In this thesis, we consider a tool replacement problem in a machining center producing multiple types of products using multiple types of tools. In the tool replacement problem, replacement schedules for tool types are determined. For the problem, we develop a new tool replacement policy to minimize total production costs composed of tool related costs (replacement and breakage costs), material costs, and tardiness costs. In the replacement policy, such costs are estimated through simulation to determine whether a tool should be replaced or not at each time an operation is to be started. The tool replacement policy is compared with other tool replacement policies that are used in practice or suggested in previous research through computational experiments on randomly generated test problems. Results of the computational tests show that the new replacement policy gives better performance than existing ones.