Agile actuator challenge

[mename]

Dear all,

I have faced a challenging problem that may be interesting for those who are in agile/precise control.

Realizing an actuation system for the following application, I appreciate any help/feedback:

• A 0.5 kg load needs to be moved up and down between positions A and B. The (vertical) distance between A and B is 1 meter. The trajectory followed for the up and down motion is free.
• The load needs to be standing still in point A for 0.2 seconds, has 0.2 seconds to move up to point B, should be standing still in point B for 0.2 seconds and has 0.2 seconds to move down again to point A.
• Specific requirements: Energy efficiency is of utmost importance and the system has to be powered via a standard net switch (230V, 16A, 50-60Hz, single phase). This motion has to be continuously repeated.

My initial approach is:

1- Regardless of the motor type, I should setup a continuously running system (no start and stop several times a second!).
2- The actuator should move the load preferably on a circular path A-->B-->A-->B... in 0.2 sec and the load should be detached from the mover at point B by a detacher (electromagnetic detach/attach mechanism?) while the mover continues to run ensuring the constant running of the motor. 0.2 sec later, the attacher (electromagnetic detach/attach mechanism?) attaches the load to the mover while the mover takes the load to point A back again, and so on.
3- The rotational moving mechanism will render unnecessary the motor to change its rotation 2-3 times a second as it does if you go with a linear actuation style (back and forth).
4- The detachment and attachment mechanism, if it works, will render unnecessary the motor to stop for 0.2 sec while the load is at point A or B. An electronically controlled timer will energize the attach/detach mechanisms at A and B.
5- For the sake of energy efficiency, a single-phase ac motor may be chosen since it is required to power the system via single-phase 230V 16A switch.
6- Or, should I design an actuation system where I control the start-stop cycle of the motor for every 0.2 sec? If so, what type of control can I employ? A PID feedback control with position/velocity sensors? I suspect any ac motor, regardless of number of phase or use of VSD, will sustain such an agile movement for long time.

What do you think?

Regards,
Maxime

 

[JimB]

What is the 0.5kg load?
What is the maximum acceleration (+ and -) it can withstand?
How accurate is the positioning required at A and B?
What is the budget? (How much money can you spend?)

JimB

 

[mename]

Hi Jim,

• The load is a lens.
• No acceleration limit is foreseen.
• Highly accurate positioning is required. (+- 1 mm error)
• Budget is no issue. Energy efficiency is important, but the capital cost is not.

Regards,
M.

 

[Tony Stewart]

Define power budget, position feedback preference. One could use a spring actuator could be unloaded in the center position then compressed and expanded to store the energy need to accelerate back and forth . from A to B and back to A.and to meet your "utmost requirement"

Rough calc.
- for acceleration a in 0.1s and -k in 0.1s to travel 0.5m in each state of +/- acceleration
- d=1/2 at² for d=0.5m , t=0.1s a=2d/t² = 100g
- F=ma = 0.5kg *100g= 50 N

There are many mechanical unknowns'
The size and fragility of 0.5kG load.
Position error is 1mm but vibration during 0.2s dwell time is undefined. 0.1mm?? or less?
Preferred axis motion only Z axis?
You basically need a smooth 1.25 Hz machine cycle
Life cycle before maintenance , given 25 million cycles per year at 100% duty cycle

 

[JimB]

Two ideas come to mind.

A linear actuator as used in 1970s hard disk drives, but they only moved about 100mm and did not have to move half a kilogram, only a few grams.

A rotary disc, driven through a slipping clutch and stopped by a disc brake mechanism. A locating pin is inserted into the disc when stopped in order to give accurate location.

All in all both of these ideas are brutal and require lots of force.

Why does the lense have to be moved?
Can you not just use two lenses, one at each location?

JimB

 

[Tony Stewart]

100g for 0.1 s means it reaches a speed of ~ 50m/s min. or 180kph .. much faster than a Tesla car ...!!

Are you serious?

 

[mename]

Why does the lense have to be moved?
Can you not just use two lenses, one at each location?

