XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX''"> Verifying the Analysis
Different analysis codes provide different means to ensure quality in your analysis results, as described in the documentation for the code. However, some guidelines for what to look for in order to ensure quality are provided here.
• Error messages. Each analysis code provides its own set of runtime error messages and feedback after the analysis has completed. These must be monitored to flag serious error conditions. For example, within MSC Nastran the optional Analysis Manager module provides extensive feedback about error conditions. It produces a measure of the numerical accuracy of a finite element model, Epsilon, that you can check after a structural analysis. An Epsilon value of less than 10-9 is acceptable.
• Convergent results. Your analysis model should return consistent results after a reasonable amount of time. If your analysis does not reach consistent results after a number of iterations, a problem with the analysis setup or with some area of the analysis model exists.
• Large gradients of results data. If large changes in displacement or stress within a small region of the model occur, go back and create a finer mesh in those areas.
• Consistency with real-world results data. If actual testing with prototypes or complete products has been done, compare your analysis simulation results with real world results.
• Consistency with your expectations for the results. You will often expect a likely result from your analysis, and are looking for precise quantification of failure threshold information. If the results from your analysis run vary substantially from your expectations, an error in the model or analysis setup may have occurred.
How to Resolve Results Problems
If one or more of the items listed above indicates problems with the quality of your analysis results, you must go back and reevaluate your model building choices. Likely problem areas will differ depending on the type of model you are testing and the analysis code you are using. Refer to your analysis code documentation for further information. Following are a few causes of results errors:
• Missing elements.
• A stiff element next to a flexible one.
• Incorrectly modeled beam/plate, beam/solid, or plate/solid connections.
• Incorrectly modeled offset beams.
• A mesh that is not divided up finely enough into constituent elements in key areas where loads are applied.