smart-materials-congress-2019-posters-abstracts

Smart Materials Congress 2019

Nano Research and Applications

ISSN: 2471-9838

Page 52

August 01-02, 2019

Dublin, Ireland

Smart Materials and

Structures

International Conference on

Nano Res Appl 2019, Volume 05

Machine learning-based optimal control for Y-shape tube

hydroforming processes

Van-Tuan Dang, Pascal Lafon

and

Carl Labergère

ICD/LASMIS/UTT Troyes, France

anufacturing of complicated components encountered

industrially associated with the control of the process are

similar to those relating to optimization. In fact, the control of a

hydroforming process requires a precise determination and

adjustment of the operating parameters so that the product

obtained satisfies precise criteria of shape and/or mechanical

properties and at the lowest possible cost. In the automotive

industry, the experimental trial and error process is replaced

by the numerical procedure, so the production time and costs

can be decreased drastically. As the result, the engineering

guidelines and finite element software would be used in the

optimal control for tube hydroforming processes. We present a

methodology of optimal control for Y-shape tube hydroforming

processes by using the machine learning technique. The main

study is to use the coupling of the optimization method and

finite element simulation at each time step to optimize the

load paths that consist of internal pressure and the axial feeds.

An optimization strategy is based on the Gaussian processes

combined with dimensionality reduction method to build the

approximation of optimization problem of the tube thickness

versus process parameters during processing. By this way, the

optimal command curves are constructed to obtain a better

quality component. The result obtained showed an efficiency

in improving thequalityof the final formof a tube.These results

achieved from numerical control can help the designers in

manufacturing a product formed. As a result, the proposed

approachhasanabilitytoreplacethetraditionalmethodsandto

useintubehydroformingprocesses.