Semiconductor Manufacturing Process: A Research Paper on the Impact of Lot Priorities on Cycle Times

Introduction

Semiconductor manufacturing is a complex process that comprises hundreds of steps in converting silicon into chips for everyday technology use. It can take weeks or months to produce each batch of wafers, and factories aim to produce large quantities in the most effective and quality-driven way possible.

Smart planning enforces production efficiency. Factories use different strategies to decide how to process wafers, such as prioritizing certain batches to speed the semiconductor manufacturing process or meet specific demands. However, it is difficult to manage these priorities because they affect the processing speed of other batches.

This semiconductor research paper explores how prioritizing some wafers, said to be 'hot lots', affects the total time taken to process other lots.

Simulation Model

The researchers used AnyLogic software to mimic the semiconductor manufacturing process and provide insights that could help improve production planning. The simulation included a part of a chip factory in Crolles, France, where each machine works as a line, with batches waiting for their turn.

Using AnyLogic, the semiconductor manufacturing process was set up to closely imitate the real operations. Batches of chips were assigned priorities—either normal or high—and then assigned to machines based on past data but with some random variation.

In the simulation, high-priority batches could skip ahead, and once a batch started processing, it would continue until it finished. Different scenarios were tested by changing the percentage of high-priority batches from none to all, and each setup was repeated 100 times to get reliable results.

The main goal of the semiconductor research paper was to observe how batch waiting times varied with different combinations of priorities. It also aimed to compare these results with simpler theoretical models.

Results

The simulation has shown utilization of the machines in the factory to be almost at capacity, with one machine showing waiting times twice as long as the others. The overall waiting times from the simulation were a bit longer than those estimated by simpler methods, with a small 11% difference.

Graph of the average waiting time of lots, depending on the priority mix

Average waiting time of lots, depending on the priority mix

Graph of the speed up of lots depending on the priority mix, compared to a

Speed up of lots depending on the priority mix,
compared to a "First In, First Out" rule without priority classes

The results of the semiconductor research paper showed that when less than 27% of batches were given high priority, the impact on other batches was minimal. However, if the priority increased to 75%, the high-priority batches would be processed much faster, significantly delaying the others.

Such findings assist the managers in setting the right number of high-priority batches so that the right balance is maintained and the rest are not enormously delayed.

相关文章