EUROPEAN CONGRESS AND EXHIBITION ON ADVANCED MATERIALS AND PROCESSES - EUROMAT 2021

Understanding how processes in metal additive manufacturing (AM) influence material-related properties (or the performance of the final part) is non-trivial. Capturing the evolving microstructure of AM parts is crucial to reveal explanatory features and enables interpretation of material properties. So far, experts utilize this knowledge to adjust the process parameters to the requested demands of the final part. At IEHK, a previously developed multi-scale integrated computational materials engineering (ICME) approach showed its suitability to derive process-structure-properties-performance (PSPP) linkages for metal AM. A drawback of such simulation frameworks manifests in time- and computationally intensive calculations. In this context, data-driven approaches can serve as promising tool to accelerate the development of novel AM metals. In particular, a closer look on feature engineering and data post-processing by unsupervised dimensionality-reduction techniques can help to reduce computational resources to enable fast predictions. In consequence, trained models on lower-dimensional data act as effective surrogate models providing a possible short-cut in conventional ICME frameworks.