Detection of fatigue damage via neural network analysis of surface topography measurements

The fatigue strength of the machined structure is significantly affected by the surface roughness because the surface topography creates the stress concentration in valleys of the surface topography; hence, causes a grain slip and a crack initiation at the surface of the machined structure, and finally leads to fatigue failures. The surface roughness usually illustrates the micro-geometric appearance variation of the machined structure. High cycle fatigue tests of polycrystalline alloy (Al7075-T6) specimens with distinctive surface textures are presented in this paper to investigate which surface textures are mostly influenced by the fatigue damage. The investigated surface textures are the average surface roughness, root-mean-square height, maximum peak height, maximum valley depth, and surface flatness. All surface textures were measured by an optical meteorology device (Alicona). This paper applies the neural network to determine the most significate surface textures that are affected by the fatigue damage, which can be determined by obtaining the best NN model performance for fatigue damage detection and classification. The results show that the best NN model is the surface flatness model, with (up to) 87.5% accuracy.