Sie sind auf Seite 1von 45

o 5-1 x

Part 5: Maintaining Optimal Wheel and Rail


Performance
Written by Mr. Michael D. Roney, member of the IHHA
Board of Directors and Professor Willem Ebershn, member
of the Technical Review Committee.
5.1 Maintaining Optimal Wheel and Rail
Performance
Rail is the single most expensive element of the track structure.
On many railways, it is behind only labor and fuel as an
expense item. The tonnage carried by a rail before it is
condemned can range from less than 100 million gross tons to
close to 2.5 gigga gross tons.
As an example of the value of rail maintenance
management, assume that a single kilometre of rail costs
$180,000 to install. Track engineers decide that the rail has a
badly fatigued surface and has reached the end of its service
life. They call for it to be replaced, gaining a salvage value of
$18,000.
But now assume that instead of replacing the rail, they did
some corrective rail grinding costing $1800 and left the rail in
track. The railway then invested the $180,000 $18,000 -
$1,800 = $160,200 in the construction of a new customer
facility at a rate of return of 20%. This earned $160,000 * 20%
= $32,000 in its first year.
The next year, the track engineers see that their rail is
approaching allowable wear limits and schedules a rail
replacement, now costing $187,200 due to cost escalation of
4%. But they have made for the railway $32,000 ($187,200
180,000) = $24,800 by deferring replacement of rail in that
kilometre, without consequence, for an extra year. And that is
why they collect a salary.
There is significant money to be made by deferring rail
replacement as much as possible without incurring risk.
Certainly it is a major responsibility of the track engineer to
ensure that he gets the most out of his rail, and rail profile
maintenance and rail testing are his most important tools to do
this.
Click Here To Go Back To Table of Contents
o 5-2 x
The catch is that the above statements assume that the
aging rail does not have a direct impact upon other expenses,
such as wheel wear or fuel consumption. Furthermore, there
must not be a significant increase in risk of rail fracture.
To optimize rail and wheel life cycle, it is imperative that
these two components are managed jointly, as it is the
wheel/rail interaction that determines the performance of each
component.
In the case of wheel and rail, rail maintenance holds the
key, as rail is static in its location along the track layout and
more accessible for maintenance. Although the wheel
requirements of a vehicle is more variable in the sense that it
moves over a variety of rail layouts and conditions, its
requirements must match those of the rail to ensure optimal
wheel rail interaction.
The joint strategies of rail re-profiling, friction
management (lubrication), control of gauge, and rail condition
measuring can protect rail-caused cost impacts.
Rail re-Profiling: Rail and wheel life is reduced if the
wheel/rail contact conditions deteriorate through migration of
the contact band or flattening of the rail. Regular metal
removal can also control surface fatigue, which is ultimately
associated with internal rail defects. Both are controllable
through regular re-profiling.
Friction Management: Maintaining the rail surface friction
within specified limits on the rail gauge face, but also on the
top of rail, can reduce rail and wheel wear and improve fuel
consumption.
Gauge Control: As the gauge face wears, widening gauge
will change wheel/rail contact locations, which can accelerate
wear and fatigue. If the rail is allowed to spread (rollover)
under dynamic gauge widening, rail contact fatigue can result.
Lubrication and attention to fastener condition act together to
control these cost impacts.
o 5-3 x
Rail Condition Measuring: Rail fatigue under high axle loads
combined with the rail section reductions through wear and
grinding is an ideal environment for condition based
maintenance practice with the measuring devices available
today. The rate of occurrence of some internal rail defects
increases and can be detected with a rail defect-measuring plan.
The cost impact does not need to be high, however, as long as
such occurrences are detected, rail defects can be changed out
in a production fashion behind the rail testing operation.
Technology advances have made rail profile measuring
accurate and repeatable to the extent that rail wear rate can be
reliably determined and used for maintenance planning.
In the same manner, wheel profile and to a certain extent
defect detection technology is also at the point where
maintenance can move from a routine based maintenance
practice to a condition based maintenance practice.
Unfortunately, wheel risk management and maintenance
practices are not at the same level as rail risk management. As
an example, one heavy haul railway reports that for the year
2000 track related failures accounted for 23% of all accidents,
where wheel failures represented 5% of the accidents. But
wheel related accidents are characteristically more severe and is
reflected by the cost of these accidents as being 31% of the
total accident costs. (FRA reported Safety Statistics shows a
ratio of 36% track vs. 3% wheel related accidents with 11% of
the track-related derailments due to broken rails
1
).
Considering the high cost of wheel related derailments, it
seems that there is a discrepancy in our management of failure
risk of rails versus wheels. As typically indicated, a wheel that
lasts 5 years on a 30 mgt line would likely be tested and
profiled 2 times, whereas a rail in the same period would likely
be re-profiled and ultrasonically tested 16 times.
Nevertheless, over the past decade, railways have increased
their control of wheel/rail contact, although not always in an
integrated manner, between the wheel shops and the track
maintenance teams.
If rail sees frequent maintenance attention, as discussed
above, how long can it stay in track? Depending on operating
o 5-4 x
conditions, costs can certainly be minimized when the rail is
ultimately removed for loss of railhead, as opposed to any
other cause such as fatigue. This gives rise to seriously
rethinking the maximum allowable rail wear limits.
Whereas rail limits in the past have been associated with
the ultimate contact of a high wheel flange with the joint bar,
continuous welded rail has virtually eliminated this barrier.
Another concern in the past, particularly for railways with 30-
ton and greater axle loads and/or standard carbon rails, has
been that the rail would be misshapen by the time such limits
were being approached, as a result of plastic flow that the
whole track structure would be suffering from the poor vehicle
tracking.
In an environment where dynamic wheel/rail interaction
and profiles are controlled, railway engineers can start to work
to get the full life potential from the yield strength of the rail
section.
The achievement of full utilization of rail generally requires
a deliberate plan to put the following supporting strategies in
place:
1. Develop target rail profiles that are seen to achieve
low fatigue and wear. Because of the heavy influence
of contact fatigue on rail, these target profiles would
typically incorporate some conformity with wheel
profiles. A mechanical representative should be on
the team to advise on implications for wheels, and to
explore the potential for joint optimization of wheel
and rail profiles. This part is addressed in previous
chapters.
2. Measure rail and wheel conditions to determine
maintenance needs.
3. Develop rail and wheel wear projection methodology.
4. Develop rail and wheel fatigue life projection
methodology.
5. Perform economic evaluation of different premium
rail options based upon condition monitoring and
execute plans to progressively balance rail strength
o 5-5 x
with service environment. Determine the use of
premium and intermediate rail steels in different
tonnage and curvature classes based on minimum life
cycle costs. Perform analyses to determine
transposition and re-use policies.
6. Perform extensive rail re-profiling to the target profile
and to correct existing rail spalling, corrugations, and
head checks.
7. Re-profile wheels to target profile.
8. Install lubricators in locations with past history of
higher gauge face wear rates.
9. Develop new rail wear limits and supply new design of
joint bar to support extended wear limits in secondary
lines.
10. Develop new wheel wear limits.
11. Implement regular and frequent ultrasonic rail
inspection as well as regular rail wear and profile
condition measurement. Quality assessment is an
important part of this strategy. Correlate rail
deterioration with track geometry condition and gauge
widening to develop joint strategies.
12. Implement regular and frequent wheel flaw inspection
as well as regular wheel profile condition
measurement.
13. Implement frequent maintenance rail re-profiling
(grinding) on regular cycles. The objective should be
to move to single pass rail grinding, with speeds
adjusted to grind as fast as possible to control rail flow
and fatigue occurring between grinding cycles.
14. Adjust profiling standards and rail-testing intervals to
match needs of rails approaching extended wear limits.
15. Implement a condition based maintenance plan for
wheel re-profiling.
Not all of these steps have to be in place to achieve major
savings in costs. It is suggested, however, that quantum
improvements in rail and wheel life require some attention to
each of the steps. Critical to the implementation of this
strategy is that all people involved clearly understand both the
o 5-6 x
end objective and the role of each of the elements. This will
undoubtedly require some education of track managers
through senior engineering managers of such aspects as role of
rail and wheel profile designs, fatigue mechanisms, and wear
rates.
5.2 Rail Structural Deterioration
5.2.1 Management of Rail Testing to Control Risk of
Rail Fracture
The occurrence of internal defects in rails is an inevitable
consequence of the accumulation of fatigue under repeated
loading. To maximize rail life, heavy haul railways live with
controlled rates of defect occurrences, relying on regular
ultrasonic or induction rail testing and strategic renewal of rail
that is obviously showing evidence of fatigue.
The consequences of internal flaws can be serious. An
unrecognized defect can result in rail breakage with
interruption of service and the potential risk of catastrophic
consequences. At the least, an isolated case means a repair
cost and introduces two unwanted welds, while a series of
defects can condemn a whole rail length.
FRA Reported Safety Statistics for 1999
1
shows 11% of the
accidents were rail and joint bar related. Figure 5.1 shows the
distribution of accidents for rail and joint bar defect types for
1999.
Train Accidents Reported for Rail and Joint Bar Defect Type (%)
1%
1%
1%
2%
3%
3%
3%
4%
4%
12%
13%
15%
17%
21%
Joint Bolts Broken or Missing
Bolt Hole Crack or Break
Mismatched Rail-head Contour
Head and Web Seperation (inside joint)
Worn Rail
Joint Bar Broken
Horizontal Split Head
Broken Field Weld
Other Rail and Joint Defects
Detaied Fracture from Shelling
Head and Web Seperation (outside joint)
Verrtical Split Head
Broken Base of Rail
Transverse/compound Fissure
Figure 5.1: Distribution of Accidents Per Rail and
Joint Bar Type
o 5-7 x
The challenge is to avoid the occurrence of service failures
