lingamanaik2013_2

8/16/2019 lingamanaik2013_2

1/8

Prediction of residual stresses in low carbon bainitic–martensitic railway

wheels using heat transfer coefficients derived from quenching

experiments

Siva N. Lingamanaik ⇑, Bernard K. Chen

Department of Mechanical and Aerospace Engineering, Monash University, Victoria 3800, Australia

a r t i c l e i n f o

Article history:

Received 27 September 2012

Received in revised form 15 February 2013

Accepted 16 April 2013

Keywords:

Residual stresses

Quenching

Heat transfer coefficient

Railway wheels

Finite element modelling

a b s t r a c t

Low carbon bainitic–martensitic (LCBM) steels have been recently developed for railway wheels and have

been shown to provide superior properties compared to conventional pearlitic railway wheel steel

grades. Pearlitic railway wheels are generally quenched at the tread region to promote the formation

of compressive residual stresses in the rim to mitigate the initiation and propagation of cracks due to fati-

gue. However, this conventional quenching method has been shown to be unsuitable for LCBM railway

wheels. Alternative quenching methods were evaluated using a FE model to develop a successful quench-

ing process to produce LCBM railway wheels. Heat transfer coefficients were determined by employing a

full scale experimental rig and were used in the FE model to model various coolant spray intensities and

configurations. The FE model was used to determine optimal quenching conditions that impart compres-

sive residual stresses to the rim of the LCBM railway wheel and the prediction of residual stresses were

verified experimentally.

2013 Elsevier B.V. All rights reserved.

1. Introduction

In the past decade, the mining industry and heavy-haul freight

service in Australia have been growing due to the increasing

worldwide demand for natural resources such as iron ore and coal.

It is estimated 352,000 railway wheels are in service across the

Australian rail network, with an estimated annual maintenance

cost of $60–$190 M [1]. This figure is expected to increase as more

tracks and railway vehicles are added to the rail network to meet

the growing demands. Hence, rail operators, driven by profitability,

have been working to reduce maintenance costs while increasing

performance, reliability and safety of railway wheels.

Most railway wheels are made using the specified Association

of American Railroads (AAR) Class steel compositions which have

a pearlitic–ferritic microstructure [2]. Over the years, their perfor-

mances have been enhanced mainly by cleaner steel production

and by micro alloying of elements to increase the strength of pearl-

itic wheel steels [2]. However, researchers agree that there is lim-

ited scope for further strength improvements in this class of steels

[2–6]. Merely increasing the carbon content to increase strength

and hardness would inevitably contribute to lower toughness

and higher sensitivity to brittle fracture and increased risk of spall-

ing failure [2].

Lonsdale and Stone [2] conducted a review of other steel com-

positions with the potential to improve the life of railway wheels

including martensitic steels and Constable et al. [4] studied the

suitability of low carbon bainitic martensitic (LCBM) steels for rail-

way wheels. Low carbon bainitic martensitic steels (0.20%C,

4.0%Mn, 0.75%Si, 0.004%Mo, 0.003%V, and 0.005%Nb) were found

to have superior strength, hardness and toughness compared to

AAR Class A to C wheel steels [4]. LCBM steels achieved strength

levels up to 1130 MPa in standard mechanical tests which is a

40% improvement over the conventional micro-alloyed AAR Class

C grade steel [4]. LCBM steels have also shown enhanced resistance

to rolling contact fatigue (RCF) and thermal fatigue, which are ex-

pected to reduce the need for wheel re-profiling and lead to sub-

stantial savings in maintenance costs. Additionally, LCBM steels

employ low cost alloying elements and are not expected to result

in additional production costs compared to what is currently used

for AAR Class railway wheels.

Constable et al. [4] have reported an improvement of 69% in

fracture stress for LCBM steels (290 MPa) compared to micro-al-

loyed AAR Class C steel (160 MPa) at a crack length of 30 mm [4].

Hence, LCBM steels are likely to provide greater safety due to im-

proved fracture toughness compared to AAR Class C grade steels.

Furthermore, fatigue studies by Peng et al. [7] have estimated a

30% increase in the service of railway wheels made from LCBM

steels compared to AAR Class B railway wheels. Rail manufacturers

in Europe have also investigated the use of bainitic type micro-

structure steels to improve wear of rails [2].

0927-0256/$ - see front matter 2013 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.commatsci.2013.04.039

⇑ Corresponding author. Tel.: +61 03 990 53647.

E-mail addresses: [email protected] (S. N. Lingamanaik), Bernard.

[email protected] (B.K. Chen).

