Disturbances compensation in high accuracy spaceborne accelerometers using multi-sensors and machine learning approach.

Abstract

Italian Spring Accelerometer (ISA) is a scientific payload of the European Space Agency’s BepiColombo mission to Mercury. It aims to measure the Non Gravitational Perturbations acting on the MPO (Mercury Planetary Orbiter) spacecraft, allowing to consider it as a test-mass free falling in the planetary gravity field and hence disclosing the possibility to study the  Mercury's interior, surface, and environment, as well as to preform tests of Einstein's General Relativity theory.ISA sensitivity to thermoelastic deformations of the spacecraft panel on which it is mounted on, is one of the limiting factors of the achievable acceleration measurements accuracy, whose target value is 10-8 m/s2 .To address this challenge, a data analysis and reduction procedure is being developed; it is based on machine learning techniques and allows to compute an acceleration measurements correction signal, starting from the data provided by multiple supplementary sensors. Specifically, we employed the temperatures recorded by several thermometers and the information about power dissipated across the MPO in order to compute the correction signal to be applied to the ISA output. Indeed, these temperatures and dissipated power variations are responsible for the thermoelastic deformations of the mounting plate housing ISA.The technique is being developed during the mission's cruise towards Mercury, exploiting also the outcomes of the GAIN “Gravimetro Aereo INtelligente” project, which developed a similar methodology for airborne gravimetry.The preliminary results related to measurement sessions during the cruise phase will be presented, and considerations on the implementation of such techniques for future space missions will be provided.Indeed, despite ISA was not specifically designed for the use of the "GAIN method”, the preliminary results are promising, underscoring its potential and allowing to envisage that future space missions could benefit of a full implementation of such a method that should go through the development of purpose built and trained multi-sensor systems.