基于改进核主元分析的故障检测方法研究

发布时间：2019-01-05 05:27

【摘要】：传统KPCA能有效捕捉数据的非线性特征、对数据在输入空间中的分布情况没有任何要求等优点,但是由于核函数的参数选择直接影响到核函数本身的性能,所以提出使用二分法进行核参数优化,获得改进核主元分析的故障检测方法。该方法引入混合核函数的定义,即多项式核函数和高斯径向基核函数的混合核方法,使用二分法分别对上述两核进行核参数优化,确定最优的混合核函数,再与主元分析方法(Principal Component Analysis PCA)相结合,得到改进KPCA。根据混合非线性主元特征计算出的T2和SPE统计量,从而实现故障检测。使用二分法进行核参数优化,选取核函数,改进了传统的核函数的选取方法。并且充分的考虑了工业过程中的非线性,更精确的描述工业过程特性,使得本文提出的改进方法可以准确有效地检测出电主轴系统故障。对田纳西-伊斯曼(TE)过程以及电主轴系统的应用实例说明了该方法的实用可行性。
[Abstract]:The traditional KPCA can effectively capture the nonlinear characteristics of the data, and has no requirement for the distribution of the data in the input space. However, the parameter selection of the kernel function directly affects the performance of the kernel function itself. Therefore, a fault detection method based on improved kernel principal component analysis (PCA) is proposed to optimize kernel parameters by dichotomy. This method introduces the definition of mixed kernel function, that is, the mixed kernel method of polynomial kernel function and Gao Si radial basis kernel function. The binary method is used to optimize the kernel parameters of the two cores, and the optimal hybrid kernel function is determined. Combining with principal component analysis (Principal Component Analysis PCA) method, an improved KPCA. is obtained. The T _ 2 and SPE statistics calculated according to the mixed nonlinear principal component features are used to realize fault detection. The traditional kernel function selection method is improved by using dichotomy to optimize kernel parameters and select kernel function. Considering the nonlinearity of the industrial process and describing the characteristics of the industrial process more accurately, the improved method proposed in this paper can detect the fault of the motorized spindle system accurately and effectively. The application examples of Tennessee Eastman (TE) process and motorized spindle system show the feasibility of this method.
【作者单位】：沈阳建筑大学机械工程学院;
【分类号】：TP277

【相似文献】