Uncertainty Quantification for Mars 2020 Powered Descent Closed-Loop Stability

摘要

Three different methods were used to capture uncertainty in the Mars 2020 PDV modal properties, with the goal of understanding the uncertainty control system stability margins. The method of varying HCB frequencies and mode shapes independently proved to generate unrealistic results, possibly due to treating the frequencies and shapes as independent random variables. The method of varying FEM parameters provided much more reasonable results but had two primary draw backs. The first is that it artificially constrained the uncertainty based on the choice of parameters and did not account for uncertainty in model form. This resulted in a variation in mode shapes and corresponding transfer functions that is artificially small relative to observed results. The second drawback was in the number of possible parameterizations that could be selected. The sensitivity of the results to the choice of parameters makes it very difficult to confidently apply this method in practice. The method of nonparametrically varying the component HCB matrices proved to be a very simple approach that captured the variability in measured data very well. While this method is nonphysical in the sense that there is no set of physical parameters in the FEM associated with each member of the random distribution, it proved to do a better job of capturing variability in measured data than the other two methods. It is also much simpler to use because of the small number of parameters involved, and much easier to "tune" based on measured data. The conclusion of this study is that the NPV method was the best approach for this particular problem.