Anomaly detection in time series data plays a crucial role in industrial manufacturing by identifying potential defects, inefficiencies, and equipment failures before they escalate into costly disruptions. With the increasing adoption of the Internet of Things (IoT), modern industrial systems generate vast amounts of sensor data, necessitating advanced detection methods. Deep learning (DL)-based anomaly detection has emerged as a state-of-the-art solution, but its practical deployment remains challenging. This study explores the application of DL for real-time anomaly detection in a real-world manufacturing system. The proposed approach demonstrates high detection accuracy, with AUC-PR scores of 1.0 and 0.9958 for two different plant sections, showcasing its feasibility in programmable logic controller (PLC)-based environments. Additionally, the research introduces a method to enhance system usability through improved visualization and anomaly clustering, making insights more actionable for practitioners. While results indicate strong anomaly detection capabilities, challenges remain in determining appropriate detection thresholds, particularly for novel or custom-built systems. The study highlights the trade-offs between detection reliability and real-world variability, emphasizing the importance of domain expertise for effective deployment. Nevertheless, for standardized industrial components, the approach presents significant benefits, enabling proactive maintenance, reducing downtime, and enhancing operational efficiency.