Fault Detection, Isolation and Remediation of Real Processes

• Author / Creator

  Arifin, B M Sirajeel

• Fault Detection, Isolation and Remediation of Real Processes",,eng,"Fault Detection, Change Detection, Leak Detection, Control Valve Stiction, Stiction-Compensation","The goal of process control and monitoring system is to operate the process plants safely, reliably and in fault-free mode. Even a small fault in the process operation or in the control logic can reduce the efficiency of the process greatly. Normal process transients and operating condition changes can affect the process monitoring and control system adversely if the system cannot identify and separate these normal condition changes from the faults. Such situations may cause a lot of false alarms or alarm-floods and operators may lose confidence in these systems. Therefore, an innovative data-driven change detection algorithm has been proposed. This novel approach will enable the operators to detect changes from the normal condition. The proposed change detection algorithm is based on the Kantorovich distance concept and it does not depend on the type of statistical distribution of the data. The method was applied successfully to the benchmark Tennessee Eastman process (TEP) to detect both abrupt and ramp type changes.

  Inherent to both the process and pipeline industries are specific faults. In this study, efficient detection and remedial algorithms have been addressed for two such faults. The first fault addressed in this research is the pipeline leak which can cause serious environmental and societal/public safety issues. Such faults can often lead to huge environmental cleanup fees, penalties plus the operation downtime and product losses and poor media coverage. The existing leak detection methods perform poorly in pinpointing the leak location when leak size is less than 10\% of the nominal flow-rate. When the leak size is less than 5\%, the existing methods take 0.5 to 1 hr or more to detect the leak. To alleviate these limitations, a novel data-driven leak detection and localization method have been proposed in this study. Unlike the model-based methods, this method does not require tedious parameter tuning. Industrial evaluation of the proposed method in real-time during regulatory leak tests at Suncor Pipeline has confirmed the efficacy of the method. The proposed method was successful in detecting and localizing leak as large as 7\% to as small as 1\%. The leak localization efficiency was comparable with the existing commercial method. The leak detection time was much faster than the existing method.

  Another fault that is commonly encountered in process industries is a sticky actuator. The control valve is the main actuating element in the control loops for both process and pipeline industries. Control valve stiction is the main cause of sustained limit cycles in most control loops and causes unwanted product variability, off-spec product, loss of energy and raw materials. So, early detection of stiction is a primary concern. A novel data-driven stiction detection scheme is proposed in this thesis based on the Kantorovich distance methodology.

  Since stiction is a physical problem, to replace or repair the valve needs an unplanned shutdown of the plant. Stiction compensation algorithm can reduce the effect of stiction till the next planned shutdown. A novel stiction compensation algorithm has been proposed in this research where a variable amplitude compensating signal is added to the control signal to move the valve stem from its sticky position. This variable amplitude compensating signal can handle uncertainty in the estimation of the stiction parameter. The proposed compensation scheme was applied successfully to a pilot plant equipped with a real sticky valve including built-in valve positioner. The method was successful for both flow and level control loops. The method can reduce the effect of the limit cycle, track dynamic setpoint changes automatically and reject disturbances with less amount of valve movement than other traditional methods.

• Subjects / Keywords

  • Fault Detection
  • Change Detection
  • Leak Detection
  • Control Valve Stiction
  • Stiction-Compensation

• Graduation date

  Fall 2018

• Type of Item

  Thesis

• Degree

  Doctor of Philosophy

• DOI

  https://doi.org/10.7939/R36D5PT0S

• License

  Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.