Fatigue crack growth analysis using Bootstrap S-version finite element model

• Published in: Journal of the Brazilian Society of Mechanical Sciences and Engineering Vol. 42; no. 4
• Main Authors: Husnain, M. N. M., Akramin, M. R. M., Chuan, Z. L., Takahashi, Akiyuki
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 2020; Springer Nature B.V
• Subjects: Accidents; Catastrophic failure analysis; Crack propagation; Engineering; Fatigue failure; Fatigue life; Finite element method; Fracture mechanics; Inspection; Life assessment; Life prediction; Lower bounds; Mathematical models; Mechanical Engineering; Normal distribution; Parameter uncertainty; Probabilistic analysis; Resampling; Schedules; Size distribution; Statistical analysis; Surface cracks; Technical Paper; Bootstrap resampling method; Fatigue; Probabilistic analysis; S-version finite element model; Crack growth
• ISSN: 1678-5878; 1806-3691
• DOI: 10.1007/s40430-020-2268-8

The objective of this paper was to predict the fatigue life, surface crack and initial flaw size distribution of fatigue surface crack growth. A statistical analysis is carried out to evaluate the distributions of initial flaw size as an input uncertain parameter. The prediction of remaining life to schedule the maintenance is important to prevent serious accidents from occurring. The three-point and four-point bendings are analysed using the Bootstrap S-version finite element model (BootstrapS-FEM) whereby the bootstrap resampling method is embedded into S-version finite element model. The validation process is conducted between the predictions, deterministic and previous experimental results. The BootstrapS-FEM with lognormal distribution shows a more accurate trend against normal distribution based on the coefficient of determination, normalised root-mean-square error and mean absolute percentage error. The prediction of fatigue life for three-point and four-point bendings by BootstrapS-FEM was well compared with previous experimental results within range from 5 to 17% of percentage errors. These errors were acceptable to the purpose of prediction which are less than 20%. The uncertainties in structural components are considered by the upper and lower bounds in probabilistic analysis. The BootstrapS-FEM results show a better agreement with previous experimental results and deterministic solutions. Initial flaw size distributions are evaluated to prior scheduled inspection for the prevention of a catastrophic failure. The risk of a catastrophic failure can be evaluated based on the initial crack size distribution and specified fatigue life for ensuring safety and reliability of the fatigue crack structures.