Point-Based Virtual Environment for Detecting Scaffolding, Wearing, and Cracks of Furnace Walls

Abstract

Maintenance of production facilities is important in various industries. Since production facilities degrade their original functions during their long life-cycle, it is necessary to periodically make deterioration diagnosis and renovate to restore functions. In recent years, the terrestrial laser scanner allows us to capture dense point-clouds from large production facilities. Point-based virtual environment is promising for supporting maintenance of production facilities. In this paper, we discuss the deterioration diagnosis of production facilities based on point-clouds when the original shapes of production facilities are unknown. As an example of production facilities, we consider the blast furnace, which is mainly used to produce metals from molten materials. We classify deterioration on the blast furnace as wearing, scaffolding, and cracks, and automatically detect them from point-clouds. In our method, the normal wall shape is estimated by fitting low-resolution B-spline surfaces to point-clouds, and deterioration is detected as the difference between the reference surface and the point-clouds. While wearing and scaffolding regions are relatively large, cracks are thin lines. In order to detect different scales of deterioration, we introduce the reference surfaces with multiple resolutions. In our experiments, the three types of deterioration could be successfully detected from dense point-clouds.