Higher Order Shape Functions Flashcards
Why use a complete quadratic
gives epsilon_x and epsilon_y to be functions of both x and y
- possible to use higher order + different ones for both rectangle and triangles
Lagrangian
- A perfect grid with 9 nodes
- Value of 1 at the specific node and zero at all others is a C0 requirement
Serendipity
-Complete quadratic but no internal node
-Shape function is quadratic along the edge of the node under question and goes linearly to zero at the other end for mid-side nodes
-For corner nodes there is bilinear variation but there is zero enforced at the adjacent nodes
Element mixing
can be bad if C0 requirements are violated @ the displacement of edges linking quadratic and linear elements
Error types
Modelling: from constructing the mathematical model of the physical problem incorrectly
Round-off: from machine precision so very small numbers shouldnt be added to very big ones
Discretisation: a finite number of DoFs will never complete the exact solution but tends to it with more elements and and more flexible response/accuracy
-geometric domain and material characteristics are correct
-continuity of shape functions
-accurate stiffness matrix and nodal F
-essential BC applied
Order of errors
Displacement: O(h to the p+1)
Stress/strain: O(h to the p+1-m)
-p = order of complete polynomial
-m = order of diff. relating disp to strains
-used to estimate exact solutions but not @ positions of concentrated loading
Ue - U_h/2 dived by Ue - U_h = (h/2) squared/h squared
p-refinement
-higher order elements of the same shape used
-imporved accuracy not guaanteed by swithing from triangular to rectangular since the shape function wont be subsumed
h-refinement
each element replaced by >=2 of the same type so that the shape functions are subsumed