Method for generating titanium alloy microstructure image by inputting multiple mechanical properties

Abstract

The invention discloses a method for generating a titanium alloy microstructure image by inputting multiple mechanical properties. The method comprises the steps: collecting a large number of titanium alloy microstructure images of different microstructure types; setting mechanical property ranges corresponding to the titanium alloy microstructure images of different microstructure types; generating corresponding random mechanical property data for the titanium alloy microstructure images of different tissue types within the mechanical property range, and obtaining the titanium alloy microstructure images of different tissue types and image sets containing the random mechanical property data; performing data amplification on the image set; taking the image set after data amplification as training set data, inputting the training set data into a preset generative adversarial network, and performing respective training according to organization categories to obtain an optimized generative adversarial network model; and inputting a plurality of mechanical property data into the optimized generative adversarial network model to generate the titanium alloy microstructure image. According to the method, a plurality of continuous variables are used as input, and the titanium alloy microstructure image with good quality can be generated.

Description

technical field [0001] The invention belongs to the technical field of computer vision image generation, and relates to a method for generating titanium alloy microstructure images by inputting multiple mechanical properties. Background technique [0002] With the development of technology, titanium alloys have been applied to various fields because of their excellent mechanical properties. At present, most of the research in the field of titanium alloy materials is the forward research of predicting the mechanical properties through the microstructure, but almost no one has realized the process of deducing the microstructure of titanium alloy materials from the mechanical properties. At the same time, there are very few studies on the microstructure in the field of image generation, and almost all of them only focus on a single label. There are few studies on generating microstructure images of titanium alloys by inputting multiple labels. At present, the conditional image...

AI Technical Summary

Problems solved by technology

[0003] 1) Conditional Generative Adversarial Networks can only be used to deal with discrete random variables

[0004] 2) The models that accept parameter input in the current existing methods are all for a single variable, and there is almost no research on multiple continuous variables as input

[0005] 3) Existing image generation methods are almost all for macroscopic objects, such as human faces and objects, and almost no research involving microscopic tissue images

Existing methods do not yield satisfactory results

[0006] 4) Another issue that needs attention is that the correspondence between mechanical properties and microstructure images is one-to-many, that is, a set of mechanical properties corresponds to multiple microstructure images with subtle differences. How to reflect this one-to-many relationship is a more difficult problem to solve

Images