Modeling of the Thermo-mechanical behavior of Additively Manufactured Thermoplastic Composites by means of full-field Numerical Homogenization

Additive manufacturing offers many possibilities in terms of the part geometry and the internal features that are achievable without using molds. Design iterations can therefore be quickly realized and tested, and for small part counts the absence of tooling makes it economically advantageous. In a method such as Fused Filament Fabrication (FFF), the combination of thermoplastic polymers and fiber reinforcements improves the mechanical performance of the parts produced, but they remain inferior to those seen in their injected counterparts. Furthermore, the printing process imparts a unique set of morphological features that makes the material properties difficult to predict. Having only limited empirical data on the mechanical behavior of short fiber reinforced polymers and their process-structure-property relationship is a problem for modelling of parts fabricated through FFF. If designers cannot predict the elastic and viscoelastic properties, and how those vary with temperature, they are forced to use conservative estimates as a substitute, and they will not be able to reliably account for the anisotropy in the response, limiting the adoption of this technique. In this thesis, the link between process and structure is investigated by taking 3D micro-tomographic (μCT) imaging of real specimens, and processing those data with a specially-built software that extracts individual constituents (fibers, matrix, voids). Using that data, homogenization models are adapted and implemented to enable the prediction of effective elastic and viscoelastic properties, and the anisotropy therein, which is then compared to experimental measurements of 3D printed materials. The homogenization method presented is demonstrated to enable prediction of tensile modulus as a function of nozzle diameter and printing pattern with an accuracy within 5% compared to experimental values. This is a significant improvement compared to leading methods in the field, which make use of artificially generated microstructure descriptions. The full anisotropic thermo-viscoelastic properties are also modelled via a dual-scale homogenization procedure which uses constitutive relations for the pure polymer as fitted from experimental data, and allows the prediction of creep compliance with a 3% mean relative error in the longitudinal direction compared to experimental measurements, both at 21 and 120°C. In addition, several morphological features that had never been reported before are presented, which provide valuable insight as to why these materials have such unique and variable mechanical response. The methodology presented can be extended to other polymer/ fiber combinations, and as a basis for predictive models of other physical properties, such as tensile strength, fracture toughness, and other phenomena like thermal or electrical conductivity, and piezoelectricity.