Development of Load Spectrum Generation Technique Using Artificial Neural Network

Aircrafts are subjected to complex and various loads repeatedly during the service life. These loads cause fatigue phenomena such as crack initiation and propagation, and decrease of structural strength. Aircraft Structural Integrity Program(ASIP) is required to operate aircraft safely without catastrophic aircraft accidents due to fatigue. It includes operational load spectrum generation to perform flight operation analysis and crack growth analysis that can calculate the change of crack growth rate at the critical points where stress is concentrated, and this enables appropriate structural maintenance and inspection intervals to be established. In this study, we developed a technique to apply an artificial neural network regression model to generation of the operational load spectrum from flight parameters of the aircraft, and compared it with a multiple linear regression model to evaluate the accuracy of the artificial neural network (ANN) regression model. According to shear force results, the ANN model's adjusted coefficient of determination was 0.999 and the relative error was 0.475%, whereas the linear model's adjusted coefficient of determination was 0.835 and the relative error was 27.761%. Through the artificial neural network regression model developed in this study, it was confirmed that the operational load spectrum required for ASIP could be obtained from flight parameters with sufficiently high accuracy.