Objective value to compare the vibration of two (similar) systems

[tralala]

Hello everyone!

For the transportation of very light objects (15gr), we are using conveyor belts with servo motors. We have noticed (after mounting an accelerometer) that the steering of the motor generates vibrations that are transfered to the belt. This makes our object slip and loose its relative position to the belt, what prevents our mesuring equipment to deliver usable results.

Now we want to find a new motor that does not generate "that much" vibration. At the moment we are considering all sorts of motors (servos, stepper, torque, etc).

Just by watching the acc-time diagram we note that some motors work smoother as other. The problem is that we need an objective way to compare the induced vibration on the belt.

I was thinking of using the RMS (root means square) acceleration in order to compare 2 systems. The system with the lower RMS should be working smoother...

Is the RMS acceleration the right value for a comparison of the behaviour of two systems?
When not, which value would be the right one?
I would also appreciate any links to literature or papers that hadle this problem.

Thank you very much,

Tralala

 

[electricpete]

My guess is you would want to look for the worst true peak acceleration.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[btrueblood]

Or possibly at the 1st time derivative of acceleration (jerk). No, that is not a snide comment, that's a standard engineering term.

 

[fetraining]

Tralala,

The RMS calculation would be a way to compare motors, but it might be misleading. If the distribution of acceleration in the time history is 'normal' in a statistical sense then the RMS value is the standard deviation (or 1 sigma) of the acceleration - there is approximatly a 68% chance of this level being seen. To get a more reliable (smoother) system you need to go to 3 sigma or greater.

If you are just comparing motors, then 1 sigma or 3 sigma is just a scaling factor and you can use either.

However it would be uncertain if the scatter of accelerations is a normal distribution. You would have to take a lot of measurements and see if they fall into this distribution.

My statistics runs out at this point. There are probably other ways of camparing non-normal distributions.

What you would really like to know though is what level of acceleration causes a slip. This is the accelaration * mass = force to overcome frictional force. Frictional force is the mass*g * friction coefficient.

With an estimate of the slipping acceleration I would be tempted to rate the motors by an exceedance count of this per minute or hour as appropriate.

I.e Motor 1 - 30 exceedences of 90% critical, 5 exceedences of 100% critical, 1 occurence of 110% critical.

If you compare graphs like this you will get a quick feel for the spread of critical accelerations, and the peak values. The RMS does a similar job statistically, but assumes the the shape of the distribution, this way you see the distribution.

Hope this helps,
Tony Abbey





Tony Abbey

http://www.fetraining.com