MAZAK 655/80N M-PLUS CONTROL POSITION ERROR

We have an older (@1990) Mazak 655/80N mill with a M-Plus control that has had two very odd crashes in the last 4 months on two different part programs. The mill runs 8 dedicated ISO programs and the operators only load parts, select the desired program and cycle start.
1: 9-15-2021 The first crash happened when the drill missed the target position and drilled into the workholding clamp. Operator hit Reset before I got there so I could not see the program status.
We G10 the work coordinates and they were correct and there was no corruption of the program (o4551) file.
Investigation revealed that the control had changed the sign of the X coordinates from positive to negative causing the the tools to attempt to engage the part in the wrong location. Programmed position X2.375 Y22.627 Actual Position X-2.3749 Y 22.627
The damaged components and tools were replaced, the mill power cycled and re-homed and it ran with no issues until 1-05-2022.

2: 1-05-2022 The program (o110488) was started and it appears the tool went to the first target X-Y position and the correct Z rapid plane then moved minus-X until it over-traveled. Operator Reset again.
Work coordinates were correct and the program was not corrupted.
Programmed position X4.391 Y-21.500 Actual Position X-25.006 Y-23.6875 X at over-travel Y unknown coordinate that is not in current program.
The damaged tool was replaced and the mill was homed then shut down for tooling repair.
The mill was powered up and homed this morning then the same (o110488) program was run with no issues and it has continued to run trouble free so far.

This appears to be some type of control processor or memory failure and I have a call in to our service provider and hope they will be able to identify the problem and get it repaired. Has anyone had this same type of failure and if so what was the cause and solution?
Thanks,
Steve

I guess one question would be... Do you power the machine off or does it always stay on?

It seems like this is a memory issue. But that can come in many forms. One is a bad memory or associated logic (such as a buffer that is iffy). That also includes things like perfectly good chips but a connection that could be getting iffy. As an example, this is an old machine that lives in a warm environment... those electrolytic caps could be leaking onto some tiny trace on the board, and may have corroded the trace. Extending that thought a bit, often terminating resistors were on signals (be it address, control, data, etc). Maybe one of those has been lost and you are getting noise on the line. That could cause a latch signal to falsely trigger, data to bounce, etc.

I guess the point I'm trying to make is that when you think "memory", that doesn't necessarily mean a faulty chip. The issue (if in fact this even *is* a memory problem) could be mechanical in nature too.

Just some thoughts to consider.... Unfortunately the problem is happening infrequently enough that it is hard to A/B test with other boards and such. This transient stuff can be a real PITA to diagnose.

Alan

Alan,
Thanks for the response. The machine is powered down daily but operates in a pretty rough environment with a lot of welding. The weld fumes get into the control cabinets and coat everything with a brown residue that is probably a least a little conductive.
Mechanical failures (connections/traces) like you mention seem very likely and the intermittent nature sure won't help with getting this solved. I should have some input from our repair service this week and will post progress.
Steve

I'm curious what the repair tech says....

The question about the power down was mainly going down the path of whether or not your battery is up to snuff. I personally think that you should be changing these out every two years or so. They are cheap enough that they are essentially a commodity, and a NiMH battery on Amazon is dirty cheap. If you were to buy the one Mitsubishi sells, it will cost a lot more, but even that is cheap compared to the downtime, fixture repair, etc. So, the calculus still comes back as, "a battery is cheap."

I wish we had better knowledge of what Mitsubishi's design is like. Specifically, if there is memory corruption, then I am wondering if they have a checksum validator that would catch this error. I am driving towards an error in the actual program here. Put differently, in the first error where you said the sign was reversed, did you see that in the *code*, upon visual inspection?

In IEEE-754 floating point format, the difference between two identical numbers, except sign, is just a flip of the sign bit. So:

2.375 -> 0-10000000-00110000000000000000000
-2.375 -> 1-10000000-00110000000000000000000

You probably get where I'm going with this. All it takes is one bit and things go south pretty quickly. Even if the memory was fine, though, when the planner read the target location, if there were noise on the line (or whatever), then it could have been written into the internal registers as the negative version of the desired value.

It will be interesting to hear if the service person recommends cleaning the boards. If so, I would argue that you should also clean the backplane into which the boards are plugged. Unfortunately, the backplane is going to take the most amount of time because it will require the NC unit itself to be removed from the machine. The cleaning part isn't too difficult, but you'll get a lot of differing opinions on how to do it. Put simply, I'm going to stir up a huge debate with the next few comments.... A manufacturing shop probably uses a high VOC type of liquid in an ultrasonic bath. If you don't have access to that (who does), many will uses something like distilled water and Simple Green to cut the grease, followed by a very good rinse. Where this gets more tricky, however, is when there are components that are not hermetically sealed. Frankly, you are not going to damage an IC if it were submersed and cleaned, so long as it is fully cleaned and dried. But a switch could be a different matter.

To be absolutely clear... I am NOT advocating you clean your boards. I cannot accept that liability, etc. You may royally, ahhh, mess them up! But, I do think that the environment in which your boards are living could well be a problem.

Power is also a possibility, by the way. If your system is as old as you say it is, the electrolytic caps are probably shot on it. Their life is a finite thing, which many do not realize. I don't know when your machine was built, but I'm guessing that an M-Plus is probably a 90s vintage machine? That would make it 30 years old... I just repaired two servo amplifiers (S11 and S12). Although the S12 had *lots* of problems on it, part of what I did was to replace the electrolytic caps. A large number of them were actually OK, but a surprisingly large number were also shot. A few of these didn't even register as capacitors anymore when I tested them, and even many of those that did register, had a series resistance of 800 ohms or more. That isn't good.... The point is, I'm wondering if this could be adversely affecting your system too.

Please do let us know what your service guy says. I'm trying to suck this stuff up into my pea brain. :)