By Wade Peterson
A servo motor on the machine has developed a whine. When you troubleshoot it, you suspect the tuning needs to be adjusted. You use the servo drive's auto-tune feature, and the whine disappears.
Now, do you manually tune the system from there or verify the tuning with a step response? In other words, do you trust that your manual tuning skills are better than an auto-tune algorithm?
As with other tools, there's a fine line between using auto-tune as a job aid and using it as a single solution. I like the auto-tuning PID feature and think its a great time saver. You dont have to worry about unit conversion calculations or writing custom software. Working parameters usually are found in one tuning iteration and often can be used without additional tweaking. So what's not to like?
There are some situations where auto-tuning falls short. Loads with large inertia mismatches or changing dynamics don't lend themselves to auto-tuning. Loads like these might require a variable gain, a gain outside ideal or might not auto-tune at all.
Auto-tuning also is very system-specific and time-dependent. Whats optimal for a system now could cause instability months or years later as the equipment wears. This complicates maintaining standard software across product lines.
At CMD, we have to consider the units in a model line as a whole in order to deliver a quality product. This means we need to determine the range of good tuning parameters and select the gains best suited for standard installations.
Ironically, the advances in motion control drives have made accessing the needed information more difficult. Until a few years ago, I was used to having the whole motion control program available for viewing. If I wanted to know exactly how the feedforward was implemented, I could trace it out line by line.
If I wanted to modify the boundaries of a notch filter, the parameters were there in black and white. If I needed to verify the tuning of a drive system, I could set a step response and record the results.
But now many of these features are built into the drive controllers, and you merely have to check a box for feedforward or just supply a center frequency for a notch filter. You can take it on faith that your drive manufacturer knows what its doing, or you can find a way to verify the information. This means making assumptions about hidden system functions or writing custom software to view information.
Ten years ago, auto-tuning routines were fairly coarse. They could get your system in the ballpark of stability, but additional work was needed to get proper functionality. The algorithms have become more refined now, and fewer tuning iterations are needed with only minor adjustments to the gains.
I've noticed my tuning skills are developing some rust, and Im concerned for the next generation of engineers coming into industry with fewer opportunities to get PID tuning experience. Will tuning become one of the niche skills that only a few learn?
If we, as an industrial OEM, were to lose these skills, it would affect our products through missed ship dates, longer production timelines and improper drive/motor selection. A diminishing knowledge base affects business results. So what's the solution?
Your business has to evaluate the skill sets needed to succeed and get those skills one way or another. In the case of tuning expertise, CMD has engineers with years of experience in motion control. We also work with our drive system supplier and consult with their motion experts when we encounter a particularly challenging system. This internal/external arrangement lets engineers like me share experiences and discover new tricks and tips for getting the most from our drive systems.
Perhaps software will make manual tuning skills optional, but until auto-tuning becomes foolproof, you'll need to keep in practice. As opportunities for practice become scarce, you will need to practice smarter. Have an inexperienced engineer help you tune in a system. If youre consulting with an expert, ask questions.
Questions from beginners and answers from experts solidify your understanding and might expose details you have overlooked. You need to take an active role in maintaining your skills and avoid relying on shortcuts.
Wade Peterson is an electrical engineer with CMD, a designer and manufacturer of high-speed bag and film converting equipment and high-performance specialty pouchmaking equipment in Appleton, Wis. Email him at Wade.Peterson@cmd-corp.com.