Flame analysis using image processing techniques

International Conference on Innovative Technology, Engineering and Sciences 2018

工业技术;自然科学

Jie, Albert Chang Her^1 ; Zamli, Ahmad Faizal Ahmad^2 ; Zulkifli, Ahmad Zulazlan Shah^2 ; Yee, Joanne Lim Mun^1 ; Lim, Mooktzeng^2

School of Engineering, Monash University Malaysia, Malaysia^1

Fuels and Combustion, TNB Research Sdn. Bhd., Malaysia^2

关键词: Fuzzy inference systems;    Image processing;    methods;    Image processing technique;    Industrial burners;    Intelligent diagnostic systems;    Power spectral densities (PSD);    Spectral parameters;    Use of fuzzy logic;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/342/1/012060/pdf
DOI  :  10.1088/1757-899X/342/1/012060

来源: IOP

【 摘 要 】

This paper presents image processing techniques with the use of fuzzy logic and neural network approach to perform flame analysis. Flame diagnostic is important in the industry to extract relevant information from flame images. Experiment test is carried out in a model industrial burner with different flow rates. Flame features such as luminous and spectral parameters are extracted using image processing and Fast Fourier Transform (FFT). Flame images are acquired using FLIR infrared camera. Non-linearities such as thermal acoustic oscillations and background noise affect the stability of flame. Flame velocity is one of the important characteristics that determines stability of flame. In this paper, an image processing method is proposed to determine flame velocity. Power spectral density (PSD) graph is a good tool for vibration analysis where flame stability can be approximated. However, a more intelligent diagnostic system is needed to automatically determine flame stability. In this paper, flame features of different flow rates are compared and analyzed. The selected flame features are used as inputs to the proposed fuzzy inference system to determine flame stability. Neural network is used to test the performance of the fuzzy inference system.