WO2021139331A1 - 一种基于瞬时频率优化vmd的轴承故障诊断方法 - Google Patents

一种基于瞬时频率优化vmd的轴承故障诊断方法 Download PDF

Info

Publication number
WO2021139331A1
WO2021139331A1 PCT/CN2020/124397 CN2020124397W WO2021139331A1 WO 2021139331 A1 WO2021139331 A1 WO 2021139331A1 CN 2020124397 W CN2020124397 W CN 2020124397W WO 2021139331 A1 WO2021139331 A1 WO 2021139331A1
Authority
WO
WIPO (PCT)
Prior art keywords
instantaneous frequency
signal
fault diagnosis
value
bearing
Prior art date
Application number
PCT/CN2020/124397
Other languages
English (en)
French (fr)
Inventor
张续
黄大荣
梁曦
李少乾
王博
Original Assignee
重庆交通大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 重庆交通大学 filed Critical 重庆交通大学
Publication of WO2021139331A1 publication Critical patent/WO2021139331A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01MTESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M13/00Testing of machine parts
    • G01M13/04Bearings
    • G01M13/045Acoustic or vibration analysis

Definitions

  • the invention relates to the technical field of fault diagnosis, in particular to a bearing fault diagnosis method based on instantaneous frequency optimization VMD.
  • the present invention discloses a bearing fault diagnosis method based on instantaneous frequency optimization VMD, which uses the instantaneous frequency of signal components after VMD decomposition to determine the optimal K value in the VMD algorithm, and uses it for bearing In the fault diagnosis, the classification accuracy of the fault diagnosis model is effectively improved.
  • the present invention discloses a bearing fault diagnosis method based on the instantaneous frequency optimization VMD, which uses the instantaneous frequency of the signal components after VMD decomposition to determine the optimal K value in the VMD algorithm, and combines it Used in the fault diagnosis of the bearing, it effectively improves the classification accuracy of the fault diagnosis model.
  • the present invention adopts the following technical solutions:
  • a bearing fault diagnosis method based on instantaneous frequency optimization VMD including:
  • step S2 Preferably, in step S2:
  • f i (t) represents the instantaneous frequency of the i-th signal component at time t, Is a single-valued function at time t, that is, a single-component signal on frequency.
  • the analytical signal corresponding to the instantaneous frequency at time t is u k (t), x(t) is the Hilbert transform of x(t); Ak (t) is the modulus of the signal, Is the phase of the signal x(t) is the real part of the analytical signal, jx(t) is the imaginary part of the analytical signal;
  • the average instantaneous frequency of the analytical signal is Z(f i (t)),
  • step S3 Preferably, in step S3:
  • the candidate K value corresponding to the coefficient vector with the smallest Euclidean distance is taken as the optimal K value.
  • the present invention discloses a bearing fault diagnosis method based on instantaneous frequency optimization VMD, including: S1, obtaining bearing detection signals; S2, based on the instantaneous frequency optimization VMD algorithm to decompose the bearing detection signals, so that candidates
  • the K value is traversed within the preset range, and the mean value of the instantaneous frequency under different candidate K values is calculated; S3, based on the mean value of the instantaneous frequency under different candidate K values, the Lagrangian polynomial is used to determine the optimal K value; S4.
  • the signal component corresponding to the optimal K value is input into the fault diagnosis model to obtain the bearing fault diagnosis result.
  • the invention optimizes the K value based on the instantaneous frequency of the VMD signal component, decomposes and processes the vibration signal of the bearing to better reflect the fault characteristics of the vibration signal, and uses it in the fault diagnosis of the bearing to effectively improve The classification accuracy of the fault diagnosis model is improved.
  • Figure 1 is a flowchart of a specific implementation of a method for diagnosing a bearing fault based on instantaneous frequency optimization VMD disclosed in the present invention
  • Figure 2 is a broken line graph of the average instantaneous frequency under different K values
  • Figure 3 is a data diagram of training accuracy after signal VMD decomposition
  • Figure 4 shows the test accuracy data diagram after the signal VMD is decomposed.
  • the present invention discloses a bearing fault diagnosis method based on instantaneous frequency optimization VMD, including:
  • the detection signal of the bearing includes but is not limited to the vibration signal of the bearing.
  • the fault diagnosis model can be divided by wavelet signals, and the EMD and EEMD signal decomposition method adopts the above-mentioned diagnosis model to perform fault diagnosis.
  • the method is the prior art, which will not be repeated here.
  • the invention optimizes the K value based on the instantaneous frequency of the VMD signal component, decomposes and processes the vibration signal of the bearing to better reflect the fault characteristics of the vibration signal, and uses it in the fault diagnosis of the bearing to effectively improve The classification accuracy of the fault diagnosis model is improved.
  • step S2 During specific implementation, in step S2:
  • f i (t) represents the instantaneous frequency of the i-th signal component at time t, Is a single-valued function at time t, that is, a single-component signal on frequency.
  • the analytical signal corresponding to the instantaneous frequency at time t is u k (t), x(t) is the Hilbert transform of x(t); Ak (t) is the modulus of the signal, Is the phase of the signal x(t) is the real part of the analytical signal, jx(t) is the imaginary part of the analytical signal;
  • the average instantaneous frequency of the analytical signal is Z(f i (t)),
  • VMD is a new adaptive processing technology, which uses the variational decomposition framework to make up for the shortcomings of EMD and LMD decomposition of modal confusion and insufficient endpoint effects, and has high decomposition accuracy.
