Enhancing individual hotelling T2 control charts in detecting excessive household water consumption / Herma Mohd Hanif

• Main Author: Mohd Hanif, Herma
• Format: Thesis
• Language: English
• Published: 2015
• Subjects: Analytical methods used in the solution of physical problems; Control theory
• Online Access: https://ir.uitm.edu.my/id/eprint/15681/1/TM_HERMA%20MOHD%20HANNIF%20CS%2015_5.pdf; https://ir.uitm.edu.my/id/eprint/15681/

Control charts are categorized into two groups; i.e univariate and multivariate control charts. Hotelling T2 control chart is widely being used in the multivariate control chart. The presence of outliers will affect the accuracy of both mean and covariance matrix and thus consequently cause wider control limit, which makes it more difficult to detect abnormalities. This study investigates the performance of classical and robustified Hotelling T2 control charts. Consequently, this study proposes to replace the ordinary mean with robust means, namely the trimmed and decile means in the first phase when computing the Hotelling T2 control chart. These robust statistics are integrated in three different Hotelling T2 control charts methods, namely the Reweighted Minimum Covariance Determinant (RMCD), the Minimum Vector Variance (M W ) and the Minimum Volume Ellipsoid (MVE). Simulation study is used to assess the performance of these control charts. The real data set used is the consumer household water consumption data that is used to detect excessive water usage in a household. The data is used due to the complexity of the problem in detecting excessive water usage. A classical chart with the implementation of median, RMCD with decile mean, MVE with median and M W with mean, gives best result in each method comparison. The results of the study indicate that the Hotelling T2 control chart with M W gives a more accurate detection among all. However, each of the methods is best for certain specification of characteristics. The proposed approach in detecting suspected excessive household water usage has great potential to be implemented as a software for online monitoring of domestic excessive water usage.