CN-115541233-B - Method for extracting characteristic frequency of rotating machinery under strong interference

CN115541233BCN 115541233 BCN115541233 BCN 115541233BCN-115541233-B

Abstract

The invention belongs to the field of big data learning models, and provides a method for extracting characteristic frequency of rotating machinery under strong interference, which is used for roughly positioning an interval where an optimal balance parameter alpha optimum for the production of a product is in VMD decomposition based on set reference mode correlation, and carrying out local optimization on the balance parameter alpha of each decomposition by combining a sparrow search optimization algorithm. The invention determines the decomposition modulus K through the self-adaption of the iterative decomposition times, and the adopted improved recursive VMD method avoids the influence of inaccurate preset decomposition numbers and balance parameters on the decomposition effect in advance. The invention is applied to the constructed simulation signal to realize the optimal decomposition effect which can be achieved by the VMD method, and is simultaneously applied to the pump cavitation flow induced vibration signal processing to successfully realize the effective extraction of the fluid mechanical flow induced vibration characteristic frequency under the condition of low signal-to-noise ratio.

Inventors

TONG ZHEMING
LIU HAO

Assignees

浙江大学

Dates

Publication Date: 20260505
Application Date: 20220819

Claims (9)

1. The method for extracting the characteristic frequency of the rotating machinery under the strong interference is characterized by comprising the following steps of: S1, setting an acquired vibration signal as a residual signal, and executing VMD operation on the residual signal; S2, setting a reference mode u ref and calculating the kurtosis of the reference mode u ref S3, setting calibration test points of the balance parameter alpha, carrying out VMD decomposition on each calibration test point, and calculating related parameters between a decomposition mode u' and a signal sequence u ref of a reference mode; s4, judging the correlation coefficient The number of calibration test points of the balance parameter alpha is reduced, the interval where the optimal balance parameter alpha optimum for the production of a product is located is further reduced, and unnecessary VMD operation is reduced; S5, refining the optimizing interval of the balance parameter alpha according to the characteristics of the decomposed signal; s6, determining an optimized fitness function; S7, selecting an optimal balance parameter alpha optimum for the production of a product of the target mode by adopting a sparrow search optimization algorithm in the optimization range of the thinned balance parameter alpha; S8, performing VMD operation by using the value of the optimal balance parameter alpha optimum for the production of a product searched in the step S7 and setting a decomposition module K=1, and extracting a unique decomposition mode as u e '; s9, removing the unique decomposition mode u e ' extracted in the step S8 by the residual signal to form a new residual signal; S10, carrying out signal reconstruction by using the unique decomposition mode u e ' extracted in the step S8; And S11, judging whether to continue the next iteration of modal decomposition or not based on the power spectrum similarity between the reconstructed signal and the original signal.
2. The method for extracting characteristic frequencies of a rotating machine under strong interference according to claim 1, wherein the step of setting the reference mode u ref in the step (2) is as follows: initializing the value of a balance parameter alpha Initially, the method comprises to be c, setting a decomposition modulus K to be 1, and taking the extracted unique mode as a reference mode u ref ; Kurtosis of reference mode u ref The calculation formula of (2) is as follows: Wherein, u ref , N, The signal sequence of the reference mode, the sampling number of the mode and the time domain signal average value of the reference mode are respectively obtained.
3. The method for extracting the characteristic frequency of the rotating machine under the strong interference according to claim 1, wherein the specific process of the step S3 is as follows: when the value of the initialized balance parameter alpha Initially, the method comprises is set as c, the influence of the value of the balance parameter alpha on the signal decomposition result is larger when the value of the balance parameter alpha is within c, so that a larger balance parameter alpha interval is selected within a range from c to 10 x c, a smaller balance parameter alpha interval is selected within c, when the values of different balance parameters alpha are selected for VMD decomposition each time to form a calibration set alpha ', the decomposition modulus K is always set as 1, the decomposition mode is marked as u ', and the related parameters of the decomposition mode u ' are calculated after each decomposition, wherein the method comprises the following steps: Correlation coefficient of decomposition mode u' and reference mode Kurtosis Kur u′ of decomposition mode u': Envelope spectral kurtosis ESK u′ of decomposition modality u': kurtosis of the decomposition modality u' and envelope entropy ratio KSE u′ : Wherein, the In is the logarithm of the base e; Is the average value of the reference mode u ref signal sequence; Is the average value of the decomposition modes u'; ES u′ (ω) is the envelope spectrum of u' at frequency ω E u′ (j) is a discrete point of the envelope signal sequence E u′ obtained by the analysis mode u' after Hilbert demodulation; P j is a normalized version of E u′ (j); J represents the sampling point of the envelope signal sequence E u′ ; j is the sample point in the envelope signal sequence E u′ at the value of j.
4. The method for extracting a characteristic frequency of a rotating machine under strong interference as claimed in claim 3, wherein the process of narrowing the interval in which the optimal balance parameter α optimum for the production of a product is located in step S4 is as follows: after VMD operation is performed on the calibration test point of a certain balance parameter alpha, if the correlation coefficient If the balance parameter alpha is greater than 0.95, the calibration test point of the balance parameter alpha is reserved, otherwise, the decomposed signal is considered to be completely different from the center frequency owned by the reference mode, the calibration test point of the balance parameter alpha is used as one boundary of the refined balance parameter alpha optimizing interval, VMD operation is not carried out on other calibration test points crossing the boundary, and the rest of calibration test points in the boundary form a new optimizing interval; the calibration test points of the balance parameter alpha are removed from the calibration test points with the value c close to the balance parameter alpha until the test points in S3 are satisfied Stopping.
