Data-Centric Yield Engineering Frameworks for Next-Generation Semiconductor Production Systems

Abstract

Next-generation semiconductor production systems require faster and more reliable yield engineering because large volumes of process, inspection, test, equipment, and yield data are produced during every manufacturing stage. This article presents a data-centric yield engineering framework that connects these data sources into one structured decision-support model. The framework integrates wafer fabrication records, wafer inspection outputs, electrical test results, equipment operation logs, manufacturing execution data, and final yield summaries. Raw records are cleaned, converted into yield-risk indicators, and used for prediction, explainable analysis, and optimization feedback. The results show that the data-centric framework gives higher yield prediction accuracy than conventional and statistical monitoring approaches. It also reduces yield loss more effectively by linking early risk signals with corrective engineering actions. The proposed framework supports early warning, root-cause review, process-window tuning, tool-maintenance guidance, and improved production control in next-generation semiconductor manufacturing systems. It improves confidence in practical, data-driven yield decisions across fabs.