Somebody please check my thinking.
Somebody please check my thinking.
(OP)
I'm looking to ease out pneumatics for electronics on part of a machine I'm working on. I'm trying to get in the ballpark with my numbers so I can talk to vendors about a replacement.
It's a transfer head that shuttles back and forth a lot - so far in three months 1.7 million times. The head moves on a two rod slide I intensely dislike. The side is subject to random binding that completely circumvents any attempts to bring the head to smooth stops at the end of travel. This results in nasty slamming.
Here's what it looks like:

Here are two flavors of the same movie in case you want to get a feel for the motion. If either work for you don't bother looking at the other - they're the same.
Transfer.mov 1.7M Quicktime movie
or
Transfer.wmv 2.4M Windows Media Player movie
With that in mind I did some numbers but I'm not comfortable with them since the final number I got seems about 2 orders of magnitude less than my gut feeling predicts. Anybody see an error in here somewhere?
DATA
------------------
3 sec per cycle (back and forth)
1.5 sec per direction
22" stroke
22"/1.5sec = 14.7in/sec AVERAGE SPEED
Preferred would be 11 inches of acceleration followed by 11 inches of deceleration.
Dynamic hardware weight is approximately 32 lbs.
Mass = 32 lbs/32ft/s^2
Mass = 1 slug
1 slug = 15kg
I need to meet the cycle time of 3 seconds. That's out and back.
But we can focus on just one direction. One direction is going to take 1/2 the time or 1.5 seconds.
This boils down to accelerating half way across then decelerating the second half to a stop.
So half of a stroke is 1.5 sec / 2 = 0.75sec.
We need to make the half stroke 22inches/2 = 11 inches
To make the schedule I need to get 11" in 0.75 secs.
Lets go metric:
11in x 0.0254 = 0.28m
Bring in the kinematics:
d = distance
Vi = initial velocity
a = acceleration
t = time
d = Vi + 1/2at^2
But if the initial velocity is 0
Vi = 0 we have:
d = 1/2at^2
a = 2 x d / t^2
a = (2 x 0.28m) / (0.75s)^2
a = 0.996m/s^2
Call it 1m/s^2
F = ma
m = 15kg
a = 1m/s^2
F = 15Newtons
Work = F x d
F = 15N
d = 0.28m
Power = Work/t
t = 0.75s
P = (15N x 0.28m) / 0.75s
P = 5.6W
This number is giving me pause. A fraction of the energy used in a night-light can't be enough to accelerate 32 lbs up to a meter per second in 3/4 of a second can it?!? This thing is run by a 5hp(3700W) air compressor that runs at a high duty cycle. The lion's share of air is used by this cylinder.
Maximum speed at center of stroke (end of acceleration):
Vf = final velocity.
Vf = Vi + at
Vf = 0 + 1m/s^2 x t
Vf = 1m/s^2 x 0.75
Vf = 0.75m/s 3ft/sec Seems reasonable.
It's a transfer head that shuttles back and forth a lot - so far in three months 1.7 million times. The head moves on a two rod slide I intensely dislike. The side is subject to random binding that completely circumvents any attempts to bring the head to smooth stops at the end of travel. This results in nasty slamming.
Here's what it looks like:

Here are two flavors of the same movie in case you want to get a feel for the motion. If either work for you don't bother looking at the other - they're the same.
Transfer.mov 1.7M Quicktime movie
or
Transfer.wmv 2.4M Windows Media Player movie
With that in mind I did some numbers but I'm not comfortable with them since the final number I got seems about 2 orders of magnitude less than my gut feeling predicts. Anybody see an error in here somewhere?
DATA
------------------
3 sec per cycle (back and forth)
1.5 sec per direction
22" stroke
22"/1.5sec = 14.7in/sec AVERAGE SPEED
Preferred would be 11 inches of acceleration followed by 11 inches of deceleration.
Dynamic hardware weight is approximately 32 lbs.
Mass = 32 lbs/32ft/s^2
Mass = 1 slug
1 slug = 15kg
I need to meet the cycle time of 3 seconds. That's out and back.
But we can focus on just one direction. One direction is going to take 1/2 the time or 1.5 seconds.
This boils down to accelerating half way across then decelerating the second half to a stop.
So half of a stroke is 1.5 sec / 2 = 0.75sec.
We need to make the half stroke 22inches/2 = 11 inches
To make the schedule I need to get 11" in 0.75 secs.
