Improving Finite Volume Methods for Industrial CFD: Adaptation, Error Quantification, and Robust Convergence

点击翻译按钮获取中文摘要

As computer power has grown, the complexity of problems simulated using computational fluid dynamics (CFD) has grown apace. Greater emphasis is also being placed on quantifying the accuracy of simulation results. Existing techniques for this task often have prohibitively large computational resource requirements.Addressing this need requires a new paradigm in which a robust flow solver with adaptive and error quantification capabilities reliably produces solutions with known error bounds with minimal human intervention. This project will prototype new techniques for estimating and reducing numerical error in simulation of flow problems, and for making commercial CFD software more efficient and able to compute solutions more robustly. We will do this work in a unified framework compatible with the second-order unstructured mesh finite volume methods employed in modern commercial CFD solvers. We expect that these advances can begin to be deployed in current commercial CFD solvers within two to five years. We will also work with ANSYS Canada to assess the use of new algorithms developed in our previous collaborative project in combination with their existing commercial flow solver. While the present work will focus on second-order methods, we expect that more accurate finite volume methods will migrate into commercial CFD solvers, and we will plan for a clear upgrade path to providing adaptation and robustness improvements for high order methods in the five to ten year time frame.These advances will provide an important competitive advantage to our industrial partner in this project, ANSYS Canada, as the vendor whose software will be the first to provide these capabilities. In turn, their Canadian customers will benefit through improved accuracy of and confidence in simulation results and increased productivity of engineers using CFD.