Sunday, February 28, 2010

Take a look at this!

Written by  By James Robeda, Senior Engineer, and Semih Kalay, Vice President R&D, TTCI, for Railway Age

In recent years, machine-vision technology has found applications in the railway industry in detectors such as wheel profile measurement (WPM) systems and brake shoe measurement (BSM) systems. This type of technology enhances the inspection process by providing historical data that can be used for defect trending and preventative maintenance. The potential to increase the effectiveness and efficiency of car inspections by utilizing advanced technologies to enhance the railcar inspection process offers an opportunity to lower costs and increase productivity.

Three machine-vision based systems are currently being developed under the Association of American Railroads’ Strategic Research Initiative Program:

• Automated Inspection of Safety Appliance System (ASAIS), which assesses the condition of a railcar’s safety appliances (e.g., ladders, hand holds, and sill steps).

• Automated Inspection of Structural Components (AISC), which evaluates the condition of the railcar’s underframe and related structural members.

• Fully Automated Train Scanning System (FATSS), which images the entire railcar—top, sides, and bottom.

The ultimate goal of the industry’s Technology Driven Train Inspection Program is to deploy a network of wayside inspection sites that would feed a centralized database to create a “ticket to ride” for every car on the North American rail network.

Two vendors, Tecnogamma and BeenaVision, have an ASAIS installed for coal cars at TTCI’s Facility for Accelerated Service Testing. Algorithm development and testing is ongoing. Additionally, Beena Vision has an ASAIS being monitored in a revenue service application on the Union Pacific at Loveland, Iowa. Tecnogamma plans on having an ASAIS installed in a revenue service application on the Norfolk Southern at Ironto, Va., in early 2010.

Testing indicates that the systems are capable of detecting safety appliance defects with a success rate of greater than 85%. Currently, work is under way to refine the detection algorithms to reduce false positive rates BY up to 10%. Additionally, detection algorithms are being developed to inspect safety appliances on boxcars, covered hopper, tank cars, and intermodal cars.

An initial proof-of-concept demonstration of the AISC system was conducted by the University of Illinois at Urbana-Champaign in 2008. Pursuant to a series of feasibility studies at UIUC, TTCI contracted BeenaVision to develop a prototype system for proof-of-concept tests at FAST. A prototype system was installed at FAST, and initial testing demonstrated the system’s ability to take high-resolution images of a railcar underframe as the train passed the system at speeds up to 40 mph. Results showed that the system was capable of capturing and processing images.

The prototype system was redesigned to handle the shock and vibrations encountered at higher speeds. Further development and testing of the system will focus on image assessment and defect detection.

The imaging and processing capabilities demonstrated by the ASIAS and the AISC system prompted the development of FATSS. This system uses an array of cameras and lights to create a complete image of the top, sides, and undercarriage of each car in a train.

The ability to image the entire railcar opens the possibility for several applications. Lower positioned cameras provide detailed images of truck, brake rigging, and draft pocket components that can be scanned for defects and unusual wear. Side-mounted cameras offer an opportunity to quickly identify imbalanced loads, unsecured lading, and various carbody defects. A top-mounted camera allows for easy evaluation of top chord and interior bracing in coal hoppers. In addition to enhancing traditional inspections, FATSS provides an excellent opportunity to add security applications such as foreign object detection and tank car assessment. A vendor’s prototype system was installed at FAST in October of 2009 and preliminary image acquisition tests were conducted.

Machine vision applications hold great promise for the future of the train inspection process. Many challenges remain but progress continues to be made in meeting and overcoming these obstacles and improving the overall inspection process.