Lets go metric:
11in x 0.0254 = 0.28m
Bring in the kinematics:
d = distance
Vi = initial velocity
a = acceleration
t = time
d = Vi + 1/2at^2
But if the initial velocity is 0
Vi = 0 we have:
d = 1/2at^2
a = 2 x d / t^2
a = (2 x 0.28m) / (0.75s)^2
a = 0.996m/s^2
Call it 1m/s^2
F = ma
m = 15kg
a = 1m/s^2
F = 15Newtons
Work = F x d
F = 15N
d = 0.28m
Power = Work/t
t = 0.75s
P = (15N x 0.28m) / 0.75s
P = 5.6W
This number is giving me pause. A fraction of the energy used in a night-light can't be enough to accelerate 32 lbs up to a meter per second in 3/4 of a second can it?!? This thing is run by a 5hp(3700W) air compressor that runs at a high duty cycle. The lion's share of air is used by this cylinder.
Maximum speed at center of stroke (end of acceleration):
Vf = final velocity.
Vf = Vi + at
Vf = 0 + 1m/s^2 x t
Vf = 1m/s^2 x 0.75
Vf = 0.75m/s 3ft/sec Seems reasonable.
Keith Cress
kcress - http://www.flaminsystems.com





RE: Somebody please check my thinking.
I did a quick check in metric units. Just to get a feel.
You need to get to .38 m/s in .75 seconds. The mass is 14.5 kilograms. The kinetic energy then is .5*.38^2*14.5=1.05 Joule - or Watt-seconds. Not much, even if you only have .75 seconds to bring it there. You need around 1.4 W mean power for that. And it can easily be produced by a system with your numbers even if you have a very lousy efficiency and lots of losses.
I would go for something a little more powerful, say a 15 or 20 W drive. Just to get some sturdiness and customer acceptance.
The lifting? Who does that? What do you know about losses in the system? Anyhow, wasting 3700 W on this movement is really bad engineering. Anything you come up with is a lot better than that. Or, is energy free your place?
Gunnar Englund
www.gke.org
--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.
RE: Somebody please check my thinking.
W = 5*.76^2*14.5 = 4.19 J
P = 4.19/.75 = 5.58 W <-same result as OP
The other implicit assumptions here are 100% efficiency; zero friction; zero air resistance.
RE: Somebody please check my thinking.
Gunnar Englund
www.gke.org
--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.
RE: Somebody please check my thinking.
How about a crank with either a yoke or a connecting rod?
Run at 20 RPM.
I would look for economy and reliability in the 1/8 to 1/4 HP range and then calculate or estimate the safety factor over actual work done.
Bill
--------------------
"Why not the best?"
Jimmy Carter
RE: Somebody please check my thinking.
It's possible to make a P&P using cranks and slots. They suffer from the same problem as Geneva drives; the shock loads associated with even tiny slot clearances eventually destroy them.
Better P&Ps have been made with (fairly fancy) cams.
Cams and cranks may be forbidden by safety rules in some outfits. ... justifiably so IMHO. The explanation is that people do inexplicable things in factory environments, including, inevitably, getting pieces of their anatomy jammed in machinery, whether the machinery is guarded or not. With a crank or cam mechanism driving the moving parts, you may have to disassemble parts of the machine in order to remove the, er, obstacle, so (s)he may be sent off for medical help. By contrast, driving each axis with its own cylinder means that you can manually retract each axis in turn, in any necessary order, for purposes of expeditious extrication of said detritus.
Check your customer's rules.
Mike Halloran
Pembroke Pines, FL, USA
RE: Somebody please check my thinking.
In addition to the mass of the shuttle, what about the mass of the object being moved? Friction losses need to be added in as well.
As far as the inefficiency of pneumatic systems, are you really that surprised? At the time the valve opens to fill the cylinder, said valve throttles the 90 psi air until the cylinder reaches line pressure, so there's a huge loss (potential energy of the pressurized air that gets wasted). The cylinder then moves at (hopefully) line pressure to the opposite stop, and the fill valve closes. Then a dump valve opens, draining that high-potential-energy pressurized air for no net work gained...then the whole process repeats. Yes, you could use timing to better utilize that air, but in my experience that is rarely done, and most cylinders are cycled as described above.
An electric linear actuator can replace the pneumatics. Probably more up-front cost, but much more economical in operating costs. No more safety issues (as noted by Mike) with these than with the cylinders.
As for the binding, get rid of one of the slide rods, and use one or two side rails to restrict horizontal off-axis motion, let one slide rod or a vertical rail support the weight. With two rods, any misalignment and/or deflection will give you binding issues as you've found.
