Optical Surface Roughness Estimation Using Artificial Intelligence

Abstract

A system for analysing steel plates is presented for an application in the wind power industry. The objective of the system is to assess the surface roughness as a preliminary step for the application of an anti-corrosion coating. A number of noncontact sensor systems were considered, with the laser triangulation scanner proving to be the optimal solution. The samples are analysed in a test stand and the roughness is measured with this optical system. At the same time, the surface is recorded with a CMOS camera. An AI model is created from the sensor fusion of both systems, which can classify individual segments of the surface as well or poorly blasted. In the following step, the surface roughness is estimated as the R_z parameter by another AI model. An error of less than 6 μm is achieved.