Strategy: Running a Global Sensitivity Study
Use the following strategies for defining a global sensitivity study and reviewing the study results:
• Each step of a global sensitivity study can take almost as much time as a standard run of the same model. Take this into account when you specify the number of intervals. A greater number of intervals will require longer run time.
• The key result you obtain from a global sensitivity study is a sensitivity graph that indicates which values of the design variables correspond to less mass, higher stress, higher frequency, or lower flux, and so forth. This information serves as a guide to help you improve the design.
• By looking at sensitivity graphs for measures you plan to use for the optimization goal and limits, you can determine which design variables are likely to have the most effect on those measures. You can also determine the portion of a design variables range that has the most effect on the measures.
As an option, you can specify that convergence be repeated. We recommend repeating convergence if the design variables cause massive distortion of the elements. However, the run will require more time.
Sometimes, by adjusting the starting and ending positions of the design variables, you can define the study such that the first step includes the most distorted elements. If the p-levels of the edges used for the worst case result in convergence, they will work for all others.