This paper goal is to present a possible approach of using Risk Graph method to determine the necessary Safety Integrity Level (SIL) that a Safety Instrumented System (SIS) must achieve. The solution proposed by this work is using fuzzy logic. To apply this method, the Risk Graph input variables (the consequence of an unsafe dangerous event, the affected area occupancy frequency, the probability to avoid the danger and the likelihood of unintended consequences) were fuzzified using trapezoidal membership functions. The same membership function was used for the Risk Graph output variable (SIL). Using Risk Graph structure, a set of rules was developed. As inference method Mamdani was considered and as defuzzification method centroid was used. The proposed approach was validated for a simple industrial process, using risk parameter values from calibrated Risk Graph.