Beginner to Robot Automation - Looking for Pointers

Context/Background at bottom of post.

I'm looking for some pointers from the experienced on rearranging a multiple-HMC (each with pallet changer) cell to have automated loading/unloading of parts. I'd like to get my ducks in a row to be able to propose a solution to management and estimate the costs and ROI (no, not concrete numbers, just a forecast) but I'm having trouble getting a good start. I've looked into various models available, capabilities, and such, but I don't know much about the specifics of integrating them into the machines.

I'm also trying to figure out the most economical/efficient approach to loading/unloading of these parts, such as what "end of arm" grippers and what not. Right now, it seems like a table-mounted arm between/adjacent the machines would be best.

It's such a big topic, I'm having trouble figuring out which bite to take out of the elephant first! Any help would be much appreciated.


------------

Context/Background:
We have a few high volume customers that keep machines dedicated to running their parts into the foreseeable future. They run in a 3 machine HMC cell, though it could be down to 2 HMCs and a little hand finishing at the end (that won't be able to go away)

I'm looking into redesigning the fixturing to better utilize the machine volume and make the loading/unloading a little faster and more importantly, much more ergonomic and comfortable for the operators.

This is an operation ripe for automation, imo, and since I'm redesigning fixtures and workholding anyways, it's no thing to accommodate this.

They are Okuma MB5000 Horizontals. Parts are aluminum, ~.12 thick iirc, and about 1/2" x 7, 8, or 9" length (different variants)

Ok, in hindsight, thinking of "how do you eat an elephant" I guess a simpler question could be - what're the best few bites to take when beginning this approach?

Some questions:

What is the heaviest part weight?
Can the 3 be arraigned at 90° (or less) angles or do they need to be side-by-side? (floor space and 7th axis question)
If they can be 90° or less, do you still have room for the robot feeding / finished goods automation?
Do you need to run the different variants at the same time? (i.e. one variant type per machine - and will doing this help with your ROI regarding WIP reduction - scheduling / process flow)
Do your fixtures need to rotate when being loaded? (i.e. be double/tri/quad sided)

Schunk grippers are my go-to for robots. Not cheap, but they have about any configuration / size / grip force you need and are dead reliable, we have thousands of them in operation in our facility and get millions of cycles out of them. Get a catalog, the website isn't the best when looking for something that might work. Gripper arms are going to be your responsibility. Do the arms need to change when you change part variants or can you design one arm to fit them all? If they do need to change, do you want to change the whole end effector automatically, or have someone manually change just the arms?
What is the cycle time of each machine? What is the volume of parts per load? What else does the robot need to do with each part? (i.e. deburr, inspect, etc?)
How will the parts be presented to the robot?
How will the robot dispose of the finished parts?

These answers will help us point you in the right direction.....

Thanks for the assistance. I figured some of those questions were relevant but didn't want to just make a wall-of-text up front.

I see no reason the HMCs can't move if necessary. I assumed it will be, as the two main machines are currently opposite eachother and some 10 ft from door-to-door, give or take. The third, which is not always part of the process, is at 90degrees to the pair.

Current fixturing has a fixture plate on two opposite sides of a tooling column. Cycle time on the fixture is ~20 minutes iirc. I'm in the data gathering process. I've done most of this months ago, before the project was back-burnered, and am having to relocate my notes.

Parts are 6061 Al and at weigh a few ounces.

The machines do not need to run multiple products at one time, and I'm unsure of the frequency of change-over. They use the same fixtures, though, currently.

Each part is approx 0.4oz / .01kg - many are machined per fixture. 12 per side, right now, I believe.

One gripper will be able to handle all the parts. Approx half the length is identical for all parts. I do not believe automatic end effector changing is necessary. Manual changes would suffice, as it'd be rare. I don't actually foresee needing to change at all while in this cell.

