Computational fluid dynamics (CFD) has emerged as a third pillar of the fluid mechanics field, joining analytical and experimental studies. When appropriately combined with experimental validation, the highly detailed solutions provide unprecedented access to small scale flow features. CFD continues to grow as a tool of discovery and design with the increasing availability of powerful computational resources.
Many engineering tasks, such as design or uncertainty quantification, require large numbers of model evaluations. CFD simulations typically require hours to days to execute. Even with modern computational assets, some very high resolution models can require run-times on the order of months. Standalone CFD simulations are not appropriate for this type of multi-run task. Researchers have tried to bridge the gap by building reduced order models (ROMs) to use as surrogates for high fidelity CFD. These ROMs are typically fast running formulations that have been trained on detailed CFD results.
AccelerateCFD© is based around a novel formulation of ROM evolution equations that allows for direct control of specific boundary conditions. With a properly constructed parameterized model, users can provide an arbitrary input transient function within clearly defined bounds. While constructing an AccelerateCFD© model will incur some upfront cost, the same model can be used for many transients and the total speedup will be several orders of magnitude for ideal applications. When many runs are required, the total speedup is likely to be several orders of magnitude. AccelerateCFD© will provide powerful model reduction capabilities while allowing users the capacity to construct ROMs through CFD packages of their choice.
