Improved Digital Engineering Techniques for Test Data Leveraging

OUSD (R&E) CRITICAL TECHNOLOGY AREA(S): Advanced Computing and Software OBJECTIVE: Develop and demonstrate a data analysis tool methodology to capture relevant test parameters from past data into a usable form and create a digital engineering-level model capable of evaluating variables and providing accurate predictive answers. Models must be able to quickly learn from new data to improve predictions DESCRIPTION: Aircraft survivability and ballistic test data have been collected for over sixty years. This data needs to be converted to a format to aid in performing data analysis and for extracting answers to assist in new test and evaluation predictions. With current test reports, which are often stored in Portable Document Format (PDF) for later reference, little regard is given to the usefulness of the data beyond the current project and often insufficient background information is documented along with the data to allow useful models to be created and tailored to new applications. In addition, new test data is often difficult to merge with this previous data without a detailed link to caveats associated with the previous testing. This may include future testing which may make use of enhanced instrumentation techniques able to capture higher resolution or details than data captured in previous testing. The lack of clarity on the past testing and the diligence required for finding and extracting the past test report data often results in duplication of effort and/or re-learning of lessons learned. This, in turn, leads to additional cost and time to get vital answers. Even when engineering level models were built upon the past test data, the addition of new test data most often results in the creation of new Monte Carlo inputs that may not properly capture parameters and constraints. The digital engineering paradigm requires more coupon, subscale, and large-scale testing to be performed earlier to support design and trade study activities before prototypes are built. Much of this test data will need to come from previous test programs and be adapted to new aircraft survivability test objectives. With only static reports available from the past and perhaps no traceability to raw data or no accurate way to interpret that data, there is often no ability to capture all the necessary information from past testing to create useful test data products for the future. In addition, there is certainly no way to improve upon the past data, as additional related test or simulation data become available. This means that digital engineering will be difficult to carry out in practice. This SBIR Phase I effort will focus on demonstrating a data encapsulation methodology to apply to coupon, sub-scale, and large-scale survivability test data relating to threat/target interactions. Target-related test data may include stress/strain, pressure, temperature, damaged structure, cracking, hydrodynamic ram effects, fuel spurt and more. Threat-related data may include flash/function probability, residual velocity, residual mass, and more. The data encapsulation methodology should include a demonstration of how various forms of data can be preserved, relevant variables are preserved, accurate predictions can be made, how the resulting tool can improve with additional test data, and how previously generated test data can be used for new and novel test programs. PHASE I: Significant work equivalent to a Phase I effort in demonstrating the feasibility of a data encapsulation methodology applicable to aircraft survivability test data must be documented in the proposal. The methodology must be able to integrate previous survivability test and analysis data along with data from current and future testing and analyses into a comprehensive analytical tool. This analytical tool will aid survivability engineers in developing future survivability test programs and analyses that will produce more reliable test and analysis results PHASE II: Development of a data encapsulation methodology should be completed and demonstrated in the form of an engineering-level model for a set of ballistic test data related to threat/target interaction. Demonstration should be conducted for a new test data being incorporated with previous test data. PHASE III DUAL USE APPLICATIONS: Digital engineering is not limited to aircraft survivability or military development efforts. With custom materials and new technologies commonly being incorporated into new designs, building efficiently on older test data is essential to both military and commercial applications. REFERENCES: D. Varas, R. Zaera, J. L pez-Puente, Numerical modelling of the hydrodynamic ram phenomenon, International Journal of Impact Engineering, Volume 36, Issue 3, 2009, Pages 363-374; Peter J. Disimile, Norman Toy, Liquid spurt caused by hydrodynamic ram, International Journal of Impact Engineering, Volume 75, 2015, Pages 65-74; KEYWORDS: test data; survivability; machine learning