Description of the linear elastic material parameters of short fiber composites using random fields. A comparison of Gaussian and non-Gaussian random fields

Modeling spatially random materials such as short fiber reinforced composites remains a persistent challenge. Existing approaches tend to either oversimplify the inherent heterogeneity and complex character of these materials or demand substantial computational resources and cost. However, as these materials gain increasing relevance across various industries, there is a growing need for efficient yet realistic modeling strategies. One such approach involves representing mechanical properties through random fields. So far the approach is limited to Gaussian random fields, which implies a normal distribution of the underlying material properties. Since this is usually not the case and e.g. negative valued material parameters are inadmissible, the method is extended to non-Gaussian random fields in this work. To assess the necessity of the presented extended approach the non-Gaussian random fields are compared with each other to derive advantages and limitations. It is shown that the numerical approach proposed here provides an effective framework for the realization of non-Gaussian random fields to model the elasticity tensor of short fiber-reinforced composites. The results demonstrate that non-Gaussian random fields in context of finite element simulations don’t enhance the representation accuracy of the input data significantly. Therefore, the use of Gaussian random fields is deemed sufficient in the context of finite numerical simulations, provided that the input data are based on a sufficiently large window size or that the negligible number of negative values is appropriately corrected afterward. random fields ,non-Gaussian ,Karhunen-Loève expansion ,Expansion optimal linear estimation method ,mcKL ,Nataf transformation ,short fibre reinforced thermoplastics