Kimberly-Clark Latin America manufactures dozens of stock-keeping units in varying quantities using a few machines, with sequence-dependent setup times and stochastic demand between planning periods.
Because the same SKU can run on multiple machines of different efficiency, and demand shifts each period, manual lot sizing and scheduling were inefficient. The team formulated an integrated inventory, production-planning and detailed-scheduling model that solves the industry-size NP-hard problem in minutes.
The solution and its approach were tested extensively and deployed in five countries, replacing manual plans.
Solve lot-sizing, scheduling and inventory as one problem
Integrating inventory, production and scheduling captured effects each silo ignored.
The solution method solved an industry-size NP-hard problem in minutes, making it usable.
Sequence-dependent setups were handled explicitly rather than averaged away.
Stochastic demand was addressed by re-optimizing each planning period.
When the constraint is shared machines and setups depend on sequence, lot-sizing and scheduling must be solved together or each is locally wrong.
The model reduced finished product inventories by up to 45%, raised yield and uptime by 2%, and improved service levels by 2.4%. It was a 2010 Daniel H. Wagner Prize finalist and a 2011 Interfaces paper.
FOLLOW THE EVIDENCE
The sources
- Kimberly-Clark Latin America builds an optimization-based system for machine scheduling Penn State Pure · pure.psu.edu
- Kimberly-Clark Latin America Builds an Optimization-Based System for Machine Scheduling Semantic Scholar · semanticscholar.org