Using predicted defects to manage grinding

Rolling contact fatigue defects have always plagued North American railway operators, and as the demand for higher axle loads and faster train speeds increase so do the occurrences of RCF defects. In the mid-1980s, the NRC-CSTT developed the concept of preventive rail grinding as a rail maintenance strategy to help control RCF defects and to prolong rail asset life. This grinding method was based on available research concerning RCF crack initiation and propagation.

Uncontrolled growth of these cracks allows them to progressively increase in length until they branch into the rail subsurface. At this point, the rate of growth in the vertical direction is accelerated and some cracks will grow to cause detailed fractures and rail service failures. Uncontrolled growth of these tiny cracks resulted in gauge corner (on the high rail of curves) shelling and spalling and, on the field side (on the low rail), flaking that rendered the rail unsafe and unusable. Recently the growth of subsurface deep seated shells in the high rail of curves has been identified as a rolling contact fatigue defect that can also result in rapid and severe rail failure.

The first preventive grinding method was designed to remove RCF cracks while they were still shallow and their propagation rate was still slow. Research, theory, in-track measurements and practical experience demonstrated that RCF cracks initiated faster in increasing degrees of rail curvature. The NRC-CSTT therefore recommended a staged grinding method whereby the grinder cycled the track frequently enough to ensure the cracks remained shallow in sharp curves.

This method cycled the rail grinder based on traffic density (tonnage) and curvature ranges of tangent, mild curves, and sharp curves. The staged grinding strategy was proven to be a successful approach in rail grinding tests during the late 1980s.¹ This approach was followed by the preventive-gradual grinding strategy^2,3 that provided a sound approach for transitioning from a corrective grinding strategy to a preventive one.

More recently, the NRC-CSTT, with technical and financial support from Loram Maintenance of Way and data available from BNSF, developed a revolutionary preventive grinding methodology that overcame the major shortcomings of the original preventive grinding protocol, which included:

• The logistics of staying on cycle for railway subdivisions with a high percentage of sharp curves that are hundreds of miles from subdivisions with predominantly mild curves.

• One-pass grinding of all curves in the mild and sharp curve categories at the same grinding speed (even though RCF cracks propagate at different rates for both high and low rails in each curvature range).

• Variances in the annual tonnages across each subdivision.

• Some grinding cycles eventually need multiple grinding passes on sharper curves when RCF cracks are too deep to control in one pass.

• Not knowing how deep the RCF cracks are at the start of grinding and subsequently not knowing if they had been completely removed when grinding was finished.

• Inefficiency of multiple-pass grinding in terms of uncertainty in grinding speeds and total passes for grinding planning and predicting traffic interruptions.

Based on practical experience, NRC-CSTT filtered the rail wear data (collected from the BNSF Track Geometry Car) and grinding data (supplied from Loram) for two 125-mile subdivisions between 1999 and 2006. The NRC-CSTT analyzed these data and selected 296 representative curves, as well as an appropriate sample of tangent locations. One subdivision was predominantly comprised of concrete ties, while the other subdivision was predominantly comprised of timber ties. Other information analyzed included: typical lubrication strategies, annual tonnage for each subdivision, detail fracture data, cost of maintenance and economic factors.

Spreadsheets were developed to represent rail wear caused by wheels and by grinding. From these data, “linearized” rates (per 100 mgt of traffic) were developed for wear resulting from the ball of the rail by train wheels (Wheel Wear Index) and by grinding (Metal Removal Index), as well as the wear attributed to the gauge face by train wheels.

The NRC-CSTT developed the generic RCF crack propagation curve (RCF Index shown in Figure 1) for heavy-haul track from an extensive literature review. This curve shows the relationship between crack depth and accumulated tonnage. This RCF Index was calibrated for heavy-haul traffic on BNSF using their current 136- and 141-pound rail sections. For BNSF, each degree of curvature (from 0.5 to 6.5 degrees and greater) and each rail position (tangent, high or low rail) now has a representative RCF Index.

