After running an analysis of a model that includes finite elements, the results show that there there is an estimated maximum mesh error of 20%. What does this mean?
The mesh stress error indicator output gives an indication of the accuracy of a shell finite element analysis. The program calculates the maximum estimated error by taking the difference in the smoothed and raw stresses anywhere in the model. The program displays the error level across the model as contours, with the 100% level indicating the maximum estimated error level. If the estimated maximum error is 20%, for example, then the 50% contour would suggest an estimated error of 10% (the smoothed stresses differs by 10% from the raw stresses) at the point considered.
Large estimated errors often spell problems, but do not necessarily mean that the stress values all across a model are inaccurate or wrong. It does, however, helps you identify zones where refinement of the finite element mesh may improve the analysis accuracy. Keep the following factors in mind:
- Error distribution: Peaks in estimated errors only in certain parts of the model suggests that the basic mesh layout is sound. Refinement of the mesh may be needed in the zones where the estimated errors are large.
- Error location: Small stress differences in critical portions of the model may be significant. Likewise, if you are interested in the stresses in a certain part of the model, large stress difference in remote portions may have not significant effect in the part considered.
- Stress smoothing: By smoothing stress, you can improve accuracy in some cases (by balancing out errors). However, you should not use stress smoothing to try hide real problems in your model.
In most cases you can be improved the accuracy of the analysis by optimising the finite element mesh, e.g. using smaller elements is zones of stress concentration. When making this decision, you should consider both the local and global characteristics of your model.