Inlet Coverage Prediction Using CFD Surrogate Flow Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computational fluid dynamics (CFD) techniques for assessing flow source coverage in industrial simulations are complex, time-consuming, and require manual post-processing, leading to inefficiencies and incomplete coverage analysis in large or complex geometries.
Innovation Solution
A computer-implemented system and method using machine learning surrogate models, specifically deep learning (DL) models, to predict and visualize the coverage of input flow sources in a simulation environment, optimizing the design process with real-time, immersive collaboration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional CFD simulation methods are used to assess flow source coverage, then measurement precision is improved, but productivity deteriorates due to extremely long simulation times ranging from hours to months
Solution Approach 1:
The patent creates a digital twin (virtual copy) of the physical flow field environment. This digital twin replicates the complex CFD simulation capabilities but runs much faster, enabling rapid assessment of flow source coverage without requiring hours or months of traditional simulation time. The digital twin is trained on CFD data and can predict flow patterns, coverage metrics, and optimization opportunities in real-time or near-real-time.
Solution Approach 2:
The system performs preliminary actions by pre-processing the complex CFD simulation data to train the digital twin model beforehand. Once trained, the digital twin can instantly provide coverage assessments without requiring full CFD simulations each time. This preliminary preparation enables rapid iterative design optimization.
2Measurement precision
If manual post-processing steps are used to identify coverage of inlets, then measurement precision is improved, but loss of time increases due to repetitive manual analysis
Solution Approach 1:
The digital twin system performs self-service by automatically analyzing flow field data, identifying coverage patterns, and generating optimization recommendations without requiring manual post-processing. The system autonomously processes simulation data, calculates coverage metrics for multiple inlets, and provides actionable insights, eliminating the repetitive manual analysis step while maintaining high measurement precision.
3Measurement precision
If the number of streamlines or particles is increased to cover the entire simulation domain, then measurement precision is improved, but use of energy increases to computationally prohibitive levels
Solution Approach 1:
Instead of increasing the number of streamlines or particles in traditional CFD simulations (which exponentially increases computational energy consumption), the patent uses a digital twin that has already learned flow patterns from training data. The digital twin can assess complete domain coverage with minimal computational resources, effectively copying the results of high-resolution simulations without the associated energy cost.
4Measurement precision
If multiple iterations are performed to decide number of streamlines/particles and injection points, then measurement precision is improved, but loss of time increases due to iterative optimization requirements
Solution Approach 1:
The digital twin system incorporates feedback mechanisms that automatically analyze coverage results and provide optimization recommendations. Instead of requiring multiple manual iterations to adjust streamline/particle numbers and injection points, the system uses feedback loops to automatically refine parameters, assess coverage improvements, and converge on optimal configurations much faster than traditional iterative approaches.
Data Source
Figure 1
Figure 2
Figure 3A~3C
AI summary
A computer-implemented method (500) to assess the coverage of input flow sources in a simulation environment is disclosed. The method (500) comprises generating, (502) flow vectors across plurality of surfaces of the simulation environment associated with predefined parameters. Further, the method (500) comprises running, (504) a flow of the fluid to simulate the flowing rates of the fluid within the simulation environment to define passive scalar functions for each zone of influence of each of a plurality of inlets within the simulation environment. Further, the method (500) comprises predicting, (506) coverage of the plurality of inlets based on simulated flowing rates of the fluid and defined passive scalar functions for each zone of influence of the plurality of inlets within the simulation environment. Further, the method (500) comprises plotting, (508) a coverage function to visualize each zone of influence of each of the plurality of inlets based on the predicted coverage.