Statistical Quality Control Analysis

                                                   Statistical Quality Control Analysis

            The authors is utilizing statistical quality control analysis by implementing the concept of Dynamic Time Warping to monitor profile trajectories.  Through the use of a time warping strategy, the author is able to develop control charts that create a baseline model, align the raw profiles and then utilize the generalized likelihood ratio for monitoring that is based on profile trajectories that have been partially observed.  In the online monitoring step, the trajectory of all the incomplete profiles are aligned with in-control baseline and a charting statistic is then evaluated for the purpose of making decisions about the process’s status.

            The authors was trying to provide a solution to a major problem facing Conventional Profile Monitoring. The problem is that in CPM, profiles for process runs and products are thought to have the same length; thus statistical monitoring can’t be initiated until one obtains a complete profile. This is a major problem as in some instances a single profile may take several days to be  generated and this makes the statistical process control process time  consuming and in some cases impractical.

 The process developed two profiles; one was regarded as conforming and the other as non-conforming.   The authors utilized two control charts to monitor the profiles. The results revealed that in the conforming profile, neither chart triggered any alarm while in the nonconforming profile, an alarm was triggered by the AEWMA chart at the 141 step and the DTW chart triggered an alarm at the 142 step. The DTW chart had a one-step delay that was attributed to the AR (1) which was placed in the residual before calculation if the charts statistics.