Pipeline pressure-cycle fatigue analysis is typically performed by analyzing pressure data in the amplitude domain and then calculating incremental fatigue crack growth in accordance with the Paris Law. Alternatively, the stochastic pressure history is converted to an equivalent number of uniform-amplitude cycles using a cumulative damage rule. The fatigue life may then be estimated by integration of the Paris Law. This second approach is computationally less involved and therefore lends itself to a probabilistic analysis because of the large number of iterations necessary with techniques such as Monte Carlo analysis.
However, studies have shown that for a broadband stochastic signal, applying linear cumulative damage can introduce large errors. The presence and magnitude of error cannot be easily determined by inspection of the pressure signal.
This paper describes the analysis of the pipeline pressure signal in the frequency domain to determine the power spectral density. The result can be used to estimate correction factors to the estimated linear cumulative damage fraction. The corrections may then be applied with a simplified integration of the Paris Law in closed form to improve both accuracy and speed for probabilistic assessment. The computation time for a probabilistic assessment may potentially be reduced by a significant factor.