Accepted Manuscript

The low-cycle fatige life prediction method for online monitoring of steam tur-
bine rotors

Mariusz Banaszkiewicz

PII: S0142-1123(18)30080-X
DOI: https://doi.org/10.1016/j.ijfatigue.2018.02.032
Reference: JIJF 4597

To appear in: International Journal of Fatigue

Received Date: 23 December 2017

Revised Date: 19 February 2018
Accepted Date: 25 February 2018

Please cite this article as: Banaszkiewicz, M., The low-cycle fatige life prediction method for online monitoring of
steam turbine rotors, International Journal of Fatigue (2018), doi: https://doi.org/10.1016/j.ijfatigue.2018.02.032

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
 THE LOW-CYCLE FATIGE LIFE PREDICTION METHOD
FOR ONLINE MONITORING OF STEAM TURBINE ROTORS

MARIUSZ BANASZKIEWICZ

The Szewalski Institute of Fluid-Flow Machinery Polish Academy of Sciences, Energy Conversion Department, Gdańsk, Poland
e-mail: mbanaszkiewicz@imp.gda.pl

The paper presents a simple method for the low-cycle fatigue life monitoring of steam
turbine rotors. The method makes use of the equivalent strain energy density rule of notch
stress-strain analysis proposed by Molski and Glinka. The method is applied to low-cycle
fatigue analyses of steam turbine rotors operating at non-isothermal conditions. A strain
energy correction factor is introduced to include the distortion strain energy density. As-
sessment of the proposed method was done by comparing their predictions with the strain
amplitudes obtained from the finite element analysis employing elasitc-plastic material
model. The inclusion of energy correction factor obtained from the elastic-plastic finite
element analysis significantly improves the accuracy of strain amplitudes and fatigue life
predictions. The strain calculation algorithm is based on an analytical solution of the
equivalent strain energy equation what makes it very useful for real time fatigue calcula-
tions.

Keywords: notch stress-strain, low-cycle fatigue, turbine rotors

1. Introduction

Steam turbine rotors are the most critical components of power plant units and experience
failures due to various damage mechanisms [1]. High-temperature rotors can fail due to
cracking under creep and/or fatigue conditions and numerous examples of cracks due to these
mechanisms have been reported over the past decades. In all these cases, high cyclic stresses
and high temperature were involved in crack initiation leading to rotor failure.
Low-cycle fatigue cracking of turbine rotors is mainly related to transient thermal stresses
occurring during turbine start-ups and shutdowns [2]. High thermal and mechanical loads
generate stresses exceeding the material yield stress at stress concentration areas, and their
repetitive occurrence may lead to fatigue crack initiation after 1000 cycles or less.
Steam turbines are more and more frequently operated in cyclic duty due to the increased
share of renewable energies in power generation [3]. The cyclic operation under variable
thermal loading generates high thermal stresses which bring about fatigue damage and
cracking at notches. In order to monitor and control the fatigue damage in real time, fast,
accurate and robust methods should be used in online monitoring systems.
The first step of fatigue life determination under stress concentration should include the
definition of local stresses and elastic-plastic strains in the notch root [4]. Non-linear finite
element modelling can be used to accurately predict the stress-temperature histories and
cyclic strain evolution. However, this approach becomes impractical when long stress
histories are to be analyzed or real time calculations performed, and consequently, analytical
stress-strain correction methods have to be considered [5-9].
The most widely used methods for elasto-plastic strain and stress analysis are the Neuber

1
rule [10] and Glinka-Molski method [11] originally developed for isothermal conditions. The
methods have been used in determining stress-strain histories and fatigue life of components
subjected to multiaxial non-proportional loading at constant temperature [12-15]. The Neuber
rule has recently been used to predict thermal stresses in turbine casings [16] and stress-strain
evolution in a notched plane specimen at non-isothermal conditions [17].
It is well recognized that in most cases the Neuber rule overestimates the notch tip stresses
and strains, while the Glinka-Molski method tends to underestimate the notch stress/strain
[18]. However, both methods were found to significantly underestimate the distributions of
stress and strain ahead of a notch tip [19].
This paper proposes a simple practical method for online monitoring of the low-cycle
fatigue life of steam turbine rotors. The algorithm is based on the Glinka-Molski equivalent
strain energy equation with a correction for distortion part of the strain energy density. Such a
formulation enables to obtain analytical solution for strain which is accurate enough for real
time simulations. The high accuracy is achieved by making use of the results of finite element
calculations with elastic-plastic material model.

2. Thermal fatigue cracking of steam turbine rotors

2.1. Cracking locations

Non-uniform loading together with frequent start-ups and shutdowns of steam turbines lead to
crack initiation in their components due to thermal fatigue. Steam turbine rotors are prone to
thermal fatigue cracking due to their specific design features and non-uniform thermal loading.
The rotors have a thick control stage disc operating at highest temperature and pressure. The
transition radii between the disc and shaft are stress raisers where cracks may initiate and
grow. The rotor ends are the areas of changing diameters where additionally heat grooves are
machined in order to reduce the negative influence of non-uniform temperature on the impulse
rotor shape. Similar grooves are present in the inter-stage glands of impulse rotors and all
these regions are the locations of potential cracking due to thermal fatigue. Reaction rotors
have deep blade grooves where rotating blades are mounted. These grooves introduce high
stress concentration at the bottom surface where thermal fatigue cracks are often found [20].
There are numerous examples of rotor fatigue cracking reported in literature [21-26]. The areas
where cracks are most frequently found are schematically shown in Fig. 1. Circumferential
cracks due to transient thermal stresses were typically found in 3 areas of rotors:
a) heat grooves in impulse rotors and blade grooves in reaction rotors
b) transition of diameters in zones of maximum temperature
c) control stage disc to shaft transition radius

An example of thermal fatigue cracks in front gland heat grooves is shown in Fig. 2. Cracks
reaching 1.8 mm depth were found in six heat grooves of the front end gland of a 200 MW
intermediate pressure turbine rotor after 126 000 operating hours and 507 start-stop cycles.
The number of cycles to cracking of heat grooves were in the range 500-1900 with maximum
crack length reaching 15 mm.

