Real-time data driven maximisation of carbonatation filter bank throughput

Abstract

Volumetric throughput at the carbonatation filters at Malalane white sugar refinery was identified as a bottleneck. Scheduling filter cleaning was found to be one of the major concerns. The maximisation of the throughput via scheduling was approached by data integration and modelling. Linear regressions of the process parameters were implemented in volume space, which allows for the prediction of the remaining run duration of each filter. These predictions are continually used as an input to an algorithm that creates a schedule. The calculations are conducted in the cloud on real-time data and the predictions reported to a web-based user interface for the plant operator to use.