Machine Learning Predicts Reactive Flow Behavior
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for simulating reacting fluid flows are computationally intensive and time-consuming, limiting their application in product development and optimization due to high costs and long simulation times.
Innovation Solution
A method combining numerical simulation techniques, such as CFD, with machine learning models, specifically recurrent neural networks like ConvLSTM, to predict the behavior of reactive flows, thereby reducing the need for extensive simulations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If CFD simulations are used to study reacting flows, then accuracy of flow behavior prediction is improved, but simulation time and computational cost increase significantly
Solution Approach 1:
The patent performs preliminary CFD simulations to train a machine learning model during an initial phase. Once trained, the ML model can predict flow behavior without requiring time-consuming CFD simulations for each new scenario, thus reducing simulation time while maintaining accuracy through the pre-trained model's knowledge
Solution Approach 2:
The patent creates a digital copy of the complex CFD simulation process in the form of a trained machine learning model. This ML model replicates the predictive capability of CFD simulations but operates much faster, effectively copying the essential functionality while eliminating the computational burden of full CFD re-simulations
2Measurement precision
If lab scale testing is conducted to gather physics insights, then understanding of underlying mechanisms is improved, but cost and complexity increase
Solution Approach 1:
The patent uses virtual digital twins and machine learning models to create simplified representations of complex physical systems. These digital copies allow researchers to study underlying mechanisms through computational experiments that are cheaper and simpler than physical lab tests while maintaining sufficient accuracy for insights
Solution Approach 2:
The patent introduces machine learning models as intermediary tools between physical lab experiments and final design decisions. These intermediaries process data from simplified experiments or existing datasets to provide insights about complex reacting flows without requiring direct complex physical testing
Data Source
AI summary
Systems and methods for transient predictions in reacting flow simulations. In one embodiment, the method includes generating an initial simulation of the reactive flows, by using a simulation technique, for a first time duration. The method includes predicting, by a machine learning model, a behavior of the reactive flows during a second time duration based on the initial simulation. The second time duration is consecutive to the first time duration. The method includes generating a subsequent simulation for the reactive flows for a subsequent time duration. The method includes providing data corresponding to the predicted behavior and the subsequent simulation as an input to the machine learning model to predict the behavior of the reactive flows for the subsequent time periods. In addition, the method includes repeating generating the subsequent simulation and predicting behavior of the reactive flows for one or more successive time durations, until the reaction of the reactive flows is completed.