This research developed an effective and efficient anomaly detection framework that addresses key industrial challenges, including noise, lack of labelled data, lack of universally applicable anomaly detection techniques, and the dynamic nature of normal and anomalous behaviour. The framework integrates various tools in an extensible architecture for diverse manufacturing datasets in learning factories and industrial environments. Framework evaluation through ablation studies assessed the contribution of each tool to overall detection capability, considering macro-averaged F1-score and computational efficiency. This research develops educational guidelines for framework development and integration in learning factories, showing how semi-automated anomaly detection can improve operational efficiency in manufacturing. The study provides practical insights for learning environments and industrial settings by explaining how to select tools and configure frameworks for effective anomaly detection in manufacturing processes.