[gar]

050826-1313 EST USA

We have a surface finish problem on a VF-3, 1998. This shows up worse at 7500 rpm.

Rather than have HAAS replace one component at a time to find the cause I want to find a means thru measurements on the machine to try to pin point the bad component.

So far we have spent about $1000 on HAAS service and talked to the factory without any clear direction.

What we observe ---

The surface finish problem is visual and is probably in the range of 10 to 100 millionths of an inch ( in the micron range ) in surface variation.

This is somewhat erratic but appears to have some component on the order of 5 revolutions of the spindle. In other words a visual pitch of .02" at 30 inch/minute feed and 7500 rpm gives us, 7500 rpm = 125 rps or 125 Hz for the once per revolution frequency, 30 in/min is 0.5 in/sec and .02" corresponds to about 25 Hz.

If we double the feed rate to 60 in/min then the .02" pitch becomes about .04". This implies that the problem is spindle related rather than X or Y axis. At 60 in/min ( 1 in/sec ) I can see an .008 to .010 pitch also. This is a frequency of 100 to 125 Hz or on the order of once per revolution. In other words we see a once per revolution component, and another at 4 to 5 revolutions.

Both end and side milling shows a surface finish problem. Changing cutters from two to three flute, or changing cutter length does not eliminate the problem. Can not say at this time whether the surface finish changed. Running the same cutters, material (6061), and program on a different HAAS ( VF2 of 1993 or VF0 of 2000 ) produces roughly a mirror surface on the side milling.

The side milling surface is identical whether X or Y axis is viewed.

HAAS tests on drawbar force were just slightly under low limit. Their ballbar test was better than some new machines (well below spec limits). Their spectrum analysis from accelerometer measuring perpendicular to bearing housing produced nothing that stood out. And their testing for end play did not indicate anything. Note, the surface finish problem is under very light load conditions.

We are clearly looking at a variation that results from once per revolution and lower. Thus frequencies below 125 Hz.

The spindle speed on the problem VF3 appears to be very stable.

The VF3 with a problem has a HAAS vector drive. The VF-2 and other VF3 are variable frequency and the VF0 is a HAAS vector drive.

Runout (TIR) on all the machines is on the order of 0.000,5". Also from my tests I am fairly sure there is at least one ball bearing cage velocity of about 1/2.3 of the spindle velocity (rpm). HAAS has been unwilling to tell me what cage velocities I might expect. My 2.3 comes from runout frequency analysis. Also this correlates approximately with an open ball bearing (arbitrary) that I measure about 2 1/3 for this value.

I have now run many tests on vibration (perpendicular on spindle housing), power input to the vector drive, and runout on the different machines. In fact the cage runout component on the VF2 is worse than on the VF3 but the VF2 produces better surface finish.

My next test is force vs displacement on the spindles.

Runout by itself does not appear to be the problem.

The problem VF3 spindle feels very free when rotated by hand.

I am not looking for an individual ball problem because this would produce a higher frequency than once per revolution.

We believe the problem may have developed over a year or two period.

I am following the instrumentation path because it is useful to have a way to pinpoint a cause rather than simply use substitution.

Is there anyone else that has had this problem and what was the cause?

.

[JPMach]

Does it have a gear box?
I don't know what the gear ratios are in a haas but I could see a bump on one gear could cause every rev and a bump on the other could cause every 4th rev.
Just a thought.

JP

[Geof]

This is a bit of a shot in the dark. Has this machine ever been shipped without the spindle immobilized? If you have brinnelling on the races due to shipping vibration this could give you an effect in synchrony with the cage rotation. The inner and outer race brinnel marks would come into opposition many times every revolution but there would only be a ball in the precisely correct position every cage revolution. (I think)

[HuFlungDung]

Personally, I'm surprised that spindles using interchangeable tooling are as good as they are. There aren't any other kinds I guess , but if I was looking for perfection, I'd be really looking closely at the fit of the tapers. If the spindle taper is a wee bit off angle (or off in any other way), you'll never notice outright looseness, but still, the tool will be able to precess and vibrate inside the spindle taper.

