Anthony D. Rollett’s research while affiliated with Carnegie Mellon University and other places

What is this page?


This page lists works of an author who doesn't have a ResearchGate profile or hasn't added the works to their profile yet. It is automatically generated from public (personal) data to further our legitimate goal of comprehensive and accurate scientific recordkeeping. If you are this author and want this page removed, please let us know.

Publications (182)


Fig. 1 Proposed grain segmentation pipeline
Fig. 10 Quickshift kernel size selection effect on segmentation accuracy
Comparison with baseline methods
Segmentation accuracy varying with kernel size selection for EBSD image
The Bhattacharyya and Hellinger distance between each grain size distribution for the as-printed (AP), stress-relieved (SR), and long-term annealed (LA) samples
Rapid Grain Segmentation of Heat-treated and Annealed LPBF Haynes 282 Using an Unsupervised Learning-Based Computer Vision Approach
  • Article
  • Full-text available

January 2025

·

6 Reads

Integrating Materials and Manufacturing Innovation

·

Junwon Seo

·

Kevin Murphy

·

Anthony Rollett

Grain size distribution is a critical factor in determining materials’ physical and mechanical properties, including thermal conductivity, hardness, and creep behavior. Understanding the distribution of grain sizes is essential for advancing the comprehension of material properties and improving materials development and design. Traditional methods for determining grain size, such as electron backscatter diffraction (EBSD), are resource-intensive, underscoring the need for more efficient approaches to grain segmentation in standard micrographs, such as those obtained via SEM and optical imaging. This paper presents a streamlined, unsupervised computer vision pipeline that employs superpixel segmentation and region adjacency merging techniques to segment and measure grain geometry from micrographs efficiently. The pipeline is validated using two methods: hand-labeled SEM images of laser powder bed fusion (LPBF) fabricated Haynes 282 Ni-alloy and open-source EBSD data of IN100 from Dream3D. Both validation approaches achieved IoU and Dice scores greater than 0.9, while processing an image with a resolution of 1000 × 1000 pixels in under 40 s, demonstrating a fast and sufficiently accurate pipeline.

Download

Sub-millisecond keyhole pore detection in laser powder bed fusion using sound and light sensors and machine learning

November 2024

·

124 Reads

·

·

Haolin Liu

·

[...]

·

Laser powder bed fusion is a mainstream additive manufacturing technology widely used to manufacture complex parts in prominent sectors, including aerospace, biomedical, and automotive industries. However, during the printing process, the presence of an unstable vapor depression can lead to a type of defect called keyhole porosity, which is detrimental to the part quality. In this study, we developed an effective approach to locally detect the generation of keyhole pores during the printing process by leveraging machine learning and a suite of optical and acoustic sensors. Simultaneous synchrotron x-ray imaging allows the direct visualization of pore generation events inside the sample, offering high-fidelity ground truth. A neural network model adopting SqueezeNet architecture using single-sensor data was developed to evaluate the fidelity of each sensor for capturing keyhole pore generation events. Our comparative study shows that the near infrared images gave the highest prediction accuracy, followed by 100 kHz and 20 kHz microphones, and the photodiode sensitive to processing laser wavelength had the lowest accuracy. Using a single sensor, over 90% prediction accuracy can be achieved with a temporal resolution as short as 0.1 ms. A data fusion scheme was also developed with features extracted using SqueezeNet neural network architecture and classification using different machine learning algorithms. Our work demonstrates the correlation between the characteristic optical and acoustic emissions and the keyhole oscillation behavior, and thereby provides strong physics support for the machine learning approach.


