Optimization of complex systems like jet engines is a process where discipline experts from several departments or even different companies have to work together. Thus, a complete system analysis code usually does not exist preventing an overall system optimization. Therefore, the system is typically split into components, and interface parameters between components are decided and fixed in early design stages based on low-fidelity information. Finally the components are optimized separately according to these fixed settings which, however, may prohibit an overall optimal system behavior even if sophisticated subsystem optimization is performed. Consequently, interface parameters should be varied in an overall coupled system optimization process and adapted to subsystem needs.
This may be supported by utilizing collaborative optimization strategies. Basically there are two types of such optimization approaches: global strategies with nested optimization loops and local subsystem optimization strategies supported by sensitivity-based approximations of other subsystems. The multidisciplinary optimization approach presented here combines the benefits of both strategies: efficient global optimization and approximations of subsystem quantities without the need of sensitivity information. It starts with an initial design of experiments for each component by varying all input parameters and evaluating the associated outputs. Due to the fact that these quantities consist of shared design parameters and outputs of other components, not all of the used input parameter combinations are feasible for the coupled system. In order to enforce consistency for the entire system, the interface regions are characterized by feasibility criteria acting as approximated constraints for further component optimizations. Beside that an approximation of the overall system objective is provided to all components to drive component design towards an overall optimal system performance.
The developed approach is demonstrated by an application to a jet engine turbine consisting of a high and low pressure part where the goal is to maximize the overall turbine efficiency. On the one hand the turbine is optimized with the proposed approach by splitting it into two coupled components; on the other hand the turbine is optimized as a complete system. It turns out that the proposed approach yields equally good results in much shorter time than overall optimization.