2022 AI Testbed Expeditions Report
Citations Over Time
Abstract
Computational Fluid Dynamics (CFD) is used in the design and optimization of gas turbines and many other industrial and scientific applications.Its practical use is often limited by the high computational cost, and the requirement of accurately resolving near-wall flow is a significant contributor to this cost.Machine learning and other data-driven methods can complement existing wall models.Nevertheless, training accurate and therefore often large models presents a large cost in itself, in part due to the computational effort and memory footprint required by back-propagation.Recent work has presented alternatives for computing gradients of neural networks that does not require a separate forward and backward sweep, nor requires storage of intermediate results between sweeps.The approach instead computes an unbiased estimator for the gradient in a single forward sweep.We implemented this approach in a framework built on top of PyTorch.In this paper we discuss the application of this approach to a CFD wall model that could potentially be used as a surrogate in wall-bounded flow simulations to reduce the computational overhead while preserving predictive accuracy.
Related Papers
- → A hovercraft testbed for decentralized and cooperative control(2004)29 cited
- → A networked and cooperative control testbed(2004)
- Susquehanna Chorale Spring Concert "Roots and Wings"(2017)
- → Ditto Subsystem : A Case Study for Building Experiment Environment in StarBED Using TopDL(2017)
- → Deterlab testbed for cybersecurity research and education(2013)