RCF Quantification - Towards Rail Grinding Best Practices for Australian Heavy Haul Rail

*Dhamodharan Raman1, 2

, Rajkumar Devadoss2, Gopinath Chattopadhyay

1 and Dwayne Nielsen

1, 2

1Central Queensland University

Centre for Railway Engineering, Faculty of Sciences, Engineering and Health

Queensland 4702, Australia 2CRC for Rail Innovation – Australia

*Corresponding author, Email: d.raman@cqu.edu.au

Abstract: Rail grinding is currently being performed by considering the variations in rail profile

measured through Grinding Quality Index (GQI) and has traditionally ignored rail surface condition.

Often this profile based grinding leads to excess grinding and reduction of rail life. Consequently, a

best practice approach for rail grinding comprising an updated index called Equivalent Quality Index

(EQI) and economic decision model was proposed in this paper. The EQI incorporates the rail surface

condition associated with rolling contact fatigue (RCF). However, there is a lack of knowledge and

comprehensive technology in the market / industry to measure and quantify the RCF cracks and thus

the surface condition. Consequently, a review and analysis was conducted on three non-destructive

testing methods and concluded that the eddy current technology has potential to detect and quantify

the cracks. In addition a data acquisition system and post processing software was developed to

overcome the practical limitations of the eddy current device, such as sensitiveness and unwanted

noise signals. The initial result from this proposed approach is discussed in this paper.

Key words: Rail Grinding, RCF detection, Eddy Current, Heavy Haul

1. Introduction Demand for freight and heavy haul services are rapidly growing in Australia [1]. This has driven rail operators to increase speed, axle load and traffic intensity, which has in turn accelerated the deterioration of rails. Consequently, this also increases the occurrence of major rail defects including rolling contact fatigue (RCF), surface initiated cracks, traffic wear and plastic flow. Therefore, rails demand continuous maintenance in order to mitigate these defects and ensure safety while achieving the expected higher productivity. Two key strategies employed to maintain rails are lubrication and rail grinding (Figure 1). The lubrication is employed largely to manage friction and reduce the wear on the rail and wheel to improve rail/wheel life, save energy and reduce noise. Train resistance on curves and tangent track can be reduced dramatically by proper lubrication of the rail/wheel interface. The Centre for Railway Engineering’s (CRE) team has been developing a lubricator placement model to optimise the lubrication practices in an Australian heavy haul coal line context [2, 3].

Wheel-Rail

Contact

Rail Defects

Wheel Defects

Grinding

Lubrication

Gauge Face

Lubrication

Top-of-Rail

Lubrication

Figure 1: Wheel-Rail management strategies

Rail grinding is predominantly conducted to mitigate the effects of rail corrugations, RCF, plastic flow and rail wear. RCF is caused by ratchetting due to heavy repeated loading and wear upon abrasion between steel wheel and steel rail [4] [5]. It was estimated that the RCF alone costs European railways around € 300 million (AUD$ 485 million) per year and these defects account for about 15% of the total cost of maintenance [6]. Rail grinding helps to maintain the rail profile and thus reduces the wheel/rail contact stress, which is the key factor causing rail defects [7]. It also helps to improve steering through curves and reduce hunting on tangent tracks. Grinding has progressed from conventional corrective grinding to preventive grinding. The various factors influencing rail grinding are rail profile shapes, surface condition, surface quality, grinding cycle and total grinding cost (Table 1). The best practice guidelines were predominantly established from the North American practices. With respect to those guidelines the current practices in Australia are compared and there is a need for a systematic approach to address each of the rail grinding factors. Currently a grinding quality index (GQI) is used to determine the amount of rail material to be removed by comparing measured profile and target profile. Often this profile only based approach leads to excess grinding and reduction in rail life [8]. In addition to the profile, RCF condition also needs to be included in grinding decisions, as RCF has the potential to initiate and propagate internal cracks to the level of a rail break. On the other hand, grinding decisions on when to grind, how much to grind and where to grind do not have a strong economic insight in the current practice. Impact of these grinding decisions on the total rail maintenance cost is not researched extensively [8]. Hence to overcome these critical issues on grinding effectiveness, a best practice approach was proposed in our research project which consists of an updated index termed as Equivalent Quality Index (EQI) [9] and economic model [10]. These approaches were developed with respect to pre-grinding and post-grinding field data collected over a period of two years and suggest a new rail grinding best practice for heavy haul networks in particular.

A key feature of the EQI is incorporating the rail surface condition through quantifying RCF cracks. Currently, there is a lack of knowledge and comprehensive technology in the market / industry to measure and quantify the RCF cracks and surface condition. Hence, a review and analyse of three non-destructive testing tools concluded that the eddy current technology has the potential to detect and quantify the cracks. In addition, the research developed a data acquisition system and post

processing software to overcome the practical limitations such as sensitiveness of the eddy current device and unwanted noise signals.

