Computational Mechanics of Knitted Textiles

Knitted textiles are hierarchically structured materials. Compared to other materials such as fiber-reinforced composites or metals, knitting provides much finer control over the manufacturing process and creates a variety of multi-scale structures using a range of input materials. Mechanical behavior of knitted textiles is difficult to be efficiently predicted by traditional computational methods due to its complicated, entangled internal structures. Flexural motion, interaction activities and hierarchical structure of or between sub-components contribute to a complex behavior with high nonlinearity and instability of knitted textiles. This dissertation aims to develop an integrated computational framework for knitted textiles. To comprehensively understand the mechanical behavior of knitted textiles, knitted textiles were firstly treated as freestanding structures. Specifically, predictions of mechanical behavior of knitted textiles were first obtained by employing direct numeric simulation (DNS) using 3D Finite Element Analysis (FEA). Given the geometrical details of the entangled yarns included in the 3D models used, the DNS approach is capable of investigating the influence of various design parameters at the yarn level, including loop architecture, material properties, as well as interfacial interactions. However, the most intractable thing for a free-standing structure with such complex internal geometry is that its behavior is dependent on the size of computational domain, namely size effect. The expensive computational cost, on the other hand, hinders the scaling up of DNS. To address this issue, DNS of the mechanical behavior of knitted textiles was for the first time conducted on High Performance Computing (HPC) using the explicit FEA method which was compared to implicit analysis. The results presented demonstrate satisfying accuracy and higher order efficiency with reduced memory requirements of the explicit method which allows for improved efficiency in simulations of larger computational domains, while also demonstrate that HPC could be a valuable resource for computational material design applied to advanced manufacturing. Furthermore, efforts to develop Reduced Order Models (ROM) for knitted textiles based on the available DNS results were also presented which were shown to provide an alternative approach to predict multi-scale behavior in addition to consist a tool that could be leveraged in future micro-structure optimization investigations. Moreover, a first order, two-scale homogenization scheme was modified and applied by considering the knitted textile models as a material point in a far field. The equivalent macro stress and consistent material stiffness are derived from the micro level where specific 3D knitted textile models are used. The macro filed with unknown material properties is eventually linked to the micro level by a user subroutine which can convey the equivalent macro stress and stiffness in a looped form of FE code. The multi-scale homogenization scheme is computationally economic and powerful in predicting the behaviors at both the macro and micro level for knitted textiles.