우천 환경 노출 시간에 따른 초고속 열차 차체의 수명평가 기법 연구: 초음파？딥러닝 주파수 분석 알고리즘 기반으로

초록

Improving the performance of high-speed railroad vehicles exposes the train body to substantial stress and fatigue, which are critical factors leading to catastrophic failures due to minuscule micro-damage. In this research, aluminum plates, commonly used in sub-sonic speed vehicles, high-speed trains, and aircraft, are employed to assess the duration of moisture exposure, a significant contributor to micro-damage and structural integrity deterioration. The appropriate guided wave mode and its frequency response are validated. A Convolutional Neural Network (CNN) detection architecture is developed and trained using images of the frequency response obtained from the selected guided wave mode. Learning samples with a consistent hydrogen ion concentration (pH 5.6) and exposure durations ranging from 1000 to 3000 hours at 200-hour intervals are provided. To validate the CNN detector, five samples exposed to the solution for durations varying from 1000 to 3000 hours at 500-hour intervals are used for exposure duration and damage assessment. The proposed algorithm demonstrates a reasonable detection error rate. Furthermore, surface conditions are evaluated using a scanning electron microscope (SEM) to ascertain the extent of surface corrosion resulting from moisture.