공정 평균과 분산을 동시에 모니터링하는 autoencoder 절차의 성능

초록

The autoencoder, an unsupervised learning method based on deep learning, has recently gained attention for its effectiveness in detecting abnormal signals in process monitoring. This study compares the autoencoder's performance with traditional monitoring charts, including the $\bar{X}$-$S$, Max, and MaxEWMA charts. Simulations are performed under scenarios involving mean and/or variance shifts in normally distributed processes. Key performance metrics, such as Average Run Length (ARL) and First-to-Signal criterion, are used to evaluate their efficiency. The results highlight the strengths and limitations of each procedure, offering valuable insights into their suitability for various process monitoring applications.

키워드