients through the hybrid particle swarm optimization algorithm, can obtain more accurate threshold than the traditional unified threshold method, recovers a one-dimensional denoising signal through the threshold processing and multi-wavelet reconstruction and post-processing, can obviously reduce the interference of background noise, and has excellent denoising and filtering effects.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the technical solutions in the prior art will be briefly described below, it is obvious that the drawings in the following description are only one embodiment of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.

Fig. 1 is a time domain diagram of a rotor imbalance signal and its envelope demodulation spectrum in an embodiment of the present invention.

Fig. 2 is a time domain diagram of a rotor misalignment signal and an envelope demodulation spectrum thereof in the embodiment of the invention.

Fig. 3 is a time domain diagram of a rotor rub-impact signal and an envelope demodulation spectrum thereof in an embodiment of the present invention.

Detailed Description

In order to make the technical means, the original characteristics, the achieved purposes and the effects of the invention easily understood and obvious, the technical solutions in the embodiments of the present invention are clearly and completely described below to further illustrate the invention, and obviously, the described embodiments are only a part of the embodiments of the present invention, but not all the embodiments.

The aircraft engine is a key component of an aircraft, and is known as a 'bright pearl' in the aviation industry, and the maintenance technology of the aircraft engine is classified as one of the most conservative secrets in all aviation countries at present. The aircraft engine has high failure probability due to the precision and complexity of the components, and the failure types are various, and the specific failure types can be seen in the common failure types and the characteristic frequencies of the aircraft engine shown in the following table 1. The rotor system fault is a main cause of shutdown of the aircraft engine, and the characteristic frequency of the rotor system fault is different according to the faults of the aircraft engine rotor system.

TABLE 1 common failure types and characteristic frequencies of aircraft engines

The implementation case is an implementation case of an extraction method of the fault signal characteristic information of the rotor system of the aircraft engine, and comprises the following processes:

A. in a system with possible rotor unbalance, misalignment and rubbing faults of an aircraft engine, data samples are collected when the rotor rotation speed is adjusted to 2700r/min, data with the rotation speed of 2700r/min are obtained through a sensor, the sampling frequency is set to 10240Hz, and signals are discrete vibration acceleration signals collected by the sensor in the horizontal direction at the bearings at the two ends of a rotor fault tester; in view of the possible error of the signals at the head and tail end points, only 4096 data in the middle of the data are adopted during the processing and calculation of the three fault signals, namely the data with stable rotor comprehensive state is adopted.

B. Carrying out time-frequency analysis on the acquired unbalance fault vibration signal of the rotor system to obtain a time domain diagram of the unbalance signal of the rotor and an envelope demodulation spectrum thereof shown in figure 1; in order to more clearly and conveniently identify the existence characteristics of a large number of interference components such as fault characteristic frequency, noise and the like in the unbalanced signal, envelope analysis is performed on the signal to obtain an envelope demodulation spectrum of the unbalanced fault signal, as shown in (b) in fig. 1; according to the unbalance failure mechanism of the rotor system of the aircraft engine, the vibration characteristics of the unbalance of the rotor are represented as follows: the fundamental frequency (frequency conversion) vibration is obvious, and no other obvious characteristic frequency exists, so that the low-frequency part in the envelope demodulation spectrum of the unbalanced signal is locally amplified, as shown in (c) in fig. 1.

It can be seen from the time domain waveform diagram of the unbalanced fault signal of the aircraft engine rotor system in the following figure that the time domain waveform is disordered, has no obvious periodic impact generated by the unbalanced fault of the rotor system and is covered by strong background noise, while it can be seen from (c) in fig. 1 of the envelope demodulation spectrum that the fundamental frequency impact response of the rotating speed frequency exists in the low frequency band (frequency range) of the unbalanced fault vibration signal, i.e. the characteristic frequency of the unbalanced fault, but at the same time, it can be seen that a large amount of noise interference frequency components exist around the characteristic frequency band cluster, so that the noise reduction processing needs to be performed on the unbalanced fault signal.

Then, carrying out noise reduction analysis on the rotor system unbalance fault vibration signals by respectively adopting Db2 single wavelets, GHM multi-wavelets and multi-wavelet adjacent coefficients and the multi-wavelet adjacent coefficient noise reduction method based on hybrid particle swarm optimization provided by the invention; it can be seen that the time domain signals after the noise reduction processing is performed on the unbalanced fault signals of the rotor system by the four noise reduction methods are clearer than the time domain waveforms of the original unbalanced fault vibration signals, and it is indicated that the four methods all have certain noise reduction effects, so that the comparison and analysis of the noise reduction effects of the four methods by the envelope demodulation spectrum can see that the four noise reduction methods all have fundamental frequency impulse response of the rotating speed frequency in the low frequency band (frequency range) of the unbalanced fault vibration signals, and the noise background signals around the fundamental frequency are suppressed to a certain extent compared with the original unbalanced fault signals. However, by comparing the four denoising methods, it can be known that some small noise signal components still exist around the fundamental frequency of the multi-wavelet neighbor coefficient denoising method based on the hybrid particle swarm optimization, but the denoising method provided herein is more significantly reduced than the Db2 single wavelet, the GHM multi-wavelet and the multi-wavelet neighbor coefficient denoising method no matter the number of noise subharmonics and the size of the peak, which shows that the denoising method provided herein can achieve effective filtering of rotor imbalance fault signals, and the denoising effect is more advantageous than the other three denoising methods.

