In order to overcome the dynamic and large-scale characteristics of the plant-wide processes, this paper proposed a distributed slow feature analysis (SFA) with inter-unit dissimilarity method for process monitoring task. Firstly, to highlight the local dynamic features, the whole process is decomposed into several units according to the prior knowledge. Based on this, SFA monitoring model is built parallelly to handle the dynamic features. Considering the possible information loss caused by the process decomposition, the inter-unit dissimilarity index is carried out to monitor the variations between adjacent units. Finally, the fusion center is conducted by Bayesian inference to combine the results of SFA monitoring models and inter-unit dissimilarity statistics. The effectiveness of the proposed method is tested on the Tennessee Eastman process and an aluminum electrolysis process.

为了克服全厂过程的动态和大规模特性，本文提出了一种分布式慢特征分析（SFA）和单元间差异方法，用于过程监控任务。首先，为了突出局部动态特征，将整个过程根据先验知识分解为几个单元。在此基础上，并行构建 SFA 监控模型来处理动态特征。考虑到过程分解可能造成的信息丢失，采用单元间相异性指标来监测相邻单元之间的差异。最后，通过贝叶斯推理进行融合中心，将SFA监测模型的结果与单元间差异统计相结合。