Cutting industries produce large quantities of wastewaters with the high content of suspended solids and turbidity. The treatments of these wastewaters are important in preventing the environmental pollution. This study aimed to optimize the use of coagulation and flocculation process by the response surface methodology (RSM) method as an effective method for the abatement of turbidity-related challenges of such wastewater using coagulants of Alum and polyaluminum chloride. This study is an analytical-applied. The synthetic wastewater with a turbidity of 170 ± 2 NTU at each stage of the jar test was used. The effect of Alum and PAC application was evaluated and compared each other as a function of settling time (40–140 min), pH (2–8), speed (20–60 rpm), and the coagulant dose (0–200 mg/l). In this study, the central composite design (CCD) was utilized to experimental design, modeling, and optimization of the turbidity removal from a stone cutting wastewater. A total of 33 experiments were designed for each coagulant. Results showed that the optimal conditions for turbidity removal were settling time, initial pH, speed, coagulant dosage, 40.98, 7.47, 55.44, 163.56–56.62, 7.93, 34.55, and 199.53 for Alum and PAC, respectively. Under these optimal values of process parameters, the dye removal efficiency of 100% and 98.84% was observed for Alum and PAC, respectively. The sum of squares model showed that the quadratic model is the best option for empirical data with the highest correlation coefficient (R2 of 0.99 and 0.98 for Alum and PAC, respectively.) and P value < 0.0001 for both models. The models can adequately describe the prominent behavior of turbidity removal process. Although a high turbidity removal was observed for application of the both coagulants, in real-scale selection, each of them is related to the operator conditions. Moreover, further investigations should be considered to generalization of this work results to the real-stone cutting wastewater. Besides, the TSS challenge of such wastewater should be solved before scaling-up the results.