Modeling of Thermal Stress and Control for steam

turbine
Wei Wang
University of Jinan
Chao Dong
University of Jinan
Yongjian Sun (  [email protected] )
University of Jinan

Research Article

Keywords: steam turbine, nite element, stress, model predictive control

Posted Date: March 6th, 2023

DOI: https://doi.org/10.21203/rs.3.rs-2595382/v1

License:   This work is licensed under a Creative Commons Attribution 4.0 International License.
Read Full License

Additional Declarations: No competing interests reported.

 Modeling of Thermal Stress and Control for
steam turbine
Wei Wang, Chao Dong and Yongjian Sun1
1* School
of Electrical Engineering, University of Jinan, West
Road of Nan Xinzhuang, Jinan, 100190, Shandong, China.

Contributing authors: [email protected];

Abstract
In this paper, the method of system identification and predictive control
is used to control the stress of steam turbine rotor. Firstly, the three-
dimensional model of turbine rotor is established by using finite element
simulation software, and then the results are identified by using ANSYS
finite element analysis software. The transfer function is obtained. Firstly,
the temperature model is identified with an accuracy of 99.73 %, and
then the stress model is identified with an accuracy of 98.2 %. Then
the transfer function is discretized, and the data is input into the dis-
crete transfer function to verify the accuracy of system identification.
After the accurate verification of the transfer function, the transfer func-
tion is transformed into the dynamic matrix control model, and then
the stress feedback controller is designed to realize the model predic-
tive control (MPC) of the rotor stress. At the same time, the output
response of the system before and after the addition of the controller
is compared. Finally, the rationality of the control model is verified.
Future work will concentrate on more complex working conditions.

Keywords: steam turbine, finite element, stress, model predictive control

1 Introduction
In today’s society, people’s demand for electricity is higher and higher[4]. In
order to ensure people’s power demand, it is necessary to improve the operation
efficiency of steam turbine unit, which means that the start-up time of the
unit needs to be shortened. However, with the shortening of the start-up time

1
2 Article Title

of the steam turbine, the safety of the unit will also be reduced[5]. Therefore,
through establish the temperature distribution of the steam turbine rotor in
the start-up process through the actual data The safety of the model after
adjusting parameters is analyzed by model predictive control method[6].
The stress control of steam turbine rotor is mainly to ensure the safety of
the unit in more efficient operation. In the unit, the rotor of steam turbine
is an important part, which carries the energy and torque[1]. The safety of
steam turbine unit is mainly determined by the quality of turbine rotor. It is a
decisive factor to reduce the unit start-up time while ensuring that the stress
value of the turbine rotor is less than the yield limit value of the rotor material.
From a few years ago to now, the quality of life has improved significantly, the
grid capacity has increased significantly, so the peak value of the grid has been
increasing[2]. Frequent peak shaving operation means frequent start-up and
shutdown of the steam turbine unit. The change of working condition of steam
turbine unit will cause the damage of rotor material, thus shortening the life
of the unit. The parameters of steam turbine will change greatly during start-
up. Among them, the change of temperature parameter is the most important.
It will make the rotor produce a force, which is called thermal stress. At the
same time, it will make the metal material deform, mainly in the form of
expansion deformation. Once the thermal stress exceeds the yield limit of the
rotor material, the high-temperature components, mainly the turbine rotor,
will produce certain damage, which will eventually bring some security risks[3].
Among the existing control technologies, predictive control is a kind of con-
trol technology which does not require high model of the whole system. In the
current research of this problem, the strategy of system identification combined
with adaptive control theory is widely used. MPC has a double-layer control
structure[7]. On the one hand, it controls the optimization of the steady-state
part of the whole system, and the other part is responsible for the dynamic
optimization of the system. At the same time, the model predictive control has
the function of global optimization and teacher tracking[8]. However, due to
the change of operating conditions and various disturbances, the speed of the
whole operation control will be reduced. In this paper, the output error model
is used dynamic matrix control is actually the representative method of MPC
method. Dynamic matrix control uses the step response curve of the model,
which can well solve the control problem with constraints. Model predictive
control combines linear programming and control problem, and solves the con-
straint problem and static optimal solution problem at the same time. The
premise of using predictive control is that a mathematical model is needed[9].
In the establishment of steam turbine system model, neural network modeling
method is commonly used, that is, neural network is used to train the collected
data, train a neural network model, and then carry out predictive control on
the model[10]. Jianxi Yu and others used the collected data when establishing
steam turbine rotor system model a hybrid modeling method based on data
and first principle mechanism is proposed and verified[11]. S. Dettori and oth-
ers used fuzzy PID control algorithm in the control phase, and compared with
 Article Title 3

