Robust Design using classical optimization

The complexity of technical products increases significantly, due to an increasing number of interacting design variables of many components and subsystems. At the same time, the need for separated development processes is increasing due to specialization and outsourcing. Solution space methods are designed to solve this conflict. The requirements from an upper level, e.g. performance measures of the whole system, can be cascaded down to requirements on a lower level, e.g. performance measures of the subsystems or components, as it is done in the V-model approach. The method does not only take the numerous interactions into account but also guarantee the resulting intervals of different parameters to be independent of each other. Unfortunately, the computational cost of the state-of-the-art stochastic approach is high. The approach in this paper shows that the computational effort can be reduced considerably using a gradient based optimization approach for constraint problems. We demonstrate that the approach reduces the required number of function evaluations for a chassis design problem by a factor of 30, but more important, the CPU time for solving the problem by a factor of 20.     «

The complexity of technical products increases significantly, due to an increasing number of interacting design variables of many components and subsystems. At the same time, the need for separated development processes is increasing due to specialization and outsourcing. Solution space methods are designed to solve this conflict. The requirements from an upper level, e.g. performance measures of the whole system, can be cascaded down to requirements on a lower level, e.g. performance measures o...     »