Application of Multivariate Control Chart Techniques to Identifying Nonconforming Pallets in Automotive Assembly Plants

The Hotelling multivariate control chart and the sample generalized variance |S| are used to monitor the mean and dispersion of underbody data including the pallet information to identify the non-conforming pallets. An iterative procedure and the Gaussian mixture model (GMM) are used to rank the non-conforming pallets in the order of severity. The multivariate Hotelling T^2 test statistic with Mason-Tracy-Young (MYT) signal decomposition method are used to identify the features that are affected by the non-conforming pallets. These algorithms were implemented in Advanced Pallet Analysis module of FCA Body Shop Analysis Tool (BSAT). The identified non-conforming pallets are displayed in a scatter plot with different color for each pallet. The run chart of an affected feature confirms the nonconforming pallets by highlighting data points from the nonconforming pallet. The analysis module has been successfully used in the body shops of FCA plants. One example is presented to demonstrate the application.