Assessment of operational risk of steam turbine valves

Andrzej Rusin
*
The Silesian University of Technology, 44-100, Gliwice, Konarskiego 18, Poland
Received 17 October 2003; revised 19 January 2004; accepted 19 January 2004
Abstract
The paper estimates the technical risk associated with the operation of power machines. The sources of risk are presented with possible
scenarios of failures. The risk involving the operation of steam turbine valves is discussed in more detail. The point estimation technique for
reliability analysis and the nite element analysis of thermal stresses are applied. Calculations of the probability of failure are made for
random loads and random properties of materials. On the basis of the real data, mean values of stresses and strains under transient operating
conditions were calculated, as well as the level of stresses under creep conditions. The calculated values were used for estimation of the
probability of valve failure and the value of technical risk.
q 2004 Elsevier Ltd. All rights reserved.
Keywords: Turbine valve; Failure analysis; Technical risk
1. Introduction
Despite the advance of technology and the creation of
modern systems of automatic control and protection,
failures of technical systems have not been eliminated.
Although awareness of the risk of such failures has been
widely disseminated, the assessment of the level of risk and
prevention is often made intuitively on the basis of
experience. Generally, the rational analysis of risk should
be based on the theory of probability, addressing the
following issues [1,2]:
How big is the risk associated with the existence of a
given technical system?
How may this risk be reduced?
What are the costs involved in the reduction of the risk in
question?
What level of risk is acceptable?
The issue of the assessment of risk associated with
conventional power stations has received more attention in
recent years [3,4]. This results from changes in the world
market for energy, involving mergers of power corporations,
sale and purchase of power stations with often unknown
history of operation. At the same time, the equipment of
many power stations has become obsolete. The operational
life of many power units has already exceeded the design life.
Furthermore, serious accidents occurring in the energy
sector, including failures of turbine rotors [3] have
contributed to the problem. In the face of erce competition
high values of reliability and availability indices and
reduction of unplanned shut-downs have become an opera-
tional priority focused on the reduction of technical risk.
The scope of this paper is to determine the technical risk
associated with the operation of power machines. In
particular, the analysis of the risk associated with the
operation of stop valves in steam turbines is discussed in
detail. Apart from rotors and cylinders, valves constitute the
most important components of turbines bearing the highest
load. Operational experience indicates that the symptoms of
wear may be observed in turbine valves, which reduces the
durability of the entire turbine. The risk associated with
the operation of valves may play an essential role in the
operational risk of a turbine and the power unit as a whole.
2. Technical risk
2.1. Quantitative assessment of risk
The analysis of the technical risk associated with
operation of a given system must follow the following
0308-0161/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijpvp.2004.01.004
International Journal of Pressure Vessels and Piping 81 (2004) 373379
www.elsevier.com/locate/ijpvp
* Fax: 48-32-2372680.
E-mail address: rusin@rie5.ise.polsl.gliwice.pl (A. Rusin).
procedures [2]:
denition of a system,
identication of the hazards and situations threatening
the entire system as well as its particular elements,
assessment of the probability of the occurrence of
hazards,
assessment of the consequences in the case of hazards,
assessment of risk,
design of activities associated with the existing level of
risk.
Technical risk should include all possible events
resulting from the existence and operation of a technical
system. In a quantitative approach, such risk is calculated as
R
X
i
R
Ei

