Magnetic bearing actuator design using genetic algorithms

Like most useful devices, magnetic bearing actuators have a large number of design parameters that must be selected in design optimization. The usual approach in the design of these devices is to limit the number of parameters by introducing relationships between them that are presumed consistent with optimal design, and then to iterate until a feasible design is found that meets all requirements. In the present work, the magnetic bearing actuator design is reformulated as a catalog selection problem and solved using a genetic algorithm (GA). The resulting designs are compared with solutions found using more conventional design procedures and are found to be superior. By challenging the embedded assumptions in the traditional design process, the CA approach reveals new and useful design concepts. In addition, the GA/catalog approach is more amenable to commercial design than is parametric optimization.