Feedforward Feedback Control Standard

There is no standard for feed forwards. There are two main ways of implementing feed forward. One is to have a feed forward gain each of the derivatives of the SP. The other method is to have an exaggerated SP. This is kind of like holding a carrot out in front of a donkey. You make an exaggerate motion with the carrot so the donkey walks exactly where you want the donkey to step.

If you only care about regulating the position of a control valve then you shouldn't need feed forwards UNLESS you need to precisely control the rate of how that valve changes position.



Peter Nachtwey
Delta Computer Systems

Peter's response is for feedforward that is used for setpoint tracking, not feedforward that is used for disturbance rejection. That follows from his knowledge of motion control systems that are trying to track a changing setpoint.

For disturbance rejection in process control, i.e., regulating a variable to a constant setpoint, feedforward is used differently. Given a known output disturbance transfer function G_d(s) that relates the disturbance input to the controlled output variable, and a known plant transfer function G_p(s), the feedforward controller transfer function C_FF(s) for disturbance rejection would be:

C_FF(s) = -G_d(s)/G_p(s)

The measured disturbance is passed through this transfer function and the resulting signal is added to the output of the feedback controller to make the total control signal sent to the plant.

There are two methods typically used to find the C_FF, both deriving from the above equation. If first-order transfer function approximations (without dead time) are known for the disturbance and the plant, the resulting controller is a lead-lag controller:

C_FF(s) = [-K_d(τ_ps+1)]/[-K_p(τ_ds+1)]

The lead-lag function is available in most commercial controllers like PLCs or DCSs. However, for many people it's too complicated to find both transfer function models, so an approximation is used that just requires the steady-state gains to be known for the disturbance and plant functions:

C_FF = -K_d/-K_p

This works fine for most applications, as the feedback controller will "trim" any remaining errors that the feedforward controller doesn't remove from disturbances.


xnuke
"Live and act within the limit of your knowledge and keep expanding it to the limit of your life." Ayn Rand, Atlas Shrugged.
Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.