AI optimisation boost for wafer fab sustainability
UK startup Flexciton is using machine learning to optimise semiconductor wafer fab operations to reduce the use of chemicals and energy, making them more sustainable without reducing the yield.
This comes as Samsung certifies its fab processes for sustainability and industry organisation Semi sets up a sustainability consortium of fab operators. The company raised $15m in October 2021 for its real time optimisation software semiconductor production and has worked with global chip companies including Renesas Electronics.
“As energy reduction becomes one of a fab’s key priorities, by working smarter and re-evaluating their production processes, it’s absolutely possible for companies to improve throughput and yield while at the same time being more energy efficient,” Jamie Potter, CEO of Flexciton told eeNews Europe.
“Fabs commonly use heuristic software where rules have been written by engineers to run the production and are based on historical data but there is no intelligence built into the software,” he said. “The intelligence comes from the skilled industrial engineers who use their years of experience to adjust and tweak the scheduling rules to allow for changing conditions and production requirements,” he said.
These algorithms are based around the Key Performance Indicators (KPIs) on the Work in Progress (WIP).
“The KPIs that fabs work to, which are primarily based on cycle time, throughput and yield, means that energy consumption has historically been very much a secondary consideration. However, as energy costs have risen and the need to reduce energy use because of its impact on climate change, this is now becoming an important KPI especially as energy consumption can account for up to 30% of a fab’s operating costs,” said Potter.
- PragmatIC: Making semiconductor fabs more sustainable
- Samsung certifies carbon footprint of its semiconductors
“To properly optimize the way in which the fab works, we have to first understand exactly what the state of the entire WIP is in real time. By mapping the current state of the fab’s operations, it’s possible to identify where bottlenecks are occurring due to sub-optimal scheduling,” he said.
The Flexciton software already knows how to operate a fab, so it does not need the time-consuming, rule-writing phase for a new fab installation or the replacement of existing scheduling software. It has the intelligence and pre-programmed knowledge to look at the data from all the tools and work out how to run them effectively and efficiently, all in real-time.
- R&D programme for sustainable semiconductor manufacturing
- imec tackles data gap in sustainable semiconductor production
“In our experience of working with different fabs, the tools where queues usually occur are involved in the most energy-intensive stages of the production process – for example, photolithography, diffusion furnace and clean room.”
AI optimisation of fab data
“Using AI-based optimization software to reduce bottlenecks by improving how wafers move through energy-intensive tools, the fab’s primary KPIs can be met and energy consumption at these tools can be reduced. For example, doing more moves with fewer tools at the photo stage means that it’s possible for some tools to be left idle. Or doing the same moves but with fewer batches at the furnace stage means fewer energy-intensive furnace runs.”
Smart optimization technology can also be used to directly control the energy consumption of less busy tools as well. As long as those areas that are prone to bottlenecks are running efficiently, and all primary KPIs are being met, tools in other areas can be optimized specifically for energy conservation – for instance, powered down because the scheduling technology has identified that they aren’t required or don’t have to be operated at their maximum rate.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
