PROBABILISTIC MODEL DEVELOPMENT FOR FATIGUE CRACK DETECTION USING ACOUSTIC EMISSION TECHNOLOGY

摘要

One of the major concerns in structures such as piping is early detection of a growing crack to prevent fracture, predict remaining useful life, schedule maintenance and reduce costly downtimes. This paper focuses on in-situ monitoring of structural health, specifically detection of small crack growth and crack initiation in structures using Acoustic Emission (AE) technology. This paper explores potential AE signal properties used in identifying the presence of an initial crack considered as the onset of a potential growing crack. Experimental investigation of uniform cyclic load tests performed on standard fatigue samples of aluminum alloys indicated that crack initiation can be identified through the multivariate statistical analysis of AE data. A probabilistic model was developed based on the observed correlation between AE event and crack initiation and small crack growth behavior during high cycle fatigue tests. Bayesian regression analysis was used for stochastic modeling of the relationship between crack growth as the dependent variable and AE signal feature as the independent variable. The concept can equally be extended and applied to stainless steel materials used in piping and pressure vessels. This paper will explain this AE monitoring technique, results from experimental tests of crack growth using AE, and probabilistic AE-based fatigue crack initiation and growth model development.