  • the present invention uses VMD to decompose the bearing fault vibration signal.
  • VMD is a new adaptive processing technology, which uses the variational decomposition framework to make up for the shortcomings of EMD and LMD decomposition modal confusion and insufficient endpoint effects, and has high decomposition accuracy.
  • the present invention uses VMD to decompose the bearing fault vibration signal.
  • VMD decomposition defines the eigenmode function IMF as a frequency modulation and amplitude modulation signal, and it is expressed as follows:
  • Ak (t) is the instantaneous amplitude of u k (t), and k represents the number of signal components after decomposition.
  • the original signal F is a multi-component signal, composed of k IMF components u k (t) with limited bandwidth, and the center frequency of each IMF is ⁇ k .
  • the bandwidth of each mode it is calculated by the following steps:
  • the analytical signal of the modal function is obtained, and the Hilbert transform is performed on each modal function u k (t).
  • ⁇ k ⁇ 1 , ⁇ 2 ,..., ⁇ k ⁇ represents the center frequency of each component.
  • a secondary penalty factor ⁇ and a Lagrangian multiplication operator ⁇ (t) are introduced.
  • the secondary penalty factor can ensure the accuracy of signal reconstruction in the presence of Gaussian noise, ⁇ (t ) Keeps the constraint conditions strict, and the extended Lagrangian expression is as follows:
  • the multiplier alternating direction algorithm is used to continuously update each IMF and its center frequency, and finally the saddle point of the constrained variational model is the optimal solution of the original problem.
  • the IMF in all frequency domains can be obtained by the following formula:
  • the above process is the adaptive decomposition process of VMD. From the decomposition principle, it can be known that VMD can well avoid the end effect and modal confusion of EMD and LMD algorithms. But from the actual decomposition process, the VMD algorithm loses the ability to decompose signals autonomously. In the VMD algorithm, it is necessary to set the K value first, that is, to set the number of VMD signal decomposition. And the reasonableness of K value in VMD algorithm determines the accuracy of VMD signal decomposition. If the K value is estimated according to the existing observation method, that is, the center frequency of the observed signal component is distinguished, the better the center frequency is, the better the K value is selected, and there is no phenomenon of over-decomposition or under-decomposition.
  • the present invention uses the instantaneous frequency to optimize the K value in the VMD, and uses the instantaneous frequency change difference between the signal components to measure the superiority of the K value.
  • step S3 During specific implementation, in step S3:
  • the candidate K value corresponding to the coefficient vector with the smallest Euclidean distance is taken as the optimal K value.
  • the K value is set too large, then the number of signal components decomposed is too large, the components will be broken and flocculent, especially at high frequencies, the average instantaneous frequency will decrease instead. If the K value is set too low, the signal will not be completely decomposed, and the superiority of the signal component cannot be reflected.
  • the original signal after denoising is decomposed by the VMD algorithm. Specifically, the K value can be traversed from 2 to 10, and the mean value of the instantaneous frequency under different K values can be calculated, and a line graph can be drawn. Use Lagrangian polynomials to fit discrete points, and extract polynomial coefficients under different K values to construct a coefficient vector, and calculate the Euclidean norm of the coefficient vector. The smaller the norm, the smoother the fitted instantaneous frequency curve , The better the K value.
  • l k (x) is a polynomial with n zeros. Therefore
  • the average instantaneous frequency of the different components is used as the discrete point to calculate the Lagrangian polynomial.
  • the coefficients of the polynomial are extracted and constructed into a vector, and the Euclidean distance of the vector of coefficients under different K values is calculated.
  • the bearing acceleration data of a certain laboratory is used for example verification analysis.
  • the normal acceleration signal of the bearing after denoising is input into VMD and decomposed, and the average instantaneous frequency line graph under different K values is obtained as shown in Figure 2.
  • the average instantaneous frequency calculation results under different K values are shown in Table 1.
  • the Euclidean norm can measure the magnitude of the vector, and also the degree of tilt of the Lagrangian polynomial. The smaller the degree of tilt, the slower the instantaneous frequency, and the corresponding optimal K value can be obtained.
  • the norm is the smallest when K is equal to 3. Then the value of K in the VMD decomposition algorithm is selected as 3 in the normal state of signal drop.
  • the original signal is compared with the signal decomposed by the VMD optimized by the K value, and 16 time-frequency domain features of the two signals are extracted and input into the fault diagnosis model for diagnosis.
  • the results are shown in Figure 3 and Figure 4 below.
  • the invention uses the change of the instantaneous frequency of the signal component after the VMD decomposition to measure the superiority of the K value. Compared with the previous observation method to judge the K value, this method is more accurate.
  • the present invention can obtain the optimal decomposition result, and the obtained signal components avoid the problem of over-decomposition or under-decomposition, and the use of these signal components for fault diagnosis can improve the accuracy of fault diagnosis.

