On the use of machine learning and genetic algorithm to predict the region processed by laser peen forming
Citation Link: https://doi.org/10.15480/882.8764
Publikationstyp
Journal Article
Date Issued
2023-11-01
Sprache
English
TORE-DOI
Journal
Volume
38
Start Page
60
End Page
64
Citation
Manufacturing Letters 38: 60-64 (2023-11-01)
Publisher DOI
Scopus ID
Publisher
Elsevier
Laser peen forming uses laser-pulse-induced strains to deform sheets by adjusting laser parameters and peening patterns. Finding an optimal pattern in a vast space of practically infinite solutions is challenging. This study presents a workflow using a simplified model to predict deformation. A machine learning-based cellular automata neural network (CANN) and genetic algorithm (GA) were used for pattern prediction. Experiments showed high process uncertainty, justifying simplified modeling. The CANN predicted patterns reliably but lacked generalization due to insufficient deformation data for various process parameters. The GA required optimization efforts to reduce computation time but was successful at generalizing pattern prediction.
Subjects
Cellular automata neural network
genetic algorithm
laser peen forming
machine learning
DDC Class
620: Engineering
Publication version
publishedVersion
Loading...
Name
1-s2.0-S2213846323002158-main.pdf
Type
Main Article
Size
1.55 MB
Format
Adobe PDF