The concept of mechanization in industry has evolved over time, gaining new capabilities with the advancement of modern technologies. Maintenance activities associated with mechanization have also become a frequently studied topic, driven by these technological developments. To align with contemporary requirements and industrial technologies, equipment maintenance practices are being modernized, moving away from traditional methods. Artificial intelligence technologies are replacing conventional maintenance approaches, enabling more efficient, reliable, predictive, and cost-effective solutions. Among these advancements, predictive maintenance activities, which are rapidly growing both in theory and practice, stand out as a key area of focus. This study was conducted in an automotive factory housing numerous machines and equipment. Specifically, the slat conveyors in the assembly unit of the factory were examined. The aim was to detect and predict failures in the slat conveyors using historical current data collected through IoT sensors. Within this scope, the EWMA (Exponentially Weighted Moving Average) method, a statistical control technique, was employed to identify data points with potential failure risks. Subsequently, the Isolation Forest method was used to assign anomaly scores to these data points. Due to the absence of pre-labeled failure data, potential failure-indicating data points were labeled based on the analysis. By evaluating the scores of the labeled data, high-risk failure points and their estimated occurrence times were identified. As a result of this study, improvements were made to the existing maintenance plans, enabling the creation of more accurate maintenance schedules. Additionally, a reduction in equipment maintenance activities and enhanced predictability of potential failures were achieved.

Keywords: Maintenance Planning, Fault Diagnosis, Exponentially Weighted Moving Average (EWMA), Isolation Forest