Corning CouldnBy Kim S. Nash | Posted 2002-10-10 Email Print
Corning's small specialty materials group tries to lead the money-losing glassmaker into the future of manufacturing.t Predict Trends">
Corning Couldn't Predict Trends
On the financial side, software redundancies and gaps caused huge inefficiencies. Five separate customer order and pricing applications, seven different item master files and four invoicing systems meant a lot of manual rekeying of data. Also, Corning couldn't predict trends because of what was missing. There were no tools to analyze orders, no master resource and capacity-planning software. No shared production-reporting system. No sales forecasting applications. The proprietary manufacturing systems were basically only repositories of test results from the factory floor. They couldn't study trends, such as cycle times, or handle ad hoc queries.
In broad strokes, the specialty materials unit wants to make all its factories act like one. (Two of the nine are due to be closed down next year.) That is, products could be started at one plant, then transferred to others as workloads shifted and capacity opened up. An integrated computer system with a view of the action at all facilities would coordinate routing.
It's called progressive manufacturing, and like most good information-technology concepts, the idea is simple but making it happen isn't, especially because turning calcium fluoride powder into saleable crystals is so complicated and unpredictable. Unlike stamping out compact disks or making cereal, growing crystals is different every time. For 30 to 45 days, the powder cooks in an industrial furnace, becoming an ingot of crystals that weighs 100 to 200 pounds. Until it comes out of the furnace and is measured, Corning doesn't know what quality it is, or whether it's even usable. Hundreds of megabytes of data must be collected and analyzed with every product run. For customers and for Corning, the history and properties of a crystal are as important as the product itself because those statistics will determine how customers can use the material.
Devices attached to the furnaces must monitor and record changes in pressure and temperature. The ingots must be measured in up to 80 ways, including how much light they absorb, how much pressure they can withstand and how uniform they are. Typically just 10% of an ingotoften lessis good enough to be used in semiconductor machines. If a crystal falls short of lithography grade, for example, rerouting it for use in an industrial microscope, which has less-demanding specifications, could still salvage value.
To do that, the right data must be readily available at any point in the manufacturing process. Real-time analysis should flag problems, then help plant managers decide whether it makes financial sense to scrap an ingot or reroute it for a customer who would accept lesser grade material. And when Corning makes a superior crystal, it wants to be able to comb the data for the next time, to try to recreate the exact conditions in which it was produced.
The existing factory systems just couldn't do all of that. To do progressive manufacturing, both the financial and manufacturing systems had to be the same in all factories. Specialty Materials chose several of PeopleSoft's financial and factory applications. PeopleSoft is installed elsewhere in Corning already; the two companies have done business for eight years.
For manufacturing execution systems, Specialty Materials bought the Virtual Factory suite from Camstar Systems in Charlotte, N.C. While PeopleSoft provides an enterprise view of such areas as order management, inventory costing and manufacturing scheduling, Camstar's software tracks and controls what happens on individual machines at the factories.