due to undetected defects. Service failures are more expensive
to repair and can lead to costly line disruptions or even
derailments. The role of rail defect testing is therefore to
protect service reliability while avoiding overly conservative rail
renewals. To illustrate the scope of rail testings contribution,
consider that North American heavy haul railways detect an
average of 0.4 defects per track km. (0.6 rail defects/mi) each
year while inspecting at intervals of 18 mgt (20 mgt) and
experience 0.06 service failures/km (0.1 service failures/mi.).
One service defect in two hundred leads to a broken rail
derailment. Rails are typically replaced when total defects are
occurring at a sustained rate of 1-2/rail km (2-3/mi.).
2
In controlling risk, the most basic control variable is the
test interval. A railway administration must decide upon the
frequency of passage of the rail testing equipment that will
balance the cost of testing and rail change-out with the
expected derailment cost to minimize the net cost of the risk.
In this delicate equation, the reliability and operating speed of
the testing system play an important role.
5.2.2 The Framework for Risk Management
The practice of rail testing has a simple objective of reducing
the annual costs incurred as a result of broken rails. But there
are many variables involved. Figure 5.2 shows the most
important of these.
o 5-8 x
Figure 5.2: Factors Controlling the Risk of
Broken Rail Derailments
The direct cost of undetected rail breaks is the difference
between the cost of replacing broken rails on an emergency
basis, and the cost of the orderly replacement of detected
defects. The cost of derailments caused by undetected broken
rails is an indirect cost of poor inspection reliability. The
derailment cost is the annualized cost of the rare but high cost
occurrence of a derailment. As the probability would be
derived statistically from past records, the annualized
derailment cost is also called the expected derailment cost.
This cost is also related to specific characteristics of the railway
o 5-9 x
such as the remoteness of the routing, the severity of the
terrain, the type of lading, and the size and speed of the trains.
The number of service failures is intimately related to the
effectiveness of the inspection. In a risk management
approach, high inspection reliability is required where long
trains are travelling fast along a watercourse in proximity to
population centers. While heavy haul lines typically have long
trains and high derailment costs, train frequencies may be less
than on mixed freight lines, defect growth rates may be more
uniform, and tonnage is easier to track for the purpose of
planning test intervals. Many heavy haul operators have
dedicated rail-testing vehicles, which they may use at frequent
intervals, even monthly.
To ensure an effective rail testing program, the test
equipment must be properly designed and calibrated to reliably
indicate defects, the equipment logic must present to the
operator only those indications that could be a rail defect, and
the operator must be experienced and diligent.
In addition, test frequencies must be matched to the
growth rate of critical defects so that at least one test, and
preferably more inspections, are made in the interval between
the development of a rail defect to a minimum detectable size
and its growth, to a size that represents a significant rate of
rapid fracture.
In practice, the growth rate of rail defects is both highly
variable and rarely known with any certainty. Rail testing on
heavy haul operations often presents some specific problems,
for as traffic loading is high, defect growth is accelerated and
the time scale for intervention is compressed. The tendency
for each wheel passage to stress the rail in a similar pattern can
increase defect growth rates. At the same time, heavy axle
loads can lead to a fatigued rail surface that may present
confusing indications from testing equipment. The use of rail
of various metallurgical qualities further complicates the task.
Most heavy haul operators attempt to control risk by
monitoring of the reliability of the test through evaluation of
failures occurring soon after testing and by comparing ratios of
service to detected rail defects.
o 5-10 x
5.2.3 Defect Occurrence Rates
The driving factor determining the risk of rail fracture is the
rate at which a population of internal flaws develops in the rail.
Internal flaws in rail have a period to initiation and a period
during which a crack will propagate. The risk is introduced
when cracks remain undetected during their growth to critical
sizes. This occurs when the period between the times the
crack reaches detectable size is significantly shorter than the
testing interval. Figure 5.3 presents the example of a rail flaw
with a long period of exposure before failure and one with a
short exposure time. An example of a long exposure time
might be a transverse fissure, which is detectable at a small size
due to its central location in the head, and which may grow
slowly. At the other extreme might be a defective weld with
poor fusion in the web area. The web cracks would typically
be large at first detection and could be expected to propagate
rapidly.
In revenue service, rail in a given routing would be
expected to have a broad population of defects of different
sizes, each growing at a different rate.
In a typical heavy haul line, the population of flaws of
different sizes can be assumed to be distributed according to
an exponential distribution (Figure 5.4), where there are many
very small flaws, but very few large defects. As fatigue cycles
accumulate in the rail from high contact stresses, more flaws
are initiated and those already present continue to grow.
Critical to the risk of a rail break is the number of defects in
the right tail of the distribution. The area of the right tail of
the distribution would represent the number of rail flaws that
are of sufficient size that they could fracture suddenly. It is a
fact that the distribution of internal defects by size varies from
location to location. A highly stressed track segment or one
laid with a dirtier steel, will have a population distribution
shifted to the right and should, in theory, be tested more
frequently to achieve the same risk level.
o 5-11 x
Figure 5.3: Size Distribution of Flaws in
Rails in a Typical Line
Figure 5.4: Transverse Defect Growth Rates
Measured under 39-Ton Axle Loads at
FAST/Heavy Tonnage Loop
Because rail flaw detection is quite effective at detecting
large defect sizes, the distribution of defects in track at any one
time is skewed to the smaller sizes. The critical objective of rail
testing programs is to both eliminate the right tail high risk
defects in the distribution at the time of testing while
o 5-12 x
attempting to detect all defects from the distribution which,
through growth, will have reached the high risk level by the
time of the next rail test.
Hence both testing reliability and test intervals are
important. But most importantly, test reliability and testing
intervals must be matched.
Presently, there is little hard evidence on either the growth
rates of different defects or their critical sizes. One notable
exception is the defect growth relationship determined in
studies by the Transportation Technology Center, Inc. (TTCI)
a subsidary of the Association of American Railroads (AAR) at
the Facility for Accelerated Service Testing Facility (FAST),
Pueblo, Colorado USA. By monitoring the growth of
transverse defects under the controlled conditions of a unit
train cycling over the test loop, TTCI measured a wide range
of different growth rates.
Figure 5.5 plots the progression of transverse defects that
developed under a consist with 35 ton (39 Ton) axle loads.
3,4
The tonnage required to initiate a defect was found to be very
difficult to predict, but once initiated, transverse defects were
found to grow non-linearly with tonnage, as would be
predicted from fracture mechanics theory. Under the uniform
heavy loading conditions of the FAST consist, some defect
growth rates were found to be quite rapid. Rapid growth rates
could also be expected where tensile residual stresses are
present in the railhead, and in low temperatures in continuous
welded rail where the rail is again in tension.
Part 3, Appendix A presents Canadian Pacific rail system
criteria for the protection of defective rail in track.
o 5-13 x
Figure 5.5: Transverse Defect Growth Rates Measured Under
39-Ton Axle Loads in Fast Heavy Tonnage Loop
5.2.4 Critical Defect Sizes
Experience has shown that rail can fracture suddenly from
transverse defects as small as 10% of the railhead. Generally,
risk is significant when a transverse defect is larger than 35%
of the head. A bolt hole crack is known to start to grow
rapidly when the length exceeds about 13 mm (in.), and rapid
fracture can usually be anticipated from a 25 mm (1 in.) crack.
In general, railways have relied upon experience to distinguish
between fractures which present a substantial risk and those
which may safely remain in track for a specified period of time.
For example, Part 3, Appendix A: Canadian Pacific Rail
Defects
5
presents the mandatory Protection Codes imposed by
Canadian Pacific Rail System on trains passing over detected
rail defects prior to their removal from track. This table
presents one heavy haul railways assessment of the risk
associated with different sizes and types of defects.
A study of dynamic fracture of rails conducted at Queens
University at Kingston, Canada,
6
has shed some more light on
the dynamic load capacity of rails, and hence the risk of
fracture under heavy axle loading. The study involved
dropping dynamic impact loads typical of those imposed by
shelled wheel treads, out-of-round wheels or wheel flats on
rails, which had been removed from track because of detected
defects of different types. The rail specimens were pulled
o 5-14 x
longitudinally to simulate tensile stresses from low
temperatures, and some specimens were tested at down to
20C.
It was found that:
1. Impact loading was far more likely to fracture defects
in the transverse plane;
2. The tensile stresses imposed by temperatures
substantially less than the neutral temperature were
important in causing rail fracture;
3. Where the rail is shelling excessively, sudden rail
fracture will occur at lower, more frequent impact
level;
4. The residual, thermal and dynamic stresses imposed by
traffic contributed equally to total stress intensity;
5. The size of the flaw is a more important risk factor
than the percent of the railhead that has fractured. In
fact, a larger railhead may fracture more easily under
dynamic loading. Because a greater rail mass must be
rapidly moved aside under a high frequency impact,
i.e. has greater inertia, a larger railhead is less
compliant and may absorb more energy in impact.
Through observation of the conditions under which a rail
with a known defect could fracture suddenly, an equation was
developed from this work, which calculates the peak dynamic
load at fracture, P
dyn
, stated in kilopounds (kips):
46 . 6 38 . 1
83 . 4
T C I
b
K
p
dyn
(1)
Where:
K is the fracture toughness of the rail steel. This
value IC is typical 38.5 MPa for standard rail steels
and 20% higher for premium rail steels;
T is the variation of the rail temperature from its
neutral or stress-free temperature in degrees
Fahrenheit.
o 5-15 x

