If I arrive on-time or early to the airport but there is a problem with the airplane and delay in the scheduled takeoff time, does it get me to my destination on time? Clearly the answer is No! Is it better for me to stay in the comfort of my home and go when I know the plane is ready to take off? Or even better, find another means of getting to my destination so that I get there on time. By going to the airport at the “wrong time” and waiting I am only increasing “WIP” or waiting time and I am also contributing to airport congestion which adds to the traffic and boarding of the flights which happen to be on-time. This problem can be avoided by an intelligent planner or, release strategy, that can figure out exactly what the right time is for me to leave home given the traffic situation, the speed of cars, the parking time and time it takes to go through security. This kind of Predictive planning is ideal for releasing lots in a semiconductor manufacturing line where depending on the mix of products, availability of resources and masks as well as WIP, it can decide which lots should be released and which ones should be held back so that we meet the following three objectives optimally:
- Cycle time
- Equipment utilization
- Delivery performance.
The following diagram shows the relationship between these three parameters and how they change with WIP increase. In a high mix environment, increase in WIP does not necessarily imply additional wait times or delay in delivery because of multiple routes and balance of allocation of jobs by the system. The grey shaded area represents optimal region of operation where the desired objectives can be achieved.
In environments where there is a high mix of products, such as foundries, we can increase the number of lots released without increasing their waiting time by ensuring that they are balanced across different bottleneck equipment such as lithography equipment. Given the complexity of such environments where each process has 400-600 steps using hundreds of equipment requiring anything from 10 minutes to 10 hours with highly sensitive set up times (implanters) or batching requirements (ovens), one has to intelligently look ahead and look behind to ensure proper balance of lots re-entering the process and or entering the process with different priorities.
Unfortunately, sequencing engines have been given too much attention in order to solve such a complex problem. We feel differently and through years of R&D have concluded that unless a proper release strategy is deployed, sequencing would not be of much value. But more importantly, in the presence of an adequate sequencing strategy, sequencing can be a liability in the sense that it would try to resolve issues locally not being aware of the potential issues it might be causing 50 steps later! Can you imagine being at the gate, and the airline personnel try to sequence your entry into the plane when the plane is not even at the gate or is being fixed!
As in our opening example, a good release strategy is aware of the right mix of products in the fab as well as the work load of each equipment, now and future, is constantly trying to balance what needs to go next such that the bottlenecks, as they are changing, will be fully utilized and at the same time keeping in mind which lots need to be ready and when for delivery. In fact, in our published papers, we have shown that in the presence of good release strategy, fairly simple sequencing algorithms can be deployed and most of the time FIFO is the best sequence for the resources. In the context of our airport example, if you left your home at the right time, as you approach your gate, without much waiting, you will show your boarding pass and get into your seat for takeoff. No need to be sequenced! In general, sequencing strategies are reactive whereas release algorithms are predictive. A combination of both is needed. But the more predictive you are the less reactive you need to be. A release algorithm provides the right mix, a sequencing algorithm executes the mix as they progress. Also, it takes care of unexpected issues such as equipment breakdowns, quality issues or priority changes. At Adexa, our patented algorithm provides both release and strategy and sequencing in order to manage an optimal operation taking into account utilization, cycle times and delivery performance.