Estimation of combustion efficiency in coal fired boilers with computer vision techniques

Coal fired-boilers are widely used in the heating of personal houses, buildings or workplaces. The efficient burning of the boiler provides an economic added value, while reducing the harmful gases released into the atmosphere and contributing positively to the environment. In this thesis, a new system is proposed that can automatically calculate the combustion efficiency of household coal boilers using computer vision techniques. This system acquires the image of the flame form inside the boiler during the combustion process and maps it to the efficiency values measured with professional flue gas analyzer devices. For this, the high-dimensional flame images obtained are reduced to low-dimensional feature vectors and the boiler efficiency is estimated by artificial learning techniques. Within the scope of the thesis, the effect of many different feature extraction and modeling approaches on match accuracy was analyzed. The main scientific contribution provided with this thesis is the development of mathematical models that provide the highest matching accuracy between flame image and efficiency measurement compared to existing methods. In addition, a mathematical model is proposed that can predict the flue gas temperature from the flame image. Real-time applications of developed prediction models on a prototype coal fired-boiler have been made.