Implementation and demonstration of autonomous ultrasonic track inspection using cloud-based AI rail flaw analyzer

This research successfully demonstrated autonomous rail inspection feasibility up to Technology Readiness Level (TRL) 7. A prototype integrating an autonomous rail vehicle and Sperry's Ultrasound Testing (UT) system was developed at Cranfield University. It was first tested at Cranfield’s Railways Innovation Test Area (RITA) at TRL 5 and tested at heritage operational railway, in Idridgehay, Derbyshire, UK achieving TRL 7. Experimental works included a 15-meter track test at RITA and nine rounds demonstration of a 250-meter track inspection at Idridgehay, showcasing inspection, localization, navigation accuracy, and defect location precision. The prototype successfully detected artificial rail defect during the demonstration and promptly communicated to command centre via email. We characterised the vehicle performance by measuring the positional error and detection rate. The positional accuracy measurements, verified through GPS and odometry, revealed an odometry-based error of 0.27-3.2 metres and an 8-metre GPS-associated error. The absence of differential GPS and a data fusion approach contributing to these errors. In addition, Weak 4G signal coverage in the fields impacted operator-vehicle communication and data uploading. Future iterations should address these limitations, exploring alternatives for enhanced accuracy and advancing defect-sizing technology.