Research And Application Of Intelligent Recognition Model For Copper-based Material Composition Based On Metallographic Diagram

Copper alloys are important materials widely used in modern industry and daily life.The distribution of their microstructure determines their performance.To improve the performance of copper alloys,it is necessary to extract and analyze the metallographic image information of the alloys.However,due to the complexity of the alloy structure caused by the diversity of the alloy composition and the problem of insufficient sample quantity,existing metallographic image information extraction algorithms cannot meet the requirements of copper alloy metallographic image classification and recognition.To address this issue,this paper proposes two algorithms.The first is an improved generative adversarial network model,which can generate more diverse microstructure images of alloys.The second is an improved convolutional neural network model used for the classification of complex copper alloy metallographic images.The main research contents are as follows:Firstly,the characteristics of copper alloy metallographic images are analyzed,and the dataset is processed using preprocessing methods such as grayscale normalization and image segmentation.Through image segmentation,the high resolution characteristics of the original image are fully utilized,and the dataset is expanded,which lays a foundation for subsequent research on copper alloy classification.Secondly,in view of the problem of insufficient sample quantity of copper alloy metallographic images and the overfitting of classification models,an improved generative adversarial network algorithm for copper alloy metallographic image data augmentation is proposed.By introducing the variational autoencoder and Wasserstein Loss,the performance of the model and the diversity of the generated images are further improved.At the same time,convolutional neural networks are used as the main building blocks of the generator and discriminator,which can generate high-quality and diverse metallographic images of the materials.Experimental results show that the SSIM value between the images generated by GAN-VAE and the original images reaches 0.80.Thirdly,a highly efficient feature fusion network classification algorithm is proposed to address the complexity of copper alloy metallographic images.To reduce the risk of overfitting during training,this method uses data augmentation techniques to expand the training set.Experimental results show that the feature fusion network classification algorithm using the expanded 30% dataset has the best classification performance,with an overall accuracy of98.34%.Fourthly,the application of the copper alloy intelligent recognition model.Based on the research in this paper,the development of the copper alloy intelligent recognition platform interface is completed using Python to achieve copper alloy metallographic organization classification recognition.The system has the advantages of automation,speed,and accuracy,greatly improving the efficiency and accuracy of material research.