Prediction of Meltpool Depth in Laser Powder Bed Fusion Using In-Process Sensor Data, Part-level Thermal Simulations, and Machine Learning

Authors

• Grant King, University of Nebraska-LincolnFollow

First Advisor

Jeffrey Shield

Second Advisor

Prahalada Rao

Third Advisor

Kevin D. Cole

Date of this Version

Fall 12-2022

Document Type

Thesis

Comments

A thesis Presented to the Faculty of The Graduate College at the University of Nebraska In Partial Fulfillment of Requirements For the Degree of Master of Science, Major: Mechanical Engineering & Applied Mechanics, Under the Supervision of Professor Jeffrey Shield & Professor Prahalada Rao. Lincoln, Nebraska: December 2022

Copyright © 2022 Grant Alan Maxwell King

Abstract

The goal of this thesis is the prevention of flaw formation in laser powder bed fusion additive manufacturing process. As a step towards this goal, the objective of this work is to predict meltpool depth as a function of in-process sensor data, part-level thermal simulations, and machine learning. As motivated in NASA's Marshall Space Flight Center specification 3716, prediction of meltpool depth is important because: (1) it can serve as a surrogate to estimate process status without the need for expensive post-process characterization, and (2) the meltpool depth provides an avenue for rapid qualification of microstructure evolution. To achieve the aforementioned objective, twenty-one Inconel 718 samples were designed and built with a variety of processing parameters ranging from a power of 200 W to 370 W and a scan speed of 670 mm/s to 1250 mm/s. These parts were characterized and the meltpool depth was measured through optical microscopy. A combination of part-level thermal simulations from a spectral graph theory method and in-process sensor data from infrared thermal camera and a meltpool imaging pyrometer were used as inputs to simple machine learning models to predict the meltpool depth. The meltpool depth was correctly predicted with an accuracy of F-Score 85.9%. This exploratory work provided an avenue for rapid prediction of microstructure evolution in metal additive manufacturing.

Advisors: Jeffrey Shield and Prahalada Rao