Domain Knowledge Assisted Gas Tomography

The presented work addresses the challenging problem of Gas Tomography (GT) – a reconstruction technique for a spatial gas distribution based on measurements with an open-path sensor. Given that such a sensor only captures an integrated gas concentration along a laser trajectory, the reconstruction problem is inherently ill-posed. The method proposed in this work assists GT reconstruction using domain knowledge in the form of a gas dispersion model, formally described by the advection-diffusion Partial Differential Equation (PDE). Specifically, the model facilitates a physics-informed interpolation in regions not covered by measurements. As a result, the proposed numerical approach demonstrates a considerably improved performance in simulation studies, as compared to state-of-the-art GT algorithms.