Intelligent Black Box Verification, Validation, and Accreditation for Rotorcraft Performance Modeling

The paper addresses process and tool development in support of qualification assessments of performance models. Details are provided regarding developing enablers in the fields of data science, uncertainty quantification, and machine learning. Models are delivered for assessment in all different shapes, for different purposes, with different pedigrees. All or part of the model may be proprietary. In an ideal case, models arrive with detailed documentation, with known data pedigrees, and have successfully passed verification testing. This would be in compliance with the standards outlined in Army Regulation 5-11, Management of Army Models and Simulations (Ref. 1). With other submissions much less is known; documentation and pedigree are absent. It may be missing documentation of requirements or detailed specifications. Traditionally these gaps in information are filled by time-consuming, resource-intensive work of subject matter experts (SMEs), if possible to be filled at all. As modeling capability and complexity continues to increase, the gaps naturally increase, to a point comprehensive assessment becomes unwieldy and unaffordable (even impossible) to be conducted with available processes and tools. Efforts contrast new applications of high-powered computational tools with the long accepted SME-driven methods to establish value and increase capabilities. Results indicate that emerging methodologies can provide valuable guidance without SME involvement.