Kanban for Shared Resource
By Jean Cunningham:
Normally when I think of Kanban, I think of external Kanban between the manufacturing cell and the external supplier. However, kanban within the plant when there are shared resources is very common too. (At least until all equipment/resources are able to be dedicated to product lines.)
For shared resources using Kanban we had to figure out a way to prioritize different kanban signals coming to the shared resource. The method I am most familar with is to segment the time available for the shared resource into time slots that each product line could utilize. This worked really well with a paint line, or pallet building area.
I would like to hear from some of our readers on visual methods used to schedule a machine shop which is shared by multiple assembly product lines. While we would like to have dedicated equipment, the shop isn't big enough yet; therefore we have to find a good way to help the machine shop to prioritize the various signals coming to them.
What have you tried successfully?
What do you think? Please scroll down (or click) to post a comment. Or please share the post with your thoughts on LinkedIn.
Don't want to miss a post or podcast? Subscribe to get notified about posts via email daily or weekly.
- Kanban for Shared Resource - January 10, 2007
- Initiating Lean with 5S? - January 9, 2007
I’ve previously used and help set up a production kanban system for a tier-1 auto supplier, this is commonly called a “supermarket” system. Rather than “scheduling” a shared resource (such as a plastic injection molding machine), you decouple the molding machine from the customer assembly lines with a supermarket (yes, inventory!).
A few keys points — you have to determine a lot/run size for the shared resource, based on the # of parts, changeover time, demand, and run time per part. You have to use SMED to get your changeover time down so you can do smaller run sizes.
Given the run sizes and all, you can estimate the lead time to replenish the supermarket with a given production lot. Then, you can also set up the supermarket with a calculated “re-order point” (based on demand and expected production lead time).
Once this is all set up, the shared resource simply responds to the production kanban cards. Ideally, the cards should flow in a FIFO manner. “Scheduling” the resource means, in this set up, planning how many hours a day you have to run the resource, rather than planning a precise MRP type schedule.
I don’t know a good reference book for this method offhand, as I learned it on the job. Feel free to email me with questions.
I have been working with a machine shop that supplies assembly cells in a make to order environment. There are some components that we can run to a supermarket, but other items are only made to order. We have set kanban sizes for the components that go to the supermarket and we use a “variable” kanban quantity for those components that are only made to order. This is to use the same ordering and priority control for all products made at the machine. We have used colored totes to denote product for each assembly cell and we use photo id and data id signals (like pharmacies use to verify the pills in the bottle)with the totes to specify the product needed.
We are still working to improve the sequencing, and the priorities are sometimes confused. We have eliminated lost or misplaced parts and we are getting better at planning our lead times on the shared machines.
Throughput has gone up and lead times and inventory have gone down, but this will require a long term commitment to continuing improvement. It is simple, but it hasn’t been easy.
I recently managed a plant that redefined “high mix” by manufacturing engineered to order products, with over 10,000 customers a year whose average order quantity was 2-3 pcs. We supported this production with a machine shop made up of primarily turning centers and CNC mills. Where we could (by product family) we established point of use supermarkets (like those already discussed), but this only scratched the surface because so much of our product was engineered to order. To deal with these we took our best shot at a real heijunka system. For example, the turning centers worked to a schedule governed by bar dimension so they would plan regular changeovers. Orders were grouped by dimension and stored together in slots and run during the scheduled capacity time slot, going into the future by a pre-determined time fence (typically a week). I would have preferred FIFO, but complexities and lead times varied widely. Caution: The biggest enemy to this system was expediting…when we allowed sales, engineering, etc to come in and expedite the shop it would all fall apart. It takes a lot of discipline…