FIGURE 1: EXAMPLE SCHEMATIC OF THE BUILD DOMAIN (Ω) COMPOSED OF THE PART (ΩP), TSS (ΩTSS), BSS (ΩBSS).
FIGURE 3: UNIT CELLS EMPLOYED FOR LATTICE SUPPORT STRUCTURE HEAT TRANSFER ANALYSIS WITH (A-C) SIZE S1 OF SIMPLE CUBIC (SC), FACE-CENTERED CUBIC (FC) AND TRANSITION CELL (TR) AND (D-F) SIZE S2 OF SIMPLE CUBIC (SC), BODY-CENTERED CUBIC (BC) AND SOLID (SOL).
FIGURE 6: 3D REPRESENTATION OF THE ADAPTER PIPE FOR A HIGH-TEMPERATURE HEAT EXCHANGER HEAT EX-CHANGER FOR THE (a) .STL FORMAT WITH PREDEFINED SUPPORT STRUCTURE DOMAIN (RED); (b) VOXEL REPRESEN-TATION WITH IDENTIFIED AND SORTED 1-MM (A SINGLE GREEN VOXEL) AND 2-MM (FOUR BLUE VOXELS) UNIT CELLS SUPPORT DOMAINS; (c) EXAMPLE OF OPTIMALLY-DIRECTED SUPPORT STRUCTURE DESIGN, OPEX.
FIGURE 7: (A) AVERAGE OBJECTIVE FUNCTION VAL-UES FOR THE HEAT TRANSFER RATE OF THE TRADI-TIONAL SA OPTIMIZER AND THE MSO METHOD FOR THE ADAPTER PIPE. SHADED REGIONS SHOW STANDARD DEVIATION WITH (B) AN EXAMPLE OPEX, WITH FINAL VALUES: OBJECTIVE = -3,742 W, VOLUME = 3,807 MM 3 , P-NORM = 0.08, UZ = 18.24 MM
A Multi-Sized Unit Cell Design Method to Design Lattice Support Structures for Complex Geometries in LPBF

Composed of individual unit cells strategically arranged to achieve a desired function, lattices are a promising solution for laser powder bed fusion (LPBF) support structure design. Despite their many advantages (e.g., multifunctionality and reduced material cost), prior work in lattice support structure (LSS) design primarily focused on parts with horizontal support domains (e.g., cantilever beams) that are not translatable to complex domains, thereby limiting their application. This work introduces a multi-sized unit cell design optimization (MSO) method to create LSS for parts with complex surfaces. The proposed method utilizes voxelization to identify and sort the locations of box-like unit cells of different sizes. It also allows for efficient design optimization through a modified simulated annealing-based optimization algorithm. The effectiveness and efficiency of the MSO method are demonstrated through the case study of an adapter pipe for a high-temperature heat exchanger. For this demonstration, LSS using multi-sized unit cells are designed to increase heat transfer rate while considering structural integrity and material cost constraints. When the optimally-directed results derived from the MSO method are compared to benchmark designs of equal-sized unit cells, it achieves an average heat transfer rate that is 16% higher while satisfying volume, stress, and deformation constraints.




Figure 2: System architecture with currently implemented models. All custom models implement our interfaces, outline color indicates which: Token Classification, Text Generation, or Image Processing. indicates components running in the same Docker container, and indicates models running in the cloud. "Materials IE" refers to materials-specific models, like ChemDataExtractor.
Figure 4: LLM Selector, as it appears in the File Upload view. Users specify an LLM to query, enter their API key, customize the prompt for an LLM, and repeat for any number of LLMs and prompts.
Figure 5: The annotations view. On the left, a screenshot showing the sidebar, allowing file and model selection, and the left pane, a visualization of the PDF with clickable regions highlighted. On the right, screenshots showing visualizations from the Table Transformer model with bounding boxes and parsed table (top), a HuggingFace transformer model with token-level tags (middle), and GPT-3.5 Turbo, with JSON output parsed into a table (bottom).
Collage: Decomposable Rapid Prototyping for Information Extraction on Scientific PDFs

October 2024

·

22 Reads

Recent years in NLP have seen the continued development of domain-specific information extraction tools for scientific documents, alongside the release of increasingly multimodal pretrained transformer models. While the opportunity for scientists outside of NLP to evaluate and apply such systems to their own domains has never been clearer, these models are difficult to compare: they accept different input formats, are often black-box and give little insight into processing failures, and rarely handle PDF documents, the most common format of scientific publication. In this work, we present Collage, a tool designed for rapid prototyping, visualization, and evaluation of different information extraction models on scientific PDFs. Collage allows the use and evaluation of any HuggingFace token classifier, several LLMs, and multiple other task-specific models out of the box, and provides extensible software interfaces to accelerate experimentation with new models. Further, we enable both developers and users of NLP-based tools to inspect, debug, and better understand modeling pipelines by providing granular views of intermediate states of processing. We demonstrate our system in the context of information extraction to assist with literature review in materials science.