They are great questions to start with. But, for the moment the situation is taken as is. If no viable solution exists, then I will step into questioning those.

Thanks.

 

[mename]

100g for 0.1 s means it reaches a speed of ~ 50m/s min. or 180kph .. much faster than a Tesla car ...!!

Are you serious?

My rough calculations are a bit different, but yes it is fast! Expected max vertical velocity is 10 m/s though.

But how to achieve such high speed (from zero speed) with an ac motor within 0.1 sec is one of the key questions now. Similarly, stopping it within 0.1 sec?

 

[Les Jones]

I think you will need some sort of position encoder so that you can start to decelerate the arm before it reaches the target position. There is this servo system design on the web that behaves like a stepper motor so all you would have to do is send it the reqired number of pulses and a direction signal. It claims to work without overshoot. You would probably have to increase the drive capabilities of the power output stage to match the motor that you used.

Les.

 

[mename]

I think you will need some sort of position encoder so that you can start to decelerate the arm before it reaches the target position.

Les,

Agreed, a PID-like feedback control is a must.

My concern is whether I can meet the time interval or not. I can yet to see any ac motor that can go from zero to rated speed in such a short time (0.1 sec to accelerate - 0.1 sec to decelerate/stop).

Any idea on that?

 

[Les Jones]

I did not realise that there was a stipulation that it had to be powered by an AC motor. I think there are AC servo motors available but I have never worked with them an never bothered to learn about them. I think this project should be possible as several years before I retired I worked on a Sony optical disc storage library and the servos in that moved the optical disc cartridges (Which probably weighed more than 0.5Kg) the full height of the cabinet in well under 0.5 seconds. (That was about 5 feet of movement.)
It should be possible to get an idea about the power reqirement by worrking out the kinetic energy in 0.5 Kg traveling at twice it's average speed. That energy would be lost every 0.4 seconds so you can work out the power. I don't think it would be practical to try to recover that kinetic energy

Les.

 

[Tony Stewart]

My rough calculations are a bit different, but yes it is fast! Expected max vertical velocity is 10 m/s though.

But how to achieve such high speed (from zero speed) with an ac motor within 0.1 sec is one of the key questions now. Similarly, stopping it within 0.1 sec?

View attachment 104420

agreed
d=1/2 at² for d=0.5m , t=0.1s
a=2d/t² = 100g = (2*0.5)/0.1²= 1/0.01=100g and v=at = 100*0.1 =10m/s not 50m/s

As in HDD servo positioners with a 100g change in acceleration the servo must switch from Acceleration mode to Position Error servo control to provide high position dampening and minimal overshoot. This requires precision position feedback near endstops and may use motor current to control acceleration.

The mechanical details are still unknown to decide if one needs a reciprocal driving piston actuator with a flat tipped camshaft or a long linear motor.
Since a "jerk is defined as j=da/dt which affects mechanical wear directly the transition from 100g to 0g almost eliminates a mechanical reciprocal solution as the position is not sinusoidal like an engine crankshaft.

Can you shed some info on the mounting and cost of this mechanism (lens?)?

I did some Fourier Analysis to indicate the Jerk stresses and compromising these would require
extending the stop time interval to allow for setting time to reduce jerk levels( only unfiltered j is shown)

 

[BobW]

I did not realise that there was a stipulation that it had to be powered by an AC motor. I think there are AC servo motors available but I have never worked with them an never bothered to learn about them.

Yes. In fact it can be done with a standard 3 phase induction motor, and performance can be quite remarkable. Essentially, a standard 3 phase AC induction motor driven by a variable frequency drive (VFD). The VFD needs to have the motion control module, which is a readily available option for most major VFD brands, and the motor (or other part of the mechanical assembly) needs to have an encoder for position feedback.

