Extraction of Pendulum Model Parameters from Steady-State Slosh Data in Diaphragm Tanks

Abstract

This study proposes a new experimental approach to extract pendulum model parameters from spacecraft propellant tank slosh data. The novelty of the approach is in the utilization of steady-state triaxial force measurements at the support points of the tank under sinusoidal excitation. This approach enables repeatable and accurate determination of pendulum model parameters in contrast to the single-axis force sensing and multiple repetitions averaging techniques associated with the more common decay approach. This approach may be used for tanks with free slosh; however, this study considers a diaphragm tank with three different diaphragm configurations under translational oscillatory lateral excitation. With the aid of a parameter optimization technique prevalently used in machine learning, the proposed methodology identifies the pendulum model parameters based on an adaptive gradient descent on the error surface defined in the parameter space. The proposed approach is validated by comparing experimental measurements against model-predicted behavior, and the numerical procedure is verified against a quasi-Newton method. Moreover, the extracted pendulum model parameters enable prediction of the frequency response functions for both the apparent mass and apparent mass moment as function of excitation frequency, and the methodology enables extraction of higher-order vibration modes.