The application of a supervised deep convolutional autoencoder was tested against partial least-squares-discriminant analysis (PLS-DA) for fault detection and diagnosis in a penicillin fed-batch process. In silico data was generated with a comprehensive simulator (IndPenSim) of an industrial-scale penicillin fed-batch simulator of operation under normal batch conditions and 8 fault batch conditions. A composite face-centered design response surface was applied to optimize key bioreactor design parameters based on a profit function that was directly dependent on fault detection results. The application of PLS-DA and the NN modeling resulted in an average fault detection rate (FDR) across all faults and process parameter configurations of 72.5% and 95.9%, respectively. In classifying complex fault conditions, the deep learning model greatly surpassed the PLS-DA model, and this improvement translated into a 25.0% increase in realized profit with the supervised deep convolutional autoencoder when compared to PLS-DA monitoring.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.