Before the parts go into the machining fixture, they are stamped blanks that have undergone a deburring via tumbling. Those blanks are then manually stacked and rubber-banded in set batch sizes to transfer to the HMC cell. That delivery will have to change obviously. The blank is asymmetric so would have to be sorted for orientation. That can be done manually, Is suppose,

Finished parts will need a hand-polish step afterwards, and are delivered individually from machinist to polisher, currently. A metal slide could be incorporated to deliver the parts to a pickup point or bin for the polisher to work from, so the robot can drop parts onto it, and have them conveyed outside the cell. I'm just beginning this process, so have mostly rough thoughts on it.

Dimensional inspection is infrequent, with one critical feature being inspected 100%. The critical feature will be human. The dimensional inspection is done with a visual/optical machine that can have fixtures located accurately... but I would not look to incorporate inspection (at this time)

I've got a copy of RIA R15_06_1999 which I know is out of date, but it was what I could find via google, and should help me keep the other important stuff straight.

Again, thank you for the help.

When I had Fanuc factory training in Detroit, they had a bunch of cam lines they were building for GM, and the robots ran back and fourth on I-beam above the machines. It's just an extra robot axis for the traverse, which is dirt simple to do. They were actually mounted upside-down (which is a setting you change in many robots so that the torque alarms can account for gravity in the proper direction). This lets one robot serve as many machines in a line as it needs to... pick from bin, clean, mount, grind, inspect, grind some more, inspect again, package on skid, etc.

Machine vision is a mature technology, and end of arm machine vision has been done for decades. If the parts tolerate it, they could be all put in a bin and the robot will pick them up and orient them properly. No big deal. Robots pick pill bottles from big gaylord type boxes and load them on the lines.

mdshunk said:

When I had Fanuc factory training in Detroit, they had a bunch of cam lines they were building for GM, and the robots ran back and fourth on I-beam above the machines. It's just an extra robot axis for the traverse, which is dirt simple to do. They were actually mounted upside-down (which is a setting you change in many robots so that the torque alarms can account for gravity in the proper direction). This lets one robot serve as many machines in a line as it needs to... pick from bin, clean, mount, grind, inspect, grind some more, inspect again, package on skid, etc.

Machine vision is a mature technology, and end of arm machine vision has been done for decades. If the parts tolerate it, they could be all put in a bin and the robot will pick them up and orient them properly. No big deal. Robots pick pill bottles from big gaylord type boxes and load them on the lines.

Click to expand...

Some units also require moving the grease zerks and vents when changing orientation.

Yes machine vision works, but if you have the parts in control keep them there.

mdshunk said:

When I had Fanuc factory training in Detroit, they had a bunch of cam lines they were building for GM, and the robots ran back and fourth on I-beam above the machines. It's just an extra robot axis for the traverse, which is dirt simple to do. They were actually mounted upside-down (which is a setting you change in many robots so that the torque alarms can account for gravity in the proper direction). This lets one robot serve as many machines in a line as it needs to... pick from bin, clean, mount, grind, inspect, grind some more, inspect again, package on skid, etc.

Machine vision is a mature technology, and end of arm machine vision has been done for decades. If the parts tolerate it, they could be all put in a bin and the robot will pick them up and orient them properly. No big deal. Robots pick pill bottles from big gaylord type boxes and load them on the lines.

Click to expand...

Yea, I've checked out a couple places that use them in exactly that fashion. While I'm not discounting anything right now - I have doubts that it'd be the most economical approach to this application.

Oh, and yea, they used a vision system as well that was pretty nifty. It was mainly used for packing and retrieving parts, though. Not precise. Mainly looked to see if a space was occupied or empty before grabbing or depositing. But we went over a few vision systems they were using in other areas, as well as some of the things they really wanted to get but couldn't justify the cost. I know enough to know how much I have yet to learn

I'm pretty eager and excited though. It'll be a lot of work and research and require some formal training, certainly, but this is one aspect I've not dived into deeply before that I always wished I could. I'm being encouraged to look into it now, and want to take this opportunity for all I can.

