«

»

Jul
08

Adaptive Tuning – Models Identified

The basis of adaptive tuning is the identification of the process step response based on operator changes in the loop operation. For example, when a control loop is in Automatic mode, changes made in the setpoint can trigger process model identification. If the loop mode has been changed to Manual, then changes by the operator in the loop output may trigger process model identification. When model identification is triggered, the value of the controlled parameter, setpoint, and control output may be automatically captured and analyzed to determine the process characterization. Self-regulating processes are characterized by process gain, deadtime, and time constant and integrating procesess by integrating gain and deadtime. As new process models are identified, they may be show in the on-line adaptive control interface. The consistency of the process model that has been identified can be graphically indicated by an icon in the interface,.

During the field trial of DeltaV adaptive tuning we found it useful to collect and examine the process models to see how they changed with process operating conditions. Based on this experience, the product was designed to automatically save the last 200 process models that are identified for each control loop. When the Model view is selected, the models that have been identified are listed along with the time of the setpoint or output change that triggered model identification. Also, the option is provided to plot some or all of the identified models against the selected source of process change, which is known as the state parameter. For example, in many cases the source of process change is associated with valve position and thus is one of the standard plot options, as illustrated below.

To achieve best performance for all operating conditions, it is necessary to change the controller tuning as a function of the state parameter. As described in Chapter 5 of Advanced Control Foundation – Tools, Techniques and Applications, when the adaptive control is enabled, the state parameter may be broken into as many as five regions. For each region the user can select the model or average of multiple models in that region to be the approved model for that region. Also, upper and lower limits can be specified in each region for each model parameter. The Adaptive Control view can be selected to view and approve a model and model limits for each region and the current value of the state parameter. Also, the manner in which adaptive control is used can be selected from this view. The options are:

  • Off – the tuning set in the PID block is always used.
  • Partial – the tuning used in control is based on the approved model for the current value of the state parameter and the selected tuning rule.
  • Full – on transition of the state parameter from one region to another, the tuning is set based on the approved model for the new state and the tuning rule. While operating in this region, the model used to determine tuning will be adapted within the model parameter limits based on new models that are identified for that state.

Through the use of partial or full adaptive control, good control performance may be achieved for all operating conditions.