A robust meta‐method for interpreting the out‐of‐control signal of multivariate control charts using artificial neural networks

Multivariate control charts are an effective mean to identify an out‐of‐control process in several (industrial or non‐industrial) fields, where the quality depends on many related variables. However, the main shortcoming of these charts is that they fail to indicate which measured variable or variables has or have shifted. In order to address this issue, several alternative analytical approaches that aim to diagnose the responsible variable or variables for the out‐of‐control signal and help identify aberrant variables have been proposed. However, there is no particular method that can be considered as panacea, since its performance depends on several parameters, such as the correlation among the variables, and so forth. In this paper, a meta‐method is proposed that combines the results of several well‐known analytical methods in order to identify robustly the out‐of‐control variables. The obtained results show that the proposed meta‐method achieves high performance and it is extremely robust under different scenarios of out‐of‐control processes.