Landscapes

  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • General Physics & Mathematics (AREA)
  • Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)

Abstract

一种基于瞬时频率优化VMD的轴承故障诊断方法,包括:S1、获取轴承检测信号;S2、基于瞬时频率优化VMD算法对轴承检测信号进行分解,令候选K值在预设范围内遍历,计算不同候选K值下的瞬时频率的均值;S3、基于不同候选K值下的瞬时频率的均值利用拉格朗日多项式确定最优K值;S4、将最优K值对应的信号分量输入故障诊断模型中得到轴承故障诊断结果。该方法能更好地体现出振动信号的故障特征,并将其用于轴承的故障诊断中,有效的提高了故障诊断模型的分类精度。

Description

一种基于瞬时频率优化VMD的轴承故障诊断方法 技术领域
本发明涉及故障诊断技术领域,尤其涉及一种基于瞬时频率优化VMD的轴承故障诊断方法。
背景技术
随着时代的发展和经济的繁荣,我国轴承工业飞速发展,轴承的品种由少到多,产品质量和技术水平从低到高,行业规模从小到大,已经形成了产品门类基本齐全、生产布局较为合理的专业生产体系。轴承作为当代工业机械设备中一种重要的零部件,它的主要功能是支撑机械旋转体,降低其运动过程中的摩擦系数,并保证其回转精度。但我国的机械轴承制造方面仍然存在着许多问题,目前我国的轴承行业生产能力较低,大多数轴承生产商来自国外,对于轴承行业来说理论基础能力较弱,研发水平不高。当前我国的设计和制造技术基本来源于对国外技术的模仿,且制造技术水平较低,我国轴承工业制造工艺和工艺装备技术发展缓慢,车加工数控率低。这些原因导致造成轴承工序能力指数低,一致性差,产品加工尺寸离散度大,因产品内在质量不稳定而影响轴承的精度、性能、寿命和可靠性,但轴承在机械运行中起着必不可少的作用,所以及时发现轴承中的故障,区分正常轴承与各类故障轴承,成为一项必不可少的研究。
综上所述,本发明公开了一种基于瞬时频率优化VMD的轴承故障诊断方法,利用经过VMD分解后信号分量的瞬时频率来确定VMD算法中的最优K值,并将其用于轴承的故障诊断中,有效的提高了故障诊断模型的分类精度。
发明内容
针对现有技术存在的上述不足,本发明公开了一种基于瞬时频率优化VMD的轴承故障诊断方法,利用经过VMD分解后信号分量的瞬时频率来确定VMD算法中的最优K值,并将其用于轴承的故障诊断中,有效的提高了故障诊断模型的分类精度。
为解决上述技术问题,本发明采用了如下的技术方案:
一种基于瞬时频率优化VMD的轴承故障诊断方法,包括:
S1、获取轴承检测信号;
S2、基于瞬时频率优化VMD算法对所述轴承检测信号进行分解,令候选K值在预设范围内遍历,计算不同候选K值下的瞬时频率的均值;
S3、基于不同候选K值下的瞬时频率的均值利用拉格朗日多项式确定最优K值;
S4、将最优K值对应的信号分量输入故障诊断模型中得到轴承故障诊断结果。
优选地,步骤S2中:
Figure PCTCN2020124397-appb-000001
式中,f i(t)表示t时刻第i个信号分量的瞬时频率,
Figure PCTCN2020124397-appb-000002
是t时刻的单值函数,即频率上的单分量信号,t时刻瞬时频率对应的解析信号为u k(t),
Figure PCTCN2020124397-appb-000003
x(t)是x(t)的希尔伯特变换;A k(t)为信号的模,
Figure PCTCN2020124397-appb-000004
为信号的相位
Figure PCTCN2020124397-appb-000005
x(t)为解析信号实部,jx(t)为解析信号虚部;
解析信号的瞬时频率均值为Z(f i(t)),
Figure PCTCN2020124397-appb-000006
优选地,步骤S3中:
将不同候选K值下的瞬时频率的均值作为计算拉格朗日多项式的离散点;
提取拉格朗日多项式的系数并构建系数向量,计算不同候选K值对应的系数向量的欧几里得距离;
将欧几里得距离最小的系数向量对应的候选K值作为最优K值。
综上所述,本发明公开了一种基于瞬时频率优化VMD的轴承故障诊断方法,包括:S1、获取轴承检测信号;S2、基于瞬时频率优化VMD算法对所述轴承检测信号进行分解,令候选K值在预设范围内遍历,计算不同候选K值下的瞬时频率的均值;S3、基于不同候选K值下的瞬时频率的均值利用拉格朗日多项式确定最优K值;S4、将最优K值对应的信号分量输入故障诊断模型中得到轴承故障诊断结果。本发明基于VMD信号分量的瞬时频率来优化K值,对轴承的振动信号进行分解处理,使其更好地体现出振动信号的故障特征,并将其用于轴承的故障诊断中,有效的提高了故障诊断模型的分类精度。
附图说明
图1为本发明公开的一种基于瞬时频率优化VMD的轴承故障诊断方法的一种具体实施方式的流程图;
图2为不同K值下平均瞬时频率折线图;
图3为信号VMD分解后训练精度数据图;
图4为信号VMD分解后测试精度数据图。
具体实施方式
为了使本发明的目的、技术方案和优点更加清楚,下面将结合附图对本发明作进一步的详细描述说明。
