Sophisticated programs enable railroad engineering departments to use available data to automate decision-making for maintenance of way.

Sophisticated technology is enabling railroads to not only find possible problems, but also to predict their occurrence so engineering departments can plan m/w programs to deal with issues before they cause trouble.

Wear-based rail replacement

"Maintenance of rail continues to be one of the major cost areas in the maintenance-of-way budget," said Dr. Allan M. Zarembski, vice president and general manager, ZETA-TECH, A Harsco Rail Business Unit. "While great strides have been taken to extend the life of the rail using a range of maintenance techniques to include grinding, lubrication and rail testing, significant quantities of rail are still replaced each year. Given the high cost of rail replacement, of the order of $500,000 a mile, this represents a major capital outlay that must be properly budgeted, planned for and scheduled. Accurate knowledge of the condition of each segment of rail and when that rail needs to be replaced allows for efficient planning and implementation of this expensive maintenance activity."

Rail is replaced for a number of reasons, including wear and fatigue, both internal and surface. While planning tools such as ZETA-TECH's RailLife rail management and replacement planning model address both wear- and fatigue-related rail replacement, a great deal of focus is currently being placed on management of rail wear and planning of wear-related replacement such as curve patching.

"That is because wear remains a primary reason for rail replacement on curves, where the rail life is significantly reduced from tangent track, even in an environment of effective lubrication and grinding," he noted. "Wear also remains a criterion for rail removal on tangent track, particularly when wheel contact-related wear is augmented by the effects of many years of rail grinding.

"By monitoring the rate of rail wear, using state of the art tools such as laser-based profile measurement systems mounted on track inspection vehicles, and combining these data with other relevant information available in most major railroad data bases, it is possible to effectively plan and forecast wear-related rail replacement programs," Zarembski said. "ZETA-TECH's RailWear is an example of such a planning tool that has been extensively used by such railroads as CSX Transportation and Network Rail (UK) to analyze tens of thousands of miles of rail profile- (and wear) related data and convert those data into rail replacement programs."

This is a multi-step process. The first step is to manage the extensive data files from the geometry car-mounted laser-based profile measurement systems, which are taken as frequently as every 10 feet, or over 500 images per mile. These data are recorded into an appropriate storage structure, categorizing the files by recording date, line, or other reference.

Figure 1. Sample Profile Analysis.

The second step is the analysis of the recorded profiles, as shown in Figure 1. In the case of the RailWear analysis process used on Network Rail, for each profile the following analysis steps are performed:

• • Initial reading of the profile - Basic checks are performed to determine if the profile meets minimum quality conditions.
• • Profile alignment - The profile is rotated to vertical and shifted to a common reference point.
• • Profile quality filtering - Geometric properties of the profile are compared against quality standards to determine if a profile is sufficient for analysis, i.e., does it look like a proper rail shape?
• • Template matching - If the rail section at the location is not reliably known, the best-fitting rail template is selected from the library of rail templates.
• • Profile overlay - The rail profile is plotted over a plot of the original rail section for visual comparison.
• • Wear calculation - The defined wear parameters are determined by comparing the rail profile to the matched template. The wear values are compared to the defined limits, to determine if there is a current exceedance (Figure 2) or if the wear is approaching an exceedance level.

Figure 2. Sample Wear Table.

In all cases where the wear value is below the exceedance limit, the wear data are stored for the determination of wear rate and projected rail life.

Among the rail wear values now used on Network Rail are:

• • Vertical or head wear
• • Gauge face (side) wear
• • Field side wear
• • Rail head width
• • Estimated rail depth (height of rail less top wear)
• • Joint bar clearance
• • Gauge corner (45º ) wear
• • Gauge face remaining
• • Gauge face angle
• • Head area
• • % loss or % remaining
• • Plastic flow or lip
• • Deviation from standard inclination (cant)

Figure 3. Sample exceedance validation window.

When any of the wear values exceeds the pre-set limits, that location is added to the wear table, together with the recorded profile and analyzed wear data (Figure 3).

The data are also used for overall curve analysis using the curve replacement decision algorithm.

In addition, a set of additional condition indices, i.e., quality indices, are determined for each profile and recorded to the Wear Table to include:

Profile Quality Index (PQI) - A measure of the general quality of the rail profile;

Match Quality Index (MQI) - A measure of the closeness-of-fit of the rail profile to the matched template;

Cross-Beam Index (CBI) - A measure of the divergence of the top of rail head measured by the two beams

The result is a set of curve relay reports in either tabular (Figure 4) or track chart (Figure 5) format.

Figure 4. Curve Relay Report.

