Work work work

[tehuatzi]

I suck less than I did last week. Specifically, I'm starting to get the hang of controller tuning. PIDs are pretty simple when you meet them in a classroom, but real-world PID tuning of 3D gain maps for a $500k twin-turbocharged loadsharing gas genset application is another matter

Comments

[sjepearson03]

Rawr. That is funny. Um, I have to say, Bryan is your friend. Jeff H. is your friend. Jeff and I have always been jealous of your groups of friends. Wish you were coming with Monica today. See you soon.

[rockyrockstar]

Ha! Thanks Sarah. I was thinking we would just leave this Dave Lin punk back in the dumpster where we found him... ;)

[tehuatzi]

See, I knew it! My friends are all just waiting to dump me at the merest provocation. Oh the pain, the pain...

[j3npayn3]

I think it's funny that I know someone who talks like this. This is my favorite part:

"real-world PID tuning of 3D gain maps for a $500k twin-turbocharged loadsharing gas genset application"

[tehuatzi]

Yeah, that's my favorite part, too.

It's too bad I never really got a chance to get to know Paul.

[darceyrene]

Yeah, I laugh too, because this is not the kind of conversation I hear from many people.

What does PID stand for in your world, Dave? You don't want to know what that abbreviation means in my world...

[tehuatzi]

PID = proportional-integral-derivative, e.g. a very popular and industrially successful class of controllers with three component signals - proportional, integral, and derivative. Say the control signal we're talking about is throttle command, and that-which-is-being-controlled is engine speed; e.g. if speed drops too low, you want to open up the throttle more, right? In a PID controller, the first part of the control signal responds to the current error in engine speed, the second to the accumulated error, and the third to the current rate-of-change of the error. Without going into the underlying theory (in which I am not well-versed), these three together are surprisingly capable of controlling a system subject to all kinds of potential disturbances, even if you don't know as much as you'd like a priori about how the system or the disturbances will behave. There are much more sophisticated, newer control strategies, but they tend to require a level of quantitative characterization of system dynamics that is simply unavailable in

[darceyrene]

Aaah, I understand it so much better now. Keep, uh, doing what you do.

(my brain hurts now, thanks a lot).

[tehuatzi]

What does PID mean in your world, anyway?

[tehuatzi]

By 3D I do mean three-dimensional, as in the proportional, integral and derivative gains are not simple numbers, but interpolants off three-dimensional maps (on such things as engine speed error, engine load, and load derivative). So tuning the controller is not simply a matter of tuning single values of Kp, Ki, and Kd, but rather multidimensional maps of hundreds of values of these gains across ranges of speed error, load, etc., for the full array of possible engine load disturbances (load steps/sheds of varying magnitudes at various load levels, block unloads, fault reclosures, etc.).

And this is just the throttle feedback control we're talking about; we haven't mentioned the throttle feedforward control, let alone control of turbo bypass, engine timing, fuel metering, etc. So you can see why we don't bang these out in a couple of days.