Advancements in our capabilities to accurately model physical systems using high resolution finite element models have led to increasing use of models for prediction of physical system responses. Yet models are typically not used without first demonstrating their accuracy or, at least, adequacy. In high consequence applications where model predictions are used to make decisions or control operations involving human life or critical systems, a movement toward accreditation of mathematical model predictions via validation is taking hold. Model validation is the activity wherein the predictions of mathematical models are demonstrated to be accurate or adequate for use within a particular regime. Though many types of predictions can be made with mathematical models, not all predictions have the same impact on the usefulness of a model. For example, predictions where the response of a system is greatest may be most critical to the adequacy of a model. Therefore, a model that makes accurate predictions in some environments and poor predictions in other environments may be perfectly adequate for certain uses. The current investigation develops a general technique for validating mathematical models where the measures of response are weighted in some logical manner. A combined experimental and numerical example that demonstrates the validation of a system using both weighted and non-weighted response measures is presented.
Accurate material models are fundamental to predictive structural finite element models. Because potting foams are routinely used to mitigate shock and vibration of encapsulated components in electro/mechanical systems, accurate material models of foams are needed. A linear-viscoelastic foam constitutive model has been developed to represent the foam's stiffness and damping throughout an application space defined by temperature, strain rate or frequency and strain level. Validation of this linear-viscoelastic model, which is integrated into the Salinas structural dynamics code, is being achieved by modeling and testing a series of structural geometries of increasing complexity that have been designed to ensure sensitivity to material parameters. Both experimental and analytical uncertainties are being quantified to ensure the fair assessment of model validity. Quantitative model validation metrics are being developed to provide a means of comparison for analytical model predictions to observations made in the experiments. This paper is one of several recent papers documenting the validation process for simple to complex structures with foam encapsulated components. This paper specifically focuses on model validation over a wide temperature range and using a simple dumbbell structure for modal testing and simulation. Material variations of density and modulus have been included. A double blind validation process is described that brings together test data with model predictions.
Accurate material models are fundamental to predictive structural finite element models. Because potting foams are routinely used to mitigate shock and vibration of encapsulated components in electro/mechanical systems, accurate material models of foams are needed. A linear-viscoelastic foam constitutive model has been developed to represent the foam's stiffness and damping throughout an application space defined by temperature, strain rate or frequency and strain level. Validation of this linear-viscoelastic model, which is integrated into the Salinas structural dynamics code, is being achieved by modeling and testing a series of structural geometries of increasing complexity that have been designed to ensure sensitivity to material parameters. Both experimental and analytical uncertainties are being quantified to ensure the fair assessment of model validity. Quantitative model validation metrics are being developed to provide a means of comparison for analytical model predictions to observations made in the experiments. This paper is one of several recent papers documenting the validation process for simple to complex structures with foam encapsulated components. This paper specifically focuses on model validation over a wide temperature range and using a simple dumbbell structure for modal testing and simulation. Material variations of density and modulus have been included. A double blind validation process is described that brings together test data with model predictions.
Accurate material models are fundamental to predictive structural finite element models. Because potting foams are routinely used to mitigate shock and vibration of encapsulated components in mechanical systems, accurate material models of foams are needed. A linear-viscoelastic foam constitutive model has been developed to represent the foam's stiffness and damping throughout an application space defined by temperature, strain rate or frequency and strain level. Validation of this linear-viscoelastic model, which is integrated into the Saunas structural dynamics code, is achieved by modeling and testing a series of structural geometries of increasing complexity that have been designed to ensure sensitivity to material parameters. Both experimental and analytical uncertainties are being quantified to ensure the fair assessment of model validity. Quantitative model validation metrics are being developed to provide a means of comparison for analytical model predictions to observations made in the experiments. This paper is one of several parallel papers documenting the validation process for simple to complex structures with foam encapsulated components. This paper will describe the development of a linear-viscoelastic constitutive model for EF-AR20 epoxy foam with density, modulus, and damping uncertainties and apply the model to the simplest of the series of foam/component structural geometries for the calibration and validation of the constitutive model.
Battery systems have traditionally relied on extensive build and test procedures for product realization. Analytical models have been developed to diminish this reliance, but have only been partially successful in consistently predicting the performance of battery systems. The complex set of interacting physical and chemical processes within battery systems has made the development of analytical models a significant challenge. Advanced simulation tools are needed to more accurately model battery systems which will reduce the time and cost required for product realization. Sandia has initiated an advanced model-based design strategy to battery systems, beginning with the performance of lithiumhhionyl chloride cells. As an alternative approach, we have begun development of cell performance modeling using non-phenomenological models for battery systems based on artificial neural networks (ANNs). ANNs are inductive models for simulating input/output mappings with certain advantages over phenomenological models, particularly for complex systems. Among these advantages is the ability to avoid making measurements of hard to determine physical parameters or having to understand cell processes sufficiently to write mathematical functions describing their behavior. For example, ANN models are also being studied for simulating complex physical processes within the Li/SOC12 cell, such as the time and temperature dependence of the anode interracial resistance. ANNs have been shown to provide a very robust and computationally efficient simulation tool for predicting voltage and capacity output for Li/SOC12 cells under a variety of operating conditions. The ANN modeling approach should be applicable to a wide variety of battery chemistries, including rechargeable systems.
Although they appear deceptively simple, batteries embody a complex set of interacting physical and chemical processes. While the discrete engineering characteristics of a battery, such as the physical dimensions of the individual components, are relatively straightforward to define explicitly, their myriad chemical and physical processes, including interactions, are much more difficult to accurately represent. For this reason, development of analytical models that can consistently predict the performance of a battery has only been partially successful, even though significant resources have been applied to this problem. As an alternative approach, we have begun development of non-phenomenological models for battery systems based on artificial neural networks. This paper describes initial feasibility studies as well as current models and makes comparisons between predicted and actual performance.