WO2019161592A1 - 一种利用聚类自动提取结构模态参数的方法 - Google Patents

一种利用聚类自动提取结构模态参数的方法 Download PDF

Info

Publication number
WO2019161592A1
WO2019161592A1 PCT/CN2018/080922 CN2018080922W WO2019161592A1 WO 2019161592 A1 WO2019161592 A1 WO 2019161592A1 CN 2018080922 W CN2018080922 W CN 2018080922W WO 2019161592 A1 WO2019161592 A1 WO 2019161592A1
Authority
WO
WIPO (PCT)
Prior art keywords
modal
class
mode
clustering
stable
Prior art date
Application number
PCT/CN2018/080922
Other languages
English (en)
French (fr)
Inventor
伊廷华
杨小梅
曲春绪
李宏男
Original Assignee
大连理工大学
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 大连理工大学 filed Critical 大连理工大学
Priority to US16/342,948 priority Critical patent/US20200074221A1/en
Publication of WO2019161592A1 publication Critical patent/WO2019161592A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/23Clustering techniques
    • G06F18/231Hierarchical techniques, i.e. dividing or merging pattern sets so as to obtain a dendrogram
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01MTESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M5/00Investigating the elasticity of structures, e.g. deflection of bridges or air-craft wings
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10Complex mathematical operations
    • G06F17/16Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/23Clustering techniques
    • G06F18/232Non-hierarchical techniques
    • G06F18/2321Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00Computer-aided design [CAD]
    • G06F30/20Design optimisation, verification or simulation
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12Classification; Matching

