Neural Network Fluid Flow Simulation for Fast Aerodynamic Prediction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for designing aerodynamic components, such as those in gas turbine engines, are inefficient due to the complexity of computational fluid dynamics (CFD) simulations and the need for extensive prototyping, which is time-consuming and resource-intensive.

Innovation Solution

The use of neural networks, specifically trained with pre-computed CFD outputs, to simulate fluid flow and predict aerodynamic performance of technical objects, allowing for faster optimization of design parameters and reduced need for physical prototypes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of time

If CFD simulations are used to model aerodynamic behaviour, then design optimisation can be achieved before prototyping, but the simulation process takes days or weeks to complete

Engineering Contradiction:
Improvedesign process timeVSAvoidsimulation speed
Core Design Contradiction:
Loss of timeVSProductivity

Solution Approach 1:

The neural network is trained in advance on a large dataset of CFD simulation results, storing the learned relationships between geometry parameters and flow characteristics. During actual design optimisation, the pre-trained network instantly predicts aerodynamic performance without running new CFD simulations, reducing design cycle time from weeks to minutes while maintaining high accuracy through the preliminary learning phase

Inventive Principle:
Principle #10Preliminary action

2Manufacturing precision

If multiple CFD runs are performed to optimise multiple design parameters, then comprehensive design optimisation is achieved, but computational resource consumption increases significantly

Engineering Contradiction:
Improvedesign optimisation accuracyVSAvoidcomputational resource consumption
Core Design Contradiction:
Manufacturing precisionVSUse of energy by stationary object

Solution Approach 1:

Instead of running multiple computationally expensive CFD simulations to evaluate different design parameters, the invention creates a neural network copy that has learned the complex flow physics from a single comprehensive CFD dataset. This neural network copy can then rapidly evaluate numerous design parameter combinations with minimal computational resources, maintaining design optimisation accuracy while dramatically reducing energy consumption

Inventive Principle:
Principle #26Copying

Data Source

PatentEP4510034A1Fluid flow simulation
Publication Date: 2025.02.19 ROLLS ROYCE PLC
  • EP4510034A1 patent drawingFigure 1
  • EP4510034A1 patent drawingFigure 2A~2B
  • EP4510034A1 patent drawingFigure 3A~3C

AI summary

A method of using a computer implemented neural network for a simulation of aerodynamic performance of a technical object having a geometry, the method comprising: Training the neural network using a plurality of sets of encodings of pre-computed computational fluid dynamics, CFD, outputs, wherein the training is generated using inputs comprising: a geometry of at least one training technical object; spatial locations of input nodes of the neural network as node attributes; a relationship between the geometry of the at least one training technical object and the neural network input node locations; associated boundary conditions; operating conditions; and computed outputs comprising flow fields and aerodynamic performance parameters; the training using a loss function that evaluates an error between a neural network output and the pre-computed CFD outputs to produce a trained neural network; using the trained neural network with new inputs to generate as output a predicted aerodynamic performance of the technical object, wherein the relationship between the geometry of the at least one training technical object and the neural network input node locations comprises a vector with at least two parameters.