A multiscale Bayesian method to quantify uncertainties in constitutive and microstructural parameters of 3D-printed composites

Abstract

3D-printed continuous carbon fiber reinforced composites (CCFRCs) are promising for various engineering applications due to high strength-to-weight ratios and design flexibility. However, the large variations in their mechanical properties pose a considerable challenge to their widespread applications. Here we develop a multiscale Bayesian method to quantify uncertainties in the constitutive parameters and microstructural parameters of 3D-printed CCFRCs. Based on the characterized microstructure of CCFRCs, a multiscale micromechanical model is developed to reveal the relationship between the properties of constituent materials, the microstructural parameters, and the macroscopic constitutive parameters. Furthermore, the joint posterior probability distribution of these parameters is formulated, and the Markov Chain Monte Carlo method (MCMC) is used to compute the posterior distributions of constitutive and microstructural parameters, enabling assessment of parameter uncertainty, correlation, and model calibration error. The inferred microstructural parameters are consistent with those measured by experiments. The posterior predictive distributions of the constitutive response are further computed to validate the probability model. Our method quantifies uncertainties in the constitutive parameters of 3D-printed CCFRCs and identifies their origins, which can optimize constituent material properties and microstructural parameters to achieve more robust composites.