AIAA 92-4791 The Suitability of Selected Multidisciplinary Design and Optimization Techniques to Conceptual Aerospace Vehicle Design

Aerospace vehicle conceptual design is dominated by interactions among various traditional engineering disciplines. Aerodynamics, propulsion, performance, weights, sizing, and others are usually highly coupled, and complete vehicle analysis requires an iterative process with efficient methods of communication among the disciplines. Progress to computerize the analysis process has been fast in recent years, producing analysis tools such as NASALangley's AVID and EASIE. Given a configuration, the capability exists to quickly analyze it in order to determine its overall characteristics and performance. However, the vehicle designer/ integrator still largely depends on intuition to make systems level changes to the configuration and components in order to improve or optimize the overall design. "What if" studies are typically performed by perturbing the design variables one at a time in an attempt to locate a better design. A complete reanalysis of the entire system is then required for each variable change. This method is a time consuming process that may or may not lead to a more desirable vehicle design. Several mathematically based design techniques have recently emerged that could help the system designer make necessary improvements. These new methods serve to bridge the gap between analysis and design. This paper attempts to give a brief overview of four such techniques, system decomposition, sensitivity analysis, Taguchi methods, and for comparison, classical optimization. References to examples of successful uses of each technique are provided. The goal of this paper is to assess the pros and cons of each technique and their applicability to aerospace vehicle conceptual design. Introduction Efforts to improve the analysis techniques used to rapidly analyze an aerospace vehicle concept (i.e. determine its performance, aerodynamic coefficients, TPS requirements, etc. for a given set of inputs) have produced great strides in efficient systems integration and data interchange capabilities. Recent experience has taught that in order to maintain a sufficient level of analysis detail, it is desirable to make use of the existing analysis tools of the disciplinary experts rather than writing a new, all-inone analysis code that may easily become outdated and unwieldy. Therefore, executive systems have been created to link the outputs of one computational analysis code to the inputs of others, to maintain a database of design variables, and to keep an audit trail of design changes (fig. 1).