Determination of the slip resistance of interspersed resin flooring with a convolutional neural network

Abstract

The ramp slip test is a reliable and widely used method to evaluate the slip resistance of interspersed synthetic resin floors, common in industrial and parking areas. Following EN 16165 and using a ramp with varying inclinations, the floor R-class can be determined. This test can, however, only be performed in a laboratory, and not on-site. Here, we propose a method to determine the floor R-class from its surface topography, which can easily be measured on-site. In this study, floors of various slip resistances were prepared following the conventional method of spreading quartz sand onto a liquid resin. The interspersed surface creates a 3-dimensional topography, which is characteristic for the anti-slip properties. After the resin hardening, the R-class was measured and, in parallel, a convolutional neural network was developed and trained with the numerous photographs taken from the specimen surfaces. The used convolutional neural network classifies 95% of the surfaces into the correct R-class. This result makes it possible, for the first time, to determine the R-class slip resistance on-site with a high probability of correctness.