X
i
P
Ei
C
Ei
1
where R
Ei
is the risk associated with event E
i
; P
Ei
is the
probability of the occurrence of event E
i
; and C
Ei
is the
consequence of the occurrence of event E
i
:
The probability of the occurrence of events may be
assessed on the grounds of the analysis of statistical data
derived from observations, opinions of experts, or prob-
ability models. As far as the assessment of the probability of
failure of a power unit is concerned, rst and foremost, as it
is a complex technical system, a detailed analysis of the risk
of failures of its particular components is required. This
paper assesses the risk associated with the operation of a
stop valve in a steam turbine, constituting one of the
principal components of a power unit.
2.2. Sources of risk
The fundamental source of the operational risk of
machines is the absence of the certainty of performance
by every single component at any operating time
Z . L 2
where Z is generalized strength and L is generalized load.
This results from the random nature of many data [5],
including for example, geometrical dimensions, material
properties, such as: yield stress, material toughness, creep
rupture strength, load, including thermal loads. These last
data are because of random temperature elds, especially in
unsteady states.
The degree of the loss of strength should also be treated
as a random variable. The above mentioned factors indicate
that the durability of a machine component should be treated
as a random variable and in its successive operating periods
the probability of damage should be assessed.
Human errors are also sources of risk. The identication
of the sources of risk makes it possible to design possible
hazards.
3. Hazards
3.1. Hazards evoked by degradation processes
The operation of power machines, including heat
turbines, has a cyclic nature. At the beginning of the
cycle there is a start-up. Once the working medium reaches
its nominal parameters, steady-state operation follows.
Shut-down occurs at the end of the operating cycle,
followed by natural or forced cooling of machine
components. At each phase of operation the material is
subjected to gradual degradation and loss of durability of
machine components leads to complete loss of machine
life. During start-up, the main process of life consumption
is low-cycle fatigue. To determine the essential properties
of the start-up processes almost 100 start-ups of a 200 MW
turbine were subjected to statistical analysis. The curves of
temperature changes of main steam at start-up from the
cold, warm and hot state were analyzed in detail, as shown
for a cold start in Fig. 1.
From computer simulations [6] it was possible to obtain
the time variations of stresses and strains at any valve point.
The processes of life consumption are mainly at stress
concentration points. Exemplary time variations of maxi-
mum effective stresses for some selected start-ups from the
cold, warm and hot state are shown in Fig. 2. The
distribution of stresses presented in Fig. 2 indicates that,
contrary to theoretical start-ups, real ones evoke several
stress amplitudes and, accordingly, strain amplitudes. Thus,
low-cycle fatigue life consumption during a single start-up
Fig. 1. Time variations of the temperature of steam at the start-up from the
cold state.
A. Rusin / International Journal of Pressure Vessels and Piping 81 (2004) 373379 374
of a turbine is equal to
Z
i

X
m
k1
Z
k
3
Z
k

1
N
fk
4
where Z
k
is the fatigue life consumption due to the kth
amplitude in a cycle, N
fk
is the number of operation cycles
to failure, m is the number of essential strain amplitudes in
one operating cycle.
Statistical analysis revealed that the average start-up is
characterized by two essential stress and strain amplitudes.
The mean values and standard deviations of the strain
amplitudes are indicated in Table 1.
In steady-state operation the life is limited by creep. This
phenomenon occurs with more intensity in stop valves in the
high-pressure section due to the highest values of the
medium temperature in contact with the valve metal.
In such case, the main load is the steam pressure.
Detailed analysis of the creep of a valve in a 200 MW
turbine was discussed in Ref. [7]. It has been estimated that
the mean value of the stress is 60 MPa and its standard
deviation is 4 MPa.
3.2. Instantaneous hazard
An instantaneous hazard may be expressed by the
following relations:
in the absence of cracks
s
M
. R
m
5
where
R
m
ultimate strength
s
M
von Mises stress
in presence of aws [8]
K
K
IC
. FL
r
6
where
FL
r
1 20:14L
2
r
0:3 0:7 exp20:65L
6
r

K Ms

l
p
L
r
s=s
y
sapplied tensile stress,
law-size dimension,
Mcoefcient dependent on the shape and location
of a defect,
s
y
yield stress,
K
IC
material toughness.
4. Probability of valve failure
4.1. Probability of valve failure from the processes
of degradation
The calculated mean values of the strain amplitude
during start-ups and the mean stress in steady creep, as well
as the estimated standard deviation of these values make it
possible to calculate the allowable number of operating
cycles and the allowable time of operation under creep
conditions. For the CrMoV cast steel the following
equation was assumed to describe fatigue
N
f
aD1
b
7
where
D1amplitude of total strains,
a; bconstants,
N
f
number of cycles to failure.
Under creep conditions the dependence between time to
rupture and stress has the form
t
f
cs
d
8
Fig. 2. Exemplary time variations of effective stresses during start-up from
the cold state (1), warm state (2) and hot state (3).
Table 1
Mean value and standard deviation of strain amplitudes
No. of amplitude
in a cycle
Mean value m (%) Standard deviation s (%)
1 0.116 0.028
2 0.07 0.016
A. Rusin / International Journal of Pressure Vessels and Piping 81 (2004) 373379 375
where
ssteady stress under the conditions of creep,
c; dconstants,
t
f
time to failure.
Thus, low-cycle fatigue life consumption in one operat-
ing cycle is:
Z
N

