Radio Frequency Fuel Gauging with Neuro-Fuzzy Inference Engine for Future Spacecrafts

A. Kumagai, T.-I. Liu, and D. Sul (USA)

Keywords

ANFIS, RF fuel gauging, subtractive clustering.

Abstract

Radio Frequency (RF) mass gauging is a technique for measuring an amount of fuel inside a cryogenic tank of a future spacecraft under a low gravity environment. In this paper, a method of predicting fuel amount from a given RF resonance tank frequency spectrum through a neuro fuzzy inference engine is described. Raw RF resonance spectrum data is first smoothed out by a moving average filter. Then a set of RF resonance modes obtained from this processed data is used as the input for predicting the output value which is the fuel amount. The subtractive clustering method is used for creating the initial membership functions and rules of the neuro-fuzzy engine. The parameters of the engine are optimized by ANFIS. Computer simulations for a wide range of RF resonance frequencies and various fuel levels indicate that the proposed neuro-fuzzy method works accurately and robustly for predicting fuel amount within 10% error based on the full tank capacity compared to pure mathematical means of interpolation such as the polynomial-fit with maximum error of 207% and the cubic spline with maximum error of 22%.

Important Links:

Go Back