Research On Cement Particle Image Detection Based On MES

Cement required by important buildings has high quality requirements and complex parameters,and generally has strict requirements on its particle size range.At the same time,the roundness of cement particles,length-width ratio and other geometric characteristics also affect the strength of concrete in the later stage.Cement grinding and powder selection is an important link in cement production,and the results of powder selection directly affect the quality of cement.At present,most cement factories still use the production mode of manual sampling detection and manual adjustment of production parameters in the production process,resulting in low production efficiency and large fluctuation of cement quality.Therefore,there is an urgent need for a quality management system that can realize on-line detection and feedback adjustment of production parameters.At present,the detection methods of cement particles,such as laser diffraction and settlement analysis,have problems of low detection efficiency and single detection parameters.In order to meet the detection technology requirements of online powder selection and solve the real-time and accuracy problems of cement particle geometric parameters detection,this paper adopts image method to detect cement raw material particles.The research of cement particle image detection based on MES(Manufacturing Execution System)is carried out.The specific work content of this paper is as follows:(1)In view of the complex operation process of quality management in cement production,high error rate,low efficiency and delay in information input,an on-line powder separation process system based on image method is built.The on-line particle real-time detection and powder selection control in the cement production process are realized.(2)In terms of particle detection technology,an improved U-Net++ network model based on deep learning is built to perform pixel-level segmentation of cement particles,and then geometric parameters such as equivalent particle size,rounded degree and aspect ratio are extracted through relevant algorithms,providing data support for MES to control the cement production process and quality informatization.(3)Based on the above research content,a MES-based quality management prototype system for cement production is designed and developed.In order to meet the requirements of intelligence and informationization of cement production quality management,the cement raw meal particle detection module and equipment parameter adjustment module are put forward.Finally,the prototype system of cement production quality management system is designed and implemented.The experimental results show that the cement particles detected by this method have better segmentation effect than the original segmentation algorithm.The accuracy of image segmentation is improved from 95% to 98%,and the accuracy is improved from 84% to 94%,which solves the problem of missegmentation of adhesive particles and missing segmentation of fine particles.After comparative analysis,the errors of the geometric parameters measured in this paper are less than 8% compared with those of manual testing,which can meet the requirements of cement production process control and quality informatization of MES.