This paper presents Structured Multi-attribute Optimization (SMO) which is a general methodology to solve complex optimization problems. It is based on the integration of general optimization concepts with specific knowledge about the problem itself in order to transform a complex optimization problem into a structure of simple optimization subproblems. SMO is the result of the experience of a research team in solving complex engineering problems, mainly related to two different areas of application: optimal design of electrical machines and systems engineering in the space industry. The example which is used throughout this paper is the design of 3-phasesquirrel-cage air-cooled induction motors, in which the manufacturing costs and the motor efficiency have to be optimized simultaneously, subject to electromagnetic, mechanical, thermal and manufacturing constraints. The problem is solved by direct search, since it presents several characteristics that make it hard to solve by classical optimization techniques. When using the SMO methodology the original complex problem is solved through a sequence of problems called optimization phase problems in such a way as the solution of phase «i» problem is used as an initial solution for the problem of phase «i+1». Furthermore, every optimization phase problem is decomposed into a structure of subproblems interrelated through recursive partial optimizations and iterative partial optimizations. SMO has been successfully applied to very diverse type of real-world engineering problems. This proves its capability to be adapted to very different contexts. Given the complexity of the addressed problems, the required computing times are very reasonable even when using personal computers.
Keywords: Optimization, computer-aided design, induction motor
Third Stockholm Optimization Days. Stockholm, Sweden, 25-26 Junio 1992
Published: June 1992.