Integrated process-structure-property modeling framework and methods for process design and performance prediction of additively manufactured material systems

W.K. Liu, Z. Gan, C. Yu, O.L. Kafka, K.K. Jones, Y. Lu
Northwestern University,
United States

Keywords: process-structure-property modeling, process design, performance prediction, additive manufacturing

Summary:

We present an integrated process–structure–property software system for process design and performance prediction of additively manufactured (AM) material systems. The key idea of the methodology and computational framework is that structural information, data that can be represented in space using a collection of volume-filling elements, sufficiently describes the software system and can thus intermediate models. In abstract form, the framework consists of various “modules” that are aggregated into “hubs;” there may be multiple hubs, each of which collects and passes a complete solution for one stage (e.g. process simulation) to the next stage (e.g. mechanical response prediction). One might think of our modules as discrete processing units capturing a unique facet of the system and hubs as data management facilities for database queries. To systematize the methods, we group modules into hubs for (1) material processing and (2) material response. We will demonstrate the process–structure–properties prediction framework for additive manufacturing that connects models for each stage and requires only basic material properties and processing conditions through two additive manufacturing blind benchmark tests. The first blind benchmark tests was conducted by National Institute of Standards and Technology (NIST) in AM-Bench challenge (2018) (https://www.nist.gov/ambench, https://doi.org/10.1007/s40192-019-00130-x) and the second one is the Air Force Research Laboratory (AFRL) in Additive Manufacturing (AM) Modeling Challenge Series (2020) (https://materials-data-facility.github.io/MID3AS-AM-Challenge/). We have demonstrated that the accuracy and efficiency are greatly improved by mechanistic machine and deep learning technologies. The developed AM digital twin would provide guideline to design new processing strategy and printable materials/alloys in a predictive manner.