43 www.drivesncontrols.com October 2021 PRECISION ENGINEERING AND MOTION CONTROL n U nderstanding the basics of servo system tuning is one thing, but the ultimate ambition is to strive for optimisation. Among the most common facilitating methodologies of servo system optimisation is feed-forward, but how is it best to apply this technique while simultaneously minimising the effects of any system non-linearities? The general goal when it comes to the effective feedback control of servo systems exhibiting some level of disturbance is clear: getting system-state variables such as position, velocity and force to follow the desired (optimum) trajectory. Traditionally, the first task is to perform a step response of the system and change the feed-forward gain to minimise system overshoot. Technicians then repeat the same process several times until they achieve satisfactory levels of optimisation. Although this trial-and-error method can be effective, other techniques can improve results further. For instance, it is possible (in the frequency domain) to scrutinise the frequency response to the plant and find a feed-forward with the matching inverse of that plant, thus eliminating any peak disturbance. Alternatively, without the ability to plot the inverse plant, another viable approach is to examine the closed loop and see howmuch peaking is present. Used as an approximation tool, feed-forward will minimise the peaks in a closed-loop response. If the frequency domain is not available, then making a step response in the time domain is the other option. This method is less preferable, although there are ways to perform proper measurement, rather than simply looking at the final response of the error. Some common questions that arise at this point include: n What feed-forward compensation structure is best? and n Where and when is the optimum place and time to insert it into the system? Feed-forward tuning If there were no disturbances in the plant, then there would no real reason to have feedback control: the feed-forward control would provide the appropriate current to move a servomotor, for example, meeting requirements in terms of position, velocity or some other more complicated trajectory. The reason feedback control is necessary is solely because of disturbances in the plant, along with the ability to accommodate plant ambiguities and the possibility of relocation at some point in the future. Based on this thinking, the ideal feed-forward is simply the inverse of the plant. Of course, there are some issues with this seemingly simple answer. For instance, maybe the plant is not invertible, which is often the case. And yet introducing more complexity into a plant model will only make it even more difficult to invert. Another school of thought is clearly required. So, rather than trying to achieve the exact inverse of the plant, it is preferable to think of the plant as a rigid body and match the inertia line, which will eliminate almost all of the errors. In relation to the amount of effort required, this method provides the greatest reward. Practical examples To demonstrate feed-forward control with multi-loop PID, consider a plant based on a two-mass system (flexible body). In this example, position measurement takes place using an encoder, which the technician can enter into a node (after differentiation) to Feed-forward control is a common approach to optimising servo behaviour but is a more complex topic than you might imagine. Dr Joseph Profeta, director of Aerotech’s control systems group, explains what is involved and the factors you need to consider. Tuning servo systems: feed-forward and non-linearities The latest stages for 3D metrology applications can be linear, rotary, lift or Z-axis, using high-specification motors to deliver smooth, reliable motion