Prediction of surface rougness of additively manufactured and machined parts via machine learning

Date

2025-8-21

Author

Usama, Muhammad

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

770
views

0
downloads

Cite This

This study addresses a key research gap in predicting surface roughness (Ra) for Polylactic Acid (PLA) components produced via Fused Filament Fabrication (FFF) and hybrid AM–CNC machining at varying inclination angles using supervised regression-based machine learning. Few prior works considered inclination angle or CNC post-processing; this study employs a comprehensive design of 54 parts via Taguchi L54 DOE, covering 10°–80° inclinations. Ten Machine Learning (ML) algorithms were evaluated, with the Explainable Boosting Machine (EBM) achieving the highest predictive accuracy (R2 = 0.983 for AM, R2 = 0.616 for hybrid). For AM, the most influential factors were Inclination Angle, Layer Height, Nozzle Diameter, and Printing Temperature. Optimal surface quality was achieved with 0.10 mm layer height, 0.25 mm nozzle, and 210 °C extrusion temperature. In hybrid machining, AM parameters coupled with depth of cut and infill density further influenced Ra as values decreased by 85–91%; a 0.20 mm depth of cut yielded 1.8 μm averaged across inclinations. Most parts ranged between 1–3.5 μm. Verification tests showed EBM predicted AM Ra within ±2 μm (MAPE 6.62%) and hybrid machining within ±1 μm (MAPE 9.17%). A preliminary coolant-assisted machining study indicated 15–21\% Ra improvement by keeping cutting temperatures below PLA’s glass transition and enhancing chip evacuation. These results provide a predictive framework for achieving target surface finishes with reduced trial-and-error, lower costs, and improved surface integrity, supporting the industrial adoption of PLA-based AM and hybrid processes.

Subject Keywords

Additive Manufacturing, Surface Roughness, CNC Milling, Machine Learning

URI

https://hdl.handle.net/11511/115643

Collections

Graduate School of Natural and Applied Sciences, Thesis