Sensitivity analysis of a large: Scale transportation simulation using design of experiments and factor analysis

This paper describes how design of experiments and factor analysis were used to conduct sensitivity analysis on multivariate output from a large scale transportation simulation model. Specifically, this research focused on the sensitivity of airlift system performance to changes or errors in a list of transportation requirements. The general approach included perturbing a time-phased list of transportation requirements according to an experimental design and using a simulation model to estimate the airlift system performance response. We used factor analysis to reduce the dimensionality of the multivariate output data and to generate sensitivity plots, which proved to be valuable graphical tools for sensitivity analysis. Additionally, we identified how factor analysis can be used as a verification and validation tool for large stochastic simulation models.