Abstract: Nearly 70 percent of Australian's raw coal is treated with dense medium cyclone (DMC) to produce up-to-standard saleable coal product. Therefore for maximizing product yield and profitability, achieving and maintaining an optimum DMC operating condition is of vital importance. Such being the case, an intelligent DMC system is specifically developed. With the use of the system, it is made possible to make on-line monitoring of medium densities of various process streams and yield of products, prediction of partition curves of DMC products based on the medium density values measured using an improved suspension-distribution model, and prediction of washability curves of feed coal based on measured product yield and partition curves. As evidenced by industrial-scale test, the DMC system and its sub-systems can all serves as an independent measuring means, and the measured values are quite high in accuracy; given the instantaneous feed coal ash curve, the product ash and incremental ash can be on-line predicted according to measured product partition curve and feed coal washability curve, and at a given product ash, the separation density of DMC at which the predicted product ash and incremental ash are consistent with the required values is then the optimum separating density as expected.