Definitions

  • the invention belongs to the field of structural health monitoring and relates to an automatic extraction method of modal parameters of engineering structures.
  • the change of structural modal parameters can reflect the health of the engineering structure.
  • the parametric modal identification method is widely used due to its explicit physical parameter model, such as least squares complex frequency domain method, random subspace method and feature system implementation algorithm.
  • most of these methods require subjective experience when using them. Taking the algorithm of the feature system as an example, the existence of environmental noise makes the model of the algorithm difficult to set. If the order is too low, the modal will be missed, and the order is too high, resulting in falsehood. Modal. To accurately extract structural physical modalities, it is generally necessary to analyze them in combination with stable graphs.
  • the stability map shows the frequency as the X axis and the calculation order as the Y axis, and displays the frequency corresponding to the mode acquired under each order. Since the same physical mode has substantially identical modal parameters (frequency, mode shape, damping ratio) under various calculation orders, and there is a large deviation between the false modes, the modal parameter deviation threshold can be set. To judge the stability of the modality. For a mode in which the modal parameter deviation is less than the threshold value in the adjacent calculation order, it can be considered as a stable mode. If a stable mode with the same modal parameter appears in each calculation order, it is considered that there is a greater possibility. For physical modality. However, the threshold of the modal parameter deviation needs to be determined based on subjective experience; in addition, the physical mode is selected from the sequence that has been determined to be a stable mode, and manual execution is required for large structures in a complex environment. Large and subjective.
  • the object of the present invention is to provide a method for automatically extracting physical modal parameters of a structure, and to solve the problem that the recognition result is subjective and difficult to identify in real time due to human participation.
  • the technical proposal of the present invention is to propose a method for automatically extracting structural physical modes, which is characterized in that natural excitation technology is combined with a feature system to implement an algorithm to obtain modal parameters of a structural random response under different calculation orders;
  • Each modal below finds the modality most similar to this modality under its adjacent calculation order, and obtains the dissimilarity (frequency, mode shape and damping ratio deviation) as the characteristics of the modality.
  • the fuzzy C-means clustering is performed on the acquired features of each mode, and the high similarity is adaptively divided.
  • the stable modal class and the low similarity unstable modal class then hierarchically clustering the obtained stable modal classes, and classifying the modalities that appear in different calculation orders and having the same modal parameters into one class, Thereby achieving automatic acquisition of each physical modality of the structure.
  • a method for automatically extracting structural modal parameters by using clustering the steps are as follows:
  • Step 1 Obtain modal parameters under different calculation orders
  • H ms (0) USV T (2) where: U and V are ⁇ arrays; S is a singular value matrix;
  • Step 2 Stable modal class and unstable modal class division
  • fuzzy C-means clustering is performed to divide the stable modal class C 1 and the unstable modal class C 2 .
  • the clustering expression is:
  • k represents the clustering category
  • ⁇ k represents the membership matrix of the fuzzy cluster, where the element ⁇ ij,k is defined as the modal i of the order j belongs to the class k Membership:
  • Step 3 Extract the physical mode from the stability map
  • Each mode in the stable modal class is a self-contained class
  • step 2) Repeat step 2) until the minimum distance between the classes exceeds the allowable value ⁇ lim ;
  • step 2) the distance between the mode i in the class g and the mode h in the class l:
  • ⁇ ig, hl d f ig, hl +1-MOC ig, hl (6)
  • the distance between classes is determined according to the average distance criterion:
  • n g and n l represent the number of samples of classes g and h, respectively;
  • n T (0.3 ⁇ 0.5) n u ;
  • the invention has the beneficial effects that by performing fuzzy clustering on the modal dissimilarity rather than the modal parameter itself, the stable and unstable modes can be adaptively divided, the parameter threshold is avoided by human intervention, and the modal parameter identification is improved.
  • the degree of automation is improved.
  • Figure 1 shows the stable mode and unstable mode distribution.
  • the mass of each layer is 1.10 ⁇ 10 6 kg
  • the stiffness of each layer is 1541.07 ⁇ 10 6 N/m
  • the damping is Rayleigh damping ⁇ M+ ⁇ K
  • Rayleigh damping coefficient ⁇ 0.3000
  • 0.0005
  • the excitation form is zero-mean Gaussian white noise excitation
  • the noise level is set to 20% of the random response variance
  • the sampling frequency is 100 Hz
  • the sampling signal is the acceleration of each layer of the frame.
  • the initial calculation order is selected as 2, and the singular value matrix S is truncated to obtain a 2 ⁇ 2 singular value matrix S n , and then the eigensystem implementation algorithm is used to obtain the frequency f i2 , the damping ratio ⁇ i2 , and the mode shape.
  • Construct dissimilarity vector Fuzzy C-means clustering is used to extract the stable modal class C 1 as the characteristic of the mode i under the order j.
  • the modal frequencies in the stable modal class are shown in Fig. 1, and according to the nearest neighbor distance distribution of the stable mode.
  • f 6 10.624 Hz
  • f 7 11.642 Hz
  • f 2 12.278 Hz
  • ⁇ 1 2.219%
  • ⁇ 2 1.254%
  • ⁇ 3 1.291%
  • ⁇ 4 1.459%
  • ⁇ 5 1.695%
  • ⁇ 6 1.903%
  • ⁇ 7 2.059%
  • ⁇ 8 2.152%.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Mathematical Physics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Pure & Applied Mathematics (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Computer Hardware Design (AREA)
  • Probability & Statistics with Applications (AREA)
  • Geometry (AREA)
  • Algebra (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Software Systems (AREA)
  • Databases & Information Systems (AREA)
  • Computing Systems (AREA)
  • Complex Calculations (AREA)
  • Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)

Abstract

一种利用聚类自动提取结构模态参数的方法,属于结构健康监测技术领域。首先,利用自然激励技术结合特征系统实现算法获取结构随机响应在不同计算阶次下的模态参数;然后,依据物理模态稳定和相似度高,而虚假模态不稳定和相似度低这一特性,以相邻计算阶次下两个最相似模态的不相似度作为较低阶次下模态的特征,进行模糊C均值聚类,自适应的获取具有高相似度的稳定模态类;最后,对其进行层次聚类,将出现在不同计算阶次且具有相同模态参数的模态划分为一类,以此类推可获取结构的每一个物理模态。通过对模态不相似度而非模态参数本身进行聚类来获取稳定模态,从而无需人工参与。

