Hi there,

I have a variable distribution needs to be smoothed on unstructured grids. Is anybody aware of any good smoothing/filtering scheme?

Thanks.

Hi Tony,

This can be done by looping over all the cells or nodes in the grid (depending on your scheme) and accumulating a sum for the variable of interest at each cell/node. The value of the neighboring cell/node is added to the sum for the current cell. At the end, this sum for each cell/node is then divided by the number of neighbors for that cell/node.

1) Determine the number of adjacent cells/nodes and their cell/node number for each cell/node 2) Save the old values 3) Compute the sum of all adjacent cell values 4) The new value for the variable is (oldvalue + epsilon*sum)/(1 + epsilon*numberofneighbors)

A few passes on this procedure will smooth the variable. Epsilon is a smoothing factor, typically 0.5.

This approach is used for residual smoothing and is described further in AIAA papers by Neal Frink, early to mid-90's.