End-to-end acceleration method for neural networks based on AUSM-like flux computing format
By introducing a neural network proxy model in fluid mechanics calculations, the flux calculation results are directly output, which solves the problem of long-term flux calculations in traditional calculations, and significantly improves the computing efficiency.
CN119443169BActive Publication Date: 2025-03-18CHINA AERODYNAMICS RES AND DEV CENT ULTRA-HIGH SPEED AERODYNAMICS RES INST
2 Cites 0 Cited by
Patent Information
- Application Number
- CN202510028613.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-08
- Publication Date
- 2025-03-18
- Estimated Expiration
- 2045-01-08
AI Technical Summary
Technical Problem
In traditional fluid mechanics, flux calculation takes a long time, especially in complex flow fields, which affects the overall computing efficiency.
Method used
Using the end-to-end acceleration method based on feedforward neural networks, the neural network proxy model is established, and the nonlinear operation in the AUSM-like flux calculation format is replaced, and the flux calculation results are directly output.
Benefits of technology
The steps of throughput calculation are significantly reduced, from 6 steps to 1 step, significantly improving the computing efficiency, especially in high concurrency environments.
✦ Generated by Eureka AI based on patent content.
Abstract
The present invention discloses a neural network end-to-end acceleration method of an AUSM-type flux calculation format, which relates to the field of fluid mechanics calculation, including: S1, realizing end-to-end input-output mapping based on a feedforward neural network to establish a neural network proxy model; S2, using the neural network proxy model of S1 to replace the AUSM-type flux calculation format in the fluid mechanics calculation process; wherein the input data of the neural network proxy model is: dimensionless physical quantities under any wind tunnel test condition of subsonic, transonic, supersonic, and hypersonic speeds; in the process of establishing the neural network proxy model, the Sobol sequence sampling method is optimized by locally encrypting the calculation-intensive area. The present invention provides a neural network end-to-end acceleration method of an AUSM-type flux calculation format, which introduces a neural network proxy model so that the flux that requires 6 steps to calculate can be directly obtained by inputting conditions in one step, significantly reducing the amount of calculation.
Need to check novelty before this filing date? Find Prior Art