| Lately, product quality control has become an important challenge facing industry, including chemical manufacturing. Improvement of product quality control can only be achieved through increased understanding of the causes for quality variations. This is usually achieved through a detailed analysis of the process variables and their relationships. This step is very time consuming because large volumes of data are generated during the operation of a process. In this work, two methods are presented for exploratory data analysis that can be used to scan the process data to suggest relationships among variables. The first method is based on a hierarchical combination of induction methods and regression analysis. The second method is developed for analyzing process data in conjunction with a prior, incomplete model. Results are presented to show the power and efficiency of the proposed algorithms.;Recent advances in artificial intelligence and process control hardware have made it possible to consider the design and deployment of intelligent control systems with the ability to learn on-line. However, the tools for learning, i.e. acquiring knowledge from routine data, are not yet fully developed and analyzed. The results of this thesis will be of interest to process engineers involved in the design and deployment of intelligent process control systems. |
