Customizable anisotropic microlattices for additive manufacturing: Machine learning accelerated design, mechanical properties and structural-property relationships

• Published in: Additive manufacturing 2024-06, Vol.89, p.104248, Article 104248
• Main Authors: Li, Xinwei, Wang, Pan, Zhao, Miao, Su, Xinran, Tan, Yong Hao, Ding, Jun
• Format: Article
• Language: English
• Subjects: Additive manufacturing; Finite element modelling; Lightweight; Machine learning; Microlattice
• ISSN: 2214-8604; 2214-7810
• DOI: 10.1016/j.addma.2024.104248

For their optimal directional mechanical properties, anisotropic microlattices hold significant importance as advanced materials in a diverse range of applications. However, a bottleneck in their developments may be encountered when the limitations of current methods have been reached. The emergence of a new technological era calls for novelties with both the materials design and their methods. Herein, we introduce a novel concept comprising of machine learning and additive manufacturing for the design and fabrication of advanced customizable anisotropic microlattices. Design methodology consists of (i) pseudo-randomizations of initial polyhedron struts geometries, (ii) subsequent characterization of their loading responses through finite element modelling and (iii) accelerated investigations through convolutional neural networks (CNN) and optimization based on defined criteria. CNN has showed the capabilities of producing high validation accuracies with further input sets that surpass the performance of initial training data. From the pool of data, structures with a diverse range of mechanical properties, based on desired applications, can be obtained. A selected few, based on defined criteria, are experimentally studied through laser powder bed fusion using stainless steel for the build. Mechanical properties are found to be highly architecture strut geometry dependent and their structural-property relationships are also fully elucidated through both experiments and finite element modelling. Overall, through this work, we aim to present a new digital manufacturing concept based on using artificial intelligence to assist in the design and customizable of microlattices for additive manufacturing fabrication.