Sparse sensor-based cylinder flow estimation using artificial neural networks

Abstract

Despite the rising use of machine learning for flow estimation problems, a standard set of flows from which estimators and predictors can be developed, compared and assessed is currently lacking within the fluid dynamics community. This work presents two challenge flow cases for the advancement of sparse sensor-based flow estimation techniques. Feedforward and long short-term memory neural networks are used in conjunction with the proper orthogonal decomposition to predict velocity fields in the wake of two cylinders in proximity using sparsely-placed pressure sensors.