r
is the residual stress determined from the opening
that develops in a saw cut test. A value of 15.7
kPa/mm (14.3 ksi) is a good estimate from the
Queens University tests.
For example, using the above empirical equation, the
following combinations of conditions could cause sudden rail
fracture:
For a transverse defect covering 8% of the railhead:
a rail temperature of 56 Celsius (100 Fahrenheit)
below the neutral temperature and a dynamic wheel
load of 356 kN (80,000 lbs).
For a transverse defect covering 10% of the railhead:
a rail temperature of 56 Celsius (100 Fahrenheit)
below the neutral temperature and a dynamic wheel
load of 311 kN (70,000 lbs).
For a transverse defect covering 18% of the railhead:
a rail temperature of 39 Celsius (70 Fahrenheit)
below the neutral temperature and a dynamic wheel
load of 311 kN (70,000 lbs).
For a transverse defect covering 40% of the railhead:
a rail temperature of 56 Celsius (100 Fahrenheit)
below the neutral temperature without any wheel
loading.
See Part 3, Appendix A: Rail Defects (Canadian Pacific
Rail & Spoornet)
5.2.5 Rail Fatigue Projection
Most railways performing regular projection of rail life use the
Weibull methodology for projecting rail fatigue occurrence
rates. The Weibull methodology is useful in identifying
locations where trends are sustained vs. the case where defects
have remained constant. The situation where rail defect
occurrence rates are increasing is more critical, as this may
signal a mature fatigue process. These projections are used to
identify consistent trends in rail defect occurrences that could
be cause for a rail renewal program.
o 5-16 x
Attention to trends identified through regular use of
Weibull projections may guide selection of a strategy to correct
a defect trend by tamping up rail joints, building up rail ends by
welding, relieving the gauge corner, or attending to flat wheels.
Rail should be changed out when the annual cost of
repairing rail defects exceeds the value of deferring the renewal
for another year. At a repair cost of only $2500 per defect, and
an annual value of $18000 in interest savings if you leave the
rail in track, it requires a strong trend line to justify a rail
replacement for defect occurrences alone. But if a significant
number of these rails are failing in service, this introduces the
possibility that leaving the rail in track may incur the high cost
of line outages during emergency rail replacements and broken
rail derailments.
The key therefore is in maintaining effective rail testing.
As shown, service reliability requires both effective testing
systems and frequent rail testing. Attainment of long rail
service lives in a heavy haul environment similarly requires a
strategy to support rail economics with effective rail testing.
5.2.5.1 Use of Weibull Distribution to Predict Rail
Flaw Occurrence Rates
( )

1
]
1

1
]
1

T
e
T
T f
1
) ( (2)
The Weibull probability density function is given by:
(T) > 0, T > , > 0, > 0, - < <
Where:
= Shape parameter
= Location parameter
= Scale parameter
T = Time, Tonnage etc.
o 5-17 x
The Weibull reliability function is given by

,
_

T
e T R ) ( (3)
and the Weibull failure function

,
_



T
e
T R T F
1
) ( 1 ) (
(4)
The failure function is manipulated into the following form:
( ) ( )

n T n
T F
n n
T
T F
n
e
T F
e T F
T
T
l l l l
l

,
_

,
_

,
_

,
_

) ( 1
1
) ( 1
1
) ( 1
1
) ( 1
(5)
This linear relationship is used for constructing Weibull
probability paper. ( ) n l is constant for a given situation.
The Weibull failure rate, (T), is given by
( ) 1
) (

