| Abstract | Machine learning can be a powerful tool for process development and part qualification. Currently, its wider adoption for AM remains a challenge absent the availability of existing process and material data for inputs. Generating these data at scale is a daunting prospect, particularly for SMEs in the AM value chain, which highlights the need for efficient methods to create process- specific data that are both time and cost effective. Polycontrols, a Canadian SME deeply involved with cold spray additive manufacturing (CSAM), is facing that challenge on a daily basis, hence the partnership with NRC to develop such a tool, and validate it in an industrial-scale environment. In this work, an AI framework is being developed to automate the analysis of multidimensional characterization data, from microstructure images to physical properties of CSAM metal specimens.
Based on a developed capability successfully applied for cast aluminum alloy, the automated extraction of complex microstructural data was also made possible by applying a deep learning algorithm on images specifically for deformed particle recognition in CSAM aluminum alloy. Predictive models were developed to fit microstructure-property relationships, which, in turn, will guide CSAM parameter optimization. Such acquisition – analysis – prediction pipeline, powered by machine learning, can be envisioned for other additive manufacturing processes.
Learning objectives:
Describe an automated extraction of the quantitative microstructural data by machine learning models
Define process-structure-property relationships for cold spray additive manufacturing aluminum alloy |
|---|