X
m
i1
1
aD1
i

b
9
The creep degradation in 1 h of operation is:
Z
t

1
cs
d
10
The total life consumption after N start-ups and an
operating time t; in accordance with a linear damage
summation rule, is:
Z
N;t
N
X
m
i1
1
aD1
i

b
t
1
cs
d
11
This life consumption is a random variable. The
probability of valve damage after N start-ups and operation
time t is calculated as [9]
p
f
pg # 0 12
g Z
c
2Z
N;t
where Z
c
is the critical value of life consumption. This is a
random variable.
Using the reliability index b; the rst order approxi-
mation of the probability of valve failure is
p
f
f2b 13
where
b
m
g
s
g
14
f is the cumulative normal distribution function.
To calculate the probability of valve damage, the
following coefcients were assumed b 21:59; d 22:9:
Coefcients a and c are random variables, as well as the
critical value of life consumption Z
c
: The mean values and
standard deviations for these random variables are indicated
in Table 2.
Moreover, the strain amplitudes and stress levels under
creep conditions are also random variables. The calculations
of the probability of failure were carried out by means of the
point estimation method PEM [10].
For function Y of one random variable X with mean value
m
x
and variance s
2
x
; the moment of kth order of variable Y
assumes the following form
EY
k
P

y
k

P
2
y
k
2
15
where
P

P
2

1
2
y

Ym
x
s
x
y
2
Ym
x
2s
x

Accordingly, the mean value of a new random variable

m
y
and its variance s
2
y
is calculated as:
m
y
EY
1
2
y

y
2
16
s
2
y

y

2y
2
2

2
17
For uncorrelated m random variables, expression (16)
assumes the following form
EY
k

1
2
m
y
k

y
k
222
18
where y

corresponds to the value of function Y

determined for the values of arguments m
x1
s
x1
; m
x2

s
x2
; whereas, y
222
denotes the value of function Y at
point m
x1
2s
x1
; m
x2
2s
x2
; for all m random variables of
random vector X:
The graphical form of the dependence between the
probability of valve failure and the operating time and
number of start-ups is presented in Fig. 3. A signicant
increase of the probability of valve failure occurs after
200,000 h of operation time.
4.2. Probability of instantaneous failure
Following Section 3.2, if a new performance function is
introduced as
g R
m
2s 19
the probability of damage is equal to:
p
f
pg # 0 20
For given values of m
s
145 MPa, s
s
38:5 MPa,
m
R
360 MPa, s
R
36 MPa, p
f
2:26 10
25
was
derived.
Another performance function may be dened as
g FL
r
K
IC
2Ms