2
 a)

b)

Fig. 1. Typical locations of fatigue cracking in impulse (a) and reaction (b) turbine rotors.

a) b)

Fig. 2. Fatigue cracks in heat grooves of intermediate pressure rotors detected after a) 1000 cycles,
b) 1800 cycles.

The second area of potential cracking are transitions of diameters in the zones of
maximum temperature. Typical crack depth in this location was about 5 mm and cracking in
120 MW turbines occurred after 300÷600 starts. These cracks had to be removed by turning
and the shape of transition between two diameters was changed what resulted in a significant
reduction of stress concentration factor.
Circumferential cracks of similar length were found in the control stage disc to shaft
transition radius of 55 MW turbines. In this area longer operation time and higher number of

3
starts are required to initiate cracking. The machines accumulated 2000-2900 start-up cycles
until the cracks were detected.
The presented examples clearly show that thermal fatigue cracks may be initiated in
various areas after considerably different number of starts ranging from 300 to 3000. It
indicates that crack initiation life is sensitive to thermal stresses generated during start-ups and
shutdowns, and online monitoring of stresses and fatigue damage is required in order to ensure
safe operation.

2.2. Crack examination

Metallurgical examinations of cracks are ususally performed as part of the root cause analysis
of rotor cracking. In order to investigate the nature of rotor cracking, the cracked locations are
machined to extract a ring of material that contains cracks [22]. Typical macro view of a
thermal fatigue induced crack in a heat groove is shown in Fig. 3. Circumferential cracks are
initiated at the groove bottom surface and propagate radially downwards into the rotor
material. The available analyses concluded that the observed cracking was either thermal
fatigue induced creep crack growth, predominantly by intergranular mechanism [22], or
transgranular thermal fatigue cracking. Hardness measurements indicated that the rotor
material was of the required strength level and the rotor microstructure was tempered bainite
which is normal for the 1CrMoV creep resistant steel.

Fig. 3. Macro view of a thermal fatigue induced crack in front gland heat groove.

Another examined case of rotor fatigue cracking was for high pressure/intermediate pressure
rotor with a circumerential crack present in the transition radius between impulse wheel and
shaft [21]. The primary cracking resulted from fatigue damage which initiated at multiple
locations around the cricumference of the rotor. Examination revealed no evidence of any
creep damage in any of the areas examined by SEM fractography and optical metallography.
The condition of the rotor microstructure was normal with no evidence that material
discontinuities played any role in the development of cracking. The microstructure was
tempered martensite typical for the 1CrMoV steel. Visual examination of crack surface using
a microscope revealed clear evidence of “beach marks” and “ratchet marks associated with the
crack initiation and propagation, typical of multiple-initiation site fatigue cracking. Typical
SEM fractographic features of the rotor material in front of the crack tip are shown in Fig. 4.

4
 Fig. 4. SEM fractographic features of the material in front of the fatigue crack tip.

3. Fatigue life analysis algorithm

Finite element modelling is commonly used to analyse steam turbine components at the
design phase. However, this technique is impractical when long stress/strain histories are to be
analysed in real time using non-linear elastic-plastic material models. For such purposes,
simplified elastic stress models and notch stress/strain correction methods are needed in order
to ensure robust and accurate results. Figure 5 presents an algorithm proposed for steam
turbine rotors fatigue life monitoring under thermo-mechanical loads. A specific feature of the
algorithm is the inclusion of a strain energy correction factor derived from elastic and elastic-
plastic finite element analysis in strain amplitude evaluation. Temperature and elastic stress
components are computed online based on the measured load history. This analysis can be
performed with high accuracy using various methods, like for example Green’s function
method and Duhamel’s integral [6, 27]. In steam turbines operation it is possible to
distinguish representative transient events like cold, warm and hot start-ups, and shutdowns
which are uniquely defined in design start-up diagrams. Alternatively, for turbines being in
service for some period of time, typical start-ups from different initial thermal states can be
selected from the available operating history. Start-ups and shutdowns are the transient events
generating highest thermal stresses in turbine components and they form the most damaging
fatigue cycles requiring online monitoring. These representative transients are then adopted as
reference load cycles for which linear elastic and non-linear elastic-plastic finite element
analyses are conducted. The main goal of the analyses is to determine the ratio of strain
energies calculated with equivalent stress and strain on one hand, and with stress and strain
components on the other hand. The ratios are computed in the elastic and elastic-plastic FEA
and are necessary for calculating the energy correction factor. This factor is further used in
combination with temperature and elastic stresses to compute the elastic-plastic strain
amplitude using the Glinka-Molski method. Finally, the number of cycles to crack initiation is
obtained employing the strain-life method and using the low-cycle fatigue data of the rotor
material.

5
 Real load history Reference load cycle

Temperature and Elastic and elastic-plastic

elastic stress history stress/strain for reference cycle

Energy correction factor

Online elastic
Off line elastic-plastic

Multiaxial notch
correction

Elastic-plastic
strain amplitude

Material fatigue Number of cycles

property data to crack initiation
Online elastic-plastic

Fig. 5. Algorithm for low-cycle fatigue life monitoring of steam turbine rotors.

