Monitoring aggregate warranty claims with dynamically designed CUSUM and EWMA charts

Statistical monitoring of warranty claims data using dynamic probability control limits has been shown to be effective in early detection unforeseen reliability problems that emerge at the design and manufacturing phases. As discrepancy between abnormal patterns normal pattern aggregate is usually small (especially stage), we develop two new schemes adopt CUSUM-type EWMA-type statistics, named DyCUSUM DyEWMA, respectively, better address problem. Three algorithms – is, Monte Carlo simulation, Markov chain, near-enumeration are proposed progressively determine for schemes. In particular, comparison studies show algorithm can attain a higher approximation accuracy with lower computational burden thus recommended. In-depth simulation experiments then conducted assess performance We find DyEWMA scheme superior robust various situations, whereas less could even ineffective certain cases, compared Shewhart-type counterpart. Some specific suggestions also provided facilitate implementation Improved by combining moving window approach mitigate ‘inertia’ problem further discussed. Finally, real case study presented.