Improving Bivariate Hotelling's T~2 Chart Using New Robust Estimators (WMOM with Q_n) And Bootstrap Data

摘要

Control charts monitor product characteristics, and are vital in product manufacturing. Different product characteristics are measured using different types of control charts. Hoteling T~2 chart is a type of control chart which can be used for most sensitive measures, for instance, sample covariance matrix and sample mean vector. Nonetheless, outliers eliminate the usability of the chart. Hence, this study proposes an innovative control chart that solves the problems linked to other control chart. In place of covariance matrix, a robust scale estimator Q_n is employed in the proposed chart, while winsorized modified one step M-estimator (WMOM) is employed, replacing the sample mean vector utilizing bootstrap sample data. The initial sample is obtained from the standard normal distribution. Simulation and real time data are used in results deduction. Probability of detection outliers and false alarm are the two major measurements of performance used. The performance of robust chart is proven to supersede that of the conventional control chart performance.