Methodology for Variable Fidelity Multistage Optimization under Uncertainty

摘要

A new methodology for solving optimization under uncertainty problems with multi-objective function, variable-fidelity, mixed-variable characteristics is proposed. Quantifying uncertainty in the design, analysis, and optimization of high cost complex systems such as launch vehicles promises significant payoffs in understanding and reducing system development costs and risk. However, the characteristics of these complex systems during the early research and development stage pose several challenges to current optimization methodologies. These unique characteristics include the presence of uncertain parameters of both aleatory and epistemic types. Some of the latter vary quantitatively in time as the design iterations progress and higher fidelity tools are applied to the system design and its design space. A review of applicable optimization under uncertainty literature is described. The capabilities of previous methodologies are compared to the characteristics of optimization under uncertainty problems typical in the design of complex systems in a multistage acquisition environment. A set of extensions to previous optimization under uncertainty methods are proposed for a new optimization algorithm and a single-stage-to-orbit engineering design problem has been formulated to test the new method. The paper is accompanied by 27 briefing charts.