Integrated Computational Materials Engineering (ICME)

The component design and fatigue life prediction of turbine disk alloys are critically dependent on the thermomechanical properties of the material. The TriBeam microscope provides a pathway to capture large, targeted 3D microstructural data volumes from turbine disk alloys in order to instantiate models that are used to simulate mechanical loading and assess material properties. The TriBeam experiments used to capture the microstructure, the analysis methods, data requirements, and future needs for 3D infrastructure will be discussed.

Integrated computational materials engineering (ICME) represents a grand challenge within materials research and development. Effective ICME involves coupling materials characterization and experimentation with simulation tools to produce a holistic understanding of the materials system, promising to accelerate the materials development enterprise. Under the Center of Excellence on Integrated Materials Modeling (CEIMM), significant strides were made in developing state-of-the-art experimental methods and simulation techniques for interrogating material structure and behavior across multiple scales. In parallel to these method developments, several advances were made in designing data structures and workflow tools that possess the required flexibility and extensibility to operate on the data produced by such advanced methods. Such software tools are a critical enabling component for effective ICME; the National Academy of Sciences noted cyberinfrastructure as a crucial factor for ICME, to include databases, software, and computational hardware [1]. Additionally, these tools enable workflows that properly integrate models and experimentation at each stage of the materials development lifecycle. Figure 1 schematically shows such a workflow for optimization of microstructure and properties in a titanium forging.

This chapter discusses fundamental aspects of the development of statistically equivalent virtual microstructures (SEVMs) and microstructure and property-based statistically equivalent representative volume elements (M-SERVE and P-SERVE) of the Ni-based superalloy at multiple scales. The two specific scales considered for this development are the subgrain scale of intragranular γ − γ′ microstructures and the polycrystalline scale of grain ensembles with annealing twins. A comprehensive suite of computational methods that can translate microstructural data in experimental methods to optimally defined representative volumes for effective micromechanical modeling is the objective of this study. The framework involves a sequence of tasks, viz., serial sectioning, image processing, feature extraction, and statistical characterization, followed by micromechanical analysis and convergence tests for statistical functions. A principal motivation behind this paper is to translate high-fidelity microstructural image data into statistics of parametric descriptors in constitutive laws governing material performance.

The desire to improve the performance and lifetime of polycrystalline components has fueled the development of advanced micromechanical modeling tools. Multiscale modeling approaches, such as Crystal Plasticity Finite Element Methods (CPFEM), now possess the ability to illuminate the link between material processing, microstructure, and properties [1]. Whereas traditional FE modeling relies on convergent macroscale properties, the ability of CPFEM to explicitly represent the morphology and local crystallographic orientations of polycrystalline microstructures requires scale-specific, quantitative microstructural information for both input and validation. The development and implementation of experimental techniques for capturing behavior and microstructural properties at salient length scales are needed to inform the determination of representative volume elements (RVEs). Here, accurately capturing microstructural details and observing size effects on material properties are both important. Simply extrapolating from average microstructure descriptors does not provide information about the relative importance of specific grain size, shape, and configuration with neighbors. These are features that can be captured experimentally through advanced characterization techniques, such as 3D serial sectioning [2].

A virtual testing methodology to obtain the mechanical response of a polycrystal as function of its microstructure is presented and applied to an Inconel 718 Ni-based superalloy. The mechanical behavior of the polycrystal for a given deformation history is obtained by the finite element simulation of the response of representative volume elements of the microstructure subjected to that particular deformation history. The microstructural information defining the representative volume elements (grain size distribution and texture) was obtained from standard metallographic characterization techniques. The behavior of the alloy crystals is given by a phenomenological crystal plasticity model, whose parameters were obtained using two different strategies, micropillar compression for the parameters defining the monotonic behavior and an inverse optimization strategy (using experimental macroscopic cyclic stress-strain curves) for the parameters controlling the cyclic deformation. From the macroscopic viewpoint, the material response under monotonic and cyclic deformation was in good agreement with the experimental data. At the micro level, the values of the local fields resolved throughout the volume elements were used to generate fatigue indicator parameters, which were able to determine the most critical points in the microstructure to initiate a fatigue crack. These fatigue indicator parameters were calibrated by comparison with a few experimental fatigue tests and then used to predict the effect of loading conditions (strain ranges and ratio) and microstructure (grain size) on the fatigue life of the superalloy. Overall, the strategy shows how a balanced combination of micromechanical and macromechanical tests together with the application of computational homogenization strategies can be used to predict the mechanical behavior of Ni-based superalloys taken into account the influence of the microstructure.

