Control loop performance monitoring: Modeling, diagnosing and compensating stiction phenomenon in process control valves.

摘要

Controller Performance Monitoring involves the following three routine activities: (1) monitor the performance of each loop, (2) isolate poorly performing loops, and (3) diagnose the cause of poor performance. Additionally corrective actions that mitigate poor performance can also be pursued. A typical processing industry installs several hundreds of control loops to achieve the desired process performance. Recent studies reveal that only about one-third of controllers provide acceptable performance, indicating significant commercial benefits exist in diagnosing and improving the remaining two thirds of the control loops. Performance degradation in control loops result in: (i) poor set point tracking, (ii) oscillations, (iii) poor disturbance rejection, and/or (iv) high excessive final control element variation. Recent surveys indicate that 20% to 30% of all control loops oscillate due to valve problems caused by static friction (also referred as stiction). In this dissertation, loops that oscillate due to stiction are studied. An approach to model, diagnose and compensate stiction is developed. Issues that arise in modeling stiction are brought out through experimental studies. A new data-driven model for stiction is discussed. Two different techniques for diagnosing stiction using routine operating data have been proposed. The success of these approaches is built on a new technique that detects and characterizes oscillations in control loops. An integrated solution strategy to compensate stiction is provided. Stiction detection and diagnosis techniques are validated on a large industrial data set. Stiction compensation is demonstrated on a laboratory liquid level system.