A two-way prediction method of material microscopic image and performance based on deep learning

Abstract

The invention relates to a two-way prediction method of material microscopic image and performance based on deep learning, and proposes an algorithm combining CNN and DNN to process parallel prediction of multiple inputs. Using CNN to extract material microscopic image features, add one-dimensional feature vector represented by material composition and process in the first fully connected layer. Integrated image features, other features are DNN input, to achieve regression prediction or classification of material comprehensive features and performance. On the basis of a large number of images generated by GAN, the performance of the generated images is predicted using the previous prediction model. In order to increase the credibility of the experimental results, three different networks can be trained for prediction, and finally the intersection of the three is output as the target image. By combining the convolutional neural network to extract features and features such as components and processes, the fitting degree of the model and performance is greatly improved, avoiding model underfitting, and the prediction method of performance classification meets the actual needs of production.

Description

technical field [0001] The invention belongs to the technical field of computer vision material image recognition, and relates to a deep learning-based bidirectional prediction method for material microscopic images and properties Background technique [0002] Materials are the material basis for human survival and development. The discovery and use of every important material in history will bring huge changes to human living standards and social productivity. Therefore, materials have always played an important role in the development of human society, and together with energy and information, they are called the three pillars of modern civilization. The research and development of new material technology plays a decisive role in the development of the country. The traditional material development method has a long cycle and high cost, and is mainly based on the trial-and-error method, that is, using the existing theory and knowledge and experience about materials, by adj...

AI Technical Summary

Problems solved by technology

Due to the high dissipation of plastic deformation, the microstructure inside the material has a large scale span and various types under different loading conditions, making it extremely challenging to construct a universal and predictable multi-scale model based on microstructure characteristics.

Images