Abstract
Customized production, quality variation of raw materials and other factors make industrial processes work in multiple operating modes. In general, complex industrial processes have strong nonlinearity under each operating mode. In this paper, a Gaussian mixture variational autoencoder (GMVAE) model, which combines with Gaussian mixture and VAE, is proposed to monitor nonlinear processes with multiple operating modes. Due to the Gaussian mixture distribution limitation in latent variable space, GMVAE can not only automatically extract features of the nonlinear system, but also make these features follow Gaussian mixture distribution. Based on Gaussian mixture distribution in latent variable space and the reconstruction error, two probability monitoring indexes are constructed, whose control limits can be determined by χ <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> distribution. TE benchmark data and real hot strip mill process (HSMP) data have been used to verify the effectiveness of the proposed method.
Highlights
Industrial process monitoring and fault diagnosis (PM-FD), an essential measure to ensure process safety and product quality stability, has been widely used in industrial production
Data-driven process monitoring methods have been attracted considerable attention in theory research and practical application [3]–[5], especially these methods based on multivariate statistical process monitoring (MSPM), such as Principal Component Analysis (PCA) and Partial Least Squares (PLS) [6]–[8]
In order to ensure the projections in latent variable space under each operating mode follow Gaussian distribution, Gaussian mixture variational autoencoder (GMVAE) model is redesigned
Summary
Industrial process monitoring and fault diagnosis (PM-FD), an essential measure to ensure process safety and product quality stability, has been widely used in industrial production. The GMVAE structure proposed in the paper [54] is modified to make it suitable for monitoring of nonlinear processes with multiple operating modes It projects the multimode data with complex distribution into the latent variable space by nonlinear mapping. Based on the probabilistic graphical model, a joint probability index of each mode component in latent variable space and a probability index for reconstruction probability distribution are further defined to detect process faults. In order to ensure the projections in latent variable space under each operating mode follow Gaussian distribution, GMVAE model is redesigned. The process data are mapped into latent variable space by model qφz (z|x), and the probability that belong to each local Gaussian components is calculated by model qφy (y|x). The process is regarded as faulty when LVPI(xi) > α
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
