The Effect of Matrix Property Smoothing on the Reliability of Fibre-Reinforced Composites

This paper presents the development and assessment of a smoothing technique for the random properties of the matrix phase of a fibre-reinforced composite as part of a multiscale reliability framework. Many sectors in the industry are using fibre-reinforced composite materials, utilising their high stiffness-to-mass density ratio. Still, many uncertainties occur in their properties because of their multiscale build-up nature. Hence, structural reliability analysis can produce efficient designs, but it requires an understanding of how all sources of uncertainty affect performance. Among other approaches, the authors developed a multiscale surrogate-based framework for reliability analysis, which uses large representative volume elements that can correlate and propagate the effect of several uncertainties, including a spatial variation of matrix properties. The framework uses a blur kernel to smooth matrix properties. In this study, the effect of using the blur kernel filter and other larger filters is examined in terms of their impact on the statistical properties of the matrix phase and the overall stiffness reliability of an analytical laminate example. The developed kernels proportionally reduce the standard deviation of the matrix property towards a lower limit as they increase in size. It is also shown that the stiffness reliability of the selected composite example (which includes the chosen parameters for laminate configuration, loading type, material, and other uncertainties in the system) is not affected by the use of the matrix smoothing technique.