General Framework for Processing Time Prediction and Machine Availability for all Fab Equipment

This paper presents a novel approach that significantly improves processing time (PT) and machine availability predictions across all tools in a GlobalFoundries semiconductor fabrication plant. An attention-based deep neural network was developed to enhance prediction accuracy. The model achieved a 50% to 80% reduction in Mean Absolute Error for PT and machine availability predictions, compared to a baseline statistical model. Deployment on three bottleneck tool families led to a 3% to 7% improvement in chamber utilization. The model is being expanded to all GlobalFoundries' facilities, with expected similar outcomes.