Transferring Process Mapping Knowledge across SS316L and IN718 in Laser Directed Energy Deposition Using Machine Learning

Date

2022

Authors

Menon, Nandana
Mondal, Sudeepta
Basak, Amrita

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

Laser-directed energy deposition additive manufacturing processes have several parameters that impact the melt pool properties, which in turn affect the microstructure of the part. Computational investigations are regularly implemented; however, these investigations must be repeated for each material of interest. In this paper, a transfer learning approach is proposed to address this challenge. Using an analytical model, input-output data pairs are generated for a nickel-based alloy, IN718, and an iron-based alloy, SS316L. A baseline neural network is trained for SS316L. The capability of transfer learning is analyzed with parametric retraining of the percentage of data used and the number of retrained layers of the SS316L baseline network on the IN718 data. With just 10% data and one hidden layer retrained, accuracies above 90% are observed. The results show that the acquired printability knowledge can be transferred across material systems without requiring a significant amount of data for a new material system.

Description

LCSH Subject Headings

Citation