Equipment Deterioration Modeling and Causes Diagnosis in Semiconductor Manufacturing

Abstract

Semiconductor equipment, which serves as the most expensive asset in the industry, is expected to function efficiently and produce qualified wafers. As the tech-node advances, the equipment utilization becomes even more critical because any faulty events occurred to the equipment will cause enormous cost. Condition-based Monitoring (CBM) is practically used to model and monitor the equipment condition for better production efficiency. It is essential to model and monitor the equipment behavior effectively in CBM. In the prognostic aspect, equipment deterioration can be traced while unexpected equipment faults are detected and investigated in the diagnostic part. In this research, the framework of equipment deterioration modeling and monitoring for batch processes is proposed. Through using the wavelet packet decomposition to transform the temporal data into macro and micro level domains, two different types of deterioration: shifted process pattern and nonstationary noise variation, will be captured. The determinant of the correlation matrix of the decomposed signals is calculated and treated as the equipment condition. The accountable factors for the deterioration are pinpointed through a stepwise searching algorithm. The results of the case study show that the proposed methodology is capable of identifying the responsible factors and forming the equipment deterioration.