Anomaly detection in Electromechanical systems using Symbolic Dynamics

Major catastrophic failures in large scaleengineering systems (e.g. aircraft power plants andturbo-machinery) can possibly be averted if themalignant anomalies are detected at an early stage.This dissertation experimentally validates a novelmethod called Symbolic Time Series Analysis(STSA) foranomaly detection in electromechanical systemsderived from time series data of pertinent measuredvariable(s).In this dissertation the performance ofthis anomaly detection method is compared with thatof other existing pattern recognition techniques fromthe perspectives of early detection of fatigue damagein Al-2024. The experimental apparatus on which theanomaly detection method is tested is a multi-degreeof freedom mass-beam structure excited by oscillatorymotion of two electromagnetic shakers. The evolutionof fatigue crack damage at one of the failure sitesis detected from STSA of the pertinent sensor signal.Industrial Application-The dissertation presents STSAof bearing acceleration derived from a dynamicsimulation model for detection and estimation ofparametric changes in flexible disc/diaphragmcouplings due to angular misalignment between shafts.