You are on page 1of 1

1.

One way can be to first calculate nodal values in each element using directly the shape
functions as proposed by David. But this is only possible if the numbers of nodes and Gauss
points of the element are equal. When using quadrilateral elements, this condition is fulfilled
if you use for example 4 node elements and a standard 4 points integration scheme. The
reference given by David proposes an extrapolation technique for second order hexahedra.
Once the nodal values of strains and stresses have been calculated in all elements, final nodal
values are obtained by calculating at each node the average of the values obtained in the
elements containing the node.

2. Another way to get nodal strains and stresses is to proceed as follows. 𝛔 being the stress
̃
field calculated by the finite element analysis, let’s define some continuous stress field 𝛔
𝐞
using the shape functions [𝐅 ] of the elements and nodal values of stresses gathered in a
vector {𝛔̃ }. Inside each element, one has :

̃ = [𝐅 𝐞 ]. {𝛔
𝛔 ̃𝒆} (1)

̃ 𝒆 } is the restriction of {𝛔
Where {𝛔 ̃ } to the nodes connected to element 𝑒.

{𝛔
̃ } is calculated in such a way that 𝛔
̃ is equal to 𝛔 in a weak sens. For each node 𝑛 and for
each component 𝜎̃𝑖𝑗 , one has :

∫𝛺 𝐹𝑛 (𝜎̃𝑖𝑗 − 𝜎𝑖𝑗 )𝑑𝑣 = 𝟎 (2)

The integral above can be rewritten :

∫𝛺 𝐹𝑛 (𝜎̃𝑖𝑗 − 𝜎𝑖𝑗 )𝑑𝑣 = ∑𝑒𝑙𝑒𝑚𝑒𝑛𝑡𝑠 ∫𝛺𝑒 𝐹𝑛𝑒 (𝜎̃𝑖𝑗 − 𝜎𝑖𝑗 )𝑑𝑣 = 𝟎 (3)

Using equation (1), equation (3) written for all the nodes of the model then leads to the
following system of equations :

(∑𝑒𝑙𝑒𝑚𝑒𝑛𝑡𝑠 ∫𝛺𝑒[𝐅 𝐞 ]𝐓 . [𝐅 𝐞 ]𝑑𝑣) . {𝛔


̃ 𝒆 } = ∑𝑒𝑙𝑒𝑚𝑒𝑛𝑡𝑠 ∫𝛺𝑒 [𝐅 𝐞 ]𝐓 . 𝜎𝑖𝑗 𝑑𝑣 (4)

All the integrals appearing in equation (4) are calculated using Gauss integration. The stress
values at Gauss points calculated by the finite element analysis are used in the right-hand
side. One may notice that the matrix appearing on the left-hand side of the above equation is
some mass matrix (with the density equal to 1). Using a lumped matrix as it is usual to do
with the mass matrix, equation (4) then reduces to N (number of nodes) scalar equations the
resolution of which is immediate.

3. A third way to calculate nodal values of stresses can be to use the diffuse approximation (or
moving least square) technique. The technique enables to calculate an approximation of the
stresses at each node using the values at the nearest Gauss points.

Details can be found, for example in the following article :

Fangtao Yang, Alain Rassineux, Carl Labergere, A hybrid Meshless-FEM field transfer
technique minimizing numerical diffusion and preserving extreme values: Application to
ductile crack simulation, Finite Element in Analysis and Design, 2018, 141:135-153.

You might also like