,
_

T
T (6)
In rail failure analyses one of two avenues for the calculation of
reliability or failure rates can be followed:
1. A maximum number of failures, defects or
occurrences, per distance of track, of a certain nature
can be decided upon beforehand. Once this level of
failures has been reached it is assumed that 100% of
occurrences had been experienced and some action
like replacing of the rail is taken.
o 5-18 x
2. No previous decision regarding the number of defects
that is allowable in the track has been taken. Here use
is made of the so called Median Rank to allocate a
value of F(T) to failures. The Median Rank will, in this
case, again be based on a unit length of track.
In order to obtain relevant results from a Weibull analysis
of rails the track must be divided up in homogeneous units.
Information required for analysis includes:
1. The type of defect (Classification of failure);
2. Tonnage to failure;
3. Time to failure;
4. History of repairs and maintenance;
5. Infrastructure data. Position in track etc.
Lengths of rail in a unit may vary upon conditions. In
general lengths from 5 km to 50 km may be used. The
considerations for lengths of rail to be identified, tested and
analysed will be discussed later.
The following example illustrates the typical use of the
Weibull function.
Failure data for heavy haul-line 20 to 40 km:
Failure type : Kidney shaped crack
Line length (km) : 20
Max. defects per km : 5 (Has to be decided on as policy)
Table 5.1 shows data from a spreadsheet program used for
the calculation of the Weibull parameters.
o 5-19 x
Table 5.1: Results from Weibull analysis
Tonnage
(mgt)
Years Failures
per period
Ave. Failures
per km
% Failed
100 2 5 0.25 5
200 4 8 0.65 13
300 6 4 0.85 17
400 8 8 1.25 25
500 10 4 1.45 29
600 12 6 1.75 35
700 14 5 2.00 40
800 16 9 2.45 49
900 18 8 2.85 57
The columns in Table 5.2 are:
F(mgt) % of failures of the Kidney shaped crack type.
Based on the max. defects allowable per km.
Tonnage (mgt) Actual gross load carried by rails.
) (
) ( 1
1
Tonnage n x
MGT F
n n Y
l
l l

,
_

Weibull parameter calculation


Location parameter, :0
Table 5.2: Calculation of Weibull parameters
F(mgt) Tonnage (mgt) Y X
5 100 -2.9702 4.6051
13 200 -1.9714 5.2983
17 300 -1.6802 5.7037
25 400 -1.2459 5.9914
29 500 -1.0715 6.2146
35 600 -0.8421 6.3969
40 700 -0.6717 6.5510
49 800 -0.3955 6.6846
57 900 -0.1696 6.8023
Regression Output: (Linear regression done on Y and X
columns)
o 5-20 x
Constant -8.50235
Std Err of Y Est 0.088236
R 0.991049
No. of Observations 9
Degrees of Freedom7
X Coefficient(s) 1.207463
Std Err of Coef. 0.043373
Shape parameter, = X coefficient = 1.207463
n l = 8.502351 (Constant)
n l = 7.041499
= 1143.099
From Table 5.2 the values of the Weibull parameters were
obtained:
= 1143.1
= 1.207
= 0
Using the Weibull parameters obtained above the
following typical calculations are now possible:
Reliability at certain life
% 14 14 . 0 ) 000 2 (
1 . 143 1
0 000 2
) 000 2 (
) (
207 . 1

,
_

,
_

R
e R
T
e T R

In terms of our model of five allowable kidney shaped


defects per km the rail will after carrying 2000 mgt have a 14%
reliability; i.e., only 0.14 x 5 = 0.7 defects per km will be
allowable or 4.3 defects per km will already exist.
o 5-21 x
Failure rate at a certain life
( )
( )
km per MGT per defects failures
T
T
/ 00111 . 0
1 . 1143
0 1500
1 . 1143
207 . 1
) 1500 (
) (
1 207 . 1
1

,
_

,
_

When a certain defect level will be reached


Should it be decided to start ordering rails when defects
have reached a level of 4.5 defects per km:
207 . 1
1 . 1143
0 2280
1
1
% 90
0 . 5
5 . 4
) (

,
_

,
_

e
T
e
T F

(By means of replacement of T in computer model).


