Predictive Simulation Sciences

Predictive Simulation Sciences concentrates on understanding and exploiting the fundamental aspects of verified and validated computational simulations that predict the response of complex Army systems and guide Army materiel design, particularly in cases where routine experimental tests are not feasible.

Computational Multiscale Methods

The computational cost of complex system M&S necessitates novel approaches. Multiscale methods provide needed fidelity while addressing cost constraints. By coupling across temporal and spatial scales it is possible to retain the physics of the building blocks of complex systems while accurately predicting full scale performance. There are many challenges to overcome towards this goal, including the need for new methodologies for scale bridging (spatial and temporal); new algorithms for automatic construction of surrogate models; uncertainty quantification between at-scale models crucial; and shorten development time and evaluation costs of novel Army systems.

Multiscale methods are applicable to many Army systems, materials, processes, etc. This list includes dislocation dynamics in micro-structured crystals, scale bridging methods for materials and modeling of transport in optical semiconductors, meso- and micro-scale forecast model, and the atmospheric boundary layer environment (ABLE) model to name a few.

Principal Investigator:

Dr. Jaroslaw Knap,, (410) 278-0420

Uncertainty Quantification

The Army has witnessed a remarkable rise in the complexity of the battlefield. In response, Army systems are rapidly becoming more complex. This added complexity comes at a time when the Army must function under tighter time and resource constraints. In order to account for the complexity, and still be able to satisfy the above constraints, robustness in design, analysis, and decision making becomes absolutely crucial. Since modeling has now become a foundation of design, analysis and decision making, it is critical that robustness and the uncertainties introduced through real-world variability are incorporated into models.

ARL is seeking collaboration opportunities in novel and efficient concepts and stochastic methodologies for high-fidelity assessment on the level of agreement in sets of models relative to input and output data, variations in interdependent models due to various physics, mathematical, and numerical assumptions (enabling. tools to (i) identify deficiencies in simulations; (ii) set guidelines for adequacy of computational results; (iii) explore the impact of known variability and uncertainty of input; and (iv) control of adaptive algorithms to achieve specified levels of accuracy to aid decisions from design to operational planning).

Principal Investigator:

Dr. Ernest Chin,, (410) 306-1988