This paper presents a novel method of pattern identification in complex systems using the tools derived from statistical thermodynamics. Complexity issues arise in natural or human-engineered systems due to behavioral uncertainties and nonlinearities involved in the process dynamics. The paper introduces a novel concept of behavioral pattern identification and anomaly detection in mechanical systems from macroscopically observed time series of the available sensor data. The theme is built upon the principles of Statistical Thermodynamics and Information Theory. The efficacy of this method is experimentally validated on a laboratory apparatus where the behavioral changes accrue from the evolving fatigue damage in polycrystalline alloy structures.

This content is only available via PDF.
You do not currently have access to this content.