Doosan DNM5700S mills and Doosan TT1800SY machines

We're about to take delivery on another 2100SY II and another TT1800SY, this TT is getting downgraded to the Oi control from 31i and is going to have thermal compensation which the other machines don't have. I hope the transition is good and the machine doesn't lose performance with the I series control because it's my favorite Doosan lineup machine.

We've had a great experience with the Doosan turning machines. They have been reliable, performing machines. The TT1800SY is my favorite machine. I kind of hate programming in mastercam, but I do enjoy applying strategy to machining and just stomping the living **** out of run times with the TT1800SY. I think the only weakness in the TT1800SY is the transfer time which is ~32 seconds. If we're running a part with a sub 2 minute run time which is a minority situation but happens, the 32 seconds hurt productivity.

We're having a problem right now with our DNM5700S mills - they can't seem to accelerate and decelerate in a high performance fashion. They are getting outperformed on the floor by a 10 year older Mazak smart control (their similar economy control). I think the Doosan mills have the ability to perform, but I don't think the parameters are allowing that to happen. We're running programs on the 15K spindle DNM5700S machines that we're posting to Mazak's dinosaur with the 12K spindle, and the Mazak is beating the DNM run times with less RPM capability in aluminum on the same chip loads per tooth (so 25% slower posted feeds). Ellison is supposed to be helping us out with that Wednesday of next week and we're looking forward to that, because we want to run faster on that side of the shop. We have to run faster to be competitive over there. Those programs involve 6" OAL parts, so when we're dynamic milling, we are often making cuts in 2" spans and the posted program running 400IPM cut feeds seems to be averaging about 100-150IPM. We see acceleration and numbers running up to around 390IPM where its just a flash and we're decelerating in milliseconds (the 390 was just a flash during a cut). This sluggish accel and deccel is affecting back feeds also so it's like the real rapids in the program are also slow. On the old Mazak, we see the numbers reach 320IPM fast and hold through the similar cuts. The mazak is if anything heavier, so it's not built lighter to move faster. The mazak has a tank of a spindle and we aren't running that optimized- we could probably cut 20% more with the mazak and put it 45% faster than the DNM5700S, but our strategy has been to just change machine defs and post to mazak with the same per tooth feeds because the DNM is harder to get chatter free cuts with and the Mazak runs anything the Doosan can cut.

We are running some operations on the DNM's that are running 5 times mastercam backplot speed. On the TT1800SY side we see about 1.5 times backplot speed.

I don't think this is a USA problem, or a Doosan or Fanuc hardware issue. I feel like and am of the personal opinion that the mills are being shipped with background settings that don't perform. I would rather wear the machine out prematurely running it faster than get poor run times, but I don't think we're even near that threshold. The mazak we bought used for substantially less than a DNM, and it has had a very hard life, you can tell by looking at it, but it's running smooth and fast and holding tolerance.

We've worked with several different arc filter settings and reached a point where we probably have pretty optimized settings- throwing Fanuc as few points as possible, and it's making the acceleration and deceleration times more obviously the problem.

I hope Doosan can fix it because I am a fan of the brand. I think Doosan should be mopping the floor with Mazak's 10 year old machine. I'm not a mill guy but I've learned over the last year that mills need speed more than any other factor. Speed gets it done in a mill especially in aluminum and the DNM5700S from looking at it, was built to cut aluminum. Letting a ten year old mazak win on the floor is like a 20 year old losing a leg race to a retiree.

We have some loyalty but with these numbers I should be buying Mazak mills- floor space is worth money.

The fourth Puma 2100SYII and the third TT1800SY were delivered this morning. I saw a settup machinist hitting the green button kind of rapidly for several seconds in single block and wondered what was up and saw one of the Puma 2100SY II's taking like 6 seconds to ramp down from 3300RPM parting in CSS G96, in simultaneous speed sync mode. That's probably where the bulk of perceived lost transfer time is located. If that's a parameter, I would like to know how to manipulate the timer on the sync, it's pretty slow. We rev limit parting to G50 S3300 so we don't increase that time unnecessarily but we need some surface speed to cut in a tool compatible manner, so it's a game of trading ideal cutting SFM for less accell and decell of the sync'd spindles, where ideal would be more like 3500 RPM and quick up and down time for a quick ideal cut and a quick ramp up and down to M05.

