Evaluation of Multidisciplinary Optimization Techniques Applied to a Reusable Launch Vehicle

Abstract

Three multilevel multidisciplinary optimization techniques, Bi-Level Integrated System Synthesis, Collaborative Optimization, and Modified Collaborative Optimization, are applied to the design of a reusable launch vehicle, evaluated, and compared in this study. In addition to comparing the techniques against each other, they are also compared with designs reached via fixed-point iteration of disciplines with local optimization and the industry accepted multidisciplinary optimization technique, All-at-Once. The new multidisciplinary optimization techniques, particularly Bi-Level Integrated System Synthesis, showed greater ability than fixed-point iteration to design for a global objective and were more applicable to complex systems than All-at-Once. This study was the first time that the novel multidisciplinary optimization methods were compared qualitatively and quantitatively under controlled experimentation practices. It is still impossible to statistically determine whether any one of the novel multidisciplinary optimization techniques is better than another, because more studies using different test problems corroborating the conclusions made here are needed. Nomenclature A = area c = characteristic exhaust velocity c F = thrust coefficient g = inequality constraint h = equality constraint/enthalpy/altitude Isp = specific impulse _ m = mass rate MR = mass ratio, W gross =W insertion P = power p = pressure Perf = performance Prop = propulsion r = fuel to oxidizer mixture ratio (propellant mixture ratio) S = wing area SF = scale factor T = thrust W = weight w = weighting factor (for BLISS) W&S = weights and sizing X = input or design variable Y = output or behavior variable V = change in velocity " = nozzle expansion ratio _ = pitch angle rate = design objective Subscripts c = combustor e = exit eng = engine loc = local o = optimized ref = reference req = required sh = shared input to multiple CAs but not calculated by any CA (for BLISS) SL = sea-level sys = system t = throat vac = vacuum veh = vehicle Superscripts pf = performance (for CO and MCO) pp = propulsion (for CO and MCO) t = target from system optimizer (for CO and MCO) ws = weights and sizing (for CO and MCO) = output passed to a CA (for BLISS) = output from a CA to system (for BLISS)

Multidisciplinary Design Optimization: A Survey of Architectures

“…However, the results of Brown and Olds [143] show strange behavior in a practical design problem. In particular, when the penalty parameter was above a threshold value, the algorithm could not improve on the initial design point.…”

“…These problems can be mitigated by the use of surrogate models or reduced-order modeling as described above. The one benchmarking result available for BLISS-2000 is the launch-vehicle design problem of Brown and Olds [143]. In this case, BLISS-2000 tended to outperform other distributed architectures, and it should be cost-competitive with other monolithic architectures when coarse-grained parallel processing is fully exploited.…”

Benchmarking multidisciplinary design optimization algorithms

“…Results from various studies have shown that the performance of a method can be dependent on its implementation, the characteristics of the problem being solved, and the optimizer employed (Alexandrov and Kodiyalam 1998;Brown and Olds 2006;Perez et al 2004). Furthermore, for some problems, specific methods may either fail to return an optimum, or may not be suited to implementation (Brown and Olds 2006). Additionally, comparing results between studies can be difficult as the performance of an MDO method can depend on specific implementation details.…”

“…Typically the selection of an MDO method is done in an ad hoc manner, since few benchmarking studies are available to make an informed decision (Alexandrov and Kodiyalam 1998). Results from various studies have shown that the performance of a method can be dependent on its implementation, the characteristics of the problem being solved, and the optimizer employed (Alexandrov and Kodiyalam 1998;Brown and Olds 2006;Perez et al 2004). Furthermore, for some problems, specific methods may either fail to return an optimum, or may not be suited to implementation (Brown and Olds 2006).…”

Development of a collaborative optimization tool for the sizing design of aerospace structures

“…The structure is a space module of the International Space Station (ISS) designed by ALCATEL ALENIA Space. An example of the use of a CO approach to the aerospace application is available in Brown [11] (where a more generic optimization on the global design of a space vehicle, with a comparison of three different methods, is reported) and in Wallace [12], where the philosophy of the CO design is applied as well as it was done for this work. The FE model 2 has an high definition degree and was created taking into account (1) the primary structure (no inside structures are presented) and (2) a large set of loads (Fig.…”