Computational Materials Science 77 (2013) 153–160

Contents lists available at SciVerse ScienceDirect

Computational Materials Science

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c o m m a t s c i

http://dx.doi.org/10.1016/j.commatsci.2013.04.039mailto:[email protected]:Bernard.%[email protected]:Bernard.%[email protected]://dx.doi.org/10.1016/j.commatsci.2013.04.039http://www.sciencedirect.com/science/journal/09270256http://www.elsevier.com/locate/commatscihttp://www.elsevier.com/locate/commatscihttp://www.sciencedirect.com/science/journal/09270256http://dx.doi.org/10.1016/j.commatsci.2013.04.039mailto:Bernard.%[email protected]:Bernard.%[email protected]:[email protected]://dx.doi.org/10.1016/j.commatsci.2013.04.039http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://crossmark.dyndns.org/dialog/?doi=10.1016/j.commatsci.2013.04.039&domain=pdfhttp://-/?-


2/8

Although, bainitic as well as martensitic type steels can offer

higher strength and hardness levels when compared to pearlitic

steels at similar carbon content, their use as railway wheels is lim-

ited because tensile stresses are formed in the rim of the wheel un-

der the conventional quenching process [2]. Most standards for

railway wheels such as AAR M-107/M-208 [8], BS 5892-3 [9] and

EN 13262 [10], require that railway wheels are manufactured with

compressive rim circumferential residual stresses to retard the ini-tiation and propagation of cracks due to fatigue and since the intro-

duction of these practices, the number of wheel-related

derailments in North America has fallen by an order of magnitude

[11].

Current AAR railway wheels are typically rim-quenched

(quenched at the tread’s surface) to promote the formation of com-

pressive residual stresses in rim of the wheel. Upon quenching,

austenite transforms to pearlite and thermal contraction which im-

parts compressive stresses in the tread region of the rim. Typical

as-manufactured compressive residual stresses in AAR wheels are

reported to be approximately 250 MPa at the tread’s surface [12].

However, during rim-quenching of martensitic type railway

wheels, the transformation from austenite to martensite in the

rim is accompanied by a volume expansion of approximately 4%.

In comparison, there is only a 1% volumetric expansion during

phase transformation of austenite to pearlite in conventional pearl-

itic railway wheels (as shown in the dilatometry cooling curves in

Fig. 1 [13]. Hence, there is a net volumetric expansion after austen-

ite has transformed to martensite in the rim of the wheel and upon

cooling to room temperature, this large volumetric expansion re-

sults in compressive stresses in the inner rim region and tensile

stresses near the tread’s surface.

Instead of applying conventional rim-quenching, Lingamanaik

and Chen [13] have shown that the quenching process can be mod-

ified to achieve compressive residual stresses in the rim of the

LCBM wheel. Their studies were based on FE modelling of the

quenching process based on estimated Heat Transfer Coefficients

(HTC) from the literature [14]. However, HTC values are known

to be highly variable depending on the actual quenching conditionssince factors such as geometry of the part, characteristics of the

coolant sprays, the temperature of the quenched surface and even

the surface finish or roughness can influence the heat transfer in a

part and its final stress distribution [15–20]. In the present study, a

full-scale experimental quenching rig was constructed and instru-

mented with thermocouples in selected regions of the railway

wheel to determine actual values of HTC during quenching of

LCBM railway wheels for different spray intensities and spray con-

figuration. These experimentally determined HTC were used in a

thermo-mechanical finite element model to develop a set of viable

quenching conditions for LCBM railway wheels to achieve com-

pressive stresses in the rim of the railway wheels.

2. Methodology

2.1. ABAQUS/DANTE thermo-mechanical finite element model

A Finite Element (FE) software ABAQUS 6.7.1 and a heat treat-

ment package DANTE 3.3 were used to model the quenching pro-

cess of railway wheels and to predict the formation and

distribution of residual stresses in railway wheels [13]. In DANTE3.3, user-defined material subroutines are used to predict and track

the volume fractions of metallurgical phases as austensite trans-

formed to pearlite, ferrite, bainite and martensite as the part is

cooled. Subsequently, thermal loadings and temperature depen-

dent material properties incorporated in DANTE are used to predict

final residual stresses [21–23].

For steels undergoing martensite phase transformation, the

transformation kinetics is written in the form of a rate equation

as shown in Eq. (1) with a strong dependency on the cooling rate

[22]:

dU

dt ¼ tM ðC ÞU

aðC Þð1 UÞbðC Þ

U ðM S hÞdh

dt ð1Þ

U ðM S hÞ, is the unit step function i.e.

U ðM S hÞ ¼ 1; hPM S

U ðM S hÞ ¼ 0; > M S

and tM ; a; b are material carbon dependent quantities determinedfrom TTT quench data [22].

By integrating the phase transformation strain rate over a time

stepDt, the phase transformation can be computed for each phase.

The dilatational transformation strain increment is given in the fol-

lowing equation:

E x p ¼ ðE p E AÞ

1 þ E Að2Þ

The transformation strain for austenite and product phases are

taken to be linear and cubic functions. For martensite, E p ¼ E M :

E M ¼ M 0 þ M 1h þ M 2h2 þ M 3h

3 ð3Þ

The Eq. (3) is temperature dependent and its coefficients are

carbon-dependent.

