Fan main shaft bearing fault diagnosis method based on IPSO-VMD

By improving the particle swarm algorithm to optimize the parameters of the VMD model, the diagnosis errors caused by the number of modal components and improper selection of punishment factors in the fan spindle bearing fault diagnosis are solved, and the accuracy and efficiency of diagnosis are improved.

CN119086066APending Publication Date: 2024-12-06CENT CHINA BRANCH OF CHINA DATANG CORP SCI & TECH RES INST CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410878486.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-02
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

In the existing fan spindle bearing fault diagnosis methods, the empirical setting of the number of modal components k and the penalty factor α can easily lead to modal aliasing and false components, resulting in diagnostic errors, and improper parameter selection leads to low recognition efficiency.

Method used

The improved particle swarm algorithm (IPSO) is used to optimize the parameters of the VMD model globally. By increasing the particle perturbation factor, the local optimization is avoided, and the combination of the number of modal components k and the penalty factor α is optimized to ensure the optimal parameter combination of the model. In turn, improve the accuracy and generalization ability of diagnosis.

Benefits of technology

It effectively avoids local optimal solutions, improves the generalization ability of the model, improves the accuracy and identification efficiency of fan spindle bearing fault diagnosis, and reduces the rate of misdiagnosis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119086066A_ABST
    Figure CN119086066A_ABST
Patent Text Reader

Abstract

The invention relates to a fan main shaft bearing fault diagnosis method based on IPSO-VMD, and the method employs an improved particle swarm optimization algorithm to carry out the global optimization of the parameters of a VMD model, prevents the later parameters of the algorithm from falling into local optimum, increases a particle disturbance factor, avoids the random selection of the number K of modal components and a penalty factor alpha according to experience in the prior art, and improves the fault diagnosis precision. According to the method, the problems that the optimal parameter combination of the VMD model is difficult to obtain and the model diagnosis fails are solved, the improved particle swarm algorithm is adopted to optimize the VMD model through global optimization given parameter combination, and finally the optimal parameter combination is found out, so that a local optimal solution caused by a random parameter optimization algorithm in the prior art is avoided, and the generalization ability of the model is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a fan main shaft bearing fault diagnosis method based on IPSO-VMD, which can be used for diagnosing the fault type of the fan main shaft bearing. Background Art

[0002] The main shaft bearing of a wind turbine is a key component in a wind turbine generator set. It bears the huge load and complex working environment of the wind turbine generator set, so it is very easy to have wear and other faults. Once there is a problem with the main shaft bearing, it will not only lead to a significant decrease in power generation efficiency, but also threaten the personal safety of the operating personnel. In addition, the cost of replacing bearings is high, so the implementation of effective online monitoring and fault diagnosis of the main shaft bearing has become a key measure to ensure the stable and efficient operation of wind turbines. Variational Mode Decomposition (VMD) has high efficiency in processing non-stationary and nonlinear vibration signals of bearings, and has adaptive decomposition characteristics, so it is widely used in the field of fault diagnosis. The two important parameters in VMD, the number of modal components k and the penalty factor α, are mostly set based on experience, and these two parameters often affect the final decomposition results of the model. When k and α are not selected properly, modal aliasing and false components are prone to occur, resulting in model diagnosis errors. Improved Particle Swarm Optimization (IPSO) is an algorithm proposed to address some shortcomings of particle swarm optimization, such as premature maturity and low convergence accuracy. It is an improved algorithm based on the PSO algorithm. Summary of the invention

[0003] In view of the above situation, in order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a fan main shaft bearing fault diagnosis method based on IPSO-VMD, which uses an intelligent algorithm to optimize the number of modal components k and the penalty factor α, and can effectively solve the problem of low efficiency of existing main shaft bearing fault diagnosis methods.

[0004] The technical solution adopted by the present invention is:

[0005] A fan main shaft bearing fault diagnosis method based on IPSO-VMD, characterized by comprising the following steps:

[0006] S1: Obtain the bearing vibration signal of the fan main shaft and use it as the input information of the VMD model;

[0007] S2: The improved particle swarm algorithm is used to determine the optimal parameter combination [α, K] of the VMD model, and the optimal parameter combination is substituted into the VMD model. The VMD model is used to decompose the bearing vibration signal to obtain several modal components;

[0008] S3: Calculate the relevant kurtosis value of each of the modal components, and select the modal component with the largest kurtosis value;

[0009] S4: performing envelope spectrum demodulation on the modal component with the largest kurtosis value, obtaining and analyzing the true characteristic frequency of the bearing, and then determining the bearing fault type.

