One of the most commonly neglected and often undetected problems in feedback control are the
effects of wind-up. Control wind-up occurs when a controller's command exceeds the physical
(saturation) limits of the system actuator, and where controller momentum is unable to immediately
respond to changes in the control error. A classic example of control wind-up, known as integrator
wind-up, is illustrated in the following diagram. If the input to this system, u, receives a sudden positive
step command, the error, e, will initially be positive as the system begins to respond to the actuator. If
the rate of integration is fast with respect to the speed of the system, the integrator output may exceed
the saturation limit of the actuator but continue to grow in size itself. When the system output finally
reaches the commanded value, the sign of the error reverses causing the integrator to begin "winding"
down. But the output of the integrator ,far beyond the operating range of the actuator, takes such a
significant amount of time to recover within the operating range of the actuator and so causes a lag in
response. This process can repeat itself as a limit cycle, or eventually converge towards the
commanded value depending on the set gain and system response.
Wind-up can occur in electronic, mechanical or software components in a control loop, and is not
limited to integrators. Wind-up can actually occur in any control element that contains memory. A first
order lag or any filter can create wind-up . If the designer permits wind-up to occur, the closed loop
system can exhibit excessive overshoot, sustained oscillations, and/or lengthy settling times.
To prevent wind-up, the operating range of control elements should be limited to the range of the
devices they are driving. This helps provide instant recovery when the control error changes sign. For
complex control algorithms, limiting specific control components in software becomes a simpler task
than with electronic controls.
There are many techniques for managing windup, known as anti-windup control mechanisms. These
mechanisms typically use the known saturation limits of the actuator and act on the memory
components in the control to limit their range relative to the non-memory components.
No matter what type of control algorithm is being used, the designer should carefully check for
wind-up. Very often a design will look good in simulation or on paper, but when implemented fails to
perform as expected. Unknowingly, the designer will attempt to fix the problem by reducing gains or
"tuning" the system. This slows integration rates to match the system speed, and can prevent wind-up
in the closed loop, but at the expensive of reduced speed in the closed loop response.