Structural Health Monitoring of Railroad Wheels Using Wheel Impact Load Detectors Brant Stratman, Yongming Liu, Sankaran Mahadevan* Vanderbilt University, Nashville, TN 37235, USA Abstract This paper proposes two quantitative criteria for removing railroad wheels from service, based on real-time structural health monitoring trends that are developed using data collected from trains while in-service. The data is collected using Wheel Impact Load Detectors (WILDs). These impact load trends are able to distinguish wheels with a high probability of failure from high impact wheels with a low probability of failure. The trends indicate the critical wheels that actually need to be removed, while at the same time allowing wheels that aren’t critical to remain in service. As a result, the safety of the railroad will be much improved by being able to identify and remove wheels that have high likelihood of causing catastrophic failures. 1 Introduction Traditional inspection techniques used in the railroad industry, such as drive-by inspections where all of the wheels on the train are glanced at while an inspection vehicle drives by, are not as accurate and reliable as more rigorous and quantitative inspection methods. Many damaged wheels aren’t found, while many useable wheels are removed when they could remain in service. By using Wheel Impact Load Detectors (WILDs), structural health monitoring trends can be developed based on the wheel impact data which indicates the actual condition of the wheels. The trends can indicate the critical * Corresponding author, Tel.: 615-322-3040; Fax: 615-322-3365; Email:[email protected]
22
Embed
Structural Health Monitoring of Railroad Wheelsyongming.faculty.asu.edu/paper/Microsoft Word - Structural Health... · Structural Health Monitoring of Railroad Wheels Using Wheel
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Structural Health Monitoring of Railroad Wheels Using Wheel Impact Load Detectors
Brant Stratman, Yongming Liu, Sankaran Mahadevan*
Vanderbilt University, Nashville, TN 37235, USA
Abstract
This paper proposes two quantitative criteria for removing railroad wheels from
service, based on real-time structural health monitoring trends that are developed using
data collected from trains while in-service. The data is collected using Wheel Impact
Load Detectors (WILDs). These impact load trends are able to distinguish wheels with a
high probability of failure from high impact wheels with a low probability of failure. The
trends indicate the critical wheels that actually need to be removed, while at the same
time allowing wheels that aren’t critical to remain in service. As a result, the safety of
the railroad will be much improved by being able to identify and remove wheels that
have high likelihood of causing catastrophic failures.
1 Introduction
Traditional inspection techniques used in the railroad industry, such as drive-by
inspections where all of the wheels on the train are glanced at while an inspection vehicle
drives by, are not as accurate and reliable as more rigorous and quantitative inspection
methods. Many damaged wheels aren’t found, while many useable wheels are removed
when they could remain in service. By using Wheel Impact Load Detectors (WILDs),
structural health monitoring trends can be developed based on the wheel impact data
which indicates the actual condition of the wheels. The trends can indicate the critical
From Table 2 it is seen that the mileage analysis agrees with the time analysis;
broken wheels’ impacts increase much more rapidly than high impact wheels. It is also
important to observe in Table 2 that high impact wheels and broken wheels, on average,
travel relatively the same distance per year. It is therefore concluded in this study that
since high impact wheels and broken wheels travel the same distance in the same amount
of time on average, then using time intervals is appropriate for developing the structural
health monitoring trends.
5 Structural Health Monitoring Trends Using Wheel Impact Load Detectors
Two trends are developed that could be used to monitor wheels in real-time. The
two trends identify wheels that suddenly get worse, thus providing a way to differentiate
wheels with a high probability of failure from high impact wheels. In general, high
impact wheels appear to degrade in a gradual manner, as seen in Figure 6.
14
From Figure 6 it is seen that wheels generally have an up-down-up impact trend,
which is caused by the train’s loaded/empty cycle for moving cargo. This means that
when a train is carrying cargo the impact is higher than when the train is empty, hence the
up-down-up impact trend. When a defect in a wheel starts to grow, the authors found that
the impacts will increase regardless of whether or not a train is carrying cargo. These
defect growth trends can be seen in Figures 7 and 8, when consecutive impacts increase
rather than following the up-down-up trend.
Figure 6 High impact wheels generally increase impact readings slowly over time.