5. The method for extracting characteristic frequency of rotating machinery under strong interference according to claim 1, wherein the rule of refining the optimizing interval of the balance parameter α in step S5 is as follows: Kurtosis of reference mode u ref The extracted decomposition mode u 'is considered to have obvious impact characteristics, and the envelope spectrum kurtosis KSE u′ of the decomposition mode u' corresponding to the calibration test points of the balance parameters alpha remained in the boundary is searched for the maximum value; Kurtosis of reference mode u ref The maximum value of the envelope spectrum kurtosis ESK u′ of the decomposition mode u 'corresponding to the calibration test points of a plurality of balance parameters alpha remaining in the boundary is searched, the value of the balance parameter alpha corresponding to the maximum value is taken as the center, one bit is moved leftwards and rightwards in the calibration test point set alpha', and the values of the two corresponding balance parameters alpha form the optimizing interval of the balance parameter alpha after final refinement.
6. The method for extracting the characteristic frequency of the rotating machine under the strong interference according to claim 1, wherein the specific process of the step S6 is as follows: If it is The extracted decomposition mode u' is considered to have obvious impact characteristics, and the maximum KSE u′ is selected as an adaptive function for optimizing; If it is The maximum ESK u′ is selected as an optimized fitness function, and the optimized fitness function fitness ESK is as follows: Wherein ESK (u k ) is the envelope spectrum kurtosis corresponding to a decomposition mode u k obtained by optimizing in the optimizing space, and u k is a decomposition mode obtained by optimizing in the refined optimizing space; ESK (u k ) is the ratio of kurtosis corresponding to the decomposition mode u k obtained by optimizing in the optimizing space to the envelope entropy.
7. The method for extracting the characteristic frequency of the rotating machine under the strong interference according to claim 1, wherein the specific process of the step S7 is as follows: s71, randomly initializing sparrow population, defining related parameters, and defining maximum iteration times; wherein d represents the dimension of the optimization problem variable, s is the number of sparrows, X represents the position of the sparrow population; s72, calculating fitness of the initial population, sequencing the fitness of the initial population and then selecting a current optimal value and a worst value; Wherein fun is determined by the fitness function in step S6; S73, updating the position of the finder, wherein the formula is as follows: Wherein m represents an mth sparrow; t represents the current iteration number, n=1, 2,3,; the item max is the set maximum iteration number, X m,n represents the position information of the mth sparrow in the nth dimension, sigma epsilon (0, 1) is a random number, R 2 and ST respectively represent an early warning value and a safety value, wherein R 2 epsilon (0, 1), ST epsilon (0.5, 1), Q represents a random number obeying normal distribution, and L represents a1×d matrix, wherein each element in the matrix is all 1; When R 2 < ST, indicating that there are no predators around the foraging environment at this time, the discoverer can search over a larger range; when R 2 is more than or equal to ST, indicating that part of sparrows in the population find predators and giving an alarm to other sparrows in the population, and all the sparrows need to fly to other safe places to find food quickly at the moment; S74, updating the position of the joiner, wherein the formula is as follows: Wherein X p is the optimal position occupied by the current finder, X worst is the current global worst position, A is a 1×d matrix, wherein each element is randomly assigned a value of 1 or-1, and A + ＝A T (AA T ) -1 is specified; When m is greater than n/2, the mth participant with lower fitness value needs to fly to other places in time to find food so as to obtain more energy; And S75, updating the sparrow position which is aware of danger, wherein the formula is as follows: Wherein X best is the current global optimal position, beta is a random number between 0 and 1 obeying normal distribution as step control parameter, Q is a random number from-1 to 1, fun m is the fitness value of the current sparrow, fun g and fun w are the current global optimal and worst fitness values respectively, epsilon is the minimum constant to avoid zero denominator; When fun i >fun g indicates that sparrows are at the edge of the population and are extremely vulnerable to predators, X best indicates the safest location in the population, and fun i ＝fun g indicates that sparrows in the middle of the population are aware of danger and need to be close to other sparrows to minimize the risk of being predated; and S76, obtaining a current optimal value, if the current optimal value is better than the optimal value of the last iteration, performing updating operation, otherwise, not performing updating operation, continuing iteration operation until the condition is met, and finally obtaining a global optimal value and an optimal fitness value.
8. The method for extracting a characteristic frequency of a rotating machine under strong interference according to claim 1, wherein the signal reconstruction function f' is:
9. the method for extracting a characteristic frequency of a rotating machine under strong interference according to claim 1, wherein the judgment in step S11 is based on: Setting the power spectrum of the original signal as PS f , and the formula is: PS f ＝|f| 2 ; Setting the power spectrum of the reconstructed signal as PS f′ , wherein the formula is as follows: PS f′ ＝|f′| 2 ; Wherein f is the original signal sequence; power spectral similarity coefficient between reconstructed signal and original signal The method comprises the following steps: Wherein, the Is the average of the original signal power spectrum; i is the total number of iterations; q is the number of iterations; If it is If the iteration is greater than the threshold value of 0.9, the iteration is stopped; If it is And if the value is less than or equal to the threshold value of 0.9, iteration is continued, and the step S1 is returned to continue.

