A method of solution for a non-linear diffusion model and for computing the parameters in the model (Q1093585)

The diffusion of a drug through a skin-like membrane which tends to partially absorbing the drug, is modelled in the following way: Let a denote the thickness of the membrance, c(x,t), z(x,t) the concentration of diffusing and absorbed drug, resp., inside the membrance \((0<x<a)\), and \(c_ 1(t)\), \(c_ 2(t)\) the drug concentration in the donor cell \((x<0)\) and receiver cell \((x>a)\), resp. Then the process is described by the system of coupled nonlinear parabolic differential equations: \[ (1)\quad c_ t=(P(z,c)c_ x)_ x-z_ t,\quad z_ t=gc \] (0\(<x<a\), \(t>0)\), with boundary conditions \(c(0,t)=kc_ 1(t)\), \(c(a,t)=kc_ 2(t)\), \(\partial c_ 1/\partial t=DPc_ x(0,t),\partial c_ 2/\partial t=-DPc_ x(a,t)\) \((t>0)\) and initial conditions \(c_ 1(0)=C_ 0\), \(c_ 2(0)=0\), \(c(x,0)=0\) \((0<x<a)\). The diffusion coefficient is assumed to be given by \[ (2)\quad P(z,x)=P_ 0\exp (-b(z/(1+c))^ 2). \] The equations (1) are solved by an implicit finite difference method employing an iterative procedure at every time step. The model parameters \(P_ 0\), b, k, g, \(C_ 0\) are computed by means of an optimization procedure: the system (1), (2) is repeatedly solved for different sets of parameters, and the sum of the squares of the differences between model outputs and experimental data is minimized.