A neural network for evaluating diagnostic evidence

Author

Sheppard, John W. ; Simpson, William R.

Author_Institution

ARINC Res. Corp., Annapolis, MD, USA

fYear

1991

fDate

20-24 May 1991

Firstpage

717

Abstract

The authors present the results of developing a trained neural network that is embedded in a portable maintenance aid called POINTER (Portable Interactive Troubleshooter). This neural network is model-independent and may be used on a wide class of diagnostic problems. The authors review, the process by which they selected the network paradigm, gathered the training and testing data set (with a discussion on the uncertainty formulation), trained and tested the network, and incorporated the resulting network into POINTER. They also discuss the overall architecture for reasoning under uncertainty as implemented in POINTER

Keywords

automatic test equipment; knowledge based systems; learning systems; maintenance engineering; neural nets; Bayes probability; POINTER; Portable Interactive Troubleshooter; architecture; diagnostic evidence evaluation; neural network; portable maintenance aid; reasoning under uncertainty; uncertainty formulation; Built-in self-test; Electrohydraulics; Neural networks; Pattern recognition; Performance evaluation; Prototypes; System testing; Test equipment; Uncertainty; Weapons;

fLanguage

English

Publisher

ieee

Conference_Titel

Aerospace and Electronics Conference, 1991. NAECON 1991., Proceedings of the IEEE 1991 National

Conference_Location

Dayton, OH

Print_ISBN

0-7803-0085-8

Type

conf

DOI

10.1109/NAECON.1991.165831

Filename

165831