In addition to the wear calculations, a Grinding Index Calculation can be performed where the RailWear model calculates a Grinding Index for each profile or each curve segment. The grinding index is a measure of the closeness-of-fit of the rail profile to a target template shape which is defined based on curvature, rail side, etc. The GI can be used to indicate grinding necessity and relative grinding priority.

Figure 5: Track chart-based curve relay report.

The third and final step in the analysis process is the Wear Analysis which is the calculation of the actual (measured) rail wear rate and the determination of the rail replacement date, i.e., the projected date when the rail will need to be replaced. This wear analysis thus determines values such as average wear rate over a curve, and remaining life of the rail based on the calculated wear rate and the exceedance limit. Tonnage and curve data are used in this wear analysis.

However, accurate lubrication or metallurgy information is not necessary since the analysis uses the actual wear rate for each curve or tangent section. The analysis of the wear rate can be used to verify the level of lubrication, since an excessively high rate of rail wear can be indicative of inadequate lubrication.

The rail wear analysis algorithms develop a statistically based rail wear loss rate to predict future rail wear and rail replacement points. The statistical analysis approach used is a multivariate regression analysis of the rail wear data, using all available rail wear inspection data.

As a supplement to this statistical analysis, an engineering-based rail wear model is also used. This model provides a "sanity check" on the forecast results and identifies segments with data problems. It also allows for the forecasting of rail replacement dates, even when the rail wear rates are not sufficiently large as to permit good rail wear inspection measurement data. It also provides for the identification of high wear rate segments for further investigation as to the cause of the high wear, e.g., inoperative lubricators.

In the actual implementation on a large scale basis, a hierarchy of analyses approach is used for the wear life calculations for each segment based on the amount of data available, the quality of the data, and the results of each incremental stage. Thus, the model includes good data in the analysis, while dismissing anomalies and outliers, in an effort to calculate an accurate wear rate. When there is insufficient data to calculate a statistical wear rate by regression, the model is supplemented by ZETA-TECH's engineering equation to provide a replacement forecast.

Figure 6A. Example of wear data and good data fit.

The multivariable linear regression is applied to each wear segment when the wear data contain multiple measurements made on more than two distinct dates. The quality of the regression line is reliant on both the quantity and quality of the data obtained. The regression analysis calculated both a vertical wear rate, as well as a gauge wear rate, by determining a trend in the data (Figure 6A).

Figure 6B. Example of regression and single point wear rates.

For cases where there are insufficient data to determine a valid wear trend or the calculated rates were not realistic. e.g., negative wear rates, a wear rate is calculated between the last wear measurement and the estimated installation date of the rail (in the RailWear and RailLife models this is called the Single Point Wear Rate). 

At the rail install date, it is assumed that the cumulative mgt on the rail is zero, and both the vertical and gauge side wear are zero. Figure 6B illustrates the data-based and partially data-based analyses that are done on segments with wear data.

For cases where there are insufficient data to determine a valid wear trend or the calculated rates were not realistic, e.g., negative wear rates, the engineering wear rate equations allow for the calculation of rail wear rates based on the specific properties of the track segment and the rail. Properties such as maximum line speed and track curvature are major factors in the engineering wear equations.

The calculated wear rates and the most recent wear measurements are input into the remaining life wear equations to calculate how much additional tonnage can be accumulated on the rail until one of the wear limits will be exceeded. Knowing the annual mgt for the segment, an estimate of the replacement date, i.e.. replacement year is determined from the last measurement date based on assumed future annual traffic (mgt) levels.

"Note, if neither the regression wear rates nor the single point wear rates produce valid results, the engineering wear rates are used to calculate the expected rail life from the date of installation," Zarembski said.

Figure 7: Summary of annual rail wear replacement, 2009-2013.

In all cases, the remaining rail wear life and the corresponding replacement year is calculated concurrently for side, head and joint bar clearance. The final output of the wear analysis is the forecast replacement date when one of the wear limits will be exceeded. Figure 7 illustrates a summary five year forecast which is also available in tabular format on a segment by segment basis. Such a multi-year forecast allows maintenance-of-way departments to develop medium and long-term rail maintenance strategies in order to most effectively plan their maintenance programs and budgets.

Bentley

Bentley's Optram^TM is an information system for railway corridor infrastructure management. The system includes an Optram Linear Data Analysis module which is a library of analysis, logical, and data alignment functions built specifically to process corridor information such as track measurements, events (such as rail defects, work records) and asset data.

LDA script-driven capability allowing users to tailor corridor data analysis to each railway's unique data types, policies, asset degradation characteristics and business processes. The Swedish national rail system, Bankveret, uses Bentley's LDA technology.

ENSCO

ENSCO, Inc., offers TrackIT^TM, a centralized Web-based system for managing track infrastructure, inspection and maintenance information. This product enables engineering personnel to retrieve the current state of the track network for viewing and analysis of data in map, tabular and track chart views through the Web browser. The product provides a means of accessing video images of the right-of-way.

Additionally, the product offers inspection history overlays in a track chart view. A simple query interface allows the user to dynamically render a track chart and rapidly view run on run result comparisons of the latest track geometry, rail flaw or Vehicle/Track Interaction automated inspections.

Over the past year, ENSCO has also focused on the development of two important automated maintenance of way planning tools that will feed into the TrackIT^TM system. ENSCO, with the support of FRA Office of Research and Development, is developing an Autonomous Track Geometry Measurement System to directly measure track conditions such as track gauge, curvature, crosslevel, warp, profile, alignment and limiting speeds. Fully-equipped with ENSCO-developed non-contact inertial and laser sensors, the ATGMS will measure track geometry parameters at speeds of up to 160 mph. A rail car or locomotive equipped with this system will be able to survey significantly more track compared to the typical dedicated track geometry car over a given period of time at a fraction of the cost per mile ENSCO notes.

Use of ATGMS will facilitate more frequent track inspections, which will improve safety and allow for more efficient maintenance planning. When defects are detected, ATGMS will transmit a notification via wireless communication from the vehicle. making survey results available in near real time via the TrackIT^TM web application. 

Many railroads are considering the use of geometry systems of this nature, and the FRA Office of Research and Development has publically addressed its vision for the role of autonomous technology in the future.

ENSCO is also planning a spring release of the Digital Track Notebook (DTN^TM) 3.0 - its latest product upgrade for managing track inspections within a Web browser on office and laptop field computers. The ENSCO DTN^TM will be configured to incorporate the FRA regulatory and internal railroad business rules into a flexible design that offers assurance that track inspection regulatory compliance is met, regulatory and non-regulatory track related defects are identified and remedial maintenance actions are taken in a timely manner.

Holland

Holland L.P.'s TrackSTAR® testing vehicles use a Split Load-Axle system that applies constant and consistent vertical and lateral loading Split-Load Axle to the rails during testing. The patented TrackSTAR® test vehicles measure track gauge twice; first unloaded gauge is measured at the front of the test vehicle, and second, loaded gauge is measured at the load axle near the middle of the vehicle. The on-board computers align the two gauge measurements in the data set and the difference between the two gauges, delta gauge, is used as a key input to determining the reserve gauge strength of the track to dynamic loadings. The on-board strip charts and exception reports provide immediate feedback showing where crossties or fasteners are weak.

Gauge Strength Testing is "performance-based" measurement of the condition of the cross-ties and fastening system. As such, it is not influenced by what crossties/fasteners look like, but rather provides objective and accurate information on how crossties/fasteners are performing under load.

Also, Holland's TrackSTAR® vehicles use lasers and cameras to measure more than 300 points on the rail surface for full railhead profile assessment and rail inclination.

Through post-processing software packages, Holland is able to quantify rail information, including: rail size, vertical and gauge-face wear, field wear, lip, gauge-face angle and cant. These data can then be used for rail wear analysis and rail replacement forecasting. Benefits include: spot maintenance indication/prioritization; rail classification by wear; long-range rail replacement planning (rail program); verification of grinding program effectiveness, maximization rail life; and locating areas of accelerated rail wear and taking corrective action.

In addition, Industrial Metrics' Rangecam software programs turn track geometry and rail and wheel measurement data into useful information. Rangecam programs such as Track View and Track Analyst provide comprehensive reporting of track geometry, GRMS, rail wear, profile and rail-flaw data. Data are graphically represented on GPS-derived route maps that can be overlaid on satellite, terrain and street-view maps. Track condition, exception and other reports are generated by pre-programmed, menu-driven search tools that can be customized for each user. Charts and graphs measure the rate of change and show how track degrades over time.

The company notes that these software programs remove the guesswork from rail and tie replacement planning. Wear rates can be calculated by time or tonnage to determine when the changeout criteria will be met for every curve or track segment. User-specified work plans and material costs take the subjectivity out of planning.

The Track Analyst Enterprise System, which connects to an Oracle or MS SQL Server database, allows railways to combine and store all of their track test data in a single database. Rangecam's comprehensive suite of analysis tools enables an Enterprise user to analyze and report on wear conditions, plan systemwide maintenance programs and extend rail and tie life.