An integrated supervised learning solution for monitoring process mean vector

When an out-of-control condition is detected by a control chart, a search begins to identify and eliminate the source(s) of the signal. Identification of the time when a process first changed is an important step in root cause analysis which helps a process engineer to eliminate the source(s) of assignable cause effectively. The time when a change takes place in the process is referred to as the change point. In multivariate environment, since there is more than one variable involved, then root cause analysis is relatively harder compared to the case of univariate because it is not clear exactly which variable has contributed to the out-of-control condition and in what direction its mean has shifted. Hence, a procedure that identifies the change point, performs diagnostic analysis, and specifies the direction of the shift in the mean of the contributing variable(s) all simultaneously could help to conduct root cause analysis effectively. Although different multivariate methods exist in the literature that allow to either estimate change point in the process mean vector or identify the contributing variables leading to the out-of-control condition, but in this research, an integrated supervised learning solution is proposed, which helps to (1) detect of an out-of-control condition, (2) identify the change point leading to shift in the mean vector, (3) specify the variable(s) contributing to the out-of-condition, and (4) identify the direction of the shift in the mean of each contributing variable simultaneously. A real case study is used to evaluate and compare the performance of the proposed integrated approach to existing methods in the literature.