Enhanced total derivative based coupling suspension in complex system optimization

The design of complex engineering systems requires an initial decomposition of the system into subsystems. These systems are linked together by couplings, which convey output data from one subsystem to the input of another. Because complex engineering systems can have hundreds or thousands of such couplings, the optimization of these systems is often computationally expensive. To reduce the optimization time, it becomes important that a system designer have the ability to select couplings that have little effect on the solution accuracy, and then temporarily remove them. Previous coupling strength analysis methods have developed models for selecting couplings for temporary removal either based on the impact of coupling suspension on solution accuracy or the local sensitivity related to the coupling. The method presented in this paper enhances the technique targeting couplings for suspension based on their impact on solution quality. The method considers both the solution impact and the local sensitivity effects of suspending a coupling. This will reduce the instabilities introduced into the system analysis phase when only considering solution accuracy. In addition, new error propagation models are used, improving the accuracy of the introduced error prediction for multiple cycle coupling suspension. The application of this new method is discussed, as well as the implementation on sample problems. The method is expected to improve solution stability and allow for longer periods of coupling suspension due to the improved error models.