Detecting and Estimating the Time of Change Point in Parameters Vector of Multi-Attribute Processes

Khedmati, Majid | 2012

586 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 43254 (01)
  4. University: Sharif University of Technology
  5. Department: Industrial Engineering
  6. Advisor(s): Akhavan Niaki, Taghi
  7. Abstract:
  8. Control charts are one of the most important statistical process control tools used in monitoring processes and improving the quality by decreasing the variability of processes. In spite of various applications for multi-attribute control charts in industries and service sectors, only a few research efforts have been performed in developing this type of control charts. The developed multivariate control charts are all based on the assumption that the quality characteristics follow a multivariate Normal distribution while, in many applications the correlated quality characteristics that have to be monitored simultaneously are of attribute type and follow distributions such as multivariate Poisson and multivariate Binomial. Despite the fact that control charts are very useful in distinguishing between common and special causes of process variation and determining the out-of-control conditions, neither they determine the real time at which the process has been moved to out-of-control state (change point), nor they provide specific information about the root causes of the process variation. Moreover, in most cases the signaling time is far away from the time of the process change. Nonetheless, in monitoring a process by a control chart, when the chart generates a signal due to the out-of-control condition, process engineers start a search to identify the special causes of process variation. In these cases, knowing the time of the process change would simplify the search for identification of the special causes and as a consequence, the quality of the process can be improved sooner; before producing a considerable number of non-conforming items. In this thesis, two transformation methods are first proposed to remove the inherent skewness involved in multi-attribute processes and also to eliminate the existing correlations between the attributes and then a multi-attribute control chart is developed based on these transformations. Then, the maximum likelihood estimator (MLE) of a step change, a linear trend disturbance and an isotonic change is derived to estimate the change point in the parameters vector of multi-attribute processes. The performance of the proposed estimators is evaluated through simulation experiments in which the results show that the estimators provide accurate and precise estimates of the change point, regardless of the shift magnitude and process dimension.
  9. Keywords:
  10. Statistical Quality Control ; Change Point ; Multiattribute Processes ; Maximum Likelihood Estimation ; Skewness Raluction ; Correlation Reduction Method

 Digital Object List