Multi-directional principal component analysis process monitoring method based on alarm reliability fusion

The invention relates to a multi-directional principal component analysis process monitoring method based on alarm reliability fusion, which comprises model data acquisition, three-dimensional data expansion, two-dimensional matrix normalization, PCA decomposition, principal component number selection, SPE and T2 statistical control limit calculation, fuzzy membership function establishment, alarmreliability and global alarm reliability weight calculation and alarm decision making. The limitation of statistics of control limit alarms by the multi-directional principal component analysis method model is overcome, alarm decision making is carried out by using a method of fusing historical alarm reliability and current alarm reliability, the false alarm rate of multi-directional principal component analysis in application of an intermittent process monitoring process can be greatly reduced, and a new way is provided for modeling and monitoring of the intermittent process. 本发明涉及一种基于报警信度融合的多向主元分析过程监测方法，包括模型数据采集、三维