A framework for finding the optimal laminate design rules using generative and explainable machine learning models

The vast design space of composite laminates poses significant challenges to the design process. Thus, establishing simple laminating rules to guide the design of layup sequences is of great importance. This paper presents a novel data-driven framework composed of two machine-learning models. The generative model is used to produce laminate designs that meet the specific design requirements, while their hidden quantitative relation is revealed by the explainable model. The effectiveness of this framework is verified through two classical composite design cases: isotropic and homogeneous laminates. The laminating rules proposed by the data-driven framework are proven to be reasonable and consistent with those in common laminate practices. This work demonstrates the great potential of artificial intelligence not only in guiding laminate design but also in providing new perspectives on discovering composite theory.