Machine Learning in Computational Fluid Dynamics

Course content

Machine Learning in Computational Fluid DynamicsBy attending the lecture, students will be able to: –describe the essential steps of the finite-volume-method, starting from the mathematical problem to discretization and iterative solution, based on a 2D transport problem in a handful of sentences, equations, or sketches for each step – visualize and describe in a few sentences the underpinning problems dealt with by supervised, unsupervised, and reinforcement learning based on a given dataset – select a suitable machine learning algorithm together with the associated inputs and outputs to solve a given fluid-mechanical problem, and outline the implementation, including the CFD simulation, as a flowchart – create a CFD-based parameter study, with the parameters being selected via latin hypercube sampling – create a surrogate model for given inputs and outputs by applying regression or classification techniques to experimental or numerical flow data – analyze the spatio-temporal behavior of large, high-dimensional CFD data by means of modal decomposition and derive a reduced-order model using cluster-based network modeling – train a neural network for active flow control by means of CFD-based deep reinforcement learning.

Basics of computational fluid dynamics with OpenFOAM and machine learning with PyTorch:

• Problem 1: predicting the behavior of rising bubbles; forces acting on fluid particles, Eo-Re-Mo diagram; classification, perceptron algorithm, logistic regression, multilayer preceptron, (neural network), automatic differentiation, regularization
• Problem 2: computing the mass transfer at rising bubbles; convection-dominated transport, scale-up strategy; regression, polynomials, splines, multilayer perceptron for regression
• Problem 3: analysis of coherent structures in transonic flows with shock-boundary layer-interaction; transonic buffets at airfoils; dimensionality reduction, sparse spatial sampling, proper orthogonal decomposition, dynamic mode decomposition
• Problem 4: model-order reduction of the flow past a cylinder; clustering, Markov chains, K-Means++, cluster-based network modeling
• Problem 5: active flow control of the flow past a cylinder; passive and active flow control; reinforcement learning, actor-critic methods, proximal policy optimization2512007Mechanical Engineering