= 0.90 = 90%
This means that the defect level of 4.5 defects/km will be
reached by the time 2280 mgt has passed over the rail.
Further refinements to this model by for instance adding
confidence limits is possible.
5.2.6 Modes of Rail Testing
An effective rail test is one in which all defects which could
present a hazard to the safe passage of trains, at the time of
testing and projected forward to the time of the next test, are
located and sized to an accuracy that permits a valid decision to
be made on its removal. A skilled test operator can perform a
reliable test at a spot location with hand probes within a few
minutes. The problem is in knowing where to look for the
defect. The solution is to use a machine to locate potential
defects at speed.
o 5-22 x
Thus, there are two distinct components to rail testing:
1. Location of defects by machine.
2. Description of defects by the operator.
The basic elements that are necessary to ensure that a
defect is correctly located and identified are:
1. The equipment used for detection must represent the
size of the defect sufficiently accurately to produce a
recognizable indication(s) on the operators display
under all rail surface conditions that could be
expected.
2. The indication presented to the operator must be
clearly identifiable as a rail defect.
3. The operator must have sufficient training, experience,
and vigilance to respond to a defect indication with
very high reliability and to correctly perform hand
testing to identify the defect.
The two aspects of detection, involving machine and
operator, explains why rail testing is carried out in a variety of
ways:
Non-stop hand testing, where an operator pushed
trolley-mounted equipment along at a walking pace and carries
out a pedestrian sweep, stopping to explore and confirm
indications. This offers the advantage that the information is
presented to the operator at a slow speed and he may perform
a simultaneous visual check of rail surface conditions. The
major disadvantage is the high cost per test kilometre due to
slow test speed and the need to test each rail separately.
While some railways still do out-of-face rail testing with
ultrasonic hand trolleys, their use is waning due to high labor
costs, the need for a chase vehicle anyway to regularly recharge
water supplies and safety considerations. Therefore, further
comments will focus on flaw detection by vehicle.
Stop and confirm machine testing, where the test
vehicle stops at each indication and the hand operator gets out
to verify and mark the defect.
o 5-23 x
Advantages are:
Less sophisticated equipment means lower capital
cost.
Rail is marked for renewal at detection.
Detection can run in sync with rail repair.
Test vehicle may be hi-rail equipped, increasing its
flexibility to move between sites over roads and to set
on and off track at road crossings.
Disadvantages are:
Slower operation than non-stop testing.
Test equipment may have to run under rules for track
equipment to permit backup moves.
Crew of hi-rail equipped ultrasonic cars must travel to
other lodging.
Non-stop machine testing, where a multi-probe machine
travels the section at typically 35 40 km/h. Real time
computing attempts to recognize signal indications, which
could possibly be defects, and to paint the location for follow
up hand testing.
Advantages are:
Less interference with traffic.
Lower unit cost of testing.
Higher productivity.
In some territories, signal systems will not allow rail-
bound equipment to back up, leaving a long walk if
immediate verification is required.
Disadvantages are:
Higher capital cost of equipment.
Ultrasonic car may leave behind more defects than can
be fixed in a day, leading to slow orders.
o 5-24 x
Longer time interval between detection and
verification.
Rail surface fatigue can cause excessive indications.
Real time detection by computer must necessarily be
conservative leading to a tendency to paint too much
rail.
Hand test results are not available to recognize for
recalibration..
Tandem machine testing is a variation on non-stop
testing. In this approach, a chase vehicle working in the
same track possession follows the principal testing vehicle.
The led test car works non-stop as fast as the rail condition will
allow, while the satellite car regulates its average speed, stops
included, to match the site advance. This method is capital-
intensive, but addresses many of the disadvantages of non-stop
testing.
5.2.6.1 Rail Testing Equipment
Various types of non-destructive methods have been employed
for testing rail in track, the main ones being:
Induction, where a low voltage eddy current is passed
along the rail between two moving probes, inducing a
strong magnetic field around the rail. An internal rail
defect causes distortions in the field around the rail,
which are picked up by search coils (see 5.2.10).
Induction methods can detect railhead defects and
certain web defects outside of the joint bar area.
Residual magnetic, where the rail is magnetized, and
search coils generate a weak current at irregularities in
the rail.
Ultrasonic, where ultrasonic waves are beamed into
the rail and the echoes are studied for irregularities.
As ultrasound is the technology most frequently used by
heavy haul railways, future comments will deal with ultrasonic
rail flaw detection only.
o 5-25 x
5.2.7 Ultrasonic Principles
While ultrasonic is a very specialized field usually left to the
experts, it is helpful for the railway user to have a grasp of the
principles involved. The test tool is a beam of electro-acoustic
energy with a frequency in the region beyond the hearing
range. The beam which can be linked to that of an electric
torch or flashlight is some 20 mm (1 in.) in diameter at its
origin and diverges from cylindrical at 35 degrees. The beam
is pulsed switched on then off at a set distance along the
rail, usually 2, 4, or 5 mm.
Ultrasonic transmitter crystals may be fitted in sliding
shoes or in rotating wheels. The sliding shoes are in closer
contact with the rail and afford a good angular stability, as the
mounting is flexible and adaptable. The wheel probes deal
better with irregular rails and offer a broad base for scanning
the railhead. A combination of both systems exist on some
machines.
Like light, the transmitted ultrasonic energy is refracted
upon changing from medium to medium, and reflected upon
meeting a suitable surface that is roughly perpendicular to its
propagation. In the flaw detection operation, two phenomena
are of interest. A beam hitting a discontinuity can reflect back
and disclose an appearance that indicates a potential defect.
And a beam masked from an expected end-echo cannot reflect
back, and thus gives a disappearance that indicates a
potential defect.
In practice, one is looking for various types of defects,
each with its own characteristics. The major characteristic
distinguishing different defect types will be the defects plane
of propagation. Thus, a transverse defect will be situated on a
plane across the railhead and sloping at some 70 to 90 degrees
to the vertical, while horizontal split heads and vertical split
heads describe themselves. In order to achieve reflection, the
search beam must meet the defect plane at about right angles.
From here, it becomes clear why rail-testing cars are fitted with
several probes on each rail.
The beams are centered on the longitudinal axis of the rail
by mechanical means with reference to the gauge face. As the
o 5-26 x
beams descend the web, and cannot radiate out from that path,
there are zones in the rail foot hat are not tested by the
ultrasonic method.
The passage of the ultrasonic energy is not as clean as one
would like. In particular, there is a disturbed zone of about 10
mm at the interface between the probe and the rail. The
passage from one medium to the other is assured by a film of
water that acts as a couplant. Nevertheless, the first several
millimetres of the entry into the rail cannot be exploited.
Other parasite effects occur also. Thus the first stage in the
recovery of the test information is to filter the returning energy
to remove the misleading effects. The second stage is to set
adjustable gates to select the areas in the rail that are of
interest.
At this point, the energy reflected can be visualized on a
cathode ray oscilloscope, an illustration referred to a an A-scan.
Defects will be recognizable from a set shape of the
oscilloscope trace. They are said to have a signature.
In principle, this information suffices to locate potential
defects. In practice, the traces are lively and require great
attention for interpretation, and there are simultaneously
several channels of information for each rail. The problem is
now one of information technology to assure the recognition
of potential defects among the mass of tested data that will
flow through the system. For example, some 10 information
points are generated every few millimetres along the rails, while
travelling at 15 25 km/hr. Another visual aid may be a
pictogram. Known as a B-scan, this is a picture of a rail
section, with diode lights or computer graphics given a "quick
glance view of suspicious echoes. An audio tone signal can
give a supplementary indication.
All of the above indications are usually presented to the
operator in real time, but are usually stored on a multi-channel
line recorder for later consultation in cases of controversial
findings.
o 5-27 x
5.2.8 Inspection Effectiveness
Rail flaw detection reliability is a system in which total
reliability is the joint result of the performance of test probes,
data processing clarity of information displays to the operator,
the operator himself, and the management of rail testing
intervals. Where there is a weak link in the system, it must be
compensated by the other elements of the system.
For example, weak performance from the ultrasonic
probes can be compensated somewhat by more sophisticated
processing. Or overly complicated or confusing information
displays must be compensated by an experienced operator.
Most importantly, poor overall performance from rail testing
can be compensated somewhat by very frequent testing
intervals.
The following discusses some of the factors affecting rail-
testing reliability.
5.2.8.1 Test Probes
To locate an internal defect in rail, ultrasonic or induction
