-
Epstein ISE Research Seminar
Wed, May 04, 2011 @ 03:30 PM - 04:30 PM
Daniel J. Epstein Department of Industrial and Systems Engineering
Conferences, Lectures, & Seminars
Speaker: Dr. Enrique del Castillo, Distinguished Professor of Industrial Engineering and Professor of Statistics, Penn State University
Talk Title: "Statistical Analysis of Shape and Profile Data in Manufacturing and Engineering Design"
Abstract: In this presentation I will summarize work performed during the last 3 years in the field of Statistical Analysis and Optimization of processes that generate complicated data. In the first part of the talk I will describe Statistical Shape Analysis (SSA) techniques and their use in Manufacturing. SSA has been used mainly to model 2 and 3-dimensional shapes of biological interest in the natural sciences. In manufacturing applications, the data is instead a cloud of points or âlandmarksâ, the locations of the measured points typically acquired with a CMM. After giving a brief review of SSA techniques, I will discuss new methods for the analysis of experiments where the responses are the shapes of manufactured parts. In practice, the usual approach to determine how controllable factors affect the shape of a part is to estimate the âform errorâ of the part and conduct an ANOVA on these errors. This, however, neglects the geometrical features of the data. Instead, I will present new ANOVA tests on the shapes themselves and I will contrast them with previously proposed alternatives. In the second part of this talk I will discuss the analysis and optimization of processes where the response is instead a one-dimensional curve or âprofileâ, a type of data of considerable relevance in Engineering Design and in certain computer experiments. I will discuss Bayesian modeling methods and the subsequent optimization of the profile responses via Spatio-Temporal Gaussian processes. These methods are illustrated with various Engineering Design examples.
Location: Hughes Aircraft Electrical Engineering Center (EEB) - Room 248
Audiences: Everyone Is Invited
Contact: Georgia Lum