Table 1: Factors affecting rail grinding best practices

Factor Best Practice (IHHA)[11] Current Practice in Australia

Target Rail Profile Ensures less wear and more stability Different profiles for high/low rails in tangent & curves

Surface Condition RCF level within safe limit Not monitored adequately for grinding

Surface Quality Surface roughness/ Facet width are within tolerance

5 – 9 microns (post grinding) / 0.8 – 3 microns (pre grinding)

Grinding Cycle Meet requirements for different track geometries & Ensures a balance between wear and fatigue

Frequency: 5 – 11 MGT (Sharp curves)/ 10 – 22 MGT (Mild Curves) Material Removal: ~ 0.2 mm per pass based on surface condition assessed through digital image/ experience

Total Grinding Cost Optimal grinding cost per finished section Lack of cost effective decisions related to grinding

Firstly, this paper discusses the various Non-Destructive Testing (NDT) technologies that were analysed to quantify the RCF cracks. Then an introduction of the proposed eddy current technology based approach. The initial results from trialling this proposed approach is also discussed.

2. Background on Grinding Quality Index

In order to overcome the shortcomings of the current GQI an extended index was proposed in this research. The proposed EQI takes into account the critical rail parameters of rail profiles, RCF condition and grinding intervals (Figure 2). This will ultimately determine amount of material to be removed. There are methods available to quantify the rail profile index i.e. GQI [5] and in this research the methodology proposed by Palese et al [12] will be utilised.

Currently, rail maintainers manually intervene with pictures of the track to assess the RCF conditions. This is due to the fact that non-destructive testing and machine vision systems, though widely applied to several industries, are not yet well developed as a user friendly and cost effective solution for monitoring, controlling and quantification of RCF cracks. The key challenges in developing a RCF quantification system are:

1. Quantification of RCF crack by a reliable NDT technique at high operating speed and varying environmental condition prevalent in Australia

2. Establish a relationship between the lengths of the cracks to the depth of the cracks

3. Develop a RCF condition index based on the above relationship

Rail Profile IndexMGT -

IntervalsRCF Condition Index

Equivalent Grinding Index (EGI)

Material to be Removed

[Depth of Cut/ Speed/ Number

of Passes]

Figure 2: Equivalent Grinding Index

3. Non-Destructive Testing for Quantification of RCF cracks

Several NDT techniques (image processing using edge detection, ultrasonic testing and eddy current testing) were investigated to check if they can quantify the RCF crack and its depth. The findings are reported in following section.

3.1. Image processing using edge detection Initially, a proposed image segmentation method using edge detection algorithms to detect and measure the transverse crack length of the RCF cracks was tested. An algorithm was developed using the image processing module in ‘Matlab’ to detect the cracks as edges. An edge is a place or the pixel value in an image where there is a rapid change in the intensity (brightness) of the image. By finding the edges that are continuous it is possible to locate the end points of the cracks and then measure the distance between them. There are various edge detection methods available in ‘Matlab’. A canny edge detection algorithm was used in the test. Two images, one taken by a high definition camera installed on a rail inspection vehicle (RIV) and another taken under controlled lighting conditions after magnetic particle testing was checked using the algorithm. The results for the image after magnetic particle testing was very good as all the edges were detected (Figure 3). However, there was a lot of variation in the intensity of the image taken from the field (Figure 4) and it was not possible to detect all the edges as the noise from the image was high.

Figure 3: Edge detection algorithm applied to magnetic particle images

Figure 4: Edge detection algorithm applied to RIV images One of the main challenges of detecting RCF edges on a RIV image is that the edges are not connected themselves and a pre-process preparation with magnetic particle or dye penetrant method is required to highlight the cracks from the rail background. Considering this key practical constraint this methodology was not considered further.

Region of Interest

3.2. Ultrasonics testing The next NDT method under review included the testing of rail samples (Head Hardened and Standard Carbon) received from rail industry partners using the ultrasonic testing set up with 0, 35 and 70 degree probes. From the results presented in Table 2, it is evident that ultrasonics cannot reliably detect the head checking cracks. This was due to the fact that ultrasonic testing tool’s measurement range is beyond internal crack’s depth range.

Table 2: Ultrasonic testing of rail samples

Rail identification Sample

identification

Probe position

0° 35° 70°

QH

– H

ea

d h

ard

ene

d, H

igh r

ail

sam

ple

AH

– S

tand

ard

Ca

rbon.