4. Mathematical models for finite element analysis

The problem under consideration is cyclic heating-up and cooling-down of a steam turbine
rotor taking place during start-up and the subsequent shutdown. Rotor heating and cooling is
caused by the hot steam flow through the turbine with varying temperature, pressure and
velocity. Non-stationary heat transfer takes place via forced convection and the thermal load
is a primary load of the rotor. The rotor rotates with a constant rotational speed  in steady-
state operation. Steam pressure and rotational body forces due to rotation are also considered
in the model. Non-uniform temperature distribution in the rotor induces thermal stresses due
to thermal expansion.

6
4.1. Thermoelasticity problem

Knowing the axisymmetric temperature field in the rotor obtained by solving the Fourier-
Kirchhof equation [28], a solution of the problem of stress state induced by temperature can
be obtained. The equilibrium conditions written for stresses are transformed taking into
account the relations between stresses, strains and displacements, to obtain differential
equations with unknown displacements u and w in the r and z direction, respectively [29]:

(1)

where the following definitions were employed:

- Laplacian

- volumetric strain

In Eq. (1)  is the Poisson’s ratio and  is the thermal expansion coefficient.
The system of partial differential equations (1) is solved by means of the finite element
method [30] with the following boundary conditions:
- at the outer cylindrical surfaces of the discs
– pressure due to centrigual forces of blades (2)
- at the outer surfaces in contact with steam
– steam pressure acting normal to the surface (3)
- at the left end face
at (4)
The relations between strains and displacements for an axi-symmetric body in the cylindrical
co-ordinate system are expressed as [29]:

(5)

Stress components are obtained from the solution of the constitutive equation of linear
thermoelasticity [28] and for an axi-symmetric body in the cylindrical co-ordinate system are
given as [29]:

(6)

7
where
- shear modulus, E – Young’s modulus

The Eqs. (6) are known as the Duhamel-Neumann relations and have a fundamental
importance in thermal stress analysis within the validity of Hooke’s law [31].

4.2. Plasticity model

There is a number of incremental plasticity models developed to estimate constitutive material

behaviour [32-35], including those capable of predicting transient hardening response.
Advanced cyclic plasticity models require many parameters, which have to be determined
from cyclic tests, and this makes these models inappropriate for practical engineering use.
In order to assess the strain amplitudes determined analytically basing on a solution of the
thermoelasticity problem by comparing them with the elastic-plastic model predictions,
numerical calculations were performed using Abaqus [36] employing elastic-plastic material
model with the Prager-Ziegler linear kinematic hardening. The plasticity surface is defined by
the Huber-Mises-Hencky function [37]:
(7)
where is the equivalent stress related to the backstress defining translation of
the yield surface. The yield function is traditionally defined as follows

(8)

where is a deviatoric part of the stress tensor , while is a deviatoric part of the
backstress tensor .
The linear kinematic hardening model assumes the associated plastic flow rule in the form
[38]:
(9)

where is a plastic flow rate and denotes here a plastic work. In the linear kinematic
hardening model, translation of the yield surface is described by the backstress tensor
whose evolution in time is determined by the Prager-Ziegler linear hardening law [39]
(10)
where is a positive scalar coefficient.

8
 5. Numerical investigation of transient stress and strain states in rotors

A typical steam turbine rotor in which thermal fatigue cracking was frequently found has
been adopted for the analysis of transient stress and strain states. First, thermoelastic strains
and stresses were computed for the representative thermal transients including cold (CS),
warm (WS) and hot starts (HS) as well as shutdown (SD) which form the closed fatigue
cycles. Instantaneous temperature distribution in the rotor during start-up from a cold state is
presented in Fig. 6. Radial temperature gradients persist in the rotor hot regions, and they are
continuously transforming into axial gradients when moving towards the ends of the rotor.
These thermal gradients together with mechanical loads due to steam pressure and centrifugal
forces produce thermomechanical stresses shown in Fig. 7. Very high thermoelastic stresses
occur at the bottom of circumferential heat grooves indicated in the enlarged box. They result
from the combination of radial gradients of temperature and very high stress concentration
due to the geometrical notch. Although all the grooves have the same geometry, meaning the
same notch factor, the elastic stresses are different. A summary of stress maxima in all six
grooves (G1÷G6) and transition radius of the control stage disc (CSD) for the three start-ups
and shutdown is given in Table 1. The differences in stress result from different axial position
of the grooves and the corresponding effect of the side free surfaces. The elastic stress ranges
determine the elastic-plastic stress and strain amplitudes used in fatigue life prediction.

Fig. 6. Instantaneous temperature distribution in rotor during cold start.

Fig. 7. Elastic Huber-Mises-Hencky stress distribution in rotor during cold start (high stress
concentration in grooves shown in zoom) .

Table 1. Maximum elastic equivalent stresses [MPa] for start-ups and shutdown.
G1 G2 G3 G4 G5 G6 CSD
Cold start 287,4 369,0 409,7 438,9 448,1 401,6 354,8
Warm start 408,1 489,5 523,5 550,6 554,1 495,4 457,3
Hot start 488,3 524,6 517,7 499,7 470,8 408,9 422,8
Shutdown 351,8 405,7 407,8 393,6 364,6 302,5 314,9

In fatigue life analysis, not only the stress or strain amplitudes are important, but also the
loading type and stress ratio influence the selection of a calculation method. The loading type
is determined by analyzing the variation of deviatoric stress components at the notch tip
during the load cycles. The results for the highly loaded heat groove 4 are shown in Fig. 8
which presents the variation of axial (DSz) and circumferential (DS) deviatoric stresses as a

9
function of radial deviatoric stress (DSr) obtained from elastic and elastic-plastic analysis.
The axial component of stress deviator changes almost linearly with the radial component
both in the elastic and elastic-plastic model. Similar behaviour is observed for the
circumferential component which is much lower than the axial one and exhibits a slightly
worse proportionality. The deviatoric stresses can thus be considered to be in nearly fixed
proportions and the use of Hencky’s equations for stress/strain analysis is justified.

