Site Search categories View by Month ranking Recent Entries
Syndicate this site (XML)

Takahashi's Scheduler Case Book: Chapter 3

The Case of the Semiconductor Plant

Production scheduling at a semiconductor plant...Speed essential for a production scheduler.

I visited a semiconductor fabrication plant. This company has put a lot of money and more than six months of time in developing their production scheduler and they have just started to put it into operation. Be that as it may, they would like to have a demonstration of our production schedulers. When I asked them why, I was told that whenever they run real data on their just-developed scheduler, schedule processing drops down to more than five hours. The semiconductor industry has to strictly observe promised delivery time even when changes are made in orders. If a change is made in the delivery time for the order, corrections have to be made in the production scheduler. Then what happens when processing takes five or more hours? That would lead to a delay in measures for countering delays in delivery time. If that happens, a human production scheduler would be faster, making determinations promptly and giving instructions to onside workers to ensure that delivery is on time. A slow production scheduler is fatal. We got the data from this plant's production scheduler and held a demonstration several days later. The semiconductor fabrication process features an exceedingly long 150 steps. Not only that it uses an enormous amount of data. We ran the production scheduling on real data with the customer in attendance. Then, what happened? We finished the entire process in five minutes! We were even questioned with "Did that really get all the calculations done just now?" Of course it's doing all the calculations. However, five hours had been turned into five minutes. In a situation like that any kind of change in order can occur, no matter whether that would be a change in delivery time or a change to an express order, and you can make changes in instructions with this production scheduler.

In the example that this company gives us, the scheduler they had developed was able to complete processing a schedule in a few minutes when data load was low, (all quite natural) but when real data was entered in large volumes there was a rapid drop off in production scheduling speed. If you're talking about the person who developed the program, they'll have an understanding of what the reason is. There are many difficulties in designing a production scheduler program that won't change speed no matter what the volume of data it handles. One thing that makes us proud about our Asprova production scheduler is that no matter how large the volume of data, its production scheduling speed remains the same. The production scheduling time for 10,000 jobs is ten times the production scheduling time for 1,000 jobs and one-tenth the production scheduling time for 100,000 jobs. This is the production scheduling logic that I came up with during a one-week sojourn in Hokkaido, traveling with the family in the summer of 2001. During that week I let the children play and I designed Asprova APS. What a wonderful memory.

After we got the system operating, I went back to check on it, and the manager told me, "We wouldn't be able to get our work done if we didn't have your production scheduler. From the schedule results we found out that leaving one semiconductor fabrication machine vacant causes a loss of 100 million yen. We would probably have bought this system even if it had cost 50 million yen. "

At any rate, the speed of a production scheduler is vital.

Case Closed. . .

Learn more about Production Scheduling at Production Scheduling System Asprova

Written by Kuniyoshi Takahashi: Production Scheduling Case Book

Leave a comment

<< Previous Entry | Next Entry >>