I suppose you could test for this with some kind of a leverage test, with a fairly long, one piece toolholder, so that you know you are not introducing another flexible connection. Perhaps an extended shell mill holder would work. Hold a chunk of bronze in the vise, and bore a hole in it, for a running fit on the pilot end of the extended toolholder. Then, for the test, lower the pilot into the bushing and apply a few thousandths of deflection by moving the X or Y axis a little bit off center. Perhaps you could place your vibration transducer somewhere on the bushing, to see what it picks up. By measuring the change in vibration from dead centered, to however far off center you dare go, you might be able to detect whether the shank of the toolholder is hammering inside the taper, ever so slightly.

[Kmed]

I am with HU..
Get a new tool holder. DY kem it then without a pull stud press it into the spindle bore, and see where the wear is. ( clean well first).. If this has lines showing up it it.. Make one out a Polyethelene on a lathe )If you need one I can sell you one of mine i made).. Then lap it back in. I did this to my bostomatic, and it made a big diffrence in the surface finish and the runout of my test pin went from .00043 to .00016. This is 3" from the collet face. ...

[gar]

050827-1957 EST USA

Thanks all. I have read all the messages and they are all important.

I have had some strange results today and will report more later.

A major result of your comments is that now I am running a feed of 120 in/min for some of the tests and the once per rev runout makes a nice visible time marker.

.

[MR_NC]

From a logical standpoint. I would start looking away from the spindle and more at the tooling, pull-studs, and draw force.
Reasons:
You stated force was below spec. I would fix that just to take it out of the picture.
The problem came over 1-2 years it seems. IMHO a bearing problem would have gotten "ugly" by now I suspect. Just an assumption however. Anything is possible of course.

Questions I have:
Are you doing all this testing with same tool holder? I would try a cross section of holders and see what the results are.
Is there a threashold speed? Meaing do you not see this problem below spindle speed X?
Is there any change with regards to result in testing machine right after power up and after full operating temp?

Best regards,
Sean

[gar]

050830-1208 EST USA

Hi all:

I have many comments to your various suggestions and no solution yet.

JPM --- yes this machine has a gear box. Our primary problem is in the range of 6000 to 7500 rpm, thus we are in high gear.

The HAAS manual provides little information on specifics. However, one can draw some conclusions. Some of the spindle drawings are about 1/2.88. Gear box drawings are much smaller.

HAAS specifies the spindle pulley as 1.875 and on the drawing it is about 1.5 or 1.5 x 2.88 = 4.32 but this does not correspond with 2 x 1.875. On the other hand if this is the id of the pulley it does not scale correctly either. I can not correlate pulley size on gear box either. What is the length of the belt? Also I do not know that.

On the gear box drawing there is an idler gear between the motor and the gear box output shaft. This is about .18" for the output and .225" for the motor, or a step-up of .225/.18 = 1.25 from motor to output in high gear. I will use this as an assumption.

I can run a test at 7500 rpm with a 1/2" two flute cutter and 120 in/min feed rate and get a good once per revolution pattern on the cut. This corresponds to 125 Hz at the cutter. Ball park runout is 0.000,5" (TIR).

Same at 6000 rpm and got a pattern that correlates with 100 Hz.

We are not concerned with this once per revolution surface finish problem. Rather the problem is a similiar pattern that occurs at 30 in/min and has a period that is about 5 times that of the once per revolution. Thus, at 7500 rpm and 30 in/min thiis corresponds to 25 Hz.

A test at 6900 rpm and 30 in/min has a pattern of about 22 lines for 25 lines on a 7500 rpm sample. This ratio is 25/22 = 1.136 and the ratio of 7500/6900 is 1.086. The fact that the pattern was not unchanged says that there may be a speed related correlation of the surface pattern with spindle speed.

At this point I can not say or find something in the motor , gear box, and belt path that correlates with a factor of 5.

As I have previously indicated there should be no reason to consider a once per ball revolution or other things of a higher than an output shaft once per revolution frequency. Their effect should be of less importance than that of the shaft once per revolution.

Geof --- primarily this would be a high frequency effect. If there were a once per cage rev effect it would not produce an output sinusoidal result. Also as best that I can determine cage velocity is about 1/2.3 of spindle. This does not correlate with a 5 to one ratio.

Hu --- toolholder loosness is not likely, very hard to remove after clamping. Looseness would be unlikely to provide a stable pattern at the 1/5 frequency.

