4th World Congress on Integrated Computational Materials Engineering (ICME 2017): Mechanical Performance Using Multi-Scale Modeling
Program Organizers: Paul Mason, Thermo-Calc Software Inc.; Michele Manuel, University of Florida; Alejandro Strachan, Purdue University; Ryan Glamm, Boeing Research and Technology; Georg J. Schmitz, Micress/Aachen; Amarendra Singh, IIT Kanpur; Charles Fisher, Naval Surface Warfare Center

Tuesday 8:00 AM
May 23, 2017
Room: Salon I
Location: Ann Arbor Marriott Ypsilanti at Eagle Crest


8:00 AM  
Scaling from Mesoscale to Macroscale in Viscoelastic Random Composites: Jun Zhang1; Martin Ostoja-Starzewski1; 1University of Illinois
    Microstructural randomness is one of basic characteristics of most solid materials. It affects their mechanical responses through properties of microconstituents, microstructural geometry, and scale of observation. While the macroscopic (homogeneous) continuum description of a material relies on the so-called Representative Volume Element (RVE i.e., macroscale level), the issue studied herein is the scale-dependent trend to it via a Statistical Volume Element (SVE, i.e. mesoscale level) in composites with linear viscoelastic, perfectly-bonded microconstituents, focusing on the quasi-static properties in the time as well as frequency domains. Requiring the material statistics to be spatially homogeneous and ergodic, the mesoscale bounds on RVE response are developed from the Hill-Mandel homogenization condition adapted to viscoelastic materials. The bounds are obtained from two stochastic initial-boundary value problems set up, respectively, under uniform kinematic and traction boundary conditions. The frequency and scale dependencies of mesoscale bounds are obtained through computational mechanics for composites with planar random checkerboard microstructures. In general, the frequency dependent scaling to RVE can be described through a complex-valued scaling function, which generalizes the concept originally developed for linear elastic random composites. This scaling function is shown to apply for all different phase combinations on random checkerboards and, essentially, is only a function of the microstructure and mesoscale.

8:20 AM  
Strategy for Coupling Lengthscales for using Predictive Computation in Polymer Development for Mechanical Performance: Steve Christensen1; E Sharp2; 1Schrodinger; 2e-Xstream Engineering
    Polymers for use as high performance engineering materials dates to the 1960’s. Nearly all of the development of polymer composite materials has been empirical. Even with the recent emergence of molecular dynamics (MD) technology for simulation of polymer chemistry and advances in finite element analysis (FEA) simulation at the continuum level, development of new materials still largely depends on the traditional empirical approach. A key obstacle to materials development via simulation is the difficulty in linking simulation lengthscales from atomic (MD) through micromechanics to continuum. A physics based method utilizing the deformation modes of dilatation and/or distortion will be shown to provide new insight into polymer formulation. The method is a hierarchical multi-scale approach that relies on a new continuum analysis concept referred to as “Onset Theory” to provide the macroscopic guidance for polymer ultimate behavior leading to a new formulation. Condensed matter physics provides additional insight in how deformation is influenced by intra and intermolecular forces within covalently bonded polymeric materials. The method was first developed for use in describing behavior of composites. In addition, Molecular Dynamics can be of use to simulate the influence of small penetrant molecules and temperature on elastic properties leading to an inference on environmental resistance. We will show that a multiscale computational system using Molecular Dynamics, micromechanics modeling, and Onset Theory continuum modeling, with an appropriate protocol, can be used to perform virtual formulation and will yield results suitable for guiding materials development, leading to experiment and polymer evaluation demonstrating a performance improvement

8:40 AM  
Hybrid Hierarchical Model for Damage and Fracture Analysis in Heterogeneous Material: Alex Vasenkov1; 1Sunergolab Inc
    Predictive damage and fracture analysis of heterogeneous material is grueling as such material consists of mesoscale components of different size and functionality. A logical approach taken by many researchers in tackling this challenge is to employ a framework that couples Molecular Dynamic (MD) and Finite Element (FE) modeling in some manner to capture damage processes occurring at different time and length scales. Unfortunately, such coupling typically suffers from lack of thermodynamic consistency between MD and FE models and the phenomena of pathological wave reflection, which occurs at the interface between MD and FE simulation regions. This works attempts to circumvent this problem with a Hybrid Hierarchical Model (HHM) involving an ab-initio based ReaxFF MD module and a peridynamic continuum module. The HHM framework was applied to perform damage and fracture analysis in a silicon carbide slab with pre-crack. The ReaxFF modeling was conducted around the crack region where the mechanical strains were the greatest and the most detailed resolution and the highest accuracy were required to predict the crack path. Peridynamic simulation was performed with increasing scale to provide more smeared and continuum-like solution for the crack path and capture correctly the waves caused by the crack propagation. The HHM was able to provide an atomistic-based description of damage and fracture mechanisms without contamination with pathological reflections of waves from the MD boundaries, which are difficult to account in traditional MD simulation because of severe limitations on the size of simulation domain.

