A bi-criteria mixed integer linear programming model for load balancing and chemical saving in wafer cleaning processes: IE: Industrial engineering

摘要

Semiconductor manufacturing or wafer fabrication process involves multiple processing steps and is highly re-entrant, where a job may visit an equipment group more than once, at various stages of the product flow. Wafer cleaning process in semiconductor fabrication comprises multiple steps at various stages of the product flow to remove particles and oxidize organic contaminants, using different chemicals. This paper focusses on front-end of the line wafer cleaning steps. We consider J equipment groups which run I clean process steps and uses M different chemicals. Each step is qualified to run on a set of equipment groups and when recipes using different chemical run back to back on a tool-there is a chemical pre-dispense to clean and condition the nozzle. Given today's increasingly cost-conscious and competitive market, this study focuses on balancing loads on the equipment groups and minimize operating cost by reducing chemical pre-dispenses. We propose a bi-criteria mixed integer linear programming model to allocate daily demands to the equipment groups with the objectives of minimizing the maximum utilization and reducing chemical pre-dispenses. We also propose a priority based dispatching algorithm to translate the model output and implement in a dispatching system. Results show that the model reduces chemical pre-dispense resulting in significant cost saving, improved throughput due to cascading, and balanced utilization across the equipment groups.