Pneumatic Shift Time Calculation

[thescreensavers]

Hello,

Can anyone point me in the right direction on how I can calculate our Shift time of our Pneumatic system?

I have the cylinders bore and stroke,the diameter of the solenoids outlet with its flow coefficient.

[Kirk Feldkamp]

Are you asking how to calculate it from logged data... or are you trying to come up with some theoretical calculation of how fast you'll be able to shift?

[thescreensavers]

Originally posted by Kirk Feldkamp:
Are you asking how to calculate it from logged data... or are you trying to come up with some theoretical calculation of how fast you'll be able to shift?

Hi Kirk,

How would I begin to calculate it? I do not have logged data or anything other than what I wrote in the OP.

[Jay Lawrence]

I would imagine that to calculate shift times theoretically, you would need to know the inertia of the various components in the shifting system... Assuming you know the available force from your pneumatic cylinder you can then work out an ideal shift time. To be accurate you would need to work out the inertias for each gear change (they will be different) and have some form of repeatable clutch actuation data. Then add whatever electrical delay that is present.

If you have no logged data to compare, how can you justify a pneumatic system over a manual one? I would either buy an off-the-shelf system and test it (from memory, our custom one was good for about 100ms shift times, compared to 150ms manual (i.e. not worth the weight/complexity)) or ask other teams/manufacturers for their times

[Kirk Feldkamp]

Ultimately, this boils down to some free body diagrams. First off, do you understand the mechanisms you have inside your transmission? Basically there's a ratcheting mechanism, a 3D cam, some shift forks, and dogs. Once you know all of your masses, inertias, and estimate some friction, it's not that crazy to come up with accelerations due to your input forces.

The realities of what you're looking at go far beyond the "homework problem" that you're trying to solve. In my experience, fast shifting is usually limited by how short you can make the ignition cut. You'll find that you can make very fast shifts... sometimes. What you need, however, is reliable shifting. If you keep adding more and more force at the actuator, you will find that you will start to wear and/or break parts fairly fast. No amount of basic calculation work is going to show you this stuff. I *highly* recommend you take a more empirical approach to this problem. It's going to be a far more efficient use of your resources to do actual physical development with a system like this. Build it, test it, take data, break it, make it better, rinse, repeat. The components are easy enough to spec initially... but the devil is in the system-level integration details. Without real-world performance data, the parameters you've already got don't mean a whole lot.

-Kirk

[tgman]

Have a look at Laplace transformations. If you set up an equation to describe each component of the system in the time domain and then convert everything to the S domain it should be possible to estimate the response time of the system. As an added bonus you can also work out the damping ratio this way.

[Carrington]

Festo has charts for each of their cylinders which indicates velocity of the piston for a given load, at a stated input pressure. Likewise, they also publish response times for their solenoids.

[Drew Price]

Originally posted by tgman:
Have a look at Laplace transformations. If you set up an equation to describe each component of the system in the time domain and then convert everything to the S domain it should be possible to estimate the response time of the system. As an added bonus you can also work out the damping ratio this way.


Did you really do this??!?!? Props I guess, don't think I'd recommend this path for anyone myself, seems like a big waste of time at this level.

In an F1 car, or really fast drag car, maybe, anything else....

[jlangholzj]

Originally posted by Drew Price:
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by tgman:
Have a look at Laplace transformations. If you set up an equation to describe each component of the system in the time domain and then convert everything to the S domain it should be possible to estimate the response time of the system. As an added bonus you can also work out the damping ratio this way.


Did you really do this??!?!? Props I guess, don't think I'd recommend this path for anyone myself, seems like a big waste of time at this level.

In an F1 car, or really fast drag car, maybe, anything else.... </div></BLOCKQUOTE>

dude seriously it takes about 20 mins for a problem like that. if your lazy, simulink can even do the whole shebang for you. This is something us sparky's learn (at least i did) in linear controls. Pretty cool stuff.

[Drew Price]

Crunching through the analytical problem isn't the issue. It's getting realistic values for all of the relative inertial and friction estimations, plus the input is on a lever with a non-linear ratio, plus there is stiction in the cylinder, plus delivery pressure is probably unsteady, plus plus plus plus....

I'm not saying that analysis method is wrong. I'm saying it might be useless data if your inputs don't correlate to reality. There's nothing wrong with putting a fish scale on the shift lever with it running on the dyno and getting a close estimate of needed input force.

That's a 5 minute measurement.

We have a good quote at work though:

"If you have a difficult problem, give it to a lazy person. They'll find the simplest/easiest solution."

I've been called a lot of things, lazy is not usually among them.