The (not syncd) single spindle RPMS are very very fast, like 0-4000 back to zero in maybe 4.5-5 seconds. We run the royals on the main side to max speed 4K RPM with JF Berns liners.

It's kind of funny but I did some math and 5 seconds on transfer is average $850 a month if only 5 turning machines of 8 are running.

https://www.instagram.com/p/B_qCD9-npez/ This is the kind of motion we need! Just getting in there and kicking ass!

We had the Ellison visit today, and the discovery made was that when we turned off Fanuc high speed look ahead, we were able to cut a 36 second dynamic path to 18 seconds and better achieve feed rates.

For roughing this benefit is tremendous, that's 50% of roughing time savings, but I guess that's heavier for guys doing huge dynamic roughing paths.

The High speed look ahead had to be turned off in the machine parameter forcing it on, and also in the probing cycles which were like 8503 and 8504 programs (I'm not sure).

When we ran the part however, we found that without G05.1 Q1 on (which is by default on unless commanded off as we did) we were unable to hold shape and tolerance. So we had a lot of operations we still had to engage G05.1 Q1 R3 to get shape/tolerance quality, and we overall reduced a 36 minute run time to 33.3 minutes with no variable except G05.1 settings. That's with probing. We also ran identical probe programs in Mazak and Doosan and the Doosan version runs 1 minute longer, so something unknown causes the Doosan to be slower than Mazak in probing as well. (Update our Ellison application guy said he believes we may be able to improve probe run times somewhat also in a few weeks when his scheduling availability allows him to get back out)

So I guess I wish the G05.1 Q1 R3 or R1 ran more like the 18 second time comparison but they were like 32 second run time functions. This kind of efficiency all over the program would probably cut the run time more. I can't say by how much, but probably at least another minute or two.

This kind of said to me that the G05.1 which fanuc suggests is for speed, is actually not helping the Doosan run faster, but actually more accurately. Where Mazak smoothing and Hurco ultimotion is making their machines run faster.

We're not giving up, we're going to try to do some program optimization with some fairly major changes to improve the run time on our end.

On the lathe side, we tried applying a G97 callout to the M204 P11 call to get initial RPM aquired faster, then firing G96 S360 M204 P11 to get CSS aquired. It seemed to help, but the run time was the same so I think optimism was to blame there. We were unable to speed spindle acceleration or deceleration in simultaneous spindle command (which is used for cutoff and part transfer). We were able to allow the Ellison guys to witness the 1 second spindle stop of P11, and the 7 second slow stop of M204 P11. This issue is affecting parting speed, and part transfer as there is no synchronous stop without confirmation. I think probably costing more like 8 seconds per part if you assume the G96 parting happened slower than optimally.

We reconstructed the program and changed the design of the part to get a run time reduction to from the "improved with G05.1 turned off at times/ optimized 30.5 minute without probing process" to 27 minutes. This put the original 36 minute with probing process to 30 minutes 10 seconds by comparison. The mastercam estimate of that program without probing is 14 minutes.

In the process we had a 1/2" tool pull out of a hydraulic holder to the un sharp top of the flutes, where it pulled a part from 6" vise workholding, so we had to push that tool to a heat shrink holder to stabilize that.

We think there is another minute there maybe but we're going to have to work on it Monday. It really shows how a fast and accurate motion setting could tremendously help reduce runtimes and process engineering hurdles.

We're excited at the prospect of dropping 3 minute probing to 2 minutes with a re-calibration around a faster settup speed. Part of this is the 5 minute load time on this job plus time to get a person at the machine. We are estimating that at 7.5 minutes and that brings the hourly rate down. This is a 4th axis job so people have said pallet options don't really work for it.

