Control structure design and advanced process control
Graph-theoretic method for integrated design and control
Control structure design, i.e. the selection and pairing of manipulated inputs and controlled outputs, is a classic problem in control that has received a lot of attention in the literature. In process control in particular, this problem has been studied extensively in the context of plant-wide control design. Traditionally, control structure design problems are considered once steady state process design is fixed. However, optimal process design does not always translate into the ease of process operation and control. It rather makes process operation and control more challenging since optimal process design generally has very small process margins. Thus, it is important to consider the process design problem and control structure design problem simultaneously. Our research aims to address this problem by developing graph-theoretic methods which can identify promising process design configurations with corresponding control structures in an automated fashion.
Associated members: Ji Hee Kim, Sunwook Kim, Hyunji Kwon
Process monitoring using machine learning techniques
Statistical process monitoring is one of the most important research problems for modern process industry to ensure sustainable operation. The key step of statistical process monitoring is to define normal operating regions by applying statistical techniques to the original data samples obtained from the process systems. Typical examples of such techniques include principal component analysis (PCA), partial least squares (PLS), independent component analysis (ICA). Any data sample which does not lie in the normal operating region is then classified as a fault, and its root cause needs to be identified through fault diagnosis. Recently, deep learning and neural networks are also widely used for the statistical process monitoring, where both supervised and unsupervised learning algorithms are implemented. We are interested in developing efficient process monitoring frameworks by utilizing various machine learning techniques.
Associated members: We need you!