Apportion the uncertainty in the output to different sources of uncertainty in the input.
General idea of sensitivity analysis
State-of-the-art computational models often need to describe very complex systems, and hence also include a large number of parameters. One of the attractive features of physics-based models is that in those models (most) parameters have a clear physical meaning. Nevertheless, the determination of these parameters is often very elaborate and costly. Hence, it is essential to identify the most important parameters (worth the effort) for a particular problem at hand. In order to distinguish parameters which have a significant influence on a specific model output from non-influential parameters, sensitivity analysis is the method of choice.
Goal
The goal of sensitivity analysis is threefold:
- Identify the most influential parameters on which experimental studies should focus.
- Identify parameters with little to no effect, which can thus be set to fixed values within their range.
- Identify and quantify the interaction between parameters.
Meta-model based Sobol analysis
QUEENS allows to use the Sobol method as a global variance-based method combined with a Gaussian process used as a metamodel. The Sobol method requires a large number of model evaluations, which is prohibitive for computationally expensive models. Employing Gaussian processes as metamodels for the underlying full model allows to solve this problem. However, metamodelling … Metamodelling introduces further uncertainty, which QUEENS also quantifies in this case.
Reprinted from (Wirthl et al., 2023), licensed under CC BY 4.0.
Features in QUEENS
- Elementary effects (Morris method)
- Sobol method
- Sobol method based on a Gaussian process used as a metamodel