Enhancing Aircraft Conceptual Design using Multidisciplinary Optimization

Research into the improvement of the Aircraft ConceptualDesign process by the application of MultidisciplinaryOptimization (MDO) is presented. Aircraft conceptual designanalysis codes were incorporated into a variety of optimizationmethods including Orthogonal Steepest Descent (full-factorialstepping search), Monte Carlo, a mutation-based EvolutionaryAlgorithm, and three variants of the Genetic Algorithm withnumerous options. These were compared in the optimization offour notional aircraft concepts, namely an advanced multiroleexport fighter, a commercial airliner, a flying-wing UAV, and ageneral aviation twin of novel asymmetric configuration. Tobetter stress the methods, the commercial airliner design wasdeliberately modified for certain case runs to reflect a verypoor initial choice of design parameters including wingloading, sweep, and aspect ratio. MDO methods were evaluated in terms of their ability to findthe optimal aircraft, as well as total execution time,convergence history, tendencies to get caught in a localoptimum, sensitivity to the actual problem posed, and overallease of programming and operation. In all, more than a millionparametric variations of these aircraft designs were definedand analyzed in the course of this research. Following this assessment of the optimization methods, theywere used to study the issue of how the computer optimizationroutine modifies the aircraft geometric inputs to the analysismodules as the design is parametrically changed. Since thiswill ultimately drive the final result obtained, this subjectdeserves serious attention. To investigate this subject,procedures for automated redesign which are suitable foraircraft conceptual design MDO were postulated, programmed, andevaluated as to their impact on optimization results for thesample aircraft and on the realism of the computer-defined"optimum" aircraft. (These are sometimes called vehicle scalinglaws, but should not be confused with aircraft sizing, alsocalled scaling in some circles.) This study produced several key results with application toboth Aircraft Conceptual Design and MultidisciplinaryOptimization, namely: MDO techniques truly can improve the weight and cost ofan aircraft design concept in the conceptual design phase.This is accomplished by a relatively small "tweaking" of thekey design variables, and with no additional downstreamcosts.In effect, we get a better airplane for free. For a smaller number of variables (<6-8), adeterministic searching method (here represented by thefull-factorial Orthogonal Steepest Descent) provides aslightly better final result with about the same number ofcase evaluations For more variables, evolutionary/genetic methods getclose to the best final result with far-fewer caseevaluations. The eight variables studied herein probablyrepresent the practical upper limit on deterministicsearching methods with today’s computer speeds. Of the evolutionary methods studied herein, the BreederPool approach (which was devised during this research andappears to be new) seems to provide convergence in the fewestnumber ofcase evaluations, and yields results very close tothe deterministic best result. However, all of the methodsstudied produced similar results and any of them is asuitable candidate for use. Hybrid methods, with a stochastic initial optimizationfollowed by a deterministic final "fine tuning", proved lessdesirable than anticipated. Not a single case was observed, in over a hundred caseruns totaling over a million parametric design evaluations,of a method returning a local rather than global optimum.Even the modified commercial airliner, with poorly selectedinitial design variables far away from the global solution,was easily "fixed" by all the MDO methods studied. The postulated set of automated redesign procedures andgeometric constraints provide a more-realistic final result,preventing attainment of an unrealistic "better" finalresult. Especially useful is a new approach defined herein,Net Design Volume, which can prevent unrealisticallyhigh design densities with relatively little setup andcomputational overhead. Further work in this area issuggested, especially in the unexplored area of automatedredesign procedures for discrete variables.