UNSPECIFIED (2002) Constraint modelling using object-based kriging. In: Engineering Design Conference 2002, JUL 09-11, 2002, KINGS COLL, LONDON, ENGLAND.

Full text not available from this repository.

Abstract

Constraint modelling, as opposed to constraint management, means the modelling of constraints using observational data, either physical or from a CAD system. Kriging is the method of function fitting using positive definite kernels, sometimes referred to as tile 'stochastic process' method, It has been used extensively by the authors and others to fit, emulators for fast optimisation under the title of DACE: the Design and Analysis of Computer Experiments. The paper combines the two methods by defining positive definite kernels on new spaces such as hyperspheres. It is able to model whole surfaces as opposed to single functions. The main example in the paper uses p.d. kernels defined on spheres to model constraint regions that are star-shaped with respect to the origin. Extensions are briefly considered where the objects' are collections of lines, planes, etc. The connection is discussed with other methods of computational geometry used for such objects such as support vector machines.

Item Type:	Conference Item (UNSPECIFIED)
Subjects:	T Technology > TS Manufactures T Technology > TK Electrical engineering. Electronics Nuclear engineering T Technology > TJ Mechanical engineering and machinery
Journal or Publication Title:	COMPUTER-BASED DESIGN
Publisher:	PROFESSIONAL ENGINEERING PUBLISHING LTD
ISBN:	1-86058-372-5
Editor:	Shahin, TMM
Date:	2002
Number of Pages:	8
Page Range:	pp. 589-596
Publication Status:	Published
Title of Event:	Engineering Design Conference 2002
Location of Event:	KINGS COLL, LONDON, ENGLAND
Date(s) of Event:	JUL 09-11, 2002
URI:	http://wrap.warwick.ac.uk/id/eprint/10295

Data sourced from Thomson Reuters' Web of Knowledge

Request changes to a record

Actions (login required)

View Item