Reliability simulation of surface crack growth and analysis of life cycles

Based on dual boundary element method, Forman modified model and back propagation neural network, a reliability simulation method for surface crack growth is presented. In consideration of the uncertainty of eight parameters such as initial crack depth, crack length, crack angle, material parameters, crack growth threshold, fracture toughness and fatigue load, the orthogonal design table L₄₉ (7^s) is built to simulate crack growth sizes of crack front and obtain residual life. 7-12-1 three-layer BP neural network is trained and obtain nonlinear relationship between random variables and crack propagation life. Finally using Monte-Carlo sampling method, crack propagation reliability model and total life reliability model are given. Three-dimensional crack growth reliability of the torsion bar as an instance is simulated, and the results of crack propagation and life calculations show that the simulation method is reasonable and effective.