Robust semiconductor production planning under yield uncertainty

Uncertainty throughout the semiconductor manufacturing process is both dependable yet inevitable. We present a robust optimization approach to production planning under uncertainty in semiconductor manufacturing. A sensitivity analysis was completed with a representative industrial dataset to verify those uncertain aspects that most affect our model: process yield out of wafer fabrication and test vendors. As the manufacturer can utilize multiple vendors for these steps, a case study was performed using actual industry data in which the number of vendors experiencing yield uncertainty was varied. Based on results, uncertainty in yield resulted in both higher first month costs and higher costs over the planning horizon; uncertainty at test results in greater cost increases than uncertainty at wafer fabrication, due to uncertainty occurring later in the supply chain. However, specific to our data, the cost of a robust solution is minimal when compared to demand lost under a deterministic solution.