Quantifying Microstructure Heterogeneity for Qualification of Additively Manufactured Materials: Roles for Modeling and Data Science
Sponsored by: TMS Materials Processing and Manufacturing Division, TMS Structural Materials Division, TMS: Additive Manufacturing Committee, TMS: Phase Transformations Committee, TMS: Advanced Characterization, Testing, and Simulation Committee
Program Organizers: Sharniece Holland, Washington University in St. Louis; Eric Payton, University of Cincinnati; Edwin Schwalbach, Air Force Research Labroatory; Joy Gockel, Colorado School Of Mines; Ashley Paz y Puente, University of Cincinnati; Paul Wilson, The Boeing Company; Amit Verma, Lawrence Livermore National Laboratory; Sriram Vijayan, Michigan Technological University; Jake Benzing, National Institute of Standards and Technology
Thursday 8:30 AM
March 23, 2023
Room: 24B
Location: SDCC
Session Chair: Edwin Schwalbach, Air Force Research Laboratory; Amit Verma, Carnegie Mellon University
8:30 AM Invited
Towards Validation of Thermo-mechanical Finite Element Modeling of the Additive Manufacturing Solidification Process: William Musinski1; Paul Shade1; Edwin Schwalbach1; 1US Air Force Research Laboratory
Next generation material advancements in the defense and energy sectors will be obtained by integrating novel characterization tools, modeling tools, and machine learning to advance our understanding of the intimate, location-specific connection between processing, structure, and properties of these materials. As a building block towards this realization, this presentation describes the integration of synchrotron X-ray diffraction measurements, finite element simulations, and machine learning techniques for the development of residual stresses in additively-manufactured (AM) thin wall nickel superalloy plates built with a range of dimensions. Local thermal histories informed by a fast acting discrete source model are used as inputs for the thermo-mechanical finite element modeling of the AM solidification process. The residual stresses induced during the solidification process are then compared to energy-dispersive X-ray diffraction measurements of the as-built plate structures. Perspectives on integration of X-ray diffraction experiments, modeling, and machine learning, and the current technology gaps will be discussed.
8:55 AM
Effects of Laser Process Parameters on Denudation Zone Width in Laser Powder Bed Fusion Additive Manufacturing: Mehdi Amiri; Eric Payton1; 1Air Force Research Laboratory
Accurately modeling the physics of laser-powder interaction is needed for predicting the probabilities of performance-limiting defects in parts additively manufactured using laser powder bed fusion. We present a model for prediction of denudation width in laser powder bed fusion additive manufacturing process. The model relates the laser power and laser scan speed to the vapor jet speed and consequently to the inward entrainment gas velocity. The effects of laser power and laser speed on denudation width are investigated and compared with experimental data. The results show that for a given laser power the denudation width increases as the laser speed decreases.
9:15 AM
3D Computer Vision and Deep Learning for Porosity Analysis in Additive Manufacturing: Daniel Diaz1; Xingyang Li1; Yuheng Nie1; Elizabeth Holm1; Anthony Rollett1; 1Carnegie Mellon University
Additive manufacturing (AM) is a promising novel technology that is revolutionizing the way we manufacture products, but properties are limited by porosity produced during processing. In order to better understand the relationship between pore morphologies and properties, it is necessary to accurately identify the characteristic classes of pores observed. To this end we leverage the tools of 3D computer vision and convolutional neural networks to examine the pore morphologies present in datasets collected using X-ray computed tomography (CT). A transfer learning approach is utilized where 3D versions of EfficientNet are initialized with weights that have been trained on community datasets, and converted to a 3D format. Segmented CT image stacks are fed into this pretrained network, and the results are used to divide the pores into clusters that aid in identifying the various morphologies. This has the potential to become a valuable tool for automating the characterization of AM products.
9:35 AM
Quantitative Analysis of Low Concentration Elements at the Nanoscale in Additively Manufactured Alloys: Pritesh Parikh1; Darshan Jaware1; Jiangtao Zhu1; Karol Putyera2; Rajiv Soman2; 1Eurofins Nanolab Technologies; 2Eurofins EAG Laboratories
Microstructural changes due to processing conditions, types of processes used, and fatigue lead to defects, segregation of elements along defects and grain boundaries, precipitate and inclusion formation. Direct material analysis with 2D techniques such as TEM (transmission electron microscopy) provide structure and chemical mapping upto ~ 1 at%. APT (Atom Probe Tomography), adds to this by providing quantitative analysis of elemental composition with high spatial resolution (nm) and high sensitivity (~ 10 ppm). In this work, we investigate additive manufactured alloys, commercial and certified that were previously studied with bulk techniques for microstructural features. Our preliminary analysis indicates segregation of P up to 0.1 at%, which cannot be detected with other techniques. APT helps to augment the quantitative microstructure information in alloys by performing a full 3D analysis specifically of low concentration elements, during the research and development phases of the processes, and during useful lifecycle of the in-use parts.