JNieman said:

They are Okuma MB5000 Horizontals. Parts are aluminum, ~.12 thick iirc, and about 1/2" x 7, 8, or 9" length (different variants)

Click to expand...

Can you repurpose the MB5000s for other jobs?

This would be perfect for your parts:

Brother Speedio with Flex-S Robot:
flexsdemo - YouTube

You could be money ahead on capital costs alone by buying 2-3 of these instead of an elaborate 3-HMC robotic retrofit, and you'd have three horizontals completely free to run other jobs. Not to mention, Speedios will smoke just about any other machine on lightweight aluminum parts and hardly take up any floor space or electrical panel capacity (30A @ 208V).

This is coming from a guy who swears by horizontals for a lot of work.

1st problem is figuring out how to turn the fixture during the load / unload process. This could be done by retrofit of a servo axis to the outside table or it might be done with the robot. This extra axis would be controlled from the automation PLC but I would figure $10k per H for that option alone. (Servo Motor, Drive, Cabling, mechanical design / mounting, etc) We did this on some H-400's years ago for robot loading and it worked very well, but you most likely will have to design all of the mechanical interface yourself and getting prints out of some MTB companies for things like this is next to impossible. You'll also have to have a pneumatic/hydraulic release for the rotation lock, if equipped for either method.

If the robot is large enough, you can just design something into the end effector and fixture to allow the robot to rotate the pallet if you have enough clearance inside the work area. Or go with some other mechanical / pneumatic indexing solution.

Your application doesn't require a big robot from a payload standpoint, but it will require a larger robot for reach considerations.

You will have to decide on whether to put the robot on a 7th axis or not given 10' between machines. If you don't use a 7th, you will need to move the machines so they are within the reach radius of the robot. It might be cheaper to just go with a 7th axis and leave the machines as they are presently located. You don't have to buy the 7th axis option from the robot vendor, you can make it yourself and use your automation controller to move it via interface with the robot. (We can build one cheaper than buy one from a robot vendor, but this depends on your experience level and abilities - in your case it may be cheaper to buy one with the robot.)

Out of the 5 robot makes I have some experience with to date (Fanuc, Nachi, Kawasaki, ABB, Ries) Kawasaki wins hands down as the most flexible, easiest to use / program, feature rich and has also been the most reliable in 24/7 production. They also have compact controller options that helps tremendously in packaging and floor space considerations.

As Miguel noted, if the parts can be pre-oriented/positioned upon presentation to the robot, go with that method over any other - without pause. That is by far the easiest to maintain and deal with in a production environment. We use cameras quite often for part location / orientation picking, but you have to have a certain level of expertise with both the robot and camera / lighting system and the interface between them. We are blessed in this regard, but most facilities aren't.

I would suggest a Kawasaki RS30 or RS50 on a 7th axis as a start for the robot. If you are re-doing fixutres, I would suggest always making the fixture value of parts an EVEN number. This lets you load / unload parts in pairs or more. With 20 minute cycle times per machine and 3 machines, you are looking at a ~6.5 minute load/unload time per pallet. Traversing to drop / get parts will cost some time, so the more you can take with you per trip the better. You will have the payload capacity due to reach, so take advantage of that.

Don't forget about things like fixture blow-off, etc that will need to be done.

Orange Vise said:

Can you repurpose the MB5000s for other jobs?

Click to expand...

High confidence that we could, sure. I might get some numbers and see about using some of the faster tool-change, faster-rapid, higher-rpm, niche machines like these and see if the payoff is there. Thanks

I would suggest a Kawasaki RS30 or RS50 on a 7th axis as a start for the robot. If you are re-doing fixutres, I would suggest always making the fixture value of parts an EVEN number. This lets you load / unload parts in pairs or more. With 20 minute cycle times per machine and 3 machines, you are looking at a ~6.5 minute load/unload time per pallet. Traversing to drop / get parts will cost some time, so the more you can take with you per trip the better. You will have the payload capacity due to reach, so take advantage of that.

