Anomaly detection in uranium conversion facilities: A method integrating C-DUCG and HDBSCAN

The vertical fluorination furnace serves as a key device for the preparation of uranium hexafluoride (UF 6 ). Its safe and stable operation is crucial for ensuring the global supply of nuclear reactor fuel. An anomaly detection algorithm for fluorination furnaces based on the integration of Cubic Dynamic Uncertain Causality Graph(C-DUCG) and Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) is proposed, which leverages the strengths of knowledge driven and data driven approaches, enabling precise diagnostics of complex fault patterns under various operating conditions. Validated with operational data from a fluorination furnace in China, the algorithm effectively diagnosed two typical faults: blockage at the UF4 feed inlet pipe and leakage in the fluorine gas inlet pipe. This confirms its efficiency and robustness in handling non-linearly distributed data and rare faults, providing theoretical guidance for the safe monitoring and preventive maintenance of nuclear industry equipment.