Description

一种利用聚类自动提取结构模态参数的方法 技术领域
本发明属于结构健康监测领域,涉及工程结构模态参数自动提取方法。
背景技术
结构模态参数的变化可以反映工程结构的健康状况,为了掌握结构的服役性能,实时获取结构的模态参数十分必要。在现有研究中,参数化模态识别方法由于其具有明确的物理参数模型而被广泛应用,如最小二乘复频域法、随机子空间法和特征系统实现算法等。然而,这些方法在使用时大多需要主观经验,以特征系统实现算法为例,环境噪声的存在使算法的模型定阶比较困难,阶次过低将造成模态遗漏,而阶次过高造成虚假模态。要准确提取结构物理模态,一般需要结合稳定图来分析。稳定图以频率为X轴,计算阶次为Y轴,将各个阶次下获取的模态对应的频率显示出来。由于在各个计算阶次下同一个物理模态具有基本一致的模态参数(频率、振型、阻尼比),而虚假模态间会具有较大的偏差,因此可以通过设置模态参数偏差阈值来判断模态的稳定性。对于相邻计算阶次下模态参数偏差小于阈值的模态可认为是稳定模态,如果具有相同模态参数的稳定模态在各个计算阶次下均出现,则认为有更大的可能性为物理模态。然而,模态参数偏差的阈值需要根据主观经验来确定;此外,从已经确定为稳定模态的序列中选择物理模态也需要手动执行,而对于复杂环境下的大型结构,人工选取的工作量大且主观性强。
为减少主观经验对选取稳定图中物理模态的影响,基于聚类的方法得到大量应用,但目前的聚类算法大多是对已经在稳定图中通过设定阈值判别为稳定的模态进行自动归类,即稳定图中阈值的设置仍然需要专业人员结合实际工程结构的特性、环境干扰程度以及模态识别算法的稳定性等多种因素依靠经验来设定。因此,在稳定图中不设阈值,自适应的区分稳定和不稳定模态具有十分重要的工程意义。
发明内容
本发明的目的是提供一种自动提取结构物理模态参数的方法,解决识别中 由于人工参与造成识别结果主观性强和难以实时识别的问题。
本发明的技术方案是:提出一种结构物理模态自动提取方法,其特点是利用自然激励技术结合特征系统实现算法获取结构随机响应在不同计算阶次下的模态参数;对于指定计算阶次下的每一个模态,在与其相邻的计算阶次下找到与此模态最相似的模态,获取两者的不相似度(频率、振型和阻尼比偏差)作为此模态的特征,依据物理模态相对稳定和相似度高,而虚假模态不稳定和相似度低这一特性,对获取的每一个模态的特征进行模糊C均值聚类,自适应地划分具有高相似度的稳定模态类和低相似度的不稳定模态类;然后对获取的稳定模态类进行层次聚类,将出现在不同计算阶次且具有相同模态参数的模态划分为一类,从而实现自动获取结构的每一个物理模态。
本发明的技术方案:
一种利用聚类自动提取结构模态参数的方法,步骤如下:
步骤一:获取不同计算阶次下模态参数
(1)利用自然激励方法处理结构响应Y(t)=[y(t),y(t+1),…,y(t+N)],其中y(t)=[y 1(t),y 2(t),…,y z(t)] T,N为样本时程点数,z为传感器个数;选定参考响应,获得各个时间延迟下的相关函数r(τ);
(2)利用相关函数矩阵构造如下形式的Hankel矩阵H ms(k-1)和H ms(k):
Figure PCTCN2018080922-appb-000001
(3)令k=1,对矩阵H ms(k-1)进行奇异值分解:
H ms(0)=USV T   (2)式中:U和V为酉阵;S为奇异值矩阵;
(4)令计算阶次j从2开始,依次增加2,对奇异值矩阵S按照计算阶次j进行截断获取新的奇异值矩阵S n,重复n u次,利用特征系统实现方法求出在各 计算阶次下的模态参数,其中,计算阶次j下的第i阶频率f ij、阻尼比ξ ij、模态振型
Figure PCTCN2018080922-appb-000002
和模态观测向量ν ij,i=1,2,…,j,j=2,4,…,2n u
(5)对任一阶次j下的每个模态i,根据频率误差与模态观测向量不相关之和最小,在其相邻阶次j+2下寻找与模态i最相似的模态p,进而获得频率误差df ij,p(j+2)、阻尼比误差dξ ij,p(j+2)、模态观测向量相关MOC ij,p(j+2),Δ ij,p(j+2)=df ij,p(j+2)+1-MOC ij,p(j+2)称为模态i的最邻近距离;
步骤二:稳定模态类和不稳定模态类划分
(6)对步骤(5)获取的每个模态与其最相似模态的频率误差序列df、阻尼比误差序列dξ、模态观测向量不相关1-MOC序列分别通过Box-Cox方法做正态变换,然后归一化变为标准正态分布序列d f s、dξ s和1-MOC s
(7)以服从标准正态分布的各偏差序列组成新的模态不相似度向量
Figure PCTCN2018080922-appb-000003
作为特征进行模糊C均值聚类,划分稳定模态类C 1和不稳定模态类C 2,聚类表达式为:
Figure PCTCN2018080922-appb-000004
