B0373
Title: Assessment of the performances of new linear profile memory-type schemes under fixed explanatory variables
Authors: Majika Jean-Claude Malela - University of Pretoria (South Africa) [presenting]
Abstract: Classical monitoring schemes are designed to monitor one or several quality characteristics that do not depend on other variables. When there is a functional relationship between the quality characteristic and one or several exogeneous variables, the use of classical monitoring schemes is not appropriate. In this case, the literature proposes the use of simple or general profiles depending on the number of exogeneous variables. A new multivariate exponentially weighted moving average (MEWMA) scheme for general linear profiles is proposed. The performance of the new MEWMA scheme is investigated using extensive simulations, and it is compared to that of the extended versions of the MEWMA, that is, the multivariate double, triple and quadruple EWMA (MDEWMA, MTEWMA and MQEWMA) schemes for general linear profiles in terms of the zero-and steady state run-length properties. It is found that the proposed MEWMA scheme is superior to the competing schemes in many cases under the zero-state and outperforms them in the steady-state regardless of the situation. A numerical real-life example is also provided to demonstrate the practicability and superiority of the MEWMA scheme over the considered competing schemes.