[0010] Compared with the prior art, the present invention has the following advantages:

[0011] 1. The present invention provides a fan main shaft bearing fault diagnosis method based on IPSO-VMD parameter optimization, which adopts an improved particle swarm algorithm to globally optimize the parameters of the VMD model. In order to avoid the parameters in the later stage of the algorithm falling into the local optimum, a particle disturbance factor is added, which avoids the problem of difficulty in obtaining the optimal parameter combination of the VMD model and causing model diagnosis failure by randomly selecting the number of modal components K and the penalty factor α based on experience in the prior art.

[0012] 2. The present invention provides a fan main shaft bearing fault diagnosis method based on IPSO-VMD parameter optimization, which uses an improved particle swarm algorithm to optimize the VMD model by globally optimizing a given parameter combination, and finally finds the optimal parameter combination, avoiding the local optimal solution caused by the random parameter optimization algorithm in the prior art and improving the generalization ability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is the improved particle swarm algorithm of the present invention, namely, the IPSO algorithm process.

[0014] Figure 2 The figure is a flow chart of the diagnostic method of the present invention.

[0015] Figure 3 The present invention applies the exception circle fault signal frequency diagram.

[0016] Figure 4 This is the envelope spectrum decomposition diagram of the application example of the present invention. DETAILED DESCRIPTION

[0017] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples.

[0018] like Figure 1-2 As shown, the present invention provides a fan main shaft bearing fault diagnosis method based on IPSO-VMD, comprising the following steps:

[0019] S1: Obtain the bearing vibration signal of the fan main shaft and use it as the input information of the VMD model;

[0020] S2: The improved particle swarm algorithm is used to determine the optimal parameter combination [α, K] of the VMD model, and the optimal parameter combination is substituted into the VMD model. The VMD model is used to decompose the bearing vibration signal to obtain several modal components;

[0021] S3: Calculate the relevant kurtosis value of each of the modal components, and select the modal component with the largest kurtosis value;

[0022] S4: performing envelope spectrum demodulation on the modal component with the largest kurtosis value, obtaining and analyzing the true characteristic frequency of the bearing, and then determining the bearing fault type.

[0023] To ensure the use effect, the step S2 specifically includes the following steps:

[0024] S2.1: The selection range of initialization parameters, that is, the selection range of the initialization decomposition layer number k and the penalty factor α;

[0025] S2.2: Improve the traditional particle swarm algorithm by adding a particle perturbation factor in the second half of the algorithm, giving particles a certain probability to mutate, and avoiding particles from falling into the local optimum. The specific improvement methods are as follows:

[0026] A particle x in a known particle group i , there are two basic information: the current velocity v of the particle i =(v i1 ,K,v id ,K,v im ), and the particle's current position x i =(x i1 ,K,x id ,K,x im ), the two basic information reflect the direction of particle search and the position of the particle search. The quality of the particle is evaluated by the fitness function, which is the objective function of the optimization problem. The particle position information x i This is the input of the objective function, and the fitness is used to determine the quality of the particle's position in the solution space. The optimal historical position of the particle is p i =(p i1 ,K,p id ,K,p im ), the optimal position of the particle group g=(g 1 ,K,g d ,K,g m ). The basic particle swarm algorithm speed and position update formula is as follows:

[0027]

[0028] Where: c 1 、c2 is the acceleration factor, usually 2; ω is the inertia coefficient; r 1 and r 2 is a random number uniformly distributed in the interval [0,1]. Generally speaking, the first term of the speed update formula is called the inertia term, which is the influence of the original search speed of the individual particle; the second term is called the individual cognition term, which is the influence of the individual search history of the particle on the particle; the third term is called the group cognition term, which is the influence of the optimal position of the population on the particle.

[0029] In the second half of the algorithm, the PSO algorithm is prone to fall into the local optimum. In order to make the particles jump out of their current positions and continue searching other areas, a certain probability is given to the particles to mutate in the second half of the algorithm iteration. In the second half of the algorithm, the particle inertia coefficient is small and the particle local search ability is strong. The mutated particles have the opportunity to conduct a detailed search of the nearby area and will not quickly return to the optimal position found by the algorithm in the first half. The specific conditions are set as follows:

[0030] In the second half of the algorithm, the iteration interval [k a ,k b ]superior( where k a is the starting point of the second half of the iteration interval, k b is the end point of the second half of the iteration interval, k max For all particles, if the fitness of the particle (i.e., the extreme value of the objective function) does not change significantly between three generations, a random number pm of [0,1] is generated. If pm < 0.5, the particle is mutated when the velocity is updated; otherwise, the particle velocity update formula with only the adaptive inertia coefficient is used. The judgment condition is calculated as follows:

