Robuste Verfahren der Prozeßkontrolle

FWF

In material processing, constancy of process parameters is often of crucial importance. Corresponding process control procedures such as the Shewhart-, CUSUM-, and EWMA-chart are based on the assumption of normality and lack of serial correlation of the control variable. However, this assumption might be not realistic in practice, resulting in erroneous decisions. Our research shows how rank-based statistics can be used to design process control procedures that are robust against violations of the mentioned assumptions. Simulation results show the relative superiority of rank-based procedures relative to the corresponding observation-based techniques.