To predict real-time distribution curve of dense mediumcyclone in the Taixi coal preparation plant, a newmethod based on field data was presented. A group of uniformly designed single machine checks (sampling test of a densemediumcyclone)were carried out for modeling field operating parameters and distribution curve indexes. Then three real-time indexes could be predicted by onsite sensor values, and the present distribution curve model was able to be calculated. Predicting part of the method was also compared with the usual approximate formula method based on the proposed models. The models were proved to be acceptable as the RMS accuracy
reached 0.73%, 0.77%, and 0.90% respectively in the training set and 0.76%, 0.63%, and 0.51% for the test example. The predicting resultswere also verified to be better andmore reasonable than the approximate formulamethod.