Study On Statistical Quality Control Strategyin Semiconductor Manufacturing

In recent years, Run-to-Run (R2R) control strategy has attracted extensiveattention in statistical quality control (SPC) research and been widely applied tovarious real manufacturing processes. It combines statistical process control andfeedback control. It has been efficient and effective in compensating process drift runby run in the case of processing one workpiece for one run in a single stage.However, in many real manufacturing processes, products are processed batch bybatch, instead of one by one. Traditional R2R process control method cannot beapplied to this case and should be adjusted according to the characteristics of batch-based manufacturing processes. In addition, a modern manufacturing processisusually divided into multiple small independent stages because of the increasingrequirement for high efficiency and effectiveness. In each stage, there are only severalprocessing actions, but many factors affect qualities. The deviation of quality from thetargets may be compensated or amplified stage after stage. The traditional EWMAmethod may not perform well in reducing the variation in the case of multistagemanufacturing processes.The paper proposes a new batch-R2R control strategy for both the single-stageand multistage processes. The quality variation can be divided into two parts: within-batch variation and batch-to-batch (B2B) variation. The new batch-R2R controlstrategy reduces with-batch variation by clustering them according to inputcharacteristics. Meanwhile, in order to reduce B2B variation, it combines feedforwardcontrol and provides an improved exponential weighted moving average (EWMA)control framework which considers the input characteristic when modeling theprocess and updating parameters.Simulation study shows that the batch-based R2R control strategy has asignificant contribution on reducing the mean square error (MSE) of process outputmeasurements for both the single-stage and the multistage manufacturing process.