轴承故障信号识别经常受到各种噪音的影响,传统K奇异值分解(K-Singular value decomposition,K-SVD)算法在稀疏表示中效果较差,通过终止准则对进K-SVD字典学习优化,设计了基于改进K-SVD稀疏表示的轴承微弱故障信号特征处理方法.将终止准则当作字典更新收敛条件,采取正交匹配追踪算法进行稀疏求解,以包络谱形式实施分析,达成对微弱故障特征的提取目标.仿真信号结果表明,添加噪声信号时域图难以对特征频率实施精准提取.通过改进K-SVD算法来学习该分量特征信息有着明显的冲击特征,通过重构误差的波动状况对更新收敛性验证.试验结果结果表明,故障特征频率被其它频率掩盖,导致故障状态难以被有效辨别.本文方法实现对微弱故障特征的高效提取,精准判断故障状态.
Bearing Weak Fault Signal Feature Processing Based on Improved K-SVD Dictionary Learning Algorithm
Bearing fault signal recognition is often affected by various noises,and the traditional K-singular value decomposition(K-SVD)algorithm performs poorly in sparse representation.Therefore,the K-SVD dictionary learning is optimized by stopping criteria.A novel feature processing method for bearing weak fault signals based on improved K-SVD sparse representation is designed.Taking the termination criterion as the convergence condition of dictionary update,an orthogonal matching tracking algorithm with optimized residual threshold is adopted to sparsely solve vibration signals,and envelope spectrum analysis is carried out to achieve the extraction of weak fault features.The simulation results show that it is difficult to extract characteristic frequency accurately by adding noise signal in the time domain map.The K-SVD algorithm is improved to learn that the component feature information has obvious impact characteristics,and the convergence of the update is verified by the fluctuation of the reconstruction error.The test results show that the fault characteristic frequency is covered by other frequencies,which makes it difficult to distinguish the fault state effectively.The scheme in this paper realizes efficient extraction of weak fault features and accurate judgment of fault state.
bearingsparse representationfault characteristicsK singular value decompositiontermination critenon
NETL
NSTL
万方数据
