If a material and manufacturing process is understood well enough, the performance of a manufactured part can be accurately predicted.

Lawrence Livermore has focused particular attention since the early 1990s on the development of predictive models for regimes where experimental data is not always available. Mature integrated modeling capabilities have been developed at a variety of time and length scales, ranging from atomic structure, through molecular and crystal mechanics, up to the continuum. These capabilities rely on fundamental theoretical models, guided and corroborated by key experimental capabilities, and sustained by significant amounts of high performance computing.

One of the challenges presented by multiscale modeling is the vast amount of data created by simulations. Knowing what data to pass from one scale to the next is crucial. Instead of trying to find a needle in a haystack, we focus on the needle from the start by defining how, and under what conditions, a part fails. Achieving this macroscale fail mode allows us to pinpoint simulations at more granular scales to get data that we can use. Refining this approach has proven to be a huge step in bringing the data gap between molecular dynamics and crystal dynamics.

Predicting performance over time

The ability to predict component performance over a lifetime, or improve and refine the manufacturing process relies on understanding fundamental physical mechanisms governing material behavior and robust modeling capabilities that represent all relevant phenomena. Such modeling is extremely computationally intensive and relies heavily on high performance computing resources.

Science-based predictive models can enable the prediction of material behavior in regimes where properties are difficult or impossible to measure directly (e.g., high strain rate behavior of a shaped charge jet). They also enable process aware modeling that accounts for the changes in properties of materials over time. Such capabilities integrated into an overall design process can lead to faster development of prototypes, and products with greater confidence and better lifetime performance.

Science-based predictive models can enable improved understand of manufacturing processes, thus making it possible to improve and optimize them. Detailed understanding of material changes during processing (microstructural changes, phase transformations) enable optimization of manufacturing processes, increasing product yield, reducing waste, improving the quality of components, and leading better understanding of the lifetime performance of components.

With greater representation of material behavior, modeling and simulation can be used to design components with less uncertainty (closer to the margins). With better understanding of failure mechanisms they can also be used to more quickly arrive at reliable component designs that meet design requirements. This approach is built upon the technical foundation that has already been established for simulation of manufacturing processes, e.g., Lawrence Livermore tools and models have been developed and used to capture the forming, rolling, and casting of complex shaped parts in collaboration with Alcoa. (Karabin et al.: 2003)