Automatic mesh generation, recognized as the "Holy Grail" of Computational Fluid Dynamics (CFD), was highlighted as a critical objective in the NASA CFD Vision 2030 study. Adaptive Cartesian grid generation has attracted significant interest due to its high level of automation and low manual intervention. However, its broad use in multicore parallel environments has been hindered by significant load imbalance. Traditional parallel techniques distribute grid cells evenly after each adaptive mesh refinement (AMR) cycle, but they overlook the significant variation in computational costs associated with geometry data retrieval. During this phase, different cells require varying numbers of iterations, resulting in substantial differences in processing time and reducing overall computational efficiency.
To tackle this main challenge, researchers from Nanjing University of Aeronautics and Astronautics introduced a dynamic partition weight (DPW) strategy. This innovative approach combines features of generated Cartesian grid cells before and after the AMR process. It also estimates the number of iterations needed for each new cell during k-d tree search and assigns its partition weight proportionally to that number. The grid-parallel repartition, which accounts for partition weights, is applied before steps that involve k-d tree searches. Experimental results demonstrate that the DPW strategy significantly speeds up grid generation using the same number of cores, providing a simple yet effective solution to parallel computational imbalance. For a complex model with 200,000 surface triangles, the DPW strategy reduces grid generation time from 610.74 seconds to just 36.50 seconds—an improvement of 94.02%.
A key achievement of this work is its progress in parallel efficiency. Numerical tests show notable performance enhancements. When generating a 1.37-billion-cell grid for the Common Research Model (CRM) wing-body setup, scaling from 128 to 1,024 cores decreases the total process time to only 44.49 seconds, while maintaining a parallel efficiency of 73.96%. This indicates that the DPW strategy maintains high efficiency and scalability in adaptive Cartesian grid generation, even with over a thousand cores.
Additionally, the study emphasizes the significance of the "iteration count inheritance" mechanism during AMR. The researchers found that new grid cells inheriting predicted iteration counts from their parent cells help sustain computational load balance. Throughout multiple AMR cycles, this inheritance mechanism enables accurate load prediction, reducing the maximum process time difference from over 100 seconds with traditional methods to just 7.52 seconds, thereby improving the stability and predictability of parallel computation. The validated dynamic partition-weight criterion provides a robust, dependable load-balancing method for parallel adaptive Cartesian grid generation. Furthermore, insights from iteration-count inheritance provide valuable guidance for enhancing the design of parallel adaptive algorithms more broadly.
Looking forward, researchers plan to expand the application of the DPW strategy. The next step involves addressing motion challenges in complex three-dimensional geometries, such as developing flow fields around rotating parts and fluid-structure interaction. This will better demonstrate its advantages in speed and efficiency. Using the DPW strategy, automatic, rapid mesh modifications at motion boundaries could enable more efficient simulation of motion problems at lower cost.
Original Source
Hang Chen, Zhenming Wang, Linlin Tian, Jianming Liu, Ning Qin, Ning Zhao. A dynamic partition weight strategy for accelerating parallel adaptive Cartesian grid generation [J]. Chinese Journal of Aeronautics, 2025, https://doi.org/10.1016/j.cja.2025.103921.
About Chinese Journal of Aeronautics
Chinese Journal of Aeronautics (CJA) is an open access, peer-reviewed international journal covering all aspects of aerospace engineering, monthly published by Elsevier. The Journal reports the scientific and technological achievements and frontiers in aeronautic engineering and astronautic engineering, in both theory and practice. CJA is indexed in SCI (IF = 5.7, Q1), EI, IAA, AJ, CSA, Scopus.
