> \pJava Excel API v2.6 Ba==h\:#8X@"1Arial1Arial1Arial1Arial + ) , * `DC,title[subtitle]title[alternative]contributor[author]contributor[alternativeName]contributor[advisor]contributor[department] publisherdate[issued]
language[iso]keywords[keyword]abstractidentifier[uri]identifier[url]Cho, Jun-Hop8|Y Y$The Graduate School, Ajou University2007-08kor7ICA;
PCA;
ɷ ȨȲ0;
Process Monitoring;
Hotelling T;41D tǩ\ Hotelling T ĳ Ϯ ɷ Ȳ0<\ T. 췘 \ ų1t| 0t H, t| tǩ\ Ȳ0t $t PCA| tǩ\ Ȳ0D X . X̹ D P Ȳ0D Ĭ<\ DP X 1D DP\ l t踴 JX.
l pt0| 1X iD $\ i X Ť Ȳ0t T @ 1D tǔ XD t DP, \.
l \ pt0є pt0@ D\ ɷ 0t tȬXՔ pt0@ t`<\ ιt ɷ ܭ| 0t pt0| 1\. pt0є 2Ĭ\ t踴 pt0\ 3X x @ 2X \ t踴.
t iD X 2 $| !\. i@ 1) XD L, 2) |ǀ XD L, 3)pt0 0 tX ȴLD L, 4)pt0X t XD L t 4X tiD \. iX T֔ 3<\ 10 X pt0ȹ 120 D \.
1 $X PCA } @ | . 2 $X |ǀ XD LŔ ICA x Ĭt @ 1D XD L@ TֈD LŔ ARx ICA } @ 1D ܭ PCA @ 1D . tX ȴLD Ɣ ARx@ PCA, ܭӔ ICA @ 1D .The objective of this research is to compare and analyze performance of PCA and ICA process monitoring techniques. Formerly, Hotelling's T2 control chart based on PCA has been used for multivariate statistical process monitoring. Recently, ICA based monitoring is introduced for process control monitoring, especially for chemical processes. However, there have been no researches conducted to compare the performance of PCA and ICA process monitoring.
In this research, simulation is carried out to generate process data and analyze the behaviors of two monitoring techniques. Two different cases are considered: in first case, data are generated by AR model and in second case, by multivariate normal distribution. The process is assumed to have three input (control) variables and two output characteristics. To find out type II error in the given process, four out-of-control cases are considered: a) all variables are changed b) some variables are changed c) sudden change in data, and d) change in variance.
The results are obtained after ten simulation test in the range of three degrees: small change, standard change, and huge change. Which shows performance of ICA is better than PCA in both AR and normal model, in case a). Similarly, in case b) and d) ICA little outperform PCA in AR model, whereas, PCA is better than ICA in normal model. As in case (c), ICA is more preferable to PCA in normal model; however, PCA is superior to ICA in AR model./https://dspace.ajou.ac.kr/handle/2018.oak/17043Whttp://dcoll.ajou.ac.kr:9080/dcollection/jsp/common/DcLoOrgPer.jsp?sItemId=000000002678-2T[})"D?adw "Km\~eA
c -
R
dMbP?_*+%" ,,??U
>@
Root EntryWorkbookSummaryInformation(DocumentSummaryInformation8