Abstract: In order to meet demand of adjusting dense medium suspension density in wide range due to change of raw coal quality, an intelligent control system for dense medium suspension density with wide domain was designed by using reverse split technology in dense medium separation process. BP neural network is used to establish liquid level prediction model of qualified medium barrel. Deviation between actual value and set value of suspension density, actual liquid level of qualified medium barrel and opening degree of shunt valve and water replenished valve are taken as input variables of the model, and predicted liquid level of qualified medium barrel is calculated through the model. According to liquid level deviation of qualified medium barrel and the density deviation, control mode switching of adding medium, steady state, density step up and density step down is realized through one-to-one multi-classification algorithm based on support vector machine. Opening degree of shunt valve, water replenished valve and water adding valve and opening time of thick medium pump and reverse shunt pump are automatically adjusted according to control mode, so as to realize wide range adjustment of density. Density fluctuation range is stable within ±0.005 g/cm3 and density adjustment time is short after application of the system.