Several years ago I worked on a similar application, except that the mass of the object was quite a bit larger, about 40kg. Because of the considerably larger mass, we were only able to get about 2 seconds from start position to end position, with an accuracy of 0.5mm. The distance travelled was about 1.2 meters. This was for moving dies into and out of position for a press operation. The drive equipment was a standard 2 HP 3 phase motor with gear reduction driving a toothed belt. The VFD was SEW Eurodrive. They claimed that for up to 3 HP, a standard induction motor with VFD performed just as well as a servo motor with servo controller. At that time (more than 10 years ago), SEW seemed to be the most advanced in motion control of standard induction motors, but nowadays just about any brand of VFD can do this. I think that moving a 0.5kg mass 1 meter in 0.2 second with this type of equipment is feasible, but would require eliminating all extraneous mass from the system.

 

[large_ghostman]

I cant find the link at the moment, But DANFOS do a motor used to turn a cosmetic machine turntable, it is ultra accurate and fast at stopping. I will go look at the motor I scavenged and see if I can find the data sheet.

 

[Triode]

I say this as someone who designs feedback control systems; I think you need a mechanical solution to pull this off efficiently.

Feedback control is good when your stopping points need to be controlled and changed. For example when they are complex or they change at run time. You need to stop in two locations with 1 degree of control, and those locations don't change.

Maybe just a good old Geneva Mechanism:

To handle the rapid acceleration I'd gear it down on the input side, with a flywheel on the fast/input end of the gearbox. Then to get your 180-degree flip, gear it up on the output side.

For precise alignment, have an alignment feature at the very end, like a spring loaded bearing roller that falls into a notch on the lens holder since there will be non-zero backlash on the gears

Yes it's complicated and the forces will be high, but so is doing this using a servo. If you use a 4 side geneva gear you only need to gear up 2:1 at the output.

All that said, I know I've used AC servos that can do this. Mitsubishi ones in my case, but I'm sure they aren't the only ones. They really aren't cheap though.

 

[BobW]

A mechanical method would no doubt be optimum, but the cost of such a mechanism could easily be orders of magnitude more expensive than a simple motor plus motion control system. For the application I mentioned above, the owner of the operation was familiar with fully mechanical versions of his equipment and said that the cost of a mechanical version would have been far more than the system that we ultimately implemented.

 

[crutschow]

What about some sort of spring resonant system.
That could be both quite simple and power efficient.
The lens could be mounted on a leaf type spring with the unloaded position half-way between positions A and B.
(It was stated that "The trajectory followed for the up and down motion is free.")
To start it would be manually moved to either end position and held by a stationary electromagnet engaging a permanent magnet on the spring next to the mirror.
The electromagnet current would then be rapidly reversed from the holding field to a give reverse field, imparting energy to the moving mass to overcome any friction as it swings to the to the other side, where it is again held by that side's electromagnet engaging the permanent magnet.

Note that the mirror will accelerate rapidly but then decelerate to near zero as it approaches the opposite side.
You would adjust the amount of reverse electromagnetic current so that the spring is given just enough added energy to reach the other side with near zero velocity.

Alternately if the electromagnet is strong enough along with a strong permanent magnet, and the friction losses are low, the electromagnetic may be able to capture the mirror even if it doesn't quite make it all the way to the opposite side.
It that case you would only need enough reverse electromagnetic current to release the permanent magnet.

The springs stiffness would, of course, have to be sufficient to move the mass from one side to the other in 0.2s.
Not sure how much stiffness that translates to.
Edit: My admittedly very suspect calculation indicates that the spring constant would be in the neighborhood of 123N/m or 8.5 lb/ft (for a massless spring).
To minimize spring mass, the spring could be made of composite material (remember cost is no object).

 

[Misterbenn]

My approach would be to use a variable frequency inverter to drive an AC permanent magnet motor and a gearing system to the actuation disk. At point be have a mechanical mechanism or solenoid to engage / disengage the load. For motor and position control i'd have an LVDT on the rotor and on the actuation disk.

I work on plane actuator motors and designing for very low inertia in the motor you get accelerations zero to 5000rpm in <0.1s quite easily. the key is minimising the inertia of the system.