Error Result Interpretation

The error results that are shown in the graphs—the numeric error values from the error plug-ins and the visual representations of the approximation error—are used to review the approximation prediction capabilities and to assess the usability of the approximation model as a surrogate for the components that it approximates.

If the approximation quality is deemed unacceptable for one or more responses, you can create a new approximation using the following methods:

  • Use a higher order RSM, or change to the RBF approximation model type. The design space may be more nonlinear than the selected approximation model type.

  • Increase the number of sample points. A highly nonlinear design space may require a larger number of points than the default or previously selected number of sampling points to approximate the space accurately.

  • Decrease the range of inputs. A highly nonlinear and high-dimensional space may be difficult to approximate over a large range (or may require an impractical number of sample points). Reducing the sample range can improve the ability to approximate the space with sufficient accuracy.