Reconstruction of Spatially Varying Random Material Properties by Self-Optimizing Inverse Method

In this paper, a new methodology for reconstructing spatially varying random material properties is presented by combining stochastic finite element (SFE) models with Self-Optimizing Inverse Method (Self-OPTIM). The Self-OPTIM can identify model parameters based on partial boundary force and displacement data from experimental tests. Statistical information (i.e. spatial mean, variance, correlation length and Gaussian normal random variables) of spatially varying random fields (RFs) are parameterized by Karhunen-Loeve (KL) expansion method and integrated into SFE models. In addition, a new software framework is also presented that can simultaneously utilize any number of remote computers in a network domain for the Self-OPTIM simulation. This can result in a significant decrease of computational times required for the optimization task. Two important issues in the inverse reconstruction problem are addressed in this paper: (1) effects of the number of internal measurements and (2) non-uniform reaction forces along the boundary on the reconstruction accuracy. The proposed method is partially proven to offer new capabilities of reconstructing spatially inhomogeneous material properties and estimating their statistical parameters from incomplete experimental measurements.