Modeling And Analysis On Shape And Texture Detection Of Burden Surface For A Blast Furnace

Measuring and monitoring the burden stocklevel and surface shape inside a blast furnace(BF)is a vital basis for the precisely automatic control of the iron making process.Industrial radars are advanced instruments that detect the bur-den surface stocklevel and shape in real time while less influenced by the harsh environment inside the BF.The mechanical swing radar(MSR)system is a newly developed industrial radar specially designed for the BF with a bell-less top.The improvement of accuracy and detectability of the MSR depends on the quantiza-tion of the measuring process which by now is an open problem.The research in this thesis is focused on this issue.The technical route of this research is formed starting from data acquisition,through modeling and finishing in measurment analysis.A framework of quantizing the measurment error of the process of detecting the burden surface radial profile(BSRP)by the MSR is finally built to achieve the goal.The research work and main achievement list as follows:(1)In parallel to the existing MSR BSRP measuring method,the proposed B mode method makes up for the deficiency in involving the implicit geometri-cal relations between sampling points.The goal of reconstructing the BSRP from discrete sampling points is then transformed into extracting a curve from the segmented region in the corresponding B mode composite image.On this formulation,in dealing with the deteriorating of the detection results in each non-charging period of the BF,a priori shape-based extraction algo-rithm is proposed to guarantee the stability of BSRP detction.For the first time,the rough texture of the burden suface is measured by a RGBD camera.In the processing of digital elevation model data of the burden rough surface,three algorithms are proposed.A morphological criterion-based algorithm is developed to extract the contours of the particles on the burden surface.And the partcle size distribution can be obtained.A method by which the root mean square height of the burden texture can be estimated from the size distribution of the particles on the burden surface is presented.On this basis,a search algorithm is developed for searching the separation frequency of the shape and texture of the burden surface.(2)A shape-texture two scale model for the burden surface is proposed for the first time.In modeling the burden surface shape,a concept of burden stack-ing density is proposed to reveal the relation between burden charging pa-rameters and burden shape after charging.A model of the charging and stacking process is proposed and verified using the detection data obtained by the MSR.In modeling the burden suface descent speed,a kinematic pre-diction model is proposed to predict both the burden suface shape after descent and the burden surface descent speed.A proposition is proposed and proved to guarantee that the BSRP after burden descent can always be calculated by numerical methods stably under a precondition of step length.In modeling the burden surface texture,four flat rough budern surface tiled from two kinds of the burden with two particle size each were examined.The properties of rough burden surfaces are summerized.And indexes of the burden surface roughness that can be used in generating simulated rough surfaces are counted and presented.(3)The kernel-induced sampling theory is introduced and adopted as an analy-sis framework.On the basis of the proposed two scale burden surface model,the measuring process of the BSRP by the MSR is analysed.The optimal reconstruction method of the BSRP from discrete detection points is ob-tained.And the corresponding error bound is given for the first time.The method of regularized kernel ridge regression is adopted to improve the re-construction accuracy which is affected by the additive noise induced by the burden surface texture.In the sense of Euclidean norm,a criterion for selecting optimal kernel and regularization parameter is given.