This paper proposed a new fuzzy approach to the multilevel lot-sizing problem (MLSP) with zero-one formulation which was constructed by Mcknew et al. Three coefficients in the crisp case was considered, namely the ordering cost of item j in the period i (Sij), per unit holding cost of item j (Cj) and the size of lot k. They are fuzzified into triangular fuzzy numbers respectively, and the fuzzy total cost was also generated. The signed distance is used for defuzzification to get the total cost estimate in the fuzzy sense. Because this type of fuzzification is difficult to defuzzify by the centroid, therefore another fuzzification way is considered. The Sij and Cj are fuzzified to triangular fuzzy numbers respectively, and we have the fuzzy total cost. Using both the signed distance and centroid for defuzzification, we obtained two estimates of the total cost in the fuzzy sense. Comparing these estimates, we can realize which method is better for defuzzification.