In multi-disciplinary optimization of flexible wings that may involve multi-fidelity structural dynamic models to be optimized concurrently, it is necessary to maintain consistency between these models of different fidelity levels. Even though a two-way communication is eventually needed in the process, this paper only presents a study to update high-fidelity structural dynamic models of flexible wings, based on the input of low-fidelity models. To address the consistency requirement between the two types of models, natural frequencies and mode shapes of their fundamental modes should be correlated. A multilayer feed-forward artificial neural network is created to map the modal information from the low-fidelity models to high-fidelity ones, which captures the impact of the high-fidelity models design variables on the desired consistency. Eventually, the high-fidelity model that maintains the desired consistency is determined by an optimization process based on the surrogate. Numerical results illustrate that flat plate and wing box models based one shell finite elements are updated based on modal information from beam representations of flexible wings. This approach has the potential to benefit the multi-fidelity and multi-stage optimization for the conceptual design of new aircraft platforms.
展开▼