l
p
21
and the probability of damage may be expressed by
p
f
pg # 0 22
where all variables, that is: K
IC
; s
y
; M; s; l are random.
Their random nature is due to the variable temperature eld
Table 2
Mean value and standard deviation of constants a; c; Z
c
Random variable Mean value m Standard deviation s
a 395 50
c 6.38 10
10
1.27 10
10
Z
c
1 0.05
A. Rusin / International Journal of Pressure Vessels and Piping 81 (2004) 373379 376
(K
IC
; s
y
; s) and indistinct location and dimensions of a
possible defect (M; l). The assumed probability character-
istics are in Table 3.
The probability of damage calculated on the grounds of
the above values is p
f
0:47 10
23
:
4.3. Total probability of valve failure
The probability of instantaneous damage of a valve as
well as damage due to the processes of degradation was
evaluated above. The probability of damage due to the
processes of degradation increases in time. The instan-
taneous damage hazard is associated with rapid changes in
stresses occurring under unsteady operation, rst of all,
during start-ups. This means that at each start-up, there are
several modes of possible failure of the valve. The total
probability p
fc
of such failure may be estimated in
the following way
X
i
p
fi
$ p
fc
$ max p
fi
23
where p
fi
is the probability of failure due to ith mode.
The above estimation is accurate for a low probability
of failure and mutual independence of possible modes.
From the results obtained for the probability of instan-
taneous damage and the damage due to the passage of
time, total probability of HP valve failure in a 200 MW
turbine at start-up phase was estimated in Fig. 4. For
example, for a turbine that has been in operation for
175,000 h and 400 start-ups, the probability of valve
failure at start-up is about 0.50 10
23
. At steady-
operation the probability of failure is considerably lower
and equals about 0.12 10
24
.
Fig. 3. Probability of valve failure in the function of operation time t and turbine start-ups number N:
Table 3
Mean value and standard deviation of K
IC
; s
y
; M; s; l
Random variable Mean value m Standard deviation s
K
IC
(MPa m
1/2
) 55 15
M ( ) 1.42 0.3
s (MPa) 66 16
l (m) 0.0025 0.001
s
y
(MPa) 165 16.5
A. Rusin / International Journal of Pressure Vessels and Piping 81 (2004) 373379 377
5. Assessment of the consequences of valve failure
The assessment of the consequences of valve failure
requires a full understanding of the mechanisms of damage
and all their effects, including, for example, the nancial
loss of income, costs of repairs, costs of replacement,
penalty fees, etc. Furthermore, the technical consequences
and impacts on the external environment, including possible
casualties, injuries to personnel, and pollution of the natural
environment or loss of corporate goodwill should also be
accounted for. The failure of the stop valve has nancial
implications, involving:
costs of purchasing a new valve,
costs of replacing the valve,
costs of start-up losses in the power unit,
loss of prots connected with stand-by,
losses associated with penalty payments for failure
to supply energy.
Other implications, for example, secondary malfunctions
of other components of the power unit caused by the
damaged valve, as well as possible impacts on
Fig. 4. Probability of value failure at start-up.
Fig. 5. Risk diagram.
A. Rusin / International Journal of Pressure Vessels and Piping 81 (2004) 373379 378
the environment, were disregarded because of their low
probability.
6. Assessment of risk
On the grounds of the calculated probability of valve
failure in different operation phases and the estimated
implications of the damage it is possible to assess the risk
involved. The variables discussed above were entered on the
risk diagram in Fig. 5. The scale of the effects was expressed
in a relative percentage scale related to the value of
replacing the power unit. The areas of negligible risk (A),
low risk (B), medium risk (C), high risk (D) and extreme
risk (E) were differentiated on the diagram. The risk
associated with the operation of the valve depends on its
operating time. For example, for the valve of a turbine
which worked for 150,000 h and was started up 400 times
during that time V
1
; the area of risk is low. Medium risk
corresponds to 250,000 h of work and 800 start-ups of the
turbine V
2
:
7. Conclusions
The calculation results indicate that stop valves are not
critical components of a power unit. The technical risk
involved in the operation of the valves increases under
unsteady operation modes, especially during start-ups.
Nevertheless, even at start-ups the risk is still low or
medium. The medium risk concerns power units operated
for over 225,000 h and started up more than 800 times.
Thus, it may be assumed that in view of such operation time,
the replacement of stop valves in 200 MW turbines is
justied. It is also possible to lower the risk by reducing the
stresses occurring under unsteady operation modes, for
example, by proper control of the turbine start-up.
References
[1] Faber MH, Stewart MG. Risk assessment for civil engineering
facilities: critical overview and discussion. Reliab Engng Syst Safety
2003;80:17384.
[2] Stewart MG, Melchers RE. Probabilistic risk assessment of
engineering systems. London: Chapman & Hall; 1997.
[3] Bissel A. Technical risk in power generation. OMMI 2002;1(1).
[4] Jovanovic A. Risk-based component life management in fossil power
plants. OMMI 2002;1(1).
[5] Miroshnik R, Jeager A, Ben Haim H. Probabilistic life assessment of
chest valve under thermal stresses. Int J PressVessel Piping 1998;75:
15.
[6] ANSYS v.7.1.-user guide; 2002.
[7] Rusin A. Numerical simulation of turbine valve creep. Arch Appl
Mech 1992;62:38693.
[8] Webster GA, Ainsworth RA. High temperature component life
assessment. London: Chapman & Hall; 1994.
[9] Sundararajan C, editor. Probabilistic structural mechanics handbook.
New York: Chapman & Hall; 1995.
[10] Rajagopalan K. Point estimations technique for the reliability analysis
at ship structures. Comput Struct 1993;48(5):83541.
A. Rusin / International Journal of Pressure Vessels and Piping 81 (2004) 373379 379