Collisions caused by smoothing

[piperVVYZ3]

I was getting collisions reported in a 3D Contour toolpath.  But watching it in the simulation, it appeared to be fine.  The moves that were colliding were a simple retracts, straight up out of a machined groove.

Eventually, I discovered that what was supposed to be a straight-up retract, was actually not.  Rather than being constant, the X and Y positions were very slightly different at the top of the retract than at the bottom.  Like, rounding-error sort of slight.  But that small lateral movement during the retract would cause the tool to collide with the wall of the groove.

On a hunch, I turned off the Smoothing option, and that fixed it.  The retracts that collided before were now perfectly vertical, with the X or Y movement at all.

I was trying to produce a simplified model that reproduced the problem, but I haven't been able to make it work yet.

[Laurens-3DTechDraw]

---

@piperVVYZ3 wrote:

I was getting collisions reported in a 3D Contour toolpath.  But watching it in the simulation, it appeared to be fine.  The moves that were colliding were a simple retracts, straight up out of a machined groove.

Eventually, I discovered that what was supposed to be a straight-up retract, was actually not.  Rather than being constant, the X and Y positions were very slightly different at the top of the retract than at the bottom.  Like, rounding-error sort of slight.  But that small lateral movement during the retract would cause the tool to collide with the wall of the groove.

On a hunch, I turned off the Smoothing option, and that fixed it.  The retracts that collided before were now perfectly vertical, with the X or Y movement at all.

I was trying to produce a simplified model that reproduced the problem, but I haven't been able to make it work yet.

---

You give a tolerance in the smoothing box by which the toolpath is allowed to be off, so this would be kind of expected if you ask me.

What were the tolerances set at?

[piperVVYZ3]

Yes, that's probably related.  But if I reduce the smoothing tolerance, that will reduce the effectiveness of the smoothing.  It would be a shame to have to do that, just to solve this problem, when all it really needs is some improvement in the algorithm.

It seems to me that if smoothing moves the bottom of a retract, the code should also adjust the top of the retract by the same amount, so the retract remains perfectly vertical and problems like this don't arise.  It should probably do the same thing with plunges.  What happens up in the rapid movement plane probably doesn't matter.  All motions up there are straight lines anyway.

[Laurens-3DTechDraw]

---

@piperVVYZ3 wrote:

Yes, that's probably related.  But if I reduce the smoothing tolerance, that will reduce the effectiveness of the smoothing.  It would be a shame to have to do that, just to solve this problem, when all it really needs is some improvement in the algorithm.

It seems to me that if smoothing moves the bottom of a retract, the code should also adjust the top of the retract by the same amount, so the retract remains perfectly vertical and problems like this don't arise.  It should probably do the same thing with plunges.  What happens up in the rapid movement plane probably doesn't matter.  All motions up there are straight lines anyway.

---

Normally I would assume there are lead-in and out's which would solve your problem as well.

Could you share this part or another one with the issue so we can see what actually is going on?

[piperVVYZ3]

The lead-ins seem to assume that you've faced the entire surface of the stock, and cleared out everything around the perimeter of  the model as well.  But I don't want to use that much machining time, just to turn a block of aluminum into chips.  I try to just cut away a groove around the model, and leave the rest untouched (except for tabs to keep it attached to the stock).

Under those conditions, lead-ins collide all over the the place, as do helical ramps.  I turned off lead-ins, and use profile or plunge ramps.

[piperVVYZ3]

I was able to produce a very simple model that exhibits the problem.  Here's a public link to it: http://a360.co/2qsaZDA

Right now, if you run the CAM simulation, it will complete with no collisions.

Edit the Contour4 step, and turn on Smoothing.  Then run the simulation again, with Stop on Collision enabled.  It should stop at the first retract move.  If you single-step the moves just before the collision, you'll see that the colliding movement is a rapid from (-3.82148, 11.3142, 4) to (-3.82064, 11.3117, 15): mostly straight up, but with a very slight lateral movement.

Turning off smoothing again, and examining the same moves, you can see that without smoothing, the rapid is from (-3.82064, 11.3117, 4) to (-3.82064, 11.3117, 15).  Perfectly straight up, no lateral movement at all.

It appears that, as I suspected, the smoothing pass very slightly moved the X and Y of the bottom of the rapid, but left the X and Y of the top of the rapid unchanged.  That causes a collision with the wall of the slot.