Hig

h r

ail

sa

mple

QL –

He

ad h

ard

ene

d, lo

w r

ail

sam

ple

QH1 10 Picking up horizontal crack - No defect detected

QH1 8 No defect detected No defect detected No defect detected

QH1 17 No defect detected - -

AH1 31 & 32 - No defect detected No defect detected

AH2 29&30 - No defect detected No defect detected

AH3 26,27&28 - No defect detected No defect detected

QH2 34 TO 39 Could not get the back signal Could not get the signal No defect detected

QH3 40 TO 44 Could not get the back signal Could not get the back

signal No defect detected

QH4 45 TO 47 No defect detected No defect detected No defect detected

QH5 33&34 No defect detected No defect detected No defect detected

QL1 50 TO 52 No defect detected No defect detected No defect detected

3.3. Eddy Current Testing The last NDT method considered was the eddy current testing method which has been a well established procedure used in aviation, automotive and process industries. Currently, the eddy current testing application in rail inspection is limited [5]. Measuring the crack depth is difficult with the eddy current system because of the complexity involved in interpreting the signal from the eddy current instrument. Furthermore, it is also very sensitive and picks up any surface discontinuities; hence directly measuring the crack depth from an eddy current signal is difficult. Consequently, a novel method that overcomes these constraints and predicts the crack depth based on the eddy current signal is proposed in this research. The eddy current signal is influenced by:

Conductivity (greater the conductivity of the test material the greater the sensitivity to the surface discontinuities)

Permeability (it has an effect of masking the eddy current signals)

Frequency (the operating frequency is 5Hz to 10 MHz – the lower the frequency the better depth of penetration)

Coil design (higher penetration can be achieved by suitable selection of probe)

Lift off and edge effect (Lift-off reduces the field density and has significant impact on sensitivity)

Figure 5: A sweep display of rail- eddy current testing

Initial testing for RCF cracks using the eddy current instrument on rail showed that even though the conductivity and permeability are considered important factors which effect the sensitivity of the signal, these factors do not appear to affect the application of detecting the surface discontinuities (as

Crack effect

Effect of material properties Lift off

the change in impedance is much higher than the noises due to conductivity and permeability) (Figure 5).

Based on this initial finding, a test setup was developed interfacing the eddy current testing equipment with a data acquisition system, followed with post processing software to process the signal. A Nortec 500D eddy current testing system was used for this test. This instrument was interfaced with a data acquisition system, a hardware unit to store the eddy current signals, which was then interfaced with the Labview post processing software. The signal processing system, developed in a LabVIEW environment, can convert the analogue output from the instrument into a sinusoidal waveform, where the signal is post processed into relative signal voltages. These digital voltage values are the corresponding eddy current signal values. This setup was used to scan the defective rail and the corresponding voltage values of the eddy current signal was read directly and stored for future analysis.

3.3.1. Lift-off effect The lift-off effect and the phase angle for the rail material need to be understood in order to eliminate the lift-off effect during testing. Hence a series of tests were conducted on rail steel material using the micro hardness machine setup. The machine setup was just used to control the measurement distance during the lift-off measurements. The measurements were then repeated on the mild steel material. The eddy current signals corresponding to various lift-off distances were plotted in Figure 6. The lift off tests shows that variations in inductive reactance value (i.e. resistance to the eddy current flow) is negligible for lift offs of 0 to 0.1mm. Thus, the test setup tolerance should be maintained at a distance between 0 to 0.1mm.

A B

C

A –Mild steel (7 & 9mm cracks)B – Mild steel (3 & 5 mm cracks)C – Rail steel

Figure 6: Lift-off measurements for rail and mild steel. Figure 7: Calibration samples

3.3.2. Crack depth measurements using eddy current device Having identified the tolerance for lift-off, calibration samples (Figure 7) were then prepared with artificial crack depths ranging from 0.5mm to 9mm in different materials. The eddy current signals were measured using the proposed test system and the resulting voltage values were plotted against the corresponding known crack depth. Figure 8: Eddy current signals for mild steel. Figure 9: Eddy current signals for rail steel.

A B

0, -1.9

3, -0.7

5, 0.25

7, 0.55 9, 0.65

y = -0.0042x3 + 0.0264x2 + 0.3818x - 1.9094-2.5

-2

-1.5

-1

-0.5

0

0.5

1

0 2 4 6 8 10

Vo

ltag

e,v

olt

s

Crack depth,mm

Mild steel

y

Poly. (y)

Lift-off 0.1mm

Initial findings show that there exists a polynomial fit of 3rd

degree for the mild steel and a polynomial fit of 4

th degree for rail steel. This shows that the crack depth can be predicted using the same

polynomial relationship, which is essential to quantify the RCF cracks. However, more samples are required to be confident of this relationship statistically. The research project is currently preparing cut samples with different angles and different crack widths to validate and develop the relationship for rail steel. Figures 8 and 9 shows the voltage verses crack depth plot for mild steel and rail steel.

