Coca-Cola uses emulation to improve throughput
Coca-Cola Enterprises (CCE) is the world's largest marketer, producer, and distributor of products of The Coca-Cola Company. For high volume products, such as Coca-Cola Classic 12 pack, orders are often found in 'stock' quantities. In other words: a full pallet of a single stock keeping unit (ie. 96 cases).
Coca-Cola Enterprises (CCE) is the world's largest marketer, producer, and distributor of products of The Coca-Cola Company.
For high volume products, such as Coca-Cola Classic 12 pack, orders are often found in 'stock' quantities. In other words: a full pallet of a single stock keeping unit (ie. 96 cases). These orders are fulfilled by forklifts retrieving full pallets from stock. At the other extreme, low volume products, such as sodium free club soda in an odd package size are found in single case quantities, and are hand picked from pallets on a picking floor.
Mid-tier products are often found in between these quantities and may be picked in 'layer' quantities. A layer is a full tier quantity of a single product. For example, a 96 case pallet of can product is built with 12 layers (or tiers) of 8 cases each. So any case quantities divisible by 8 is a candidate for layering picking.
In a large number of our sales centers, CCE has implemented forklift grippers that are capable of picking up full layer quantities of our core packages. Up to 25% of our order volume can be handled by these layer pickers.
Pickers think ahead
The performance of these layer pickers is dependent on the skill of the forklift operator. A combination of precise hand-eye coordination, speed, and thinking are required to achieve high productivity. The thinking involves looking ahead to complete order requirements with a minimum of product movement.
Building off this experience with mechanized layer picking, CCE developed a fully automated layer picker that employs a monorail gantry robot and a layer gripper adapted from the forklift version. Complex logic is incorporated in the robot controller for deciding the best method of assembling a given 'layered' pallet. This logic was developed using AutoMod simulation software from AutoSimulations Division of Brooks Automation (Salt Lake City, Utah).
AutoMod's Model Communication Module, a tool that enables users to read and/or transfer model data from multiple models in order to 'communicate' with multiple points in a complex simulation system. CCE uses this to emulate and determine the best sequence in which to build a series of 'layered' pallets. Our intent is to use this 'Dynamic Scheduler' on the live system to achieve an additional 17% gain in throughput.
By emulating or 'testing' hardware and control software with a simulation model, users can eliminate any bugs in the system before actual implementation. Control system developers can test control software by connecting to accurate models, which provide the same responses as the real hardware, well before it becomes available.
The original CCE simulation model looked at a given pallet, and the real time inventory available on the stock pallets within the system. It then used logic to determine whether to build this pallet in a normal mode, negative mode, or combine mode (plus some variations on these basic methods). The normal mode followed the pallet requirements in lock step. The negative mode identified situations where instead of picking up, for example, five individual layers from a pallet containing six layers to create a pallet with five layers - a single layer was picked from the stock pallet, thus satisfying the requirement for five with the balance in the stock pallet location. The combine mode would create the five layers by combining two layers from one stock pallet with three layers from another, where two stock pallets of the same article existed in the system. Using the negative and combine modes improved performance by completing work with fewer moves by the robot.
Using emulation, engineers look for a sequence of pallets that create opportunities to use the negative and combine logic. The Dynamic Scheduler looks at the next five pallets which may be built, at the time to execute the five and identifies the sequence with the lowest execution time.
Once the best sequence is identified, the first pallet of the sequence is run forward from the initial condition. The ending condition from this first pallet are then captured and become the new initial conditions for the next sequence group. (Which will be the 2nd, 3rd, 4th, and 5th pallet in the best sequence, plus a 6th pallet now added to the mix.)
Emulation is providing CCE with a powerful tool for scheduling an automated system on the fly to achieve higher levels of productivity. It may well be the difference between automation which works and automation which pays.