PROTECTION ENGINEERING CONSULTANTS LLC — Department of Defense SBIR Phase I: Protection Engineering Consultants (PEC) proposes to develop a coarse-grain neural network

Description

Protection Engineering Consultants (PEC) proposes to develop a coarse-grain neural network modeling approach for efficient simulation of shock propagation through geological media including soil-structure interaction. In the coarse-grain approach, fast-running computational elements are developed which are an order of magnitude larger than typical FE elements, and the elements are driven by neural network-based equations as opposed to physics-based equations. Typically FRMs cannot extrapolate to problems beyond the boundaries of the data set used to develop the models. Moreover, the size of the data set required increases geometrically with the number of independent variables needed as input variables to the FRMs. A more robust approach will help to increase the applicability of the developed FRM, while still maintaining the requirement for rapid execution of the problem. The coarse-grain neural network modeling approach is a novel approach to develop a tool that is versatile and still retains the FRM requirement of rapid execution.