抄録
Typically, a multivariate process disturbance can be monitored and revealed by the use of multivariate statistical process control (MSPC) charts. The starting time of a process disturbance occurs first, and then the MSPC signal is subsequently triggered. Although MSPC charts are useful in monitoring a multivariate process, it may still take time to trigger a signal when a disturbance has occurred. This delay would often create a situation in which the process personnel are more difficult to identify the root causes of an underlying disturbance. This study aims at developing an effective procedure to determine a starting point for a multivariate process disturbance. Using the proposed combination of support vector machine (SVM) and the Hotelling T^2 Control Chart, the starting time of a disturbance can be accurately and reliably determined.