In DANTE’s mechanics module, a hypoelastic response for each

phase is assumed and the effective stress to cause plastic flow in

each phase is given in Eq. (4). The inelastic deformation rate in

Eq. (5) is expressed in terms of deviatoric stresses:

jnij ¼ jr0ðiÞ aij kðiÞ ð4Þ

DðiÞ p ¼ f ðiÞðhÞsinh

jnðiÞj Y ðiÞðhÞ

V ðiÞðhÞ

! r0ðiÞ aðiÞ

jr0ðiÞ aðiÞj ð5Þ

where f ðiÞðhÞ and V ðiÞðhÞ describe the rate dependence of the yield

stress at constant temperature while the function Y ðiÞðhÞ is the

rate-independent yield stress. Mechanical properties for the mate-

rial for the various metallurgical phases such as modulus of elastic-

ity and yield strength are obtained from tension and compression

tests as functions of metallurgical phase, temperature, carbon con-

tent, strain level and strain rate and are implemented in DANTE’s

user-defined subroutines [22].

Phase transformation kinetics parameters can be obtained by

several sources which includes CCT diagrams, TTT diagrams, Jomi-

ny Hardness test and dilatometry data. While TTT diagrams are

mainly used for diffusive transformations such as pearlite, CCT dia-

grams offer data for both diffusive and martensitic transformation

as reported by Li et al. [23]. Jominy tests alone are not adequate fordetermining kinetic phase transformations since strain–time data

1.000

1.002

1.004

1.006

1.008

1.010

1.012

1.014

0 100 200 300 400 500 600 700 800 900 1000

L / L o

Temperature (oC)

Pearlitic

Bainitic -Martensitic

Fig. 1. Dilatometric cooling curves showing phase transformation characteristics of pearlitic and bainitic–martensitic steels [13].

154 S.N. Lingamanaik, B.K. Chen / Computational Materials Science 77 (2013) 153–160


3/8

cannot be obtained as shown by Li et al. [23]. However, Li et al. [23]

demonstrated that Jominy and TTT can be used to verify kinetic

parameters. In DANTE, dilatometry data is preferred over the above

sources as time, temperature and strain can be obtained from dila-

tometry experiments and cooling transformation kinetic parame-

ters can be easily verified against TTT and CCT data as described

by Li et al. [23].

A set of dilatometry experiments were undertaken to determineand quantify volumetric changes associated with martensite phase

transformation for a number of different cooling rates for LCBM

steels [13]. Thermal expansion data and kinetic rate equations

from the dilatometry experiments were incorporated into DANTE

in a similar fashion as described in Warke et al. [24]. A CCT diagram

and dilatometry curves for LCBM steel are shown in Figs. 2 [4] and

3 [13] respectively.

DANTE has an inbuilt fitting utility which can be used to obtain

kinetic parameters from dilatometry experiments. For martensitic

steels, phase transformation kinetics for martensite have been

determined using the DANTE fitting function and good agreement

has been found between predicted and experimental dilatometry

data as reported by Ferguson et al. [25,26]. Since TTT and CCT dia-

grams are available for common steels, they provide further checks

on the fitting model performance as shown by Li et al. [23]. Fur-

thermore, dilatometry data can be used and has shown to provide

thermal expansion and various phase transformation strains at dif-

ferent temperatures as described by Li et al. [23].

2.2. Development of a quenching rig for LCBM railway wheels

An experimental quenching rig has been constructed with the

flexibility to apply different coolant spray intensities at selected

locations on LCBM railway wheels. The main components (Fig. 4)

of the experimental set-up are described as follows:

- A turntable to support and rotate the wheel during quenching

(purple item)

- A variable-speed motor to rotate the table through a speedreduction and friction wheel drive system which allowed uni-

formity of the spray nozzles during quenching when the rail-

way wheel is rotated.

- Inner and outer fixed lower coolant manifolds (orange and

green items respectively).

- Hinged upper coolant manifold that allows the wheel to be

loaded onto the turntable, prior to positioning of the top spray

nozzles (yellow item).

- A centralised control system to operate the turntable and

pumps.

- Articulated nozzles to allow different regions of LCBM railway

wheels to be quenched independently (Fig. 4a).

Fig. 2. CCT diagram of LCBM steel 0.20%C, 4.0%Mn [4].

Fig. 3. Dilatometry curve for LCBM steel (0.21%C) at 0.9 C s1, 1.5 C s1, 3 C s1

and 9 C s1 [13].

(a)

(b)

Fig. 4. (a) Cut-away CAD model of the main experimental quenching rig

highlighting the principle components; and (b) uniformly distributed spray nozzlesand rotation of wheel platform.

S. N. Lingamanaik, B.K. Chen / Computational Materials Science 77 (2013) 153–160 155

http://-/?-http://-/?-


4/8

- Control valves and pressure gauges fitted to the pumps enable a

range of coolant spray flow rates.

