Fatigue cracking is a key type of defect for liquid pipelines, and managing fatigue cracks has been a top priority and a big challenge for liquid pipeline operators. The existing inline inspection (ILI) tools for pipeline defect evaluation have large fatigue crack measurement uncertainties. Furthermore, the current physics-based methods are mainly used for fatigue crack growth prediction, where the same or a small range of fixed model parameters is used for all pipes. They result in uncertainty that is managed through the use of conservative safety factors such as adding depth uncertainty to the measured depth in deciding integrity management and risk mitigation strategies. In this study, an integrated approach is proposed for pipeline fatigue crack growth prediction utilizing ILI data including consideration of crack depth measurement uncertainty. This approach is done by integrating the physical models, including the stress analysis models, the crack growth model governed by the Paris’ law, and the ILI data. With the proposed integrated approach, the finite element (FE) model of a cracked pipe is built and the stress analysis is performed. ILI data are utilized to update the uncertain physical parameters for the individual pipe being considered so that a more accurate fatigue crack growth prediction can be achieved. Time-varying loading conditions are considered in the proposed integrated method by using rainflow counting method. The proposed integrated prognostics approach is compared with the existing physics-based method using examples based on simulated data. Field data provided by a Canadian pipeline operator are also employed for the validation of the proposed method. The examples and case studies in this paper demonstrate the limitations of the existing physics-based method, and the promise of the proposed method for achieving accurate fatigue crack growth prediction as continuous improvement of ILI technologies further reduces ILI measurement uncertainty.
Stochastic Filter-based Fatigue Crack Growth Prediction for Pipelines considering Unknown Model Parameters and Measurement Uncertainty
