The goal of this research award is to solve process control and optimization problems where the response of interest of a system is a profile, that is, values of a function of interest, as opposed to a single observation measured at each of some given experimental conditions. Control of this type of systems needs to be performed by adjusting the controllable factors in the presence of uncontrollable noise factors, achieving in this way a process performance that is insensitive, or robust, to noise factor variation. This class of Robust Parameter Design (RPD) problem for profile responses abounds both in manufacturing and in non-manufacturing. For example, in manufacturing, machining settings result in geometric profiles of parts that are measured at several positions over a plane or space, and a target geometry needs to be achieved by varying the machine tool conditions in the presence of uncontrollable sources of variability. The research will propose, test and implement new statistical techniques useful when responses are profiles, based on the Statistics sub-disciplines of Functional Data Analysis and Statistical Shape Analysis. Manufacturing laboratories at both Penn State and at Politecnico di Milano, Italy, will allow testing the methods developed in this project.
The main outcome of this research will be a new set of statistical optimization and control techniques aimed at solving on-line RPD problems for profile responses. To allow technology transfer, software that implements the methods developed will be written and distributed at the PI's lab web site. Research opportunities for a PhD student and for undergraduate students, incorporation of the research results in courses at Penn State, and dissemination via the technical literature and through the PI?s book on Process Optimization will take place. Collaboration with industrial researchers (Intel, GlaxoSmithKline) will provide practical expertise and valuable experiences for participating students.