9:00 AM  
Development of Integrated Computational Materials Engineering (ICME) Approach for Compression-Molded Chopped Carbon Fiber Sheet Molding Compounds (SMC) in Vehicle Light-weighting Applications: Yang Li1; Hongyi Xu1; Zhangxing Chen2; Danielle Zeng1; Jeffery Dahl1; Mansour Mirdamadi3; Xuming Su1; 1Ford Motor Company; 2Chongqing University; 3The Dow Chemical Company
    While the applications of continuous carbon fiber composites have been widely seen in many industries, chopped carbon fiber sheet molding compounds (SMC) made through compression molding process, featuring improved balance between weight reduction, mechanical properties and manufacturing cost, is regarded as an alternative promising material system for light-weighting products in automobile industry. However, the inhomogeneous and anisotropic mechanical behavior of the material hinders its deployment in vehicle part design. Furthermore the material properties are observed to be highly sensitive to the processing conditions, creating additional challenges for real-life applications. In the present study, an Integrated Computational Materials Engineering (ICME) approach is applied to develop a modeling framework which enables the computer-aided optimal design of the carbon fiber SMC parts without the necessity of performing a massive number of experiments in traditional trial-and-error procedures. The simulation of compression molding process realized in Autodesk Moldflow provides local microstructure information of the SMC material, which is then utilized in a multi-scale material modeling tool and generates homogenized material properties for macroscopic structural finite element analysis (FEA) to predict the part performance. Validations of the ICME models at different stages are provided through comparison between modeling prediction and experimental measurements. The workflow is automated with assistance from homemade MATLAB scripts and embedded into an optimization platform built in Esteco modeFrontier. An exemplar optimal design procedure on a vehicle subframe utilizing the developed ICME models is demonstrated.

9:20 AM  
Automated Composite Material Model Development with Multiscale Designer: Colin McAuliffe1; Robert Crouch1; Jeff Wollschlager1; Jacob Fish2; Venkat Aitharaju3; Roger Ghanem4; 1Altair Engineering Inc; 2Columbia University; 3General Motors Company; 4University of Southern California
     Development of accurate computational models for composite materials presents a challenge for industry practitioners. The input parameters to a multiscale model of a composite are the properties of the constituent materials, e.g. carbon fiber and epoxy. It is often not possible to fully characterize a complex constituent material like carbon fiber from performing experiments on them individually. Even if it were, manufacturing processes may result in different in situ properties. Model development therefore must be accomplished through inverse identification, where experimental observations on the composite are used with a model to infer the constituent properties. This presents its own set of challenges, since inverse problems are notoriously difficult to solve correctly. Additionally, the success of the inverse solution requires that each of the constitutive properties be constrained by at least one experimental observation. Due to the expense associated with testing of composites, there is strong motivation to minimize the number of tests.Multiscale Designer employs both deterministic and stochastic approaches to the model development problem for linear and nonlinear property identification. These approaches are automated and highly accessible to practicing engineers. The stochastic approaches provide crucial insight into the reliability of the inverse solution, and can inform the analyst when the lack of a particular experimental observation negatively impacts this solution. In this case, a purely deterministic approach may give the false impression of a good solution. In this presentation, the technical details of these approaches are presented, and their ability to successfully identify composite properties is demonstrated.

9:40 AM  
Creep Damage Calculation of Weld Joint of 9Cr-1Mo-V-Nb Steel Tube for Creep Lifetime Prediction (P-17): Kozo Koiwa1; Masaaki Tabuchi1; Masahiko Demura1; Takaaki Matsuoka2; Keisuke Torigata2; Masayoshi Yamazaki1; Makoto Watanabe1; 1National Institute for Materials Science; 2IHI Corporation
    Prediction of creep lifetime for weld joint of 9Cr-1Mo-V-Nb steel (ASME Gr. 91) tube was conducted by using computational simulation. The creep lifetime was predicted from the creep damage accumulation computed with the creep parameters obtained from the creep tests of base metal and heat-affected zone. Two type of creep damage mechanics analyses were applied, as follows: The Hayhurst-type analysis and the time exhaustion analysis assuming Norton creep law. The computed creep lifetime was compared with the experimental results to investigate the prediction accuracy of the computational creep simulation with the module can assess the creep lifetime of structural parts including varied materials.

10:00 AM Break

10:30 AM  
Understanding the Role of Chemical Composition in the Formability and Mechanical Properties of Ni-base Superalloys: Enrique Galindo-Nava1; Catherine Rae1; 1University of Cambridge
    Nickel-base superalloys are widely employed in high-temperature structural components due to their excellent mechanical and environmental properties. These features can only be achieved by tailoring sophisticated microstructures through different thermomechanical routes, which are usually tailored for specific alloying contents. This work presents a systematic study on the role of chemical composition in the hot deformation and mechanical properties of Ni-base superalloys. A number of modelling strategies are integrated as a function of chemical composition to link microstructure evolution with tensile and creep strength. Evolution equations for dynamic recrystallization and grain growth including multicomponent solute drag and particle pinning are presented; it is explored how typical alloying elements present in superalloys affect the strain to reach complete recrystallization, altering the processing window of a component. An overview of compositional effects in precipitation evolution using existing simulation tools is presented. Microstructure-sensitive models for yield and creep strength developed recently are also introduced. The models are based on estimating the relative strengthening contributions of solid solution, grain boundary and multimodal particle shearing. The chemistry not only dictates the strength of each constituent but it also affects the values of interfacial energies; the latter alter the occurrence of different deformation mechanisms. The methodology is applied to identify optimal processing routes for improving mechanical properties in commercial Ni-base superalloys.

