Prediction Assessment and Validation of Multiscale Models for Additively Manufactured Lattice Structures Under Uncertainty

Abstract

In the design of lattice structures fabricated by additive manufacturing, a multiscale modeling process is usually required to effectively account for fine scale uncertainties. The validation of the multiscale model predictions, on the other hand, is a challenging task. In this research, two prediction assessment approaches, namely the area validation metric and the Kolmogorov-Smirnov test, are presented in a systematic validation pyramid approach with u-pooling method to address this issue. The use of these two approaches are evaluated in terms of being an unbiased decision criterion for the prediction assessment and validation of the multiscale models. The fine scale material and geometry uncertainties are propagated onto homogenized properties using a stochastic upscaling method at each scale of interest. The homogenized model predictions are validated using the experimental data obtained for the lattice structure example fabricated by material extrusion process. The results indicate that the presented approach is capable of effectively validate the predictions of the multiscale models under uncertainty.