Hammering Test for Tile Wall Using Deep Learning

Economic activities heavily rely on social infrastructure, such as bridges, tunnels, public structures, and buildings. In Japan, regular inspections are mandated by law to ensure these assets remain functional. This kind of requirement is rare worldwide. Similar regulations exist only in major cities in the United States and Canada. These inspections often focus on detecting issues that are not visible to the naked eye, such as cavities within concrete walls. The most widely used method for such inspections is the hammering test, in which inspectors analyze the sound variations produced when a hammer strikes a surface. By interpreting these auditory changes, they can assess the structural integrity and identify hidden defects within a structure. While effective, the accuracy of this method strongly depends on the inspector’s sensory perception, which varies with individual experience and skill. Furthermore, because the test is conducted manually, it is time-consuming, especially for large-scale structures. The most critical problem is the shortage of skilled inspectors due to retirement. It is difficult to meet the demand for inspections. To solve this problem, this paper proposes an AI-based hammering test system that analyzes and evaluates structural health by detecting abnormal sounds during inspections. This system allows even less-experienced workers to accurately identify defective areas. The focus of this study is on the application of AI-driven sound analysis for inspecting adhesive-applied tile walls, a common feature in modern apartment buildings. By integrating AI technology, this approach promises to improve the consistency and quality of inspections, reduce reliance on human expertise, and significantly enhance the overall efficiency of the inspection process.