Abstract In this study, a three-level Box–Behnken factorial design combined with response surface methodology (RSM) for modeling and optimizing of some operations parameter of Multi-Gravity Separator (MGS) to produce a celestite concentrate was developed. The three significant operational parameters of MGS, which are drum speed, tilt angle and shake amplitude, were varied and the results evaluated with the Box–Behnken factorial design. Second-order response functions were produced for the celestite grade and recovery of the concentrate. Taking advantage of the quadratic programming, a drum speed of 150 rpm, tilt angle of 6° and shake amplitude of 20 mm have been determined as optimum levels to achieve the maximum SrSO 4 concentrate grade of 96.91%, whereas the maximum level of grade was 95.69% in the tests conducted or predicted. In the same way, a drum speed of 250 rpm, tilt angle of 2° and shake amplitude of 10 mm has been determined as optimum levels to achieve the maximum recovery of 98.35%, whereas it was 95.83% in the tests conducted or predicted.