Performance Prediction of Antenna Control Servo System based on LSTM Network

Unmanned ground stations have become popular for satellite tracking, but lack of human presence causes difficulties to address equipment failures. Predictive maintenance through equipment performance monitoring observes performance trends to detect equipment degradation and failures beforehand. This study applies performance monitoring to Antenna Control Servo System (ACSS), wherein current drawn by servo elevation motor during satellite passes is analyzed. Using a sliding window technique to process sequential data from servo logs, a stacked Long Short Term Memory (LSTM) network predicts elevation current drawn for future satellite passes. The developed model produces promising performance, achieving a mean absolute error of 0.06.