9:55 AM Break
10:20 AM
Predicting Crystallographic Texture in Laser Powder Bed Fusion via a Machine Learning Approach: Gregory Wong1; Elizabeth Holm1; Anthony Rollett1; Gregory Rohrer1; 1Carnegie Mellon University
Metal parts produced using Laser Powder Bed Fusion (LPBF) have crystallographic texture that has been shown through experiments and computational modeling to depend on the parameters of the LPBF build and resulting solidification. This work shows a machine learning approach to predict the as-printed orientation distribution functions of LPBF parts through a method of reconstruction via linear fitting using commonly known texture components as fitting variables. Training and testing data consist of experimental EBSD data that has been fitted to standard texture components (e.g. cube, Goss, etc.) to produce a ground truth output. These data are used to simultaneously train and test a densely connected neural network and random forest model both with an output vector mirroring that of the ground truth data. This talk will discuss the input data fitting method, model structures, and compare the results of the two models relative to each other and traditional computational methods.
10:40 AM
Effects of Processing Conditions and Build Geometry on Microstructure Development in Laser Powder Bed Fusion and Wire Arc Additively Manufactured 316L: Charles Smith1; Olivia Denonno1; Matthew Schreiber1; Anthony Petrella1; Amy Clarke1; Jonah Klemm-Toole1; 1Colorado School of Mines
Microstructure development in fusion-based additive manufacturing (AM) starts with solidification. The vastly different microstructures observed across various AM processes of a given alloy have their origins in thermal gradients (G) and interface velocities (V) during solidification. These values depend strongly on parameters such as laser power, scanning velocity, and layer height. However, other factors can affect the values of G and V; one such factor is the surrounding environment around the molten pool. For example, powder material can have different thermal properties from a solidified material of the same composition and can affect G and V of the solidifying front. In addition, material solidified from previous passes can provide residual heat, altering G and V. In this presentation, we show how different surrounding environments can influence G and V and thus the melt pool geometry and microstructure of AM parts in laser powder bed and wire arc additive manufacturing.
11:00 AM
Additive Manufacturing Beyond the Gaussian Beam: Insights from Microstructure-based Modeling Studies: Daniel Moore1; Theron Rodgers2; Sergio Turteltaub3; Daniel Moser2; Heather Murdoch4; Fadi Abdeljawadf1; 1Clemson University; 2Sandia National Laboratories; 3Delft University of Technology; 4Army Research Laboratory
In laser-powder-bed-fusion additive manufacturing (LPBF AM), Gaussian laser beams are typically used as the energy input source. Recent experimental studies revealed that the use of laser beam shapes could expand the AM processing windows and result in desired microstructures. However, the role of laser beam shapes in the evolution of AM microstructures is not well understood. Leveraging a recently developed model coupling thermal transport and microstructural evolution during LPBF, we examine the impact of Ring and Bessel-like laser beams on the spatial distribution of temperature fields and the evolution of grain microstructures. Computational studies reveal that such beams expand processing windows to power and speed combinations not commonly used with Gaussian beams. To expound on the differences between the different beam types, we employ multiscale modeling techniques to describe the mechanical responses of representative microstructures. The results demonstrate enhanced control of process-structure linkages by considering beam shapes in AM.
11:20 AM
The Impact of Volumetric Energy Density on Mechanical Properties of Additively Manufactured 718 Ni Alloy: Benjamin Stegman1; Anyu Shang1; Luke Hoppenrath1; Anant Raj1; Hany Abdel-Khalik1; John Sutherland1; David Schick2; Victor Morgan2; Kirti Jackson2; Xinghang Zhang1; 1Purdue University; 2Proto Precision Additive LLC
Parameter normalization is essential in understanding how parameters affect the microstructure and resulting mechanical properties in laser powder bed fusion (LPBF). Volumetric energy density (VED), a leading measurement tool, estimates the amount of energy inputted per volume and compares the mechanical properties. In this work, alloy 718 tensile bars were printed in a design of experiments (DOE) testing VEDs from 30-93 J/mm3. ~46 J/mm3 was identified as the threshold value between the lack of fusion regime and the conduction regime. Tensile properties, such as yield strength and ultimate tensile strength, decreased with higher VEDs (up to ~200 MPa difference) , while other properties, such as uniform elongation, final elongation and work hardening exponent, increased with higher VEDs (up to ~0.1 and ~26% MPa difference). In-depth microstructural analysis, utilizing residual stress measurement techniques, chemical segregation analysis, electron backscatter diffraction (EBSD) analysis and fracture surface investigation, explains these trends.