Microscopic Image Processing And Analasis Of Polymer Composites

The macroscopic properties of materials are determined by its microscopic structure, so control of macroscopic properties of the material must recognize its microscopic structure and further to describe and characterize them, which have become a very important aspect during the preparation of the materials. The quantitative description and analysis of the microstructure of three kinds of polymer composites were investigated in this paper, and wheat straw/polypropylene composite were highlighted which combined with the other two composites, resin/carbon friction composites and SiC/PTSE nanocomposite. Wheat straw/polypropylene composite material is a new type of green, environmental protection and recycling material, but the composite material of resin matrix is subject to aging phenomenon which will make the mechanical properties decrease substantially with significant changing of its structures. To select the quantitative description method of microstructure of materials and to establish the mathematical model between the microstructure and the macroscopic mechanical properties of material are of great significance to study the aging properties of composite materials. In the accelerated UV aging experiment on wheat straw/polypropylene composite materials SEM pictures of microstructure image of the composites were taken by using JSM-6000 scanning electron microscope under different aging stages (non-aging, twentieth, fortieth, sixtieth, eightieth and 100th aging cycles). Microstructure image processing of wheat straw/polypropylene composites was designed in MATLAB programming environment as an example, including image smoothing filtering, gray level transformation and histogram processing, which can implement image sharpening, image edge detection and image threshold segmentation and so on. AMT method was introduced to describe the wheat straw/polypropylene composite microstructure and the relationships between microscopic structure and macroscopic mechanical properties of the composites were studied. The key achievements and conclusions are listed as follows:(1) Based on digital image processing technology, the corresponding programs of the microscopic image processing of straw/polypropylene composites were programmed, which can realize the image enhancement, image edge detection and image segmentation and so on according to the SEM photos of the material microstructure structure. Adaptive mask filtering method was used to make smooth filtering to the material original images, and this method not only can get the original ones smoothing, but also can preserve the contour of image edges or even make them enhance in some degree. Unsharpened mask method was used to make the original image edges sharpen, and good results of the sharpened edges could be achieved. The aging cracks of edge detection and extraction of material original microscopic images were studied also. The aging cracks of the target edge were extracted by using SUSAN algorithm, and targets are rather clear and complete. K-means clustering method were used to realize the composites microstructure image segmentation and the effects are ideal with high clearness of the image aging crack edges and strong contrast of the targets and background, consequently the processing can give the observation of the image segmentation and bring great convenience to interpretation;(2) The quantitative method of description and characterization of microstructure images of straw/polypropylene composites in different aging period were studied. Creatively AMT method was applied to quantitatively describe the texture complexity of microstructure structure SEM images of straw/polypropylene composites in different aging cycle and MA spectrum of the composite in different aging period were derived according to the principle of this method. The results show that with the increasing of aging periods mechanical properties of composites are on the decline as a whole, while characteristic angle & characteristic scale of MA spectra of them tend to increase overall, there is a high or very high negative correlation between them. Negative correlations are existed between the characteristic angle & characteristic scale of MA spectrum of the composites macrostructure and its macro mechanical properties, especially the extremely high negative correlation in the elastic modulus and flexural strength (-0.882 and-0.974, respectively). Principal component analysis was used to reduce the dimension of MA scale spectrum of the composite in different aging cycle, MA spectrum can be principal component scale map data will be different aging cycle. With PCA approach different aging cycle can be effectively distinguished. The research shows that AMT transform method, as a application of signal transform in scale domain, can be well characterized the complex degree of image texture microstructure of wheat straw/polypropylene composite in different aging period;(3) characteristic parameter extraction method of microstructure image texture of wheat straw/polypropylene composites was researched in different aging period, and principal component analysis (PCA) method was used to extract MA spectrum of microscopic image characteristic parameters of material;(4) Extreme learning machine (ELM) and support vector machine (SVM) were selected to build the intelligent pattern classifier, and studied class recognition of microscopic images of composites in different aging cycle. The results show that both the classification accuracy of classifiers was very high, their classification accuracy of test set reached 95% and 95% respectively. From the point of classification accuracy, extreme learning machine forecast effect slightly higher than the support vector machine classification accuracy, as well as the parameter setting process of ELM is better than SVM. Nevertheless, number of hidden layer neurons of extreme learning machine may take a long time, relatively the training and predict process of SVM is rather simpler and faster than ELM;(5) Extreme learning machine and regression support vector machine (SVR) two algorithms were used to establish regression model between composite microstructure and its flexural strength in different aging period. The results showed that two kinds of the regression fitting precision of regression model were fairly high. Determination coefficient (R=0.9821) of regression model using ELM neural network is slightly higher than that of regression model using SVM (R2=0.9795), but the mean square error (MSE) of regression model using ELM of is much bigger than SVM. Regression model using SVM to prediction of flexural strength of composite materials, the predicting values can closely approximate to real values in a very small error.In the overall, the performance of regression model using SVM to predict output values is superior to that of ELM.