Stochastic Modelling and Prediction of Fatigue Crack Propagation Based on Experimental Research

3rd World Multidisciplinary Civil Engineering, Architecture, Urban Planning Symposium

土木建筑工程;文学

Omishore, Abayomi^1

Brno University of Technology, Faculty of Civil Engineering, Veveí 331/95, Brno

602 00, Czech Republic^1

关键词: Correlation measurement;    Latin hypercube sampling methods;    Linear fracture mechanics;    Modelling and predictions;    Pearson correlation coefficients;    Quantitative sensitivity;    Rank correlation coefficient;    Spearman rank correlation;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/471/10/102037/pdf
DOI  :  10.1088/1757-899X/471/10/102037

学科分类：土木及结构工程学

来源: IOP

【 摘 要 】

Numerical analysis of the propagation of an edge crack in a plate was performed in this study. The theoretical model of fatigue crack progression is based on linear fracture mechanics. Calibration functions for short edge cracks are applied in stochastic models and the stochastic dependencies between input random variables and the fatigue resistance are described. Attention is focused on the domain of the relative crack length. Results are obtained using the Latin Hypercube Sampling method. Sensitivity analysis is evaluated using methods ranging from the screening method to quantitative techniques based on correlation measurements. Pearson correlation coefficient, Spearman rank-correlation coefficient and Kendall rank correlation coefficient are used for the evaluation of sensitivity analysis. The study demonstrates the application of several numerical simulation procedures covering both qualitative and quantitative sensitivity analysis using a one-sample base. The effects of non-linear stochastic dependencies and outliers on the results of sensitivity analysis are discussed.