Abstract
Laminar flow has been characterized for its smooth and linear behavior concerning its flow. When a non-linearity is introduced, the flow becomes turbulent. Visually it is possible to distinguish between laminar and turbulent flow by identifying areas in the flow that become chaotic or non-smooth. We used a Convolutional Neural Network (CNN) trained on labeled laminar and turbulent images to automate the process. This form of machine learning is called supervised machine learning. The training data set is labeled for the neural network to identify distinguishing parameters that make them different. In this study, we take simulated contours from a flow past cylinder simulation, train it, and test it to see how effective the CNN is at distinguishing between the laminar and turbulent flow.
