Multi-objective Fault Monitoring for Semiconductor Manufacturing Process with DEWMA Run-to-Run Controller

Abstract

Double exponentially weighted moving average (DEWMA) is a popular algorithm to handle the drifted disturbance in run-to-run (RtR) control of semiconductor manufacturing process. However, due to the varying environment and the equipment aging, there are faults in the process. In this paper, a multi-objective monitoring approach is proposed to monitor the Semiconductor Manufacturing Process with DEWMA run-to-run Controller. An autoregressive and moving average (ARMA) model is set up to represent a drifted semiconductor manufacturing process with DEWMA controller. A recursive extended least-squares (RELS) algorithm is used to identify the coefficients of ARMA model. The proposed multi-objective monitoring approach indices are controller performance, system stability and the coefficients of the ARMA model. Finally, the faults are detected by applying SPC on these multi-object indices instead of the process data. Our proposed approach not only can be used to monitor the non-stationary drifted process, but also can reduce the miss rate. The simulation results demonstrate that the proposed approach is effective for fault detection in general semiconductor manufacturing process.