機械学習を用いた配管の打音検査手法に関する研究

抄録

In Japan, deterioration of industrial plants, built during the period of high economic growth in the middle of the 20th century, has become a social issue recently. Above all, Corrosion Under Insulation (CUI) of piping in the plant is a pressing issue. As conventional methods, X-ray inspection and ultrasound inspection have been utilized for that but still have limitations because of consuming time and inspection cost. Therefore, easy and low-cost screening techniques for CUI are required. We developed a hammering-type inspection robot that moves inside the piping and proposed an acoustic analysis method to identify anomaly parts from the hammering sound using machine learning techniques. By using testing pipes, we could successfully identify anomaly parts by the acoustic analysis using Uniform Manifold Approximation and Projection (UMAP) as a dimensionality reduction method and a neural network as a classification method.