2.3. Determination of values of HTC for LCBM railway wheels

Thermocouples (K-type-1.5 mm dia.) were embedded at several

locations in LCBM railway wheels (Fig. 5) to measure temperature

histories under three sets of nozzle spray pressures delivering a‘Low’, ‘Medium’ and ‘High’ spray intensities (Table 1). The values

shown in Table 1 have been normalised against the maximum

pressure measured for each set of nozzles. In each experiment, a

LCBM railway wheel was uniformly heated to 920 C and then

transported to the quenching rig located close to the furnace. The

thermocouples were then connected to t he data acquisition system

and the coolant sprays switched to one of the three spray intensi-

ties to cool the web region of LCBM railway wheel.

A ‘trial and error’ approach is commonly used for determining

the values of transfer coefficients in quenching experiments [27–

30] and this approach was also used in the present study of heat

transfer coefficients during quenching of LCBM railway wheel. At

the start of the analysis, the values heat transfer coefficient as func-

tion of the part’s surface temperature were estimated at the part’ssurface. The temperature history predicted for the node at the ther-

mocouple location is then plotted against the temperatures mea-

sured by the thermocouple. The values of heat transfer

coefficients were adjusted and the analysis repeated until the pre-

dicted results showed good agreement with the measured

temperatures.

In the FE model, the boundary of the two dimensional FE model

of the LCBM railway (Fig. 5) were divided into several sections and

HTC values were specified around the whole boundary surface. For

surfaces being cooled due to the overflow of coolant, a HTC value of

1500 W m2 K and 750 W m2 K were determined when the high-

est and lowest spray intensities were used respectively. For ambi-

ent convection and radiation, a HTC value of 48 W m2 K was

determined and a surface emissivity of 0.95 and Stefan–Boltzmannconstant of 5.67 108 W m2 K4 was assumed in the FE model.

2.4. Modelling of quenching process for LCBM railway wheels

In this work, a new quenching process is designed for quench-

ing LCBM railway wheels (see Fig. 6) whereby the web and inner

rim regions have been quenched separately. The FE model was

used to investigate three different quenching intensities ‘Low’,

‘Medium’ and ‘High’ labelled Cases L, M and Case H respectively.

Boundary conditions and thermo-physical quantities assumed in

the model are listed below:

(1) At the start of the analysis, the temperature of the railway

wheel was assumed to be 920 C. Time from furnace to

quenching test rig was averaged over multiple quenching

experiments and taken as 240 s. The thermocouples were

connected to the data acquisition system at approximately

150 s prior to the start of the quenching process which cor-

responds to t = 0 s in Fig. 8.

(2) The values of heat transfer coefficients were assumed to be

dependent on the part’s surface temperature (Fig. 9). In the anal-

ysis, the initial coolant temperature was assumed to be 19C.(3) Thermal conductivities of individual phases were tempera-

ture dependent; austenite = 0.016 + 1.3 105 (T ) W mm1

C; martensite = 0.025 + 3 106 (T ) W mm1 C.

(4) Specific heat capacities individual phases were temperature

dependent; austenite = 370 + 0.298 (T ) J kg1 C; martens-

ite = 450 + 0.387 (T ) J kg1 C.

2.5. Measurement of residual stresses in LCBM railway wheel

Verification of the presence of residual stresses in the rim of the

quenched LCBM railway wheel was undertaken by a radial saw cut

made to a depth at least one inch deeper than the rim inner diam-

eter as specified in AAR M-107/M-208 [8]. Prior to the cut, gaugemarks were stamped on the outer surfaces of the rim and initial

gauge length readings were taken. Cutting commenced from the

outer rim of the wheel (centrally between the gauge marks) and

progressed radially towards the centre of the wheel (Fig. 7). When

the cut has reached the web of the wheel (approx. 25 mm beyond

the inner rim’s surface), a final reading of the gauge lengths was ta-

ken and the net displacement computed; a negative reading or

‘closure’ indicating compressive residual stress in the rim and po-

sitive or ‘opening’ of the gap indicating the presence of tensile

residual stress.

Fig. 5. A half cross-section of 920£ freight S plate LCBM railway wheel showing the locations of embedded thermocouples.

Table 1

Normalised nozzle spray intensities.

Spray locat ion Nozzle sp ray int ens ities

LOW MEDIUM HIGH

Web 0.44 0.73 1

Rim Region #1 0.25 0.56 1

Rim Region #2 0.22 0.47 1


http://-/?-http://-/?-


5/8

3. Results

3.1. Temperature measurements in LCBM railway wheels under

different quenching conditions

Fig. 8 shows the temperature histories measured at thermocou-

ple Location #1 in the web section of LCBM railway wheels under

the different quenching intensities i.e. ‘Low’, ‘Medium’ and ‘High’.

