LITERATURE REVIEW SUMMARY PLATFORM TITLE: AUTHOR: PUBLICATION: ISSUE, TIME & Page: TOPIC: Application of multivariate statistics (Multiway Kernel Principal Component Analysis) in chemical process...

1 answer below »

LITERATURE REVIEW SUMMARY PLATFORM



TITLE:




AUTHOR:





PUBLICATION:





ISSUE, TIME & Page:




TOPIC:

Application of multivariate statistics (Multiway Kernel Principal Component Analysis) in chemical process monitoring. Only fault detection (monitoring) issue is considered and discussed, no diagnosis concern.


SUMMARY:

In this paper, the MKPCA, in stead of the popular MPCA, is employed to indentify the nonlinear structure among the variables of the measurement data and to setup the model to monitor the future sample (batch). The reviewer believes that he main contributions of this paper include:

  1. Introducing the method of MKPCA and its applicability in identifying the nonlinear structure of the variables.

  2. Developing and testing the procedure of model constructing process monitoring

    1. Off-line analysis for constructing the MKPCA model which reflecting the Normal Operating Condition (NOC) and setting up the control limit of Hotelling T2 and Q statistics.

    2. Monitoring future batchs with the existing model and control limits.



  3. Not only mean centering but also variance scaling is taken into account for applying MKPCA to process monitoring.

  4. Comparing the performance of the MKPCA with that of the traditional MPCA in a case study.





METHOD SELECTION:




WHAT:

MKPCA, the basic idea of this method is

  • First map the input (data) space into a feature space via a nonlinear map (kernel function)

  • Apply traditional MPCA in the mapped feature space.



WHY:



  • Traditional MPCA is not capable in reflecting complex nonlinear characteristic of some industrial chemical and environmental processes.

  • “The principal curve algorithm is limited to identifying structures that exhibit additive type behavior” (Jia, Martin & Morris, 2000).

  • Principal curve involves nonlinear optimization problem




PRO/CON:




  • Pro:

    be able to represent the nonlinear characteristics.



  • Pro:

    avoid computational effort in nonlinear optimization.



  • Con:

    does not consider the reconstruction issue in the feature space.



  • Con:

    face problematic size of the kernel matrix when the sample size getting large, which is contradicting to the common sense that the larger the sample size the better.



  • Con:

    be affected significantly by the selection of the nonlinear kernel function., which means that knowledge of the characteristic of the process is critical to the performance.

005_drodazo-14nvbb24.pdf 005_bbsdazo-yieaymfs.doc
Answered Same DayDec 26, 2021

Answer To: LITERATURE REVIEW SUMMARY PLATFORM TITLE: AUTHOR: PUBLICATION: ISSUE, TIME & Page: TOPIC:...

David answered on Dec 26 2021
122 Votes
LITERATURE REVIEW SUMMARY PLATFORM
LITERATURE REVIEW SUMMARY PLATFORM
TITLE: Projections and the U2 Multivariate Control Chart
AUTHOR: George C. Runger
PUBLICATION: Journal of Quality Technology
ISSUE, TIME & Page: Vol. 28, No. 3, July 1996, 313-319
TOPIC: Developing a U2 chart that are sensitive to specific assignable causes such as mean shift that occur in a subspace, reducing the dimensionality of control problem
SUMMARY: In this paper, a multivariate Control U2 chart is employed by applying projection approach to simplify the construction of the multivariate control chart. The U2 is developed to detect an assignable cause that shifts the mean vector in s subspace of variables. A signal from a U2 chart suggested that the anticipated...
SOLUTION.PDF
SOLUTION.PDF

Answer To This Question Is Available To Download

Related Questions & Answers

More Questions »

Submit New Assignment

Copy and Paste Your Assignment Here
Have files?