如图1所示,本发明公开了一种基于瞬时频率优化VMD的轴承故障诊断方法,包括:
S1、获取轴承检测信号;
在本发明中,轴承的检测信号包括但不仅限于轴承的振动信号。
S2、基于瞬时频率优化VMD算法对所述轴承检测信号进行分解,令候选K值在预设范围内遍历,计算不同候选K值下的瞬时频率的均值;
S3、基于不同候选K值下的瞬时频率的均值利用拉格朗日多项式确定最优K值;
S4、将最优K值对应的信号分量输入故障诊断模型中得到轴承故障诊断结果。
在本发明中,故障诊断模型可采用小波信号分,EMD、EEMD信号分解方法采用上述诊断模型进行故障诊断的方法为现有技术,在此不再赘述。
本发明基于VMD信号分量的瞬时频率来优化K值,对轴承的振动信号进行分解处理,使其更好地体现出振动信号的故障特征,并将其用于轴承的故障诊断中,有效的提高了故障诊断模型的分类精度。
具体实施时,步骤S2中:
Figure PCTCN2020124397-appb-000007
式中,f i(t)表示t时刻第i个信号分量的瞬时频率,
Figure PCTCN2020124397-appb-000008
是t时刻的单值函数,即频率上的单分量信号,t时刻瞬时频率对应的解析信号为u k(t),
Figure PCTCN2020124397-appb-000009
x(t)是x(t)的希尔伯特变换;A k(t)为信号的模,
Figure PCTCN2020124397-appb-000010
为信号的相位
Figure PCTCN2020124397-appb-000011
x(t)为解析信号实部,jx(t)为解析信号虚部;
解析信号的瞬时频率均值为Z(f i(t)),
Figure PCTCN2020124397-appb-000012
根据平稳相位原理,式
Figure PCTCN2020124397-appb-000013
的积分在频率f i(t)处存在最大值,f i(t)需要满足
Figure PCTCN2020124397-appb-000014
Figure PCTCN2020124397-appb-000015
这一结论说明,非平稳信号的能量主要集中在瞬时频率处,这一结论标明瞬时频率在信号的识别、检测、估计和建模中起到很大的作用,同时也可以用来作为VMD分解信号的评价指标。
VMD是一种新的自适应处理技术,它利用变分分解框架很好地弥补了EMD以及LMD分解 模态混淆和端点效应不足的缺点,具有较高的分解精度。鉴于针对轴承故障信号非平稳,非线性的本质特征,因此本发明利用VMD对轴承故障振动信号进行分解。
VMD是一种新的自适应处理技术,它利用变分分解框架很好地弥补了EMD以及LMD分解模态混淆和端点效应不足的缺点,具有较高的分解精度。鉴于针对轴承故障信号非平稳,非线性的本质特征,因此本发明利用VMD对轴承故障振动信号进行分解。
VMD分解将本征模态函数IMF定义为一个调频调幅信号,并表达如下式:
u k(t)=A k(t)cos[φ k(t)],k=1,2,…,K.
其中,A k(t)为u k(t)的瞬时幅值,k表示分解后的信号分量个数。
设原始信号F为多分量信号,由k个有限带宽的IMF分量u k(t)组成,且各IMF的中心频率为ω k。为确定每个模态的带宽,通过如下步骤求取:
求取模态函数的解析信号,对每个模态函数u k(t)进行希尔伯特变换。
Figure PCTCN2020124397-appb-000016
对各模态解析信号预估中心频率
Figure PCTCN2020124397-appb-000017
进行混合。将每个模态的频谱调制到相应的基频带。如下所示:
Figure PCTCN2020124397-appb-000018
计算以上解调信号的梯度的平方L 2范数,估计出各个模态分量的带宽。建立的约束变分模型为:
Figure PCTCN2020124397-appb-000019
Figure PCTCN2020124397-appb-000020
式中,u k={u 1,u 2,…,u k}表示分解得到的K个IMF分量,ω k={ω 12,…,ω k}表示各分量的中心频率。
为求解上述约束变分模型,引入二次惩罚因子α和拉格朗日乘法算子λ(t),其中二次惩罚因子可在高斯噪声存在的情况下保证信号的重构精度,λ(t)使得约束条件保持严格性,扩展后的拉格朗日表达式如下:
Figure PCTCN2020124397-appb-000021
利用乘子交替方向算法不断更新各IMF及其中心频率,最终所求式约束变分模型的鞍点即为原问题的最优解。所有频域中的IMF可通过下式获得:
Figure PCTCN2020124397-appb-000022
其中,
Figure PCTCN2020124397-appb-000023
为当前剩余量f(ω)-∑ i≠ku i(ω)通过Wiener滤波的结果;算法中各IMF功率谱的中心更新式如下:
Figure PCTCN2020124397-appb-000024
