Modelling and Prediction of Fatigue Crack Growth in Part-Through Cracked Pipes under Bending Load

Abstract

Pipe installations experience fluctuating loading condition due to changes in fluid pressure and momentum and occasional high amplitude seismic vibrations. Design against fatigue in case of cylindrical structures like pipes is mostly done based on the models derived from standard and non-standard specimens fabricated from the pipe members. The results from these specimens were reported to be conservative due to the difference in constraint effect. Conventional design philosophy of leak-before-break can be disastrous for pipes carrying hazardous fluids. In such cases there is a need to replace the pipe before the crack penetrates the thickness.

In this work, fatigue crack growth in part-through cracked pipes under bending load has been studied. Three sets of part-through circumferentially notched pipe specimens with different notch angles have been tested. No crack length-compliance correlation is available for pipe specimens for in-situ crack length measurement. Therefore, crack opening displacement (COD) gage has been calibrated and used for calculation of crack growths. Power law equations with a constant rate parameter have been used earlier for modelling the crack growth rate in pipes. These models use integration schemes for finding crack growths and fatigue lives. Not much work has been done for the direct modelling of crack growths in pipes.

In the present work it has been observed that under four-point bend cyclic load, initially the crack propagates in the transverse plane in the radial direction followed by propagation in the circumferential direction. The crack propagation in the circumferential direction resulted in retardation of crack growth in radial direction.

The fatigue crack growth has been modeled using modified exponential model, Weibull function and gamma function. Artificial neural network based programming has also been applied for the prediction of crack growths. The crack growth parameters in the different models have been expressed as functions of crack driving forces, material properties and specimen geometry. The predicted results have been found in good agreement with the experimental values.