The new TT1800SY is being a challenge to get operational on the floor. With the OI control the ladder needs a little help to get tool life and load monitoring functional. We've had it job ready for about 2 weeks with a part run off, and the machine isn't quite operational on account of the OI control being new. The Ellison guys are working at trying to get it resolved. Right now we're waiting on high pressure flow monitors to test the machine to see if the load monitoring is working.

We re-calibrated DNM5700S probes today around 300IPM feed and that didn't resolve the issues. We tried 100, 200, 300, and 400 ipm calibrations, but the issues with probing speed seemed to center around the 1.18IPM touch feed for the probe, and the backoff distance value being one value for two different conditions, the spindle Z and the XY travel accell and decell. So we got probe run times from 3:06 (2 second improvement, to 3:32 24 second loss). It seemed the sweet spot was at 200IPM fast travel speed, with backoff about .160". The Z seems to need more backoff, and that means if you use a higher fast feed, the backoff irregularity will cost you the time in the 1.18IPM feed which is miserably slow. We hit 30 points in the program, and it also seems that the machine stops about 1 second to think, twice per point.

An Ellison technician from Illinois put some numbers into the G05.1 settings and we ran off a 14 minute program at a time savings of 30 seconds on G05.1 Q1 R3, and that was pretty positive- a savings of 3.5% or 2 minutes per hour based on that program cutting 17-4 stainless steel. We were told fanuc Nano smoothing was engaged at G05.1 Q3 and is slower than G05.1 Q1 R1 which is the fastest mode. We didn't have time to try the 30 minute program in aluminum due to working with probes all day after the tech left.

We ended up calling Renishaw and were told the locked program 9726 probably controls the feed value in IPM and 1.18IPM or 30mm per minute was compatible with machines about 25 years old, and that's where it came from. They said the OI should handle 4IPM, but might work up to 40 IPM for touch feeds at .00008" accuracy. They mentioned that they have newer software called Inspection Plus for Fanuc Control, and it gives the user the ability to change the touch feedrate and also has programs for optimizing settings for productivity.

They said the Mazak probably has the inspection plus software. We asked for a quote of the inspection plus. We're hoping to see if it can get the DNM5700S to par the 2011 Mazak probing time. The inspection plus will apparently make the probing language identical for all the mills on the floor- Mazak, Hurco, and Doosan.

We did a file compare and figured out what the G05.1 AICC AI contour MCS (machining condition settings) optimized values were. We also integrated the purchased Renishaw inspection plus with Supa-Touch technology for optimizing probing cycles with an automation - macro program. The probing cycle dropped from 3:08 to 1:55 on the part. That saves 2:26 per hour on the process. We had gotten kind of exhausted from re-calibrating the probe GUI all afternoon in an effort to optimize manually prior to purchasing renishaw inspection plus, and our probe cycles the last few days were actually 4 minutes prior to that due to our leaving the probe GUI calibration in an un-optimized state.

The optimization of the MCS settings brought the run time of the original 36 minute program with probing to 27:30 with the probing. We finally beat the 2011 Mazak both in probing speed and program run time. We were able to quit turning off G05.1 for roughing to save time, and now run G05.1 Q1 R1 for roughing, and the motion is smoother and more controlled, at a cost of only 4 seconds per minute. With that control, we probably could increase the feed rate to eliminate the 4 seconds per minute, making it a total wash with obvious benefit to the machine in G05.1 with acceleration and deceleration happening more smoothly.

6 minutes 47 seconds of the optimization or 18% of the time saved from the 36 original minutes, were in modifications to machine and probing setup. The other change was due to process optimization of the program for the part.

The changes represent 18.54 more minutes of work in a 60 minute hour, or about 31% of efficiency improvement, 80% of which (in total 25% improvement) came from machine and probing optimization.

On the lathe side we learned that the lathes do not have the function MCS, so we need to pursue knowledge of the cost of that for the sake of pursuing the ability to use our R settings in the lathe milling operations. We want to capture some of the process benefits on lathe, even though we understand they are probably more like 6 minutes theoretically per hour of efficiency gain.