Date of Award


Degree Name

Master of Science


Computational Science


Vinod Kumar

Second Advisor

Son-Young Yi


Traditional models of poroelastic deformation in porous media assume relatively homogeneous material properties such that macroscopic constitutive relations lead to accurate results. Many realistic applications involve heterogeneous material properties whose oscillatory nature require multiscale methods to balance accuracy and efficiency in computation.

The current study develops a multiscale method for poroelastic deformation based on a fixed point iteration based operator splitting method and a heterogeneous multiscale method using finite volume and direct stiffness methods. To characterize the convergence

of the operator splitting method, we use a numerical root finding algorithm to determine a threshold surface in a non-dimensional parameter space separating convergent & divergent problems. We also use the method of manufactured solutions to verify the proposed

multiscale algorithm.

Results suggest that non-dimensional parameter values above the threshold surface ensure convergence, with increasing rate of convergence as the non-dimensional parameter approaches infinity. For a given spatial discretization stepsize, convergence can be ensured by choosing larger time stepsizes.

The proposed multiscale algorithm converges for the decoupled solid deformation PDE with analogous behaviors as observed in Chu et al. (2012). We observed divergence in our multiscale algorithm for the decoupled fluid equation in the heterogeneous case and attribute it to lack of monotonicity preservation induced by the reaction term in the microscale model.. An alternative cross-sectional flux estimator is proposed to improve convergence.




Received from ProQuest

File Size

121 pages

File Format


Rights Holder

Paul M. Delgado