Skip to content

Optimizing the Semiconductor Enterprise: Accounting for the Unplanned

March 17, 2009

From the time an order hits the books until it is shipped, how do we get the best performance possible out of a semiconductor enterprise? Sophisticated concepts and tools abound to answer this question. Most tools center on bringing orders, WIP, tool states, staffing, to name a few components, into a central tool or database. This data is then simulated forward though the factory processes and the outcome evaluated. Central to getting a realistic outcome, (predicted cycle time, lot out dates, etc.), is accounting for the unplanned events. The unexpected is what destroys cycle time and on-time delivery. It comes from all parts of the enterprise: customers, tools, staff, and so on. Predicting the unplanned events is a central feature of ManSim product offerings. Mathematical distributions are chosen that best represent the availability patterns of the tools in fab, test and assembly. During a simulation run, a random number generator uses the distribution to mimic tool behavior. The ManSim model forces lots to stop and wait for wafer processing, (at down equipment).1 The resulting variation builds WIP and increases cycle time. In this way ManSim products predict unplanned tool or staffing events. In a simulation run, orders are started, lots assigned to tools for processing, material is consumed, unplanned events are generated and the finished product allocated to its order. Statistical reports are generated predicting cycle time, lot out dates and tool utilization. By accounting for unplanned capacity losses or material shortages a realistic outcome is predicted. Problem-solving what-if analysis solutions can be tested with the model to optimize cycle time and on-time delivery. The best solution can found and implemented to improve the fab metrics.

  1. Using unexpected cycle time values to manage change” 9/30/2008 By Jim Schumacher and Dick Zuanich, of ManSim Inc.
Advertisements

Comments are closed.

%d bloggers like this: