Thursday, June 14, 2012

Maintenance-of-Way Challenges: Automated tie inspection

Written by 
GREX's Aurora Track Inspection System on track just outside a tunnel. GREX's Aurora Track Inspection System on track just outside a tunnel.
Automated crosstie evaluation systems work toward removing  subjectivity and improving accuracy of crosstie inspections.
North American railroads spend a significant amount of their capital budgets each year on crossties. While inspection and evaluation are aimed at making sure those ties last as long as is safely possible, walking tie inspectors have many variables, such as experience, training, workload and environment that can affect their perception of tie condition.

“Automation of the process using machine vision technology removes the variables and provides consistency of judgment,” said Lynn Turner, vice president sales and marketing with Georgetown Rail Equipment Company.

With the Federal Railroad Administration’s rule requiring automated inspection of concrete crosstie track, the automation of tie inspection will continue to grow in importance. While the systems available today vary in the technology used in tie evaluation, they all aim to produce results that are an accurate and objective assessment of a crosstie’s condition. 

ENSCO

ENSCO, Inc., offers several technologies for the automated evaluation of crossties, including Deployable Gage Restraint Measurement Systems (D-GRMS) and Track Component Imaging Systems. Both of these technologies can be further complimented with the measurement of rail cant and the movement of the rail under loads using ENSCO’s Rail Profile Measurement System (RPMS) to assess rail seat conditions.

The D-GRMS, a contact measurement system provided jointly by ENSCO and Plasser American Corporation, uses a hydraulically-loaded split axle to laterally push downward and outward on each rail to expose gauge restraint weaknesses. The D-GRMS utilizes a deployable fifth axle on a full-sized railbound vehicle, is capable of testing at speeds up to 50 mph and, according to ENSCO, can uniformly apply 14,000-lb. lateral loads and 21,000-lb. vertical loads.

ENSCO’s Track Component Imaging System offers high-resolution imaging of the track bed from a moving vehicle, allowing for both automated and manual inspection of ties, track bed and rail components. This technology was developed in direct response to industry demand for faster and more cost-effective approaches to grading of wood and concrete ties, detection of cracks in concrete ties and detection of missing fasteners or other components. Track Component Imaging services commenced in spring 2012 with ENSCO’s latest Comprehensive Track Inspection Vehicle (CTIV), which includes combined track geometry, rail profile and rail cant measurements with imaging technology to get a comprehensive assessment of infrastructure condition.

Eric Sherrock, senior staff engineer with ENSCO’s Applied Technology and Engineering Division, says the company devotes a significant portion of its research and development to improving key aspects within its machine-vision inspection systems including lighting and feature detection algorithms.

“ENSCO is continually developing enhancements to its automated image processing algorithms in order to provide operators with consistent, objective and an efficient means to inspect tie and fastener conditions,” said Sherrock. “The industry is learning that by combining the results of different measurement and inspection systems they can get a clearer picture about the condition of the infrastructure. ENSCO is currently evaluating the combined results of rail profile measurement, including rail cant, with those provided by the D-GRMS and track component images to provide objective assessments of fastener effectiveness and rail seat conditions.  Comprehensive inspection of track conditions using complimentary technologies will play a vital role in the industry’s ability to address critical track component issues, such as rail seat deterioration on concrete ties.”

Sherrock points out that D-GRMS is a well-established technology that produces accurate and repeatable data, especially on concrete ties, and the ultimate goal of any machine-vision-based system is the same.

Image
From GREX, showing an evaluation of opposite truck side camera/laser measurement performance on common rail side.

“To achieve this goal, the quality of the image is critical. ENSCO has directed significant resources to optimizing image quality, which involves striking the right balance between lighting, image resolution and other important factors. For automated detection of particular track features, the development of reliable machine vision algorithms capable of providing accurate and repeatable results is a complicated process in which the designers have to strike the right balance between maximizing event detection while minimizing false ‘positive’ identifications and missed events.  The success of algorithms of this nature is highly dependent on both robustness of the methodology and the quality of images that are used to develop the processes. ENSCO has devoted much of its development resources on these issues in order to ensure the reliability of the system and the confidence of the user,” said Sherrock. 

He also points out that the impact of test conditions on inspection results must be minimized, noting that the D-GRMS operating speed is capped at 50 mph because contact-based systems are less susceptible to conditions, such as day and night, but more susceptible to conditions, such as speed.

“For a number of reasons, D-GRMS is not typically run through switches. It is capable of operating in all weather conditions with little adverse effects. Non-contact, optical based systems are more susceptible to weather conditions and day and night conditions and less susceptible to speed conditions. For that reason, ENSCO has spent considerable effort in developing a lighting system that allows the use of its Track Component Imaging System in both day and night conditions.  As with all optical inspection technology, care must be taken to minimize the impact of dirt, debris, water and other environmental conditions on windows and lenses. Work continues on the optical-based systems to minimize the effects of weather conditions and night or day inspection,” said Sherrock.

GREX

“Anticipating North American railroad demand for a comprehensive and repeatable method of inspection for the second most costly capital expense, crossties, GREX committed a large amount of resources to deliver a system capable of providing a solution,” said Turner.

Georgetown Rail Equipment Company says its Aurora Track Inspection System began in 2003 as a push cart equipped with cameras and after talking with Class 1 railroads to understand their challenges, has developed into a hi-rail-mounted system that uses machine vision to create a three-dimensional view of the track for the purpose of grading wood ties and collecting rail seat abrasion (RSA) measurements on concrete ties. The system collects data at up to 42 mph and can operate day or night.

Greg Grissom, vice president engineering at GREX, says that for wood tie grading, Aurora measures 25 variables (including plate cut, splits, warpage, roughness and other variables) to determine a tie score for each tie based on a statistical method. According to Grissom, for concrete ties, the system records height measurements at regions of interest on the tie and on the base of the rail and then performs calculations to output RSA measurements. Data processing occurs onboard and reports are provided within 48 hours that contain customized information to identify clusters of bad ties, plate cut information, enhanced ballast detection and component inventory.

“GREX has invested substantially in interviewing and conducting field testing with customers to identify avenues of development and enhancements for the Aurora Tie Inspection System,” said Turner. “The results are improved grading models, on-board diagnostics, enhanced ballast detection, enhanced tie plate and plate cut algorithms and enhanced GPS identity, which currently provides an accuracy of +/- 15 cm 95 percent of the time when good signals are present. New developments include GPS filtering that incorporates odometer and rate gyro data to interpolate GPS readings when signals are weak or do not exist.”

To make sure the Aurora system produces consistent results, Grissom says the Aurora inspection vehicles undergo a rigorous annual re-certification process, in which the Aurora team verifies and/or corrects the camera/laser configurations, performs annual maintenance, then collects data on track for comparative analysis with all other vehicles before being approved for production. 

“For RSA, the repeatability of opposite direction scans and one vehicle versus another vehicle scans has been well studied. Table 1 shows results from a recent opposite direction analysis of the same truck. The difference in RSA measurements from one direction to another was less than 1/8 inch for all measurements and for 99 percent of the ties (on both rails) was less than 1/16 inch. Figure 1 shows that the calculated difference in opposite direction measurements almost always fall within the range of +/- 1/16 inch and fall within +/- 1/32 inch better than 82 percent of the time,” said Grissom.

He continued, “When comparing multiple trucks, RSA measurements have been repeatable within 1/16 inch better than 95 percent of the time.  For wood tie grading, the repeatability, on a four-scale grading system, of one vehicle versus another in the same direction is better than 90 percent. The repeatability of one vehicle versus another in opposite directions is better than 80 percent.”

Image
From GREX, showing opposite direction measurement data.

Grissom also said the Aurora team performs regular audits to verify its grading and measurements on track alongside its customers, which produces information to make adjustments to become more accurate. 

“Much like ultra-sonic rail testing development in the early 90’s and the recognition of the benefits gained in rail grinding, tie condition reporting using automated methods appears to be an integral part of railroads physical plant condition assessment for now and the future,” said Turner.

Rail Radar®

Tom Keogh, president of Rail Radar®, says the company’s Tie Inventory and Inspection System is based on an expert, rules-based system using criteria and thresholds that can be adjusted for each job, each location and each class of track. The system can assess ballast and tie conditions, either together or separately, and utilizes a combination of optical imagery, illumination lasers and measurement lasers. Keogh notes that the system is less vulnerable to ambient conditions, such as shadow and brightness, because it does not use infrared technology.

The system evaluates both wood and concrete ties and identifies potential defects including crack severity, plate cut on wood ties and rail seat abrasion on concrete ties.

“We have built into the system that the railroad can establish its own site-specific tie threshold parameters, which means the railroad can determine what the index is that determines whether a tie is defective or not, which varies by railroad and class of track,” said Keogh. “We also collect data for fastening systems, so we can tell the railroad what the pattern is and whether the fasteners are there or not. We can also identify rail joints and our system separates between curves and tangent.”

Keogh says the company has spent a lot of time developing precise positional referencing. He says that once a potential defective tie is found, it is marked by milepost, lineal distance, landmark and tie number and an online video is then given to the railroad. The railroad can take the data and load it onto a GPS device, which a field crew or contractor can use to precisely identify the tie or fastening system component that is to be replaced.

According to Keogh, the system has been in development since 2005 and was put in production about three and a half years ago. Rail Radar used actual field data and input from the railroads to validate findings flagged by the technology. Ties that were identified by the system were put up against the railroad’s walking tie inspectors. Keogh says that one Class 1 railroad that has used the system for some time spends less and less time validating its results and is now using the system for tie and fastener replacement planning and rehabilitation.

As far as the future of this technology goes, for Keogh, it’s primarily in the evolution of the software.

“The fundamental aspects of this technology is not new, what is new is the way we are putting it together and how it works together,” said Keogh. “The data processing and recognition software and the way we do the coding are the strength of our systems and our future.”

There is one more key aspect in the future of automated tie assessment and that is the approach and advancement of the railroads, notes Keogh.

“There is no doubt in our view that automated tie and ballast assessment is going to replace the walking tie and ballast inspectors. It’s just a question of when,” said Keogh. “It really depends on the progressiveness of the railroad and a willingness to adopt the use of this technology.” 

blog comments powered by Disqus