Veera Sundararaghavan

A predictive machine learning approach for microstructure optimization and materials design.

This paper addresses an important materials engineering question: How can one identify the complete space (or as much of it as possible) of microstructures that are theoretically predicted to yield the desired combination of properties demanded by a selected application? We present a problem involving design of magnetoelastic Fe-Ga alloy microstructure for enhanced elastic, plastic and magnetostrictive properties. While theoretical models for computing properties given the microstructure are known for this alloy, inversion of these relationships to obtain microstructures that lead to desired properties is challenging, primarily due to the high dimensionality of microstructure space, multi-objective design requirement and non-uniqueness of solutions. These challenges render traditional search-based optimization methods incompetent in terms of both searching efficiency and result optimality. In this paper, a route to address these challenges using a machine learning methodology is proposed. A systematic framework consisting of random data generation, feature selection and classification algorithms is developed. Experiments with five design problems that involve identification of microstructures that satisfy both linear and nonlinear property constraints show that our framework outperforms traditional optimization methods with the average running time reduced by as much as 80% and with optimality that would not be achieved otherwise.

Study of temperature dependence of thermal conductivity in cross-linked epoxies using molecular dynamics simulations with long range interactions

In this work, we demonstrate the use of the Green–Kubo integral of the heat flux autocorrelation function, incorporating long-range corrections to model the thermal conductivity versus temperature relationship of cross-linked polymers. The simulations were performed on a cross-linked epoxy made from DGEBA and a curing agent (diamino diphenyl sulfone) using a consistent valence force field (CVFF). A dendrimeric approach was utilized for building equilibrated cross-linked structures that allowed replication of the experimental dilatometric curve for the epoxy system. We demonstrate that the inclusion of a long-range correction within the Ewald/PPPM approach brings the results close to experimentally measured conductivity within an error of 10% while providing a good prediction of the relationship of thermal conductivity versus temperature. This method shows significant promise towards the computation of thermal conductivity from simulations even before synthesis of the polymer for purposes of materials by design.

Utilization of a Linear Solver for Multiscale Design and Optimization of Microstructures

Microstructures have a significant effect on the performance of critical components in numerous aerospace metallic material applications. Examples include panels in airframes that are exposed to high temperatures and sensors used for vibration tuning. This paper addresses the techniques to optimize the microstructure design for polycrystalline metals. The microstructure is quantified with the orientation distribution function that determines the volume densities of crystals that make up the polycrystal microstructure. The orientation distribution function of polycrystalline alloys (e.g., hexagonal close-packed titanium) is represented in a discrete form, and the volume-averaged properties are computed through the orientation distribution function. The optimization is performed using the space of all possible volume-averaged macroproperties (stiffness and thermal expansion). A direct linear solver is employed to find the optimal orientation distribution functions. The direct solver is capable of finding ex...

A statistical learning approach for the design of polycrystalline materials

An improved semiconductor device and method of manufacturing employs interconnecting films on film circuit as ground lines which extend to the periphery of the film circuit where there is a further connection to a conductive reinforcing plate 25. Advantageously, the conductive reinforcing plate reduces electrical noise from interfering with the semiconductor device and prevents the semiconductor device from radiating undesired signals. The interconnecting films also reduce cross-talk between signal lines of the semiconductor device.

Reconstruction of three-dimensional anisotropic microstructures from two-dimensional micrographs imaged on orthogonal planes

A pervasive method for reconstructing microstructures from two-dimensional microstructures imaged on orthogonal planes is presented. The algorithm reconstructs 3D images through matching of 3D slices at different voxels to the representative 2D micrographs and an optimization procedure that ensures patches from the 2D micrographs meshed together seamlessly in the 3D image. We show that the method effectively models the three-dimensional features in the microstructure using three cases (i) disperse spheres, (ii) anisotropic lamellar microstructure, and (iii) a polycrystalline microstructure. The method is validated by comparing the point probability functions of the reconstructed images to the original 2D image, as well as by comparing the elastic properties of reconstructed image to the experimental data.

