<ul><li>A novel machine learning-based initialization method has been developed to accelerate transient computational fluid dynamics (CFD) simulations in industrial applications.</li><li>The method reduces the time-to-convergence by 50% compared to traditional uniform and potential flow-based initializations.</li><li>The study evaluated three machine learning-based initialization strategies, with two strategies recommended for general use: a hybrid method combining ML predictions with potential flow solutions, and an approach integrating ML predictions with uniform flow.</li><li>The proposed methods enable CFD solvers to achieve convergence times similar to computationally expensive steady RANS initializations, requiring only seconds of computation.</li></ul>

Accelerating Transient CFD through Machine Learning-Based Flow Initialization

Discover more