Inverse Engineering: A Machine Learning Approach to Support Engineering Synthesis
Rao, R. Bharat
This item is only available for download by members of the University of Illinois community. Students, faculty, and staff at the U of I may log in with your NetID and password to view the item. If you are trying to access an Illinois-restricted dissertation or thesis, you can request a copy through your library's Inter-Library Loan office or purchase a copy directly from ProQuest.
Permalink
https://hdl.handle.net/2142/72004
Description
Title
Inverse Engineering: A Machine Learning Approach to Support Engineering Synthesis
Author(s)
Rao, R. Bharat
Issue Date
1993
Doctoral Committee Chair(s)
Lu, Stephen C-Y
Department of Study
Electrical Engineering
Discipline
Electrical Engineering
Degree Granting Institution
University of Illinois at Urbana-Champaign
Degree Name
Ph.D.
Degree Level
Dissertation
Keyword(s)
Engineering, Electronics and Electrical
Artificial Intelligence
Computer Science
Abstract
This research presents a knowledge processing methodology called inverse engineering, that uses machine learning techniques for early stage design in parameterized domains. This methodology functions as a model translator, changing the representation of analysis knowledge embedded in a unidirectional simulator, into a multidirectional model that supports design synthesis. This methodology requires addressing two issues.
The first is the task of learning models from data in specified representations. This thesis describes an empirical learning algorithm called KEDS, the Knowledge-based Equation Discovery System. The user selects a restricted hypothesis space bias in the form of a class of parameterized (polynomial) model families, and KEDS learns accurate models that are restricted to those forms. In addition to being a model-driven empirical discovery system, KEDS is also a conceptual clustering system that partitions the problem domain based upon the relationships that it discovers among the problem variables. The use of the minimum description length (MDL) principle as a preference bias for KEDS provides a foundation for learning the "best" models (i.e., those that minimize predictive error on unseen data). KEDS has been applied to model three real-world domains: a diesel engine combustion chamber, a CMOS circuit for an operational amplifier, and a turning process on a lathe.
The second issue is that of supporting early stage design. Current computer-aided methods for product and process design require the iterative use of computer-based analysis models in a generate-and-test fashion. While this process is essential to optimize performance during the final stages of design, it has a number of disadvantages during early design. By restricting the models families used by KEDS to forms that can provide synthesis support (hyperplanes), the user can learn a multidirectional model. The user can use this model to propagate constraints in the analysis as well as the synthesis direction. This avoids the time-consuming traditional procedure of iteratively using analysis models to support synthesis. Further this multidirectional model provides the user with great flexibility during early stage design, and the valuable ability to perform "What-if?" analysis. The inverse engineering methodology has been successfully applied to learn models to support early product design of combustion chambers for diesel engines, and to support process design for a turning machine.
Use this login method if you
don't
have an
@illinois.edu
email address.
(Oops, I do have one)
IDEALS migrated to a new platform on June 23, 2022. If you created
your account prior to this date, you will have to reset your password
using the forgot-password link below.
