Control loop performance benchmarking of an NGL plant using achievable PID bounds

Abstract

It has been noted by many researchers and industrial practitioners that Minimum variance control benchmarking of industrial control loops underestimates performance, and does not provide practical and achievable assessment. A benchmark is needed that takes into consideration the restricted PID control structure used in the industry and the requirements for robustness and healthy actuator movement. This thesis first studies the most recent developments in achievable performance assessment by implementing these techniques using real industrial data for an NGL fractionation plant in order to examine the reliability and practicality of these methods. This study will then propose a new performance assessment technique using convolution and robust IMC tuning rules that can provide more reliable and consistent assessment. The performance assessment problem will be treated in combination with fault detection techniques to study the influence of common control loop problems on control performance.