Robust Phase I monitoring of high-dimensional processes with individual observations

Phase I is essential in statistical process control. In the presence of outliers, traditional robust control charts designed to monitor Phase I multivariate processes with individual observations are unsuitable for high-dimensional processes, where the number of quality characteristics is greater than the number of observations in the historical data set. This paper proposes a robust T2 control chart based on the minimum regularized covariance determinant estimator to monitor high-dimensional processes with individual observations. A simulation study uses the signal probability to evaluate the performance of the proposed control chart in the absence and presence of outliers. Two examples illustrate the application of this control chart in real situations.