In each case, on quenching, the temperature in the web section

of the LCBM railway wheel was shown to first decrease rapidly

in a non-linear fashion. After, approximately 40 s, the temperature

continued to decrease at a fairly constant cooling rate of about

1.2

C/s. The start of martensite phase transformation is expectedto occur much sooner in the web section at the highest quenching

intensity since it results in the highest cooling rate. For LCBM rail-

way wheels, martensite phase transformation start temperature

was shown to be approx. 330 C [13], occurring in this case approx.

50 s after the start of the quenching process. At the lowest quench-

ing intensity, the start of martensite transformation is expected to

be delayed further by approx. 70 s compared to quenching at the

highest intensity. Therefore, the quenching intensities were found

to critically influence the kinetics and distribution of martensite

phase transformation and the resultant residual stresses.

Support frame

Quenching Sprays in 1st

Stage Quenching Sprays in 2nd

Stage

Fig. 6. Schematic diagram of the alternate quenching process sequence shown on half cross-section of a 920£ freight S plate railway wheel.

Fig. 7. Radial cut of an as-cast experimental LCBM railway wheel using water-jet

cutting system.

S p r a y s w i t c h e d o n

Air cooling (during wheel transit)

Fig. 8. Temperature histories in the web section of LCBM railway wheel measured

at thermocouple Location #1 in Fig. 5 of LCBM railway wheel for differentquenching intensities.

Fig. 9. Experimentally determined heat transfer coefficients for different sprayintensities and those determined by Lee [15].


http://-/?-http://-/?-http://-/?-


6/8

3.2. Experimentally determined heat transfer coefficients for

quenching of LCBM railway wheels

The shape of the curve (Fig. 9) was found to be consistent with

the different regimes known to exist during quenching, allowing

for a shift in the position of the nucleate boiling regime (region

of highest heat transfer) due to factors such as the formation of

oxide layers, surface roughness and the coolant characteristics[15–20]. Wendelstorf et al. [16] found that the HTCs were influ-

enced by the formation of oxide layers during quenching. They

showed that the surface temperature at which nucleate boiling re-

gime occurred was dependent on the oxide layer; the deeper oxide

layer, the higher surface temperature of the nucleate boiling re-

gime. Also, in another study by Wendelstorf et al. [16], the nucleate

boiling regime during quenching was shown to be a function of the

part’s surface roughness in which the temperature at which nucle-

ate boiling occurred was shown to increase with a coarser surface.

There is a large variation in the published values of heat transfer

coefficients and this is not unexpected since the different regimes

namely, film boiling, nucleate boiling and convection have been

shown to depend on several parameters such as surface part tem-

perature, geometry, nozzle distance, surface roughness and coolant

temperature. Lee [15] reported values of HTC ranging from 500 to

3300 W m2 K and Karwa et al. [17] reported a range of values for

heat transfer coefficients between 7000 W m2 K and

25,000 W m2 K. Krause et al. [18] found the HTCs varied with dif-

ferent nozzle distances and reported values between 4200 W m2

K and 10,000 W m2 K. In a study by Chen and Tseng [19], a wide

range of HTCs (1000–100,000 W m2 K) was reported for acceler-

ated cooling of hot steels using spray-like nozzles and water jet

cooling.

Fig. 10 shows the temperature history recorded by the thermo-

couple located at the web region of LCBM railway wheel (Location

#1 in Fig. 5) under the ‘High’ spray intensity. Also shown are two

sets of predicted temperature histories, one assuming an average

(constant) HTC (Case A) and another assuming HTCs that are

dependent on surface temperature (Case B). A constant HTC valueis commonly assumed for simplicity in FE models used by Lingam-

anaik and Chen [13] and Khulman and Gallagher [14] but here, it

leads to an incorrect prediction of a greater heat loss at the early

stages of cooling (t < 375 s); therefore resulting in poor prediction

of the start of martensite phase transformation and volume frac-

tion of transformed martensite. Agreement with experimental

measurements requires the use of a surface dependent HTC which

was found in this study to vary between 1000 and 3500 W m2 K

depending on both the spray intensity and the part’s surface tem-

perature. The results demonstrate the importance of applying a

temperature dependent HTC in the FE model for predicting the

cooling histories, the kinetics of martensitic transformation and

the development of residual stresses during and after quenching

of LCBM railway wheels.

3.3. Prediction of phase transformations and residual stress

distributions in LCBM railway wheel for two different quenching

processes

Fig. 11a and b show the predicted distribution of temperature,

martensite phase distributions (Fig. 11c and d) after the first stage

of the quenching of LCBM railway wheel and the resultant residual

stresses (Fig. 11e–h). The highest and lowest spray intensities for

LCBM railway wheels were modelled as Case H and Case L respec-

tively. As expected for the highest spray intensity, the web section

was predicted to cool much quicker when compared to quenching

at the lowest intensity. For Case H, martensite transformation was

predicted to start at the web’s surface after approx. 60 s into the

quenching process and the distribution of martensite was pre-

dicted to extend deeper in the central region of the web after

150 s. At the end of the first quenching stage, the entire web region

of LCBM railway wheel (Fig. 11d) was predicted to have trans-

formed into martensite.

