DIAGNOSTICS OF CAROUSEL-TYPE AUTOMATED PICKING SYSTEMS

Carousel-Type Automated Picking Systems (CTAPS) are critical for high-speed multistation assembly, yet they remain susceptible to performance degradation caused by mechanical wear, sensor instability, pneumatic delays, and servo drift. While classical diagnostic methods rely on static thresholding of physical signals, modern manufacturing demands predictive capabilities that can account for complex, nonlinear fault patterns. This study proposes a novel 5-layer integrated diagnostic framework that combines classical signal processing with hybrid Artificial Intelligence (AI) architecture. The proposed methodology integrates Autoencoders for anomaly detection, CNN-LSTM networks for Remaining Useful Life (RUL) prediction, and XGBoost for precise fault classification. By synthesizing these outputs into a unified Health Index (HI), the system enables real-time condition monitoring and root-cause analysis. Experimental results demonstrate that this AI-driven approach increases system reliability by 27–35% and enables the early detection of incipient faults 3–5 cycles earlier than traditional methods.