Predictive modelling of surface roughness for double vibropolishing in trough system

• Main Authors: Alcaraz, Joselito Yam Tomacder, Mankar, A.V., Ahluwalia, Kunal, Mediratta, Rijul, Majumdar, Kausik Kumar, Yeo, Swee Hock
• Other Authors: School of Mechanical and Aerospace Engineering
• Format: Article
• Language: English
• Published: 2018
• Subjects: Numerical Modelling; Vibratory Finishing; DRNTU::Engineering::Mechanical engineering
• Online Access: https://hdl.handle.net/10356/89945; http://hdl.handle.net/10220/47158
• Institution: Nanyang Technological University

Vibratory finishing is a ubiquitous surface finishing process administered to components of various functionalities. Alongside the development of more complex finishing techniques such as drag finishing and abrasive flow machining, significant progress on numerical simulation has also been achieved, e.g. computational fluid dynamics, discrete element method. Yet, search into predictive roughness modelling has been insipid. In this study, multi-variable regression and artificial neural network modelling was done using experimental data obtained from subjecting rectangular test coupons to double vibropolishing in a vibratory trough. Two regression models, i.e. exponential and power, and several Multi-Layer Perceptron (MLP) architectures were trained using experimental data, and were subsequently evaluated for generalization ability. Model selection was done by comparing the mean-absolute percentage error and r-squared values from both training and testing datasets.