Remarks:

•  The feedforward controller ideally does not get any feedback from the process output. Hence, it solely works on the merit of the model(s). The better a model represents the behavior of a process,the better would be the performance of a feedforward controller designed on the basis of that model. Perfect control necessitates perfect knowledge of process and disturbance models and this is practically impossible. This inturn is the main drawback of a feedforward controller.

•  The feedforward control configuration can be developed for more than one disturbance in multi-controller configuration. Any controller in that configuration would act according to the disturbance for which it is designed.

•  External characteristics of a feedforward loop are same as that of a feedback loop. The primary measurement (disturbance in case of feedforward control and process output in case of feedback control) is compared to a setpoint and the result of the comparison is used as the actuating signal for the controller. Except the controller, all other hardware elements of the feedforward control configuration such as sensor, transducer, transmitter, valves are same as that of an equivalent feedback control configuration.

•  Feedforward controller cannot be expressed in the feedback form such as P, PI and PID controllers. It is regarded as a special purpose computing machine .

•  Let us consider a system where process delay is higher than disturbance delay, eg. and ; in such case, . That means one needs to know the future values of disturbance in order to decide present control action. This is physically unrealizable controller.

 

V.I.3 Combination of Feedforward-Feedback Controller

The following table provides a comparative assessment of feedforward and feedback controllers.

Table V.1: Merits and demerits of feedforward and feedback controllers
Merits	Demerits
Feedforward controllers
Takes corrective action before the process “feels” the disturbance	Requires measurement of all disturbances affecting the system
Good for sluggish systems and/or system with large deadtime	Sensitive to variation in process parameters
Does not affect the stability of the process	Requires a “near perfect” model of the process
Feedback controllers
Does not require disturbance measurement	Acts to take corrective action after the process “feels” the disturbance
Insensitive to mild errors in modeling	Bad for sluggish systems and/or system with large deadtime
Insensitive to mild changes in process parameters	May affect the stability of the process