Combining Ica And Mewma For Monitoring Between- Part And Within-Part Variation Of Product Measurements  

Oleh:

Cheng, Chuen Sheng ; Chen, Pei Wen ; Ko, Huang Kuo ; Lee, Hung Ting

Jenis: Article from Proceeding
Dalam koleksi: Asian Network for Quality (ANQ) Congress 2011, Ho Chi Minh City, Vietnam, 27-30 September 2011, page 1-9.
Topik: Independent Component Analysis; Statistical Process Control; MEWMA
Fulltext: TW12.pdf (212.75KB)

Isi artikel

Independent component analysis (ICA) is a signal processing technique for transforming observed multivariate data into statistically independent components. In this paper, ICA is used to monitor within- and between-part variations in the monitoring of product measurements. We assume each variation source causes a distinct spatial variation pattern in the measurement data and might also reveal interesting temporal pattern over the data sample. The spatial variation pattern and temporal pattern caused a variation source may turn out to be the observed within- and between-part variations in the monitoring of product measurements. A new approach which combines ICA with multivariate exponentially weight moving average (MEWMA), called ICA-MEWMA, is developed to better monitor processes small shifts. In the proposed approach, independent components were first obtained by FastICA. MEWMA model is applied to the I d2 , I e2 , SPE statistics calculated using the ICA monitoring method. Various selections of dominant independent components are investigated in this research. The average run length (ARL) is used as a measure of abnormalities detection performance. An extensive comparison based on simulation study indicates that the ICA-MEWMA control scheme performs better than ICA-based charts in terms of ARL.

Opini Anda

Klik untuk menuliskan opini Anda tentang koleksi ini!

Kembali