Abstract
The assumptions made to estimate confidence limits for control charts in MSPC based on asymptotic statistical distributions do not always hold. This mismatch can cause the limits to be unrealistic in both an optimistic (limits estimated too wide for the process under monitoring, not detecting true process upsets) and pessimistic (too narrow, raising false alarms) direction. New bootstrap confidence limits based on prediction are suggested for the control charts in different online PCA-based batch MSPC strategies as an alternative to asymptotic methods. Performance of the bootstrap and asymptotic confidence limits for five commonly applied methods are compared based on Overall Type I and II errors. A simulated bio-production batch process will be used to evaluate the theoretical performance of different monitoring schemes while the everyday potential is assessed by a real data industrial batch process dataset.
| Original language | English |
|---|---|
| Journal | Chemometrics and Intelligent Laboratory Systems |
| Volume | 127 |
| Pages (from-to) | 102-111 |
| Number of pages | 10 |
| ISSN | 0169-7439 |
| DOIs | |
| Publication status | Published - 15 Aug 2013 |