On the other hand, for Case L, martensite phase transformation

was predicted to start in the web approx. 130 s after the com-

mencement of the quenching process. At the end of the first

quenching stage, martensite phase distribution was predicted to

vary through the thickness of the web. The lower portion of the

web was completely transformed to martensite but the upper

web’s surface was only partially transformed (volume fraction of

approx. 0.3, Fig. 11c). For Case L, in which the lowest spray inten-

sity was used in the quenching process of LCBM railway wheel,

tensile residual stresses (Fig. 11g) were predicted to form near

the tread’s surface and in the flange of the wheel.However, when the spray intensity was increased (Case H), the

residual stress in the tread’s surface was predicted to change from

tensile (64 MPa) to compressive (20 MPa) (Fig. 11h). From a basic

conceptual perspective, compressive residual stress acts to resist

cracking by pushing the material together, while tensile residual

stress pulls the material apart increasing the propensity for crack

development.

In the new quenching process, the residual stresses in the web

of LCBM railway wheel (Fig. 11e and f) were predicted to be in

compression in the central region of the web with isolated tensile

stresses (400 MPa) along the inner rim’s surfaces and inner web’s

surface. The resultant stresses in the web from mechanical load-

ings were found to be significantly lower than the contact stresses

in the tread region due to the interaction between the wheel andrail. Besides, the web of the wheel generally tends to be into com-

pression as the load from the wheel axle is transmitted to the rim

which is expected to reduce the near surface tensile stresses pre-

dicted along the inner web’s surface.

Local stresses due to the wheel-rail contact (combination of sta-

tic and dynamic loads) and thermal stresses from tread braking

(and wheel skids) in the tread region have a greater influence on

the fatigue life of the railway wheel than fatigue stresses in the

web region (contact pressure in the tread of the wheel can be as

high as 1944 MPa [7]. Therefore, the high stresses are far more

likely to result in the initiation and propagation of cracks due to fa-

tigue in the tread region [7]. Furthermore, in the literature, tread

defects are far more common in railway wheels and fatigue failures

of railway wheels originate from cracks due to high mechanicaland thermal fatigue loads in the region of the tread [7].

Fig. 10. Predicted temperature histories for Case A (constant HTC) and Case B

(surface dependent HTC) in the web region of LCBM railway wheel plotted againsttemperature measurements at the thermocouple Location #1 in Fig. 5.



7/8

3.4. Evaluation of residual stress measurements for LCBM railway

wheel

The results of the radial cut made on the LCBM railway wheel

quenched using Case H quenching conditions showed a general

net closure of about 0.5 mm in the entire tread, flange and front

and back faces of the rim, confirming the presence of the desiredcompressive residual hoop stress in the as-cast experimental LCBM

wheel (Fig. 12). This work has demonstrated the successful devel-

opment of a quenching process for LCBM railway wheels. Quench-

ing the web region at the highest spray intensity imparts high

compressive residual stresses in the rim of LCBM railway wheel

and significantly reduces the risks of initiation and propagation

of fatigue cracks. Further work is being conducted to optimise

the quenching process for LCBM railway wheels through adjust-ments of the quenching periods to increase the level of compres-

Case L Case H Legend

Temperature

distribution

at the end of

1st

quenching

stage

(a) (b)

Martensite

distribution

at the end of

1st

quenchingstage

(c) (d)

Final Min.

Principle

residual

stress

distribution

in the web

(e) (f)

Final Min.Principle

residual

stress

distribution

in the rim

(g) (h)

Fig. 11. Temperature histories, martensite phase distribution and residual stress distribution during quenching process of LCBM railway wheels under quenching condition

Case L and Case H respectively.


http://-/?-http://-/?-


8/8

sive residual stresses in the rim, and the option of using another

grade of LCBM steels with a lower carbon content is being

evaluated.

4. Conclusions

A thermo-mechanical finite element model has been developedto study the effects of quenching conditions on the formation of

residual stresses in new low carbon bainitic–martensitic (LCBM)

railway wheels. Full size railway wheels quenching experiments

have been carried out for different quenching spray intensities and

values of heat transfer coefficients have been determined and ap-

plied in the FE model to predict the thermo-mechanical behaviour

during quenching of LCBM railway wheels. Quenching at the highest

spray coolant intensity is recommended to promote favourable com-

pressive residual stresses in the rim of LCBM railway wheels.

References

[1] http://www.annualreport.qrnational.com.au/.[2] C. Lonsdale, D. Stone, Some possible alternatives for longer-life locomotives

wheels, in: Proceedings of IMECE2002, New Orleans, Louisiana.[3] P.J. Mutton, R. Boelen, Trans. Inst. Eng., Aust. Mech. Eng. 15 (1989) 110–117.[4] T. Constable, R. Boelen, E. Pereloma, Improved performance for wheels through

the development of low carbon bainitic–martensitic steels, Conference onRailway Engineering, Brisbane, 2006.

