Inventory control is one of the important issues in supply chain management. The present study deals with designing and comparing a multi-level multi-product inventory simulation model in Iran steel industries. The divergent supply chain network model is considered with several final products, several middle products and one primary product. The purpose is to minimize cost function by maintaining the minimum level of service offering for each facilitation that is measured by means of fill rate. It is tried in the proposed model to achieve a local optimal point by having a possible point and second-order localization of the target function and linear constraints around that point as well as the use of genetics algorithm. Since point estimations of the target function and fill rates are carried out with the help of Monte Carlo simulation, statistical hypothesis testing is employed to test the possibility and improve the responses. After validation is fulfilled, the model is implemented in a three-level network via the information of Mobarakeh Steel Company. Given that linear localization is a specific state of second-order localization, it can be expected with more confidence that the achieved point in this model is better than the linear localization state.