Interpretable Analysis of Traffic Congestion Relationships between Interface Modes in Clustered Semiconductor Fabrication

Clustered semiconductor fabrication (CFAB) has recently been introduced in South Korea, connecting multiple FABs to meet the high demands of large-scale integration chip products. To make these FABs connected effectively, CFAB utilizes various types of interface modes (IMs) such as lifters and conveyors to transfer wafers between FABs. However, due to the operational algorithm limitations of IMs and the frequent wafer transfers between FABs to meet production requirements, traffic congestion among IMs often occurs unexpectedly. These instances of congestion can cause unpredictable traffic conditions in other IMs, making it more challenging to manage traffic in a CFAB. In this paper, we propose three analysis methods based on data mining and graph network analysis approaches to explore reciprocal relationships between IMs regarding traffic congestion and describe the reasons behind traffic congestion in each of them. To demonstrate the effectiveness of the proposed method, we conduct numerical experiments using a dataset collected from the real-world CFAB in South Korea. The findings through experiment results suggest that most traffic congestion occurs in only a few specific IMs in the CFAB, indicating that managing these IMs can reduce overall traffic congestion. Moreover, we provide visual insights into managerial implications to support decision-making for enhancing IM operation performance and mitigating traffic congestion bottlenecks in complex CFAB environments.