Our study concerns the supply planning of assembly systems. We consider a finished product assembled from several components. These components are themselves obtained from other components. Therefore, we interest for the two level assembly systems. We assume that the finished product demand is deterministic but the time required to supply components (lead time) are random. Our objective is to minimize the average costs consist of components holding costs and finished product backlog costs. Our optimization approach investigates the properties of dominance properties to reduce the search space of solutions as well as lower and upper bounds of the cost function. The resolution is based on a Branch and Bound algorithm, which allows us to find the optimum cost. We present an example to illustrate our approach.