Global virtual factory simulation for energy efficiency

Abstract

Over the past decades industries have had to tackle the issue of sustainability as a matter of increasing urgency to mitigate greenhouse gas emissions and abide by government regulation policies. The semiconductor industry has come under scrutiny for its significant carbon and water footprints, largely attributed to substantial energy consumption. Notably, machine utilisation accounts for a major portion of this consumption, while the rest is shared by factory facilities. Thus, it is critical to optimise machine utilisation to attain desired throughput while minimising energy consumption. A flexible job shop model is developed for the six step Intel Mini fab using a discretised time approach to find an optimal schedule. A multi objective formulation is solved using the epsilon constraint method to minimise energy consumption and make-span for a constant throughput. This approach affords stakeholders the flexibility in choosing machine combinations based on energy implications, a crucial consideration, especially in situations of machine unavailability. Modifying this formulation, the throughput is varied, and an additional objective is incorporated to maximise it. The energy-throughput trade-off is critically analysed, and the effect of dynamically switching off machines is well demonstrated. Energy consumed per unit wafer is observed to decrease non-linearly with make-span. The effect of optimal scheduling on energy consumption is highlighted using the FIFO heuristic strategy. When employing optimal scheduling, there was a marked reduction of 26.72 percent in energy consumption compared to the maximum energy consumed in the FIFO approach.