基于稀疏核主角的在线故障检测  

Online Faults Detection Based on Sparse Kernel Principal Angles

作　　者：甘良志[1] 

机构地区：[1]徐州师范大学泉山校区电气工程及自动化学院,徐州221116

出　　处：《微计算机信息》2010年第25期147-149,共3页Control & Automation

摘　　要：为了满足对于复杂生产过程监测的需要,本文提出了一种基于稀疏核主角的在线故障检测方法。核函数方法能够通过非线性映射在高维特征空间中解决原始空间中的非线性问题,但是可能导致过学习。稀疏核主角方法首先将原始空间中的数据集映射到高维特征空间中,然后在高维特征空间中求这组数据的基(或近似基),用这组基或近似基的线性组合来描述核主角,不但解具有稀疏性,而且能够有效提高学习机器的推广能力。实验结果证明,稀疏核主角能够在线检测TE过程中的故障发生情况。In this paper,an online nonlinear process detecting technique based on sparse kernel principal angles（SKPA） is proposed.Kernel method can efficiently solve nonlinear problem with nonlinear kernel functions,but can also lead to over learning.The basic idea of SKPA is to first map the input data into a Reproducing Kernel Hilbert Space via nonlinear mapping and then to compute the basis or approximate basis of the data in feature space.The basis or approximate basis can represent all the samples by the way of linear combination.By this way,the principal angles can be represented by the basis or approximate basis,so the solution is sparse.According to Vapnik＇s theory,SKPA can get better generalization.In this paper,SKPA are applied to fault detection in one example system：the simulation benchmark of the Tennessee Eastman（TE） process.The proposed online approach effectively captures the faults in the TE process.

关 键 词：核主角 主分量分析 再生核希尔伯特空间 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]