Integral Windup occurs when the integral controller reaches its limit and no more affecting the controlled system (feedback). In this case, the integral term continues accumulating and increasing the controller output. The integral windup problem is recognized when the feedback (error) falls down and get back into the controller effective zone. The controller signal will take longer time to fall down and no more following the feedback. To solve this problem, you have to stop the integral controller from accumulating when the controller reaches its limits. There are several anti-windup methods. For example, Back-calculation method simply limits the controller output (using its saturation values), then as long as the controller is out of range, it internally discharges the integral term using an internal feedback. You need to adjust the back-calculation gain (Kb) according to your system dynamics. One way is: Kb=1/Tt, where Tt=1/sqrt(Ti*Td), where Ti and Td are the integral and derivative time constants of the PID controller.

Hi Bashir,

Mr. Kassem has basically explained the phenomenon of reset windup. It is commonly caused by the inappropriate selection of control gains involving a controller with integral action. If the gains are very high, the controller commands the actuator to produce an output greater than its limit, and thus, the actuator becomes saturated. Because the integral action (I) does not respond as quickly as the proportional action (P), it continues to build up the error signal as long as the error has not changed sign. Consequently, a large overshoot is resulted and the response time is much longer.

The natural way to prevent a reset windup is to determine the maximum value of P gain without overdriving the actuator to saturation, based on the maximum error. Then, the I gain is designed by determining the smallest possible dominant time constant of the closed-loop system.

If the reset windup is inevitable no matter how the gains are tuned, then consider implementing an anti-windup system. There are several anti-windup techniques. So, take your time to learn.

Dear Bashir,

Previously answers explain very good windup phenomenon. Maybe, a possible solution can be funded in:

Article Stabile algorithms switching for multiple models control systems

or in

Article Multiple-models control systems - Switching solution

Best regards,

Ciprian

Thank you all.

@ Kasem, Tt=sqrt(Ti*Td).

We are interested in an anti-windup compensator that prevents the error from building up excessively in the integral term of the controller.

Dear all,

When using back calculation on a PI controller, how can I tune the back-calculation gain? I’m working in an already existing chemical plant. I don’t have any model for my process, rather I have the process already built and are able to do step changes.

Thanks in advance,

Best regards

Lukas Vikander