Description

Method for extracting characteristic frequency of rotating machinery under strong interference Technical Field The invention belongs to the field of big data learning models, and particularly relates to a method for extracting characteristic frequency of rotating machinery under strong interference. Background Pumps, hydraulic turbines, propellers and the like are typical hydraulic rotating machinery, and flow-induced faults are inevitably generated in the operation engineering of the hydraulic rotating machinery. Abnormal flow such as tip vortex cavitation, axial flow pump blade tip clearance cavitation, turbine draft tube vortex strip and the like generated by the propeller usually induces accompanying phenomena such as vibration and the like. The vibration signal carries a large amount of flow induced failure information. Characteristic signals such as axial frequency, leaf frequency and the like excited in the operation process of the hydraulic rotary machine have typical low-frequency characteristics. Under complex operating conditions, these features are often contaminated with intense background noise and are superposition of multifrequency feature information. Therefore, it is necessary to extract the fault signature by signal decomposition. How to accurately realize the demodulation of vibration signals under the condition of low signal-to-noise ratio is a key step of the fault diagnosis and target identification of the hydraulic rotating machinery. Various methods have been devised for this problem, such as wavelet decomposition, wavelet packet decomposition, empirical Mode Decomposition (EMD), and Local Mean Decomposition (LMD), etc. However, wavelet decomposition and wavelet packet decomposition are non-adaptive signal analysis methods because the wavelet basis functions are pre-selected. While EMD and LMD are adaptive signal processing methods, their application is limited due to the presence of mode mixing. Noise auxiliary technologies such as integrated EMD and integrated LMD alleviate the mode mixing problem to a certain extent, but the computational complexity is increased sharply, and the added white noise cannot be eliminated effectively The Variational Modal Decomposition (VMD) is a novel adaptive signal decomposition method proposed in recent years. By means of the advantages of strict mathematical theory guidance, high convergence speed, strong noise robustness, effective avoidance of end-point effects, over-envelope, under-envelope and the like in decomposition, the method is widely applied to the field of fault diagnosis. However, the effect of VMD decomposition is highly dependent on the decomposition modulus K and the balance parameter alpha. There is currently no unified way of determining these two parameters. Although some studies have proposed improved VMD methods that combine optimization algorithms and fitness functions, most of them optimize alpha and K simultaneously, resulting in all sub-modes sharing the same alpha value. Because different sub-modes have different bandwidth characteristics, sharing alpha results in the occurrence of unreasonable under-decomposition or over-decomposition. Disclosure of Invention In order to solve the problems that the number of decomposition modes is difficult to set and all modes share balance parameters in VMD application, the invention adaptively determines the decomposition modulus K by referring to the EMD recursion decomposition idea, combines a sparrow search optimization algorithm to locally optimize the balance parameter a of each decomposition, and provides a fluid mechanical flow induced vibration demodulation method for optimizing the recursion VMD on the basis of the combination of the two. In order to achieve the above object, the present invention provides the following technical solutions: the invention provides a method for extracting characteristic frequency of rotating machinery under strong interference, which comprises the following steps: S1, setting an acquired vibration signal as a residual signal, and executing VMD operation on the residual signal; S2, setting a reference mode u ref and calculating the kurtosis of the reference mode u ref S3, setting calibration test points of the balance parameter alpha, carrying out VMD decomposition on each calibration test point, and calculating related parameters between a decomposition mode u' and a signal sequence u ref of a reference mode; s4, judging the correlation coefficient The number of calibration test points of the balance parameter alpha is reduced, the interval where the optimal balance parameter alpha optimum for the production of a product is located is further reduced, and unnecessary VMD operation is reduced; S5, refining the optimizing interval of the balance parameter alpha according to the characteristics of the decomposed signal; s6, determining an optimized fitness function; s7, selecting an optimal balance parameter alpha optimum for the production of