Checking Convergence Quality...

The three things to review to judge the solution accuracy are the Summary Report and Un-averaged Fringe Plots for both SPA and MPA convergence algorithms and the Measure Convergence Plots for the MPA algorithm.



From the Run data form you may choose to look at the Summary report file (this will be a *.rpt file type in the design study folder). For MPA solutions check for messages about measure convergence. If the MPA solution does not converge there will be a message about that. This will let you know if your convergence criteria were met. You may not always achieve convergence. For both MPA and SPA solutions check the RMS Stress Error estimate. If this is greater than 10% there is probably someplace in the model with a significant error. If it is less than 10% you may still have significant errors or maybe not. This number is calculated based on the overall RMS stress of the model. It is a good indication of error if the stress has a very small gradient through the model. If there are one or a few hot spots with very high gradients, this error estimate is not as accurate.


The next thing to look at is an un-averaged stress fringe plot. These are created using the Results Definition data form. Select Von Mises stress for the Quantity and Fringe option under Display. Next, un-select the Average button under the display area. Set the Feature Angle to zero (this will make element edges visible). This will result in plots showing the raw element-by-element stress calculation. You will see that the stress fringes do not line up across element boundaries. The more continuous the stresses are across the mesh, the better the solution is.





For MPA solutions (not SPA) you need to look at measure convergence plots. These are crated using the Results Definition data form. Select Measures for Quantity and Graph for Display. These plots will show the measure values from pass-to-pass during the convergence process. The MPA algorithm only compares the last two passes for the percent change. This alone does not mean you have a good solution. You need to look at the shape of the curve. You would like to see an asymptotic convergence curve over several points.



<TSD Engineering HOME>

Copyright 2003, TSD Engineering