Abstract: The objective of steel-iron raw materials purchasing lot-sizing problem is to identify the purchasing quantity of each item in each period over a given horizon SO that the total sum of inventory cost and purchasing cost is minimized under considerations. The lot-sizing problems have been proved to be NP-hard. The existing optimal approaches can only solve small-sized problems. In this paper we formulate the steel-iron raw materials purchasing lot-sizing problem and propose a new approach combining column generation with GUB(the generated upper bound)to solve the problem. The computational results of a practical problem show that the approach is effective and has potential application values in making purchasing decisions.