Mechanical parts wear out over time, leading to different types of surface damage. Characterizing this damage helps understand what caused the failure. Traditionally, human experts visually inspect the worn surfaces, but this process is slow and subjective. Automated methods using image analysis have been explored, but these methods overlook the depth of the worn surfaces and need many images to work. Alternatively, assessing the worn topography in three-dimensions (3D) can help to identify the different types of damages. This thesis presents a new framework to identify wear damage on metal parts. It uses 3D measurements of worn surfaces combined with an artificial neural network. Two new metrics are introduced: surface motion orientation (Smo) and surface stratification ratio (Ssr). Smo shows how damage is aligned with the direction of movement, and Ssr measures the ratio of peaks to valleys on the surface. These metrics serves as input to a neural network developed to classify between indentation, grooving, pitting and adhesive failure. The proposed framework exhibited an accuracy of 94% in classifying these wear failures in samples originated from micro-abrasion and gear tests. This new framework offers a reliable and efficient monitoring tool to identify wear damage on metal parts.