I had previously setup a 3/4'" Oilite bearing in the vise and with our test 3/4 rod in this and varied the side load. Could not produce the 5 factor effect, but did get the expected 2.3 effect in runout. And of course the once per revolution factor.

Kmed --- our spindles on all our machines are quite good for HAAS, .000,1 to .000,2" (TIR). I do not currently have a way to get my LVDT to measure this, and therefore only used a dial indicator. Our toolholders mate very well with the spindles, but do have runout that exceeds the HAAS spindle.

On my runout measurements from the 3/4 shaft at 1300 rpm, can not currently run faster with the LVDT, I had a 55 db signal at 21.698 Hz (once per revolution), and 14 db at 9.399 Hz (assumed from the cage) ( 21.698/9.399 = 2.309). Obviously the frequencies are not as accurate as the decimal places imply, but this is the FFT result. This is quite a small signal compared to the fundamantal. This was done with an 4 minute average of the inout signal.

On the other hand we do need better runout on some of our toolholders. I doubt that reduced toolholder runout would solve our cuurent problem. But since I do not know the cause it can not be totally ruled out. However, the same tool and toolholder, or different tools and tool holders produce the same results in this 1998 VF3, but produce good results in other rmachines.

I would like to know more about your lapping procedure. If you are lapping the spindle, then how do you force the lapped cone to be concentric with the centerline of spindle rotation? Also the same question relative to the toolholder, and its relation to its collet?

MR_NC --- there is nothing so far that I can pin on the spindle. If I run a vibration test with the accelerometer mounted on the spindle bearing housing measuring the radial vector and no toolholder, then I can not see the fundamental or the 1/2.3 components at 7500 rpm, but there is a strong 100 Hz component. By not see I mean they are not identifiable as meaningful. Is this 100 hz from the motor? I do not know. Adding heavy unbalance then the fundamental component shows up.

I do not think the clamp force is a factor because HAAS previously used a lower value, and the toolholder appears to be locked in the spindle by the amount of force to remove it.

Force vs displacement measurements vertically on the spindle do not indicate a lack of preload on the bearings.

We have used different toolholders and tools.

We have done heavy unbalanced tests with two 1/4-20 screws radially. One screw head is 2.5 from center and the other is 1.5". At 7500 rpm this shakes the machine. With this unbalance the once per rev fundamental shows a strong signal at 125.099 Hz and 60 db. The 101.292 Hz component is about 46 db. 89.626 is 53 db, 21.537 is 50 db, and 13.442 is 51 db.

Below 6000 rpm the surface finish problem is mostly gone.

Have not been able to find a difference between initial startup and after running.

Most tests have been run with X, Y and Z still.

Over the last weekend I ran comparative tests cutting 6061 over a straight path 5" long. Side milling .25" in Z and DOC usually .02" some times .04". These were done at both 120 in/min and 30 in/min. Accelerometer tests were done at 30 in/min.

At 7500 rpm, 30 in/min, .02 cut with accelerometer on spindle housing perpendicular to x-axis we get a strong component at 25.05 Hz (69.3 db), but the fundamental at 124.97 is only 54.5 db. 100.94 is stronger at 63.6 db.

Shifting to 6900 rpm and we do not get what would be expected, yet the pattern on the cut is as previously seen. Then I moved the accelerometer to the vise, both perpendicular and parallel to the x-axis, and no correlating components showed up.

I need to drop this experiment for a while. But later I want to revisit power measurement to the motor, a way to measure speed and runout at 7500 rpm on the shank of the tool while cutting.

Thank you everyone.

Do not not hesistate with new ideas. Your various comments have caused me to redo experiments and to think of new ones.

.

[mxtras]

I love this thread!

Gar - you have compiled some heavy duty information there - awesome analytical work! I like your approach. With testing of this caliber, you will isolate the problem I am sure.

Could this be a line frequency/motor/machine harmonics thing? Or do you feel it is straight mechanical?

Keep us posted, please. Very interesting. Now if you'll excuse me, I need to go re-read most of your posts.

Scott

[miljnor]

most of this is over my head but as far as frequencies go, It could be a harmonic of you 100/125hz couldn't it?

