基于KICA-KFDA的集成故障识别算法  

Integrated Fault Identification Algorithm Based on KICA and KFDA

作　　者：许洁[1] 赵瑾[1] 刘如成[1] 胡寿松[2] 

机构地区：[1]南京师范大学电气与自动化工程学院,南京210042 [2]南京航空航天大学自动化学院,南京210016

出　　处：《数据采集与处理》2013年第6期812-817,共6页Journal of Data Acquisition and Processing

基　　金：国家自然科学基金重点(60234010)资助项目;航空科学基金(05E52031)资助项目;国家自然科学青年基金(61203092)资助项目;江苏省高校自然科学研究(11KJB510007)资助项目

摘　　要：针对复杂的化工过程,提高过程监控能力,提出基于核独立成分分析(Kernel independent component analysis,KICA)和核Fisher判别分析(Kernel fisher discriminant analysis,KFDA)的过程监测与故障识别方法。通过利用核独立成分分析建立正常工况模型,得到检测故障信息。在发生故障的情况下,利用Fisher判别分析方法在高维的特征空间的特点和优势,可求出满足最大分离程度的核Fisher判别向量和特征向量,根据当前故障的判别向量和历史故障数据集中所含故障的最优核Fisher判别向量的相似度进行故障识别。仿真结果验证了所提方法的有效性。To improve the statistical monitoring performance of complex chemical process, a new statistical process monitoring and fault identification method based on kernel independent component analysis （KICA） and kernel fisher discriminant analysis （KFDA） is proposed, hav- ing the character of nonlinear. KICA is used to establish the normal operating conditions and i- dentify the fault. If a fault occurs, the nuclear fisher discriminant vector and feature vector of the process data are extracted from the Fisher subspace. Thus, it can be detected if the batch is normal by comparing the distance with the predefined threshold. Comparing the present dis- criminant vector and the optimal one of fault in historical data set, the similar degree can be i- dentified. According to it, the perform fault can be diagnosed. The simulation results demon- strate that the proposed method effectively detect and diagnose the malfunctions.

关 键 词：故障识别 过程监控 核独立成分分析 核FISHER判别分析 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]