A method for identifying stochastic information of a heterogeneous material utilizes physical loading measurements that are input into a global optimization process. The optimization process executes, in parallel, a force-driven non-linear finite element simulation and a displacement-driven finite element simulation of a constitutive model of the heterogeneous material. The constitutive models model the spatially varying random material properties (i.e. stochastic properties) using the Karhunen-Loeve expansion, thereby introducing the stochastic parameters, including spatial mean, spatial variance, and correlation length for example into the models. Stress and strain values for both the force-driven and displacement driven finite element analyzes are input into an objective function, whereupon the finite element simulations are updated after each iteration of the optimization process is performed until the objective function is minimized to a desired level. This results in the identification of optimized stochastic parameters associated with the heterogeneous material under investigation.