Fig. 8. Deviatoric stresses variations in 4th heat groove in cold start cycle.

The strain-life method can be successfully applied to multiaxial fatigue damage modelling
in the low cycle fatigue regime where plastic deformation is significant [40]. In the presence
of large plastic strain the mean stress has relatively less effect on the fatigue damage, and it is
well recognized that compressive mean stress has beneficial effect on fatigue life [41]. The
low cycle fatigue data are usually presented for a constant stress ratio R and this parameter
was assessed for typical stress cycles shown in Figs. 9-11. They present variations of stress
components and equivalent stress in groove 4 for a cold, warm and hot start cycle obtained
from elastic and elastic-plastic solution. The stress state at the groove bottom is bi-axial with
compressive stresses present during heating-up (start-up phase) and tensile during cooling-
down (shutdown phase). During the start-up, the notch tip is thus subjected to bi-axial
compression which changes into bi-axial tension during the shutdown. Both at stress minima
and maximum the axial component is c.a. as twice as the circumferential component. The
stress ratios are in all cases lower than minus 1 and higher values are observed in elastic-
plastic state. The values of equivalent stress are always between the absolute value of axial
and circumferential stress components both in elastic and elastic-plastic state.
The variation of mechanical strain components at the groove bottom for the same load
cycles are shown in Figs. 12-14. The strain state at the notch tip is tri-axial with all normal
strain components different than zero. Axial strain at elastic state is c.a. as twice as the radial
strain and they are of opposite sign while the circumferential strain is of the same sign as the
axial strain component. All the mechanical strain components change sign when thermal
loading changes from heating-up to cooling down.
Two major consequencies follow from the state of stress and strain investigated above:
1) Hencky’s relation can be used for cyclic stress-strain evaluation which is justified by
the nearly proportional behaviour of deviatoric stress components
2) Strain-life approach with fatigue data for R=-1 can be safely used for fatigue life

10
 prediction of notched rotors which is justified by the compressive mean stress of the
start-stop cycle

Fig. 9. Stress variations in heat groove43 in cold start cycle.

Fig. 10. Stress variations in heat groove 4 in warm start cycle.

Fig. 11. Stress variations in heat groove 4 in hot start cycle.

11
 Fig. 12. Mechanical strain variations in heat groove 4 in cold start cycle.

Fig. 13. Mechanical strain variations in heat groove 4 in warm start cycle.

Fig. 14. Mechnical strain variations in heat groove 4 in hot start cycle.

During transient heating and cooling, the thermo-mechanical stresses reach the material
yield stress at the bottom of the grooves and produce significant plastic deformation. The
level of plastic strain plays an essential role in lifetime of a part [42, 43] As it is seen from
Fig. 15, the zone of plastic deformation is highly localized and confined to the groove bottom
surface. Peak plastic strains occur at the groove surface near its axis of symmetry where
highest elastic stresses are generated. The shape of the plastic zone is the same in all grooves
but its size is different due to the different values of stresses generated by thermal expansion.
The plastic zone at the analysed groove of radius r = 1.5 mm extends axially to maximum c.a.

12
2 mm and radially to c.a. 0.7 mm below the groove bottom. The small size of the plastic zone,
comparable with the notch radius, justifies the use of analytical notch stress-strain correction
method assuming small scale plastic yielding.

Fig. 15. Equivalent plastic strain distribution in heat groove 1 (left) and 4 (right) after 3rd hot start
cycle.

6. Notch stress-strain correction

6.1. Glinka-Molski method

Thermomechanical stresses at notches calculated using the linear-elastic material model

can attain very high values, significantly exceeding the material yield stress as it is shown in
Figs. 9-11. In traditional approach to fatigue life estimation, in order to determine the number
of cycles to crack initiation, it is necessary to know the stresses or strains at the notch tip. For
describing the elasto-plastic material response based on the elastic solution, the most
commonly used methods are Neuber’s rule and Glinka-Molski equivalent strain energy
density method. Both methods were derived for a simple state of stress where only one stress
component is present at notch tip. For multi-axial state of stress and strain, the methods have
been extended either by defining the strain energy using stress and strain components or using
the equivalent values of stress and strain tensor [15, 19, 44]. For the Glinka-Molski method
the two definitions result in:
(11)

(12)
where: – stress component, – strain component, – stress component obtained from
elastic solution, – strain component obtained from elastic solution, – equivalent stress,
– equivalent strain, – equivalent stress obtained from elastic solution, –
equivalent strain obtained from elastic solution.
The equivalent stress is defined by the deviatoric stress components [37]:

(13)

and the equivalent strain is given by the deviatoric strain components with energy
conjugate strain definition:

(14)

Under multiaxial stress and proportional loading, the equivalent strain can be expressed as
a sum of equivalent elastic strain and equivalent plastic strain [45]
(15)

13
 The equivalent elastic strain is defined as
(16)
and the equivalent plastic strain in the linear kinematic model can be expressed as

(17)
where is material yield stres, and C is a kinematic hardening parameter.

