Wear particles and surfaces are three-dimensional (3-D) objects and their numerical characterization and classification is still largely an unresolved problem. Usually a set of various parameters is employed to describe the surface topography. These parameters are of limited use, especially when dealing with anisotropic surfaces. To solve this problem a modified Hurst orientation transform (HOT) method has been developed and applied to characterize the surface anisotropy. However, despite the apparent success this method does not yet provide a full description of the surface topography. It is known that complex structures observed in nature can be described and modeled by a combination of simple mathematical rules. It is therefore reasonable to assume that, in principle, it should also be possible to describe any surface by a set of such rules. The problem is in finding those rules. For this purpose, a modified partitioned iterated function system (PIFS) was developed and applied to encode the 3-D surface topography information, i.e. to obtain full description of surface topography of wear particles and surfaces. Importantly, PIFS information gained from individual wear particles or surfaces allows to classify them in groups which are characteristic to a particular failure type. This, in turn, allows to ascribe an 'unclassified' particle or surface to a particular group/category which is characteristic to a specific failure type or wear mechanism. This forms the basis of a system, which when fully developed, would allow an automated recognition of particles and surface morphologies without the need for experts. The system then can be developed further to include diagnosis of the type of failure. In this paper an overview of recent advances and developments in the characterization, classification and recognition of wear particles and surfaces is presented. (C) 2001 Elsevier Science B.V. All rights reserved.