Parametrization of flow processes in porous media by multiobjective inverse modeling

A multiobjective optimization algorithm was developed and applied to parameterize biofluid flow processes in partially saturated porous media. The forward problem was formulated as a nonlinear partial differential equation and solved by an efficient Galerkin finite element method. The numerical simulator was validated with reference and analytical solutions. The inverse problem was formulated in a nonlinear optimization framework in which model parameters were estimated by minimizing a complex penalty function, representing the discrepancies between the observed and predicted attributes of the physical system. A physical model was designed, built, and operated to collect the experimental datasets needed for model calibration. The proposed strategy was then coupled with the fluid pressure head, fluid content, and fluid flux density time–space series to estimate model parameters. Several optimization scenarios were investigated and it was concluded that while single objective optimization approach has limited accuracy and results in reasonable outputs for one or some attributes of the physical system, the proposed multiobjective optimization shows excellent agreements with the experimental datasets for all state variables.