Detection of Change in Nonlinear Profiles using Kriging and Comparison with Self Organizing Clustering Method

Please enable javascript in your browser.

Seifi Shishavan, Hadi | 2017

660 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 49808 (01)
University: Sharif University of Technology
Department: Industrial Engineering
Advisor(s): Mahlooji, Hashem
Abstract:
Control Charts are the most popular monitoring tools for profiles. The time that a control chart gives an out-of-control signal is not the real time of change. The actual time of change is called the change point. This study suggests two new algorithms to find the change (shift) in parameters of nonlinear profiles. First, using ordinary Kriging method, new points are estimated. Then, with the help of Bernoulli hypothesis test, the probability of detecting the change for new points is tested. Nonlinear profiles in this study follow the exponential family of distributions; in particular, Exponential, Poison and Gaussian distribution structures are used as nonlinear profiles. The proposed Kriging algorithm has an acceptable rate of accuracy in detecting the change for the aforementioned profiles. To evaluate the performance of the proposed Kriging algorithm, a new and novel algorithm using clustering method named self-organizing map (SOM) has been proposed. Both algorithms have same input data and the outputs are somewhat the same, yet the proposed self-organizing map algorithm has a better performance than the Kriging algorithm
Keywords:
Statistical Process Control ; Nonlinear Profile ; Clustering ; Self-Organizing Map (SOM) ; Kriging Metamodel ; Parameter Change

No TOC