About this Abstract |
Meeting |
2020 TMS Annual Meeting & Exhibition
|
Symposium
|
Purveyors of Processing Science and ICME: A SMD Symposium to Honor the Many Contributions of Taylan Altan, Wei Tsu Wu, Soo-Ik Oh, and Lee Semiatin
|
Presentation Title |
Zoning Thermomechanical Process History Data Using Unsupervised Machine Learning |
Author(s) |
Sean P. Donegan, Dennis Dimiduk, Michael Groeber |
On-Site Speaker (Planned) |
Sean P. Donegan |
Abstract Scope |
Understanding and quantifying process-structure relationships in thermomechanical processing through modeling and simulation is a foundational aspect of ICME. Such modern process models are capable of producing data streams having high dimensionality both in space and time; and, developing workflows that relate such complex data to resulting measured microstructure remains a challenge. We present a procedure for reducing the overall complexity of process model data by zoning: grouping together regions of a component having similar process histories. This zoning procedure utilizes unsupervised techniques from machine learning to develop this lower dimensional representation without expert intervention. We demonstrate this zoning procedure in two application areas: i) relating process output from a DEFORMŽ pancake forging simulation of a Ti-6242Si cylinder to measured microstructure; and ii) comparing process histories computed from a fast-acting thermal model of powder bed fusion additive manufacturing across different component geometries. |
Proceedings Inclusion? |
Planned: Supplemental Proceedings volume |