Open AccessAnalogy of support vector machine and linear regression models in surface roughness prediction
Author(s) -
Zvikomborero Hweju,
Khaled Abou-El-Hossein
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1710/1/012005
Subject(s) - support vector machine , surface roughness , artificial intelligence , gaussian , computer science , taguchi methods , surface finish , machine learning , linear regression , predictive modelling , diamond turning , linear prediction , gaussian function , algorithm , engineering , materials science , mechanical engineering , machining , physics , quantum mechanics , composite material
Accurate surface roughness prediction models are an indispensable part of an automated manufacturing system. Since most of the studies on surface roughness prediction have focused on large datasets, it is necessary to shift the focus towards the effectiveness of these models on small datasets. Thus, this study is an analogy of Support Vector Machine (SVM) and Linear Regression Models in surface roughness prediction of single point diamond turned RSA443, using a small dataset. Taguchi’s L9 orthogonal array-based experimental results from single point diamond turning of RSA443 have been integrated into the SVM model development. Primary cutting parameters (cutting speed, feed, and depth of cut) have been utilized, while a fine Gaussian SVM classifier is introduced for the analysis of the input features and surface roughness prediction using the MATLAB classification learner toolbox. The results indicate that the linear regression model is superior to the fine Gaussian SVM model. However, the high prediction accuracy of the fine Gaussian SVM model confirms the suitability of the model as an alternative surface roughness prediction tool. Hence both models can be reliably used for surface roughness prediction of RSA443 using small datasets.