Linear Solution Scheme for Microstructure Design with Process Constraints

This paper addresses a two-step linear solution scheme to find an optimum metallic microstructure satisfying performance needs and manufacturability constraints. The microstructure is quantified using the orientation distribution function, which determines the volume densities of crystals that make up the polycrystal microstructure. The orientation distribution function of polycrystalline alloys is represented in a discrete form using finite elements, and the volume-averaged properties are computed. The first step of the solution approach identifies the orientation distribution functions that lead to the set of optimal engineering properties using linear programming. This step leads to multiple solutions, of which only a few can be manufactured using traditional processing routes such as rolling and forging. In the second step, textures from a given process are represented in a space of reduced basis coefficients called the process plane. This step involves generation of orthogonal basis functions for rep...

Atomistic modeling of thermomechanical properties of SWNT/Epoxy nanocomposites

Molecular dynamics simulations are performed to compute thermomechanical properties of cured epoxy resins reinforced with pristine and covalently functionalized carbon nanotubes. A DGEBA-DDS epoxy network was built using the ‘dendrimer’ growth approach where 75% of available epoxy sites were cross-linked. The epoxy model is verified through comparisons to experiments, and simulations are performed on nanotube reinforced cross-linked epoxy matrix using the CVFF force field in LAMMPS. Full stiffness matrices and linear coefficient of thermal expansion vectors are obtained for the nanocomposite. Large increases in stiffness and large decreases in thermal expansion were seen along the direction of the nanotube for both nanocomposite systems when compared to neat epoxy. The direction transverse to nanotube saw a 40% increase in stiffness due to covalent functionalization over neat epoxy at 1 K whereas the pristine nanotube system only saw a 7% increase due to van der Waals effects. The functionalized SWNT/epoxy nanocomposite showed an additional 42% decrease in thermal expansion along the nanotube direction when compared to the pristine SWNT/epoxy nanocomposite. The stiffness matrices are rotated over every possible orientation to simulate the effects of an isotropic system of randomly oriented nanotubes in the epoxy. The randomly oriented covalently functionalized SWNT/Epoxy nanocomposites showed substantial improvements over the plain epoxy in terms of higher stiffness (200% increase) and lower thermal expansion (32% reduction). Through MD simulations, we develop means to build simulation cells, perform annealing to reach correct densities, compute thermomechanical properties and compare with experiments.

Uncertainty Quantification of Microstructural Properties due to Experimental Variations

Electron backscatter diffraction scans are an important experimental input for microstructure generation and homogenization. Multiple electron backscatter diffraction scans can be used to sample th...

Fiber Path Optimization of Symmetric Laminates with Cutouts for Thermal Buckling

Automated fiber placement technology has pushed for the need to explore nonconventional fiber paths in laminated composites. This paper investigates optimal spatially varying fiber paths in a symmetric linear orthotropic laminate, which could increase the critical buckling temperature under uniform applied thermal loads. The key idea here is to achieve gains in buckling performance yet focus on manufacturability of the obtained optimal fiber path. The subject of this study is a four-layer symmetric orthotropic laminated plate, with a central circular cutout that is clamped on all the edges. A novel finite element algorithm is proposed, which imposes a condition on the definition of discrete fiber angles within each element. The main objectives of the proposed finite element approach are to maintain continuity of the fiber paths and to use a computationally efficient model by reducing the number of optimization variables.

A method to predict fatigue crack initiation in metals using dislocation dynamics

Abstract A theory is proposed to predict the initiation of fatigue cracks using cyclic dislocation dynamics (DD) simulations. The evolution of dislocation networks in a grain is simulated over several cycles. It is shown that the dislocation density and the energy stored in the dislocation networks increase with the number of cycles. The results of the DD simulations are used to construct an energy balance expression for crack initiation. A hypothetical crack is inserted into the grain, and the Gibbs energy consisting of the energy of the dislocation structure, the surface energy of the hypothetical crack, and the reduction in continuum energy is evaluated. Once the Gibbs energy attains a maximum, the dislocation structure becomes unstable, and it becomes energetically more favorable to form a real crack. The proposed method is applied to oxygen-free high conductivity copper, and the results are compared against experiments. Finally, it is shown how the method can be amended to account for environmental effects.