Application of Gaussian mixtures in the fracture identiffication

Fracture surfaces of X70 steel DWTT broken samples are analyzed using statistical methods and fractal concept. A new approach for identifying the fracture is presented. To distinguish fracture surface area the deviations between normal vectors of two neighbouring elementary triangles has been evaluated. The distribution of the angular deviation can be described as a Gaussian mixture or as a lognormal distribution. Gaussian mixtures parameters are obtained using well-known EM algorithm. Calculated fracture surface map of brittle fracture probability highly corresponds to real placement of the ductile and brittle fracture areas. The results of the statistical analysis are compared to the fractal geometry approach. Applying the first return probability method the Hurst exponent of fracture areas has been found.