Crystal plasticity constitutive models are frequently used with finite elements for modeling metallic grain-scale phenomena. The accuracy of these models is directly a function of the calibrated parameters, which fully define a crystal plasticity model. A number of techniques exist for the calibration of these parameters. In the current study, a comparison of results using deterministic and non-deterministic calibration methods is made. Additionally, the effect of the type of measured data on calibrated material parameters, global (homogenized) or local, is also presented. Included in the study is a new approach to parameter calibration based on combined digital image correlation and high angular resolution electron backscatter diffraction. Utilizing data from these experimental techniques allows for local evaluation of both strain and relative stress: essentially giving stress-strain curves from numerous point locations in a single coupon. The overall result is that calibration based on sub-grain-scale measurements is preferable when sub-grain-scale phenomena are of primary interest.

Accurately representing the process of porosity-based ductile damage in polycrystalline metallic materials via computational simulation remains a significant challenge. The heterogeneity of deformation in this class of materials due to the anisotropy of deformation of individual single crystals creates the conditions for the formation of a damage field. The work reported upon here is interested in the formation of porosity in the body-centered cubic metal tantalum. This chapter reports on the soft-coupled linkage between a macroscale damage model and mesoscale calculations of a suite of polycrystal realizations of tantalum. The macroscale model is used to represent a tantalum on tantalum plate impact experiment and predict the point in time in the loading profile when porosity is likely to initiate. The 3D loading history from the macroscale calculation is then used to define the probable loading history profile experienced within the experimental sample. Tantalum displays non-Schmid behavior in the motion of the dominant screw dislocations during deformation. This introduces directionality in the magnitude of stress required to propagate glide of these screw dislocations. A model is presented which provides representation of non-Schmid effects in tantalum. This model is employed in performing of meso-scale calculations of statistically equivalent microstructures of the tantalum material to provide local-scale stress condition at the time of the loading profile where initiation of porosity is anticipated. The results of these simulations suggest that non-Schmid effects significantly impact the local stress conditions within the microstructure and are very important to represent. The results also suggest that vonMises stress conditions at grain boundaries and grain boundary triple lines are highly variable close to those features but the variability is reduced with distance to the grain center. The computational results also suggest that the stress traction conditions at the grain boundary are a strong function of the orientation of each boundary with respect to the shock direction. Grain boundaries whose surface normal is parallel to the shock direction have a significantly higher normal tensile traction than other grain boundaries. Grain boundaries whose normal is at 45 or 135 degrees to the shock direction have relatively higher magnitudes of shear stress.

Three-dimensional microstructural information has become increasingly important to advanced computational modeling, and future developments in the field rely on continued maturation of the technology. In this work, a framework for analyzing the error associated with the collection of 3D data sets is proposed. The framework allows users to evaluate errors introduced into the data collection process through the selection of various experimental parameters. Synthetically generated microstructures, called phantoms, are used as a baseline. A simple model for microstructural data collection via electron backscatter diffraction is established. This model is used to simulate data collection from the phantom microstructures, in order to make observations about the effects of individual characterization parameter selection. By comparing simulations to the original phantoms, direct error measurements can be made. Results show how resolution, sample size, and noise can affect the quality of data sets. Finally, the framework is used to show how errors in computational models based on reconstructed microstructures are dependent on the choice of characterization parameters used to generate these microstructures.

