Characterization of powder materials for additive manufacturing (3D printing ability) using machine learning methods

Metal additive manufacturing (AM), also known as 3D Printing, is a technology that produces three-dimensional metallic parts layer by layer as opposed to conventional subtractive techniques. The qualities of 3D parts are significantly influenced by the characteristics of the feedstock materials, which depend on the manufacturing processes and also vary between batches. Powder pack density and flowability are often used to evaluate the 3D printability of the powder. The two metrics are currently determined by experiments and/or numerical simulations. However, the existed methods are often time/labor consuming due to the repeated sampling/testing processes and/or generate poor knowledge in the predictions of a new powder batch due to the sophisticated physics behind particle contact/interaction. The machine learning (ML) methods, especially deep learning (DL) method has already been widely used in object detection and segmentation. In particular, neural networks (NNs) are able to learn the relationship between the input features and output targets based on previous data. Compared with traditional methods (experiments and simulations), ML methods feature high objectivity and prediction accuracy and efficiency. Once the dataset between powder characteristics and processing is established, ML can realize in-situ monitoring of powder characteristics and real-time control the powder manufacturing processes. TOBECONTINUED

Faculty Supervisor:

Yu Zou

Student:

Partner:

AP&C

Discipline:

Engineering

Sector:

Manufacturing

University:

University of Toronto

Program: