Reinforcement Learning Turbulence Model for RANS Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current turbulence models, particularly those using the Reynolds-Averaged Navier-Stokes (RANS) equation, struggle with high accuracy in calculating complex flows at high Reynolds numbers, leading to significant deviations from actual conditions, making it difficult to implement accurate turbulence field updates.
Innovation Solution
A turbulence field update method that combines reinforcement learning with the RANS equation, where sample turbulence data is used to train a reinforcement learning model, which then processes initial turbulence data to predict Reynolds stress, thereby improving calculation accuracy and reducing the influence of differences between high- and low-Reynolds-number training data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional RANS equation with simple flow turbulence models is used, then the device complexity is low, but the manufacturing precision (calculation accuracy) deteriorates for high-Reynolds-number separated flows
Solution Approach 1:
A neural network is introduced as an intermediary component between the RANS equation solver and the turbulence model. The neural network processes flow field data and predicts turbulence parameters, acting as a mediator that bridges the gap between simple RANS formulations and complex turbulence physics, thereby improving calculation accuracy without requiring complete reformulation of the underlying equations
Solution Approach 2:
The turbulence modeling approach combines traditional RANS equation framework with machine learning components to create a hybrid composite model. This composite structure integrates the physical basis of RANS with the pattern recognition capabilities of neural networks, achieving superior performance compared to either approach alone
2Reliability
If turbulence models derived from simple flows are used, then the ease of manufacture is high, but the reliability deteriorates for complex high-Reynolds-number flows
Solution Approach 1:
The neural network is pre-trained using high-quality turbulence data from direct numerical simulations (DNS) and experimental measurements before being integrated into the RANS solver. This preliminary training phase allows the model to learn accurate turbulence relationships in advance, ensuring reliability when applied to complex high-Reynolds-number flows without requiring complex development procedures
Solution Approach 2:
The approach copies accurate turbulence behavior patterns from high-fidelity DNS simulations and experimental data into the neural network model. By learning from these accurate reference solutions, the simplified RANS-NN model reproduces complex turbulence physics without requiring direct simulation of the full complexity
3Manufacturing precision
If reinforcement learning turbulence model is used to improve generalization ability, then the manufacturing precision improves, but the device complexity increases
Solution Approach 1:
The reinforcement learning turbulence model implements a feedback mechanism where the neural network predictions are continuously evaluated against the RANS equation solutions and flow field evolution. The model learns to adjust its predictions based on the feedback from how well they contribute to accurate turbulence field updates, improving precision through iterative learning while managing complexity through focused training objectives
Data Source
AI summary
Provided are a turbulence field update method, apparatus, and device, and a computer-readable storage medium. The method includes: obtaining sample turbulence data; performing model training by use of the sample turbulence data to obtain a reinforcement learning turbulence model; calculating initial turbulence data of a turbulence field by use of a Reynolds Averaged Navior-Stokes (RANS) equation; processing the initial turbulence data by use of the reinforcement learning turbulence model to obtain a predicted Reynolds stress; and performing calculation on the predicted Reynolds stress by use of the RANS equation to obtain updated turbulence data.