Application of Bayesian Estimation to Structural Health Monitoring of Fatigue Cracks in Welded Steel Pipe

Abstract

Vibration induced fatigue is a well-known problem in oil and gas piping systems [1]. Various vibration-based monitoring techniques have been implemented in the field of condition (CM) and structural health monitoring (SHM). However, the major challenge in monitoring technology is to detect failure with high confidence. In principle, vibration-based techniques evaluate structural condition from physical and dynamic characteristics. This method is useful particularly for damage detection and localization. However, the information obtained may be insufficient for complex problems of detection and localization of cracks, where the scale of damage is relatively small compared to the overall size of piping system. The difficulty arises due to negligible changes in structural stiffness, and thus no observable change in natural frequencies. Unmeasurable local stresses at crack tips further increase the risk of fatigue failure. In this paper, acoustic emission (AE) is used for damage detection and localization in a welded pipe subjected to dynamic loading by resonance fatigue testing. Acoustic emission is a passive non-destructive testing (NDT) technique due to rapid changes of stress state such as during crack extension. The energy is released as elastic waves that travel within material in the form of microscopic displacements and are then converted into electrical signals by AE sensors. The acoustic emission method is very sensitive to crack evolution. However, the signals are also susceptible to extraneous noise. Therefore Bayesian estimation, which uses a probabilistic approach to estimate the unknown parameters for damage evaluation, is proposed. The estimation was carried out using a signal-based method where the parameter used for Bayesian estimation is derived directly from the signals.