Analysis of stop-hole effects on mode I-II fatigue crack behavior for Q420 steel using experiments, FEM and variable length RNN approaches

• Published in: Theoretical and applied fracture mechanics Vol. 124; p. 103823
• Main Authors: Gong, Hao, Jin, Zeng-gui, Yang, Feng-peng, Mao, Wen-tao
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.04.2023
• Subjects: Artificial neural network; Fatigue life; Numerical simulation; Stop-hole, Fatigue crack path; Fatigue life; Numerical simulation; Artificial neural network; Stop-hole, Fatigue crack path
• ISSN: 0167-8442; 1872-7638
• DOI: 10.1016/j.tafmec.2023.103823

•A series of experiments were conducted to investigate the effects of stop-hole repair technique. The compact tension-shear specimens with different stop-hole sizes were employed under various loading angles. The results show that the fatigue life increases with the loading angle, and also increases with the hole diameter.•The numerical simulation which consists of the initial damage and the following FCG was performed using FEM. A databased for the neural networks training and prediction was built by 637 experimental conditions and 17,836 sets of data. The database shows that the better range of stop-hole diameter is from 4 mm to 6 mm.•A variable-length inputs recurrent neural network (VRNN) model was developed which considers the association of the current crack sections and the former section during the networks learning. The predicted crack path and the corresponding life is since the R squares are larger than 0.95. The VRNN turns out to be more accuracy and efficient than the ordinary BPNN model, since the MSEs of both training set and prediction sets are lower than BPNN.•This research provides an inspiration to predict fatigue life and crack path by using VRNN. The corresponding results can be used to estimate the residual life and crack path, as well as evaluate the safety of equipment after using the stop-hole repaired technique. To investigate the stop-hole effect on the fatigue crack path and life under mixed I-II cyclic loads, the compact tension-shear specimens with different stop-hole diameters were adopted to perform a series of tests under various loading angles. The results show that the fatigue life increases with the loading angle, and also increases with the hole diameter. Numerical simulation was used to enhance the experimental database. Based on the database, a variable-length recurrent neural network is proposed which memories the crack segment information at previous learning step. The results are encouraging and provide an inspiration to evaluate the safety of components.