Production costs and inventory control in supply chain and are very important elements for companies. Finished products have to be produced at lowest costs. For this reason, planners must reduce average stock levels and determine optimal safety lead times. In this paper we deal with a model of a multi-levels assembly system. Several types of component and one type of finished products are considered. The demand in each period is known in advance and procurement lead times are independent random discrete variables with known probability distributions. In this study we try to determine optimal order release dates for components at last level of BOM to minimize the sum of the average inventory holding cost for components and the average backlogging and inventory holding costs for the finished product. A simulation model is proposed and a genetic algorithm is used to analyze and optimize the planned lead times.