- **Main CFD Forum**
(*http://www.cfd-online.com/Forums/main/*)

- - **discretization of variable parameters**
(*http://www.cfd-online.com/Forums/main/89757-discretization-variable-parameters.html*)

discretization of variable parametersDear all,
in a porous layer I have a certain liquid s(x) distribution. The relative permeability is function of it: k(x) = s(x)^3. I need to determine the derivative of k, dk/dx. Should I discretize k: dk/dx|i = (k(i+1) - k(i-1))/(2h) or reduce the derivative as follow: dk/dx|i = 3*s(i)^2 * ds/dx|i = 3*s(i)^2*((s(i+1) - s(i-1))/(2h)) ? They lead to different results, and can not find a valid reason to choose one of them. thanks and regards Mauri |

Dear friends:
The difference come from the discretisation error.for example, if we set s(x)=x; x|i=0.1; then dk/dx|i=(x+h)^3-(x-h)^3/(2*h); using the reduced form we have dk/dx|i=3*x^2*ds/dx=3*x^2; if we set h=0.1, the first one gives a value of 0.40 while the second one gives 0.030. if we reduce the h to 0.01, then the first one gives a avalue of 0.0301; |

All times are GMT -4. The time now is 07:34. |