How Do You Not Bottle Neck A CMM?

[acgarcia]

Anyone here use a CMM for in process inspections? For multiple CNC machines? How do you overcome bottle-necking the CMM?

At my sister plant, the CMMs are used in cells so they check the same part over and over with no variation. Each CMM has a probe calibration program and maybe 1 or 2 inspection programs with dedicated fixture setups. We reduced the number of hand measuring tools to a minimum and the CMM determines pass or fail. Pretty simple and straight forward. The purpose for the CMMs are to eliminate the need for the part to go to the QC area to be inspected by hand. The CMM is QC so once is passing report, it moves on in the cell to the next operation. 1 CMM per cell.

Now at my currently location, this 1 CMM is suppose to cover multiple CNC machines that make a large variety of parts in various quantities. The CMM can already run a couple of parts however I haven't fully released it for production use just yet. I need to make it full proof so I don't get called in the middle of the night to come fix something simple. Some of the things i'm doing is making 3D printed fixtures to eliminate the need for random standoffs. I'm trying my best to create multi part fixtures so i can hold a part family in 1 3d printed fixture to try and keep the number of fixtures down as well. I've added operator comments with pictures on where to place the 3d printed fixtures on the baseplate for every part program. I'm also using pcdmis to determine which operation the part is currently in using the onerror command to determine which features are finished on the part because some parts will see the CMM twice. This way there are less programs to scroll through.

The CMM is centrally located in the shop so machine operators will have to walk back and fourth to run their parts. Now i'm worried about when 2 or more operators meet at the CMM to check their parts. People are going to have to wait their turn thus creating a bottle neck at the CMM. Since we plan on using the CMM as QC, they will need the CMM report to make machine adjustments. Now we are also slowing down production of parts because the CMM is a bottle neck.

What are some methods to solve this or reduce this to a minimum?

[louisd]

To effectively answer your question, you would need to conduct cycle time study (studies). This would encompass the time it takes for operator to load parts, execute routine, remove parts and assess results, giving room for next operator/part to repeat.
From there, you could break down each step in the process and grab at whatever is the lowest hanging fruit, in regards to time savings.

If it takes an operator 5 minutes to setup part on granite, you should look into standardizing or streamlining the process. I've seen people "barcode" fixtures to measure a diameter, cut to a specific size at a common location for all fixtures. if diameter is X, subroutine X will run checking Part X...If diameter is L, subroutine L will run, checking part L. each fixture would have a unique diameter for each unique part config. this would eliminate the time it takes to find and open the correct routine. operator just sets part on posts (could even use magnetic posts) that have two sides to them fenced off, then hits go. no more time consumed looking for the correct routine to open.

If it takes operator 10 minutes to physically print and review the results.. Maybe you could digitize the data, log it on a network location, and add code to alert operator of any OOS dimensions at the end of each routine? You could potentially bring this 10 minutes down to just 2.

Other tips relating exclusively to code improvements to increase the actual routine cycle time: increase movespeeds, turn on flymode, reduce prehit/retract distances, reduce numhits, use fastprobemode (if scanning head), automate input prompts by implementing barcodes that operators would scan. make barcode cheat sheets for commonly used scans (yes/no,OP#, etc).

[acgarcia]

I use a generous prehit/retract when doing the alignments and then move the prehit/retract down for speed. The only prompt I use is to scan the barcode serial number.

[KIRBSTER269]

Depending on the material being machined, first piece needs to be verified. After everything is verified. I myself get the machinist as close to nominal as possible, technically if everything is close to nominal, I then can verify every 3rd, maybe 5th part, this depending on the material and tolerance, is the tolerance tight?, is the material titanium, or aluminum? etc. This will effect tool wear. Even if a customer requires 100% check required, you can do this at a different time. Technically the machinist is supposed to be checking his or her parts, not relying on the CMM. The closer you get them to nominal, the less you have to worry about the machinist wondering off of their tolerance.

[acgarcia]

material is typical steel 4140, handful aluminum parts, waiting for some silicon nitride probes to come in before I start those. Tolerances range from ±0.015" to 0.0005". I usually don't measure the 0.015" stuff because that feature won't affect the parts function most of the time.

[KIRBSTER269]

I was just using the types of material as an example of what would be revelant or not. The material will determine tool wear and if you need to worry about little change in the tolerance or possible a major change.

[acgarcia]

KIRBSTER269 I was thinking of getting pcdmis NC and using the machine probe for measurements and the CMM as the validator. That way we can work on how many parts we can skip going to the CMM.

[BKulpa]

Do you need to check all your features 100% on each program.
Typically here we run the full program at start of shift then have flow controlled short version for additional checks during the shift and at tool changes.

[kingsld1]

Thinking out loud....

Probably way expensive to implement, but ......


