Bayesian Estimation for Crack Monitoring of a Resonant Pipe using Acoustic Emission Method

Abstract

Vibration induced fatigue is a well-known problem in the oil and gas piping and pipeline systems. However the use of vibration data to detect damage is not an easy task without a priori knowledge about the undamaged condition. In this paper, acoustic emission monitoring of a resonance fatigue test of a welded carbon steel pipe has been carried out to monitor the weld condition from healthy until failure. The information provided by acoustic emission monitoring is useful in evaluating the condition of the pipe during the test and the occurrence of cracking before failure. However, acoustic emission signals are also susceptible to extraneous noises. Therefore in this work, the acoustic emission signals from different combination of sensors were recursively correlated, which provides parameter for Bayesian estimation to discriminate the true signals and avoid false detection and localization errors. The method provides indicator for damage detection, which shows a strong relationship of the acoustic emission energy and the estimated coefficients. High correlation of signals was found to be associated with cracking, which has an effect of the increase of acoustic emission energy. Likewise low correlation of signals was found to be associated with random signals or noises. The method will be useful in monitoring piping or pipelines condition to reduce the risk of vibration induced fatigue failure.