CFD solution acceleration method and system based on parallel computing and load dynamic balancing

By dynamically partitioning subdomains into grids in a parallel computing cluster and performing load assessment and adaptive grid migration, the problems of unbalanced load and low communication efficiency in traditional CFD computing are solved, achieving efficient computational flow field solution.

CN121187806BActive Publication Date: 2026-06-09SHENZHEN Y& D ELECTRONICS CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202511727903.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-11-24
Publication Date
2026-06-09
Estimated Expiration
2045-11-24

AI Technical Summary

Technical Problem

Traditional CFD parallel computing methods suffer from uneven computational load, low communication efficiency, and lack of adaptability in hypersonic flow fields, resulting in long solution times and low resource utilization efficiency.

Method used

By probing the hardware information of the parallel computing cluster, dynamically dividing the subdomain grid data, and performing load assessment based on physical characteristic parameters, adaptive grid migration and asynchronous computing and communication overlap are achieved, thereby optimizing load balancing and communication efficiency.

Benefits of technology

It achieves efficient utilization of computing resources, reduces communication waiting time, improves the solution speed and resource utilization efficiency of computational flow fields, and solves the problems of load imbalance and communication bottleneck.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121187806B_ABST
    Figure CN121187806B_ABST
Patent Text Reader

Abstract

The application relates to the technical field of computation, and provides a CFD solution acceleration method and system based on parallel computation and load dynamic balancing. Physical characteristic quantities such as CFL numbers and residuals are quantified as load indexes, so that load balancing decisions are closely coupled with flow field evolution rules. By using an asynchronous communication overlap framework which is deeply coupled with a CFD solver, blocked communication waiting time is converted into effective calculation time, and communication delay is systematically hidden. Based on a real-time load evaluation incremental grid migration strategy, high-load grid blocks in hotspot processes can be intelligently identified and migrated, and differentiated and efficient use of computing power is realized.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • A supercomputer-based overlap optimization method for parallel CFD computation and communication is proposed

    CN109088776A

  • System and method for load balancing for parallel computations on structured multi-block meshes in cfd

    US20140365186A1