Detecting Anomalies in Sensor Data using Neural Networks

There exists a need within the Department of Defense (DoD) to provide a highly responsive and costeffective anomaly detection capability for satellite health monitoring. The solution must address current Battle Management Command and Control (BMC2) deficiencies. A soft-computing solution is robust in that it could be engineered to detect anomalies without requiring pre-set thresholds and thus enhance the situational awareness of satellite and sensor assets monitored by operators in Space Command and Control (C2). This paper details our solution for detecting problems present in telemetry and sensor data by employing a soft-computing system that could continuously monitor and detect anomalies within a potentially large number of data streams.