Development of Soft-Sensors for Flow Pattern Detection Guided by Discrete Element Models
Online monitoring of the flow patterns of milling processes is of considerable interest to industry, as it is often desired to operate the processes in certain flow pattern modes to maximize breakage efficiency. However, direct monitoring of comminution processes is not feasible due to the hostile environment inside the mills. It is possible to study particle flow in milling processes with discrete element method (DEM) models to gain a greater understanding of the internal conditions, and hence assist the development of soft-sensors.
Exemplified by the study on a rotating drum, this paper illustrates how DEM models can be used to guide the design of soft-sensors for flow pattern detection. This approach makes use of DEM simulations to provide both ‘hidden’ internal information which is not physically measurable and data which is can be measured by physical sensors at the interface of the drum and the outside. The physical measurements of the impacts between particles and the drum wall are used to extract the feature variables, which are in turn analyzed using multivariate statistical techniques. This leads to soft-sensor models which can be used to infer the internal flow patterns.