AOI: Digital Engineering Tools
The key innovation of this is effort is development of a machine learning framework for creating digital engineering models of hypersonic vehicles with quantified prediction uncertainty. This capability will be provided in a form compatible with generic integrated digital environments (IDE) so that direct benefit to a broad range of digital engineering efforts within the DoD can be provided. The key capabilities added to digital engineering programs through this effort are:
- Implementation of the model validation hierarchy for a representative hypersonic system.
- Formulation of surrogate models for SRQs with quantified uncertainty within the hierarchy
- Affordable uncertainty propagation upward through the hierarchy to assess impacts at the system level.
- Integration with industry standard MBSE representations for broad applicability.
The capability will initially be applied to a representative advanced hypersonic vehicle to demonstrate the impact of the methodologies toward improving decision support for vehicle design.
The objective of this effort is to implement a machine learning framework for digital engineering of hypersonic vehicles with quantified prediction uncertainty. The framework will integrate model-based system engineering (MBSE) concepts; physics-based modeling; and machine learning within a software framework for advanced hypersonic vehicles. In combination, these capabilities will enable digital representation of hypersonic systems with quantified uncertainty metrics that can be provided to decision makers.
A digital transformation is revolutionizing the DoD’s and AFRL’s approach towards engineering design and acquisition. The digital engineering (DE) paradigm employs multi-physics models to understand the behavior of engineering systems during the full life-cycle of development and deployment. When successfully implemented, DE can significantly reduce costs and time to deployment for new systems. However, many DE initiatives do not make it to the point of deploying digital twins to supporting decision makers in the critical functions of designing and maintaining complex weapons systems. This shortfall is generally attributable to two underlying technical challenges in: 1) generating high-fidelity response predictions of a large-scale complex systems engineering model within the time of relevance, and 2) quantifying the uncertainty of SRQs propagated through multiple tiers of a system hierarchy model. If these two challenges can be met, then DE can improve decision making with reasonable computational costs. A critical need exists to address these challenges.