Analysis of space shuttle main engine data using beacon-based exception analysis for multi-missions

This paper describes analysis of Space Shuttle Main Engine (SSME) sensor data using Beacon-based Exception Analysis for Multimissions (BEAM), a new technology developed for sensor analysis and diagnostics in autonomous space systems by the Jet Propulsion Laboratory (JPL). The BEAM anomaly detection system has been applied to SSME in a joint effort between JPL and the Marshall Space Flight Center (MSFC). MSFC is evaluating BEAM as an automated tool for rapid analysis of SSME ground-test data. BEAM is an end-to-end method of data analysis intended for real-time (on-board) or non-real-time anomaly detection and characterization. For the SSME application, a custom version of BEAM was built to analyze data gathered during ground tests. Since BEAM consists of modular components, a custom version call be tailored to address specific applications and needs by mixing-and-matching if components. The initial build of BEAM for the SSME focuses on signal processing and contains three , components: Coherence-based Fault Detector (CFD). Dynamical Invariant Anomaly Detector (DIAD), and Symbolic Data Model (SDM). This paper describes the software environment, its training steps, and the analysis results of the SSME data using the DIAD module. The DIAD module was the first applied due to its particular suitability to SSME data features. It detects anomalies by computing coefficients of an auto-regressive model (i.e. dynamical invariants) and comparing them to expected values as extracted from previous data. The DIAD is particularly sensitive to anomalies caused by faulty sensors, subtle and sudden performance shifts, and unexpected transients. This functionality will be illustrated by several examples from actual test data.