Nearest neighbor difference rule-based kernel principal component analysis for fault detection in semiconductor manufacturing processes

In semiconductor manufacturing processes, effective fault detection techniques are needed to reduce scrap, increase equipment uptime, and reduce the use of test wafers. To this end, multivariate statistical fault detection methods, such as kernel principal component analysis (KPCA) and partial least squares, have drawn increasing interest in the semiconductor manufacturing industry. The KPCA has emerged as an effective method to tackle the problem of nonlinear data. However, the unique characteristics of the semiconductor processes, such as multimode batch trajectories due to product mix and process steps with variable durations, have posed some difficulties to KPCA. So as to enhance the detectability of KPCA, a nearest neighbor difference- based KPCA (NND- KPCA) is proposed in this paper. The basic idea of the proposed method is first adopting the NND rule to eliminate the multimode structure and guarantee process variables to follow multivariate Gaussian distribution; next, the conventional KPCA monitoring statistic is applied to detect system state in the NND space. The NND- KPCA does not require process knowledge and multimode modeling; it requires only a single model for multimode process monitoring. The efficiency of the proposed monitoring scheme is implemented in a simulated multimode case and in the semiconductor manufacturing processes. The experimental results indicate that the proposed method outperforms the traditional KPCA monitoring schemes.