energy must be transmitted along a pathway and at an energy
level sufficient to produce a clearly anomalous reflection from
the fracture surface. This reflection must be distinguishable
from the base of the rail itself and from railhead surfaces, and
the signal received from the fracture surface must be
substantially greater than the overall noise level introduced by
the grain structure of the rail or from the probe itself. And the
signal received must be sampled sufficiently frequently to have
captured sufficient pulses or echoes to have seen the
defect.
Fortunately, the larger the defect, the greater the signal
reflected. The difficulty arises in the detection of threshold
defect sizes. The pulse count from a small transverse defect,
for example, can be very similar to the noise floor,
particularly in older, less clean rail steels.
Furthermore, the full size, or length of the defect, is almost
never seen. This is because the full size of the defect is only
seen if it happened to be oriented at exactly 90 degrees to one
of the search beams. As there are an infinite number of
o 5-28 x
possible defect orientations and a finite number of probe
orientations, this is rare. What is usually seen as the reflecting
surface is therefore the projection of the fracture surface onto
the plane of the probe.
True sizing of the defect only occurs when the defect is
linear and oriented at 90 degrees to one of the search beams,
As there are an infinite number of defect orientations and a
finite number of probe orientations, this is rare. In the
example shown in Figure 5.6, a 32 mm (1-1/4 in.) curvilinear
bolt hole crack emanating at 45 degrees from the bolt hole is
seen as a 25 mm (1 in.) defect by the ultrasonic probe
arrangement. A particular problem is experience in those rare
circumstances where the crack emanates vertically from the
bottom of the bolt hole. This type of crack is susceptible to
pull-apart, but is transparent to the 0 degree probe and under-
represented by 43% by the 35 probe.
Figure 5.6: Length of a Bolt Hole Crack 32 mm Long and
Growing at 45 as seen by 35 and 0 Ultrasonic Probes
Another example of undersizing by ultrasonic occurs when
defects are located in the extreme corners of the railhead, the
initiation point for detail fractures from shell (Figure 5.7).
7 8
Here the problem is due to the diverging angles of the
ultrasonic beam. A centrally-located ultrasonic beam, 20 mm
thick and diverging at 35 degrees, is unable to illuminate the
parts of the crack surface near the gauge corner and in the
o 5-29 x
head-web fillet area. Flaws larger than 65% of the head area
are therefore characteristically undersized. This can be
counteracted to some degree by the addition of 70 degree
probes on field and gauge side, but the lateral separation of
such probes is constrained by the possibility of taking air
when encountering a severely worn rail gauge corner.
Figure 5.7: Illustration of Characteristic under Sizing of a Large
Transverse Defect by a Single 70 Ultrasonic Probe
5.2.8.2 Signal Processing
The electronic signals received by the test probes must be
adjusted to discern a recognizable signal, while eliminating
spurious noise. Signal processing may affect overall testing
reliability if filters are not adjusted to recognize returning signal
thresholds that could constitute valid defects. For example,
processing logic that does not lower the threshold level for
defects found deeper in the railhead whose reflected energy
will be less than a shallow defect, would potentially miss
smaller defects deeper in the rail.
o 5-30 x
5.2.8.3 Displaying Indications to the Operator
Most testing systems use a gating logic whereby an ultrasonic
echo is divided into reflections from head, web, base, and
possible defect. An indication of a potential defect is
presented to the operator only if sufficient signal pulses or
echoes have been received within a time interval that would
constitute the defect zone. System test reliability depends
upon the successful definition of the time interval through
which echoes from the probes should be counted as an
indication of both the presence and size of a defect.
Recent developments have also sought to assist the
operator by recognizing patterns typical of rail ends, bolt holes,
bond pin drillings, etc. The objective is to present to the
operator only those patterns, which cannot be, explained by
typical track features. This prevents the operator from being
flooded with information that he must mentally process.
5.2.8.4 Operator Vigilance
Current testing systems continue to be operator sensitive. The
ideal operator can maintain mental vigilance over extended
periods of time, using his training and experience to identify
suspicious pen indications or patterns, in spite of various
distractions within the test car. Such operators exist, but there
are an equal number of excessively conservative operators who
frequently stop and hand test and may mark a rail for
unnecessary removal where they do not recognize the pattern
of indications. On the other hand, some operators are
production-oriented, or are perhaps too quick to attribute
unusual indications to a rail surface condition. Operator
performance should be reviewed regularly be selecting random
samples of recorded signal indications, ideally in territories
where the number of detected defects has changed dramatically
between tests. These recordings should be reviewed with the
operator to locate areas where he may have missed a potential
rail defect.
o 5-31 x
5.2.8.5 Estimates of Rail Testing Reliability
The effect of defect size on the probability of finding a defect
is well illustrated in Figure 5.8.
8
It was compiled by the
Transportation Systems Center and the AAR, and is an
assessment of typical testing capabilities of current contractors
equipment. The AAR model estimates that a defect covering
60% of the railhead has a 90% chance of detection in a given
test. At the same time, a flaw covering 10% of the head is
likely to have a probability of detection of only 45%.
Figure 5.8: Estimated reliability of conventional
rail test equipment
The American Railway Engineering Association is more
demanding in their recommended minimum performance
guidelines for rail testing. These guidelines, summarized as
Table 5.3,
9
define the minimum acceptable percentage of
defects that must be detected by a single ultrasonic test. The
detection rate recommended by the AREA as indicative of a
fair to good ultrasonic inspection varies with the type of defect,
its size range and the class of track. For example, the AREA
recommends that rail testing services be considering to be
operating below acceptable performance if more than 65% of
o 5-32 x
transverse defects in the 5-10% range go undetected. On the
other hand, 98% of transverse defects covering 60% of the
railhead must be correctly identified and marked for removal.
The specification also defines the minimum sizes of defects
that are considered both worthy of reporting and within the
size range for reliable detection.
Such a specification invites questions to as how the
required testing performance can be verified. The best method
is to have a section of test track containing defects of know
sizes. This tests the capabilities of the test equipment, but is
not a realistic test of the vigilance of the operator in normal
service. Some railways run tandem tests where two test cars
will alternate running in the trailing position.
Defects found by one test car and operator and not the
other, after verification by breaking open or lab inspection,
would be considered missed defects for the purpose of
verifying performance to the specification.
o 5-33 x
Table 5.3: Minimum Performance for Rail Testing
Defect Type
Size
(Length or % of head Area)
Reliability Ratio (% of such
defects properly indicated as
flaws in any single test)
Category I
II
5-10% 65% 55%
11-20% 85% 75%
21-40% 90% 85%
41-80% 98% 95%
Transverse Defects in the Rail
Head eg. transverse fissure
compound fissure engine
burn/welded burn fracture
81-100% 99% 99%
10-20% 65% 55%
21-40% 85% 75%
41-80% 95% 85%
Detail Fracture from Shelling
or Head Check
81-100% 98% 95%
3-5% 65% -
6-10% 75% 65%
11-20% 85%
75%
21-40% 90%
85%
41-80% 95%
95%
Defective welds
Plant Welds (Head)
81-100% 99%
99%
12-25 mm 75%
65%
25-50 mm 95%
90%
- Plant Welds (Web)
more than 50 mm 99%
95%
5-10% 75%
65%
11-20% 80%
70%
21-40% 85%
80%
41-80% 95%
90%
- Field Welds (Head)
81-100% 99%
95%
12-25 mm 75%
65%
25-50 mm 90%
85%
- Field Welds (Web)
more than 50 mm 99%
95%
50-100 mm long 80%
70%
100 mm - 1 m 95%
95%
Longitudinal Defects in the
Rail Head eg. horizontal split
head vertical split head
more than 1 m 99%
99%
Web Defects * eg. head and
web separation split web
50-100 mm 95%
90%
more than 100
mm
99%
95%
Piped rail
more than 200
mm any
85%
-
Any size with non vertical orientation,
evidence of bulged web or progression into weld.
85%
75%
12-25 mm 75%
65%
25-50 mm 75%
65%
50-100 mm 90%
85%
Web Defects in
Joint Area *
eg. bolt hole crack
head and web separation
more than 100
mm
99%
99%
* defects must have progressed more than halfway through the web.
o 5-34 x
5.2.9 Selecting Rail Testing Intervals
It can be seen from the above that management of rail testing
incorporates two abstract disciplines:
Risk management: an outlay of known prevention
costs to achieve a hypothetical reduction in the
probability of damage.
Statistical performance: the evaluation of detection
success rates in relation to probability tolerances.
Furthermore, the subject matter is not definitive. The
growth rates of known types of defects are not narrowly
predictable. There are sources of rail breakage that are clearly
not predictable, such as defects in the rail foot, and there are
random events such as infrequent but significant impacts from
wheel defects.
The well documented defect growth experience of the
FAST Heavy Tonnage Loop presents a unique opportunity to
explore the theoretical relationship between rail testing
frequency and the possibility of a service break. For illustrative
purposes, assume that the 11 defects shown in Figure 5.5
represents the full population of defects initiated and growing
over a one year period in a 20 km line carrying 40 million gross
ton annually. Based upon practical experience, it can further
