Global Advanced Research Journal of Agricultural Science (GARJAS) ISSN: 2315-5094
October 2015 Vol. 4(10): pp. 711-724
Copyright © 2015 Global Advanced Research Journals

Full Length Research Paper

Nonlinear Fuzzy Robust PCA for Fault Detection of Environmental Processes

Majdi Mansouri^a, Marie-France Destain^b, Hazem Nounou^a, Mohamed Nounou^c

^a Electrical and Computer Engineering Program, Texas A&M University at Qatar, Doha, Qatar,

^b University of Liege - Gembloux Agro-Bio Tech Faculty Department of Biosystems Engineering, Gembloux, Belgium.

^c Chemical Engineering Program, Texas A&M University at Qatar, Doha, Qatar.

*Corresponding Author’s E-mail: majdi.mansouri@qatar.tamu.edu;

Tel:+974.7773.4583; Fax: +974.4423.0065.

Accepted 20 October, 2015

Abstract

Fault detection is often utilized for proper operation of environmental processes. In this paper, a nonlinear statistical fault detection using nonlinear fuzzy robust principal component analysis (NFRPCA) -based generalized likelihood ratio test (GLRT) is proposed. The objective of this work is to extend our previous work (Mourad and Bertrand-Krajewski 2002), to achieve further improvements and widen the applicability of the developed method in practice by using the NFRPCA method. It is well known that the principal components are often affected by outliers, thus may not capture the true structure of the data. Therefore data reduction based on PCA becomes unreliable if outliers are present in the data. To relieve the noise sensitivity, to obtain accurate principal components of a data, and to reduce the effective system dimension, we propose to use the nonlinear fuzzy robust principal component analysis. The objective of this paper is to combine the GLRT with NFRPCA model in an attempt to improve the performance of fault detection. GLRT-based NFRPCA is a multivariate statistical method utilized in fault detection. Here the fault detection problem is addressed so that the data are first modeled using the NFRPCA analysis algorithm and then the faults are detected using generalized likelihood ratio test. The data is collected from the crop model data in order to calculate the NFRPCA model, the thresholds and the fault detection indices (Hotelling statistic , Q statistic). It is demonstrated that the performance of faults detection can be improved by combining GLRT and NFRPCA.

Keywords:  Environmental processes, fault detection, Generalized likelihood ratio test, Nonlinear fuzzy robust, Principal component analysis.