In many features under Results, such as surface plots and line graphs, there is a section named
Quality, through which details of how the results are computed can be specified. In some situations, you need to modify these settings to obtain optimal accuracy or visually attractive presentations.
First, it can be noticed that stresses, strains, and any other variables that are based on derivatives of the shape functions are not continuous over element borders. The Smoothing control is used for improving plots by applying an averaging between neighboring elements. The default when you add a result feature is to use smoothing, and this will in general improve the results. It does however also hide any jumps in the solution, which could indicate a too low resolution in the mesh. As soon as you are investigating the quality of the solution, it is a recommended to avoid smoothing. In the default stress plots that are generated by the structural mechanics interfaces, an intermediate path is taken: the
Smoothing threshold is set to
Manual, with the
Threshold value set to 0.2. This means that as long as the values from adjacent elements do not differ by more than 20%, smoothing is applied. If, however, there are significant jumps in the solution, they will be clearly visible.
The Resolution setting determines at how many points in the element the result quantity is evaluated. Essentially, a local finer mesh is used within the element for the visualization. Using a high resolution can be problematic and lead to local overshoots and artificial ‘waviness’ of the solution. For smooth expressions, like a stress in the absence of inelastic strains or a displacement, this is not a problem. If, however, the function has sudden variations within the element, a high resolution will give results having artificial variations.
Using a low resolution will remove the sudden variations, but may still not be an optimal choice. If you set the quality to No refinement, the expression is evaluated only at the corners of the element. This is, for example, rather far from the Gauss points where many types of inelastic strains are stored. Also, a thermal strain with a quadratic distribution will be rather far from its best linear approximation at the element corners.