Concurrent nonstationary process analysis model and its application in nonstationary process monitoring

Over the past decades, effective monitoring and high product quality of nonstationary process have been considered as challenging tasks due to its sophisticated data characteristics. In this paper, a concurrent nonstationary process analysis (CNPA) strategy is proposed for nonstationary process monitoring. Firstly, the original data space is decomposed into several subspaces according to the nonstationary characteristics and their relationship with quality variables. Secondly, corresponding latent variables are developed to describe the behavior of each subspace, which will then be applied in process monitoring. The efficacy of the proposed method is confirmed by a numerical case and an application in penicillin fermentation process. Compared with previous researches, this method exhibits better performance in fault detection for nonstationary processes.