Aluminum Alloy Inclusions Identification And Automatic Recognition

In aluminum alloy, different kinds of inclusions are harmful to the microstructure and properties of alumina alloy. It’s necessary to analyze the type and origin of these inclusions in order to improve the properties of aluminum alloy. As a quality control tool, PoDFA system can do qualitative and quantitative analysis of inclusions in alumina alloy. The inclusions are concentrated by filtrating the molten aluminum and then analyzed under metalloscope. This method is very effective and accurate, but it takes metallographers plenty of time. With the development of digital image processing, we consider to apply digital image analysis to the analysis of inclusions. The basic idea is that first we collect metallographic images of alumina alloy by image acquisition system, and then these metallographic images are analyzed through the software designed by ourselves instead of metallographers. The work efficiency can be improved and cost can also be save to some extent.The features of inclusions in aluminum alloy we can observe under metalloscope have been studied and the inclusions have been classified and identified by metallographic analysis and electronic probe analysis; characteristics of metallographic images of alumina alloy inclusions have been studied and then image preprocessing program have been made according to these image features; best image segmentation algorithm has been confirmed after trying several ordinary image segmentation algorithms; the features of inclusions have been extracted and the results of features extraction have been analyzed; finally, automatic recognition program for recognizing aluminum inclusions has been developed and then the automatic program has been tested. The major conclusions in this paper are shown as follows:(1) Ordinary inclusions have been identified by the metallographic image analysis and electron probe microprobe analysis and the relationships between digital image features and types have been obtained. (2) The metallographic images in this experiment present uneven illumination. The effect of background light compensation by manual collection and homomorphic filtering isn’t obvious. The top-hat transformation method with a150μm diameter disc-shaped structural element achieve a remarkable effect.(3) Otsu segmentation, watershed segmentation and edge detection segmentation can not segment the image effectively. According to the features of the metallographic images, edge detection-closing operator has been proposed. The specific process is as follows:first, edge detection with Sobel operator is used to process the image; then closing operation with a3μm diameter disc-shaped structural element is used to process the image; afterwards, union set of the first two steps and Otsu segmentation image is taken; the regions obtained from Otsu segmentation are marking in order to segment the inclusions that accompany with each other This method can achieve remarkable effect.(4) The gray value of aluminum martix has been defined in this paper and algorithm for counting gray value of aluminum alloy has been given; after conforming the gray value of aluminum alloy region of different metallographic images, gray value features, shape features, texture features and accompanying relationship features.(5) According to the features of each kind of inclusion we obtained, the automatic recognition program for recognizing aluminum inclusions has been developed; then the automatic recognition program has been tested by analyzing the metallographic images and satisfying result have been achieved.