The NRC-CSTT has developed optimization and wear progression algorithms (a result of the analysis is shown in Figure 2 for BNSF’s Lakeside subdivision) for both the subdivisions on BNSF showing the breakdown of the track curvature, the annual tonnage and the Loram metal removal index at the ball of the rail. Two grinding strategies, with or without good lubrication, were developed based on optimal number of pass-miles or optimal number of spark hours. Each strategy has the calculated requirements for:

• Grinding cycle in mgt.

• Number of grinding passes per grinding cycle.

• Grinding speed for each rail (Note: Where grinding speed differs from rail to rail in a curve, the rail with the higher speed will have to have a power reduction to accommodate the lower speed of the other rail.).

• Estimated rail life for each strategy.

The RCF Index can be calibrated for any railroad based on the specific history of wear and grinding for each subdivision or territory. Using the tonnage since the last grind, the curvature of the track and the rail position, the new predictive preventive grinding model can recommend the grinding strategy for each rail position.

In contrast to the old model, the RCF Index predictor enables the grinding manager or supervisor to predict RCF crack depths and to remove them during the next grinding cycle. Thus the predictive preventive grinding method is truly preventive, removing only the amount of material required for a given curve based on historical data; not on simple generalizations as a “mild” or “sharp” curve.

Based on a return on investment analysis of track time, grinding cost and rail life for two grinding strategies, one with good lubrication and the other with poor, the new grinding strategy can extend the rail life by a further 22 percent to 58 percent due to improved rail grinding efficiency.

The predictive strategy:

• Helps the railroad to justify predictive preventive rail grinding as the most-cost-effective maintenance strategy in terms of savings in rail maintenance costs and rail life as compared to other preventive or corrective rail grinding strategies.

• Helps the railroad understand cost implications of changing the grinding cycle from the optimized grinding strategy.

• Helps the railroad justify grinding budgets based on ROI for each subdivision that is not being ground at optimal cycles.

• Enables better planning of rail grinding for each mile of track and propose alternative solutions for the optimized management of track segments.

• Allows for the calculation of grinding costs using smaller rail grinders to cover small pockets of sharp curves that now require long travel distances by the large production grinders.

• Takes the guesswork out of grinding each track segment by predicting the depth of RCF cracks and optimizing the metal removal for RCF control and profile production at each track segment.

• Allows grinding quality indices to be based on both RCF control and profile quality as compared to the current indices which are based entirely on rail profile quality.

Loram, BNSF and NRC-CSTT envision that the optimization and wear progression algorithms will be used in Loram Grinding Data Management software and/or by the BNSF Rail Maintenance Manager to produce the grinding plan for each subdivision and territory. The software may also be used onboard the Loram rail grinder and the Loram Rail Inspection Vehicle in real time to determine the day-to-day grinding plan.

For example, each track location can be input into the spreadsheet with the respective tonnage since last grind and the program will calculate the depth index of metal removal from the ball of the rail to remove/control RCF crack. Loram metal removal software, along with an NRC-CSTT calibrated RCF Index, can be used to overlay the NRC-CSTT optimized rail grinding templates onto the in-track rail profile to calculate the number of passes, the horsepower delivered to the pattern and the grinding speed for each pass.

References

1. J. Kalousek, C. Hegelund, P. Sroba, “Analysis of Rail Grinding Tests and Implications for Corrective and Preventive Grinding,” International Heavy Haul Conference, Brisbane, Australia, 1989.

2. J. Stanford, P. Sroba, E. Magel, “Burlington Northern Santa Fe Preventive Gradual Initiative,” AREMA, Chicago, Ill., September 1999.

3. J. Stanford, P. Sroba, E. Magel, “Transitioning from Corrective to Preventive Rail Grinding on the BNSF Railway,” International Heavy Haul Conference, Brisbane, Australia, 2001.