On-line Detection Of Fuel Particle Size Using Machine Vision

Circulating fluidized bed(CFB)boiler has strict control for the fuel entering the furnace.Especially,the particle size of the fuel is directly related to the economic and safe operation of the boiler.The traditional fuel particle size detection adopts the screening method,but the process is cumbersome and difficult to guide the industrial production.Image processing-based machine vision has the advantages of fast detection,low cost and real-time online,and it is the best solution for on-line detection of particle size distribution.However,commercial products mainly rely on imports from the United States,Finland,etc.,and are expensive,and the maintenance of the products is not timely.It is difficult to be promoted in the power plants in China.To this end,supported by the National Natural Science Foundation of China and the Shanxi Province Science and Technology Major Projects,the research is based on the development of low cost,independent intellectual property,and national production of machine vision-based fuel particle on-line detection technology and systems to better serve China's power industry.This paper focuses on the significant demand for fuel particle size control in CFB boilers.And the research results and technological breakthroughs have been made in the key technology of on-line fuel particle detection in the coal blending process:1.An image particle segmentation method based on convex hull analysis is proposed.To solve the problem that there are some of under-segmented particles after traditional watershed segmentation,a new image segmentation algorithm based on convex hull analysis is developed,which identifies the adhesion particles existing in the image by comparing the particle convexity with the threshold value,and further separate the undivided particle areas effectively.Compared with the traditional method,using the segmentation method proposed in this paper,the average of the measurement error and the uncertainty of particle size distribution detection are reduced from 3.78% and 2.27% to 2.23% and 1.91%.2.The self-adjusting calibration technology is developed.The calibration coefficient of the pixel is dynamically adjusted by measuring distance of the ultrasonic.The particle size measurement data is corrected according to the fuel thickness on the belt in real time.And regular cleaning of the module is developed,which regularly sprays water and erases dust on the window in order to solve the problem of stable operation of equipment in industrial harsh environment.3.Machine vision-based on-line fuel particle detection system is integrated.The on-line detection system of fuel particle includes the online detection equipment and the data analysis software.And the system realized industrial applications.The system takes 1 minute to complete detection of one image.The real-time and historical data such as the fuel particle size distribution,average particle size,and the mass ratio of each level are output by the analysis software.