Combustion Regime Monitoring by Flame Imaging and Machine Learning
- 9 Downloads
A method for automatic determination of combustion regimes using flame images on the basis of a convolutional neural network on labeled data is under consideration. It is shown that the accuracy of regime classification reaches 98% on the flame images of a gas burner. The results of the operation of the convolutional neural network and classification using different linear models are compared.
Keywordsimage classification monitoring computer training convolutional neural network flame
Unable to display preview. Download preview PDF.
- 1.S. M. Borzov, V. I. Kozik, O. I. Potaturkin, and N. N. Shushkov, “Optimization of Combustion Processes for Gaseous Hydrocarbons Through Variation of Fuel Mixture Content,” Teplofizika Aekomekhanika 11 (2), 309–315 (2004) [Thermophysics Aeromechanics, No. 2, 299–304 (2004)].Google Scholar
- 2.V. D. Antsygin, S. M. Borzov, S. T. Vas’kov, et al., “Combustion Optimization Based on Flame Parameters,” Avtometry, No. 5, 3–11 (1999).Google Scholar
- 3.A. González-Cencerrado, B. Pe˜na, and A. Gil, “Coal Flame Characterization by Means of Digital Image Processing in a Semi-Industrial Scale PF Swirl Burner,” Appl. Energy. 94 (C), 375–384 (2012).Google Scholar
- 6.I. Witten, E. Frank, M. Hall, and C. Pal, Data Mining: Practical Machine Learning Tools and Techniques (Morgan Kaufmann, San Francisco, 2017).Google Scholar
- 7.M. T. Hagan, H. B. Demuth, M. H. Beale, and O. de Jesus, Neural Network Design (Martin Hagan, USA, 2014).Google Scholar
- 11.A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet Classification with Deep Convolutional Neural Networks,” in Proc. of the 25th Intern. Conf. on Neural Information Processing Systems, Nevada, USA, Dec. 3–6, 2012.Google Scholar