Enhancing surface quality of metal parts manufactured via LPBF: ANN classifier and bayesian learning approach

• Published in: International journal on interactive design and manufacturing Vol. 18; no. 6; pp. 4093 - 4101
• Main Authors: Arunadevi, M., Veeresha, G., Kharche, Anil W., Suryawanshi, Vinayak P., Sollapur, Shrishail B., Mhatre, Mitali S., Kapadani, Kaustubh R., Nalawade, Dattatraya
• Format: Journal Article
• Language: English
• Published: Paris Springer Paris 01.08.2024; Springer Nature B.V
• Subjects: Additive manufacturing; Aeronautics; Algorithms; Artificial neural networks; Bayesian analysis; Business metrics; CAE) and Design; Composite materials; Computer-Aided Engineering (CAD; Cost analysis; Cost control; Efficiency; Electronics and Microelectronics; Engineering; Engineering Design; Friction stir welding; Heat exchangers; Industrial Design; Instrumentation; Lasers; Machine learning; Manufacturing; Mechanical Engineering; Mechanical properties; Metal fatigue; Original Article; Parameter sensitivity; Performance measurement; Performance prediction; Powder beds; Python; Residual stress; Surface properties; Surface roughness; Turbomachinery; Turbo machinery; LPBF; Metal additive manufacturing; Machine learning; Bayesian learning
• ISSN: 1955-2513; 1955-2505
• DOI: 10.1007/s12008-024-01942-8

One of the metal additive manufacturing techniques, Laser Powder Bed Fusion (LPBF), is utilised to fabricate several metal composites, including S30 and AlSi10Mg, which are extensively utilised in the automotive and aerospace sectors. The main objective of this manufacturing is to achieve high surface quality for the complex dimensional parts specially heat exchangers and turbomachinery components. In this research, the 30 datasets of S30 alloy cube samples are collected from literature for analysis. The supervised classification algorithm (Bayesian learning) is used for the analysis and surface roughness prediction in term of current, Line offset and scans speed. The manual calculations of Bayesian leaning are performed to obtain the probability prediction for the objective function. Then the same input and output parameters are trained and modelled by ANN classifier using sklearn library from python. The performance metrics for classifier such as sensitivity, specificity, precision and accuracy are calculated for Bayesian learning and compared with ANN classifier. ANN classifier Prediction of performance characteristic gave accurate results which plays very important role in LPBF method because of high experimentation cost.