Quantification of Dynamic Model Validation Metrics Using Uncertainty Propagation from Requirements

Abstract

The Space Launch System, NASA's new large launch vehicle for long range space exploration, is presently in the final design and construction phases, with the first launch scheduled for 2019. A dynamic model of the system has been created and is critical for calculation of interface loads and natural frequencies and mode shapes for guidance, navigation, and control (GNC). Because of the program and schedule constraints, a single modal test of the SLS will be performed while bolted down to the Mobile Launch Pad just before the first launch. A Monte Carlo and optimization scheme will be performed to create thousands of possible models based on given dispersions in model properties and to determine which model best fits the natural frequencies and mode shapes from modal test. However, the question still remains as to whether this model is acceptable for the loads and GNC requirements. An uncertainty propagation and quantification (UP and UQ) technique to develop a quantitative set of validation metrics that is based on the flight requirements has therefore been developed and is discussed in this paper. There has been considerable research on UQ and UP and validation in the literature, but very little on propagating the uncertainties from requirements, so most validation metrics are "rules-of-thumb;" this research seeks to come up with more reason-based metrics. One of the main assumptions used to achieve this task is that the uncertainty in the modeling of the fixed boundary condition is accurate, so therefore that same uncertainty can be used in propagating the fixedtest configuration to the free-free actual configuration. The second main technique applied here is the usage of the limit-state formulation to quantify the final probabilistic parameters and to compare them with the requirements. These techniques are explored with a simple lumped spring-mass system and a simplified SLS model. When completed, it is anticipated that this requirements-based validation metric will provide a quantified confidence and probability of success for the final SLS dynamics model, which will be critical for a successful launch program, and can be applied in the many other industries where an accurate dynamic model is required.

Uncertainty Quantification of Inducer Eigenvalues Using Conditional Assessment of Models and Modal Test of Simpler Systems

“…As the focus of this study was only to quantify the uncertainty of the final model of the full-scale inducer rather than update the model, which had been previously performed, it was deemed excessively complex to apply the techniques shown in the literature, which focus on improving the primitive random variable posterior distributions and generating nondeterministic models that would then be propagated. The initial concept was generated based on studies by the lead author identifying the uncertainty in the new NASA Space Launch System Flight vehicle primary mode based on a ground modal testing of the vehicle [9]. In that study, a quartile linear regression technique was used to obtain the flight mode purely as a regression on a single ground mode, as opposed to this case, where dependence on a number of modes from different configurations was sought.…”