A general framework for experimental design, uncertainty quantification and sensitivity analysis of computer simulation models

Rigorous design of experiment (DOE) is essential to conduct validation, uncertainty quantification (UQ), and sensitivity analysis (SA) of computer simulation models. However, executing the process often involves knowledge of data management, statistical design, running simulation model, data analysis, and so on. It is a non-trivial task even for domain experts without solid computing backgrounds. Besides, the lack of standardization of data formats, configuration specifications, model invocation and execution mechanisms makes the process a harder undertaking. In this paper, we propose a comprehensive framework to support efficient experimental design, and UQ/SA in a domain and model independent manner. The data management and model execution issues are handled transparently from the users so that they can focus on the analysis itself. An application example is provided as an illustration of the concepts and basic use of this framework.