Real-time active-gas imaging of small gas leaks

Abstract

Abstract. To tackle global warming, the reduction of greenhouse gas leaks is of great public interest. While state-of-the-art optical gas imaging (OGI) cameras can visualize larger gas leaks with magnitudes of liters per minute in the case of methane, a much more sensitive laser-based approach is introduced here. This is accomplished using an infrared camera in combination with an interband cascade laser (ICL) as active illumination. The laser beam diverges such that it covers roughly half of the camera's field of view. Three-image batches are recorded to perform classic direct absorption spectroscopy (DAS) at the image scale. The obtained concentration length in parts per million meter (ppm m) is validated using measurements with varying known methane concentrations, different reflective elements, and varying distances. The real-time camera was able to record and quantify a methane leak as low as 40 mL min−1. Possible incorrect information due to moving objects is taken into account using an adapted frame-difference approach.

Active gas camera mass flow quantification (qOGI): Application in a biogas plant and comparison to state-of-the-art gas cams

“…This effect is analyzed and discussed in Ref. 9. We here propose a procedure to address this by tuning the sensitivity of the active gas camera while still obtaining reasonable flowrate predictions.…”

“…However, 1 l/min can serve as an order of magnitude for a minimum detectable leakage for a pure methane leak. 7,8 This paper aims to compare the conventional gas cameras with the active gas camera approach that we have previously published 9 under real-world conditions. In doing so, we introduce a methodology to extend the applicability of our neural network 10 to much larger flow-rate scales, obviating the need for additional training data.…”

“…A detailed description has been published in real time active gas imaging of small gas leaks. 9 In many real-world applications, the flowrate of a gas leak is of larger interest than the gas concentration. We use a 3D Convolutional Neural Network (3D-CNN) for the mass-flow quantification estimation.…”

“…More details on the entire procedure can be found in the works of Strahl et al 12 and in Bergau et al 9 By selecting the wavenumber ν to be at a suitable absorption strength, the sensitivity of the camera can be tuned to fit the leak of interest. This is described in detail in Sec.…”