Multiscale SPC in the Presence of Multiresolution Data

Multiscale statistical process control (MSSPC) has the ability of handling the multivariate and multiscale nature of data arising from industrial processes. However, it assumes that such data is available at a single resolution. In this paper, we present an approach for conducting MSSPC that adequately integrates data available at different resolutions (multiresolution data) and demonstrate its merits though simulated examples. The proposed approach, called MR-MSSPC (standing for multiresolution and multiscale statistical process control), proved to be able to provide a clearer definition of the regions where faults occur, when compared to alternatives based on single resolution data, without compromising detection speed.