Abstract:
The use of high-fidelity physics-based simulation models has become increasingly important in aircraft design process, including models based on computational fluid dynamics, finite element structure analysis, and system simulation methods. It is now common to formulate a problem with large number of design variables with complex and nonlinear relations among them. The increase in the number of variables, coupled with high computational cost associated with those high fidelity models, has led to a challenging task of global optimization even on high performance computing facilities within the time constraint and of visualizing the results with the typical 2D and 3D visualization methods currently available. In this paper, the common visualization methods for low dimensional data are first reviewed, and then a couple of new visualization techniques are proposed and implemented in a modular and hybrid visualization tool, based on a combination of design of experiments, response surface models and some data analysis methods. The tool is implemented in MATLAB and is capable of systematically presenting the high-dimensional design space, multi-dimensional response functions and their relationships. The applications on test functions and airfoil aerodynamic optimization indicate that the proposed visualization method is useful for designers to analyze high-dimensional data set and to increase the efficiency and quality of high-dimensional optimization solution.