Skip to main content

Predicting settling behaviour of asymmetric particle using image analysis and machine learning

Academic lead
Prof Andrew Bayly, School of Chemical and Process Engineering,
Prof Jeff Peakall, School of Earth and Environment, , Dr He Wang, School of Computing,, Yi He, School of Chemical and Process Engineering,
Project themes
Environmental Flows, Industrial Processes, Underpinning Methods for Fluid Dynamics

Settling of solid particles in fluids are seen in numerous industrial and natural situations, ranging from pharmaceutical processing to sedimentation in oceans. Whilst empirical and semi-empirical correlations can be used to determine settling rate for spherical and symmetric particles, for real particles of random shape, and especially asymmetric particles and those with high aspect ratios, correlations are poor.  There is therefore a need to develop better approaches for these systems.

This project will look to develop predictive models of settling behaviour by using machine learning approaches based on the shape of the particles captured by image analysis.   The initial focus of the project will be the behaviour of single particles, and will involve: 1: developing and apply an imaging system to capture settling dynamics of settling particle, e.g velocity and rotation.  2: developing techniques and algorithms to capture shape and extract features to link to settling behaviour. 3: developing machine learning to link shape to dynamics.