式中:k表示聚类类别;b表示模糊度因子,b=2;η k表示模糊聚类的隶属度矩阵,其中元素η ij,k定义为阶次j下的模态i属于类k的隶属度:
Figure PCTCN2018080922-appb-000005
聚类中心:
Figure PCTCN2018080922-appb-000006
步骤三:从稳定图中提取物理模态
(8)对获得的稳定模态类C 1进行层次聚类,具体步骤为:
1)稳定模态类内各个模态自成一类;
2)两个距离最近的类归为同一类;
3)重复步骤2),直到各个类之间的最小距离超过容许值Δ lim
4)将类内样本数量超过阈值n T的选为物理类;
步骤2)中,类g中模态i与类l中模态h间距离:
Δ ig,hl=d f ig,hl+1-MOC ig,hl   (6)
同时,根据平均距离准则确定类间距离:
Figure PCTCN2018080922-appb-000007
式中:n g和n l分别表示类g和h的样本数;
各个类之间的最小距离超过容许值Δ lim根据步骤(5)中获得的稳定模态对应的最邻近距离分布的95%置信水平来确定,p(Δ≤Δ lim)=95%,样本数量阈值n T=(0.3~0.5)n u
(9)选取频率与物理类内所有频率平均值最接近的模态,作为最终的物理模态。
本发明的有益效果:通过对模态不相似度而不是模态参数本身进行模糊聚类,可自适应地划分稳定和不稳定模态,避免人为干预设定参数阈值,提高了模态参数识别的自动化程度。
附图说明
图1是稳定模态与不稳定模态分布。
具体实施方式
以下结合技术方案和技术方案,进一步阐明本发明的实施方式。
取一个8层剪切框架,各层质量均为1.10×10 6kg,各层刚度均为1541.07×10 6N/m,阻尼采用瑞利阻尼αM+βK,瑞利阻尼系数α=0.3000, β=0.0005,激励形式为零均值高斯白噪声激励,噪声水平设为随机响应方差的20%,采样频率为100Hz,采样信号为框架各层的加速度。
具体实施方式如下:
(1)对结构加速度响应Y(t)=[y(t),y(t+1),…,y(t+N)]进行自然激励法处理,以测点1作为参考点,获得各延时的相关函数向量,r(τ)=[r 11(τ) r 21(τ) … r 81(τ)] T,其中r ij(τ)表示测点i和测点j处加速度响应之间的互相关函数。
(2)令m=80,s=160;选取τ=1~239处的相关函数矩阵构造Hankel矩阵H ms(0);选取τ=2~240处的相关函数矩阵构造Hankel矩阵H ms(1)。
(3)对Hankel矩阵H ms(0)进行奇异值分解,获取奇异值矩阵S,其维数为640×160。
(4)初始计算阶次选为2,对奇异值矩阵S进行截断获得2×2的奇异值矩阵S n,进而利用特征系统实现算法获得频率f i2、阻尼比ξ i2、模态振型
Figure PCTCN2018080922-appb-000008
和模态观测向量ν i2;令计算阶次按2递增到80,共进行n u=40次循环,截断奇异值矩阵S并求出各计算阶次j所对应的频率f ij、阻尼比ξ ij、模态振型
Figure PCTCN2018080922-appb-000009
和模态观测向量ν ij
(5)对任一阶次j下的每个模态i,按照频率偏差与模态观测向量不相关之和最小,在j+2阶次下寻找模态i的最相似模态p,获得频率误差df ij,p(j+2)、阻尼比误差dξ ij,p(j+2)、模态观测向量相关MOC ij,p(j+2)和最邻近距离Δ ij,p(j+2)=df ij,p(j+2)+1-MOC ij,p(j+2)
(6)对(5)获取的每个模态与其最相似模态的频率误差序列df、阻尼比误差序列dξ、模态观测向量不相关1-MOC序列,分别通过Box-Cox方法做正态变换,然后再进行归一化变为标准正态分布序列d f s,dξ s和1-MOC s
(8)构造不相似度向量
Figure PCTCN2018080922-appb-000010
作为 阶次j下模态i的特征进行模糊C均值聚类来提取稳定模态类C 1,稳定模态类内的模态频率如图1所示,同时根据稳定模态的最邻近距离分布p(Δ≤Δ lim)=95%自动给出后续层次聚类的容许值Δ lim=0.154%。
(9)对获得的稳定模态类C 1按照公式(6)和公式(7)进行层次聚类,样本数量阈值n T=0.5n u获得8个物理模态类,选取频率与类内平均频率最近的模态作为物理模态代表,获得各阶固有频率和阻尼比分别为,f 1=1.152Hz,f 2=3.419Hz,f 3=5.571Hz,f 4=7.536Hz,f 5=9224Hz,f 6=10.624Hz,f 7=11.642Hz,f 2=12.278Hz;ξ 1=2.219%,ξ 2=1.254%,ξ 3=1.291%,ξ 4=1.459%,ξ 5=1.695%,ξ 6=1.903%,ξ 7=2.059%,ξ 8=2.152%。

