Additive manufacturing has significantly expanded the achievable geometric design space for components to be built from a wide set of materials. At the same time, it has come with complex process parameters that must be chosen carefully to ensure optimal build quality, often on a part-by-part basis. This has created the need for methods to take full advantage of the new design freedom in a manner that is compatible with the capabilities of additive processes, and a need for intelligently selecting the interrelated build parameters of the process. Meanwhile, mechanical engineers routinely face a range of design scenarios, each with its own level of complexity. Optimal engineering efforts in these scenarios would each suffer from considerable compromises if addressed by some single computational approach. Therefore, a series of design optimization approaches is required, with differences where necessary and similarities where possible. There is a corresponding need to computationally optimize the AM processes themselves. Collectively, these efforts are motivated by the goal of fully utilizing AM’s design freedom to achieve complex parts with the exact desired behavior, which ranges from lightweight strength to tunable secondary stable configurations. An ensemble of approaches to this end has been emerging in the literature. However, the following challenges persist, which the present work is intended to address, through the careful integration of machine learning with traditional design and manufacturing principles.·&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; High computational expense, even associated with optimization with linear materials.o&nbsp;&nbsp; We employ a generative adversarial network (GAN) to reduce iterative finite element analysis (FEA) requirements.o&nbsp;&nbsp; Dimensionality reduction, such as uniform manifold and projection (UMAP), enables the development recommendations for further efficiency improvements in future work.·&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Scarcity of considerations for strain rate effects.o&nbsp;&nbsp; We exploit Bayesian optimization’s abilities in non-convex, gradient-free conditions.·&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Limited ability to handle nonlinear materials—especially in multi-stable structures.o&nbsp;&nbsp; We propose objective and optimization formulations to ensure multi-stable optima while maintaining computational efficiency.·&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Uncertainty quantification in process-driven quality prediction.o&nbsp;&nbsp; We develop a Bayesian network-driven predictive framework of an AM process, including a first-order approximation of the process window.As illustrated in Fig. 1 (please see .PDF file for figure), addressing the above points requires work spanning the product development process, from initial physics-based design to manufacturing process setup.

Optimal Design and Processing for Additive Manufacturing Using Machine Learning

Paper Proceedings: https://asmedigitalcollection.asme.org/MSEC/MSEC2021/volume/85062

2021 International Manufacturing Science and Engineering Conference (MSEC2021)

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.