6.2. Strain energy density correction

The strain energy formulation based on the equivalent stress and strain was used in [19] for
notch tip fields analysis and it was analytically prooved that for the power hardening material
the maximum difference between the two formulations of strain energy density is less than
5% under plane stress condition and increases up to 17% in plane strain condition. Although
the formulation based on the stress and strain components is theoretically precise as it
includes both the dilatation and distortional part of the strain energy density, the formultion
employing the equivalent stress and strain is easier to compute. This feature makes it very
attractive for application in real-time computations of stresses, strains and fatigue life of
notched components. In particular, when bi-linear material behaviour is assumed, the
analytical solutions for equivalent stress and strain in elastic-plastic condition can be obtained
with an input of elastic stress or strain only.
The Neuber rule based on the equivalent formulation of strain energy density has been
used by many authors for elastic-plastic notch tip strain analysis under different loading
conditions [44, 46]. The first application of the Glinka-Molski method to the strain analysis of
a steam turbine rotor revealed that it underpredicts the notch tip strain amplitudes by
maximum 18%. At the same time, the Neuber rule overpredicted the equivalent strain
amplitudes by 20% [46]. This resulted in fatigue life overestimation by a factor of 3 in the
first case, and underestimation by 0.33 in the second case. This behaviour is consistent with
the general observation that in most cases Neuber’s rule overestimates the notch tip stresses
and strains, while the Glinka-Molski method tends to underestimate the notch stress/strain
[19]. Furthermore, it is well recognized that the accuracy of the methods depends strongly on
the level of the nominal stress relative to the material’s yield stress, the material constitutive
law, the stress concentration factor as well as the nature of the stress state.
The reported inaccuracy of the prediction of strain amplitude and fatigue life can be partly
attributed to the adopted formulation of the strain energy density and the resulting neglection
of the dilatation energy. For the linear elastic material, the strain energy density is given by
the relation [19]

(18)

where the equivalent stress is given by Eq. (13) and the mean stress is defined as:

(19)

The first term in Eq. (18) describes the distortional part and the second term is the dilatation
part of the strain energy density.

14
The elastic strain energy formulated using the equivalent quantities is as follows:

(20)

and neglects the distortional part.

For the bi-linear elastic-plastic material with plastic part described by Eq. (17), the elastic-
plastic strain energy density is defined as

(21)

while the equivalent definition in elastic-plastic condition corresponding to Eq. (20) is

(22)

The two definitions differ by the dilatation term, but the difference is small for low
hydrostatic stress .

The ratio of strain energies in elastic and elastic-plastic conditions were

computed using the results of finite element analysis presented in section 5. The results for
elastic and elastic-plastic conditions are presented in Fig. 16a and 16b, respectively. In case of
the linear elastic material model, the ratios are practically the same for different loading
conditions (start-up type) and remain relatively constant at different grooves. During heating-
up the strain energy ratio is on the level of 1.45, and for cooling down slightly drops to 1.35.
For the bi-linear elastic-plastic material, the strain energy ratios are clearly lower and change
more from groove to groove. A higher scatter of calculation points is also seen as compared
with the linear elastic material model. The general observation is that the ratios of elastic
strain energy density are visibly higher than those obtained for the elastic-plastic energy what
may have an effect on the predicted strains and stresses in the elastic-plastic condition. In
order to include this effect, the strain energy ratios are considered in Eqs. (11) and (12)
resulting in the modified Glinka-Molski rule with a strain energy correction factor cE:

(23)

where cE is given by

(24)

This factor describes the different contribution of the dilatation energy in elastic and
elastic-plastic condition.

15
 a) b)

Fig. 16. Strain energy ratios in elastic (a) and elastic-plastic (b) condition.

7. Modelling of notch strain amplitude

Strain amplitude at the notch tip is used as a damage parameter for the low-cycle fatigue
life prediction of rotors with heat grooves. Hysteresis loops for stress-strain were a basis for
strain amplitude determination in the elastic-plastic condition. Typical plots of the hysteresis
loops for the cold, warm and hot start-up cycle obtained from the elastic-plastic finite element
analysis are shown in Fig. 17. They describe complete cycles including the start-up and
shutdown phase. The largest strain range and the most regular shape of the hysteresis loop
was found for the hot start cycle, while the lowest range was obtained in the cold start. It
should be noticed that the differences in strain ranges are due to the start-up sections
(compressive stress) of the curves, while the shutdown sections (tensile stress) are practically
the same as there is no difference in the shutdown run for different start-ups.

Fig. 17. Stress-strain hysteresis loops for a cold, warm and hot start-up calculated using elastic-plastic
material model.

In the case of linear kinematic hardening and with the hypothesis of proportional loading
[47], equation (23) can be transformed into a quadratic equation using the strain definitions
given by equation (16) and (17):

16
 (25)

The above equation can be solved analytically using the determinant method to obtain the
equivalent strain history in the elastic-plastic state which is further used to evaluate the
strain amplitude of the analyzed cycle. This is a very important advantage of the strain
energy formulation based on the equivalent stress and strain as the need for iterative
numerical solution of the system of nonlinear equations is avoided. This property facilitates
the use of this method in real time calculations where, in addition to accuracy, the calculation
time and reliability are of the most importance. The accuracy of this approach is similar to
that achieved by solving the system of nonlinear Hencky’s equations.
Based on the elastic-plastic hysteresis loops and pseudo-elastic stress variations, the strain
amplitudes were obtained for all the considered grooves and three types of starts using the
Glinka-Molski method with and without strain energy correction. The dilatation strain energy
correction factor cE was obtained from previously performed FE calculations and was
assumed different for heating (start-up) and cooling phase (shutdown). The results are shown
in Fig. 18, which also provides information about the maximum deviation of each correction
method from the elastic-plastic FE predictions. Each point in the plot corresponds to a strain
amplitude computed in one heat groove. All six grooves in which plastic deformation due to
thermal and mechanical stresses occurred were considered in the analyses.
When a stabilised hysteresis loop is attained in the elastic-plastic analysis, the
corresponding strain amplitudes obtained in FEA drop as compared with the first cycle. For
the stabilised cycle, the original method overpredicts the strain amplitude at lower strain
levels and the maximum deviation does not exceed +5%. At higher strain levels the predicted
the strain amplitudes are lower than those obtained from elastic-plastic FEA and the
maximum underestimation error reaches -18%. This results in the scatter of strain amplitude
predictions of (-18%; +5%) with nearly all points located below the FE line.
Fig. 18b presents a similar comparison between the strain amplitudes obtaind using the
corrected Glinka-Molski method and FEA. It is clearly seen that the inclusion of the effect of
dilatation strain energy significantly improves the performance of the Glinka-Molski method.
The maximum errors in strain amplitudes are within the range (-3%; +10%) with more
number of points lying above the FE line. The accuracy of this method in terms of the strain
amplitudes is thus 2-3 times better than that of the original Glinka-Molski rule with energy
formulation based on the equivalent stress and strain. The level of accuracy found in the
analyzed notches is in good agreement with the results presented by Glinka [48] showing that
in most cases the disagreement between the measured and calculated strains was below 10%,
regardless of the material stress-strain curve, stress concentration factor and notch geometry.
a) b)

