Research And Realization Of Metallographic Analysis Method For Semi-continuous Casting Al-Si Alloy Images

The components and structure can determine the properties of the metal alloy.The metallographic analysis technology evaluates the properties of the metal alloy through the observation and analysis of microstructures.The metallographic analysis technology is an essential procedure in the process of alloy material production.In this paper,an integrated approach based on the advanced theory of image processing,machine learning and deep learning is proposed.The accurate segmentation and analysis of microstructures in semi-continuous casting Al-12.7Si-0.7Mg alloy images are realized.The specific research contents are as follows:(1)The research status of metallographic analysis methods both here and abroad is reviewed.From the traditional metallographic image analysis technology to the metallographic image analysis system,and then to the metallographic analysis method based on image processing technology,the structure and causes of different microstructures in semi-continuous casting metallographic images are summarized.This paper introduced the theoretical basis of metallographic analysis for semi-continuous casting Al-Si alloy images.(2)A method for evaluating properties of microstructures in semi-continuous casting Al-Si alloy images based on V-MOB net is presented.V-MOB net combines the traditional convolution layers of VGG-16 and the depthwise separable convolution layers of MobileNet v1.The aim of V-MOB net is to realize the accurate evaluation of the microstructures and properties of semi-continuous casting Al-Si alloy images.In order to verify the validity of the method,a classification data set semi-continuous casting Al-Si alloy images was constructed.A large number of experimental results show that the classification accuracy of V-MOB net is up to 85.0%.Compared with the existing classification network,the training speed of V-MOB net is improved,and the classification accuracy is guaranteed at the same time.(3)By combining the traditional threshold segmentation method with Faster-RCNN network,a new defect removal method for semi-continuous casting Al-Si alloy images is proposed.In this method,the defects in the semi-continuous casting Al-Si alloy images are detected by Faster-RCNN,and the threshold value is set to realize the pretreatment of the semi-continuous casting Al-Si alloy images.The interference of defects to metallographic analysis is greatly reduced.In order to verify the effectiveness of the method,a detection data set of defects in semi-continuous casting Al-Si alloy images was constructed.A large number of experimental results show that this method can accurately detect and remove defects in semi-continuous casting Al-Si alloy images.(4)A method of microstructure segmentation and post-processing for semi-continuous casting Al-Si alloy images is proposed based on K-means and morphology method,and the statistical information of the segmented microstructures is obtained.In this method,K-means clustering method is used to segment the microstructures.Canny operator edge detection method and morphological method are used to post-process the segmentation result image.Finally,the region growth method is used to calculate the statistical information of the segmented microstructures.A large number of experimental results show that this method can achieve accurate segmentation and analysis of the microstructures in semi-continuous casting Al-Si alloy images.