It's quite long winded...please bear with me as I kept adding as I typed.
The Horizontal cells are quite capital intensive, so keep that in mind when kicking it around. I would follow DD's advise and do some times studies to see how much time is being spent in set up, part change out/loading, etc. to see how much your actual spindle time uptime really is, of course, some controls will have this available. I think you will be amazed how little spindle uptime you actually use. Use this information to help calculate your ROI. Makino has the best webinars on looking at horizontal cells that I have found, you just have to register, and they generally won't throw the hard-sell at you either.
When exploring the pallet pool, make sure that it is really expandable (some small pallet pools are NOT expandable), and that the machine(s) you are considering will marry up to a cell. You will also want to take a look a the cell controller software, but it seems as most of the major ones will all do the same thing, just calling it a bit differently. Don also mentioned tool capacity above, verticals are certainly limited, and we have found even 120 tools in our horizontal to be limiting. As you attempt to add part families to the cell, you find you quickly run out of tool pockets, and since you cannot move those projects into the cell, you limit the effectiveness of operating the cell.
If you were to purchase X number of machines to produce your family of parts, consider more tools than you will think you will ever want, and tool the machines exactly the same so the cell controller will place the job into the first available machine. This will keep you from having to schedule the machines based on tools available and cycle times - the cell controller will do it all for you.
Another issue to consider is how the cell is controlled. Mazak uses machine #1's (usually) CNC control as its "brain" for the servo drives in the cell, so if machine #1 is down for maintenance (or worst case a crazy bad crash), the entire cell is down. They do offer an optional, stand-alone cell controller to combat that issue. Also, if for some reason the robot loader is down, the cell is down and you will be loading parts through the machine door, but you won't be shuffling pallets.
You most likely are producing in "batches", overproducing what is needed at any point in time. With the cell set up correctly, you will eliminate set up, you can produce exactly what is required on a daily basis, and reduce raw material and WIP more easily than running the verticals. All this with less operators, and a more robust process by eliminating the set up/first article labor variable.
Tony mentioned the amount of real estate available - on a 500mm machine you can pretty easily get 4 sides of a tombstone with 15" x 25" of usable space if your parts allow it. If you need to access the sides of the parts, you may only get one part per face, or maybe only one part per tombstone (you may want to use a window fixture to get some tight tolerance parts in spec - especially if you have some tight TP features on opposite sides of the parts.) We run a family of parts that we are moving to our cell - most of them are "flat", 2 op parts, so they lend themselves to our crazy big tombstones - 20" x 20" square. We can get 12-20 parts per face depending on their size, but we have OP10 and OP20 on one face, so every pallet cycle will yield finished parts. Since all of this family of parts are used on a one-to-one basis, we will schedule the cell controller to make XX number of parts each day/week/month to meet the demand, and as long as we keep loading, the cell controller will keep cycling the pallet into the machines to meet the production schedule.
Using your times studies, you can estimate the total time required for loading parts. Take this information to determine how many load stations (work setting stations) you will need - you may find you want an extra load station to eliminate a shift. I know of a shop running two load stations for a four machine cell, where two operators can keep the machines running 24/7. This is certainly based on your parts and cycle times, but the cell controller can help determine when parts will run based on attended vs. unattended assignments.
Plan on 3-5 days of training just for getting the most out of the cell controller. If you can prove out all of your parts and fixtures before hand, then you can have the apps guy train you on the finest details of the cell controller. You will want to know cycle times of each tool on each part, production schedules for every part number, tool life for each tool, and any redundant tooling, etc. When you have all of this data in the cell controller, you will be able to use the cell to its fullest extent. For example, if your 1/2" end mill is out of tool life or it breaks (because you put tool breakage checks in your program), the cell controller can pull the pallet, place it back on the stocker and flag it for inspection, then puts in the next job that does not use that 1/2" end mill (or uses the redundant tool), and keeps making parts wile you are snoozing on Sunday morning.
There is a lot to consider and you will certainly need a project lead - you will probably also want to reprogram all of your parts to get the most out of machine and to make sure you are making robust programs. The implementation of the cell is a special project, visit other shops and ask a lot of questions of them about their methodology and what they would do different.
Steve
