Antonino Galletta

Summary Large low Earth orbit satellite constellations require machine learning methods for enabling autonomy in health keeping of the satellites. Autonomy in health keeping entail's fault detection, isolation and reconfiguration. However, prior to model building, it becomes imperative to conduct exploratory data analysis of the data to gain an intuition of data and to decide the best model. Univariate exploratory data analysis has been carried out on a BUS CURRENT sensor of electrical power system of a low Earth orbit satellite to gain an understanding of data. Various aspects of data like presence of outliers, sampling frequency, missing values, comparison of imputation methods to fill missing values seasonality and trend analysis, stationarity test on data, rolling mean and autocorrelation and partial auto correlation plots have been made, and a detailed statistical analysis of results has been conducted.