Particle Size Detection Of Green Ball Based On Image Processing And Pelletizing Water Addition Control

Pellet ore is one of the main raw materials for blast furnace ironmaking and its quality directly affects the production efficiency of blast furnace.The pelletizing process of most of the disc pelletizing machines in China is dominated by manual operation.This control method has low production efficiency and unstable pellet quality.The particle size distribution of green ball is one of the key indexes affecting the quality of pellets.In this thesis,the image processing technology is used to detect the particle size of green ball,and a fuzzy controller is established based on the particle size information to realize the automatic control of pelleting production.The main research contents of this thesis are as follows:(1)By analyzing the motion characteristics of the green ball,the image of the green ball was collected in the static layer,the hardware system was designed,and the required instruments were selected.(2)The green ball image is segmented by traditional image segmentation algorithm.It is found that most of the green ball objects in the image after adaptive threshold segmentation are not accurately segmented and the overall segmentation effect is poor.The improved watershed algorithm has the disadvantages of incomplete segmentation and excessive segmentation.Therefore,the above traditional image segmentation algorithm has a poor effect on the segmentation of green ball image,and the accurate information of green ball size cannot be obtained.(3)Deep learning network is used for semantic segmentation of green ball image.First of all,data sets are made and green ball image segmentation is completed in Pytorch deep learning framework based on U-Net network and U-Net++ network.Experimental results show that the segmentation effect of U-Net++ is better.Then,the final binary image of the green ball was obtained from the trained U-Net++ model,and the geometric parameters such as the contour of the target area were obtained through the method of connected domain marking to calculate the particle size percentage.Relevant data show that the granularity information obtained by this method is more accurate.Finally,Based on Py Qt framework,a user interface for online detection of green ball particle size was constructed,and relevant algorithms were called to detect green ball particle size.Select the corresponding function reading module through the visual interface,detect the particle size of green ball and display it in real time on the window interface,realize the real-time detection of the particle size of green ball.(4)Using the granularity information obtained,such as the percentage of green balls<8mm and the percentage of green balls >16mm,fuzzy rules are designed to realize automatic control of water addition.The results show that the designed fuzzy control rules can realize the automatic control of fog water quantity.In this thesis,the image-based online particle size detection method can replace the manual screening method,which can not only detect the information of green ball size in time,but also realize the automatic control of water addition of pelletization machine by using the real-time feedback of green ball size distribution information,and improve the automation degree of pelletization production.