The development of revolutionary aerospace vehicles presents significant challenges. Cross-disciplinary integration must be considered early in the design cycle, and to do so accurately and efficiently requires merging multiple levels of fidelity within an engineering discipline, as well as understanding the necessary degree of coupling between disciplines. The purpose of this work was to develop methods for representing, managing, and fusing information of various levels of fidelity within an engineering discipline and across disciplines for a wide range of analysis and design tools. An efficient, radial basis function-based alternative to multifidelity sequential Kriging optimization was developed. This method, referred to as multifidelity sequential radial basis optimization (MFSRBO) addresses multicriteria optimization involving more than a single type of model representing more than a single discipline, and takes into account both the location in design space and the fidelity of further data acquisition/infill. Preliminary numerical results indicate that methods based on radial basis functions can compete with Kriging methods in relation to efficiently and repeatedly identifying global optimal solutions. As a result, the radial basis functions are preferred, due to their reduced computational overhead and improved numerical stability. The application of MFSRBO is illustrated on a UAV wing design function integrating information from structural and fluid models. For a fixed computational budget, the numerically computed probability of finding feasible high-fidelity solutions exceeding a design target was shown to be an order of magnitude greater than when using comparable high-fidelity-only optimization.
In the context of virtually any major computational (CAD/FEM) optimization, the proposed methods offer substantially reduced computational cost to arrive at the optimum solution, compared with solvers that use only a single level of fidelity. As a result, engineers and researchers can solve more design problems in less time and achieve higher quality results at reduced costs. Applications of the technology include the design of advanced sensor platforms, UCAVs, and space launch vehicles. A diverse and broad range of potential design applications also exists beyond the aerospace field. These include applications to renewable energy, the chemical and metallurgical industries, pharmaceuticals, macroeconomics, and optimization of manufacturing processes.
References
Reisenthel, P. H., Allen, T. T., Lesieutre, D. J., and Lee, S. H., "Development of Multidisciplinary, Multifidelity Analysis, Integration and Optimization of Aerospace Vehicles," NEAR TR 657, Nielsen Engineering & Research, Santa Clara, CA, 2010.