Risk-based fault prediction of chemical processes using operable adaptive sparse identification of systems (OASIS). (September 2021)

• Record Type: Journal Article
• Authors: Bhadriraju, Bhavana; Kwon, Joseph Sang-Il; Khan, Faisal
• Is Part Of: Computers & chemical engineering. Volume 152(2021)
• Publication Date: 2021-09
• Subjects: Sparse identification -- Adaptive model -- Risk assessment -- Fault prediction; Chemical engineering -- Data processing -- Periodicals; 660.0285
• DOI: 10.1016/j.compchemeng.2021.107378
• Languages: English
• ISSNs: 0098-1354

Abstract

Highlights: OASIS framework is used to develop a risk-based fault prediction framework. Sparse regression and deep learning are combined to predict nonlinear dynamics. Risks are assessed and monitored dynamically to predict faults. The developed framework can handle uncertainties and abrupt changes. Abstract: Fault prediction has arisen as a basic monitoring strategy that predicts an abnormal event occurring in near future based on the current symptoms observed in a process. Such a proactive approach helps in taking an appropriate action beforehand so as to mitigate the impact a fault can have on a process. Recently, data-driven modeling techniques have been widely used due to an increased accessibility to process data. Though the offline trained models are successful in modeling complex dynamics, they have limited ability in capturing the dynamic process behavior, especially under abnormal conditions. To address this issue, we utilize an adaptive modeling technique called operable adaptive sparse identification of systems (OASIS) that can cope with any dynamical changes. Based on the forecasted process behavior using OASIS, we perform risk-assessment to predict faults and assess risk. In the proposed method, risk is used as a criteria to monitor and manage process operation.