Computer simulation of ball mill cicuits has been found to be very valuable for optimisation where existing plant data can be used to calibrate the mathematical models used. However, as by definition no existing plant is available for calibration in greenfield scenarios, the use of simulation in ball mill circuit design has been limited. Running a pilot ball mill circuit can overcome this problem but this requires a reasonably large sample and therefore can be costly. In addition a valid scale-lip procedure is required. To overcome these limitations in using modelling and simulation for designing ball mill circuits, a programme of research was initiated at the JKMRC in which it was proposed to use laboratory ball mill test results to calibrate a suitable ball mill mathematical model for use in predicting full scale plant performance. To validate the technique, laboratory tests were conducted on a range of ore samples from existing plants, whose ball mill circuits were surveyed. The model, which was derived from the laboratory test results was then used to predict the full scale plants＇ performance. This paper describes the technique used and gives initial results from the comparison between the model predictions and the results obtained from the full scale plants.