Abstract: Correctly identifying the in-control profile set and determining the profile control parameters are the basis of profile monitoring. When the within-profile data is correlated, the existing profile parameters identification method is greatly affected by abnormal profiles. Then, a density-based identification method is developed, including profile modeling based on the linear mixed model, determining the initial in-control profile set based on the data density, identifying the in-control profile set based on successive iteration method, and calculating the in-control parameters. Then, based on the Monte Carlo simulation, the influence of density parameter and the number of profiles in initial in-control profile set is analyzed. Finally, the identification performance of the proposed method is compared with the existing methods. Simulation results show that the identification performance of the proposed density-based method is better than the existing methods.

Key words: profile monitoring, correlation, data density, statistical process control