Managing uncertainty in CFD simulations

In computational fluid dynamics (CFD), managing uncertainty is key to ensuring the reliability and accuracy of our simulations, especially in complex fluid dynamics scenarios. Given the many variables that can influence results, it's important to apply practical methods to quantify and mitigate these uncertainties. Here are some key steps I’ve explored in my work, highlighting effective techniques like Monte Carlo analysis and sensitivity analysis, alongside comparisons to experimental data.

1. Evaluate Discretization Error ??

? Mesh Refinement ???: Start with a coarse mesh and refine it in critical areas. Compare results with experimental data to ensure accuracy.

? Convergence Testing ??: Run simulations at different mesh levels. If the drag coefficient stabilizes around 0.35 and matches an experimental value of 0.34, it indicates good convergence.

2. Quantify Uncertainty ??

? Monte Carlo Analysis ??: Process: Vary input parameters (e.g., air density) to estimate uncertainty. Use a normal distribution with a mean of 1.225 kg/m3 to see how variations affect outcomes. Note: After establishing a baseline with Monte Carlo analysis, stratified sampling can be applied by dividing the ranges of your input parameters into equal intervals. For example, if examining density values, create strata for specific ranges (low, medium, high) and randomly draw points within each stratum. This ensures that each section of the space is represented, improving the accuracy of your results.

? Sensitivity Analysis ??: Process: Identify key parameters and assess how variations in these parameters affect the results.

3. Documentation and Reporting ??

? Clear Reports ??: Include methodology and results with comparisons to experimental data for transparency.

? Graphs ??: Use graphs to illustrate results, making conclusions more accessible to stakeholders.

Conclusion: Building Trust in Our Simulations ??

By systematically applying these methodologies, particularly Monte Carlo and stratified sampling analyses, and comparing results with experimental data, we can better manage uncertainty in CFD and enhance the reliability of our simulations. These steps are crucial for ensuring the robustness and relevance of our results in engineering applications. Integrating these techniques not only improves the precision of our simulations but also contributes to advancing the field toward more reliable and insightful outcomes.

#CFD #Uncertainty #Simulation #Engineering #DataAnalysis @ ALTEN MAROC