Get a fixture plate and standoff set for each Machine.
Make someone responsible for setting up the CMM fixture for their machine. Some kind of sign off procedure.
Figure out out to locate the fixture plate in the machine in the place every time. May have to get creative if you wind up with multiple sized fixture plates.
Have the operator clean and load their part in the fixture and then load the fixture onto the CMM.
Write the programs so that you don't have to do a manual alignment each time. Just work off of a corner of the fixture.

This should eliminate setup time at the CMM. Should make it load and go.

Write your programs so they will go as fast as possible. Use the optimize options in the software if it's available in your version.

[dwade]

Buy another CMM!! We have 6 CMMs in the lab and all they do is run production parts. We are in the process of speeding up the existing programs movespeed (probes changes and long moves around the parts), prehit and retract are big factors. Reducing probe changes is a big contributor in program run time. We've got a family of parts that is machined in at least 3 different operations and each operation requires CMM measurement. In 2 of the operation programs we have reduced CMM cycle time to 10 minutes from 12 and 13 minutes, while that doesn't sound like much when you run 200+ parts across the CMMs it shows that you are saving a lot of time and $$.

[Schlag]

Inprocess programs. Dont check 32 /32 holes in your bolt pattern. Check 2. Simple things like that. Once the part is "running" on the CNC and everything is verified, all of the subsequent inspections can be more inprocess. Check a few features / locations on each CNC offset.

[acgarcia]

Well yes buying another CMM would help out. I have 5 at my other plant and this is the first 1 at my current plant. (7.10.7 indexing tough trigger).

Also we have not CMM operators as in we don't pay people to sit at the CMM and watch it go beep. All CMMs are place and go. All responsibility of placing fixtures and selecting programs is on the machine operator.

For the most part, we already only check the minimum critical features and not 100% of the part.

I like the corner fixture idea. I should establish that now since I don't have very many programs on there yet. At least the machine operators would know to place the corner of the 3d printed fixture in the same stop no matter the size of the part.

[dwade]

At a previous employer we had 8 B&S CMMs and 1 Zeiss Contura all being operated by machinists. We used R&R Fixtures (now part of Renishaw) for 99% of all of setups, most were quick and fairly idiot proof which was necessary (same as your needs) because every Tom, Dick and Harry on the shop floor was supposed to be able to read the setup instructions and run the parts. It can be done and you will need to implement good instructions either embedded into the program or in a BIG notebook.

[Ego Murphy]

To effectively answer your question, you would need to conduct cycle time study (studies). This would encompass the time it takes for operator to load parts, execute routine, remove parts and assess results, giving room for next operator/part to repeat.
(snip)

Buy another CMM!!
(snip)

Clear path right here. The numbers found from following louisd plan will either make the business case for buying another CMM (or 5) as dwade mentioned, or at the least some nifty quick-change fixturing as kingsld1 brought up.

The most important thing is to present everything to mgmt. in terms of dollars, as this is the only language they understand. Things that actually translate to dollars, such as "efficiency" or "less setup problems" are completely meaningless and confusing to the pointy-haired types - best to do the translation beforehand.

[CodeWarrior]

Clear path right here. The numbers found from following louisd plan will either make the business case for buying another CMM (or 5) as dwade mentioned, or at the least some nifty quick-change fixturing as kingsld1 brought up.

The most important thing is to present everything to mgmt. in terms of dollars, as this is the only language they understand. Things that actually translate to dollars, such as "efficiency" or "less setup problems" are completely meaningless and confusing to the pointy-haired types - best to do the translation beforehand.

I agree with Ego Murphy, when I was going through this process I had to break down everything to return on investment and time.

For example, I started with a TTP and our programs took about 6-8 minutes to pick up the datums (lots of diameters) before measuring the applicable features. So I told them we could increase the CMM throughput if I had a scanning head to take this timing down to a couple minutes or so. I made the case that we wait 20-25 minutes to approve first pieces for production and could reduce that by at least half by buying the scanning head. Also, there wasn't the need to catch up on things on third shift so they didn't need to train and pay an operator.

[CodeWarrior]

For me I broke this down to 3 categories:

1) Set-Up -> Repeatable fixturing, I use a rayco plate and multiple magnet chucks
Programs are all parametric with a barcode system

2) CMM Runtime -> Optimizing the CMM by using the best hardware - upgraded from ttp (have a vision system on my wishlist)
Optimizing the programs by reducing unnecessary hits, move/all (be careful), less wasted moves around part, etc.

3) Data Management -> I use scripts, batch files, and excel macros to make the report process automated. The operators have a pass fail report that shows green, red, and purple for values close to the edge of being in tolerance so its as easy to digest as possible. Also, everything is saved/formatted in excel databases in the background

[ProbFlex]

At SpaceX, because most of the parts are big. Most of our in-process inspections are with a Romer Arm. doing the inspection there on the CNC. Help with not having to break down the CNC set-up in the machine in case it needs to rework or re-align on the machine. At final, the CMM would only inspect Critical features for most of it would be already inspected. then you also have go-no go gage and auditing inspection. that helps out.