上述过程即为VMD的自适应分解过程,从分解原理中可以知道,VMD很好地规避EMD以及LMD算法的端点效应和模态混淆。但从实际分解的过程来看,VMD算法丧失了自主分解信号的能力,在VMD算法中需要优先设定K值,即设定VMD信号分解的个数。且VMD算法中K值的合理性决定了VMD的信号分解精度。若根据现有的观察法预估K值即观察信号分量的中心频率区分情况,中心频率区分的越好说明K值选择越好,没有出现过分解或者欠分解的现象。但这种预估方法难以存在很大的误差,难以保证信号的分解精度,也会影响故障诊断精度。因此本发明利用瞬时频率优化VMD中K值,利用信号分量之间的瞬时频率变化差异衡量K值的优越性。
具体实施时,步骤S3中:
将不同候选K值下的瞬时频率的均值作为计算拉格朗日多项式的离散点;
提取拉格朗日多项式的系数并构建系数向量,计算不同候选K值对应的系数向量的欧几里得距离;
将欧几里得距离最小的系数向量对应的候选K值作为最优K值。
如果K值设定过于大,那么分解出的信号分量个数过大,则分量会出现断断絮絮地现象,尤其是在高频平均瞬时频率反而会降低。若K值设定过低,则信号不会被完全分解,无法体现信号分量的优越性。将除噪后的原始信号通过VMD算法进行分解,具体可令K值从2遍历到10,计算不同K值下的瞬时频率的均值,并画出折线图。利用拉格朗日多项式拟合离散点, 并提取出不同K值下的多项式系数构建成系数向量,计算系数向量的欧几里得范数,范数越小,拟合的瞬时频率曲线越平滑,则K值越好。
在计算出不同候选K值下的瞬时频率的均值,我们需要采用一种指标来衡量瞬时频率的均值的变化走势,可以避免造成主观判断所产生的误差。通过对瞬时频率的均值的拟合分析可以计算其拉格朗日多项式并比较其系数构成的系数向量范数大小,对于K值的优劣性进行评价。
令x i=Z(f i(t)),对于插值节点(即平均瞬时频率点)x 0,x 1,…,x n中任一点x k(k=0,1,…,n)做一n次多项式l k(x),满足
Figure PCTCN2020124397-appb-000025
拉格朗日插值法的基函数即为l k(x),节点为x i(i=0,1,…,k-1,k,k+1,…,n)。从而l k(x)为有n个零点的多项式。故
Figure PCTCN2020124397-appb-000026
式中l k(x)(k=0,1,…,n)为在n+1个插值结点上的n次基本插值多项式或n次拉格朗日插值基函数。利用n次基本插值多项式可写出满足插值条件P n(x i)=f(x i)=y i(i=0,1,2,…,n)的n次拉格朗日多项式为:
Figure PCTCN2020124397-appb-000027
将不同分量的平均瞬时频率作为计算拉格朗日多项式的离散点。通过计算得到拉格朗日多项式的最简形式后,提取多项式的系数并构建成向量,计算不同K值下系数的向量的欧几里得距离,对于系数向量v=(v 1,v 2,…,v 3),其向量的欧几里得距离为
Figure PCTCN2020124397-appb-000028
为验证本发明的有效性,采用某实验室的轴承加速度数据进行实例验证分析。将去噪后的轴承正常加速度信号输入VMD中分解,得到不同K值下是的平均瞬时频率折线图如图2所示,同时不同K值下的平均瞬时频率计算结果如表1所示。
表1 不同K值下的平均瞬时频率
Figure PCTCN2020124397-appb-000029
Figure PCTCN2020124397-appb-000030
将每个K值下的瞬时频率的点看作为离散的点并构建拉格朗日多项式,简化后得到多项式的每项系数并计算出系数向量的欧几里得范数如表2所示。
表2 欧几里得范数
Figure PCTCN2020124397-appb-000031
从表2中可知欧几里得范数可衡量向量的大小,也就可以衡量拉格朗日多项式的倾斜程度,倾斜程度越小则瞬时频率较缓,即可得到对应的最优K值。K=2时只能分解出两个分量信号,与实际情况不相符,故排除K值取2。除K值等于2之外发现K等于3时范数最小,则在正常状态信号下降VMD分解算法中的K值选定为3。
将利用原始信号和经过K值优化的VMD分解后的信号进行对比,分别提取两种信号的16种时频域特征后输入故障诊断模型中进行诊断。结果如下图3及图4所示.
表3 对比实验故障诊断精度
Figure PCTCN2020124397-appb-000032
在表3的对比实验结果中我们可以看到经过VMD分解后的原始信号经过故障诊断模型后得到的训练精度可以达到96%,而未经处理的原始信号的诊断精度只有88%。所以,该实验证明了本发明中的发明方法可以对原始信号进行进一步的优化处理,并且使得轴承的故障诊断结果更加精确,由此证明本发明具有有效性和实用性,且可以用于轴承故障分类中。
综上所述,本发明与现有技术相比,具有以下技术效果:
本发明在基于VMD算法的基础上,利用VMD分解过后的信号分量的瞬时频率的变化来衡量K值的优越性。与以往的观察法来判断K值的方法来说,本方法更具有准确性。
本发明选择出VMD算法的最优K值后,可以得到最优的分解结果,得到的信号分量避免了过分解或者欠分解的问题,采用这些信号分量进行故障诊断可以提高故障诊断的精度。
最后说明的是,以上实施例仅用以说明本发明的技术方案而非限制,尽管通过参照本发明的优选实施例已经对本发明进行了描述,但本领域的普通技术人员应当理解,可以在形式上和细节上对其作出各种各样的改变,而不偏离所附权利要求书所限定的本发明的精神和范 围。

