In computational mechanics, a probabilistic/Bayesian approach means that the parameters and variables of numerical simulations are “dressed” with posterior distributions conditioned on available real-world observation. The Bayesian hierarchy can be extended to integrate and link predictive models which describe the mechanics at distinctive scales and physics.
The framework should serve as a coupling/model selection tool for all computational models relevant to granular matter, and provide uncertainty propagation capabilities to understand model uncertainty affects the industrial application. Ideally, when launched on a open-access Cloud platform, the framework can utilize a hybrid data-driven/physical-based modeling approach, with a sufficiently large database contributed by users. We are currently developing a prototype of such a framework, starting from an iterative algorithm that efficiently sample parameter/solution space in a Bayesian setting.
The framework should serve as a coupling/model selection tool for all computational models relevant to granular matter, and provide uncertainty propagation capabilities to understand model uncertainty affects the industrial application. Ideally, when launched on a open-access Cloud platform, the framework can utilize a hybrid data-driven/physical-based modeling approach, with a sufficiently large database contributed by users. We are currently developing a prototype of such a framework, starting from an iterative algorithm that efficiently sample parameter/solution space in a Bayesian setting.