5.1 First Structural Health Monitoring Trend
The first trend developed is designed to capture wheels that were running at a
high dynamic impact but increased greatly over a short period of time. This trend
differentiates wheels with a high probability of failure from high impact wheels since
0
10
20
30
40
50
60
70
4/2
0/2
003
6/9
/2003
7/2
9/2
003
9/1
7/2
003
11/6
/2003
12/2
6/2
003
2/1
4/2
004
4/4
/2004
5/2
4/2
004
Time
Dyn
am
ic (
kip
s)
High Impact Wheel Dynamic
Mate Wheel Dynamic
15
high impact wheels’ impacts increase slowly over time. Generally all of the wheels
caught by the first trend have peak impacts greater than 90,000 pounds so they are
already condemnable by AAR standards.
The first wheel removal criterion is found to be as follows, based on analysis of
the data. The wheel needs to have three consecutive increasing impact readings, each
reading with an impact increase of 15 kips or more than the previous reading, and all
three readings need to occur in a period of less than 50 days with the peak reading having
a dynamic impact greater than 70 kips. An example of the first trend is given in Figure 7.
Of the failed wheels included in the database, 10.53% (14.28% of Vertical Split
Rims, and 8.33% of Shattered Rims) had impact histories following the trend discussed
above. This means that 10.53% of wheel failures would have been prevented by the first
structural health monitoring criterion had it been implemented at the time of the wheels’
service.
It is estimated from the data that approximately 200 wheels per year in North
America are captured by this wheel removal criterion. Examining data from previous
years concludes that of the wheels that would have be caught by the first criterion had it
been implemented, 78.2% were removed for another reason (i.e. high impact wheel as
detected by wheel impact detector, visual inspections, etc.) after the criterion would have
detected them. This means that because the criterion was not in place 156 damaged
wheels were left in service and caused additional damage to the rail and train
components. According to the data, the damaged wheels would have been removed 31.7
days earlier on average had the first criterion been implemented at the time of their
16
operation, which in effect would prevent 31.7 days of additional damage from those
individual wheels.
Implementing this criterion would increase the number of removed wheels by
approximately only 44 wheels/year; this equals the 200 wheels/year that are detected by
the criterion minus the 156 wheels/year that would be removed regardless of whether the
criterion was in place or not.
Figure 7 An example of the first trend showing rapid increase in impact readings.
The number of shattered rims has been estimated to be around 700 wheels per
year in North America12; implementing the proposed criterion would prevent 8.33% of
those. This means approximately 58 shattered rim failures per year will be prevented by
increasing the number of wheels removed per year by only 44 wheels. As discussed
above, actually the trend detects approximately 200 wheels/year, approximately 58 of
0
10
20
30
40
50
60
70
80
90
100
110
3/1
2/2
00
3
4/1
2/2
00
3
5/1
2/2
00
3
6/1
2/2
00
3
7/1
2/2
00
3
8/1
2/2
00
3
9/1
2/2
00
3
10
/12
/20
03
11
/12
/20
03
12
/12
/20
03
1/1
2/2
00
4
2/1
2/2
00
4
3/1
2/2
00
4
4/1
2/2
00
4
5/1
2/2
00
4
Time
Dy
na
mic
(k
ips
)
Failed Wheel Dynamic
Mate Wheel Dynamic
17
which are shattered rims, but because some of the shattered rims wouldn’t have been
detected until after they failed the proposed criterion actually prevents more shattered
rims than additional wheels it would require to be removed. It is important to note that
included in the 200 wheels detected by the trend each year are a significant number of
vertical split rim failures that will also be prevented.
The average cost of a wheelset (including new and refurbished wheelsets) is
approximately $830. Therefore, the additional 44 wheelsets that would be removed by
this tread would cost North American railroads approximately $36,520. Considering the
damage these failed wheels cause to the rail and train components, and their high
probability of causing derailments, implementing this trend would be financially
beneficial and at the same time increase the safety of the railroad.
5.2 Second Structural Health Monitoring Trend The second trend is developed to capture wheels that are running in a normal
impact reading range, and suddenly increase in a very short period of time. Wheels
following the second trend have a dynamic impact history that looks nothing like the
dynamic impact history of a high impact wheel.
The second criterion for wheel removal is found to be as follows, based on
analysis of the data. The wheel needs to have three consecutive increasing impact
readings, each reading with an impact increase of 2 kips or more than the previous
reading with an average increase of 10 kips or more for the three readings, and all three
readings need to occur in a period of less than or equal to 20 days with the peak reading
have a dynamic impact greater than 40 kips, no reading 30 days prior the initial impact
18
deviation reading can have an impact reading greater than 5 kips or less than 15 kips from
the initial deviation dynamic value. An example of the second trend is given in Figure 8.
