Numerical identification of parameters for a model of sedimentation processes
An bal Coronel, François James, Maurício Sepúlveda · Inverse Problems · 2003
In this paper we present the identification of parameters in the flux and diffusion functions for a quasilinear strongly degenerate parabolic equation which models the physical phenomenon of flocculated sedimentation. We formulate the identification problem as a minimization of a suitable cost function and we derive its formal gradient by means of an adjoint equation which is a backward linear degenerate parabolic equation with discontinuous coefficients. For the numerical approach, we start with the discrete Lagrangian formulation and assuming that the direct problem is discretized by the Engquist–Osher scheme obtain a discrete adjoint state associated with this scheme. The conjugate gradient method allows us to find numerical values of the physical parameters. It also allows us to identify the critical concentration level at which solid flocs begin to touch each other and determine the change of parabolic to hyperbolic behaviour in the model equation.