Abstract: Considering the insufficiency of traditional monitoring methods to neglect dynamic information in the process industry, the study proposes a novel online feature reordering and feature selection based on slow feature analysis algorithm (improved FROSSFA), which can expand the SFA fault detection method to the field of quality-related fault detection. Finally, the proposed method is utilized in the process of Tenness-Eastman, and the results show that the improved FROSSFA method has higher fault detection rate, and it can determine whether the fault is related to the quality accurately.

Key words: Slow Feature Algorithm, quality-related fault detection, online monitoring

摘要： 考虑到传统工业过程监测方法容易忽略掉过程动态信息的现象，提出了基于慢特征分析（Slow Feature Analysis, SFA）的在线特征选择与重排序的质量相关的故障检测方法（改进FROSSFA），将SFA故障检测扩展到质量相关的故障检测领域。最后将所提方法应用在TE过程上，仿真结果表明改进的FROSSFA方法具有较高的故障检测率，并能准确判断所发生故障是否与质量相关。

关键词: 慢特征分析, 质量相关的故障检测, 在线监测