Expert System and Model-Based Self-Tuning Controllers

The necessity of self-tuning3 controllers is best illustrated by an example. Consider a heat exchanger that uses saturated steam to heat water that flows through its tube bundle. A simple control scheme senses the outlet water temperature and attempts to position the steam valve so that the actual water temperature equals the desired water temperature. Effects of both nonlinearities in the steam valve and changing steam pressure can be reduced by using a second control loop to control the steam flow.4 The slower-acting temperature controller now adjusts the set point of the faster-acting steam flow controller. Unfortunately, a fixed parameter temperature controller has difficulty because of the nonlinear, time-varying behavior of the process. A change in the water flow rate changes the effective delay time and heat transfer characteristics of the process. Gradual fouling of the heat exchanger tubes also changes the process dynamics over time. Good control performance at one operating condition can give way to very poor performance (overdamped or unstable response) at another operating condition.