Webinars

Sensitivity analysis to understand the relative importance of the input parameters of an engineering simulation on its outputs is an important machine learning process. It is frequently featured in simulation guidance and best practices documents created by many manufacturing companies and various governmental agencies and professional organizations such as the FAA, FDA, AIAA, and ASME. This process can provide many benefits.

• Identification of insignificant parameters can reduce the dimensionality of the system being studied and hence its complexity, saving computational cost and allowing for easier to interpret results.
• Sensitivity analysis is essential to overall uncertainty quantification and management (UQM) efforts, e.g. with respect to which parameters does the engineer need to be most concerned about uncertainty?

Sensitivity analysis can be performed by directly sampling a simulation model; however, for simulations with even moderately long run times, non-linearities in the response, and/or many input parameters this approach can be computationally infeasible. Often the solution is to apply the same techniques, but using a predictive model trained using machine learning to act as a surrogate of the simulation.

This webinar will provide an introduction to both sampling and predictive modeling based sensitivity analysis. Examples will be used to highlight the benefits of a machine learning approach when dealing with complex real-world applications.