be assumed that a transverse defect left in track with a size
greater than 60% could represent a significant risk of sudden
fracture, and that it would be the objective of ultrasonic testing
to prevent this eventuality.
It can be seen that in the absence of testing, from 13 of
these defects could be expected to have covered more than
60% of the railhead by 20 million gross ton of accumulated
traffic; 24 more would reach this threshold by 30 million
gross ton. By the time 40 million gross ton had passed over
this rail, it could be speculated that two broken rails would
have been experienced, representing a 1 in 100 chance of a
broken rail derailment using U.S. average statistics.
This case study can be used to illustrate the reduction in
risk that can be expected with rail testing. For example, if
testing at 9 million gross ton intervals (10 mgt), the rail flaw
o 5-35 x
detector car would pass over the defect identified with the
asterisk at the time when it would cover 23% of the railhead.
According to AREA specs in Table 5.3, it would have a 90%
chance of detecting and marking this defect for removal. If the
defect were missed and the flaw detector care were to again
pass over the site at 13 million gross ton (15 mgt), the flaw
would now cover 55% of the railhead and should be detected
with 98% probability. The net probability of detecting this
particular defect before it reaches the 60% size is therefore
calculated as 0.90 + 0.10 (.98) = 0.998. Of course, this makes
the perhaps gross assumption that there is no particular
recurring condition that is preventing detection.
Using this same methodology, one can calculate the net
effect of different test intervals on the probability that one of
these defects will reach the 60% level before being detected
and marked for removal. Using the AREA Minimum
Performance Guideline as an assessment of the typical
detection performance of the test car, the results shown in
Table 5.4 are obtained for the year.
Table 5.4: Effect of Test Interval on Expected Number of
Undetected Defects in a Hypothetical 20 km Line with
Transverse Defect Growth Rates as Measured in the FAST
Heavy Tonnage Loop
Test Interval Expected No. of Defects that
will reach the 60% Level
Undetected
36 mgt (40 mgt) 11
18 mgt (20 mgt) 0.700
9 mgt (10 mgt) 0.234
4 mgt (5 mgt) 0.004
It can be seen that the probability of leaving a transverse
defect undetected to a size representing a high risk of failure is
very dependent upon the testing frequency. Again, this
assumes that the probability of success is independent from
test to test. On the surface it would appear that very frequent
rail testing is economical, but there is a case of diminishing
returns.
o 5-36 x
Assume for example that it costs $2,500 to replace a rail
detected behind a rail flaw detector car costing $50 for the test.
An emergency replacement, on the other hand would involve
delay of trains and could cost $10,000 per occurrence. For
transverse defects, perhaps 1% of service failures may be
expected to lead to derailments costing an average of $400,000.
Using these cost numbers, the value of the different test
intervals can be calculated from the above probabilities for the
hypothetical 20 km. The results are tabulated in Table 5.5.
This example would indicate an optimal testing interval of
518 million gross ton, with 9 million gross ton as potentially
the most economical.
To illustrate the value of reliable testing, assume that a rail
flaw detector vehicle is used that does not meet the AREA
Performance Guidelines, but instead performs as predicted by
Figure 5.9. When testing for defects in this line at 9 million
gross ton (10 mgt) intervals such a car would be calculated to
leave an expected 2.02 defects that would progress to the 60%
level. A comparison of the economics of the two vehicles is
included as Table 5.6.
Table 5.5: Economics of Test Frequency in a Hypothetical Line
Annual Cost of Defect
Repairs
Test
Interval
Cost of
Testing
per year Detected Service
Expected
Annual
Broken Rail
Derail. Cost
Total Cost
per year
40 mgt $1000 $2500 $111,000 $44,000 $158,500
20 mgt $2000 $30750 $7000 $2800 $44,550
10 mgt $4000 $31915 $2340 $936 $39,191
5 mgt $8000 $32490 $40 $16 $40,546
Figure 5.9: Decision Matrix to Direct Risk Reduction
o 5-37 x
Table 5.6: Economics of Rail Testing Performance in a Hypothetical
Line at 9 mgt Testing Interval
Annual Cost of Defect
Repairs
Specification
Defining Car
Performance
Cost of
Testing
per
year
Detected Service
Expected
Annual
Broken Rail
Derail. Cost
Total
Cost per
year
AREMA $4000 $31915 $2340 $936 $39,191
TSC/AAR
Model
$4000 $27443 $20,228 $8092 $59,763
This example calculates that the less accurate test vehicle
would cost this heavy haul line an additional $20,000 each year
in this 20 km line segment, or $1000/km. There is clearly
value in maintaining good quality control on testing. In this
example the difference is chiefly due to the differences in the
testing performance in the 30 60% size. As a general
statement, if a rail flaw detection vehicle is to be cost effective,
it must be very good at detecting defects such as transverse
defects in the 30 80% size range, as this likely will represent
that last time the defect is seen by the detector car before crack
out unless test intervals are exceedingly tight.
5.2.9.1 Performance-Based Adjustment of Test
Intervals
In light of the inexact nature of the science, most heavy haul
railways control risk by monitoring the occurrence of both
detected and service defects.
In North America heavy haul railway practice, risk is
typically judged to be sufficiently high to merit tightening test
intervals when:
Service defect rates exceed 0.17 service failures/km
(0.1 service failures/mi/yr).
Service plus detected rail defects exceed 0.04
failures/km/million gross ton (0.06 failures/mi./mgt).
The ratio of service to detected defects exceeds 0.2.
In fact, risk can be the result of track condition that is not
matched to service demand, test intervals that are not matched
to the reliability of testing systems, or both. The appropriate
o 5-38 x
course of action can be determined by comparing service
failure statistics and the number of detected failures.
Figure 5.9 illustrates the decision matrix that could be used
to direct an effort to reduce risk. For example, if service
failures are exceeding 0.04/km/million gross ton (0.06
failures/mi./mgt), it is apparent that the railway property is
living with a significant risk of a broken rail derailment. The
obvious question is whether ultrasonic testing is reliable and is
being performed frequently enough to find the defects. Should
it also be the case that service failures represents more than
20% of all rail defects recorded, it can be surmised that testing
intervals must indeed be tightened as a first step to reducing
risk. On the other hand, if service failure rates are low, but
detected defects are high, it can be presumed that rail testing is
effectively compensating for a track that may have stress
problems or cumulative fatigue damage.
5.2.9.2 A Parametric Approach
In 1991, Committee 4 of the American Railway Engineering
Association developed a quantitative guidance for specifying
recommended ultrasonic testing intervals. The emphasis was
not on specifying the intervals themselves, but on illustrating
how different railways have perceived the relative effects of
different parameters on risk, and hence the resultant test
interval.
The results represent the experience of two major US
railroads that have developed inspection interval planning
equations. The multipliers suggested to account for different
conditions are given in Table 5.7:
Table 5.7: Inspection Multipliers per Parameter
Significant Parameter Parameter Range Inspection
Interval
Decrease
Annual Tonnage Rate 10:1 ratio increase 70-80%
Track Class (as defined by
maximum allowable freight
speed)
U.S. FRA Class 1 to 6
(16 km/h 133 km/h
60%
Existence of Passenger
Trains
vs. exclusive freight
line
50%
Rail section Size 68 kg/m vs. 45 kg/m 70%
Prior Rail Defect Rate 10:1 ratio increase 50-70%
o 5-39 x
5.2.9.3 Cluster Testing
When addressing the risk of a rail-caused derailment, it is wise
to look at the rail plant comprising a routing as a series of
shorter sections of track, with different defect-producing
potential. In older railway lines, lack of homogeneity is a
natural result of the relaying of sections of track in different
years, resulting in rails with different accumulated service
tonnage.
Curved track also generally produces more defects due to
the additional stresses imposed by lateral loading. This occurs
even if the sections of track has rail with the same accumulated
service tonnage with uniformly good track support conditions.
Finally, variations in track and sub-grade support
conditions, rail metallurgical cleanliness and rail weld quality
can have profound influences on defect occurrence rates. As
an example of the impact of metallurgical cleanliness in the
period 1972 1980, Canadian Pacific Rail found that fully 38%
of transverse defects had occurred in rails from the A or top
position of the ingot, which potentially has the greatest density
of non-metallic inclusions. A ingot rails would have
constituted only 18% of the population of rails in track.
8
It follows then, that the greatest risk reduction pay-off
from rail testing will result from tests in those locations with
higher defect occurrence rates. The practice of scheduling
additional tests in high defect locations is called cluster
testing. To reduce the additional cost of testing intervals
governed by high defect locations, rail-bound rail testing cars
may deadhead without testing over intervening track segments
with acceptable defect occurrence rates.
Many railways schedule cluster testing on the basis of
some known characteristic of track. For example, a high
curvature section of an otherwise tangent routing might receive
an extra test, as could a length of older rail, or jointed rail
within a continuous welded rail routing. Other railways
monitor service and detected rail failures and schedule testing
based upon the limits of high defect rate locations.
o 5-40 x
Hi-rail based cars are particularly adapted to cluster testing,
where road access and frequent level crossing permit easy
