| dc.description.abstract | The advancement in technology and computational power has enabled large amounts of data collection in real time, which has initiated the "Big Data" era. Big data analytics is playing an essential role in academy, business as well as government, providing assistance in decision-making in numerous fields. In this work, selected challenges in process systems engineering are addressed through advances of and applications in big data analytics. First, challenges in chemical process monitoring, such as fault detection and diagnosis, are addressed by exploiting the industrial data abundance. Data-driven process monitoring has become one of the key approaches in industry to maintain a safe and robust operation while increasing process efficiency to ensure high standards in product quality. In this work, a novel fault detection and diagnosis framework based on nonlinear Support Vector Machine-based feature selection and modeling algorithm is developed for the simultaneous fault detection and diagnosis of chemical processes (s-FDD framework) in both continuous and batch modes. The major advantage of the s-FDD framework is its ability to identify the optimal number of process variables diagnosing the fault while providing highly accurate models for fault detection. The s-FDD framework is further improved with the integration of (i) maintenance optimization strategies, and (ii) multi-parametric model predictive control (mp-MPC) in order to maximize the process profitability and resilience while minimizing process downtime. A novel "parametric fault-tolerant control" concept has been developed for chemical/biochemical processes that serves as an active fault tolerant strategy.
This work can serve as an online decision support tool during process operations to enable (i) early detection and diagnosis of process faults, and (ii) rapid actions to adapt altering process or controller conditions to achieve smarter operation. Secondly, we address challenges in understanding the environmental health impact of complex substance/mixture exposures during environmental emergency-related contamination events (i.e. hurricanes). A data-driven framework is developed to group complex substances with known chemicals by analyzing high dimensional analytical chemistry data, and predict their impact on the environmental health. This facilitates the communication of substance characteristics and decision-making via read-across in order to mitigate the adverse environmental health effects. | en |