Claims (3)

  1. 一种基于瞬时频率优化VMD的轴承故障诊断方法,其特征在于,包括:
    S1、获取轴承检测信号;
    S2、基于瞬时频率优化VMD算法对所述轴承检测信号进行分解,令候选K值在预设范围内遍历,计算不同候选K值下的瞬时频率的均值;
    S3、基于不同候选K值下的瞬时频率的均值利用拉格朗日多项式确定最优K值;
    S4、将最优K值对应的信号分量输入故障诊断模型中得到轴承故障诊断结果。
  2. 如权利要求1所述的基于瞬时频率优化VMD的轴承故障诊断方法,其特征在于,步骤S2中:
    Figure PCTCN2020124397-appb-100001
    式中,f i(t)表示t时刻第i个信号分量的瞬时频率,
    Figure PCTCN2020124397-appb-100002
    是t时刻的单值函数,即频率上的单分量信号,t时刻瞬时频率对应的解析信号为u k(t),
    Figure PCTCN2020124397-appb-100003
    x(t)是x(t)的希尔伯特变换;A k(t)为信号的模,
    Figure PCTCN2020124397-appb-100004
    为信号的相位
    Figure PCTCN2020124397-appb-100005
    x(t)为解析信号实部,jx(t)为解析信号虚部;
    解析信号的瞬时频率均值为Z(f i(t)),
    Figure PCTCN2020124397-appb-100006
  3. 如权利要求1所述的基于瞬时频率优化VMD的轴承故障诊断方法,其特征在于,步骤S3中:
    将不同候选K值下的瞬时频率的均值作为计算拉格朗日多项式的离散点;
    提取拉格朗日多项式的系数并构建系数向量,计算不同候选K值对应的系数向量的欧几里得距离;
    将欧几里得距离最小的系数向量对应的候选K值作为最优K值。
PCT/CN2020/124397 2020-01-08 2020-10-28 一种基于瞬时频率优化vmd的轴承故障诊断方法 WO2021139331A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010018271.5 2020-01-08
CN202010018271.5A CN111189639B (zh) 2020-01-08 2020-01-08 一种基于瞬时频率优化vmd的轴承故障诊断方法

Publications (1)

Publication Number Publication Date
WO2021139331A1 true WO2021139331A1 (zh) 2021-07-15

Family

ID=70708049

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/124397 WO2021139331A1 (zh) 2020-01-08 2020-10-28 一种基于瞬时频率优化vmd的轴承故障诊断方法

Country Status (2)

Country Link
CN (1) CN111189639B (zh)
WO (1) WO2021139331A1 (zh)

Cited By (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113552447A (zh) * 2021-07-27 2021-10-26 上海电机学院 一种基于随机森林的串联电弧故障检测方法
CN113589795A (zh) * 2021-08-02 2021-11-02 湖州师范学院 一种基于智能寻优的非线性啁啾模态分解算法的多重振荡检测方法
CN113878613A (zh) * 2021-09-10 2022-01-04 哈尔滨工业大学 一种基于wlctd与oma-vmd的工业机器人谐波减速器早期故障检测方法
CN114118147A (zh) * 2021-11-17 2022-03-01 西安交通大学 基于改进鲸鱼优化vmd的扭振信号瞬时频率特征提取方法
CN114215733A (zh) * 2021-11-02 2022-03-22 西安交通大学 嵌入式压缩机活塞杆螺纹松动故障自监测诊断方法及系统
CN114235408A (zh) * 2021-12-17 2022-03-25 哈尔滨工程大学 基于改进级联变分模态分解的轴承故障诊断方法及系统
CN114371005A (zh) * 2021-12-17 2022-04-19 江苏核电有限公司 一种滚动轴承的冲击特征提取方法及装置
CN114563181A (zh) * 2022-01-10 2022-05-31 浙江工业大学之江学院 基于改进变分模态提取的旋转机械故障特征提取方法
CN114563189A (zh) * 2022-02-28 2022-05-31 西北工业大学 基于瞬时转速的无人机发动机故障诊断方法
CN114994517A (zh) * 2022-07-04 2022-09-02 哈尔滨理工大学 一种用于模拟电路的软故障诊断方法
CN115687862A (zh) * 2022-10-19 2023-02-03 北京科技大学 一种基于时变滤波的旋转机械信号时频分析方法
CN116522269A (zh) * 2023-06-28 2023-08-01 厦门炬研电子科技有限公司 一种基于Lp范数非平稳信号稀疏重建的故障诊断方法
CN117691631A (zh) * 2024-02-04 2024-03-12 西安热工研究院有限公司 一种基于混合储能装置的电力调频方法和系统
CN117977636A (zh) * 2024-03-28 2024-05-03 西安热工研究院有限公司 一种基于样本熵的熔盐耦合火电机组的调频方法和系统
CN117977634B (zh) * 2024-03-27 2024-06-07 西安热工研究院有限公司 一种基于双分解的熔盐耦合火电机组的调频方法和系统

