动态广义主成分分析及其在故障子空间建模中的应用

Dynamic generalized principal component analysis with applications to fault subspace modeling

针对传统故障子空间建模方法未考虑故障数据中同时包含正常工况信息和故障工况信息的实际情况,或未考虑故障数据中的动态因素而导致的对故障子空间提取不够准确的问题,提出了一种动态广义主成分分析方法。通过将带延迟的输入数据进行空间重组,采用广义主成分分析方法提取正常工况和各故障工况之间的动态特征信息,实现对故障子空间的准确建模,并进一步建立故障库实现故障诊断。仿真结果表明,所提方法能够准确提取动态过程的故障子空间,并可用于动态工业过程的故障诊断。

In order to solve the problem of inaccurate modeling of fault subspace,traditional fault subspace modeling method did not consider the fact that fault data contain both normal and fault condition information,or did not consider the dynamic factors in the fault data,these flaws may lead to the case that the fault subspace cannot be extracted accurately,a dynamic generalized principal component analysis(DGPCA)method was proposed.By reorganizing the lagged input data,the dynamic characteristics between normal and fault data were extracted by the proposed DGPCA method,and then the fault subspaces could be modeled for further fault diagnosis.Finally,simulation results confirm the availability of the proposed method for fault subspace modeling and fault diagnosis.