Enhancing Flame Analysis in Industrial Combustion: A Comparative Evaluation of Spectral Emission Reconstruction Techniques Using Low-Cost Sensor

Flame reconstruction, an indispensable tool in combustion diagnostics, improves the understanding of the radiative characteristics of the flame and can be derived from the flame emission spectrum. This paper investigates the feasibility of spectral reconstruction from low resolution sensor data, namely from digital colour cameras capturing visible light (RGB). Two spectral reconstruction techniques are evaluated: the Maloney-Wandell method, which requires an accurate characterisation of the spectral sensitivity function (SSF), and a kernel-based regression method, which avoids the need for an explicit characterisation of the SSF. The study focuses on flames generated in industrial environments, including the combustion of hydrocarbons and copper concentrates. This research contributes to the advancement of spectral reconstruction techniques for industrial combustion monitoring and process optimisation, because although the results show that both methods effectively reconstruct spectral emissions, kernel-based regression proves to be a cost- effective and practical solution for industrial environments as it does not require characterisation of the SSF, which is a difficult task in practice.