Location-Dependent Phase Transformation Kinetics During Laser Wire Deposition Additive Manufacturing of Ti–6Al–4V

September 2024

·

68 Reads

Metallurgical and Materials Transactions A

This work models the microstructural evolution as a function of position in laser-wire deposited Ti–6Al–4V parts using a classical multi-component JMAK nucleation and growth model with additive isothermal time-steps. Model predictions are compared to experimental observations. The model can be used to interpret nucleation and growth kinetics of various characteristic features of the microstructure that are inaccessible by experiments. It explains the presence of “layer bands” at specific locations unrelated to the weld bead structure. The last re-heating (of multiple thermal cycles) of a solidified layer, and where it peaks, plays a key role in increasing the nucleation density at specific locations, resulting in full α\alpha α -colony microstructures constituting the “layer bands,” while the rest of the build is predominantly basketweave. Additionally, changes in the basketweave α\alpha α -lath thickness as a function of the distance from the substrate are studied and compared with an Arrhenius-type function, with results highlighting a more complex relationship than an Arrhenius-type function would suggest.


A Multi-Sized Unit Cell Method for the Design of LPBF Lattice Support Structures Concerning Complex Geometries

September 2024

·

53 Reads

Journal of Computing and Information Science in Engineering

Composed of individual unit cells strategically arranged to achieve a desired function, lattices are a promising solution for laser powder bed fusion (LPBF) support structure design in additive manufacturing. Despite their many advantages (e.g., multifunctionality and reduced material cost), prior work in lattice support structure design primarily focuses on horizontal support domains that are not translatable to support domains for complex geometries, thereby limiting their application. This work introduces a multi-sized unit cell design optimization (MSO) method to create lattice support structures (LSS) for parts with complex geometries. The proposed method utilizes voxelization to generate LSS using box-like unit cells of different sizes. It also allows for efficient, high-dimensional design optimization for the types and locations of user-specified unit cells through a modified simulated annealing-based optimization algorithm. The effectiveness and efficiency of the MSO method are demonstrated through the case study of an adapter pipe for a high-temperature heat exchanger. For this demonstration, LSS using multi-sized unit cells are designed to increase heat transfer rate while satisfying structural integrity and material cost constraints. The case study results indicate that the design of the LSS derived from the MSO method fulfills all constraints, including the design constraint of 50% material cost reduction, compared to the solid support structure. In contrast, the lattice support structure designs derived from equal-sized unit cell methods either cannot satisfy all design constraints or have a lower heat transfer rate than the design of the MSO method.


Citations (66)


... 8 The criterion for full melting across subsequent melt pools is defined as Equation 2 where computed values above the threshold can result in unfused powder. 8,9 Balling defects occur due to the hydrodynamic capillary instabilities that occur during high scan speeds and the resulting grooves along the sides of the melt pool result in formation of voids if not remelted. 10 and DistilBERT. ...

Reference:

AdditiveLLM: Large Language Models Predict Defects in Additive Manufacturing
Impact of melt pool geometry variability on lack-of-fusion porosity and fatigue life in Powder Bed Fusion - Laser Beam Ti-6Al-4V
  • Citing Article
  • October 2024

Additive Manufacturing

... An aspirational goal is to integrate monitoring data and modeling to predict local microstructure, its evolution, and, consequently, local and global properties, and performance (Figure 4b). Such a goal has been realized to different degrees of success in which researchers have integrated thermal models and physical processes such as evaporation and microstructural evolution to predict microstructure and have used sensor data to calibrate the models [115,213,289,290]. An example of a more fully integrated workflow is the ExaAM project that has produced multiple different models targeted towards metal AM processes, one of which is aimed towards predicting local microstructure; however, the problem that exists in all models, i.e., verification, remains reliant on the use of statistics, thus requiring additional computational resources. ...

The Effect of Interlayer Delay on the Heat Accumulation, Microstructures, and Properties in Laser Hot Wire Directed Energy Deposition of Ti-6Al-4V Single-Wall

Materials

... In L-PBF, the process window is governed by parameters such as laser power, scanning velocity, hatch spacing, layer thickness, and beam spot size, which influence defect formation like keyhole, LoF, and balling. Modeling approaches, such as defect structure process maps, normalized model-based diagrams, and data-driven models like Gaussian process regression, help predict optimal parameters and minimize defects, ensuring consistent fabrication of dense parts [9], [10]. Extensive research has focused on optimizing the processing window to minimize defects, streamline parameter development, and reduce time and cost but primarily for L-PBF. ...