[0031] Let particle x i Fitness function fit(x i ) Change value among three generations

[0032] Where: k is the number of algorithm iterations;

[0033] Particle x i The probability of randomly generated mutations in a certain iteration:

[0034] pm=rand(0,1)

[0035] Where: rand(0,1) represents a random number in the interval [0,1];

[0036] If in the iteration interval [k a ,k b ], particle x i The corresponding G i <0.01, and the random number pm generated at this time is <0.5, this particle xi Use the velocity update formula with adaptive inertia coefficient and variation term:

[0037]

[0038] Where: c 3 is a constant, and the maximum particle speed can be taken; r 3 is a random number uniformly distributed in the interval [-1,1]; ω n is the adaptive inertia coefficient;

[0039] If the above conditions are not met, the particle uses the original update formula with adaptive inertia coefficient:

[0040]

[0041] The step S4 is specifically as follows:

[0042] Perform envelope spectrum demodulation on the modal component with the largest kurtosis value to obtain the true characteristic frequency of the bearing, and compare the true characteristic frequency of the bearing with the fault characteristic frequency corresponding to each bearing fault type. When the true characteristic frequency of the bearing is consistent with the fault characteristic frequency corresponding to one or more bearing fault types, it is determined that the bearing has a fault of that bearing fault type.

[0043] The following are application examples of the method of the present invention:

[0044] The fault diagnosis model is verified for the main shaft bearing of a 2.0MW direct-drive wind turbine generator set in a wind farm. The bearing model is SKF6205-2RS, the number of rolling elements is 9, the sampling frequency is 12KHz, the motor speed is 1797rpm, and the nominal inner diameter, nominal outer diameter, ball nominal diameter, nominal width and pitch circle diameter are 25mm, 52mm, 7.938mm, 15mm and 39mm respectively. First, the theoretical fault frequency of each part of the bearing is calculated, and the calculation formula is as follows:

[0045] Cage failure frequency f t =n / 120(1-D / dcosα)

[0046] Outer ring fault frequency f o =n / 120(1-D / dcosα)Z

[0047] Inner ring fault frequency f i =n / 120(1+D / dcosα)Z

[0048] Rolling element failure frequency f b =nd / 120D[1-(D / d) 2 cos 2 α]

[0049] Where: the number of rolling elements is Z, n is the bearing speed, D is the rolling element diameter, d is the outer ring diameter and α is the contact angle. Substituting the above bearing parameters into the above formula, the theoretical failure frequencies of the inner ring, outer ring and rolling element are 162.18 Hz, 107.36 Hz and 141.16 Hz respectively.

[0050] Next, IPSO is used to optimize the VMD parameter combination [α, K], using the amplitude spectrum as the fitness function. The IPSO initial parameter c 1 、c 2 The values ​​of K = 6 and α = 1992 are respectively 1.2 and 1.5, the population size is 30, the number of iterations is 200, and 10 groups of wind turbine main shaft bearing outer ring fault vibration signals are processed respectively. The results of multiple optimizations are obtained and the average value is taken as the final result. The final optimized result is K = 6, α = 1992. Taking the outer ring fault as an example, the outer ring fault feature extraction of the bearing is performed, and the obtained frequency diagram is as follows Figure 3 shown.

[0051] In the rolling bearing signal diagram, the outer ring fault signal waveform shows obvious impact components, and the intervals are relatively uniform. Figure 3 The spectrum shows that there is an amplitude peak in the frequency range of 3000-4000Hz, and no resonance frequency band is found in the low frequency band, making it difficult to make a clear judgment on the bearing status. VMD is used to decompose the bearing outer ring fault signal. The parameters [α, K] in VMD are selected after IPSO optimization and the values ​​are K = 6 and α = 1992. The bearing fault signal is decomposed into 6 intrinsic mode functions IMF by VMD. Each intrinsic mode function contains different fault information. It is necessary to select a suitable evaluation index to select the IMF containing the most fault information.

[0052] Kurtosis can be used to indicate the peak degree of the signal waveform. The kurtosis value of the vibration signal of the fan bearing in the healthy state and the fault state is completely different. The vibration signal of the fan bearing with fault will be enhanced due to its impact component, and the kurtosis value will also increase. Therefore, the intrinsic mode function contains more fault information. The kurtosis values ​​of each modal component after IPSO-VMD decomposition are shown in Table 1.

[0053] Table 1 Kurtosis values ​​of each modal component obtained by IPSO-VMD

[0054] Quantity IM1 IM2 IM3 IM4 IM5 IM6 Kurtosis 2.97 3.85 3.46 3.06 2.88 2.71

[0055] From Table 1, we can see that the kurtosis value corresponding to IM2 is the largest, so the envelope spectrum of this component is decomposed, and the results are shown in Figure 4 .

