Materials knowledge system for nonlinear composites

Abstract

In this contribution, we present a new Materials Knowledge System framework for microstructure-sensitive predictions of effective stress-strain responses in composite materials. The model is developed for composites with a wide range of combinations of strain hardening laws and topologies of the constituents. The

Three-dimensional Analysis and Reconstruction of Additively Manufactured Materials in the Cloud-Based BisQue Infrastructure

“…Capabilities for raw data storage, reconstruction, and analysis of 3D microstructures discussed here are also useful because they can be combined with tools for predicting microstructure-sensitive properties. Such tools which may include analytic bounds (e.g., Voigt/Reuss) or emerging data-driven models for microstructure-property relationships [50] can also be incorporated into BisQue, which is discussed in detail elsewhere [51].…”

Autonomous Development of a Machine-Learning Model for the Plastic Response of Two-Phase Composites from Micromechanical Finite Element Models

“…Mean-field theories approximate the effective response of composites by assuming uniform stress and/or strain fields within the constituents. 30,31 The specific approach employed in this work builds on the approach described initially by Stringfellow et al, 32,33 and more recently by Latypov et al 19 The problem of interest here is the homogenized equivalent stress-equivalent plastic strain response of a composite material made of two isotropic constituents (i.e., thermodynamic phases), both following a J 2 -based associated flow rule 34 with distinct hardening responses. The composite microstructure is assumed to be adequately captured in a voxelized 3-D representative volume element (RVE).…”

“…In the recent work of Latypov et al, 19 the approach of Stringfellow and Parks 33 was extended to include the higher-order details of the material microstructure. This was accomplished by extending the MKS framework 2 as shown schematically in Fig.…”

“…Of particular relevance to this work are the approaches used to build the reduced-order models after the feature engineering step (i.e., after the low-dimensional representations of the material structure are obtained employing n-point spatial correlations and PCA). In most of the prior work, 16,17,19,20 surrogate models were built using standard regression techniques. These techniques are indeed ideally suited when there is a reasonably large dataset available, and there is already evidence (or physical insights) suggesting a specific model form for capturing the governing physics of the problem.…”

“…This specific case study was selected because the viability of the standard MKS regression techniques for this problem has already been established. 19 Therefore, it offers an opportunity to evaluate critically the computational cost of training the new GPR-based MKS models and their predictive accuracy compared to the previously established models based on standard regression methods.…”