Question

Consider a semiconductor wafer fabrication facility. The input into the facility is wafers that are released into the facility in lots (or batches) of 25 wafers. Each wafer can produce a maximum of 50 die or chips. The output from the fabrication facility goes to a testing operation that determines which chips (or die) on the wafer are good and which are bad. The good chips go into an inventory to serve customer demand, and the bad chips are discarded. The production lead-time through fabrication and test is 50 days on average with a standard deviation of 3. The daily demand for chips is 300 per day with a standard deviation of 30. The yield of good chips from each wafer is 20 good chips on average with a standard deviation of 8. Suppose we want to have a 95% Type I service from the chip inventory. Describe an inventory policy to achieve this (You can assume all uncertainty follows a normal distribution and you probably want to make some assumptions of independence, as well as some simplifications and approximations). In particular, you should decide how to determine when to release a lot of wafers into the wafer fabrication facility. If it is helpful, you might consider a particular instance: suppose the number of batches in progress is 30 and the current position of the finished goods inventory is 1500 chips. Then the question is whether or not we should launch any new batches into production, and if so how many?

Answer 1

complexity of the related supply chains suggest that simple, intuitive, manual techniques are unlikely to perform well.

Simulation models for semiconductor wafer fabrication are considered important tools for supporting the decision-making processes in manufacturing operations.

For the selection of a lot to be released, two groups of dispatching rules are used in this paper. One is a set of existing rules that are reported to give good performance for lot release control and the other is a set of rules that are newly developed in this paper. Priorities of waiting lots are determined using these dispatching rules and a lot with the highest priority is selected.  

The jobs have to be scheduled on the various types of tool groups in the wafer fab and processed together. This class of integrated job formation and scheduling problems are called “multiple orders per job” scheduling problems.

when to release a lot of wafers into the wafer fabrication facility, In numerous wafer fabrication facilities (wafer fabs) the load changes dynamically. Current load release control schemes are usually developed and tested in a constant load environment. In this paper, we discuss four CONWIP-like lot release policies and assess their ability to buffer variability generated by the load changes and to keep the lot release variability as low as possible. All control approaches reduce the WIP and cycle time variability inside the fab but at the cost of increasing the total WIP and cycle times of the lots compared to an uncontrolled fab.

suppose the number of batches in progress is 30 and the current position of the finished goods inventory is 1500 chips. we should launch new batches into production, i.e (1500*30)/25=1800 batches.

we should launch 1800 new batches into production.