10:50 AM  
A Strain Energy Based Damage Model for Fatigue Crack Initiation and Growth (P-5): Peter Huffman1; 1John Deere
    A strain energy based fatigue damage model is proposed which uses the strain energy from applied loads and the strain energy of dislocations to calculate stress-life, strain-life, and fatigue crack growth rates. Stress ratio effects intrinsic to the model are discussed, and parameterized in terms of the Walker equivalent stress and a fatigue crack growth driving force. The method is then validated using a variety of different metals with strain-life data and fatigue crack growth rate data available on the SAE Fatigue Design & Evaluation subcommittee database.

11:10 AM  
Image-based Micromechanical Fatigue Simulation of Cast Aluminum Alloy by Parallel Finite Element Method: Osamu Kuwazuru1; Masaki Teranishi1; Shota Gennai1; Masataka Uchida1; Masakazu Kobayashi2; Hiroyuki Toda3; 1University of Fukui; 2Toyohashi University of Technology; 3Kyushu University
    Low-cycle fatigue test of cast aluminum alloy and its in-situ observation were performed in the synchrotron radiation facility SPring-8. The inclusions such as silicon and intermetallic particles were visualized by the X-ray CT. The crack initiation site and the particle of crack source were identified by the chronological CT observation. Then, the particle fracture life was also determined. A variety of the particle shapes and the particle fracture lives were selected, and the micromechanical finite element model of those particles was semi-automatically constructed with a lot of surrounding particles by the image-based modeling technique. The cyclic loading corresponding to the actual load in the experiment was applied to the finite element model. The material nonlinearity and the geometrical nonlinearity was considered, and the kinematic multi-linear hardening law was employed. Since the number of elements exceeded a ten million, we adopted a massively-parallel computing with a PC cluster system. Through these simulation results, the fatigue crack initiation mechanism and the effect of particle shape and distribution on the particle fracture life were discussed.

11:30 AM  
Fatigue Performance Prediction of Structural Materials by Multi-Scale Modeling and Machine Learning: Takayuki Shiraiwa1; Fabien Briffod1; Yuto Miyazawa1; Manabu Enoki1; 1The University of Tokyo
    Structural materials having higher performance in strength, toughness, and fatigue resistance are strongly required. In the conventional materials development, many fatigue tests need to be conducted to validate statistical behavior of fatigue failure. Accordingly the evaluation of fatigue properties with shorter time becomes quite essential. Based on such background, we are developing fatigue prediction methods for wide range of structural materials by multi-scale finite element analysis (FEA) and machine learning. The multi-scale FEA consists of the following procedures: i) mechanical and thermal properties are estimated by using commercially available software and database; ii) temperature field, residual stress and distortion generated during a manufacturing process is calculated on the macroscopic model by thermo-mechanical FEA; iii) macroscopic stress field under cyclic loading condition is calculated with a hardening constitutive model; iv) the microscopic stress field is derived by finite element model with the polycrystalline structures and the cycles for a fatigue crack initiation is analyzed by Tanaka-Mura model; v) the cycles for fatigue crack propagation is analyzed by extended finite element method (X-FEM) and the total number of cycles to the failure is obtained. The second approach is to use empirical equation and fatigue database accumulated over the years. Materials and fatigue data have been aggregated from public databases, published papers and academic resources. Several empirical equations have been derived by applying machine learning techniques to the database. The accuracies and the features of our prediction methods will be discussed.

11:50 AM  
Multiscale Modeling of Deformation Response of Polycrystalline Alloys using Orientation Distribution Functions: Ali Ramazani1; Veera Sundararaghavan1; 1University of Michigan
    An ICME approach to model texture evolution and mechanical behavior of polycrystalline alloys during deformation processing is presented. EBSD measurements were made in all surfaces (L-LT, L-ST and LT-ST) of samples in forged and compressed specimens. We have quantified the percentage of recrystallization in the forged specimens as well as the heat treated ones using grain orientation spread maps. Based on the experimental results, an orientation distribution function (ODF) model is developed to simulate the texture evolution at different intermediate strains in various loading conditions. Here, the polycrystalline microstructure is represented through a finite element discretized orientation distribution function and texture evolution is modeled using ODF conservation laws via Taylor assumption. Additionally, an elasto-visco plastic crystal plasticity framework is utilized to predict crystal reorientation. The ODF model is coupled to every integration point in a macro-micro Lagrangian finite element algorithm to predict the deformation response of the bulk material under compression. The approach results in a multiscale prediction where texture evolution is identified at every point in the macro-specimen. The ODF model is fully coupled to the macroscale mesh and can be used to predict spatial distribution of texture components. Numerical results show very good agreement with the experimental texture measurements.

12:10 PM Break