Claims (1)

  1. 一种利用聚类自动提取结构模态参数的方法,其特征在于,步骤如下:
    步骤一:获取不同计算阶次下模态参数
    (1)利用自然激励方法处理结构响应Y(t)=[y(t),y(t+1),…,y(t+N)],其中y(t)=[y 1(t),y 2(t),…,y z(t)] T,N为样本时程点数,z为传感器个数;选定参考响应,获得各个时间延迟下的相关函数r(τ);
    (2)利用相关函数矩阵构造如下形式的Hankel矩阵H ms(k-1)和H ms(k):
    Figure PCTCN2018080922-appb-100001
    (3)令k=1,对矩阵H ms(k-1)进行奇异值分解:
    H ms(0)=USV T  (2)式中:U和V为酉阵;S为奇异值矩阵;
    (4)令计算阶次j从2开始,依次增加2,对奇异值矩阵S按照计算阶次j进行截断获取新的奇异值矩阵S n,重复n u次,利用特征系统实现方法求出在各计算阶次下的模态参数,其中,计算阶次j下的第i阶频率f ij、阻尼比ξ ij、模态振型
    Figure PCTCN2018080922-appb-100002
    和模态观测向量ν ij,i=1,2,…,j,j=2,4,…,2n u
    (5)对任一阶次j下的每个模态i,根据频率误差与模态观测向量不相关之和最小,在其相邻阶次j+2下寻找与模态i最相似的模态p,进而获得频率误差df ij,p(j+2)、阻尼比误差dξ ij,p(j+2)、模态观测向量相关MOC ij,p(j+2),Δ ij,p(j+2)=df ij,p(j+2)+1-MOC ij,p(j+2)称为模态i的最邻近距离;
    步骤二:稳定模态类和不稳定模态类划分
    (6)对步骤(5)获取的每个模态与其最相似模态的频率误差序列df、阻尼比误差序列dξ、模态观测向量不相关1-MOC序列分别通过Box-Cox方法做正态变换,然后归一化变为标准正态分布序列df s、dξ s和1-MOC s
    (7)以服从标准正态分布的各偏差序列组成新的模态不相似度向量
    Figure PCTCN2018080922-appb-100003
    作为特征进行模糊C均值聚类,划分稳定模态类C 1和不稳定模态类C 2,聚类表达式为:
    Figure PCTCN2018080922-appb-100004
    式中:k表示聚类类别;b表示模糊度因子,b=2;η k表示模糊聚类的隶属度矩阵,其中元素η ij,k定义为阶次j下的模态i属于类k的隶属度:
    Figure PCTCN2018080922-appb-100005
    聚类中心:
    Figure PCTCN2018080922-appb-100006
    步骤三:从稳定图中提取物理模态
    (8)对获得的稳定模态类C 1进行层次聚类,具体步骤为:
    1)稳定模态类内各个模态自成一类;
    2)两个距离最近的类归为同一类;
    3)重复步骤2),直到各个类之间的最小距离超过容许值Δ lim
    4)将类内样本数量超过阈值n T的选为物理类;
    步骤2)中,类g中模态i与类l中模态h间距离:
    Δ ig,hl=df ig,hl+1-MOC ig,hl  (6)
    同时,根据平均距离准则确定类间距离:
    Figure PCTCN2018080922-appb-100007
    式中:n g和n l分别表示类g和h的样本数;
    各个类之间的最小距离超过容许值Δ lim根据步骤(5)中获得的稳定模态对应的最邻近距离分布的95%置信水平来确定,p(Δ≤Δ lim)=95%,样本数量阈值n T=(0.3~0.5)n u
    (9)选取频率与物理类内所有频率平均值最接近的模态,作为最终的物理模态。
PCT/CN2018/080922 2018-02-26 2018-03-28 一种利用聚类自动提取结构模态参数的方法 WO2019161592A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US16/342,948 US20200074221A1 (en) 2018-02-26 2018-03-28 Automatic method for structural modal estimation by clustering