I mean a fly landing on the spindle causes some deflection right so a small loosness somewhere will cause a small harmonic and when you hit that sweet spot all hell breaks loose (or in your case a small surface finish defiation). Hell it could probably be something as minor as a loose encoder belt on the spindle setting an osilation up in the spindle controler.

just throwing some random stuff your way to maybe give you some ideas. Just making you think this guy don't know $h!t might bust something loose!

[gar]

050830-1521 est usa

mxtras and miljnor:

Both good comments.

I have constantly swung back and forth on whether this is a mechanical or electrical problem. It should be noted that the same DC supply, contained in the spindle vector drive, is used to supply the brushless servos.

I am looking for a measurable signal that is a predictor of the end results. Frequency analysis is extremely useful, especially when I can average the signal over a long time, and the signal is stationary. Stationary means that the statistical properties do not change over the interval of interest.

Anything in the gear or belt train that would cause this problem should not be particularly sensistive to the speed (meaning amplitude with speed), unless there was a resonance. Its frequency should be directly related to speed.

When I am running at a spindle speed of 7500 rpm the once per revolution frequency is 125 Hz derived from 7500/60 = 125 rev/sec. This in itself won't generate 100 Hz. Further when there is nothing hanging on the spindle (it is very well balanced) I do not see a 125 HZ signal from the accelerometer. When I put a major unbalance on the spindle, then the 125 Hz signal pops right out.

But running the spindle at 7500 rpm I see a strong 100 Hz signal. If total gearing from the motor to the spindle is 1 to 1.25, then the motor is running at 6000 rpm for a 7500 rpm spindle speed. Now I have a source of 100 Hz. Likely the motor is not as well balanced as the spindle.

From the magnitude of the unbalance I added to the spindle to get a good solid 125 Hz signal I do not think that a minor cutter unbalace would have any effect on my problem, and unbalance does not correlate with the 25 Hz.

On my above assumption, the 1.25 ratio of gears, the motor is running at 100 Hz for 7500 rpm. 25 Hz is a subharmonic of this and it might originate in the motor control.

If I had a small speed variation in the motor could it produce the surface pattern I get? I have no idea.

A cage velocity of 1/5 would be a far better explanation, but then the surface finish problem should not drop away around 6000 rpm.

.

[Geof]

Are you possibly experiencing something like belt slap? The motor/spindle drive is a toothed belt I believe; this could generate several fundamentals from the tooth spacing, the different pulley diameters and the different natural frequencies for the tight and loose sides and could give rise to beat frequencies that could come and go over a narrow rpm range.

And just a question for interest sake; have you done any tests with a lefthanded cutting running the spindle counterclockwise?

[HuFlungDung]

Could it possibly be just the vibration of the whole head on the Z axis rails. Is there some method to calculate the resonant frequency that the head would have?

It might be a good idea to take the spindle motor out and test it alone. Might be time for some new bearings in it anyways

[JPMach]

Just another shot in the dark but could it be a hiccup in the carrier frequency of the drive to the motor. Thus slower rpm equals small change in each pulse of dc so hiccup is not seen at higher speed the change in pulses of dc is greater which might make the hiccup stronger. This hiccup might translate into a ever so slight change in chipload which would cause a mark on the part.
Just a thought.

JP

[gar]

050830-1943 EST USA

Geof:

The HAAS serviceman tightened the belt with no change.

My unloaded tests on spindle rpm were made with a 100 tooth encoder on the 3/4 test shaft at 7500 rpm. I first counted down by 5, then 2 using a 7490 to generate a balanced square wave. Then I used the Tektronix expaned sweep and looked at the tenth count. There was no jitter. Next time I will count by 50, then two and average for 4 minutes, then do the FFT. This should do better than my viewing the scope.

I currently do not think there is an unloaded belt or drive train problem.

Next time I need to measure from the cutter as I previously mentioned. Then I can get data while cutting.


Hu:

That is a thought. I tried tapping the head with a soft hammer and the damped frequencies generated were higher than 25 Hz. But this may not represent the head on the rails. My guess is that the head on rails will resonant lower than 25 Hz, there is a lot of mass in the head.

When I had the accelerometer on the vise I tapped the coolant catch tray attached to the X Y table and got as much signal as when I was cutting the sample material.


