The topography of multi-scale surface textures developing upon ultrashort pulse laser irradiations can easily be controlled by the irradiation conditions. In particular, the wavelength and polarization of the laser beam as well as laser peak fluence, number of pulses irradiating per laser spot area and scan number have greatly influence on the specific morphology of the laser made self-organizing surface features ranging from nano to micro meter feature size. This allows a great variety of laser machining combinations towards the development of diverse laser surface textures consisting of manifold microscopic surface features. As a consequence, a high dimensional parameter field must be processed and evaluated in a time consuming manner for gaining understanding of the interplay between the influencing factors, and also to identify appropriate surface features for specific surface functionalities.

In this work, a novel method is presented utilizing fractal analysis algorithms in combination with machine learning procedures and database processing to categorize laser made surface textures in a cost and time efficient multi parameter matrix approach. The 3D datasets of the laser processed surfaces were captured with a laser scanning microscopy and further analysed by applying fractal box counting and k-means clustering algorithms. The fractal and lacunarity parameters enable similarity search on the surface features, differentiation by spatial complexity or heterogeneity and for anomaly detection. As a result, on the basis of the fractal analysis, the surface features can be differentiated, categorized and further linked to their process parameters. All the data are collected in a database can be used, i.e., to identify optimal parameter settings for functional topographies or rather highest processing speed etc. at minimum man power and laboratory costs.