Craig Burkhart

Efficient 3D porous microstructure reconstruction via Gaussian random field and hybrid optimization

Obtaining an accurate three‐dimensional (3D) structure of a porous microstructure is important for assessing the material properties based on finite element analysis. Whereas directly obtaining 3D images of the microstructure is impractical under many circumstances, two sets of methods have been developed in literature to generate (reconstruct) 3D microstructure from its 2D images: one characterizes the microstructure based on certain statistical descriptors, typically two‐point correlation function and cluster correlation function, and then performs an optimization process to build a 3D structure that matches those statistical descriptors; the other method models the microstructure using stochastic models like a Gaussian random field and generates a 3D structure directly from the function. The former obtains a relatively accurate 3D microstructure, but computationally the optimization process can be very intensive, especially for problems with large image size; the latter generates a 3D microstructure quickly but sacrifices the accuracy due to issues in numerical implementations. A hybrid optimization approach of modelling the 3D porous microstructure of random isotropic two‐phase materials is proposed in this paper, which combines the two sets of methods and hence maintains the accuracy of the correlation‐based method with improved efficiency. The proposed technique is verified for 3D reconstructions based on silica polymer composite images with different volume fractions. A comparison of the reconstructed microstructures and the optimization histories for both the original correlation‐based method and our hybrid approach demonstrates the improved efficiency of the approach.

Multiscale modeling of polyisoprene on graphite

The local dynamics and the conformational properties of polyisoprene next to a smooth graphite surface constructed by graphene layers are studied by a multiscale methodology. First, fully atomistic molecular dynamics simulations of oligomers next to the surface are performed. Subsequently, Monte Carlo simulations of a systematically derived coarse-grained model generate numerous uncorrelated structures for polymer systems. A new reverse backmapping strategy is presented that reintroduces atomistic detail. Finally, multiple extensive fully atomistic simulations with large systems of long macromolecules are employed to examine local dynamics in proximity to graphite. Polyisoprene repeat units arrange close to a parallel configuration with chains exhibiting a distribution of contact lengths. Efficient Monte Carlo algorithms with the coarse-grain model are capable of sampling these distributions for any molecular weight in quantitative agreement with predictions from atomistic models. Furthermore, molecular dynamics simulations with well-equilibrated systems at all length-scales support an increased dynamic heterogeneity that is emerging from both intermolecular interactions with the flat surface and intramolecular cooperativity. This study provides a detailed comprehensive picture of polyisoprene on a flat surface and consists of an effort to characterize such systems in atomistic detail.

A Transfer Learning Approach for Microstructure Reconstruction and Structure-property Predictions

Stochastic microstructure reconstruction has become an indispensable part of computational materials science, but ongoing developments are specific to particular material systems. In this paper, we address this generality problem by presenting a transfer learning-based approach for microstructure reconstruction and structure-property predictions that is applicable to a wide range of material systems. The proposed approach incorporates an encoder-decoder process and feature-matching optimization using a deep convolutional network. For microstructure reconstruction, model pruning is implemented in order to study the correlation between the microstructural features and hierarchical layers within the deep convolutional network. Knowledge obtained in model pruning is then leveraged in the development of a structure-property predictive model to determine the network architecture and initialization conditions. The generality of the approach is demonstrated numerically for a wide range of material microstructures with geometrical characteristics of varying complexity. Unlike previous approaches that only apply to specific material systems or require a significant amount of prior knowledge in model selection and hyper-parameter tuning, the present approach provides an off-the-shelf solution to handle complex microstructures, and has the potential of expediting the discovery of new materials.

Stochastic Reassembly for Managing the Information Complexity in Multilevel Analysis of Heterogeneous Materials

Efficient and accurate analysis of materials behavior across multiple scales is critically important in designing complex materials systems with exceptional performance. For heterogeneous materials, apparent properties are typically computed by averaging stress-strain behavior in a statistically representative cell. To be statistically representative, such cells must be larger and are often computationally intractable, especially with standard computing resources. In this research, a stochastic reassembly approach is proposed for managing the information complexity and reducing the computational burden, while maintaining accuracy, of apparent property prediction of heterogeneous materials. The approach relies on a hierarchical decomposition strategy that carries the materials analyses at two levels, the RVE (representative volume element) level and the SVE (statistical volume element) level. The hierarchical decomposition process uses clustering methods to group SVEs with similar microstructure features. The stochastic reassembly process then uses t-testing to minimize the number of SVEs to garner their own apparent properties and fits a random field model to high-dimensional properties to be put back into the RVE. The RVE thus becomes a coarse representation, or “mosaic,” of itself. Such a mosaic approach maintains sufficient microstructure detail to accurately predict the macro-property but becomes far cheaper from a computational standpoint. A nice feature of the approach is that the stochastic reassembly process naturally creates an apparent-SVE property database. Thus, material design studies may be undertaken with SVE-apparent properties as the building blocks of a new material’s mosaic. Some simple examples of possible designs are shown. The approach is demonstrated on polymer nanocomposites.Copyright

A Hybrid Approach to 3D Porous Microstructure Reconstruction via Gaussian Random Field

Obtaining an accurate three-dimensional (3D) structure of a porous microstructure is important for assessing the material properties based on finite element analysis. While directly obtaining 3D images of the microstructure is impractical under many circumstances, two sets of methods have been developed in the literature to generate (reconstruct) 3D microstructure from its 2D images: one characterizes the microstructure based on certain statistical descriptors, typically two-point correlation function and cluster correlation function, and then performs an optimization process to build a 3D structure that matches those statistical descriptors; the other method models the microstructure using stochastic models like a Gaussian random field (GRF) and generates a 3D structure directly from the function. The former obtains a relatively accurate 3D microstructure, but the optimization process can be very computationally intensive, especially for problems with a large image size; the latter generates a 3D microstructure quickly but sacrifices the accuracy. A hybrid optimization approach of modeling the 3D porous microstructure of random isotropic two-phase materials is proposed in this paper, which combines the two sets of methods and hence maintains the accuracy of the correlation-based method with improved efficiency. The proposed technique is verified for 3D reconstructions based on silica polymer composite images with different volume fractions. A comparison of the reconstructed microstructures and the optimization histories for both the original correlation-based method and our hybrid approach demonstrates the improved efficiency of the approach.Copyright

Interfacial properties of carbon-rubber interfaces investigated via indentation pull-out tests and the JKR theory

Properties of the interface between the filler and the matrix of a composite material draw research attention due to their contributions to the overall properties of the material, especially when the filler and the matrix differ significantly from each other. The work reported in this paper investigates the interface between amorphous carbon and a cross-linked synthetic natural rubber. The interface was experimentally simulated with the surfaces of a sputtering-coated carbon on a spherical Al2O3 tip and a flat synthetic natural rubber sample. Step-loading and pull-out tests with a micro-/nano-indentation instrument were conducted. Fully relaxation of the samples occurred during both test procedures. The penetration depth, applied load, and experimental time were recorded during each test. The Johnson–Kendall–Roberts theory was used to analyze the data at the initial point (step-loading) and the final surface separation point (pull-out) to obtain the initial equivalent modulus, infinite equivalent modulus, work of adhesion, and the average normal interfacial strength at separation. It is found that the pull-out force and the work of adhesion depend on the unloading rate, but the infinite equivalent modulus and the average interfacial strength in the normal direction of the carbon–rubber interfaces are independent of the unloading rate in current experimental domain.