4. Conclusion

Rail grinding effort is currently performed by considering rail profiles and has traditionally ignored rail

surface condition. This profile based grinding often leads to excess grinding and a reduction of rail life.

Consequently, a new approach for rail grinding comprising of an EQI and economic decision model

for rail grinding cycle was proposed in this research project. The EQI incorporates the rail surface

condition associated with rolling contact fatigue (RCF). However, there is a lack of knowledge and

comprehensive technology to measure and quantify the RCF cracks and thus the surface condition.

Consequently, a review of three non-destructive testing tools: image processing, ultrasonic testing

and eddy current technology was conducted to determine the method best suited to measuring RCF.

The study revealed that image processing techniques of images from an RIV cannot be effectively

processed to detect the RCF condition. Pre-process preparation by magnetic particle or dye penetrant

method is required to highlight the cracks from the rail background, which provides a major practical

constraint for image processing. Whereas the ultrasonic test could not detect RCF defects due to its

detection range being outside RCF crack depth level. Finally, a review of an eddy current tool showed

positive results. Consequently, a test set-up was developed to measure the eddy current signal

corresponding to the cracks. The lift-off effect and corresponding tolerance was also studied and

identified. In addition, a data acquisition system was developed with post processing software to

overcome the practical limitations, such as sensitiveness of the eddy current device and unwanted

noise signals. This preliminary study confirmed that the crack depth can be predicted using a

polynomial relationship. However more samples are required to be confident of this relationship

statistically. The research project is currently preparing cut samples with different angles and different

crack widths to validate and develop the relationship for rail steel. The Shuohuang railway network in

China also has operating conditions similar to this CRC project in terms of tonnage and rail

characteristics. From their field study a preventive grinding interval of 25 to 30 MGT with a metal

removal of 0.2 – 0.3mm was established to extend the rail life. The CRC project also established that

using particular template on sharper curves the grinding interval can be extended to a range from 25

to 30 MGT compared to the current grinding interval of 15 MGT. Therefore it is evident that the RCF

quantification method proposed in this CRC project can be extended to Shuohuang railway network

and other similar network in other countries.

Acknowledgements

The authors are grateful to the CRC for Rail Innovation (established and supported under the

Australian Government's Cooperative Research Centres program) for the funding of this research

Project No. R3.109. Project Title: “Rail Grinding Best Practices”. The authors also acknowledge the

support of the Centre for Railway Engineering, Central Queensland University and the industry

partners that have contributed to this project.

Reference

1. Australian Rail Transport Facts- 2008. 2008, Apelbaum Consulting Group Pvt. Ltd 2. Uddin, M.G., G. Chattopadhyay, R. Moahmmad, and P. Sroba. Friction Management Best

Practices for Australian Heavy Haul Lines. in International Heavy Haul Association Conference. 2011. Canada: International Heavy Haul Association.

3. Uddin, M.G., G. Chattopadhyay, P. Sroba, R. Moahmmad, and A. Howie. Wayside lubricator placement model for heavy haul lines in Australia. in Conference on Railway Engineering (CORE 2010). 2010. Wellington, NZ.

4. F.J. Franklin, G.-J. Weeda, A. Kapoor, and E.J.M. Hiensch, Rolling contact fatigue and wear behaviour of the infrastar two-material rail. Wear, 2005. 258(7-8): p. 1048-1054.

5. Rolling Contact Fatigue: A Comprehensive Review. 2011, US Department of Transport. 6. Cannon, D., K. Edel, S. Grassie, and K. Sawley, Rail Defects: an overview. Fatigue and Fracture

of Engineering Materials and Structures, 2003. 26: p. 865-887. 7. Sroba, P. and M. Roney. Rail Grinding Best Practices. in AREMA. 2003. 8. Chattopadhyay, G., V. Reddy, and P. Larsson, Decision on Economical Rail Grinding Interval

for Controlling Rolling Contact Fatigue. International Transactions in Operational Research, 2005. 12(6): p. 545-558.

9. Devadoss, R., G. Chattopadhyay, N.K. Mandal, P. Sroba, M. Turner, and G. Creese. Rail quality grinding assurance based in profile correction and RCF control. in Conference on Railway Engineering (CORE 2010). 2010. Wellington, New Zeland.

10. Devadoss, R., G. Chattopadhyay, and P. Sroba. Economic modelling for preventive grinding to support grinding decisions for heavy haul lines. in 27th International Conference COMADEM 2011. 2011. UK.

11. Guidelines to Best Practicves for Heavy Haul Railroad Operations: Wheel and Rail Interface Issues. in International Heavy Haul Association. 2001.

12. Palese, P., T. Euston, and A. Zarembski. Use of Profile Indices for Quality Control of Grinding. in Annual Conference and Exposition, AREMA. 2004. Tennessee.