Quantitative fractography of fatigue cracks: a new solution in 3D

摘要

A general method of statistical classification for estimating the fatigue crack growth rate on the basis of 3D features of fracture surfaces is proposed. Three CT specimens from stainless steel AISI 304L were loaded by a constant cycle, the same with regular overload, and a periodically repeating block of random cycles. Crack growth was recorded in detail. Fracture surfaces were documented by 3D reconstructions of their morphology along central axis in crack growth direction. 3D reconstructions were decomposed in 8 levels by means of sequential wavelet transformation. Each level was represented by 1662 features - parameters of roughness and morphometry computed from line profiles and areal data. Reconstitution of crack growth history on the basis of a single 3D feature at all levels of decomposition was investigated. In all cases, several tens of features enabled this with an acceptable precision. Quality of results is documented by graphs comparing crack growth rates and crack growth history estimated on the basis of fracture surface with real values known from the experiment. Training and testing approach has been applied.