Coal Preparation » Gravity Separation
This project has shown that it is possible to make useful predictions of dense medium cyclone performance with a dynamic model of the DMC running side-by-side with the real plant. Two different DMC models were evaluated, the Wood coal washing model, and the Dunglison general purpose model. As might be expected both gave similar results, the Dunglison model is more complex and is also able to provide predictions of medium classification behaviour.
The models read operating data from the plant (medium density, feed pressure, feed tonnes) and predict cut point and Ep on a size-by-size basis at 10 second intervals. Other data such as underflow and overflow medium densities, and medium to coal ratio are also predicted. If reliable washability data is available, the models can estimate the cyclone yield and product and reject ash.
The project field work was carried out at Anglo Coal's Moranbah North CPP. The DMC models were linked to the one metre cyclones in both modules. An OPC client/server network link was used to read the data from the Moranbah North DCS. After an initial site visit to prove the concept, a second site visit was timed to coincide with a routine sampling audit undertaken by SGS. The models were run reading data and making predictions while the same cyclone was being audited.
The predicted results which were available immediately showed the cyclone tested had an overall cut point (12.5 x 0.71mm) of 1.742 and Ep of 0.033. The predicted yield was 92% and the ash 8.0% using two month old washability data. When contemporary washability data was used, the yield was 91% and the clean coal ash 7.8%. The audit results which were available two months later showed the cyclone had an overall cut point of 1.730, and an Ep of 0.031. Within experimental error these results are the same. Audited yield was 91.2% and ash 8.6%.
At present the model assumes the medium density and pressure signals are absolutely correct and that the cyclone is well maintained. Further development of this approach by making additional plant data available will improve the 'believability' of the model predictions, and allow for changes in operation that occur due to cyclone wear.