Learn how to validate CFD and ML models for combustion problems, by choosing appropriate data sets, applying suitable validation metrics, performing sensitivity analysis, and comparing and integrating CFD and ML models.

Test different turbulence models to identify their effect on combustion predictions. In a gas turbine combustion simulation, we assessed the sensitivity of our results to various turbulence models, including k-ε, k-ω & LES. By comparing the predicted turbulent flame speeds and temperature distributions, we selected the model that best matched experimental data. Evaluating turbulence model sensitivity ensures that your model accurately captures the complex interactions between turbulence & combustion Evaluate the influence of different time steps on transient combustion simulations. In a study on engine combustion, we tested various time step sizes to determine their effect on the accuracy of ignition timing & flame propagation predictions

What are the best practices for validating CFD and machine learning models for combustion applications?

Combustion is a complex physical and chemical process that involves high temperatures, turbulent flows, and multiple species. Computational fluid dynamics (CFD) is a powerful tool to simulate combustion phenomena and optimize the design and performance of engines, turbines, and reactors. However, CFD models often require high computational resources and rely on assumptions and simplifications that may introduce errors and uncertainties. Machine learning (ML) is an emerging technique that can complement CFD by providing data-driven insights, reducing computational costs, and enhancing predictive capabilities. But how can we ensure that CFD and ML models are reliable and accurate for combustion applications? In this article, we will discuss some of the best practices for validating CFD and ML models for combustion, based on the latest research and industry standards.

1 Define the validation problem

Validation is the process of assessing how well a model represents the reality of a physical system or phenomenon. Validation is essential for ensuring the quality and credibility of CFD and ML models, especially for complex and nonlinear problems like combustion. However, validation is not a straightforward task, as it involves many challenges and uncertainties. Some of the questions that need to be addressed are: What are the objectives and criteria of validation? What are the sources and types of uncertainties in the model and the data? How can we quantify and propagate the uncertainties? How can we compare and evaluate the model predictions with the experimental or operational data? How can we identify and correct the sources of errors and discrepancies?

Add your perspective

Raj Saini, PhD

Technology Specialist at R.K.S. Technology & Services® | Customized Solutions & Services | Energy Technology | R&D | CFD: AI-ML integrations | Alma Mater-IITB & IITM
Report contribution
Validating CFD and machine learning models for combustion applications involves several best practices: 1. Benchmarking: Compare results against established experimental data or well-validated models. 2. Sensitivity Analysis: Test how variations in input parameters affect outputs. 3. Cross-Validation: Use different subsets of data to train and test models, ensuring they generalize well. 4. Uncertainty Quantification: Assess and quantify the impact of uncertainties in model inputs and assumptions. 5. Validation Metrics: Employ metrics like mean absolute error, root mean square error & correlation coefficients to evaluate model performance. The above practices help ensure the accuracy and reliability of models in combustion studies.

Like

Load more contributions

2 Choose appropriate data sets

One of the key aspects of validation is to select appropriate data sets that can provide relevant and reliable information for the model evaluation. Data sets can be obtained from experiments, measurements, or other sources, depending on the availability and quality of the data. The data sets should cover a range of conditions and scenarios that are representative of the combustion problem of interest, such as different fuels, operating parameters, geometries, and regimes. The data sets should also provide sufficient details and accuracy for the model inputs and outputs, such as boundary conditions, initial conditions, flow variables, species concentrations, and flame characteristics. The data sets should be documented and standardized to facilitate the comparison and analysis of the results.

Add your perspective

Load more contributions

3 Apply suitable validation metrics

Another important aspect of validation is to apply suitable validation metrics that can measure and compare the performance and accuracy of the model predictions with the data sets. Validation metrics can be qualitative or quantitative, depending on the nature and purpose of the validation. Qualitative metrics can include visual inspection, graphical comparison, or pattern recognition, which can provide a general overview of the model behavior and identify major trends and features. Quantitative metrics can include statistical measures, error indicators, or uncertainty bounds, which can provide a numerical assessment of the model accuracy and uncertainty. Validation metrics should be chosen carefully to reflect the objectives and criteria of validation, as well as the characteristics and uncertainties of the model and the data.

Add your perspective

Load more contributions

4 Perform sensitivity analysis

Sensitivity analysis is a useful technique to investigate how the model predictions depend on the model inputs, parameters, or assumptions. Sensitivity analysis can help to identify the most influential factors that affect the model performance and accuracy, as well as to explore the range and variability of the model outputs. Sensitivity analysis can also help to detect and correct potential errors, biases, or inconsistencies in the model formulation, implementation, or calibration. Sensitivity analysis can be performed by varying one or more factors at a time, or by using more advanced methods such as global sensitivity analysis, which can account for the interactions and correlations among the factors.

Add your perspective

Cmdr (Dr.⁹) Reji Kurien Thomas , FRSA, MLE℠

I Empower Sectors as a Global Tech & Business Transformation Quantum Leader| Stephen Hawking Award 2024| Harvard Leader | UK House of Lord's Awardee | Fellow Royal Society | CyberSec | CCISO CISM CCNP-S CEH
Report contribution
Test different turbulence models to identify their effect on combustion predictions. In a gas turbine combustion simulation, we assessed the sensitivity of our results to various turbulence models, including k-ε, k-ω & LES. By comparing the predicted turbulent flame speeds and temperature distributions, we selected the model that best matched experimental data. Evaluating turbulence model sensitivity ensures that your model accurately captures the complex interactions between turbulence & combustion Evaluate the influence of different time steps on transient combustion simulations. In a study on engine combustion, we tested various time step sizes to determine their effect on the accuracy of ignition timing & flame propagation predictions

Like

Load more contributions

5 Compare and integrate CFD and ML models

CFD and ML models can have different strengths and weaknesses for combustion problems, depending on the level of complexity, fidelity, and generality of the models. CFD models can provide detailed and physics-based descriptions of the combustion phenomena, but they can also be computationally expensive and sensitive to the model assumptions and parameters. ML models can provide fast and data-driven predictions of the combustion behavior, but they can also be limited by the availability and quality of the data, as well as by the interpretability and robustness of the models. Therefore, it can be beneficial to compare and integrate CFD and ML models, by using methods such as coupling, hybridization, or emulation, which can combine the advantages and overcome the limitations of each model type.

Add your perspective

Load more contributions

What are the best practices for validating CFD and machine learning models for combustion applications?

1

2

3

4

5

1 Define the validation problem

2 Choose appropriate data sets

3 Apply suitable validation metrics

4 Perform sensitivity analysis

5 Compare and integrate CFD and ML models

Computational Fluid Dynamics (CFD)

Rate this article

Thanks for your feedback

More articles on Computational Fluid Dynamics (CFD)

More relevant reading