Fig. 18. Strain amplitudes determined with the original Glinka-Molski method (a) and taking into
account strain energy correction (b).

17
Another form of validation of the corrected Glinka-Molski method is the comparison of strain
amplitude distribution in the grooves with the distribution of crack lengths shown in Fig. 19.
The maximum strain amplitudes were computed in grooves No 2, 3 and 4 (Fig. 19a) where
the deepest cracks were found during non-destructive inspections of two rotors with different
numbers of starts. In these grooves the fatigue cracks are expected to initiate first and further
grow with the highest rate, so the largest crack depths revealed in these grooves confirm the
correctnes of the applied models and methods.
a) b)

Fig. 19. Distribution of strain amplitude (a) and crack length (b) in heat grooves.

8. Fatigue life determination

Critical locations of high temperature steam turbine rotors may be subject to the combined
accumulation of cyclic damage arising from strain transients generated during start-up and
shutdown, as well as creep damage resulting from primary (directly applied) and secondary
(self-equilibrating) stresses during operation [49]. Time fraction method is currently most
widely used to determine the creep-fatigue damage accumulated [50-52] and finds application
to steam turbines lifetime assessment [5]. The creep-fatigue damage for a cycle type i is
determined as follows:
th
th dt 1
Di    (26)
t R  p , T  0
t R  s , T  N  a , T 

where the first term on the right hand side covers the creep damage due to primary loading
(p), the second term describes the creep damage due to secondary loading (s), and the third
one accounts for the fatigue damage.
The creep damage due to secondary loading can be accounted for as part of the fatigue
damage fraction calculation by using fatigue endurance to crack initiation with hold time
N  a , t h 
. The creep damage due to primary loading is negligible in heat grooves as they are
located on the outer surface where centrifugal loads and the resulting stresses are at minimum
[53]. In such circumstances, a simplified version of Eq. 26 can be used:
1
Di 
N  a , t h  (27)

18
The effect of hold time duration on fatigue endurance depends mainly on the temperature and
cycle type. The number of cycles to cracking is reduced more at higher temperatures and at
cycles where the hold time at the maximum tensile strain is introduced into the cycle. The
total damage due to all cycles is calculated using the linear damage accumulation rule by
summing up damage fractions due to each individual cycle.
Fig. 20 presents the estimation errors of the two analytical methods in terms of the number of
cycles to crack initiation relative to the number of cycles predicted with the FEA strain
amplitudes. The largest errors are found for the hot starts where the calculated strain
amplitudes were largest and the numbers of starts lowest. Using the original Glinka-Molski
method the predicted number of cycles is within a factor of 3 as compared with the FEA
predictions. When the correction factor is applied, the overestimation error is significantly
reduced to +35% with simultaneaus rise of the lower bound error to 40%. For the majority of
points representing different grooves and different cycles, the numbers of cycles calculated
using the analytical Glinka-Molski method are lower than those predicted by elastic-plastic
finite element analysis.

a) b)

Fig. 20. Fatigue lives determined for strain amplidues obtained with the original Glinka-Molski
method (a) and taking into account strain energy correction (b).

For the rotor 1 presented in Fig. 19b, the split of start-ups into cold and hot was available and
it was possible to estimate its fatigue damage using Eq. (27). Mean value fatigue data with
hold time were used in calculations to consider the creep damage due to secondary loading.
The stress ratio R = -1 was conservatively assumed as the mean stress correction models are
validated mainly for tensile mean stresses and the available low-cycle fatigue data for
1CrMoV steels are with zero mean stress (R = -1). The results of lifetime calculations for the
most damaged heat groove 3 are summarized in Table 2. As it is seen the total fatigue damage
reached nearly 1 which exceeds the critical value of 0.75 being the lower scatterband limit of
linear damage accumulation rule in the range of a low number of cycles [54]. The maximum
lifetime exhaustion predicted in groove 3 exceeding the limit of crack initiation is in good
agreement with the non-destructive test results shown in Fig. 19b, where the longest crack
was found in the same groove 3. The longest crack was also found in groove 3 of the rotor 2,
and this finding additionally supports the predictive capabilities of the presented method.

