Dissertations and Theses @ UNI

Availability

Thesis (UNI Access Only)

Abstract

Investment casting is an important aspect for modern manufacturing to make metal parts. The traditional version of that is where AM (Additively Manufactured) patterns for an investment are made from FDM (Fused Deposition Modeling), PLA (Polylactic Acid), PMMA (Polymethyl methacrylate) etc, and not a wax pattern from a die. In the case of the traditional version, an investment is inserted into a furnace so that the AM pattern is burnout and preheating of the shell takes place before molten metal is poured into the empty shell to get a desired metal shape. Within investment casting the specific process of AM pattern burnout is important but has had issues specifically with how long it takes and how to monitor it to make it more efficient.

In this research the issue of identifying and signifying when the finishing time of an AM pattern burnout process took place was investigated. This was carried out via the use of machine learning methods and other statistical methods for interval-censored data to narrow down the typical time for the AM pattern burnout process. The methods of C5.0 decision trees were the best at identifying the usefulness of the most important factors of ending temperature, rate of change for temperature, and rate of change for oxygen. Based on these findings, a real-time system was created in Node-RED. This realtime system displays gauges, graphs, and values for indicating whether the AM pattern burnout process has finished or not.

Year of Submission

2025

Degree Name

Master of Science

Department

Department of Applied Engineering and Technical Management

First Advisor

Jin Zhu

Date Original

5-2025

Object Description

1 PDF (ix, 91 pages)

Language

en

File Format

application/pdf

Download
Off-Campus Download

Comments?

Share

COinS