[5] C. P. Lonsdale, L. W. Ishler, Development of locomotive wheels for improvedadhesion and life, in: Proceedings of the 2001 IEEE/ASME Joint RailroadConference, 2001, pp. 29–34.

[6] M.R. Zhang, Mater. Sci. Technol. 23 (8) (2007) 970–974.[7] D. Peng, R. Jones, T. Constable, S.N. Lingamanaik, B.K. Chen, Theor. Appl. Fract.

Mech. 57 (2012) 1–13.[8] AAR Specification M-107/M-208, Wheels, Carbon Steel, Association of

American Railroads, (2011).[9] BS 5892-3, Railway rolling stock materials. Specification for monobloc wheels

for traction and trailing stock, British Standards Institution, 1992.[10] EN 13262, Railway Applications – Wheelsets And Bogies – Wheels – Product

Requirement, 2004.[11] D.H. Stone, K. Sawley, D. Kelly, W. Schust, Wheel/rail materials and

interaction: North American heavy haul practices, Proceedings, InternationalHeavy Haul Association STS Conference, Moscow, Russia, 1999, pp. 155–168.

[12] P J Mutton, M.R.L., Improving the safety of railway wheels through non-destructive measurement of residual stresses, Conference on, RailwayEngineering, 2004.

[13] S.N. Lingamanaik, B.K. Chen, J. Mater. Process. Technol. 211 (2011) 1547–1552.

[14] C. Khulman, M. Gallagher, The significance of material properties on stressesdeveloped during quenching of railwayroad wheels, in: Proceedings of the1988 Joint ASME/IEEE Railwayroad Conference, Chicago, Illinois, 1998.

[15] J. Lee, ISIJ Int. 49 (2009) 1920–1925.[16] R. Wendelstorf, K.H. Spitzer, J. Wendelstorf, Int. J. Heat Mass Trans. 51 (2008)

4892–4901.[17] N. Karwa, S.R. Kale, P.M.V. Subbarao, Exp. Thermal Fluid Sci. 32 (2007) 571–

579.[18] C. Krause, E. Wulf, F. Nurnberger, F.W. Bach, Forsh Ingenieurwes 72 (2008)

163–173.[19] S-J Chen, A.A. Tseng, Int. J. Heat Fluid Flow 13 (1992) 358–369.[20] H.P. Li, G.Q. Zhao, S.T. Niu, Y.G. Luan, Finite Elem. Anal. Des. 42 (2006) 1087–

1096.[21] V.C. Prantil, M.L. Callabresi, J.F. Lathrop, G.S. Ramaswamy, M.T. Lusk, J. Eng.

Mater. Technol. 125 (2003) 116–124.[22] D. Bammann, V. Prantil, A. Kumar, J. Lathrop, D. Mosher, M. Callabresi, M. Lusk,

G. Krauss, H. -J. Jou, W. Elliott, Jr., G. Ludtka, W. Dowling, Jr., D. Nikkel, T. Lowe,D. Shick, Development of a carburizing and quenching simulation tool: amaterial model for low carbon steels undergoing phase transformations, in:American Society of Metals 2nd International Conference on Quenching andControl of Distortion, 1996.

[23] Z. Li, B.L. Ferguson, A.M. Freborg, Mater. Sci. Technol. 2 (2004) 219–226.[24] V.S. Warke, R.D. Sisson, M.M. Makhlouf, Mater. Sci. Eng. A. 518 (2009) 7–15.[25] B.L. Ferguson, Z. Li, A.M. Freborg, Comput. Mater. Sci. 34 (2005) 274–281.[26] B.L. Ferguson, Z. Li, A.M. Freborg, G. Petrus, Adv. Mater. Process. (2000) H31–

H36.[27] G. Comini, S.D. Guidice, R.W. Lewis, O.C. Zienkiewicz, Int. J. Numer. Methods

Eng. 8 (1974) 613–624.[28] E.L. Wilson, R.E. Nickell, Nucl. Eng. Des. 4 (1996) 276–286.[29] G. Aguirre-Ramirez, J.T. Oden, Int. J. Numer. Methods Eng. 7 (1973) 345–355.[30] R.E. Beckett, S.-C. Chu, J. Heat Trans. (1973) 126–129.

