Affiliation
Meeting ID: 281 491 465 746 Passcode: YNdNny
Event Type:
MSE Grad Presentation
Date:
Talk Title:
​​​​​​​Investigation of Process Structure Property Relationships in Additively Manufactured IN718 Through Machine Learning
Location:
​​​​​​​In-person 4211 MRDC or Via Teams Meeting

Committee Members: 
Prof. Rick Neu, Advisor, ME/MSE
Prof. Josh Kacher, MSE
Prof. Chris Saldaña, ME
Prof. Aaron Stebner, MSE/ME
Dr. Xuan Zhang, Argonne National Lab

Investigation of Process Structure Property Relationships in Additively Manufactured IN718 Through Machine Learning

ABSTRACT: Given its good fatigue strength at elevated temperatures, IN718 is an ideal material for hot gas path turbine components for energy and aviation. In both applications, additive manufacturing (AM) can be implemented to improve current turbine designs by increasing efficiency through light-weighting and implementation of cooling paths unattainable by conventional manufacturing methods. With its added benefits, AM comes with its own set of obstacles when it comes to applications that undergo cyclic loading and where fatigue is a critical property. Porosity created by the additive manufacturing process can be detrimental to these fatigue critical applications if not controlled and monitored properly. This work is being conducted to understand and model the effect of AM introduced porosity and microstructural anomalies on the fatigue behavior of IN718 through a combination of ex-situ characterization and machine learning methods.

In this work, using laser powder bed fusion (LPBF), a set of 10 walls were constructed on the same build plate. Each build wall was constructed using a different combination of laser scanning speed and power to explore the process regime space near the default IN718 process parameters set by the EOS manufacturer. The range of process parameters in this work were chosen to explore the different AM process regimes of weld pool morphologies: conduction (shallow), transition, and keyhole (deep). The build walls, after removal from the build plate, underwent a direct age heat treatment, 720Image removed. Image removed. for 8 h with furnace cool to 620Image removed. Image removed. for 8 h with air cool, to maintain as built microstructure while increasing strength to increase sensitivity to porosity. These build walls were sectioned into dogbone fatigue specimens, polished, characterized by XCT and tested in high cycle fatigue at 1000Image removed. Image removed./ 538Image removed. Image removed. with a Image removed. Image removed.of 690 MPa, a stress ratio of 0.1, and a cycling frequency of 20 Hz. Separate sections of the build walls were mounted and polished then characterized via SEM based EBSD. The XCT and EBSD data generated through these ex-situ characterizations and their relationships to the fatigue properties of AM IN718 will be analyzed using machine learning methods as a part of the proposed research.