Loading...
Thumbnail Image
Publication

Acceleration of alloy design and manufacturing via machine learning and automated optimization

Liu, Sen
Citations
Altmetric:
Editor
Date
Date Issued
2021
Date Submitted
Keywords
Research Projects
Organizational Units
Journal Issue
Embargo Expires
2022-03-10
Abstract
Unveiling the composition – process – structure – property (CPSP) relations is challenging and a perpetual pursuit for alloy design and manufacturing processes. The applications of machine learning (ML) or artificial intelligence (AI) techniques to these research fields are emerging and expected to be a common practice in the future. The key behind this expectation is those unknown domains can be accurately estimated by learning from sufficient training examples. However, on the one hand, the experimental data are usually insufficient, highly sparse, and used as private resources, which poses a significant challenge in applying ML/AI approaches to real-world engineering problems. Alloy design or metal additive manufacturing (AM) are typical examples and often conduct limited amounts of high-cost experiments. The samples produced have high CPSP relations in terms of multiple length/time scales, multiple physics fields, and the “curse of dimensionality” in statistical learning. On the other hand, with the advancement of characterization instruments, a wealth of data is generated, such as microscopy and X-ray diffraction images that far outpace domain knowledge can interpret and analyze in real-time.Based on these challenges, the research of this thesis explores the state-of-the-art ML/AI techniques to assist materials manufacturing, property design, process optimization, and characterization. Specifically, the methods in the thesis automate the microscopy image/data analysis and make ML with limited experiments data to speed up scientific discovery and improve modeling performance. The results highlight the possibility or potential that a large fraction of manual work in the alloy design and additive manufacturing processes will be assisted with ML/AI-based automation systems. The methods developed will benefit a broad range of materials fields and enable high-throughput materials design and manufacturing.
Associated Publications
Rights
Copyright of the original work is retained by the author.
Embedded videos