Integrated Maintenance Scheduling for Semiconductor Manufacturing

Abstract

Tools in a manufacturing plant require regular maintenance over their lifespan (e.g. cleaning, calibration, safety checks) in order to keep them running smoothly. In capital intensive industries, such as semi-conductor manufacturing, the scheduling of maintenance operations on the tools used in production is a critical function. Maintenance operations can be expensive to perform, so we should only perform them when necessary. However if maintenance is delayed too long, tools may run sub-optimally or break down (thus requiring even more expensive unplanned, corrective maintenance). Furthermore a tool that is undergoing maintenance may be partly or wholly unavailable for (revenue generating) production operations.

We have developed a system to generate maintenance schedules for the IBM East Fishkill, New York 300mm semiconductor manufacturing plant. In the sections which follow, we give a description of the maintenance scheduling problem in semi-conductor manufacturing and discuss some of the challenges in solving it. We present a goal programming approach that incorporates both constraint programming and mixed-integer programming solution technologies. A system we have developed based on this approach is now in use within IBM.