Neural network based sensor fusion for on-line prediction of delamination and surface roughness in drilling AS4/PEEK composites

Ugo E. Enemuoh, A. Sheriff El-Gizawy, Chukwujekwu A. Okafor

Research output: Contribution to journalArticlepeer-review

Abstract

The inhomogeneous nature of composite materials has made mathematical modeling of their delamination (Da) and surface roughness (Ra) impossible. In this paper, an intelligent sensor fusion technique is used to estimate on-line Da and Ra during drilling of an advanced fiber reinforced composite beam (AS4/PEEK). A method for designing an artificial neural network based sensor fusion is presented. The fusion model design will include two drilling parameters, two process conditions, and two sensor signals. In order to minimize the effects of training parameters, resilient back propagation technique is used to train the networks. The final network predicts Da and Ra with errors ranging from 0.2% to 5%.

Original languageEnglish (US)
Pages (from-to)MS99-187-1 - MS99-187-6
JournalTechnical Paper - Society of Manufacturing Engineers. MS
Issue numberMS99-187
StatePublished - 1999

Fingerprint Dive into the research topics of 'Neural network based sensor fusion for on-line prediction of delamination and surface roughness in drilling AS4/PEEK composites'. Together they form a unique fingerprint.

Cite this