An Observer-Based Fault Detection Scheme for Distributed Parameter Systems of Hyperbolic Type and Its Application in Paper Production Process Haiyang Hao * , Steven X. Ding * , Adel Haghani * and Shen Yin *,** * Institute for Automatic Control and Complex Systems, University of Duisburg-Essen, Bismarckstrasse 81 (BB), 47057 Duisburg, Germany (e-mail: haiyang.hao@uni-due.de) ** Research Center of Intelligent Control and Systems, Harbin Institute of Technology, Harbin, Heilongjiang 150080, China Abstract: In this paper, an observer-based fault detection scheme is developed for one special kind of distributed parameter systems (DPS) of hyperbolic type. An alternative order reduction approach is utilized to handle the infinite-dimensional nature of DPS for fault detection purpose, compared with the existing modal analysis based method which relies on the existence of eigen- decomposition of the spatial differential operator. The developed fault detection scheme is finally implemented on a simulation model of the paper drying process in paper production industry, and the achieved results demonstrate the effectiveness of the proposed scheme. Keywords: Observer-based fault detection, distributed parameter systems, order reduction, finite element method, paper drying process. 1. INTRODUCTION The problem of fault detection and diagnosis (FDD) for industrial processes is a crucial issue which has been investigated with different approaches. Among them the majority of contributions are made since 1970s in the framework of lumped parameter systems (LPS), i.e. sys- tems whose dynamic behavior is governed by ordinary differential equations (ODE), stimulated by the newly es- tablished observer theory (Ding (2008)). Generally speak- ing, these approaches can be classified into several cat- egories, including signal-based, knowledge-based, model- based and data-driven ones. The first two belong to the earliest methods which have been utilized for fault diag- nosis purpose and received great success, especially for mechanical systems (Isermann (2006)). Model-based tech- niques, which serve as powerful tool for multi-input-multi- output (MIMO) dynamic systems, have been well estab- lished (Blanke et al. (2006); Ding (2008); Gertler (1998); Isermann (2006); Patton et al. (2000)). Recently, data- driven approaches have received significant attentions both in academic research and industrial applications, focusing on large-scale complex processes where the derivation of first principle models is not feasible (Ding et al. (2011); Russell et al. (2000)). Nevertheless, FDD of another kind of frequently encountered practical engineering systems (Ucinski (2005)), i.e. distributed parameter systems (DPS) or infinite-dimensional systems whose spatiotemporal dy- namic is governed by partial differential equations (PDE), ⋆ This research has been founded by the EU FP7 research project “PAPYRUS”. The research consortium is acknowledged for the support. has received very limited attention (Demetriou et al. (2007); Ghantasala et al. (2009)), although an enormous development in the theory of control and estimate has been achieved. To solve the FDD problem of DPS, the applica- tion of LPS-based FDD techniques on the approximated DPS is of great interests. Demetriou et al. (2007) and Ghantasala et al. (2009) have developed FDD systems for DPS based on the approximated slow finite-dimensional system, which captures the dominant dynamic of original DPS assuming that the omitted infinite-dimensional fast system is stable. One prerequisite of this approach is the existence of eigen-decomposition of the spatial differential operator. In the current work, an alternative numerical ap- proach is suggested, which is supposed to be more flexible and easier to implement. For the order reduction of original DPS, the well known finite element method (FEM) is used, which is optimal in the sense of energy norm for given basis functions (Gockenbach (2002)). Then, an observer-based fault detection scheme, i.e. fault detection filter (FDF), is implemented on the approximated system, and generated residual is evaluated with the well known T 2 index. The proposed approach is validated on a simulation model developed based on the work of Berrada et al. (1992) and Berrada et al. (1997). The rest of this paper is organized as follows: in Section 2, we introduce some preliminaries on fault detection theory and formulate the detection problem. The solution to the detection problem is obtained in Section 3, Section 4 is devoted the derivation of PDE model of the paper drying process and the application of proposed fault detection scheme. Finally, the achieved results are discussed in Section 5 and Section 6 concludes this study. 8th IFAC Symposium on Fault Detection, Supervision and Safety of Technical Processes (SAFEPROCESS) August 29-31, 2012. Mexico City, Mexico 978-3-902823-09-0/12/$20.00 © 2012 IFAC 1047 10.3182/20120829-3-MX-2028.00124