Kinetix 6000- Could you please explain Maximum Acceleration and Deceleration Jerk

Good Afternoon ,

We have a number of Kinetix 6000 Servo Drives and motors. I'm trying to understand these a little ( really alot ) better . Could you explain the following in simple terms.......

What is Acceleration and Deceleration Jerk ?

What is Maximum Acceleration and Deceleration in Servo terms ?
( I understand in Frequency Drive Terms )

We keep breaking a ball screw on a machine. The Acceleration and Deceleration Jerk is set for 333037.2 inches /s^3 .
Does that seem high to you ?
and what does the inches / s^3 mean ?

Thanks so much .

The maximum numbers are values that the motion profiler won't violate. That is to say, you can ask for higher values in the program but the system will only command values up to the maximum. These numbers are often calculated by the development software as a result of a velocity or position autotune based on the available peak torque and the calculated system inertia.

These ARE NOT necessarily the accel and jerk values that the machine is seeing. These are hard limits that the software will not intentionally violate.

Jerk is the rate of change of acceleration, just like acceleration is the rate of change of velocity and velocity is the rate of change of position. If your position units are inches, velocity is in inches/second (in/s), acceleration is in inches/second squared (in/sec^2) and, you guessed it, jerk in in inches/second cubed (in/sec^3). Jerk defines the shape of the rounded part in an S-curve acceleration.

The sepoint values of acceleration and jerk are set in the motion instructions used to make things move. These can be entered in a few ways but one way is percent of maximum and the other is in axis units per time. If your motion instructions use percent of maximum for these values then the maximum acceleration and jerk have real program consequences. If you are entering these items in axis units per time then the maximums aren't as important.

Are you thinking you are twisting the ballscrews in half due to acceleration?

Keith

The jerk is way too high. The jerk is so high it is almost like using a trapezoidal velocity profile with no jerk time. 1g=386 in/sec^2. If the load is accelerating and decelerating at 200 in/sec^2 and the jerk is 20000, it would take only 10ms to accelerating from 0 to 200 in/sec^2. jerk time = acceleration/jerk.

The jerk value should be at most 100 times the acceleration or deceleration rate. 40-50 times is much better.

Hopefully this will provide you with a 'feel' for what the values should be.

What are the acceleration and deceleration values?

Always nice to see motions programmed with engineering units rather than something arbitrary. Makes it easy to relate to when moving across platforms, or commissioning new mechanisms.

Make sure axis configuration takes into account drive ratios. For a ball screw, lead, and whatever gear reducer may be in place... Verify that a one inch move command results in a physical move of one inch...

When doing startups on new machines, I always use 1g = 386 in/s^2 for initial values accel/decel, and modify as required, almost always using multiples. Peak velocity isn't always as important. Using motion planner to simulate moves prior to having actual hardware helps to visualize move profiles. Most controls folks without a physics/engineering background typically end up with some pretty arbitrary units...furlongs per fortnight is a valid unit, but doesn't translate well.

Apologies if this sounds obvious...

Originally posted by Peter Nachtwey:

The jerk is way too high.

Click to expand...

Again, keep in mind that the MAXIMUM values don't come into play UNLESS the accel/jerk are defined in percent of maximum at the motion instructions OR the accel/jerk are defined in user units AND they are greater than the maximum. Outside of that, it is just a number that, depending on whether autotunes were run, reflect the physical limits of the system.

Keith

Who you calling a jerk?!

Thanks for all your information. This ballscrew only moves about 2" Foward and Reverse. This is an existing machine , and we have had a history of breaking ballscrews. We have two loaders that we call them. They are somewhat identical . I made a comparision sheet with all the tabs. I need to look in the program also. I have the Comparision Sheet attached. Wrapper 1 is the one we keep breaking ballscrews . We never have these issues with Wrapper 2 . You will see many settings are higher with Wrapper 1 than they are with Wrapper 2 .

Also , just for learning sake , Why , and in what case would you use Acceleration and Deceleration Jerk ?

Thanks so much for your help,

kamenges said:

Again, keep in mind that the MAXIMUM values don't come into play UNLESS the accel/jerk are defined in percent of maximum at the motion instructions OR the accel/jerk are defined in user units AND they are greater than the maximum. Outside of that, it is just a number that, depending on whether autotunes were run, reflect the physical limits of the system.

Keith

Click to expand...

So when and why would a lower value of jerk be used? I know that most people have no clue what the jerk should be.

The acceleration and deceleration values are very high. If I round the numbers a bit, the jerk is 100 times the acceleration rate which means the time to the acceleration rate is about 10 ms. This is OK but the acceleration rate is probably too high. I will stick this into my python program. We get many tech support calls from customers hoping that the motion controller can defy the laws of physics.

