Prediction of NOx emissions from a biomass fired combustion process through digital imaging, non-negative matrix factorization and fast sparse regression

This paper presents the development and evaluation of an algorithm for the prediction of NOx emissions from a biomass fired combustion process based on flame radical imaging, image processing and soft computing techniques. The investigation was performed on a biomass-gas fired test rig. An algorithm which combines texture analysis and non-negative matrix factorization (NMF) is studied for the image feature extraction. Fast sparse regression with convex penalties is then employed to establish the relationship between the image features and NOx emissions. The predicted NOx emissions from a fitted model are in good agreement with the measurement results. The results demonstrate that the proposed technical approach to the prediction of NOx emissions is effective.