Research and Development of Efficient Acquisition of Damage Data for Shinkansen Civil Engineering Structures

Abstract

 Faced with aging and population decline, streamlining labor-intensive visual inspections is essential. This study explored the use of AI for extracting deformations in civil structures within Shinkansen track areas and developed new equipment for continuous image capturing. Results showed that AI models trained on images taken under specific conditions could extract deformations with reasonable accuracy. Additionally, the newly developed equipment successfully captured images continuously and allowed propulsion via a rail bicycle, and functioned effectively during tests on operational Shinkansen lines, enabling efficient data acquisition.