Industries such as railways are moving from traditional inspection methods to using artificial intelligence (AI) and machine learning (ML) to perform automated inspection of railcars. Data streaming from the edge at high rates requires the compute power of a High Performance Computing (HPC) cluster, storage and advanced analytics to return results in real time. The system also must effectively enable remote administration and management to ensure reliable system operations. Duos Technologies, in collaboration with Dell, Kalray, and NVIDIA, has pioneered the Railcar Inspection Portal (RIP), which harnesses innovative AI and ML at the edge for wayside mechanical railcar inspections.
Challenges of existing railcar inspection methods
Railways are mandated to perform inspection of railcars for every trip to correct any identified issues. Manual railcar inspections can typically inspect one railcar every two minutes. The average freight train has 120 railcars, so the time to inspect an entire train can take between 3 and 4 hours.
Introducing the Duos Technologies Railcar Inspection Portal (RIP®)
Duos deploys cutting-edge technologies that transform railroading, logistics, and intermodal transportation operations. Duos RIP® is a modular visualization system providing real-time detailed 360-degree imaging of railcars while they are in transit at speeds up to 125MPH. Data is generated by a large array of high-resolution cameras mounted on a truss that spans the track. The RIP solution employs an array of highspeed machine vision cameras generating in aggregate just over 60 Gigabytes per second of data. The system needs to ingest raw data in near real time, while producing and storing usable images at a rate of 20GB/s. In addition, the system must be able to immediately start feeding the images through AI and ML pipelines as fast as possible while ingesting and pre-processing the remaining train images. RIP offers 895TB usable capacity – the ingest of data at over 80GB/s – and has read performance of over 100GB/s.
“If the current railcar data throughput was running on an average home computer, the computer would be full in about 4 seconds,” according to Derrick Schmenk. VP of Research, Development and Innovation, Duos Technologies.
Using the RIP system reduces the railcar inspection time from 3-4 hours to a few minutes with a significant reduction in inspection time and increased accuracy in finding potential defects.
Innovative technologies used in the Railcar Inspection Portal
The speed of the data capture and the volume of railcar data being captured requires special storage and capture systems due to the enormous amount of data and very high-resolution images. The Duos Technologies RIP system uses a Dell Technologies solution incorporating technology from Kalray, and NVIDIA. A bank of 14 Dell acquisition servers captures images from the camera array and then writes the data to the storage platform. All of the servers in this solution (with the exception of a pair of storage management servers that are not involved in the processing) are Dell PowerEdge R750-servers. Multiple NVIDIA GPUs provide the lightning-speed throughput needed to process and analyze data in microseconds. The hardware is all connected via an NVIDIA InfiniBand Network. The solution uses NVIDIA SDKs for high speed image compression via NVIDIA GPUs and NVIDIA CUDA SDKs in AI algorithms.
RIP solution server management software
The RIP system uses the Dell Open Management system and Kalray server management software with 2x Ngenea NG-Stor management nodes and 4 x Ngenea NG-Box NVMe nodes. The solution allows users to remotely manage and administer the system and conduct health and performance checks on the RIP system. Schmenk states, “I am very proud of what we have been able to achieve with the Dell and Kalray teams.”
RIP solution benefits
- Allows fast reliable near real time artificial intelligence driven mechanical and safety inspection of trains
- Safely increases rail throughput
- Decreases cost for the rail companies and car owners
- Helps prevent train derailments
Summary
Traditional methods of manual railcar inspections are time consuming and costly. Duos Technologies created the Railcar Inspection Portal (RIP) and selected Dell, NVIDIA, and Kalray technology platform to utilize innovative AI and ML technology at the edge, automating railcar inspections. Progressive Railroading announced the Duos Technologies Railcar Inspection Portal and Artificial Intelligence Detection Factory solutions as the 2023 Winner for Rail Safety & Technology Products.
“The Duos RIP system is an excellent example of a new HPC solution driving game-changing business value. Traditional enterprise storage systems cannot meet the performance demands required by HPC solutions. Dell Technologies and Kalray have joined forces to deliver an optimized HPC storage solution that delivers the throughput and capacity needed to manage rapid data growth and increased demands from data-intensive HPC and AI workloads like the Duos RIP system,” states Elliott Berger, Sr. Solution Architect, Data Science Infrastructure Solutions at Dell Technologies.
Call to Action
Download this brochure for more information on high performance storage solutions from Dell Technologies and Kalray.
