Sensitivy analysis and graph-based methods for black-box functions with on application to sheet metal forming.

摘要: The general field of the thesis is sensitivity analysis black-box functions. Sensitivity studies how variation output can be apportioned to input sources. It an important tool in construction, analysis, and optimization computer experiments.The total interaction index presented, which used for screening interactions. Several variance-based estimation methods are suggested. Their properties analyzed theoretically as well on simulations.A further chapter concerns models that take functions variables return a scalar value output. A very economical sequential approach not only discovers those functional whole but identifies relevant regions domain.As third concept, support functions, indices over distribution support, suggested.Finally, all three successfully applied sheet metal forming models.