Using gradient boosting regressor to predict stress intensity factor of a crack propagating in small bore piping

Abstract

Estimating the value of the stress intensity factor (SIF) for a crack propagating in small bore piping is essential for predicting the remnant fatigue life (RFL) of the aforementioned asset. Currently, the finite element method (FEM) is utilized to predict the SIF. The main shortcomings of SIF prediction using the FEM are high computational cost and considerable time-consumption. In this manuscript, the authors propose using gradient boosting regressor (GBR) as an alternative to FEM for predicting the SIF of a propagating crack. The GBR is firstly trained and then validated by using 70 and 30 SIF values, respectively, obtained by FEM. During the validation process, the coefficient of correlation (R <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> ) between the SIF values obtained by FEM and by GBR is 0.977, indicating good agreement between the two. The time required to predict the SIF of 30 data points is reduced from 30 mins (for FEM) to one second with the help of the proposed GBR. Good prediction accuracy and less time-consumption of GBR make it a suitable alternative to FEM for SIF prediction.