Soft computing in engineering design optimisation

The implementation of Soft Computing methodologies in two aerospace design problems is presented, one being the design of quiet and efficient aircraft propellers, and the other being the manoeuvre control of a satellite. They were chosen as they present very challenging engineering design problems with nonlinearities and discontinuities in the design space. The methodologies used include Simulated Annealing for design optimisation, Neural Networks for system representation and Fuzzy Logic for system control. The choice of these methods over conventional analytical techniques is shown to enable the solution of these design problems. The propeller design methodology described produces designs that have equivalent or improved performance and significantly reduced noise when compared to commercial off-the-shelf designs. The optimal satellite controller described shows significant improvements in reduced settling time and overshoot when compared to conventional controllers.