摘要
针对传统时频分析方法分解不准确、效率低下的问题,提出一种改进的自适应变分模态分解(AVMD)方法,该方法预先使用短时傅里叶变换预估模态数量,并对原始信号频谱与分量叠加频谱进行谱相关分析筛选最优惩罚因子,提高变分模态分解(VMD)的精确性,与经验模态分解(EMD)、聚合经验模态分解(EEMD)、小波变换相比,该方法分解速度快、准确度高。之后,结合AVMD和谱相关分析提出一种新的滚动轴承故障诊断方法,该方法首先采用AVMD将已知故障信号分解成若干本征模态,并使用主要成分分析(PCA)降维去噪后构成故障模型库;然后对新采集的检测信号进行相同处理得到检测特征向量;最后将检测向量和故障库故障库特征向量分别进行频域内谱相关性分析和判别,实现故障诊断。使用西储大学实验台轴承数据和实际风场采集数据对该方法进行验证,诊断结果表明该方法相比于传统方法,识别率有明显提高。
Most traditional algorithms based on the frequency domain are not accurately and efficiently decompose signals.A new method for fault diagnosis is proposed based on adaptive variational mode decomposition(AVMD)and spectral correlation analysis.Short-time Fourier transform is utilized to estimating the number of intrinsic mode functions(IMFs)in AVMD method.In order to select the optimal penalty factor,correlation analysis is carried out on the raw signal spectrum and spectrum of the signal components.The results indicate that AVMD not only improves the accuracy of VMD but is more accurate and faster than EMD,EEMD and wavelet transform.By using the AVMD raw signal is decomposed into several IMFs,and then PCA is performed to simplify and de-noise IMFs to construct the fault feature library.Then the detection signal is processed by the same procedure to obtain the feature vectors.Finally,spectrum correlation analysis of detection vector and fault feature library vectors are carried out to realize fault diagnosis.The results of experimental analysis show that the proposed method performs better on diagnosis accuracy than the traditional method.
作者
齐咏生
白宇
高胜利
李永亭
Qi Yongsheng;Bai Yu;Gao Shengli;Li Yongting(Institute of Electric Power,Inner Mongolia University of Technology,Huhhot 010080,China;Inner Mongolia North Longyuan Wind Power Co.,Ltd.,Hohhot 010050,China)
出处
《太阳能学报》
EI
CAS
CSCD
北大核心
2019年第7期2053-2063,共11页
Acta Energiae Solaris Sinica
基金
国家自然科学基金(61763037
21466026)
内蒙古自治区自然科学基金(2017MS0601)
