Small Burning Hydrogen Leaks in Industrial Plants: A Comparison of UV and IR Remote Imaging Detection

Hydrogen is crucial for sustainable energy systems but poses significant safety challenges due to its high flammability, low ignition energy, and invisible flames. Detecting burning hydrogen leaks is difficult, requiring specialized sensors and optical methods, such as ultraviolet (UV) and infrared (IR) detection. UV cameras are particularly effective due to hydrogen flames emitting strong UV signals while being less affected by hot surfaces. This study investigates the detection and imaging of hydrogen flames in outdoor industrial settings using three camera systems: a standard visual camera, a UV camera, and an IR camera. The UV system, optimized for OH* chemiluminescence at 310 nm, used a high-sensitivity sCMOS sensor with narrow-band UV-filtering to suppress non-OH* signals. The correct choice of filter was verified by spectroscopic OH* chemiluminescence measurements. The commercial IR system captured mid-IR radiation (7.5–14 µm) emitted by the flame’s heat. A small hydrogen flame was generated between industrial pipes with a 6 mm Swagelok tube at different flow rates from 2 to 100 l/min, resulting in hydrogen flames between 1.7 cm and 26.5 cm in length. Measurements were conducted under realistic outdoor conditions in direct sunlight from a distance of 40 m in both mid-infrared (MIR) and UV light. The angular size of the flame was between 1/3° and 1/40°, i.e., near the resolution limit of the human eye. In all cases, the flame could be unambiguously detected in the UV and in the IR. The results demonstrate the feasibility of hydrogen flame detection in challenging environments. Depending on the ambient conditions, one or the other method may be more sensitive, since direct sunlight affects the UV measurement while heat sources and incandescent lamps hamper the MIR measurement. Considering the standard deviation between several images can further increase sensitivity. The system is lightweight and could be mounted on commercial drones for regular automated inspections. Its implementation would be a separate project that is not the subject of this paper.