AI-Based Spacecraft Health Monitoring System: A Machine Learning Approach for Telemetry-Based Health Classification

Abstract

Spacecraft continuously generate telemetry data from onboard sensors and subsystems to support monitoring of health, performance, and operational status. Manual observation of large volumes of telemetry can be difficult and may delay recognition of abnormal conditions. This paper presents an AI-Based Spacecraft Health Monitoring System developed during a research internship at Acceleron Aerospace. The proposed prototype combines ESP8266-based sensor acquisition, synthetic telemetry generation, Python-based data processing, a Random Forest machine learning classifier, and an interactive Streamlit dashboard. The dataset contains temperature, humidity, light, three-axis acceleration, and three-axis gyroscope measurements, with spacecraft health classified into HEALTHY, WARNING, and CRITICAL states. A balanced synthetic dataset of 3,000 samples was generated and divided into training and testing subsets. The trained model was integrated into the monitoring application to provide rapid health classification and graphical telemetry visualization. The work demonstrates a modular foundation for AI-assisted spacecraft monitoring and can be extended toward real telemetry, anomaly detection, fault diagnosis, predictive maintenance, and intelligent alerting.