Tile Adhesion Condition Diagnosis Method Employing Machine Learning Considering Auditory Perception Characteristics

- Effect of Noise Reduction Processing -

Abstract

In this study, we considered a method to perform automatic hammering inspection using machine learning based on auditory perception characteristics for diagnosis of tile adhesion condition in drone noise. In the hammering test at quiet condition, the condition could be evaluated accurately by machine learning. However, the accuracy was not enough at the noise condition. We then applied a noise reduction method using several microphones in addition to employing auditory perception characteristic of human in the estimation model. As the result, the accuracy could be improved significantly, and it was more accurate than the correct response rate by the subjective diagnosis.