Machine learning as an alternative to thresholding for space radiation high current event detection

The space environment is known to be the seat of radiation of different kinds to which satellites in orbit are subjected. These include cosmic rays that come from stars and radiation belts that come from the Earth’s magnetic field. The impact of radiation on electronic components results in anomalies called “Single Event Effects” which can lead to the destruction of equipment. Various protection methods exist, like hardening of components or satellite shielding, but they are often costly and/or difficult to implement. This is why space designers try to circumvent these processes with an efficient software protection method. This paper reports a set of experiments based on machine learning tools that will provide the basis to design and develop an anomaly detection method for Single Event Effects. The data sets that were used are issued from emulated radiations obtained by laser tests on a SAM3X microcontroller, complemented by data obtained by simulation.