Multimode Process Monitoring Based on Modified Density Peak Clustering and Parallel Variational Autoencoder

Clustering algorithms and deep learning methods have been widely applied in the multimode process monitoring. However, for data with unknown mode, traditional clustering can hardly identify number of modes automatically. Further, learn effective features from nonlinear data, while extracted cannot follow Gaussian distribution, which may lead to incorrect control limit fault detection. In this paper, a comprehensive monitoring method based on modified density peak parallel variational autoencoder (MDPC-PVAE) is proposed processes. Firstly, novel algorithm, named MDPC, presented mode identification division. MDPC without prior knowledge information divide whole into multiple modes. Then, PVAE established distinguished generate features, generated each VAE distribution. Finally, feature representations obtained by are provided construct statistics H2, limits determined kernel estimation (KDE) method. The effectiveness evaluated Tennessee Eastman semiconductor etching process.