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111189639B (zh) * 2020-01-08 2021-09-14 重庆交通大学 一种基于瞬时频率优化vmd的轴承故障诊断方法
CN112367063B (zh) * 2020-11-13 2022-02-01 苏州大学 自适应中心频率模式分解方法及系统
CN112633371A (zh) * 2020-12-22 2021-04-09 河北建投能源投资股份有限公司 一种基于vmd-msst的轴承故障诊断方法
CN113204850A (zh) * 2021-05-28 2021-08-03 重庆交通大学 一种桥梁挠度监测中的温度效应分离方法
CN113625164A (zh) * 2021-08-02 2021-11-09 南京航空航天大学 航空发电机故障特征提取方法、系统、介质及计算设备
CN114662548B (zh) * 2022-04-12 2023-06-20 安徽中安昊源电力科技有限公司 一种基于动作异常的断路器诊断方法及系统
CN115979649A (zh) * 2023-02-06 2023-04-18 安徽大学 一种自适应复数域amfm模型优化的轴承故障诊断方法

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106777626A (zh) * 2016-12-07 2017-05-31 西安科技大学 一种离散变量桁架非概率可靠性优化设计方法
CN109829402A (zh) * 2019-01-21 2019-05-31 福州大学 基于gs-svm的不同工况下轴承损伤程度诊断方法
CN110061792A (zh) * 2019-04-04 2019-07-26 西安电子科技大学 一种基于变分模态分解的频谱感知算法
CN110333285A (zh) * 2019-07-04 2019-10-15 大连海洋大学 基于变分模态分解的超声兰姆波缺陷信号识别方法
US20190325608A1 (en) * 2018-04-24 2019-10-24 Canon Kabushiki Kaisha Calibration apparatus, calibration method and storage medium
CN111189639A (zh) * 2020-01-08 2020-05-22 重庆交通大学 一种基于瞬时频率优化vmd的轴承故障诊断方法

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5331980B2 (ja) * 2009-10-30 2013-10-30 旭化成エレクトロニクス株式会社 回転角度センサ及び回転角度算出方法
CN107589454A (zh) * 2017-07-25 2018-01-16 西安交通大学 一种基于vmd‑tfpf压制地震勘探随机噪声方法
CN108168924B (zh) * 2017-12-20 2020-04-14 桂林航天工业学院 一种基于vmd和mfss模型的往复压缩机寿命预测方法
CN109596349B (zh) * 2018-12-06 2020-08-25 桂林电子科技大学 一种基于vmd和pct的减速器故障诊断方法
CN110363130B (zh) * 2019-07-08 2023-01-13 国网四川省电力公司电力科学研究院 基于变分模态分解的电压暂降源辨识方法及辨识装置
CN110309817B (zh) * 2019-07-19 2020-10-02 北京理工大学 一种参数自适应优化vmd的脉搏波运动伪影去除方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106777626A (zh) * 2016-12-07 2017-05-31 西安科技大学 一种离散变量桁架非概率可靠性优化设计方法
US20190325608A1 (en) * 2018-04-24 2019-10-24 Canon Kabushiki Kaisha Calibration apparatus, calibration method and storage medium
CN109829402A (zh) * 2019-01-21 2019-05-31 福州大学 基于gs-svm的不同工况下轴承损伤程度诊断方法
CN110061792A (zh) * 2019-04-04 2019-07-26 西安电子科技大学 一种基于变分模态分解的频谱感知算法
CN110333285A (zh) * 2019-07-04 2019-10-15 大连海洋大学 基于变分模态分解的超声兰姆波缺陷信号识别方法
CN111189639A (zh) * 2020-01-08 2020-05-22 重庆交通大学 一种基于瞬时频率优化vmd的轴承故障诊断方法

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
LUO XIAOYAN, LU WENHAI;YOU YIPING;HU XIANNENG: "A Method for Ball Mill Vibration Signal Random Noise Suppression based on VMD and SVD", NOISE AND VIBRATION CONTROL, vol. 39, no. 6, 1 December 2019 (2019-12-01), pages 169 - 216, XP055827787, ISSN: 1006-1355, DOI: 10.3969/j.issn.1006-1355.2019.06.030 *
ZAN TAO, PANG ZHAOLIANG;WANG MIN;GAO XIANGSHENG: "Early Fault Diagnosis Method of Rolling Bearings Based on VMD", BEIJING GONGYE DAXUE XUEBAO - BEIJING UNIVERSITY OF TECHNOLOGY.JOURNAL, BEIJING GONGYE DAXUE, BEIJING, CN, vol. 45, no. 2, 1 February 2019 (2019-02-01), CN, pages 103 - 110, XP055827786, ISSN: 0254-0037, DOI: 10.11936 /bjutxb2017090012 *

