The ability to relax a macroscopically applied stress is often associated with molecular mobility, or the possibility for a molecule to move outside the confines of its current position, within the material of which the stress is applied. Here, a viscoelastic constitutive analysis is used to investigate the counter-intuitive experimental observation of “mobility decrease with increased deformation through yield” [1] for a glass forming polymer during stress relaxation while under compressive and tensile loading conditions. The behavior of an epoxy thermoset is examined using an extensively validated, thermorheologically simple, material “clock” model, the Simplified Potential Energy Clock (SPEC) model.[2] This methodology allows for a comparison between the linear viscoelastic (LVE) limit and the true non-linear viscoelastic (NLVE) representation and enables exploration of a wide range of conditions that are not practical to investigate experimentally. The model predicts the behavior previously described as “mobility decrease with increased deformation” in the LVE limit and at low strain rates for NLVE. Only when loading rates are sufficient to decrease the material shift factor by multiple orders of magnitude is the anticipated deformation induced mobility or “mobility increase with increased deformation” observed. While the model has not been “trained” for these behaviors, it also predicts that the normalized stress relaxation response is indistinguishable amongst strain levels in the “post-yield” region, as has been experimentally reported. At long time, which has not been examined experimentally, the model predicts that even the normalized relaxation curves that exhibit “mobility increase with increased deformation” “cross back over” and return to the LVE ordering. These findings demonstrate the ability of rheologically simple models to represent the counter-intuitive experimentally measured material response and present predictions at long time scales that could be tested experimentally.
Sylgard® 184/Glass Microballoon (GMB) potting material is currently used in many NW systems. Analysts need a macroscale constitutive model that can predict material behavior under complex loading and damage evolution. To address this need, ongoing modeling and experimental efforts have focused on study of damage evolution in these materials. Micromechanical finite element simulations that resolve individual GMB and matrix components promote discovery and better understanding of the material behavior. With these simulations, we can study the role of the GMB volume fraction, time-dependent damage, behavior under confined vs. unconfined compression, and the effects of partial damage. These simulations are challenging and push the boundaries of capability even with the high performance computing tools available at Sandia. We summarize the major challenges and the current state of this modeling effort, as an exemplar of micromechanical modeling needs that can motivate advances in future computing efforts.
Previous numerical studies of Sylgard filled with glass microballoons (GMB) have relied on various microstructure idealizations to achieve a large range of volume fractions with high mesh quality. This study investigates how different microstructure idealizations and constraints affect the apparent homogenized elastic constants in the virgin state of the material, in which all GMBs are intact and perfectly bonded to the Sylgard matrix, and in the fully damaged state of the material in which all GMBs are destroyed. In the latter state, the material behaves as an elastomeric foam. Four microstructure idealizations are considered relating to how GMBs are packed into a representative volume element (RVE): (1) no boundary penetration nor GMB-GMB overlap, (2) GMB-GMB overlap, (3) boundary penetration, and (4) boundary penetration and GMB-GMB overlap. First order computational homogenization with kinematically uniform displacement boundary conditions (KUBCs) was employed to determine the homogenized (apparent) bulk and shear moduli for the four microstructure idealizations in the intact and fully broken GMB material states. It was found that boundary penetration has a significant effect on the shear modulus for microstructures with intact GMBs, but that neither boundary penetration nor GMB overlap have a significant effect on homogenized properties for microstructures with fully broken GMBs. The primary conclusion of the study is that future investigations into Sylgard/GMB micromechanics should either force GMBs to stay within the RVE fully and/or use periodic BCs (PBCs) to eliminate the boundary penetration issues. The implementation of PBCs requires the improvement of existing tools in Sandia’s Sierra/SM code.
This work was done to support customer questions about whether a Sylgard/Glass Microballoon (GMB) potting material in current use could be replaced with pure Sylgard and if this would significantly change stresses imparted to internal components under thermal cycling conditions. To address these questions, we provide micromechanics analysis of Sylgard/GMB materials using both analytic composite theory and finite element simulations to better understand the role of the GMB volume fraction in determining thermal expansion coefficient, elastic constants, and behavior in both confined and unconfined compression boundary value problems. A key finding is that damage accumulation in the material from breakage of GMBs significantly limits the global stress magnitude and results in a plateau stress behavior over large ranges of compressive strain. The magnitude of this plateau stress is reduced with higher volume fractions of GMBs. This effect is particularly pronounced in confined compression, which we estimate bears the most similarity to the application of interest. This stress-limiting damage mechanism is not present in pure Sylgard, however, and the result is much higher stresses under confined compression. Thus, we recommend that some volume fraction greater than 10% GMBs be used for confined deformation applications.
We present selected results from a series of Open Stack thermal battery tests performed in FY14 and FY15 and discuss our findings. These tests were meant to provide validation data for the comprehensive thermal battery simulation tools currently under development in Sierra/Aria under known conditions compared with as-manufactured batteries. We are able to satisfy this original objective in the present study for some test conditions. Measurements from each test include: nominal stack pressure (axial stress) vs. time in the cold state and during battery ignition, battery voltage vs. time against a prescribed current draw with periodic pulses, and images transverse to the battery axis from which cell displacements are computed. Six battery configurations were evaluated: 3, 5, and 10 cell stacks sandwiched between 4 layers of the materials used for axial thermal insulation, either Fiberfrax Board or MinK. In addition to the results from 3, 5, and 10 cell stacks with either in-line Fiberfrax Board or MinK insulation, a series of cell-free “control” tests were performed that show the inherent settling and stress relaxation based on the interaction between the insulation and heat pellets alone.
We calibrate a linear thermoviscoelastic model for solid Sylgard 184 (90-10 formulation), a lightly cross-linked, highly flexible isotropic elastomer for use both in Sierra / Solid Mechanics via the Universal Polymer Model as well as in Sierra / Structural Dynamics (Salinas) for use as an isotropic viscoelastic material. Material inputs for the calibration in both codes are provided. The frequency domain master curve of oscillatory shear was obtained from a report from Los Alamos National Laboratory (LANL). However, because the form of that data is different from the constitutive models in Sierra, we also present the mapping of the LANL data onto Sandia’s constitutive models. Finally, blind predictions of cyclic tension and compression out to moderate strains of 40 and 20% respectively are compared with Sandia’s legacy cure schedule material. Although the strain rate of the data is unknown, the linear thermoviscoelastic model accurately predicts the experiments out to moderate strains for the slower strain rates, which is consistent with the expectation that quasistatic test procedures were likely followed. This good agreement comes despite the different cure schedules between the Sandia and LANL data.