Data-based Fault-tolerant Control of the Semiconductor Manufacturing Process based on K-Nearest Neighbor Nonparametric Regression

Run-to-run (R2R) control is the most commonly method in semiconductor manufacturing process. Generally, it is based on mathematical model, but for the complexity of the practical manufacturing process, it is difficult to set up the mechanical process model. This paper presents a data-based fault tolerant approach. Taking the disturbance and the fault into account, it adopts a large amount of historical data to predict the output of the single-product and multi-products semiconductor manufacturing process by the K-nearest neighbor (K-NN) nonparametric regression method. Then fault detection is achieved and a alarm is given, furthermore the traditional exponent weight moving average (EWMA) controller is improved to achieve fault-tolerant control. The results of simulation show that the approach is effective.