access to spot locations without tying up track in deadheading.
As there is a cost to deadheading between sites to be
cluster tested, for scheduling purposes, high defect locations
should be 1020 km in length, with sufficient defect
occurrence rates, when averaged over this length, to trigger the
selected threshold for an additional test.
Therefore, when selecting test intervals for a routing, these
should be somewhat related to the potential for a service
failure, in turn leading to a derailment. Test intervals should
target specific longer track sections with significantly different
characteristics of rail age or quality, rail weight, jointed vs.
welded rail, track support quality and curvature. When, after
tailoring testing intervals to these characteristics of track,
defect occurrence rates in any homogeneous grouping of track
are still high, an additional test should be performed in the
offending location to control overall risk.
5.2.9.4 Special Care in Special Track Work
The turnout area represents a particularly difficult area to test
due to change in the cross-section of the rails and castings.
This means that the ultrasonic echo will return at a different
time than expected or that some probes will contact the
running surface at unusual angles. As a further complication,
castings have a considerably coarser grain structure than the
surrounding steel leading to different base echoes. As a
general rule, only the standard rail cross-sections within the
turnout area and railway diamonds are effectively tested with
flaw detection vehicles.
In at grade crossings, the fouling of the rail surface by
road-borne materials, particularly salt, can obstruct a good
ultrasonic indication. This can be overcome by sweeping out
the crossing in advance, slowing down the test and reversing if
an unusual indication is seen. Welds are another problem.
Because of the change in grain structure and the fact that
fractures can propagate rapidly from very small cracks or stress
raisers, welds are very difficult to test either with ultrasonic or
induction. One possibility is to have automated ultrasonic
o 5-41 x
recognition of the weld upset, which could trigger a change in
the signal gain and the use of tighter inspection tolerances.
Most heavy haul railways ensure more careful testing
through special track work. Some have retained a program of
hand testing, however hand testing typically uses the same
probes that are used by rail flaw detector cars.
5.2.9.5 Rail Testing Intervals Canadian Pacific
Approach
Canadian Pacific Rail System uses a risk management approach
whereby rail-testing intervals are adjusted according to
different categories of risk. The approach used is to first
segment all tracks into homogenous sections with the same
tonnage, weight of rail, type of traffic and rough levels of past
defect occurrence rates. These segments must be at least 16
km (10 miles) long to be practical for an additional test.
The approach is to first select a basic testing interval that is
dependent upon tonnage, which is a proxy both for the rate of
accumulation of fatigue in the rails and the probability that a
service failure will be encountered by a train. As Table 5.8
shows, there are six basic testing intervals based upon the level
of tonnage.
The testing interval for each track segment may then be
upgraded to the test frequency corresponding to the next
highest risk class if there is an additional element of risk
associated with the track segment. The factors that will qualify
for a more frequent risk are:
o 5-42 x
Higher Risk Traffic: Line carries passenger trains
Line carries hazardous materials
Lower Standard Rail: Non control cooled rail is being used in a
line
Carrying more than 1 million gross ton
per year 50 kg/m (100 lb/yd.) or lighter
rail is being used in a line carrying more
than 2.7 million gross ton/year.
50 kg/m (100 lb/yd.) or lighter rail is
being used in a line where train speeds
exceed 67 km/h (40 mph).
Evidence of Rail Fatigue:Detected rail defects exceed 0.7 defects
per km
(1.2 defects/mi.) per test
Evidence of Low
Inspection Effectiveness:Service failures exceed 0.12 failures per
km
(0.2 failures per mile) per year.
Table 5.8: CP Rail Testing Standards are Based Upon Eight
Testing Frequencies
Traffic Density +
(mgt/yr.)
Traffic + Rail Type + Defects = Test Class
< 0.5
0.5 2.7
2.8 7.2
7.3 13
14 27
> 27
Hazardous
Materials
Passenger
> 70 km/h
Non
Cooled
< 50 kg/m
> 0.7/km/yr.
Service/detect
ed ratio > 0.2
5 yrs.
3 yrs.
2 yrs.
annual
2/yr.
3/yr.
4/yr.
5/yr.
If any three of the above factors are in evidence in the line
segment, the testing interval is tightened by two classes.
Therefore, in the example of Table 5.9, a line carrying 10
million gross ton per year would be tested once per year. But
if it carried hazardous goods a well, it would be tested twice
per year. If in addition to this the line was laid with lighter
than 50 kg/m (100 lb/yd) rail and had a service to detected
defect ratio of 0.25, it would be tested three times per year, or
after every 3 million gross ton of traffic.
o 5-43 x
Table 5.9: Test Interval Class is Tightened Based
Upon Risk Factors
Traffic Density +
mgt/yr.
Traffic + Rail Type + Defects = Test
Class
7.3 13 annual
7.3 13
Hazardous
Materials
2/yr.
7.3 13
Hazardous
Materials
< 50 kg/m
Service/detected
ratio > 0.2
3/yr.
5.2.10 Induction Measuring Principles
The induction testing technique requires the injection of a
direct current into the rail. The current is generally around
3600A. The injection takes place through the application of
two sets of brushes that are placed on the railhead. The spacing
between the brush sets is of the order 120cm (4ft). The current
flows into the rail through the leading brush set and out
through the trailing brush set. The rail thus becomes part of an
electrical circuit.
Once motion is introduced, a magnetic field associated
with the current flow in the rail is induced. The magnetic field
is the means by which information about the condition of the
rail is coupled to the sensor unit. The sensor unit is located
between the two sets of brushes. The sensor unit is set up to
maintain a constant lift-off between the underside of the unit
and the surface of the railhead. If this is not done, the data
recorded will be noisy and thus very difficult to interpret.
The mechanism by which rail condition is inferred starts
with the current. In general for modern rail weights, only the
head and the top part of the web is filled with current. In the
past with smaller rail sections, the whole rail section has been
filled with current. As the current flows through the rail, if any
features such as a defect block the current path, the current
will take the shortest possible route to get around the
obstruction. This distortion of the current flow will also lead to
a distortion of the associated magnetic field. It is this distortion
of the magnetic field that is detected by the sensor unit (see
Figure 5.10).
o 5-44 x
Figure 5.10: Distortion of Induced Magnetic Field around
Two Types of Flaws
The sensor unit itself houses multiple coils or Hall Effect
devices. Often the arrangement is differential in nature to help
keep the number of false indications down. By differential it is
meant that two identical sensors located next to each other
across the railhead will be wired together. Thus it is only when
one sensor sees a disturbance and the other doesnt that a
signal will be sent to the test system. For example, a rail end is
essentially a gross transverse defect, both sensors will see the
rail end so no signal will be sent to the test system. A
transverse defect will generally only be seen by one sensor, so
the asymmetrical disturbance will send a signal to the test
system. Multiple sensors are used to allow the detection of all
of the components of the field disturbances.
Considering the current flow through the rail, as it is
longitudinal, current distortion will not occur as a result of
longitudinal features in the rail. The features that will produce
the most current disturbance are those that are transverse in
the railhead. Unlike the ultrasonic technique, the induction
technique does not have trouble with inspecting right to the
top surface of the railhead. The nature of the current flow is
o 5-45 x
such that it is the very center of the railhead that is likely to be
missed if the system is unable to fill the railhead with energy.
The signals sent to the system are generally observed to
determine if they exceed a set threshold. If they do, a count is
started. The number of threshold exceedances then determines
whether the data is presented to the inspecting operator as a
potential defect or not. With increasing computer power new
analysis algorithms, some combining information from
different channels (both induction and ultrasonic), are
becoming more common.
The data can be presented in many different formats. Most
often it is a combination of processed (counted) data and raw
analog data side-by-side. The processed data is often the
mechanism that highlights the problem area and then the
subtle features of the indication can be extracted from the
analog waveform.
5.2.11 Conclusion
While recent research has provided some clues to assessing
risk, it is not yet possible to develop a solid mechanistic
relationship between the risk of derailment and the frequency
of tests. To control the likelihood of service failures, rail can
either be tested very frequently with lower accuracy equipment,
as it will be likely that the rail is scanned while the defect is
larger and more easily detectable. Alternatively, a more
accurate test system can be employed on a longer cycle, as it is
likely that even if the test happened to coincide with the early
appearance of the defect it will still be detected. In fact, heavy
haul operators have found that the costs of poor service
reliability are such that it is profitable to both use very effective
testing systems, with particular emphasis on high reliability in
detection of larger defects, while also maintaining frequent
testing.
5.3 Rail Wear Measurements
5.3.1 Rail Wear Measurement Techniques
It is true in the rail industry as in any other that What Gets
Measured Gets Managed. A regular program of rail cross
sectional/wear measurements is critical to the ability to

Das könnte Ihnen auch gefallen