The Parameter Analysis And State Prediction Of KR Desulfurization Based On Image Processing

With the increasing demands of industrial production,molten iron desulfurization has become a key step in steel production,and KR desulfurization method is widely used in steel enterprises due to its good kinetic conditions and low cost.Currently,the desulfurization status,or desulfurization rate,of the KR iron desulfurization process with different stirring parameters is mainly obtained through simulated hydrodynamic experiments.However,the commonly used method has a large error in the analysis of the experiment,and the experiment also has the disadvantages of long cycle,high cost and strong dependence on personnel.In this paper,the influence of stirring parameters on the desulfurization rate of KR during desulfurization is analyzed by image processing method,and the prediction of KR desulfurization status is carried out by using shallow convolutional neural network,which has certain theoretical significance and practical value.Firstly,in view of the commonly used KR desulfurization stirring process parameter analysis method based on the desulfurizer entrainment depth,only the defects of the longitudinal diffusion range of the desulfurizer are considered in the analysis process.The parameter analysis method of KR desulfurization stirring process based on image segmentation is proposed.K-means clustering segmentation algorithm is used to segment the water model experimental image under each stirring process parameter to obtain the target region containing the desulfurizing agent;determine the desulfurization rate corresponding to each image target region by the defined desulfurization rate judgment criterion;and then according to each image target region corresponding desulfurization rate is used to analyze the influence of stirring process parameters on desulfurization rate.Then,according to the AlexNet network structure,a KR desulfurization state prediction based on the improved AlexNet network is proposed.This method introduces the segmentation and classification method in image processing,taking the water model test pictures under different stirring process parameters as input and the desulfurization state defined according to the desulfurization rate criterion as the output,a prediction model of KR desulfurization state based on the shallow convolutional neural network was constructed.The results show that the KR technology parameter analysis method based on image segmentation can reflect the relationship between the technology parameter and the desulfurization rate more accurately.The prediction accuracy of KR desulfurization status prediction model based on improved AlexNet network can reach 96.58%.