Modelling the surface roughness of steel after laser hardening by using 2D visibility network, convolutional neural networks and genetic programming

Abstract

The surface characterization of materials after Robot Laser Hardening (RLH) is a technically demanding procedure. RLH is commonly used to harden parts, especially when subject to wear. By changing their surface properties, this treatment can offer several benefits such as lower costs for additional machining, no use of cooling agents or chemicals, high flexibility, local hardening, minimal deformation, high accuracy, and automated and integrated process in the production process. However, the surface roughness strongly depends on the heat treatment and parameters used in the process. This article used a network theory approach (i.e., the visibility network in 2D space) to analyze the surface roughness of tool steel EN100083-1 upon RLH. Specifically, two intelligent methods were merged in this investigation. Firstly, a genetic algorithm was applied to derive a relationship between the parameters of the robot laser cell and topological surface properties. Furthermore, convolutional neural networks allowed the assessment of surface roughness based on 2D photographic images.

Keywords

• robot laser hardening (rlh)
• roughness, genetic algorithm
• 2d visibility network
• convolutional neural networks