The primary goal of this project is to develop a search system for engineering
parts. It is expected that this system can be used in conjunction with
PLM systems or can be customized as a standalone feature in CAD repositories.
A significant amount of knowledge generated during design and manufacturing
is associated with the 3D model. However, text-based searching is not
robust primarily for the following reasons: (a) All models will not have
a well-defined attached context. (b) Keywords such as project names or
part names may be unknown to the user. (c) Context may be too narrow
or too broad to retrieve relevant models. (d) Context changes with time,
such as when designers or naming conventions change. Thus, a search system
which is capable of retrieving similar 3D models based on their shape
will also retrieve related knowledge.
Designers spend a significant amount of time searching for information
that is availablebut
cannot be located through traditional methods. Rectification of errors
that have been committed due to lack of information is a costly way to
learn. Nevertheless this has become a de facto process for new product
design. A significant amount of information generated during the lifecycle
of a product is associated with 3D models. Reuse of this information can
significantly shorten lead times and reduce costs during a product's lifecycle.
This project deals with an innovative approach to search for 3D models.
3D search offers an alternative means to retrieve design knowledge that
is intimately associated with 3D geometry. 3D models are represented by
a hierarchical skeletal graph representation rich with local information.
The skeletal graph representation preserves geometry and topology of the
model with good fidelity. Critical issues such as algorithms for converting
a model into a skeletal graph, search system efficiency, semantic gap reduction
and the subjectivity of the similarity definition are addressed.
Dr. Karthik Ramani,
Kuiyang Lou, Subramaniam
Jayanti, Yagnanarayanan Kalyanaraman,
Suyu Hou, Min Liu,
Dr. Sunil Prabhakar
Dr. Zygmunt Pizlo
Indiana 21st Century Research and Technology Fund
Purdue University Faculty Scholar Award