SENSITIVITY ANALYSIS OF FREQUENCY RESOLUTION AND FREQUENCY DOMAIN FOR HAMMERING SOUND TEST USING K-NEAREST NEIGHBOR

Abstract

In recent years, many of the civil infrastructures that were constructed intensively during the period of rapid economic growth have passed 50 years since their construction, and the number of deteriorated and damaged parts due to the aging of the structures has increased. About 60% of the civil infrastructure, the road bridges will be 50 years old within the next 10 years and it means condition of bridge will be deteriorated further. As a result, in Japan, the Road Law Enforcement Regulations were partially amended in 2014, requiring road bridges to be visual inspection, with bridges were inspected by inspector once every five years. Only close visual inspection is not possible to detect deterioration and damage inside the structure. Therefore, it needs to hammer a sound test by “test hammer”.

The close visual inspection is a method to detect the spalling in the concrete structure by the sound of hammering the concrete structure with test hammer. In order to reduce the dispersion of the identification results and to improve the identification accuracy, we attempted to use machine learning to automatically determine the results of hammering test by computer.