RE: Somebody please check my thinking.
Just mill an oval bore in the bushings over one rod, so variations in center distance between the rods can't bind anything.
( The classy way to do that is to mount one set of bushings in/on a structure that has substantial compliance toward/away from the other rod. )
Mike Halloran
Pembroke Pines, FL, USA
RE: Somebody please check my thinking.
RyreInc; I haven't noticed you around here before. Thanks so much for jumping into my thread. Thanks for the confirmation.
waross; Several times!! Gaah! I'll keep that in mind.
Most of the linear drives and linear motors offer far more than 15N of force. Numbers like 400N so over-sizing a few X isn't a large price increase.
Mike; Thanks for naming it. I didn't know that. And the prohibition was news to me. I was reviewing a movie what this replaced and it had a huge quarter-gear running back and forth against a pinion in plane site. (shudder) This new system, if you power it down, you can manually slide anything off of body parts..
btrueblood; Fear not I included the mass of the product. The inefficiency.. Yeah I have no doubt. As is common I was brought in after everything was already easing over the lip of the elevator shaft. This machine was entirely run (via a truly torturous gear train) using a single 3/4hp motor. After air made its appearance we have the 5hp air compressor running an 80% duty cycle. I was told, on asking why the shuttle was air and not a linear actuator, "that was considered too complicated to control but now that you're supporting us it's obvious that decision was a poor one".
Thanks for confirming my rods-are-bad discovery on this. I told them this rod unit is way too precise for this application. I suggested we switch them out for greased broom handles. I'll be changing it somehow.
Keith Cress
kcress - http://www.flaminsystems.com
RE: Somebody please check my thinking.
Couple other ideas/thoughts..... 1+million strokes/3mo.... yes, some/most ballscrews are rated to go at rated load/speed "1million inches" for life so you should calculate the stroke per time too and compare to screws; gut feel is a screw is not right solution due to wear. and screws are used in the electric actuators, unless you get a belt driven one.... since I am familiar with certain brands, I can tell you possible sources to check out for either screw or belt driven electric actuators:
http://www.kollmorgen.com/en-us/products/linear-po...
but I think best idea/thought I could offer is you check out free motion sizing software. again, I know and work daily with one that I like:
http://www.kollmorgen.com/en-us/service-and-suppor...
this software lets you play all these what-ifs and try the different accels, times, etc. but TWO main thing it does is 1) prevents you from making math errors (lets you mix and match all units on the way too), & 2) makes you NOT forget extra stuff like efficiencies along the way, loads like part AND slide weight, thrust for that darn cable track, etc, so you don't forget those little details that can be big. Then it actually picks the product from a look up library too, whether it is a plain servo, servo actuator, pulley/belt system, or even linear motor solution!
Last idea is with this motion it seems you should consider a linear motor solution too.
RE: Somebody please check my thinking.
http://www.kollmorgen.com/en-us/products/motors/di...
RE: Somebody please check my thinking.
I'm wondering if changing things like btrueblood suggested so I essentially have a trolley, the type with a couple of vee rollers running along a rail with a matching vee edge. Another set on the bottom to deal with any upward forces. Then use a timing belt on a servo motor with the belt attached to the trolley. Have the motor run the belt back and forth - like a monster print head. That or the same trolley with a linear motor instead of a belt.
Thanks much.
Keith Cress
kcress - http://www.flaminsystems.com
RE: Somebody please check my thinking.
Bill
--------------------
"Why not the best?"
Jimmy Carter
RE: Somebody please check my thinking.
Another option may be to install small reservoirs ahead of each solenoid. Each reservoir would be filled through a restriction and hold enough volume at line pressure to drive the cylinder a complete stroke. You would still close the valve at the end of each stroke to avoid bleed through.
You may be able to increase the efficiency of that kludgy air system by an order of magnitude.
It sounds as if you are on track solving the jamming issues.
Bill
--------------------
"Why not the best?"
Jimmy Carter
RE: Somebody please check my thinking.
we who sell and apply servos say air cylinders are great for stuff that goes between 2 points without limit switch positioning. add a 3rd or end position that is not mechanical end of stroke position and it is time to go to servos as that historically does not work well if at all.
RE: Somebody please check my thinking.
RE: Somebody please check my thinking.
Mike thanks for the education you're providing, it's most helpful.