[0056] from Figure 4It can be seen that the envelope spectrum of the IMF2 component contains a fault characteristic main frequency of 106.9Hz, and a small amount of double frequency 215.3Hz, which is basically consistent with the above-calculated rolling bearing outer ring fault characteristic frequency 107.36Hz and its double frequency, and the other frequency components account for a very small proportion and can be ignored. Due to a small error in the calculation process, the 106.9Hz finally obtained in this example can correspond to the bearing outer ring fault characteristic frequency, thereby determining that the bearing outer ring has a fault.

Claims

1. A fan main shaft bearing fault diagnosis method based on IPSO-VMD, characterized in that: The following steps are involved: S1: Obtain the bearing vibration signal of the fan main shaft and use it as the input information of the VMD model; S2: The improved particle swarm algorithm is used to determine the optimal parameter combination [α, K] of the VMD model, and the optimal parameter combination is substituted into the VMD model. The VMD model is used to decompose the bearing vibration signal to obtain several modal components; S3: Calculate the relevant kurtosis value of each of the modal components, and select the modal component with the largest kurtosis value; S4: performing envelope spectrum demodulation on the modal component with the largest kurtosis value, obtaining and analyzing the true characteristic frequency of the bearing, and then determining the bearing fault type.

2. The wind turbine main shaft bearing fault diagnosis method based on IPSO-VMD according to claim 1 is characterized in that: The step S2 specifically includes the following steps: S2.1: The selection range of initialization parameters, that is, the selection range of the initialization decomposition layer number k and the penalty factor α; S2.2: Improve the traditional particle swarm algorithm by adding a particle perturbation factor in the second half of the algorithm, giving particles a certain probability to mutate, and avoiding particles from falling into the local optimum. The specific improvement method is as follows: A particle x in the known particle group i , there are two basic information: the current velocity v of the particle i =(v i1 ,K,v id ,K,v im ), and the particle's current position x i =(x i1 ,K,x id ,K,x im ), the two basic information reflect the direction of particle search and the position of the particle search. The quality of the particle is evaluated by the fitness function, which is the objective function of the optimization problem. The particle position information x i It is the input of the objective function. The fitness is used to determine the quality of the particle's position in the solution space. The best historical position of the particle is p i =(p i1 ,K,p id ,K,p im ), the optimal position of the particle group g=(g1,K,g d ,K,g m ), the basic particle swarm algorithm speed and position update formula is as follows: Where: c1 and c2 are acceleration factors, usually 2; ω is the inertia coefficient; r1 and r2 are random numbers uniformly distributed in the interval [0,1]; In the second half of the algorithm, the iteration interval [k a ,k b ]superior( where k a is the starting point of the second half of the iteration interval, k b is the end point of the second half of the iteration interval, k max For all particles, if the fitness of the particle does not change much in three generations, a random number pm of [0,1] is generated. If pm < 0.5, the particle is mutated when the speed is updated; otherwise, the particle speed update formula with only adaptive inertia coefficient is used. The judgment condition is calculated as follows: Let particle x i Fitness function fit(x i ) Change value between three generations G i : Where: k is the number of algorithm iterations; Particle x i The probability of randomly generated mutations in a certain iteration: pm=rand(0,1) Where: rand(0,1) represents a random number in the interval [0,1]; If in the iteration interval [k a ,k b ], particle x i The corresponding G i <0.01, and the random number pm generated at this time is <0.5, this particle x i Use the velocity update formula with adaptive inertia coefficient and variation term: Where: c3 is a constant, which can be taken as the maximum velocity of the particle; r3 is a random number uniformly distributed in the interval [-1,1]; ω n is the adaptive inertia coefficient; If the above conditions are not met, the particle uses the original update formula with adaptive inertia coefficient:

3. The wind turbine main shaft bearing fault diagnosis method based on IPSO-VMD according to claim 1 is characterized in that: The step S4 is specifically as follows: Perform envelope spectrum demodulation on the modal component with the largest kurtosis value to obtain the true characteristic frequency of the bearing, and compare the true characteristic frequency of the bearing with the fault characteristic frequency corresponding to each bearing fault type. When the true characteristic frequency of the bearing is consistent with the fault characteristic frequency corresponding to one or more bearing fault types, it is determined that the bearing has a fault of that bearing fault type.

Citation Information

Cited By

  • Vehicle transmission system bearing fault diagnosis method based on IPSO-Wav-KAN

    CN121723274A