Support vector machine-based unified learning system for prediction of multiple responses in AWJM of borosilicate glass and SEM study Online publication date: Sat, 19-Mar-2016
by Ushasta Aich; Simul Banerjee; Asish Bandyopadhyay; Probal Kumar Das
International Journal of Mechatronics and Manufacturing Systems (IJMMS), Vol. 9, No. 1, 2016
Abstract: Modelling of responses in any manufacturing process is helpful for working in virtual world. As such, effective model development of stochastic processes working on heterogeneous materials is reasonably difficult. Hence, a robust unified learning system, multi-objective modelling with SVM, is proposed in this work to study the gross erosion behaviour of borosilicate glass in abrasive water jet machining. In this study, experiments are conducted on borosilicate glass with variation of the control parameters - water pressure, abrasive flow rate, traverse speed and standoff distance. Two process responses - material removal rate (MRR) and depth of cut (DOC) are trained through support vector machine (SVM)-based learning system for regression. An optimised single set of internal parameters of SVM, that would predict both MRR and DOC with their respective Lagrange multipliers, is estimated by minimising the training errors with the help of particle swarm optimisation (PSO) procedure. A modification of PSO is also proposed in this article. Further, scanning electron micrographs of cut wall are qualitatively examined to reveal the possible erosion behaviour of the amorphous material - borosilicate glass.
Existing subscribers:
Go to Inderscience Online Journals to access the Full Text of this article.
If you are not a subscriber and you just want to read the full contents of this article, buy online access here.Complimentary Subscribers, Editors or Members of the Editorial Board of the International Journal of Mechatronics and Manufacturing Systems (IJMMS):
Login with your Inderscience username and password:
Want to subscribe?
A subscription gives you complete access to all articles in the current issue, as well as to all articles in the previous three years (where applicable). See our Orders page to subscribe.
If you still need assistance, please email subs@inderscience.com