As for the binding. I'm perplexed myself but have my theories. If you look at the structure the slide is bolted to, it's a pretty large structure with long arms all made from quarter inch stainless. I can't imagine this big structure is perfectly square. When the slide is bolted up to it with four 1/2" bolts the slide becomes the conduit of forces as it tries to pull the structure into square.
When I first saw this thing it was crazy. When the shuttle was on its way back the air would hit it and it would sit there for a fraction of a second then it would start to move slowly without accelerating for about 4 inches. Then it would look like it was shot out of a cannon rocketing to the other side to SLAM into the stop shaking all the windows in this 1/2 acre builing. Once I saw that I pulled the air lines to run the shuttle by hand. It took everything I had - both arms to move the shuttle in those 5 inches. That meant the shuttle just sat there until the pressure built to 80psi before the shuttle could move.
I promptly unbolted three of the four bolts and double jam nutted them loose. The shuttle immediately became one-hand not too much force to move. I still feel that's too much. I continued loosening the bolt holding one end of the bottom rod and we could then slide the shuttle with one finger (still with a little effort). I think this proved my theory. It does seem that the force changes with time though. Sometimes it's stiff sometimes not.
Keith Cress
kcress - http://www.flaminsystems.com
RE: Somebody please check my thinking.
I think from all I see & and have heard so far, that spending the additional thousands going to servos of any kind will not help your basic issue.... I do not see any basic design issue using air cylinder to do what you are doing. Just my 2 cents worth.
RE: Somebody please check my thinking.
It does seem that your air cylinder is greatly over-sized. A properly designed electric mechanism would be more energy efficient than a poorly designed pneumatic one. Crank mechanisms are the most energy efficient way to create reciprocating motion.
RE: Somebody please check my thinking.
Mike Halloran
Pembroke Pines, FL, USA
RE: Somebody please check my thinking.
The mass (32 lb) will undergo a constant acceleration for 0.75 seconds until 11 inches is reached
then a constant deceleration until 1.5 seconds and 22 inches is reached.
Using this the needed acceleration is.
A=(2*d)/T^2 = 22/(0.75^2) = 39.11 in/sec^2
The peak power would occur when the load is moving the fastest under the constant acceleration.
This would be at the middle with a velocity of
V=39.11(in/sec^2) * 0.75 (sec) = 29.33 in/sec
At this speed the force is still 32 lb.
peak power would be 32*29.33 = 938 ft*lb/sec or
Power= 938 ft*lb * 1.335 (ft*lb/joule) = 1253 joules/sec or 1.68 hp
Not entirely sure about this but is sounds much closer to something that would move the load.
RE: Somebody please check my thinking.
RE: Somebody please check my thinking.
Silly error in the prior.
V=39.11(in/sec^2) * 0.75 (sec) = 29.33 in/sec
At this speed the force is still 32 lb.Force would be 32*(39.11/(12*32)=3.26 lbf
peak power would be 3.26*2.44= 8 ft*lb/sec or
Power= 8 ft*lb * 1.335 (ft*lb/joule) = 10.68 joules/sec or 10.68 watts.
Still seems low..
RE: Somebody please check my thinking.
I see that off center piston rod! I'll have to ask what the thinking was on that. Thanks for the input.
Compositepro; Indeed this cylinder is controlled by exactly that kind of 3 pos valve. Rate is controlled by restricting the outlets. (Hydraulic input side - pneumatic output side). The pressure is the same on both sides, at least until the rod seals start leaking. This replaced a crank system - I know not why....
Mike: I'll suggest a smaller cylinder if I can get the shuttle happy.
2dye4; You are incorrectly using weight instead of mass..
btrue; I'm trying to get some feel for the cost of linear motor. I haven't yet been successful. I can't believe a linear motor would cost much more than a cylinder and slide assembly.
2dye4; Ahhhhh You caught it. And.. You're getting the same power we got and I agree, it sure looks like too little.
Keith Cress
kcress - http://www.flaminsystems.com
RE: Somebody please check my thinking.
I thought your issue was random jambing of mechanics; if it is air supply and control smooothness issues also, then changing from air to servo definitely would help that portion of it.
You are absolutely correct if you compare 2 round shafts to TWO profile rails, shafts are a LOT more forgiving. But I thought you might be able to design with ONE profile rail and hang this down from it; one rail should make misalignment issues go away, no? But may not be robust enough for your application? although I have seen quite a few single larger rails handle some mighty big offset loads.
RE: Somebody please check my thinking.
Keith Cress
kcress - http://www.flaminsystems.com