Events Calendar

06 Dec
Swanson School of Engineering
Event Type

Defenses

Topic

Research

Target Audience

Faculty, Graduate Students, Postdocs

University Unit
Department of Mechanical Engineering and Materials Science
Subscribe
Google Calendar iCal Outlook

PhD Dissertation Proposal Presentation, Seth T. Strayer

This is a past event.

Title:

"Efficient and Robust Thermal Process Simulation for Laser Powder Bed Fusion Additive Manufacturing Using Data-Driven and Probabilistic Heat Source Models

 

 

ABSTRACT:

 

Laser powder bed fusion (L-PBF) is notarized for manufacturing geometrically complex metallic components not feasible via traditional subtractive techniques. However, uncertainties associated with part performance currently hinder the qualification and certification requirements regarding their use in safety-critical applications. These uncertainties depend on the formation of defects and unique microstructure and properties that emanate from local microscopic thermal field variations induced by the complex layer-by-layer deposition process. Thus, predicting the thermal history and melt pool geometries throughout a part is critical to understanding expected part performance, quantifying its variability, and promoting L-PBF’s efficacy.

        Physics-based numerical simulation offers a cost-effective and detail-oriented method for predicting the thermal history of an L-PBF process. However, obtaining high-fidelity simulation results at the part scale is currently unfeasible due to the intense multiscale multiphysics nature of the process. Lower fidelity solvers, such as those based on the finite element (FE) method, can help address this computational limitation by abstracting the relevant physics into an effective heat source model but do not readily address the substantial thermal field variation observed through experiments.  

        This research aims to address these challenges by integrating data-driven and probabilistic heat source models with existing thermal process simulation techniques. First, a dimensionality reduction, deep learning-based heat source model is incorporated with a matrix-free, graphical processing unit (GPU) based FE model to accelerate high-fidelity thermal process simulation of L-PBF. This methodology is extended to practical scenarios by interfacing adaptive remeshing with a parallel simulation control algorithm that can account for the transient changes in thermal fields throughout the deposition process. Next, a novel probabilistic FE heat source model that samples from a time-correlated distribution of calibrated heat source parameters is proposed to help statistically mimic experimentally observed variations in melt pool size. This methodology is then extended to the part scale by incorporating different scanning orientations and preheat temperatures into the stochastic model. The development of these methods offers practical solutions addressing two significant challenges associated with existing thermal process simulation techniques. Consequently, this research is expected to provide efficient and robust thermal process prediction and ultimately help promote further adoption of L-PBF for next-generation manufacturing applications.

 

 

Join Zoom Meeting:

Link: https://pitt.zoom.us/j/2258283738 

Passcode: 374188

Meeting ID: 225 828 3738

Dial-In Information

  

Tuesday, December 6 at 9:00 a.m.

Benedum Hall, Benedum Hall, Room 611
3700 O'Hara Street, Pittsburgh, PA 15261

PhD Dissertation Proposal Presentation, Seth T. Strayer

Title:

"Efficient and Robust Thermal Process Simulation for Laser Powder Bed Fusion Additive Manufacturing Using Data-Driven and Probabilistic Heat Source Models

 

 

ABSTRACT:

 

Laser powder bed fusion (L-PBF) is notarized for manufacturing geometrically complex metallic components not feasible via traditional subtractive techniques. However, uncertainties associated with part performance currently hinder the qualification and certification requirements regarding their use in safety-critical applications. These uncertainties depend on the formation of defects and unique microstructure and properties that emanate from local microscopic thermal field variations induced by the complex layer-by-layer deposition process. Thus, predicting the thermal history and melt pool geometries throughout a part is critical to understanding expected part performance, quantifying its variability, and promoting L-PBF’s efficacy.

        Physics-based numerical simulation offers a cost-effective and detail-oriented method for predicting the thermal history of an L-PBF process. However, obtaining high-fidelity simulation results at the part scale is currently unfeasible due to the intense multiscale multiphysics nature of the process. Lower fidelity solvers, such as those based on the finite element (FE) method, can help address this computational limitation by abstracting the relevant physics into an effective heat source model but do not readily address the substantial thermal field variation observed through experiments.  

        This research aims to address these challenges by integrating data-driven and probabilistic heat source models with existing thermal process simulation techniques. First, a dimensionality reduction, deep learning-based heat source model is incorporated with a matrix-free, graphical processing unit (GPU) based FE model to accelerate high-fidelity thermal process simulation of L-PBF. This methodology is extended to practical scenarios by interfacing adaptive remeshing with a parallel simulation control algorithm that can account for the transient changes in thermal fields throughout the deposition process. Next, a novel probabilistic FE heat source model that samples from a time-correlated distribution of calibrated heat source parameters is proposed to help statistically mimic experimentally observed variations in melt pool size. This methodology is then extended to the part scale by incorporating different scanning orientations and preheat temperatures into the stochastic model. The development of these methods offers practical solutions addressing two significant challenges associated with existing thermal process simulation techniques. Consequently, this research is expected to provide efficient and robust thermal process prediction and ultimately help promote further adoption of L-PBF for next-generation manufacturing applications.

 

 

Join Zoom Meeting:

Link: https://pitt.zoom.us/j/2258283738 

Passcode: 374188

Meeting ID: 225 828 3738

Dial-In Information

  

Tuesday, December 6 at 9:00 a.m.

Benedum Hall, Benedum Hall, Room 611
3700 O'Hara Street, Pittsburgh, PA 15261

Event Type

Defenses

Topic

Research

Target Audience

Faculty, Graduate Students, Postdocs

Powered by the Localist Community Events Calendar ©