Smart Additive Manufacturing for Microstructure Tuning via Insitu and Exsitu Machine Learning

Metal additive manufacturing (AM) has unlocked several innovative opportunities to enhance and tailor product performance across a wide spectrum of industrial applications. Within this context, high-value products are increasingly defined by novel physical, mechanical, and geometrical attributes. Achieving groundbreaking material properties strongly relies on the control and consistency of microstructural characteristics. Consequently, microstructural analysis assumes a central importance in process qualification and certification procedures. Currently, the industrial practice for analysing microstructural image data primarily relies on human experts' assessments. This study introduces a novel approach suitable to automatically extract informative features from complex 2-D microstructural data and utilize them for robust microstructure classification via machine learning approach. It encompasses the extraction and utilization of both morphological grain properties and crystal orientation distribution features. The proposed methodology is demonstrated using electron backscattered diffraction (EBSD) measurements on additively manufactured parts. We also present a further level of innovation, aimed at anticipating microstructural assessment from ex-situ (post-process) measurements to in-situ (in-process) predictions based on in-line thermography. The aim consists of predicting the salient microstructural properties of the part while it is being manufactured. Such capability allows tuning the microstructure to meet advanced performance requirements, as well as the rapid identification of possible departures from target properties. Real case studies in metal powder bed fusion applications are presented and discussed.