Intelligent Particle Size Inspection System For Iron Ore Green Pellets

Iron ore pellets are important raw material for modern iron-making industry using blast furnaces.Disc pelletizing is an important process in iron ore pellet production,in which iron ore concentrate is agglomerated in disc pelletizers into green pellets.Green pellets are then processed by sieving,drying,roasting etc.to become iron ore pellets(final product).Therefore,the quality of green pellets greatly influence the energy consumption and the quality of final products.Particle size distribution(PSD)is one of the important quality indices of green pellets.According to current standard,80% of the pellets should be in the size range of 10mm-16 mm.At present,PSD is mainly measured by manual sampling and sieving,which is time-consuming and inefficient.Moreover,the measuring result lags behind,which may result in improper control of the pelletizing process and fluctuations of product quality.In the framework of National Natural Science Foundation of China(Project No.61973108),an intelligent system for pellet size inspection was designed and implemented,which consists of a manipulator for grasping the pellets and a machine vision system for pellet size inspection.With it,the pellet size inspection process can be fully automated with high accuracy and efficiency,which has promising application in industry.The main contributions of the present work are as follows:(1)A hardware system for the intelligent particle size inspection system was designed and implemented on a laboratory disc pelletizer.The system mainly includes a manipulator module for pellet sampling and a machine vision inspection module for pellet size.The novelty is the design of a magnetic end-effector,which can avoid damaging of pellets during sampling process.(2)Two algorithms for particle size inspection are studied based on previous work of our research group,including a new regionally constrained watershed algorithm and an improved LU-Net based algorithm.The main difference between them is the image segmentation step: the former adopts the traditional image segmentation method with better computing efficiency,while the latter adopts the deep learning method with higher segmentation accuracy.(3)A software system for intelligent particle size inspection was designed implemented using CSharp and C++ programming language.It has the functions of camera control,control of the manipulator for pellet sampling,machine vision inspection of particle size,data storage etc.(4)Comprehensive tests were performed to evaluate the software and hardware system,including module testing,function testing and system performance testing.Results show that each module of the system works well and the system runs stably.The pellet sizes measured by the automatic system was compared with those by manual sieving.It is found that the average error of the system is 10.0% when the regionally constrained watershed algorithm is selected,the error can be reduced to4.4% if the LU-Net based algorithm is selected.The average time for processing a pellet image on an ordinary computer configuration is 0.72 s and 13.63 s,respectively.Both of them are suitable for automatic size inspection of green pellets.The users can select one of them according to their specific application requirement.