约束KICA信号分离方法及在滚动轴承故障诊断中的应用研究

发布时间：2019-01-27 20:39

【摘要】：滚动轴承是旋转机械的核心部件，其性能直接影响到整个机械系统运行状态。但是由于加工工艺复杂，工作环境恶劣等因素的存在，导致了轴承具有较高的故障率，因此轴承的故障诊断一直是一个重要的研究课题。滚动轴承实测振动信号往往会有多个振动源的同时存在，表现出非平稳、非线性，这些都严重的影响诊断的准确性和可靠性。独立成分分析（ICA）是近几年发展起来的基于信号高阶统计特性的分析方法，它能仅通过多路混合信号就能将相互独立的分量分离出来，但是目前该算法对于处理非线性变化的信号有一定的局限性。而本文使用的核独立成分分析(KICA)是将核方法的非线性解决能力以及ICA的高阶统计特性结合起来的全新算法，它能有效的捕捉数据的非线性特征，在处理高阶相关性数据时表现出特有优势。尽管核方法在很多问题上有着良好表现，但是核函数和核参数的选择目前缺少统一定则，在实际应用中只能凭借经验确定；同时基于KICA的信号处理存在着不确定性、源信号估计以及信号信息使用单一问题，针对上述问题本论文开展以下工作：（1）归纳了旋转机械核心部件滚动轴承的典型故障机理，开展了滚动轴承故障振动信号采集实验，，同时论述了滚动轴承故障诊断中ICA、核方法以及KICA方法原理，作为后续研究的理论基础。（2）针对机械故障诊断中KICA算法的核函数及核参数难以确定问题，构造了基于改进向量夹角余弦值的相似度函数，提出以KICA分离前后信号相似度作为KICA信号分离效果评判标准，对滚动轴承故障诊断中KICA的核参数进行寻优研究。通过仿真实验确定最适于滚动轴承故障诊断的核函数类型及其参数取值范围，并用于指导实测滚动轴承振动信号的分析与诊断。（3）针对信号的KICA分离中的不确定性、源信号估计问题以及信号信息使用单一的问题提出提出了基于约束KICA的滚动轴承故障诊断模型，旨在通过滚动轴承故障机理先验知识构建脉冲参考信号用于KICA信号分离的约束条件，提高故障信号分离的效果。研究了信号通道数以及脉冲参考信号频率、相位以及脉冲宽度对算法性能的影响，并结合滚动轴承故障实验台测试的数据对该方法的特征信号提取进行验证，证明了基于脉冲参考信号的约束KICA对滚动轴承故障诊断的有效性。
[Abstract]:Rolling bearing is the core component of rotating machinery, and its performance directly affects the running state of the whole mechanical system. However, because of the complex processing technology and poor working environment, the bearing has a high failure rate, so the bearing fault diagnosis has been an important research topic. The measured vibration signals of rolling bearings often have multiple vibration sources at the same time, which show non-stationary and nonlinear, which seriously affect the accuracy and reliability of diagnosis. Independent component analysis (ICA) is an analytical method based on the high order statistical characteristics of signals developed in recent years. It can separate independent components from each other only by multichannel mixed signals. However, the algorithm has some limitations in dealing with nonlinear signals. The kernel independent component analysis (KICA) used in this paper is a new algorithm which combines the nonlinear solution of the kernel method and the higher-order statistical characteristics of the ICA. It can effectively capture the nonlinear characteristics of the data. It has a unique advantage in dealing with high-order correlation data. Although the kernel method has good performance on many problems, the selection of kernel function and kernel parameters is lack of uniform rules, and can only be determined by experience in practical application. At the same time, there is uncertainty in signal processing based on KICA. To solve the above problems, the following works are carried out: (1) the typical fault mechanism of rolling bearing, which is the core component of rotating machinery, is summarized, and the vibration signal collection experiment of rolling bearing is carried out. At the same time, the ICA, kernel method and the principle of KICA method in rolling bearing fault diagnosis are discussed. (2) aiming at the difficulty of determining the kernel function and kernel parameters of KICA algorithm in mechanical fault diagnosis, a similarity function based on the cosine value of the improved vector angle is constructed. In this paper, the signal similarity before and after KICA separation is used as the criterion for evaluating the effect of KICA signal separation, and the kernel parameters of KICA in fault diagnosis of rolling bearing are studied. The type of kernel function and the range of parameters are determined by simulation experiments, which are used to guide the analysis and diagnosis of vibration signals of rolling bearings. (3) aiming at the uncertainty in KICA separation, the source signal estimation problem and the single problem of signal information use, a rolling bearing fault diagnosis model based on constrained KICA is proposed. The purpose of this paper is to construct the constraint condition of pulse reference signal for KICA signal separation by prior knowledge of rolling bearing fault mechanism, and to improve the effect of fault signal separation. The effects of signal channel number, pulse reference signal frequency, phase and pulse width on the performance of the algorithm are studied. The validity of constrained KICA based on impulse reference signal for rolling bearing fault diagnosis is proved.
【学位授予单位】：湖南科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TH133.33;TH165.3

【相似文献】