Back-end Factory Scheduler
Bring Powerful WIP Scheduling to Assembly, Package, and Test
SMART back-end Scheduling
WIP management and scheduling for semiconductor Back-End operations is becoming increasingly important. Line linearity, increased throughput and utilization, and accurately meeting ship targets is as important in the Back-End as it is in the Front-End.
Smart scheduling techniques developed for complex Front-End environments can be applied to the more linear operations in Assembly, Package, and Test facilities and achieve the same results.
The FPS Back-End Factory Scheduler uses advanced algorithms leveraging practical constraints and real time information to make a schedule based on current conditions and priorities. The schedule updates every few minutes and generates the full end-to-end schedule for the entire facility. This guarantees continuously relevant results amidst the dynamic conditions of the back-end facility.
Substrates entering a back-end factory will complete all process steps in typically 7 – 10 days. This allowed INFICON to modify the front-end scheduler to account for the entire back-end process flow.
The end result is that you are able to see where any lot will process, based on real time tool status, recipe qualifications, and lot status, through the entire back-end flow.
Scheduling is not Dispatch
Dispatch rules have worked well. When maintained by skilled programmers and operators, they provide impressive efficiency. They are generally fast to execute, comprehend local priorities, and provide more consistent behavior than purely manual decisions. However, their drawbacks are considerable:
Dispatch rules do not forecast results. If I run this lot now, will it be able to finish its queue timer 6 steps from now?
Each local step must have its own logic, which can be out of sync or in conflict with upstream or downstream rules.
Rules are difficult to quickly adjust to changing needs
Global business goals are difficult to formulate in the language of local dispatch rules
Rules typically only consider lots at station, and not those arriving soon or downstream.
The results of changing dispatch rules are very hard or impossible to test in advance
Look ahead. A scheduler will stop a lot if a queue timer will expire.
See the whole line. Factors like line balance, down toolsets, WIP holding are comprehended.
Incorporate setup changes, reticle switches, movement penalties, etc.
Comprehend upcoming maintenance events
Provide true balancing of local and global priorities
Facilitate centralized business logic, and give Fab managers the controls
Allow changes to be tested hypothetically, without affecting production
Open up smart manufacturing integration opportunities - A system that "knows" where WIP will be creates opportunities for managing pumps, abatement units, etc.
The Scheduler Gantt Chart shows when and where lots will run, but equally importantly, it shows how the decision was made.
The UI will display the reasons that led to the placement of the lot on the schedule, broken down by the importance of each priority. If you disagree with the Scheduler's decision, you can see how to adjust priorities in order to change it.
Intuitive Controls, Easy Testing
Scheduling priorities are set with a simple UI with clear explanations, The UI includes a full sandbox to test the effects of changing priorities. Compare this to the problem of quantifying the effects of changing dispatch rules. It has truly never been this easy.
When was the last time you had a strong lever for line balance? Or a quick way to stop a WIP bubble from forming due to an unexpected bottleneck tool downtime? Or an easy way to run your tool that requires a particular pattern of lots?
We have seen impressive achievements with variations on dispatch logic, and there is no question that full factory scheduling is a difficult problem requiring a lot of hard work. However, this is all the more reason to choose the right tool to invest the time in. Installing INFICON schedulers in a large fab can take up to a year. But that year of time pales in comparison to the equivalent years spent updating and maintaining less powerful solutions only to deliver less optimal results.