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810159710.7 2018-02-26
CN201810159710.7A CN108388915A (zh) 2018-02-26 2018-02-26 一种利用聚类自动提取结构模态参数的方法

Publications (1)

Publication Number Publication Date
WO2019161592A1 true WO2019161592A1 (zh) 2019-08-29

Family

ID=63069250

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/080922 WO2019161592A1 (zh) 2018-02-26 2018-03-28 一种利用聚类自动提取结构模态参数的方法

Country Status (3)

Country Link
US (1) US20200074221A1 (zh)
CN (1) CN108388915A (zh)
WO (1) WO2019161592A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021142901A1 (en) * 2020-01-15 2021-07-22 Dalian University Of Technology Method of physical mode extraction for engineering structure flexibility identification
CN116400244A (zh) * 2023-04-04 2023-07-07 华能澜沧江水电股份有限公司 储能电池的异常检测方法及装置
CN116796214A (zh) * 2023-06-07 2023-09-22 南京北极光生物科技有限公司 一种基于差分特征的数据聚类方法
CN117851464A (zh) * 2024-03-07 2024-04-09 济南道图信息科技有限公司 一种用于心理评估的用户行为模式辅助分析方法

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110059353B (zh) * 2019-03-14 2022-10-14 长安大学 一种模态参数自动识别简化实用方法
CN111144206B (zh) * 2019-11-21 2021-02-09 东南大学 一种柔性结构中立方非线性系统识别方法
CN111709350B (zh) * 2020-03-03 2022-05-13 天津大学 基于fcm聚类的低频振荡模态参数识别方法及系统
CN111898664B (zh) * 2020-07-22 2023-02-21 福建农林大学 一种基于Block-Bootstrap和多阶段聚类的桥梁模态参数自动识别方法
US11983278B2 (en) * 2021-03-18 2024-05-14 Tata Consultancy Services Limited System and method for data anonymization using optimization techniques
CN116755532B (zh) * 2023-08-14 2023-10-31 聊城市洛溪信息科技有限公司 一种计算服务器通风装置用智能调控系统
CN117725394B (zh) * 2024-02-18 2024-05-07 浙江浙能技术研究院有限公司 基于分层内嵌模态分解的风电场宽频振荡辨识方法