Fig. 12. Net negative displacement (indicative of the presence of compressive

residual stresses) was measured on the outer rim of the LCBM railway wheel after a

cut made was made through the rim. Yellow markers indicate the location of one

the gauge marks. (For interpretation of the references to colour in this figure legend,

the reader is referred to the web version of this article.)

http://www.annualreport.qrnational.com.au/http://refhub.elsevier.com/S0927-0256(13)00211-5/h0005http://refhub.elsevier.com/S0927-0256(13)00211-5/h0010http://refhub.elsevier.com/S0927-0256(13)00211-5/h0010http://refhub.elsevier.com/S0927-0256(13)00211-5/h0010http://refhub.elsevier.com/S0927-0256(13)00211-5/h0015http://refhub.elsevier.com/S0927-0256(13)00211-5/h0015http://refhub.elsevier.com/S0927-0256(13)00211-5/h0020http://refhub.elsevier.com/S0927-0256(13)00211-5/h0020http://refhub.elsevier.com/S0927-0256(13)00211-5/h0020http://refhub.elsevier.com/S0927-0256(13)00211-5/h0025http://refhub.elsevier.com/S0927-0256(13)00211-5/h0025http://refhub.elsevier.com/S0927-0256(13)00211-5/h0030http://refhub.elsevier.com/S0927-0256(13)00211-5/h0035http://refhub.elsevier.com/S0927-0256(13)00211-5/h0035http://refhub.elsevier.com/S0927-0256(13)00211-5/h0040http://refhub.elsevier.com/S0927-0256(13)00211-5/h0040http://refhub.elsevier.com/S0927-0256(13)00211-5/h0045http://refhub.elsevier.com/S0927-0256(13)00211-5/h0045http://refhub.elsevier.com/S0927-0256(13)00211-5/h0045http://refhub.elsevier.com/S0927-0256(13)00211-5/h0050http://refhub.elsevier.com/S0927-0256(13)00211-5/h0055http://refhub.elsevier.com/S0927-0256(13)00211-5/h0055http://refhub.elsevier.com/S0927-0256(13)00211-5/h0060http://refhub.elsevier.com/S0927-0256(13)00211-5/h0060http://refhub.elsevier.com/S0927-0256(13)00211-5/h0065http://refhub.elsevier.com/S0927-0256(13)00211-5/h0070http://refhub.elsevier.com/S0927-0256(13)00211-5/h0070http://refhub.elsevier.com/S0927-0256(13)00211-5/h0075http://refhub.elsevier.com/S0927-0256(13)00211-5/h0080http://refhub.elsevier.com/S0927-0256(13)00211-5/h0080http://refhub.elsevier.com/S0927-0256(13)00211-5/h0085http://refhub.elsevier.com/S0927-0256(13)00211-5/h0085http://refhub.elsevier.com/S0927-0256(13)00211-5/h0085http://refhub.elsevier.com/S0927-0256(13)00211-5/h0090http://refhub.elsevier.com/S0927-0256(13)00211-5/h0090http://refhub.elsevier.com/S0927-0256(13)00211-5/h0095http://refhub.elsevier.com/S0927-0256(13)00211-5/h0100http://refhub.elsevier.com/S0927-0256(13)00211-5/h0100http://refhub.elsevier.com/S0927-0256(13)00211-5/h0095http://refhub.elsevier.com/S0927-0256(13)00211-5/h0090http://refhub.elsevier.com/S0927-0256(13)00211-5/h0085http://refhub.elsevier.com/S0927-0256(13)00211-5/h0085http://refhub.elsevier.com/S0927-0256(13)00211-5/h0080http://refhub.elsevier.com/S0927-0256(13)00211-5/h0080http://refhub.elsevier.com/S0927-0256(13)00211-5/h0075http://refhub.elsevier.com/S0927-0256(13)00211-5/h0070http://refhub.elsevier.com/S0927-0256(13)00211-5/h0065http://refhub.elsevier.com/S0927-0256(13)00211-5/h0060http://refhub.elsevier.com/S0927-0256(13)00211-5/h0060http://refhub.elsevier.com/S0927-0256(13)00211-5/h0055http://refhub.elsevier.com/S0927-0256(13)00211-5/h0055http://refhub.elsevier.com/S0927-0256(13)00211-5/h0050http://refhub.elsevier.com/S0927-0256(13)00211-5/h0045http://refhub.elsevier.com/S0927-0256(13)00211-5/h0045http://refhub.elsevier.com/S0927-0256(13)00211-5/h0040http://refhub.elsevier.com/S0927-0256(13)00211-5/h0040http://refhub.elsevier.com/S0927-0256(13)00211-5/h0035http://refhub.elsevier.com/S0927-0256(13)00211-5/h0035http://refhub.elsevier.com/S0927-0256(13)00211-5/h0030http://refhub.elsevier.com/S0927-0256(13)00211-5/h0025http://refhub.elsevier.com/S0927-0256(13)00211-5/h0020http://refhub.elsevier.com/S0927-0256(13)00211-5/h0020http://refhub.elsevier.com/S0927-0256(13)00211-5/h0015http://refhub.elsevier.com/S0927-0256(13)00211-5/h0010http://refhub.elsevier.com/S0927-0256(13)00211-5/h0010http://refhub.elsevier.com/S0927-0256(13)00211-5/h0010http://refhub.elsevier.com/S0927-0256(13)00211-5/h0010http://refhub.elsevier.com/S0927-0256(13)00211-5/h0005http://www.annualreport.qrnational.com.au/

lingamanaik2013_2

Documents