CAREER: Bridging Geometric Design and Aerodynamic Simulation of Turbomachinery: An Integrative Design-Through-Analysis Framework Enabled by Embedded Domain Methods

The transfer from computer-aided geometric design to physics-based computer simulation requires geometry processing and mesh generation procedures, which often constitute significant barriers to creating rapid design-through-analysis workflows. The objective of this project within the Faculty Early Career Development (CAREER) Program is to remove these barriers by developing new computational methodologies that enable seamless integration and automation. Turbomachinery is a key enabling energy technology, e.g., for electricity production and air transport. The new design-through-analysis methodologies envisioned in this project aims to fundamentally facilitate simulation-based turbomachinery design and optimization. These can lead to critical improvements in design procedures that serve the national interest, promote national prosperity, and enable unconventional turbomachinery designs that improve sustainability and energy efficiency. The project results also promote the progress of computational science, in particular by contributing significant methodological advances to the current drive towards unsteady full-wheel simulations. The project includes an education component that combines a new summer camp with research internship opportunities for high school and undergraduate students. It provides opportunities for energizing K-12 and undergraduate experiences that motivate participants to join advanced degree programs in engineering, with particular emphasis on minorities and underrepresented groups. This will directly strengthen the infrastructure for STEM outreach at the University of Minnesota and contribute to developing and maintaining a well-trained STEM workforce, which is important for the future competitiveness of related industries in the United States.

The research approach revolves around the idea of combining parametric geometry modeling with new embedded domain finite element methods for integrated computational aerodynamics. While parametric modeling techniques are available in commercial design tools, embedded domain methods remain plagued by serious technology gaps that prevent their application for high-fidelity aerodynamics. Examples include the detrimental impact of elements with small cuts on conditioning and time step size, low-order accuracy of quadrature rules in cut elements, and algorithms that are unsuitable for emerging heterogeneous computing architectures. The research approach focuses on devising new techniques that effectively close these gaps, incorporate key computational aerodynamics paradigms such as the Discontinuous Galerkin concept and the variational multiscale method for large eddy turbulence modeling, and enable high-order accuracy beyond established second-order codes. The superior design-through-analysis capabilities are demonstrated for multistage unsteady aerodynamic simulations of turbine components, involving complex blade geometries, turbulent high-Reynolds-number flows and moving fluid domains.