Click to expand...

Yea, the goal of the fixture redesign is to get run-time per pallet increased by adding more parts via better use of space. That would also give me more time for the robot to work before a pallet change is ready.

Your point on turning the pallet is something I was thinking about earlier - I assumed retrofitting a motor and controller was going to be the best option but I'll figure that out.

I guess I need to get a more accurate layout of the machines and compare an in-situ solution with one requiring machine relocation. Relocation wouldn't be a terrible ordeal, though.

I think the biggest hurdle is figuring out the best fixture design with all these variables in play. It'd be worth remaking punches for our die that stamps the blanks if extra geometry would make clamping or workholding better.

My main concern is the inefficiency of handling single blanks one at a time (or two or four even) when you might have ~32 on each side of a fixture that might end up being 4-sided. Machine cycle time will increase as well, but it seems like a lot of small repetitious movement for a robot. Would defeat the benefits if I required a worker to load and unload fixtures just for the robot to put them on the tooling columns.

I might have to solve that hurdle first before even worrying about the robot details.

JNieman said:

My main concern is the inefficiency of handling single blanks one at a time (or two or four even) when you might have ~32 on each side of a fixture that might end up being 4-sided. Machine cycle time will increase as well, but it seems like a lot of small repetitious movement for a robot. Would defeat the benefits if I required a worker to load and unload fixtures just for the robot to put them on the tooling columns.

Click to expand...

This is a fairly straight forward time study.

JNieman said:

My main concern is the inefficiency of handling single blanks one at a time (or two or four even) when you might have ~32 on each side of a fixture that might end up being 4-sided.

Click to expand...

High density and robotic tending generally don't mix. You only have so many circuits available for powered clamping.

Orange Vise said:

High density and robotic tending generally don't mix. You only have so many circuits available for powered clamping.

Click to expand...

There are ways around that.
Field IO can greatly expand your options and is inexpensive now.

Miguels244 said:

There are ways around that.
Field IO can greatly expand your options and is inexpensive now.

Click to expand...

Yea, you can mount a Festo CPX up on the arm and have 12 valves and 24 inputs with only a 10 mm air line, 24V supply and an industrial ethernet cable going up the arm. The Kaw RS series is available with 4 valves and 12 inputs within the arm and the 30 & 50 size have an additional 16 conductor cable IIRC routed internal through the arm for user use.

JNieman said:

I might have to solve that hurdle first before even worrying about the robot details.

Click to expand...

Definitely need to figure out how you are going to feed the beast before bringing him home. You might need to marry the two concepts as we've done on some systems. Use a camera for fine orientation and let a cleated conveyor or dunnage serve as the rough alignment / separator. Something like a plain-Jane roach flatbed conveyor that they put a single layer on, with the correct side up, which feeds off to a flexlink type conveyor to separate them into a single lane, end to end. They can pretty much bulk-load the table with 100-200 or more parts turned a certain way. When it gets to the end of the flexlink conveyor inside the cell, a camera can fine align the robot for picking.

Tonytn36 said:

Yea, you can mount a Festo CPX up on the arm and have 12 valves and 24 inputs with only a 10 mm air line, 24V supply and an industrial ethernet cable going up the arm. The Kaw RS series is available with 4 valves and 12 inputs within the arm and the 30 & 50 size have an additional 16 conductor cable IIRC routed internal through the arm for user use.

Click to expand...

And available out to IP67 in metal cases.
CPX is a pretty bad-ass platform.
Fair disclosure, I used to work for festo.

They have other less expensive platforms for remote io as well.

Miguels244 said:

And available out to IP67 in metal cases.
CPX is a pretty bad-ass platform.
Fair disclosure, I used to work for festo.

They have other less expensive platforms for remote io as well.

Click to expand...

We use a ton of CPX units (IP67, Metal). It reduced field wiring costs tremendously, is robust and not that badly priced for what you get.