(404a) Use of Principal Components with Parallel Coordinates for Early Detection of Compressor Surge

Authors: 
Dunia, R. - Presenter, The University of Texas at Austin


Parallel Coordinate Plots [1][2] (PCP) are used as visualization tool for multidimensional data representation into two dimensions. Its main function is to identify homogeneous patterns of data among many different variables. It is a non-projective mapping where hyper-surfaces are represented by their planar images. In contrast, Principal Component Analysis (PCA) [3][4][5] does project into a lower dimensional space in order to compact the operating information of the process. Although data projection permits to extract the most important process variations, the process may require more than three principal components. Plots that intend to illustrate more than three dimensions are not appropriate for monitoring visualization, detection and identification of process faults. 

This work proposes the use of PCP with PCA for the early detection of compressor surge [6] in industrial facilities. The proposed method of Principal Components in Parallel Coordinates (PCPC) brings the advantage of data projection and visualization in one technique.  PCPC shows to be useful in cases where more than three principal components are needed to represent the main operating variations of a process system, as it is the case of compressor surge.

References

[1] Inselberg, Alfred, Parallel Coordinates: Intelligent Multidimensional Visualization, INTELLIGENT COMPUTER GRAPHICS 2009, Studies in Computational Intelligence, 2009, Volume 240.

[2] Albazzaz, H and Wang, XZ, Historical data analysis based on plots of independent and parallel coordinates and statistical control limits, JOURNAL OF PROCESS CONTROL, 2006, Volume 16, Number  2.

[3] Dunia, R, Qin, SJ,  Edgar, TF and McAvoy, TJ, Identification of faulty sensors using principal component analysis, AICHE JOURNAL, Volume  42, Number 10.

[4] Qin S., Statistical process monitoring: basics and beyond, JOURNAL OF CHEMOMETRICS, Volume 17, Number 8.

[5] Westerhuis J., Kourti T. and MacGregor J., Analysis of multiblock and hierarchical PCA and PLS models, JOURNAL OF CHEMOMETRICS, Volume 12, Number 5.

[6] Hafaifa, A., Laaouad, F. and Laroussi, K, A Numerical Structural Approach to Surge Detection and Isolation in Compression Systems Using Fuzzy Logic Controller, SO INTERNATIONAL JOURNAL OF CONTROL AUTOMATION AND SYSTEMS, Volume 9, Issue 1.