Thursday 2:00 PM

March 2, 2017

Room: 10

Location: San Diego Convention Ctr

Successful modeling of a material’s deformation behavior is dependent on the development of realistic constitutive models. In-depth knowledge about the parameters in a constitutive model will lead to a better understanding of the relationship between flow stress and deformation conditions and will eventually aid in better design of the deformation process. Curve-fitting is generally employed to train constitutive models using experimental data. However, relative impact of constitutive model parameters on the mechanical response for different models has not been explored much. In this study, both local and global sensitivity analysis methods are used. Canonical correlation analysis (CCA) has been used to understand the effect of constitutive model parameters on the flow stress behavior of titanium alloys. The limitations of local sensitivity methods have been highlighted and it has been shown that CCA provides a measure of both individual variable’s contribution and the effectiveness of the parameter set as a whole.

A critical risk assessment for handheld consumer electronic products is the probability of fracture for glass components, namely the display, during drop impact. Glass strength can be determined through four point bend testing and combined with product level simulations for multiple drop orientations, producing a temporal and spatially varying stress state on the glass. The simplified technique of using only peak values in the worst orientations is generally over-conservative and not effective for competitive design where reliability must compromise with cost, weight and industrial design requirements. An effective risk assessment method is discussed which uses the FEA results directly and addresses the size and rate effects on glass strength, as well as the temporal and spatial nature of the stress results to produce a probability of failure. The method is shown to have excellent correlation with multiple product test cases using both EPD and LCD systems.

In this study, we investigate ion transport behavior in a dual phase system influenced by distribution and shape variation of the phases. Charge transport is represented by Nernst-Planck equation in a finite element model developed using COMSOL Multiphysics. We use a mathematical model to generate normally distributed phases to be used as geometric domain representing the dual phase system. Shape variation is obtained by using different regular geometric shapes as the representative phases. Charge transport can occur by diffusion, migration and convection and driven by chemical or electric potential and velocity. This ion transport behavior study gives insight to different multiphase systems, i.e., ceramic separation membrane, ceramic waste form system for nuclear waste materials.

Nickel based superalloys, whose strength is derived from its γ-γ’ microstructure, have long been used for high temperature applications. To better understand the effect of the microstructure on the mechanical response, we present an uncertainty quantification framework for the calibration of a dislocation density based hierarchical crystal plastic model for the nickel based superalloy Rene’88DT. The framework consists of a genetic algorithm that can use two sub algorithms: A method to generate statistically equivalent sub-grain microstructures based on microstructural parameters, and a dislocation density based CPFEM model. The genetic algorithm perturbs parameters of the constitutive models as well as the microstructural parameters. The genetic algorithm fits constitutive model parameters, while simultaneously gaining understanding of the uncertainty and sensitivity of both microstructure, and constitutive parameters. Decomposition of variance determines which parameters are relevant to the mechanical response. Irrelevant microstructural parameters will not be perturbed in subsequent rounds of the genetic algorithm.

Combining theory, models and simulation, computational materials science has become the third leg of materials discovery and development, along with materials synthesis and characterization. While the analysis of errors is quite well developed for experiments, such analysis for simulations, particularly for simulations linked across length and time scales, is much less advanced. In this talk we will discuss the two extremes of the types of multiscale simulations, sequential multiscale and concurrent multiscale, and the role that uncertainty quantification (UQ) can play in each. We will base the discussion on specific examples from our work and the rather sparse literature on UQ in materials simulations. We will end by identifying needs for conceptual advances, needs for the development of best practices, and needs for specific implementations.

A factorial Design of Experiments (DOE) parametric study using Finite Element Analysis (FEA) of Polymer/Clay Nanocomposites (PCN) was completed to expose the most crucial parameters affecting the performance of PCNs under quasi-static tension. Three-dimensional FEA featuring a Cohesive Zone Model (CZM) based on results of Molecular Dynamics (MD) simulations was used to perform a DOE parametric study on Polyvinyl Alcohol (PVA)/Montmorillonite (MMT) nanocomposite systems. The DOE process utilized an Analysis of Variance (ANOVA) technique to analyze the relative influence of four parameters related to nanoclay particles within a polymer matrix (aspect ratio, orientation, intercalation, and gallery strength) with respect to the overall PCN mechanical performance while including uncertainty principles. Additionally a Monte Carlo (MC) routine featuring a Radial Basis Function (RBF) provided quantification for the uncertainty related to the multiscale modeling methodology used in this endeavor.

Two fundamental sources of error in macroscale solid-mechanics modeling are (1) the assumption of a separation-of-scales in homogenization theory and (2) the use of a macroscopic material model that represents, in an average sense, the complex processes occurring at the microscale. Macroscopic material models attempt to approximate the response of the material under complex loading conditions, and are typically in error when exercised outside of the calibration regime. These approximation errors may be particularly significant in welded regions of a structure and for additively manufactured (AM) structures. In order to quantify these approximation errors on macroscale quantities-of-interest, we adopt an a posteriori error-estimation framework. We demonstrate how this framework can be used in an uncertainty-quantification mode that considers microstructural variability in both grain morphology and texture. For an example, we will study the multiscale mechanical response of a 304L stainless steel tube manufactured using a selective laser melting AM process.

Phase field modeling has become significantly more popular in materials science and engineering and is becoming a mainstream technique. As a result, the growing phase field community continues to develop a wide variety of codes; but it lacks benchmark problems to consistently evaluate, validate, and verify new implementations. Following the example set by the micromagnetics community, the Center for Hierarchical Materials Design (CHiMaD) and the National Institute of Standards and Technology (NIST) are developing benchmark problems that test individual numerical or physical aspects of the codes. We discuss the benchmark problems developed to date, which focus on the diffusion of solute and the growth and coarsening of a second phase, linear elastic solid mechanics, and solidification. We demonstrate the utility of these problems by comparing the results of simulations performed with different numerical techniques. Finally, we discuss the needs of future benchmark problems and how the community can be involved.

Quantification of uncertainties with disparate origins is imperative for predictive simulations in materials science and engineering. Functional uncertainty quantification (FUQ) described in this work focuses directly on uncertainties originating from input constitutive functions as opposed to its parameters as it is typically done. These uncertainties primarily stem from approximations of the physics in a given model. Calculation of functional derivatives yields the sensitivity of output quantities of interest (QoI) to changes in the input function and, with additional possible input functions, QoIs can be directly corrected from one function to another without additional simulation. Examples of materials modeling and functions of interest showing the breadth and utility of the method are described: molecular dynamics (interatomic potentials), viscoplasticity self-consistent methods for texture evolution (grain interactions), and solidification modeling (permeability functions).