A Method for Monitoring Variance in Industrial Processes with Individual Observations

摘要

The paper considers an industrial process described by a number of physical quality characteristics, each measured with a hard sensor, with a certain error, at different operating loads and external conditions. Monitoring such a process is in general performed with the Hotelling T~2 chart on which an out of control condition is signaled when the control limit is exceeded. One assumption for validating the chart is that the covariance matrix of the process remains unchanged, and in phase II of the process this assumption needs to be verified. A method for testing the process covariance matrix is discussed and compared with other procedures. The method is illustrated with an example from a real situation. The application highlights the advantage of implementing multivariate quantitative tools in a process quality system.