Prediction of fatigue crack growth rate in aircraft aluminum alloys using optimized neural networks

Abstract

In aerospace industry, Fatigue Crack Propagation pose a serious threat to the professionals involved in designing mechanical assembly of the aircraft structures. In these structures crack growth is a problem that needs to be handled seriously, as human life risk is concerned in addition to economic loss. Fatigue Crack Growth (FCG) Rate is the rate at which crack grows with number of cycles subjected to constant amplitude loading. FCG curve is drawn between crack growth rate on y-axis and Stress Intensity Factor (SIF) range on x-axis. It needs to be predicted accurately to avoid losses. Upon analyzing the curve, it becomes obvious that the correlation between Stress Intensity Factor (SIF) ranges “ΔK” with FCG rate “da/dN” is deviating linear relationship considering region II of the curve that is also called Paris Region. Empirical formulation methods cannot deal with linearity factor satisfactorily. Other hybrid techniques are also found incapable of dealing with non-linearity suitably. In contrast to the prior methods, machine learning algorithms are capable to deal with the non-linearity issue in a much better way owing to their admirable learning ability and flexible nature. In this research work three distinct MLA based Optimized Neural Networks are utilized for prediction of FCG rate. The used algorithms include Genetic Algorithm based Optimized Neural Network, Hill Climbing based Optimized Neural Network and Simulated Annealing based Optimized Neural Network. The algorithms presented in the proposed technique are validated by testing on different aluminum alloys used for aerospace industry that includes 2324-T39, 7055-T7511 and 6013-T651 aluminum alloys. The least predicted MSE for 2324-T39 aluminum alloy is achieved by Hill Climbing based Optimized Neural Network that is 3.1069×10-8. For 7055-T7511 alloy, minimum predicted MSE is 1.4284×10-9 which is achieved by Hill Climbing based Optimized Neural Network. Finally, the least predicted MSE for 6013-T651 is 1.0559×10-9 that is achieved by Simulated Annealing based optimized Neural Network. Taking all alloys on which experiments were held with used algorithms, the least predicted MSE that is attained is 1.0559×10-9 for 6013-T651 Aluminum Alloy with Simulated Annealing based Optimized Neural Network. Moreover, the results demonstrate an exceptional conformity to the data conceived during experimentation process.