Integrating causal models and trend analysis for process fault diagnosis

Mano Ram Maurya, Purdue University

Abstract

Quick fault detection and diagnosis is very important for effective decision making during abnormal events to safely operate process plants, to ensure safety of the plant personnel and to minimize the impact of such events on product quality. Among various models used for fault diagnosis (FD), signed digraphs (SDG) have been quite popular. The development of SDG models is time consuming and error-prone. Further, due to the qualitative nature of SDG models, SDG-based FD generates a number of spurious faults. To overcome the problem of generation of an excessive number of spurious faults, integration of SDG-based FD and qualitative trend-analysis (QTA) is proposed. This thesis deals with systematic development of signed digraphs, their analysis, and integration of SDG-based FD with QTA. Integration of QTA and PCA (principal component analysis) also has been investigated. Digraphs were originally used by Iri et al. [2] for process FD. There are mainly two drawbacks in the application of digraphs for large flowsheets. One, the development of digraphs is error-prone and time consuming, and two, a large number of spurious solutions are generated. This thesis discusses various algorithms for digraph generation and proposes suitable methodologies for digraph analysis. A framework for SDG-based analysis of flowsheets is also presented. All the methods presented are explained through case studies involving single units as well as process flow sheets. Inspired by the success of SDG-based incipient FD, an SDG-QTA based integrated framework is proposed for performing incipient FD. The proposed framework is tested on the Tennessee Eastman flowsheet. Various issues in real-time application of QTA for FD have been discussed, and a comprehensive framework has been proposed for the same. Integration of QTA with PCA has been proposed and implemented to reduce computational complexity in QTA-based FD. Thus, this thesis is in four parts that deal with (i) discussion on the existing research work in the area of FD, (ii) discussion on a systematic framework for the development and analysis of SDG, (iii) discussion on QTA-based FD, integration of QTA with SDG and integration of QTA with PCA, and (iv) summary and future recommendations, respectively.

Degree

Ph.D.

Advisors

Venkatasubramanian, Purdue University.

Subject Area

Chemical engineering

COinS