Of the failed wheels included in the database, 5.26% (8.33% of Shattered Rims)
had impact histories following the trend discussed above. This means that 5.26% of
wheel failures would have been prevented by the second structural health monitoring
criterion had it been implemented at the time of the wheels’ service.
Figure 8 An example of the second trend showing sudden increase in impact readings.
It is estimated from the data that approximately 1400 wheels per year in North
America are captured by this wheel removal criterion. Examining data from previous
years concludes that of the wheels that would have be caught by the second criterion had
it been implemented, 47.5% were removed for another reason after the criterion would
have detected them. This means that because the criterion was not in place 665 damaged
0
10
20
30
40
50
3/1
5/2
003
4/1
5/2
003
5/1
5/2
003
6/1
5/2
003
7/1
5/2
003
8/1
5/2
003
9/1
5/2
003
10/1
5/2
003
11/1
5/2
003
12/1
5/2
003
Time
Dynam
ic (kip
s)
Failed Wheel Dynamic
Mate Wheel Dynamic
19
wheels were left in service and caused additional damage to the rail and train
components. According to the data, the damaged wheels would have been removed 66.6
days earlier on average had the second criterion been implemented at the time of their
operation, which in effect would prevent 66.6 days of additional damage from those
individual wheels. Implementing this criterion would increase the number of removed
wheels by approximately 735 wheels/year, thereby preventing damage from these wheels
to the train and rail components.
The average cost of a wheelset (including new and refurbished wheelsets) is
approximately $830, and the average cost of a derailment is approximately $340,000.
This means only two derailment preventions per year from the 1400 wheels following
this trend would cover the cost of the additional 735 removed wheelsets detected by this
criterion, and at the same time would increase the safety of the railroad. In addition,
when the additional damage to the rail and train that is prevented from this criterion is
included in the cost analysis it is clear that this structural health monitoring trend
provides a cost efficient criterion for removing wheels.
From the above discussions it is clear that both criteria are cost efficient methods
for removing wheels. Even so, when implementing the two criteria together the cost
efficiency is further increased; a very small percentage (0.57%) of the wheelsets are
identified by both criteria. The combined implementation of the two criteria will result in
approximately 775 removed wheelsets per year. This slight overlap in wheel detection
reduces the increase in removals by 4 wheelsets/year; making the two trends, when
implemented together, an even more cost efficient method for removing wheels.
20
6 Conclusion
Traditionally, wheels are removed based on drive-by visual inspection and the
90,000 pound high impact condemning limit. Unfortunately many damaged wheels are
not found with the drive-by inspection method, while many useable wheels are removed
when they could remain in service. The structural health monitoring trends developed in
this paper provide a quantitative decision method based on the wheel impact data which
indicates the actual condition of the wheels. Therefore the trends indicate the critical
wheels that actually need to be removed, while at the same time allowing wheels that
aren’t in critical condition to remain in service.
The two wheel removal criteria proposed in this paper will remove 15.8%
(16.67% of Shattered Rims, and 14.28% of Vertical Split Rims) of wheels with a high
probability of failing by indicating that the wheels are about to fail before they actually
fail. The two criteria will increase the number of wheels removed, for the accumulation
of railroads in North America, by approximately 775 wheelsets/year. Considering the
damage these failed wheels cause to the rail and train components, and their high
probability of causing derailments, implementing these criteria would be financially
beneficial while at the same time increase the safety of the railroad.
7 Acknowledgments
The research reported in this paper was supported by funds from Union Pacific
Railroad and Meridian Railroad (Research Agreement No. 18140, Monitors: Rex Beck
and Todd Snyder). The support is gratefully acknowledged.