Fatigue-based process window for laser beam powder bed fusion additive manufacturing
  • Citing Article
  • June 2024

International Journal of Fatigue

... Model-based material definitions can guide establishment of local volumes (SERVEs) that exhibit unique properties based on local processing paths/conditions. Efforts have been and continue to be conducted that provide physics-based and empirical, data-driven models that link local processing parameters with output structure, including build-induced defects (keyhole porosity, unmelt porosity, balling-based porosity, and surface connected porosity) and microstructure (grain size distributions, phase selection, retained metallurgical strain, residual stress, etc.) [29][30][31][32]. This is extremely useful in developing AM processing (build) parameters and controls holistically within a component, material, and process design framework using the associated models as part of MBMDs. ...

Model-Based Material and Process Definitions for Additive Manufactured Component Design and Qualification
  • Citing Article
  • May 2024

Integrating Materials and Manufacturing Innovation

... At high temperatures, Cr species evaporate and move through the cell. The resulting Cr vapor can react with oxygen to produce high valence Cr oxides, forming a barrier that slows the oxygen reduction reaction, which is critical for SOC efficiency [6]. The kinetics of Cr poisoning are determined by the rates at which Cr vaporizes, diffuses, and accumulates on the air electrode of a SOC. ...

Systematic and predictive trends to chromium poisoning in solid oxide fuel cell cathodes
  • Citing Article
  • May 2024

Journal of Power Sources

... Many experimental researches have strongly combined the LoF defects with insufficient energy input, which results in shallow and narrow molten pools that implement material bonding during processing [4,37]. Moreover, it is recommended to focus on the molten pool boundaries between adjacent tracks to simplify the numerical calculation, which has been widely adopted in considerable numerical research on LoF defects [38][39][40]. However, there is currently no alternative method for addressing layer thickness (h) and hatching space (t) in the context of LoF criterion using 3-D Rosenthal equation. ...

Inference of highly time-resolved melt pool visual characteristics and spatially-dependent lack-of-fusion defects in laser powder bed fusion using acoustic and thermal emission data
  • Citing Article
  • February 2024

Additive Manufacturing

... The authors mentioned the need for further work addressing the mechanical performance of the heat exchanger in terms of creep and fatigue. Several designs of AM heat exchangers have been proposed in recent years and shown to out-perform baseline designs significantly (Sabau et al. [14], Moon et al. [15], Ning et al. [16], Gerstler and Erno [17], Das et al. [18], Samad and Lai [19], Zhang et al. [20]). ...

Design and techno economic optimization of an additively manufactured compact heat exchanger for high temperature and high pressure applications
  • Citing Article
  • February 2024

Applied Thermal Engineering

... Other models for melt pools include smoothed particle hydrodynamics (SPH) [204], which can also be used to help with defect modeling [155]. FEM and similar modeling methods are particularly well suited for heat flow because of energy source-material interactions and are commonly used to simulate and study the complex thermal histories associated with AM processes [205][206][207][208][209]. ...

Understanding the Role of Geometry and Interlayer Cooling Time on Microstructure Variations in LPBF Ti6Al4V through Part-Scale Scan-Resolved Thermal Modeling
  • Citing Article
  • February 2024

Additive Manufacturing Letters

... (5) Figure 4: Infinite plate with elliptic hole under tension [19] Furthermore, the stress concentration factor Kt can be calculated from the following formula [20]: ...

Analytical verification of FFT-based micromechanical simulations near an elliptical crack tip by using composite voxels
  • Citing Article
  • January 2024

Mechanics of Advanced Materials and Structures

... Du Plessis [171], examined the effect of process parameters on the formation of defects in L-PBF Ti6Al4V. The results showed that while higher scanning speeds offered a safer processing window for avoiding pore formation, the possibility of keyhole formation increased with increasing laser power. ...

Deep learning approaches for instantaneous laser absorptance prediction in additive manufacturing

npj Computational Materials