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2012130237A1 (en) * 2011-03-30 2012-10-04 Rune Brincker Method for improving determination of mode shapes for a mechanical structure and applications hereof
CN106777763A (zh) * 2016-12-31 2017-05-31 大连理工大学 一种工程结构虚假模态准确判别方法
CN107609291A (zh) * 2017-09-22 2018-01-19 哈尔滨工业大学 一种基于密度聚类的虚假模态剔除方法

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2012130237A1 (en) * 2011-03-30 2012-10-04 Rune Brincker Method for improving determination of mode shapes for a mechanical structure and applications hereof
CN106777763A (zh) * 2016-12-31 2017-05-31 大连理工大学 一种工程结构虚假模态准确判别方法
CN107609291A (zh) * 2017-09-22 2018-01-19 哈尔滨工业大学 一种基于密度聚类的虚假模态剔除方法

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021142901A1 (en) * 2020-01-15 2021-07-22 Dalian University Of Technology Method of physical mode extraction for engineering structure flexibility identification
CN116400244A (zh) * 2023-04-04 2023-07-07 华能澜沧江水电股份有限公司 储能电池的异常检测方法及装置
CN116400244B (zh) * 2023-04-04 2023-11-21 华能澜沧江水电股份有限公司 储能电池的异常检测方法及装置
CN116796214A (zh) * 2023-06-07 2023-09-22 南京北极光生物科技有限公司 一种基于差分特征的数据聚类方法
CN116796214B (zh) * 2023-06-07 2024-01-30 南京北极光生物科技有限公司 一种基于差分特征的数据聚类方法
CN117851464A (zh) * 2024-03-07 2024-04-09 济南道图信息科技有限公司 一种用于心理评估的用户行为模式辅助分析方法
CN117851464B (zh) * 2024-03-07 2024-05-14 济南道图信息科技有限公司 一种用于心理评估的用户行为模式辅助分析方法

Also Published As

Publication number Publication date
US20200074221A1 (en) 2020-03-05
CN108388915A (zh) 2018-08-10

Similar Documents

Publication Publication Date Title
WO2019161592A1 (zh) 一种利用聚类自动提取结构模态参数的方法
CN109657945B (zh) 一种基于数据驱动的工业生产过程故障诊断方法
CN106658590B (zh) 基于WiFi信道状态信息多人室内环境状态监控系统的设计与实现
WO2020248228A1 (zh) 一种Hadoop平台计算节点负载预测方法
CN111562108A (zh) 一种基于cnn和fcmc的滚动轴承智能故障诊断方法
CN111985533B (zh) 一种基于多尺度信息融合的增量式水声信号识别方法
CN112633098B (zh) 一种旋转机械故障诊断方法、系统及存储介质
WO2019173943A1 (zh) 一种自动追踪结构模态参数的方法
CN108415884B (zh) 一种结构模态参数实时追踪方法
CN108171119B (zh) 基于残差网络的sar图像变化检测方法
CN107798351B (zh) 一种基于深度学习神经网络的身份识别方法和系统
WO2019104904A1 (zh) 一种基于稀疏分量分析模态识别中的阶数确定方法
CN111612130B (zh) 一种频移键控通信信号调制方式识别方法
Scionti et al. Tools to improve detection of structural changes from in-flight flutter data
CN114355298B (zh) 一种雷达复合调制脉冲信号识别方法
CN112070130A (zh) 一种基于动态时间规整的跨道车辆匹配方法及系统
CN111274630A (zh) 一种用于工程结构柔度识别的物理模态提取方法
CN113515678A (zh) 一种异常数据筛选方法
CN111273126B (zh) 一种配电网拓扑快速感知方法
Chauhan et al. Clustering approaches to automatic modal parameter estimation
CN114781083A (zh) 一种发动机稳态数据层次分析与稳态数据特征值提取方法
CN112016539B (zh) 信号识别方法、装置、电子设备与存储介质
Jazizadeh et al. Self-configuring event detection for electricity disaggregation
Fomin et al. Construction of diagnostic features space using Volterra kernels moments
CN112215246B (zh) 道路交通诊断方法、装置、电子设备及机器可读存储介质

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: 18907393

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: 18907393

Country of ref document: EP

Kind code of ref document: A1