The concept of Integrated Computational Materials Engineering (ICME) is aimed at accelerating the development and insertion of new materials in engineering applications. ICME approach relies on the development and use in design of relationships between processing and structure, and its corresponding property/performance. This poses a constraint on computational speed, which is difficult to achieve without losing the physics. The concept of building data-driven, material agnostic models to describe process-structure-property linkages has the potential to satisfy this need. In recent works, this has been introduced on a wide variety of materials at multiple length scales of interest. We review these developments. More specifically, a review of the fusion of material science and data science is presented. The framework addresses curation of materials’ knowledge from the available datasets in computationally efficient manner to extract and use the processing-structure-property relationships.

Epoxies play an important role in determining the performance of epoxy-based composites, coatings, and adhesives. Multiscale modeling methods emerge as a complementary tool to conventional experimental and theoretical approaches and are widely used to study the relationships between processing, structure, and property of polymer materials. This paper aims to provide a review of multiscale modeling efforts on epoxy and epoxy-based materials with a main focus on DGEBA and DGEBF systems. Material’s structural, thermal, mechanical, and interfacial properties are discussed in details.

A computational framework is developed to model the transverse failure of fiber-reinforced polymer-matrix composites, with an emphasis on capturing fiber debonding with a cohesive failure model along the fiber/matrix interfaces. We introduce a nonlinear material sensitivity formulation to quantify how variations in the interfacial cohesive zone properties affect the transverse failure response. The analytic sensitivity formulation is implemented in an interface-enriched generalized finite element method (IGFEM) framework that allows for the simulation of transverse failure in a composite layer consisting of hundreds of closely packed fibers discretized with finite element meshes that do not need to conform to the composite microstructure.

This manuscript presents a computationally efficient method based on a geometric model to simulate the transverse cracking of a 90^∘ cross-ply in a composite laminate. The model expands on existing homogenized solutions of transverse cracking by accounting for the random microstructure of the transverse ply extracted from optical micrographs of a hybrid [0∕90∕0]_T glass/carbon/epoxy composite laminate. The chapter summarizes the three steps of the method, which allows to model the creation of multiple transverse cracks in realistic transverse plies composed of tens of thousands of fibers. The model is calibrated against experimental measurements of the critical values of the applied transverse strain corresponding to the appearance of transverse cracks and then used in a statistical analysis of the impact of the interface strength distribution on the evolution of the transverse cracking process.

The response of heterogeneous materials subjected to extreme dynamic loads is complicated by meso-scale phenomena which manifests a bulk response to the percolating dynamic event. The microstructural arrangement of phases, the extrinsic properties of microconstituents, and the property contrasts and various length scales affect the ability of stress waves to propagate through a material and affect the material’s inherent dissipative behavior. Measuring the effects of these meso-level phenomena is very challenging when considering extremely fast events occurring at multiple spatial and temporal scales. The advent of high performance computing and massively parallel computations allows for highly resolved phenomena to be modeled via hydrocode simulation with relative ease. Combining microstructural and mechanistic understanding of the relevant physics of the dynamic processes can lead to a tractable solution to the problem of shock compression response in heterogeneous materials. This chapter discusses the challenges in understanding the dynamic behavior of heterogeneous materials in particular, which are of interest due to their fascinating and useful properties and in part because of the richness and diversity of phenomena activated under extreme dynamic conditions. A brief literature survey on the shock compression of heterogeneous materials is provided, with attention given to granular media, reactive powder mixtures, energetic and composite materials, and multiphase materials. Case studies from the authors’ work on reactive materials are presented which employ Integrated Computational Materials Science and Engineering (ICMSE) tools to understand the connection between observed experimental behavior and meso-level phenomena. A discussion is presented on possible ways of exploring topology, property contrasts, and microstructural morphology to link dynamic response to micro- and meso-scale behavior.