Research On Flameout Diagnosis And Analysis Based On Furnace Image And Parameter

The purpose of this paper is judging whether there is any flame-out signs and analysing the causes of fire out by image detection.the main methods of study are detecting features of flame image in combustion process in three cases collected under the boiler,collected the black area,volatile combustion area,volatile combustion energy difference,characteristic parameters of flame color and temperature,the preliminary data processing of combustion rate,combustion wave amplitude and frequency characteristic parameters.Then the principal component analysis and dimension reduction is used before training and classification with BP neural network and support vector machine(SVM),that analysis could help us to find the reason of the fire instability which even cause the flame out.In the combustion process,volatile combustion is during carbon in coal fire and no fire,so in different causes of flameout before the volatile combustion state can reflect the combustion state of flame instability,and the causes of instability.Through the observation and analysis of the collected data,it is found that the volatile combustion area is affected by the primary air ratio,and the shaking phenomenon in the combustion area can directly reflect the relationship between the primary air and the flame stability.This paper mainly used digital image,The flame image is divided into 4 areas :unburned zone,primary combustion zone,combustion zone and burn area.Primary zone at the beginning of the main combustion zone is volatile combustion zone,unburned zone and primary combustion zone composed of black area.Collected 4 area,the length of the black dragon,burning zone temperature in the double color method measuring the characteristic parameters.Combined with the temperature measurement of furnace parameters,correct the imaging parameters of the camera.The parameters of flame image data collected directly is complex for analysing,This paper use the method of big data processing to reduce the dimension of the data and eliminate the correlation,which make the data expression is more efficient,The application of the BP neural network and parameter optimization of support vector machine for training and classification,reduce the data redundancy after the input,improve the accuracy of the neural network,comparative test results show that BP neural network optimized by genetic algorithm make the output accuracy for the fire more accurately...