Real-Time Anomaly Detection in Cement Kilns: A Deep Learning Framework for Industrial Process Monitoring

Abstract

This memorandum presents a predictive intelligence system for the rotary kiln at the SCAEK cement plant. The project addresses the limits of the existing reactive Siemens WinCC RT v7.2 supervision system by introducing real-time anomaly detection based on industrial time-series data and deep learning. The proposed solution uses a GRU AutoEncoder model to detect early signs of equipment degra-dation from sensor data such as temperatures, gas flows, kiln speed, and power consumption. A real-time architecture built with Apache Kafka, InfluxDB, Grafana, and Streamlit enables continuous monitoring, alert management, and operator feedback. The study also examines secure integration strategies within Siemens PCS 7 / WinCC environ-ments, emphasizing cybersecurity and operational continuity. Overall, the system aims to reduce unplanned downtime, improve reliability, optimize maintenance resources, and support the digital transformation of industrial maintenance