NEURAL-NETWORK APPLICATIONS IN DETERMINING THE FATIGUE-CRACK OPENING LOAD

A neural network approach is developed to determine the crack opening load from differential displacement signal curves. A backpropagation n eural network of three layers was employed. In order to examine the me asurement accuracy and precision of the neural network method, compute r simulation was extensively performed for various combinations of cra ck opening levels and signal-to-noise (S/N) ratios. For all crack open ing levels examined, the method shows good accuracy and precision. The proposed method was applied in practical to constant amplitude loadin g tests and is found to provide good results. (C) 1998 Elsevier Scienc e Ltd.