Systems Engineering Approaches to the Study of Industrial Processes and Biological Systems
Abstract
Large-scale industrial processes and biological systems share many similarities at the systems level: they both
consist of many individual components; they both have built-in feedback control/regulation mechanisms; and
the properties of the overall systems are determined by the complex interaction among different components. Their complex nature makes the integrative systems approaches essential in understanding, controlling and
optimizing these systems. As a result, many process systems engineering principles and techniques have been
extended into the emerging field of systems biology. However, despite their commonalities at the system level, large-scale industrial processes and biological
systems have their unique characteristics and challenges that existing systems approaches cannot fully
address. In this talk, our most recent progress in both research areas (process systems engineering and systems
biology) will be presented. For large-scale industrial processes, one of our focuses is process monitoring. The specific challenge we aim to address is how to effectively handle process nonlinear dynamics,
non-Gaussianity, frequent process changes driven by manufacturing on-demand, but without the heavy
computational burden of available nonlinear methods. The solution we developed is a new multivariate
framework named statistics pattern analysis (SPA) and we use the benchmark Tennessee Eastman Process to
demonstrate the effectiveness of the new framework. For biological systems, one specific challenge we aim
to address is how to effectively utilize genome-wide metabolic network models and extract biological
meaningful information from them. The solution we developed is a system identification based approach
where we use the metabolic models as a high fidelity simulator to conduct carefully designed in silico experiments. We will use scheffersomyces stiptis (the yeast with the strongest native capability to ferment
xylose) as the model system to illustrate our developed approach.