Multidiscipline Design Optimization

Abstract

While formal optimization techniques are seeing increasing use within individual disciplines, application of this technology to the more general multidiscipline design problem is less common. This is due to both the inherent complexity of multidiscipline design and the fact that design is traditionally separated along discipline lines. The application of optimization to multilevel and multidiscipline design is discussed here. It is seen that, computationally, multilevel design within a single discipline and multidiscipline design across disciplines have similar features, and so are generally treated the same. Multidiscipline optimization at the conceptual level is first discussed and it is seen that this has been done successfully for some time. Then the more general case is discussed where formal mathematical decomposition of the larger problem is required to make optimization practical. Here, the state of the art is still relatively undeveloped. Two basic approaches are briefly described to indicate the concepts, and a simple example is offered. The key idea in persuing the multidiscipline design problem is that the optimum system is seldom the sum of optimum components. It is necessary to properly account for the coupling that exists among the subsystems, while still allowing the individual designer to work with relative freedom within his discipline. It is concluded that to achieve this, considerable research remains ahead.