Reinforcement Learning CFD Automation for Blade Flow Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The complexity of flow around blades in fluid machinery makes it difficult to determine ideal analysis settings for computational fluid dynamics (CFD) simulations, leading to time-consuming and costly iterative processes with variable results dependent on engineer judgment.
Innovation Solution
A computational fluid dynamics (CFD) automation method using reinforcement learning to generate an analysis automation model that predicts CFD flow analysis results, optimizing and automating the process by determining analysis setting parameters such as computational mesh generation, turbulence model selection, and numerical method choice.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If iterative simulations are used to find appropriate CFD settings, then analysis accuracy can be improved, but time consumption and costs increase significantly
Solution Approach 1:
The patent applies preliminary action by training a reinforcement learning model in advance to learn optimal CFD settings from historical simulation data. Once trained, the model can directly predict appropriate settings for new blade geometries without requiring iterative simulations, thus achieving high accuracy while eliminating time-consuming trial-and-error processes.
Solution Approach 2:
The system implements self-service by enabling the CFD analysis process to automatically determine optimal settings through the reinforcement learning model without human intervention. The model autonomously selects mesh generation parameters, turbulence models, and numerical schemes based on input blade geometry, making the system self-sufficient and eliminating repetitive manual configuration.
2Reliability
If iterative simulations are performed to adapt settings to different flow conditions, then analysis reliability can be improved, but the process becomes more complex and costly
Solution Approach 1:
The patent applies parameter changes by using the reinforcement learning model to automatically adjust CFD simulation parameters (mesh density, turbulence model selection, numerical schemes) based on input blade geometry and flow conditions. The model learns optimal parameter configurations from training data and adapts them to different cases, ensuring reliable results while simplifying the process compared to manual iterative adjustment.
Solution Approach 2:
The system implements feedback mechanisms through the reinforcement learning training process, where simulation results are fed back to update and improve the model's predictions. The model learns from historical simulation data and evaluation metrics, continuously refining its ability to predict optimal settings, thereby improving reliability while maintaining process simplicity.
3Measurement precision
If professional settings are manually configured for CFD analysis, then analysis accuracy can be maintained, but the process requires significant expertise and time
Solution Approach 1:
The system applies self-service by enabling automatic configuration of CFD settings through the reinforcement learning model. The model independently determines optimal mesh generation parameters, turbulence models, and numerical schemes based on input geometry, eliminating the need for manual expert configuration while maintaining high analysis accuracy.
Solution Approach 2:
The patent applies mechanics substitution by replacing the manual expert judgment and mechanical configuration process with an automated reinforcement learning-based prediction system. The ML model substitutes human expertise in setting CFD parameters, transforming a skill-dependent manual process into an automated intelligent system that maintains accuracy while dramatically improving ease of operation.
4Measurement precision
If settings are optimized for specific conditions, then analysis accuracy improves, but the settings become less effective when flow conditions or geometry change
Solution Approach 1:
The patent applies universality by training the reinforcement learning model on diverse training data encompassing various blade geometries and flow conditions. The model learns generalizable patterns that allow it to predict optimal CFD settings for a wide range of applications, making the system universally applicable rather than condition-specific. This enables the same model to accurately configure simulations for different geometries and flow regimes without requiring condition-specific tuning.
Data Source
AI summary
Disclosed is a computational fluid dynamics (CFD) flow analysis method for blades includes: generating an analysis automation model that predicts CFD flow analysis results according to input analysis conditions, the generating an analysis automation model comprising inputting a blade shape and flow conditions, determining analysis settings for flow analysis over blades, performing CFD simulation based on the determined analysis setting parameters to conduct flow analysis over blades, evaluating results of the performed flow analysis over blades, and training through reinforcement learning using artificial neural networks to satisfy predetermined evaluation criteria, inputting arbitrary blade shape conditions and flow conditions, and determining analysis setting parameters by applying the analysis automation model to the input blade shape and flow conditions, predicting CFD simulation, and outputting the predicted CFD simulation as flow analysis results.