Particle system macroscopic stress fluctuation prediction method based on convolutional neural network

Abstract

The invention relates to a particle system macroscopic stress fluctuation prediction method based on a convolutional neural network, and the method comprises the following steps: converting the microcosmic plastic deformation of a particle system into a regular three-dimensional voxel matrix based on the microcosmic plastic behavior of particles and a coarse graining method of a Gaussian kernel function; establishing a deep learning data set which takes the three-dimensional voxel matrix as input and takes actual macroscopic stress fluctuation of the particle system as output; and training a convolutional neural network based on the deep learning data set, and establishing a cross-scale prediction model of the quantitative relationship between the microscopic behavior and the macroscopic response of the particle system. According to the particle system macroscopic stress fluctuation prediction method based on the convolutional neural network, the quantitative relation of the particle system from the particle microcosmic plastic behavior to the macroscopic mechanical response can be established, and a brand-new and effective way is provided for macroscopic stress fluctuation prediction and macroscopic and microcosmic cross-scale research of particulate matter.

Description

technical field [0001] The invention relates to the field of granular material mechanics research, in particular to a method for predicting macroscopic stress fluctuations of granular systems based on convolutional neural networks. Background technique [0002] Granular material is a complex system composed of a large number of discrete solid particles interacting with each other. It is ubiquitous in engineering construction, industrial production and all aspects of nature. Debris flows and dams widely exist in nature, and block stone materials used in civil engineering , Coal and sand in industrial production all involve the accumulation and movement of granular materials. Granular materials have characteristics ranging from discrete to continuous, from microscopic to macroscopic, from disordered to ordered, and from flow to clogging, and are one of the key research objects in basic disciplines such as physics, materials, and mechanics. [0003] In the related technology, ...

AI Technical Summary

Problems solved by technology

[0006] Embodiments of the present invention provide a method for predicting macroscopic stress fluctuations in granular systems based on convolutional neural networks to solve the problem that most of the macroscopic and microcosmic connections established in related technologies are qualitative, and it is impossible to quantitatively consider the dynamic behavior and macroscopic mechanical response of the system at the microscopic level relationship problems

Images