Methods for Rapid Pore Classification in Metal Additive Manufacturing

Robert Snell*, Sam Tammas-Williams, Lova Chechik, Alistair Lyle, Everth Hernández-Nava, Charlotte Boig, George Panoutsos, Iain Todd

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

Abstract

The additive manufacturing of metals requires optimisation to find the melting conditions that give the desired material properties. A key aspect of the optimisation is minimising the porosity that forms during the melting process. A corresponding analysis of pores of different types (e.g. lack of fusion or keyholes) is therefore desirable. Knowing that pores form under different thermal conditions allows greater insight into the optimisation process. In this work, two pore classification methods were trialled: unsupervised machine learning and defined limits. These methods were applied to 3D pore data from X-ray computed tomography and 2D pore data from micrographs. Data were collected from multiple alloys (Ti-6Al-4V, Inconel 718, Ti-5553 and Haynes 282). Machine learning was found to be the most useful for 3D pore data and defined limits for the 2D pore data; the latter worked by optimising the limits using energy densities.

Original languageEnglish
Pages (from-to)101-109
Number of pages9
JournalJOM
Volume72
Issue number1
Early online date4 Sep 2019
DOIs
Publication statusPublished - 1 Jan 2020

Fingerprint

Dive into the research topics of 'Methods for Rapid Pore Classification in Metal Additive Manufacturing'. Together they form a unique fingerprint.

Cite this