Similarly, the aero-engine rotor system is utilized to perform the misalignment fault experiment, and the acquired misalignment vibration signal of the rotor system is subjected to time-frequency analysis to obtain a rotor misalignment signal time domain diagram and an envelope demodulation spectrum thereof shown in the following fig. 2; the mechanism of the misalignment fault of the rotor system of the aircraft engine is known, and the vibration characteristics of the misalignment fault of the rotor are mainly represented as follows: even-order frequency multiplication such as 2 frequency multiplication and 4 frequency multiplication is obvious, and for this reason, local amplification processing is performed on a high-frequency part in a signal envelope demodulation spectrum, as shown in (c) in fig. 2.

The noise reduction effect of the four denoising methods is contrastively analyzed through the envelope demodulation spectrum, as can be seen from (c) in fig. 2 of the envelope demodulation spectrum: in a high frequency band (frequency range) of the misaligned fault vibration signal, four noise reduction methods can easily find a fundamental frequency impact response of a rotating speed frequency, and the fundamental frequency impact response comprises a frequency multiplication impact response and a frequency multiplication impact response with a certain impact amplitude.

And finally, performing time-frequency analysis on the rub-impact vibration signal acquired by the rub-impact fault experiment of the rotor system to obtain a rotor rub-impact signal time-domain diagram and an envelope demodulation spectrum thereof as shown in fig. 3. Periodic impacts in the original rub-on fault signal are masked from recognition by strong background noise, and can be seen in fig. 3 (c) of the envelope demodulation spectrum: in the low frequency range (frequency range) of the rub-impact fault vibration signal, there is a fundamental frequency impulse response of the rotational speed frequency and an impulse amplitude which is not a particularly pronounced frequency-doubled impulse response, because of the strong noise interference of the low frequency range of the rotor rub-impact signal. And by (d) in the envelope demodulation spectrogram 3, it can be seen that: four groups of frequency band clusters with equal intervals can be seen in a high frequency band (frequency range) of the rub-impact fault vibration signal, the interval frequency is 45Hz, but a large amount of noise interference frequency components exist around the four groups of frequency band clusters with equal intervals.

The analysis shows that the time domain signal after the noise reduction processing is clearer than the time domain waveform of the multiple wavelet de-noising of GHM, the periodic impact is obvious, a certain de-noising effect is shown, and the time domain signal can be seen through each envelope demodulation spectrum region: fundamental frequency impulse response and impulse response of rotating speed frequency exist in a low frequency band (frequency range) of a noise reduction signal, and wave crest obvious interference components do not exist around the fundamental frequency, which shows that noise background signals in the low frequency band are effectively suppressed, while frequency multiplication impulse response is not obvious and cannot be effectively identified as can be seen from Db2 single wavelets, GHM multiple wavelets and signals subjected to noise reduction by multiple wavelet adjacent coefficients, and meanwhile, in a high frequency band (frequency range) of the noise reduction signal, final dryness removal result analysis can be obtained, namely, after noise reduction processing is respectively carried out on unbalanced, misaligned and rubbing fault signals of a rotor system containing noise, the influence of background noise interference components in the signals can be effectively reduced by four noise reduction methods, and the signal-to-noise ratio of the fault signals is improved. However, the noise reduction method provided by the invention is superior to other three-medium noise reduction methods in the suppression of noise subharmonics in different frequency bands, so that the noise reduction effect of the multi-wavelet adjacent coefficient noise reduction method based on the hybrid particle swarm optimization is better, and the effective noise reduction of rotor fault signals can be realized.

Having thus described the principal technical features and basic principles of the invention, and the advantages associated therewith, it will be apparent to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, but is capable of other specific forms without departing from the spirit or essential characteristics thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

Furthermore, it should be understood that although the present description is described in terms of various embodiments, not every embodiment includes only a single embodiment, and such descriptions are provided for clarity only, and those skilled in the art will recognize that the embodiments described herein can be combined as a whole to form other embodiments as would be understood by those skilled in the art.

Claims

1. A denoising method for a fault vibration signal of an aircraft engine rotor system is characterized by comprising the following specific steps:

step three, setting constant factors in the hybrid particle swarm optimization algorithm, wherein learning factors c are set₁＝c₂1.4995, the number m of particle groups is 40, the maximum number of iterations T is 200, and the maximum value ω of the weighting coefficients_max0.9, minimum value ω of weight coefficient_minMaximum velocity v of the particle equal to 0.4_max＝0.2·λ_maxMaximum value λ of particle group position parameter λ_maxAnd a minimum value λ_minThe method is carried out by the following formula,

λ_j＝(2·logn_j)/q

x_ij(t+1)＝x_ij(t)+v_ij(t+1)

wherein w represents an inertial weight coefficient, c₁And c₂Represents a learning factor (also called acceleration constant), i ═ 1, 2, ·, D, v_ijDenotes the velocity, v, of the particle_ij∈[-v_max，v_max]，v_maxIs a constant number r₁And r₂Are all uniform random numbers between 0 and 1, r₁And r₂To enhance the randomness of particle search;