WHAT’S DIFFERENT ABOUT hp MESHES?
Error/Computation Cost Targeted Adaptive Strategy
- a priori error estimate driven strategy for computing “optimal” new mesh using non-linear programming techniques to solve :
- Find (hK,pK) such that
min Cost = SK=1, nel [1 + 2 g (pK -1) + d pK2]/hK2 - a
sub. to. ||e||K = C hKm pK-n < b
-
hK,pK : mesh parameters, a:target compute cost, b: max. error/element, e:error
g,d : empirical constants, m,n: convergence rates in a priori estimates
=> very high local changes in mesh (degrees of freedom can increase by order of magnitude from mesh to mesh)