Using Machine Learning to Forecast Time Series in Spacecrafts

Spacecrafts are intelligent and highly sensitive systems that are constantly exposed to environmental impact. Complex systems like those found in spacecrafts are designed to reduce faults and fatalities utilizing machine learning - more specifically known as Telemetry Mining. This analysis predicts satellite behavior which allows spacecrafts to correct their course or take other measures to mitigate potential negative impact. This paper aims to implement the statistical Autoregressive Integrated Moving Average (ARIMA) algorithm, used to forecast time series to predict spacecraft failure with the aim of saving investment and lives. To identify potential failure, results from predictions are evaluated through mean, standard deviation, covariance and Pearson's correlation square using the ARIMA algorithm. The data received from the Egyptsat-1 satellite's battery temperature is used as the parameter for input into Matlab This paper summarizes the performance and health of the spacecraft with the result of the implementation of the ARIMA algorithm on the basis of input parameters. Finally, the output from this algorithm is compared. This paper proposes a new framework for Machine Learning to Forecast Time Series in Spacecrafts. Prediction accuracy helps to decrease the failure rate of and improves the performance of the spacecraft.