An overview of computer based design synthesis and optimisation

In general the optimisation problem will involve many competing disciplines, e.g. field quality, internal stresses, heating, costs etc. and these problem´s may be competing. Also the variables are usually constrained and the global minimum is often required. Useful automatic optimisation techniques have been developed that should have an impact on the design of magnetic devices. The problem of searching for global minima has been partially overcome by use of stochastic methods but these are expensive to apply as there are often many objective functions to calculate. What can be recommended is that the stochastic method, evolution or simulated annealing can be used to survey the parameter space and identify trends toward a deeper optimum and then speedily close the solution by using a direct search method. Many problems involve multiple conflicting objectives that must be mutually reconciled. This conflict can often be resolved by the use of decision making methods which include some procedures for defining and assuring a minimally acceptable outcome for each objective. It must be emphasised, however, that the use of optimising methods to minimise objectives is only part of the total problem of design. The techniques briefly reported in this paper must be supplemented by sensitivity analysis in order to obtain good starting points in parameter space and also the processes of automatic synthesis based on design rules and engineering knowledge