19
 Table 2. Low-cycle fatigue damage evaluation for heat groove 3.

Number of Number of
Damage Damage Total
Number of Number of cycles to cycles to
due to cold due to hot fatigue
cold starts hot starts cracking in cracking in
starts starts damage
cold start hot start

378 576 5296 634 0.090 0.909 0.999

9. Summary

A simple method for predicting the low cycle fatigue life of notched steam turbine rotors
has been presented. The method has been developed for practical use in lifetime monitoring
systems performing calculations in real time on industrial controllers.
The proposed algorithm uses the Glinka-Molski equivalent strain energy density method
for elastic-plastic strain analysis and strain-life approach for the low-cycle fatigue life
prediction. The strain energy density is formulated using the equivalent stress and strain with
a correction factor modelling the dilatation energy related with the mean stress.
It was shown by means of numerical calculations that the assumption of equivalence of the
strain energy density in the actual elastic-plastic and pseudo-elastic state is fulfilled for the
rotor grooves with a reasonable accuracy when the the dilatation energy is taken into account.
This finding justifies the use of analytical methods of stress-strain correction basing on this
assumption.
The plastic strain fields at the heat grooves are highly localized and confined to the groove
bottom surface. The damage under non-isothermal loading of the grooves is thus associated
with small-scale yielding.
Comparison of the strain amplitudes predicted by the Glinka-Molski method with energy
correction factor and those obtained from elastic-plastic FEA shows that the maximum error
does not exceed 10% which is in good agreement with other available results. The resulting
error in fatigue life obtained using the strain-life method is below 40%.
The results of FE analyses show that the modified Glinka-Molski method provides
reasonably accurate predictions of strain amplitudes and fatigue lives of circumferential heat
grooves where fatigue cracking has been observed. The distribution of strain amplitudes in the
grooves computed numerically well correspond with the distribution of fatigue crack lengths
found by NDT in real rootors. The highest low-cycle fatigue damage was obtained for the
groove where the longest cracks were found on many rotors. This confirms the predictive
capabilities of the method in low-cycle fatigue life estimation and online monitoring of steam
turbine rotors subjected to thermomechanical loads. The proposed concept consisting in the
use of the modified Glinka-Molski method with the strain energy correction factor determined
by elastic-plastic FEA was validated for the circumferential U-notch and should be further
studied for different notch geometries present in steam turbine rotors.

20
 References

[1] Viswanathan R. Damage Mechanisms and Life Assessment of High-Temperature Components,

ASM International, Metals Park, Ohio; 1989.
[2] Leyzerovich AS. Steam Turbines for Modern Fossil Fuel Power Plants, The Fairmont Press Inc.,
Lilburn; 2008.
[3] Helbig K, Banaszkiewicz M, Mohr W. Advanced lifetime assessment and stress control of steam
turbines. PowerGen Europe 2013, June 21-23, Milan.
[4] Łagoda T, Ogonowski P. Fatigue life estimation of notched specimens under bending and torsion
with strain energy density parameter, Journal of Theoretical and Applied Mechanics 2007; 45 (2):349-
361.
[5] Banaszkiewicz M. Multilevel approach to lifetime assessment of steam turbines, International
Journal of Fatigue 2015; 73:39-47.
[6] Banaszkiewicz M. Online monitoring and control of thermal stresses in steam turbine rotors,
Applied Thermal Engineering 2016; 94:763-776.
[7] Ince A, Glinka G. A numerical method for elasto-plastic notch-root stress-strain analysis, Journal
of Strain Analysis 2014; 48(4):229-244.
[8]Ince A, Glinka G. Innovative computational modeling of multiaxial fatigue analysis for notched
components, International Journal of Fatigue 2016; 82:134-145.
[9] Ince A, Glinka G, Buczyński A. Computational modeling of multiaxial elasto-plastic stress-strain
response for notched components under non-proportional loading, International Journal of Fatigue
2014; 62:42-52.
[10] Neuber H. Theory of stress concentration for shear-strained prismatical bodies with arbitrary
non-linear stress-strain law, ASME Journal of Applied Mechanics 1961; 28:544–550.
[11] Molski K, Glinka G. A method of elastic-plastic stress and strain calculation at a notch root,
Material Science and Engineering 1981; 50:93-100.
[12] Buczyński A, Glinka G. 1997, Elastic-plastic stress-strain analysis of notches under non-
proportional loading, 5th International Conference on Biaxial/Multiaxial Fatigue and Fracture Cracow.
[13] Buczyński A, Glinka G. 2001, An analysis of elasto-plastic strains and stresses in notched bodies
subjected to cyclic non-proportional loading paths, 6th International Conference on Biaxial/Multiaxial
Fatigue and Fracture Lisbon.
[14] Ince A. Numerical validation of computational stress and strain analysis model for notched
components subject to non-proportional loadings, Theoretical and Applied Fracture Mechanics 2-16;
84:26-37.
[15] Moftakhar A, Buczyński A, Glinka G. Calculation of elasto-plastic strains and stresses in notches
under multiaxial loading, International Journal of Fracture 1995; 70:357-373.
[16] Gehlot S, Mahadevan P, Kannusamy R. 2012, Analytical correction of nonlinear thermal stresses
under thermo-mechanical cyclic loadings, Proceedings of ASME Turbo Expo Copenhagen,.
[17] Gordon AP, Williams EP, Schulist M. 2008, Applicability of Neuber’s rule to thermomechanical
fatigue, Proceedings of ASME Turbo Expo Berlin,.
[18] Shin CS, Man KC, Wang CM. A practical method to estimate the stress concentration of notches,
International Journal of Fatigue 1994; 16:242-256.
[19] Guo W, Wang CH, Rose LRF. Elasto-plastic analysis of notch-tip fields in strain hardening
materials, Aeronautical and Maritime Research Laboratory Report 1998; DSTO-RR-0137, 1-36.
[20] Vogt J, Schaaf T, Helbig K. Optimizing lifetime consumption and increasing flexibility using
enhanced lifetime assessment methods with automated stress calculation from long-term operation
data, GT2013-95068, Proceedings of the ASME Turbo Expo 2013, June 03-07, 2013, San Antonio,
Texas, USA.
[21] Davies M. Life Extension of a High Pressure ST Control Wheel by Means of Weld Repair and
Assessment of Remaining Useful Life, Power-Gen Europe 2015, Amsterdam, June 2015.
[22] Fowler AD, Turner S. Life Extension of Steam Turbine Rotors Through Application of Repair
Technologies, Alstom paper T3D804, July 2015.
[23] Dobosiewicz J. Thermal fatigue of 200 MW steam turbine rotors, Power Engineering 1977;
3:101-104 (in Polish).

