Machine Learning
Machine Learning algorithms are developing fast and provide us with new tools to enhance accuracy, improve efficiency and automate CFD solvers, see our recent review (available online).
We are working on a variety of topics involving Machine Learning including Unsupervised Clustering, Neural Networks, Autoencoders and Deep Reinforcement Learning:
Unsupervised clustering to detect regions that necessitate mesh adaptation (here the wind turbine wake is adapted using unsupervised clustering)
Reinforcement learning to automate CFD simulations (e.g., dynamic mesh adaptation)
Neural Networks (DNN, CNN, LSTM) to accelerate high order solvers
Variational autoencoders to enhance optimisation algorithms (in this work to reduce aeroacoustic emissions)