21
References 1 Lee, M.L. and Chiu W.K. (2005). A comparative study on impact force prediction on a railway track-like structure. Structural Health Monitoring, Vol. 4, No. 4, 355-376. 2 Gustavson, R. (2000). Static and dynamic finite element analysis of concrete sleepers. Licentiate of Engineering Thesis, Chalmers University of Technology, Division of Concrete Structures, Göteborg. 3 Union Pacific’s Lab Database. 4 Tanaka, T., Kinoshita, K. and Nakayama, H. (1992). Effect of loading time on high cycle range impact fatigue strength and impact fatigue crack growth. JSME International
Journal, Series 1, 35(1), 108-116. 5 Cheng, Y., Chen, D. and Nogata, F. (1994). Fatigue behaviour of a rail steel under low and high loading rates. Fatigue and Fracture of Engineering Materials and Structures, 17(1), 113-118. 6 Association of American Railroads, 2005 Field Manual of the AAR Interchange Rules, Rule 41, AAR, Washington, DC. 7 S. Kalay and A. Tajaddini, “Condemning Wheels Due to Impact Loads: Preliminary Survey – Six Railroads Experience,” AAR Report R-754, February, 1990. 8 Barke D, Chiu WK. Structural Health Monitoring in the Railway Industry: A Review. Structural Health Monitoring 2005; 4(1):81-93. 9 Johansson, A. and Nielsen, J. (2001). Railway wheel out-of-roundness – Influence of wheel-rail contact forces and track response. 13
th International Wheelset Conference,
Rome. 10 Kalay, S., Tajaddini, A. and Stone, D.H. (1992). Detecting wheel tread surface anomalies. Rail Transportation – 1992 American Society of Mechanical Engineers, Rail Transportation Division (Publication) RTD, ASME, New York, NY, USA: 165-174. 11 Partington, W. (1993). Wheel impact load monitoring. Proceedings of the Institution of
Civil Engineers – Transport, 100(4): 243-245. 12 Lonsdale, C., Pilch, J. and Dedmon S. (2002). Stress effects of wheel impact loads. –unknown-. 13 Bladon, K. (2003). Barke, D.W. (ed.), Function of Wayside Wheel Impact Monitors, Adelaide, Australia. 14 Tournay, H.M. and Mulder, J.M., “The transition from the wear to the stress regime”, Wear, Vol.191, pp.107-112, 1996. 15 Stone, D.H. and Moyar G.J., “Wheel shelling and spalling – an interpretive review”, in Rail Transportation 1989, ASME, pp.19-31, 1989. 16 Marais, J.J, “Wheel failures on heavy haul freight wheels due to subsurface effects”, Proc 12
th International Wheelset Congress, Qingdao, China, pp. 306-314, 1998.
17 Mutton, P.J., Epp, C.J. and Dudek, J., “Rolling contact fatigue in railway wheels under high axle loads”, Wear, Vol. 144, pp. 139-152, 1991. 18 Giammarise, A. W., and Gilmore, R. S. “Wheel Quality: A North American Locomotive Builder’s Perspective”, GE Research & Development Center, CRD140, Sep, 2001. 19 Stone, D.H., Majumder, G. and Bowaj, V.S. (2002). Shattered rim wheel defects and the effect of lateral loads and brake heating on their growth. ASME International
22
Mechanical Engineering Congress & Exposition, Nov 17-22 2002, New Orleans, Louisiana: 1-4. 20 Ekberg, A., E. Kabo, and H. Andersson., “An engineering model for prediction of rolling contact fatigue of railway wheels”, Fatigue & Fracture of Engineering Materials
and Structures, Vol. 25, pp. 899-909, 2002. 21 Stone, D.H. (2000). Wheel shattered rims – and interpretive review. Wheels and Axles,
Cost Effective Engineering. ImechE Seminar Publication: 75-84. 22 Beretta, S., Roberti, A., Ghidini, A. and Donzella G. Deep shelling in railway wheels. Proceedings, 13th International Wheelset Congress, Rome, Sept 2001, Paper 23. 23 Stone, D.H., Kalay, S.F., and Lonsdale, C.P. Effect of wheel impact loading on shattered rims. Proceedings, 13th International Wheelset Congress, Rome, Sept 2001, Paper 2. 24 Berge, S., “Shattered Rim Fracture Research”, Proceedings of 2000 Brenco Rail Conference, LaQuinta, California, October 19-20, 2000. 25 Stone, D.H., and Geoffrey, E.D., “The Effect of Discontinuity Size on the Initiation of Shattered Rim Defects,” ASME Transportation Division- Vol. 19, pp. 7-14, ASME Spring 2000. 26 Gordon, J., and Perlman, A. B., “Estimation of Residual Stresses in Railroad Commuter Car Wheels Following Manufacture”, Proceedings, International Mechanical
Engineering Congress and Exhibition, ASME RTD Vol. 15, pp.13 – 18, 1998.