The Multiscale Spectral Stochastic Finite Element Method for Chloride Diffusion in Recycled Aggregate Concrete

In this study, the multiscale stochastic finite element method (MsSFEM) was developed based on a novel digital image kernel to make analysis for chloride diffusion in recycled aggregate concrete (RAC). It is significant to study the chloride diffusivity in RAC, because when RAC was applied in coastal areas, chloride-induced rebar corrosion became a common problem for concrete infrastructures. The MsSFEM was an efficient tool to examine the effect of microscopic randomness of RAC on the chloride diffusivity. Based on the proposed digital image kernel, the Karhunen-Loeve expansion and the polynomial chaos were used in the stochastic homogenization process. To investigate advantages and disadvantages of both generation and application of the proposed digital image kernel, it was compared with many other kernels. The comparisons were made between the method to develop the digital image kernel, which is called the pixel-matrix method, and other methods, and between the application of the kernel and various other kernels. It was shown that the proposed digital image kernel is superior to other kernels in many aspects.