Fatigue Crack Growth Predictions in Aging Aircraft Panels Using Optimization Neural Network

An optimization-based neural network method is developed to predict fatigue crack growth and fatigue life of multiple site damage panels found in aging aircraft. The method utilizes an optimization solution to predict the probable crack path based upon the initial panel cone guration and accounts for lead crack spanning, small multiple site damage, and plasticity zones. The approach of the neural network was motivated by the optimization analysis and the time-consuming computational analyses for multiple site damage problems. The present neural network method was able to predict crack propagation and fatigue life and compares well with the experimental data of fatigue tests on the 2024-T3 aluminum panels.