JPMach:

This is the area I keep thinking about. The power meter I used to monitor the moter was a three phase unit using Hall devices. This does not have the bandwidth that is really required, but it was interesting and I did see a low frequency component. I can not run it at 7500 rpm because it is saturated with just the spindle power losses.


I will report back when I get some photos on my web site.

.

[ViperTX]

I would side with Hu....I suspect an imbalance in a rotating part.

[gar]

050901-1449 EST USA

Hu and ViperTX:

An unbalanced member rotating at constant speed will produce a sinusoidal force vector at one fixed reference angle. At any other angle relative to the reference angle the vector magnitude will be the same but the timing will have a phase shift.

A tooth on an end mill will produce a force pulse which is not a sine wave. It roughly can be defined as an impulse.

Since at 7500 rpm and 30 in/min I see what appears to be a 25 Hz variation on the surface. This won't result from an unbalanced member unless it is about 1500 rpm. There is nothing in the drive train that results in a frequency this low unless it is the length of the belt. The gear box idler shaft is lower than 6000 at an output of 7500, but not as low as 1500.

However, as Hu has suggested does something resonant in the 25 Hz range that can be excited by a higher frequency source.

Hu's comment has forced me to try to find head resonance that might relate to the finish problem. I was wondering where I would find a suitable exciter. Then it became self evident even though it is not the best.

25 Hz corresponds to 1500 rpm. Thus, if I use my heavy weight unbalance test detail, then I can run the spindle at various speed and produce a sinusoidal force excitation above and below this 25 Hz. A resonant circuit, whether electrical or mechanical, will have a higher amplitude at resonance than on either side.

I am in the process of running some tests. So far I have found little at 25 Hz, but a relatively large 20.032 Hz signal that does not change with motor rpm. The various frequency components are different, for the same excitation frequency, between vectors in the x axis direction vs y axis. The 20.032 Hz signal is almost always large, but is it some unrelated instrumentation artifact.

At an excitation in the 1000 rpm range I have generated a very large 5 Hz component. Later I will supply the exact rpm.

.

[Geof]

Your first post has this sentence: "The side milling surface is identical whether X or Y axis is viewed." Which I take to mean you see the same effect whether you are passing the cutter along the X or Y axis. If this is the correct interpretation then if the source is a resonant head vibration it implies the vibration has the same frequency and amplitude in both directions. That would be surprising. Possibly I am not following things correctly.

On another tack; have you tried boring a hole with a single point tool? Granted it might be difficult to get the tool dynamically balanced but this may not be essential. You bore down taking a very light cut then retract rapidly with the spindle still running; the tool trace during retraction will not be even if the head is vibrating.

[gar]

050901-2056 EST USA

Geof:

When I initially said the effect was the same in both X and Y this was not a precise statement in the sense of the greater analysis I have done cutting along the X axis. Looking at the part that started this analysis the patterns look very similar. I count 5 cycles per 1/10". That is 25 Hz at 30 in/sec, and I do not think that I am off by 20% to get to 20 Hz.

The resonant frequencies of the head are unlikely to be the same in both the X and Y directions. And the measurements I have made show different patterns.

I have revisited test speeds above and below 1500 rpm and nothing significant occurs at 25 Hz. Interesting that at 1500 rpm I got a very large spike at 0.208 Hz ( one cycle every 5 seconds ). Nothing like this at +/- 1% of 1500. The rest of the spectrum at 1500 rpm had no noticable spikes.

No I have not tried a hole boring experiment.

.

[gar]

050905-2104 EST USA

Have run a lot of tests.

Does not look like a 325 vdc supply coupling between vector drive and brushless servos is our problem. The 325 v is fairly clean.

It is very clear that there is a 25 Hz modulation of the surface finish at 7500 rpm.

I have now created a crude mechanical exciter for 50 hz down. With the machine unpowered I can generate a strong resonance around 50 Hz and some other frequencies. This is providing an excitation in the x-axis direction directly to the table. The excitation is sinusodial. The accelerometer is mounted on the vise sequentially in both the x and y axes. I do not have a triaxial accelerometer, thus separate runs are required.

It appears the cutter is exciting the lower frequency resonances. Possibly the 50 Hz resonance is not dominate from the 7500 rpm input because this is a 2.5 divisor, rather than an integer divisor.

.