Abstract: Surface tension is one of the major parameters for mineral wool production, and it may influence burdening and processing parameter selection. The surface tension of melts for mineral wool production using blast furnace slag as a major material was systematically investigated by experiment measurements and model forecasting. Firstly, a series of surface tension values of SiO₂ (40%-60%)-Al₂O₃ (5%-20%)-CaO(20%-30%)-MgO(5%) quaternary systems were measured, and they showed in the range of 350 to 500 m·m^-1. Then in combination with data from reports in literature, an artificial neural network (ANN) model was constructed to calculate the surface tension of melts in an extension system of SiO₂ (35%-60%)-Al₂O₃ (5%-20%)-CaO(20%-45%)-MgO(0-10%). The average error of the developed model is 9.32%, proving a higher accuracy for predicting the surface tension of those melts in the extension system.