21
[24] Dobosiewicz J. Lifetime of intermediate and high pressure steam turbine rotor shafts, Power
Engineering 1985; 11: 448-450 (in Polish).
[25] Dobosiewicz J. Operating conditions of steels in selected power generation equipment, Power
Engineering 1975; 7/8.
[26] Mamontow NI, Pugacziewa TN. Determination of residual service life and lifetime extension of
turboset K-200-130 unit 9 at Luganskaja P.P., Power and Thermal Processes and Equipment 2008;
6:137-144 (in Russian).
[27 ] Banaszkiewicz M. On-line determination of transient thermal stresses in critical steam turbine
components using a two-step algorithm, Journal of Thermal Stresses 2017; 40:690-703.
[28] Hetnarski RB, Eslami MR. Thermal Stresses – Advanced Theory and Applications, Springer;
2009.
[29] Orłoś Z. Thermal Stresses, PWN, Warsaw (in Polish); 1991.
[30] Zienkiewicz OC, Taylor RL, Zhu JZ. The Finite Element Method: It’s Basis and Fundamentals,
Elsevier Butterworth-Heinemann, Burlington; 2005.
[31] Nowacki W. Thermoelasticity, PWN, Warsaw (in Polish); 1986.
[32] Armstrong P, Frederick C. A mathematical representation of the multiaxial Bauschinger effect,
CEGB Report No. RD/B/N 731, 1996.
[33] Chaboche JL. Time-independent constitutive theories for cyclic plasticity , International Journal
of Plasticity 1986; 2 (2):149-188.
[34] Garud YS. A new approach to the evaluation of fatigue under multiaxial loadings, Journal of
Engineering Materials and Technology -Transactions ASME 1981; 103 (2):118-125.
[35] Mróz Z. On the description of anizotropic work hardening, Journal of the Mechanicsand Physics
of Solids 1967; 15:163-175.
[36] ABAQUS 6.13 User’s Manual, 2013.
[37] Kleiber M, Kowalczyk P. Introduction to Non-linear Mechanics of Deformable Bodies, IPPT
PAS, Warsaw; 2011.
[38] Olszak W, Perzyna P, Sawczuk A. Theory of Plasticity, PWN, Warsaw (in Polish); 1965.
[39] Ziegler HA. Modification of Prager’s Hardening Rule, Quarterly of Applied Mathematics 1959;
17:55-65.
[40] Ince A, Glinka G. A generalized fatigue damage parameter for multiaxial fatigue life prediction
under proportional and non-proportional loadings, International Journal of Fatigue 2014; 62:34-41.
[41] Ince A, Glinka G. A modification of Morrow and Smith-Watson-Topper mean stress correction
models, Fatigue & Fracture of Engineering Materials & Structures 2011; 34:854-867.
[42] Banaś K, Badur J. Influence of strength differential effect on material effort of a turbine guide
vane based on a thermoelastoplastic analysis, Journal of Thermal Stresses 2017; 40:11:1368-1385.
[43 ] Badur J, Ziółkowski P, Kornet S, Stajnke M, Bryk M, Banaś K, Ziółkowski PJ. 2014, The Effort
of the Steam Turbine Caused by a Flood Wave Load, 2017, AIP Conference Proceedings 1822,
020001.
[44] Hoffman M, Seeger T. A generalized method for estimating multiaxial elastic-plastic notch
stresses and strains – Part I and II, ASME Journal of Engineering Materials and Technology 1985;
107:250-260.
[45] Harkegard G, Mann T. Neuber prediction of elastic-plastic strain concentration in notched tensile
specimens under large-scale yielding, Journal of Strain Analysis 2003; 38:79-94.
[46] Banaszkiewicz M. Numerical investigations of crack initiation in steam turbine rotors subjected
to thermo-mechnical fatigue (accepted to Applied Thermal Engineering).
[47] Lemaitre J, Desmorat R. Engineering Damage Mechanics, Springer-Verlag, Berlin Heidelberg;
2005.
[48] Glinka G. Calculation of inelastic notch-tip strain-stress histories under cyclic loading,
Engineering Fracture Mechanics 1985; 2(5):839-954.
[49] Holdsworth SR, Creep-Fatigue interaction in Power Plant Steels, EMPA 20100672
[50] ASME Boiler and Pressure Vessel Code,, Section III, Division 1: Sub-section NH, Claas 1
Components in Elevated Temperature Service, ASME, New York, 2001.
[51] TRD 301, Annex 1 – design: Calculation of cyclic loading due to pulsating internal pressure or
combined changes of internal pressure and temperature, Technical Rules for Steam Boilers, 1978.

22
[52] RCC-MR, Design and construction rules for mechanical components of FBR Nuclear islands,
AFCEN, Paris, 1985.
[53] Banaszkiewicz M. Analysis of rotating components based on a characteristic strain model of
creep , Journal of Engineering Materials and technology – Transactions of the ASME 2016; vol. 138,
No. 3:031004-1-11.
[54 ] Szala J, Ligaj B, Szala G. Sources of differences in calculations and experimental test results of
fatigue life of structural elements, Scientific Journal of Silesian University of Technology. Series
Transport 2014; 83:271-277 (in Polish).

23
Highlights

 Simple practical method for online assessment of the low-cycle fatigue life of steam
turbine rotors presented
 Glinka-Molski method with a correction for distortion strain energy proposed and suc-
cessfully appliedd for strain amplitude estimation
 Good accuracy was confirmed by comparisons with elastic-plastic FEA and crack
findings in real rotors

24