Cited By (22)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113552447A (zh) * 2021-07-27 2021-10-26 上海电机学院 一种基于随机森林的串联电弧故障检测方法
CN113589795A (zh) * 2021-08-02 2021-11-02 湖州师范学院 一种基于智能寻优的非线性啁啾模态分解算法的多重振荡检测方法
CN113878613B (zh) * 2021-09-10 2023-01-31 哈尔滨工业大学 一种基于wlctd与oma-vmd的工业机器人谐波减速器早期故障检测方法
CN113878613A (zh) * 2021-09-10 2022-01-04 哈尔滨工业大学 一种基于wlctd与oma-vmd的工业机器人谐波减速器早期故障检测方法
CN114215733A (zh) * 2021-11-02 2022-03-22 西安交通大学 嵌入式压缩机活塞杆螺纹松动故障自监测诊断方法及系统
CN114118147A (zh) * 2021-11-17 2022-03-01 西安交通大学 基于改进鲸鱼优化vmd的扭振信号瞬时频率特征提取方法
CN114118147B (zh) * 2021-11-17 2024-02-13 西安交通大学 基于改进鲸鱼优化vmd的扭振信号瞬时频率特征提取方法
CN114235408A (zh) * 2021-12-17 2022-03-25 哈尔滨工程大学 基于改进级联变分模态分解的轴承故障诊断方法及系统
CN114371005A (zh) * 2021-12-17 2022-04-19 江苏核电有限公司 一种滚动轴承的冲击特征提取方法及装置
CN114235408B (zh) * 2021-12-17 2023-08-29 哈尔滨工程大学 基于改进级联变分模态分解的轴承故障诊断方法及系统
CN114563181A (zh) * 2022-01-10 2022-05-31 浙江工业大学之江学院 基于改进变分模态提取的旋转机械故障特征提取方法
CN114563181B (zh) * 2022-01-10 2023-06-27 浙江工业大学之江学院 基于改进变分模态提取的旋转机械故障特征提取方法
CN114563189B (zh) * 2022-02-28 2024-01-12 西北工业大学 基于瞬时转速的无人机发动机故障诊断方法
CN114563189A (zh) * 2022-02-28 2022-05-31 西北工业大学 基于瞬时转速的无人机发动机故障诊断方法
CN114994517A (zh) * 2022-07-04 2022-09-02 哈尔滨理工大学 一种用于模拟电路的软故障诊断方法
CN115687862A (zh) * 2022-10-19 2023-02-03 北京科技大学 一种基于时变滤波的旋转机械信号时频分析方法
CN116522269A (zh) * 2023-06-28 2023-08-01 厦门炬研电子科技有限公司 一种基于Lp范数非平稳信号稀疏重建的故障诊断方法
CN116522269B (zh) * 2023-06-28 2023-09-19 厦门炬研电子科技有限公司 一种基于Lp范数非平稳信号稀疏重建的故障诊断方法
CN117691631A (zh) * 2024-02-04 2024-03-12 西安热工研究院有限公司 一种基于混合储能装置的电力调频方法和系统
CN117691631B (zh) * 2024-02-04 2024-04-30 西安热工研究院有限公司 一种基于混合储能装置的电力调频方法和系统
CN117977634B (zh) * 2024-03-27 2024-06-07 西安热工研究院有限公司 一种基于双分解的熔盐耦合火电机组的调频方法和系统
CN117977636A (zh) * 2024-03-28 2024-05-03 西安热工研究院有限公司 一种基于样本熵的熔盐耦合火电机组的调频方法和系统

Also Published As

Publication number Publication date
CN111189639A (zh) 2020-05-22
CN111189639B (zh) 2021-09-14

Similar Documents

Publication Publication Date Title
WO2021139331A1 (zh) 一种基于瞬时频率优化vmd的轴承故障诊断方法
CN109829402B (zh) 基于gs-svm的不同工况下轴承损伤程度诊断方法
CN107784325B (zh) 基于数据驱动增量融合的螺旋式故障诊断方法
Qin et al. Adaptive bistable stochastic resonance and its application in mechanical fault feature extraction
CN109542089B (zh) 一种基于改进变分模态分解的工业过程非线性振荡检测方法
Chen et al. Maximum likelihood estimation for uncertain autoregressive model with application to carbon dioxide emissions
CN104881567A (zh) 一种基于统计模型的桥梁健康监测数据小波降噪方法
CN112199888B (zh) 一种基于深度残差网络的旋转设备故障诊断方法、系统及可读存储介质
CN112394642B (zh) 一种基于超参数优化的机器人铣削加工颤振辨识方法
CN113094893A (zh) 晶圆品质虚拟测量方法、装置、计算机设备和存储介质
CN111260776B (zh) 一种自适应正态分析的三维形貌重建方法
CN111538309B (zh) 一种基于多变量非线性调频模态分解的工业过程厂级振荡检测方法
CN110333054B (zh) 一种针对白车身焊接设备的缓变微小故障检测方法
CN112200060A (zh) 一种基于网络模型的旋转设备故障诊断方法、系统及可读存储介质
CN104731875B (zh) 一种获取多维数据稳定性的方法和系统
Wang et al. A novel time-frequency analysis method for fault diagnosis based on generalized S-transform and synchroextracting transform
Liu et al. Bearing failure diagnosis at time-varying speed based on adaptive clustered fractional Gabor transform
CN114800036B (zh) 一种设备健康状态评估方法
Hwang et al. A new machine condition monitoring method based on likelihood change of a stochastic model
CN114993671A (zh) 一种基于q因子小波变换的振动故障诊断方法及其系统
Zhang et al. Real-time remaining useful life prediction based on adaptive kernel window width density
CN114755010A (zh) 一种旋转机械振动故障诊断方法及其系统
Chiuso et al. Learning sparse dynamic linear systems using stable spline kernels and exponential hyperpriors
CN114235408A (zh) 基于改进级联变分模态分解的轴承故障诊断方法及系统
Ganesan et al. A multiscale Bayesian SPRT approach for online process monitoring

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20912373

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20912373

Country of ref document: EP

Kind code of ref document: A1