Intelligent Predictive Failure Detection in Automated Guided Vehicle Operations

Automated guided vehicle (AGV) systems have become a key component in modern smart manufacturing environments due to their ability to streamline material handling and improve operational efficiency. However, unexpected failures in AGV operations can cause significant production delays and require human intervention, disrupting the continuity of the workflow. This study proposes a deep learning approach using long short-term memory neural networks to predict such failures in advance. Unlike previous prediction models based on regression, this study formulates the problem as a binary classification task to directly detect whether a failure occurs within a specified future time window. The proposed model integrates techniques such as class imbalance handling, early stopping, and binary classification activation based on sigmoid functions to improve the reliability of predictions. High-accuracy results demonstrate the feasibility of a classification-based predictive maintenance framework to minimize downtime and enable proactive decision-making in Industry 4.0 production systems.