Global optimization methodologies, belonging to both the deterministic and the stochastic class, are discussed. A higher-order-deterministic optimizer and a zeroth-order-stochastic one are compared for the solution of a practical case study, taken from electromechanics. A non-linear constrained design problem is formulated and solutions, in terms of optimal shapes of the device considered, are determined. Some improvements in generating feasible random configurations have been suggested. On average, stochastic zeroth-order strategies were superior to their deterministic higher-order counterparts.
Get full access to this article
View all access options for this article.
References
1.
SalonS.J. and IstfanB., IEEE Trans. Magn.MAG-22 (1986) 817.
2.
RudnickiM. and SaviniA., Proc. Int. AMSE Conf., Istanbul, Turkey (1988) vol. 2A, pp. 51–58.
3.
GottvaldA., PreisK.MageleC.BiroO. and SaviniA.IEEE Trans. Magn.MAG-28 (1992) 1537.
4.
AngerG., Inverse Problems in Differential Equations (Plenum Press, New York, 1988).
5.
VanderplaatsG.N., Numerical Optimization Techniques for Engineering Design (McGraw-Hill, New York, 1984).
6.
GottvaldA., PreisK. and FriedrichM. in: Proc. IGTE Symp., Graz, Austria (1990) pp. 23–28.
7.
ScalesJ.A., SmithM.L. and FischerT.L., J. Comput. Phys.103 (1992) 205.
8.
BrodanG.S. and HaydenJ.T., Advanced Electrical Power and Machines (Pitman, London, 1966).
9.
PapoulisA., Probability, Random Variables and Stochastic Processes (McGraw-Hill, New York, 1965).
