On-Demand Tuning

During the commissioning of feedback control based on the PID algorithm, the PID tuning parameters (that is, the proportional, integral, and derivative gain) must be set to specific values to achieve the best controller response to setpoint and disturbance input changes. To minimize process variations and the response time to setpoint and disturbance input changes while providing stable operation, the PID tuning should be based on the observed process gain and process dynamics for each control loop. The instrumentation engineer involved in plant commissioning may not have correctly set the loop tuning, or the process operating conditions may have changed, and as a result plant operation does not achieve maximum efficiency. Fortunately, most distributed control systems (DCS) sold today include on-demand tuning support that can be used to automatically establish the correct loop tuning.

In most cases the on-demand tuning capability of a DCS, also known as auto-tuning capability, is based on the identification of the process model that matches the observed process response for a step change in a manipulated input. To identify the process response, it is necessary to test the process by initiating a change in the manipulated process input. As addressed in Chapter 4 of Advanced Control Foundation – Tools, Techniques, and Applications, the manner in which this is done varies with the on-demand identification technique used by the DCS manufacturer. The size of the change introduced in the manipulated process input must be large enough to easily distinguish the process response from any process noise that may be present. For a self-regulating process, the process response is most often approximated as first order-plus-deadtime, that is, the response is characterized by gain, time constant, and deadtime as illustrated below.

The response of an integrating process can be characterized by the process integrating gain and deadtime as illustrated below.

Based on the identified response, tuning rules are applied to determine the recommended PID tuning.