Optimization of Production Planning in IBM's
Microelectronics Supply Chain
Brian Denton
IBM Microelectronics Division, 69B Center Road
Essex Junction VT 05452, USA, bdenton@us.ibm.com
OR methods are used extensively
for decision support of production planning in IBM's Microelectronics Division.
The division consists of several plants including wafer fabrications plants,
test plants, and ceramic substrate plants, geographically dispersed around
the world. Division central planning systems for coordinating the production
at each of the plants, and interplant logistics throughout the division,
require the solution of very large-scale supply-chain planning problems.
We describe the practical application of linear programming and heuristics
to optimization of these extended enterprise supply-chains in semiconductor
manufacturing. We give an overview of the manufacturing process, discuss
some of the model requirements specific to semiconductor manufacturing,
and describe important issues in the practical implementation of these
models in a production environment. Some structural properties of the model
are explained and decomposition methods for handling the resulting large-scale
linear programming problems are described. Computational experience and
future directions will also be discussed.