Vision-based Rail Asset Inspection Solution

Background

The rail industry is widely recognized as the most environmentally friendly mode of transportation for goods and people. However, the adoption of rail transport faces challenges from road-based freight transportation and passenger vehicles. To provide reliable service and sufficient capacity, the industry requires a well-maintained rail infrastructure to prevent breakdowns and safety incidents.

A leading rail operator collaborated with Tata Elxsi to develop an image-based, automated rail wayside asset inspection solution. This solution promises cost and time savings through the intelligent use of technology, increasing the automation of inspection processes.

Challenge

With a vast rail network made up of variety of parts sourced from different suppliers over a long period of its existence, the challenge was to identify the best technologies and technical approaches to further improve the level of efficiency in inspection of the rail wayside assets while maintaining safety.

Solution

Tata Elxsi’s robust solution leverages image analytics supported by AI/ML models to detect and classify various types of defects in wayside rail infrastructure. The solution requires only a set of images and automatically executes a series of steps to determine if the asset is normal or faulty. The solution to detect defects is robust enough without any performance degradation while operating under a full range of environmental scenarios(shadows, vegetation and other image parameters).

The various AI/ML models developed are reusable. The solution delivers high accuracy in detection of faults.

Impact

The solution developed delivers high performance in wayside asset inspection leading to predictive health monitoring and delivers high degree of automation. The solution can also integrate with backend workorder systems to realize a full end to end process automation thus improving the overall maintenance processes.

Services Rendered

• Image based solution development for automated detection of broken parts, detection of missing or misaligned parts, measurements of critical distances between parts/gaps pertaining to wayside rail infrastructure
• Development of the ML models for object detection classification, segmentation
• Development of Computer Vision algorithms