Checklist for Cascade Control
Cascade control is an effective way of providing better feedback and feedforward control. The peak error in the primary loop can be reduced by more than an order of magnitude for disturbances originating in the secondary loop. The nonlinearity a control valve can be removed from the primary control loop. Flow feedforward and ratio control can become more effective. The following checklist provides some guidance on how to get this and more from cascade control.
You may not realize your single loop is already a cascade loop if you have a positioner on the valve. Positioners are closed loop controllers of actuator shaft position. Smart positioners today like the digital valve controller (DVC) have something equivalent to a PID controller. In the following discussion we will refer to a primary controller as being the higher process controller (e.g., concentration, level, pH, and temperature controller). The secondary controller in this checklist will be considered to be a flow controller.
There are other important secondary loops. The secondary loop of jacket, coil, or heat exchanger temperature compensates for coolant disturbances and nonlinearities in the process (i.e., process gain inversely proportional to process flow) and in the valve (valve gain proportional to slope of installed flow characteristic) so they do not affect a primary vessel temperature loop. Similarly, the secondary loop of static mixer pH or recirculation line pH compensates for neutralization disturbances and for nonlinearities in the process and in the valve so that they do not affect a primary vessel pH loop. Here the pH process gain is proportional to the slope of the titration curve besides being inversely proportional to process flow. Crystallizers, fermenters, bioreactors, neutralizers and reactors are examples of vessels that would benefit from a temperature to temperature or a pH to pH cascade control system. The open loop gain for the primary loop simply becomes 1.0 if the scale of the secondary loop is the same as the scale of the primary loop (e.g., 0-14 pH). There may be a secondary flow loop creating a triple cascade. If you have a positioner, you have a quadruple cascade.
Nearly all loops end up manipulating a flow. A secondary flow loop can compensate for pressure disturbances and valve nonlinearities before they affect the primary loop. To get the most benefit, the secondary loop should have a closed loop response that is five times faster than the primary loop. This relative speed can be achieved by a lambda of the secondary loop that is less than 1/5 the lambda of the primary loop.
To prevent a burst of oscillations, the output of the primary loop must not change faster than the secondary loop can respond. This can be prevented by external-reset feedback of the secondary loop PV (e.g., dynamic reset limit enabled in primary PID with PV for BKCAL_OUT set for secondary PID). For slow valves, external-reset feedback of the valve position readback is needed (e.g., dynamic reset limit enabled in secondary PID with readback PV for BKCAL_OUT set for the Analog Output). Be careful to make sure the readback of actual valve position is fast enough. Second, third, or fourth HART variables are not fast enough.
One or more integrators in the control system or in the process with valve stiction will cause limit cycling. Two or more integrators in the control system or in the process with valve backlash will cause limit cycling. Level, vessel pressure, and batch temperature are common integrating processes. Integral deadband should be added to the controllers as necessary to prevent oscillations. If the positioner has integral action enabled, an integral deadband should be considered in the positioner as well.
An enhanced PID developed for wireless with a threshold sensitivity setting to ignore noise that is enabled when the PID output is within normal range, will suppress limit cycles. An enhanced PID should also be used when the primary loop uses an at-line or offline analyzer to suppress oscillations from the excessive time interval between updates.
If the secondary flow loop gets noisy, erratic, or goes to zero at a low flow, you may need to switch from cascade flow control to direct manipulation of the control valve. This situation commonly occurs for secondary flow loops that are differential head meters and vortex meters. This problem has been noted to occur in 3 element boiler drum level control control. If you have a magmeter or Coriolis meter, you probably have enough rangeability (turndown) in the flow measurement to stay on cascade flow control. Logic is added to provide bumpless transfer between cascade control and direct manipulation of the valve.
The other considerations are simpler. So without further ado here is a checklist.
•(1) Do the primary PID output units match the secondary PID setpoint units?
•(2) Do the primary PID output limits match the secondary PID setpoint limits?
•(3) Do the secondary PID control options have SP Track PV in MAN, ROUT, etc?
•(4) Is the secondary PID structure PI on error D on PV?
•(5) Is logic needed for direct manipulation of control valve when below the flow meter rangeability?
•(6) Is the valve positioner sufficiently faster than the secondary PID?
•(7) Is the secondary PID sufficiently faster than the primary PID?
•(8) Has external-reset feedback of the secondary PV been set up properly for the primary PID?
•(9) Has integral deadband been enabled as needed in the positioner, secondary PID, & primary PID?
•(10) Has the positioner, secondary PID, & primary PID been tuned in this order?