LLM-Assisted Sensor-Log Fusion for Predictive Maintenance in Nanofabrication Equipment

Abstract

Predictive maintenance (PdM) is increasingly important for nanofabrication equipment because unexpected equipment faults may cause process interruption, wafer loss, production delay, and increased operational cost. Conventional corrective and preventive maintenance strategies are limited because they either react after failures have occurred or rely on scheduled inspections that may introduce unnecessary downtime. Recent advances in machine learning (ML) and large language models (LLMs) provide new opportunities for PdM by enabling structured sensor analysis and unstructured maintenance-log interpretation. This paper presents a review-based LLM-assisted sensor-log fusion framework for PdM in nanofabrication equipment. The proposed framework integrates sensor data, maintenance records, logs, alarms, manuals, ML-based prediction, LLM-based interpretation, and human-AI decision support into a unified workflow. The reviewed studies show that ML models are effective for anomaly detection and fault classification, while LLMs can support contextual diagnosis and maintenance recommendation. Key deployment challenges, including data scarcity, model trust, domain adaptation, privacy, computational cost, and human verification, are also discussed.