I just checked the motion profile. It looks good if the system can really accelerate and decelerate at 8.8g without breaking.

A trend showing the target and actual positions and velocities with the control output would be nice.

atc222 said:

Always nice to see motions programmed with engineering units rather than something arbitrary. Makes it easy to relate to when moving across platforms, or commissioning new mechanisms.

Click to expand...

Agreed, I usually ignore posts that don't have units. Units are extremely important or one can crash probes into Mars. ooops.

I also ignore posts about PID gains when the units are not specified.

Another possibility is having an axis fault while load is traveling at maximum speed and coasting into hard stop. Instantaneous deceleration is usually an effective way to break stuff...

On closer inspection, the acceleration and deceleration values are never quite reached. My program computes 7 3rd order motion segments. the first segment the acceleration is increasing, the second segment the acceleration is constant but because the jerk is not high enough the acceleration of 3400 is not reached before the jerk must be negative to ramp the acceleration down to 0 when the speed is 30 inches.
It would be better to increase the speed to 34 inches per second of the motor can go that fast. This way the acceleration can reach 3400 in /s^2 before ramping down to 0.
Either that or reduce the acceleration to 3000 in/sec^2 and the jerk to 300000 in/sec^3

This likes like a rounding error problem. in what ever calculated the parameters.
Another alternative is to accept the fact that the acceleration will not reach 3400 in/sec^2. The computer should calculate out a good value. In this case there will be no constant acceleration segment.

Thanks so much for taking the time to look into this. So how did you get this information just from the Bayonet Properties ? Does your software show a trend , that was developed by that data that you entered ? If so can you display that ?

I looked in the RS Logix 5000 program for Wrapper 1. The only instructions was a MAM ( For Open Bayonet ) , a MAM ( For Close Bayonet ) , and a MAS for stopping .

I have those 3 instructions attached.

Thanks again. I know you guys are busy.

I assume you are seeing the maximum jerk in the Axis properties.

Looking at your instructions, your jerk rates are set as % of time (not the same as percent of maximum). AB recommends that you use at 100% of time as a starting point. The applications that I have done have performed fine with that setting.

Your "close" instruction is set to 10% the other two are 100%. You could try decreasing the other two instructions bit by bit to see if that causes any functionality issues.

What really grabs my attention is that your MAS instruction has a decel rate of 100% of maximum. That might be destructive.

I was just about to ask for the motion instructions info. Glad you posted it.

For the moves the profile type is set to "trapeziodal". This means that the jerk value in the motion instruction and the jerk limit value in the axis configuration don't come into play.

When you look at the axis set-up values you can kind of get a feel for where they come from by the significant digits that are displayed. For example, in the Dynamics tab in the comparison document you provided the max speed, accel and decel values have no digits to the right of the decimal point for Wrapper 1.
For Wrapper 2 there are several. This tends to indicate that someone typed in the numbers for Wrapper 1 and the values for Wrapper 2 were calculated in the development software. The problem is the accel value is based on the inertia values from the Output tab. Wrapper 1 has a load inertia ratio that looks pretty reasonable. Wrapper 2 does not. So while the values for Wrapper 2 Dynamics look more reasonable, those values appear to be based on incorrect information.

So if we assume the Wrapper 1 inertia numbers are right and the resulting Wrapper 2 accel rate is right, the true physical accel limit is probably around 3100 in/sec^2.

The programmed accel rates are 2000 in/sec^2 for close and 3000 in/sec^2 for open. That is definitely aggressive, especially for an axis with a trapezoidal profile. I assume the recipes are the same between both wrappers? As for the MAS, it looks like the instruction is changing the the accel rate to 10% of the limit value (340 in/sec^2 for Wrapper 1 and 520 in/sec^2-ish for Wrapper 2) and the jerk value to 100% of the accel time. This is more gentle than the MAM parameters by a fairly large margin so I don't know why this would be done for an abort, but that is what the instruction says.

Another thing I notices in the axis config compare document is that Wrapper 2 has velocity feed forward set to 100% while Wrapper 1 is set to 0. I would think this would cause some position lag in Wrapper 1 that is not present in Wrapper 2. In addition, neither of them is using any acceleration feed forward. With accelerations as aggressive as those set here I would think accel feed forward would be a big help. More so with Wrapper 1 than Wrapper 2 since I think Wrapper 2 total inertia is closer to correct. But it would help both.

Again, what is the failure mode of the ballscrew systems? Are you tearing the ball races out of the nuts, tearing the ball tracks off the screws or twisting the screws in half?

Keith

my bad - the MAS decel rate is 10% as Kamenges stated.

I do not ever use the %of max option. I always like to use a set accel/decel rate so I don't have to go look elsewhere for what the max is.