Autonomous classification and quantification of abrasive wear ‐ for efficient materials development allowing long service lives

Research question

Reducing the need for replacing components and tools used by industry and society by giving them longer service life is paramount for both a sustainable society and efficient production. One way to achieve this is to use materials wisely, selecting and tuning the material to resist wear in whatever form it occurs. The problem is that when examining worn surfaces, they often look chaotic
and cannot give clear indications of what was the mechanisms causing wear. The proposed project is
literally an attempt to bring order out of such chaos.
The project will use dedicated tribological tests that allow individual damage and scratches to be characterized, classified and counted when imaged at high magnification using electron microscopy. Typically, an understanding of what mechanisms that dictate the wear requires many images, and a huge amount of scratches that need to be characterized and classified. To make this a
viable process route, we need to take advantage of software and machine learning. For those techniques, the problem largely transforms into an autonomous decision challenge. In this project, a neuro-symbolic artificial intelligence (AI) approach will provide the grounds for the automated decision processes. Characterizing worn surfaces in this way has not been tried before.

Sustainability aspects

The vision is that the procedures developed in this project can efficiently bring a crucial understanding of how to select materials and how to design materials to reduce particular modes of wear, in many areas. Manufacturers can then optimize their materials to give their tools and components long service lives. This reduces the need for raw materials and energy,  contributing to sustainability, benefiting both industry and society.