the traditional PID algorithm[12]. In the control of steam turbine system, PID
control is the most widely used at present, because the control method is very
mature, and it is very convenient to use. In this paper, MPC control algorithm
is used, which can make the efficiency faster and more accurate[13]. Hugo A. et
al. considered the constraint problem when using adaptive MPC[14]. Yan Z. et
al. proposed a stochastic model predictive control architecture[15]. Shifan W.
and others proposed a model predictive control (MPC) method to regulate the
grid connected microgrid system[16]. Soroush R. et al. analyzed the advantages
and disadvantages of MPC method applied to nonlinear systems[17]. Luwei Y.
et al. proposed a multi cell model predictive control theory, which was solved by
a peer-to-peer communication algorithm based on distributed projection[18].
Morteza M. et al. introduced the design, hardware implementation and hard-
ware in the loop simulation of MPC algorithm for turbofan engine control. In
addition, the feedback correction technique is used to compensate the effect of
model mismatch[19].
In all of the above studies, any work assumes that the initial temperature
is horizontal, which is a common assumption since there is no temperature
data at the time of turbine shutdown and the process is long. Turbine rotor
is the most important bearing component, can not be installed measuring
device. The initial temperature field of the turbine rotor is very different due
to the different initial state of cold start. Different initial thermal states will
affect the initial stress field. The optimal inlet temperature of gas turbine in
cycle is discussed with reference[20]. The entire analysis step takes 1 minute to
calculate the stress, thus capturing the transient characteristics during startup.
In this paper, the method of system identification and predictive control is
applied to the stress control of steam turbine rotor. Firstly, a 3-D model of
steam turbine rotor is established by using finite element simulation software.
Then, the ANSYS finite element analysis software is used to identify the results
and get the transfer function. Then, a stress feedback controller is designed
to realize the model predictive control of rotor stress. At the same time, the
output responses of the system before and after the addition of the controller
are compared. Finally, the rationality of the control model is verified.

2 Establishment of temperature model

The temperature mathematical model plays an important role in the stress
control. Only after the temperature model of stress is established, can the
stress model be established based on the results of the temperature model, so
as to carry out the work of stress control. The temperature model is essentially
a single input single output model.
During the whole operation process of steam turbine, the steam containing
heat works on the blades of the rotor through the steam nozzle, and drives
the rotor body to rotate by working on the blades of the rotor. At the same
time, the steam with high temperature will cause the theme temperature of
the steam turbine rotor to rise. When the material properties of the rotor
4 Article Title

are selected differently, the temperature rise rate of the rotor surface is also
different From the point of view of theory, the model from steam temperature
to rotor surface temperature can be regarded as a single input and single output
process, which can be realized by using the output error model as shown eq.1.

B(z −1 )
T (k) + e(k) Ts (k) = (1)
F (z −1 )
The system identification tool used in this paper uses the Smodel identifi-
cation toolbox in MATLAB to identify the model from steam temperature to
rotor surface temperature through the system identification toolbox. The data
of steam temperature adopts the cold start-up curve of steam turbine under
standard working conditions, as shown in Figure1.

80 600 25
Temperature
70 Speed
Pressure
20
Power occupancy
60

400
Power occupancy/%

Temperature/¡æ

50

Pressure/Mpa
3000 15

Speed/r
40

2000 10
30
200

20
1000 5
10

0 0 0 0
0 100 200 300 400 500 600

Fig. 1 Start-up parameter curve of steam turbine

The temperature of the rotor surface can not be obtained directly by mea-
surement, so the method adopted in this paper is to establish the 3-D model
of the turbine rotor of ANSYS finite element analysis software, load the input
data into the model, and obtain the temperature data of the rotor surface
through the calculation of the finite element analysis software. The 3-D model
of ANSYS is shown in Figure2.
By inputting the temperature data of cold start-up curve into the 3-D
model of ANSYS, the temperature data of rotor surface can be obtained.
The data result of steam temperature rotor surface temperature is shown in
Figure3.
 Article Title 5

Fig. 2 Finite element analysis model

Calculate the slope of each stage of the start-up curve, a total of four stages,
in which the slope of 2 and 4 stages is 0, and the temperature is 280 ◦ C and
540 ◦ C respectively. The slope of the first stage is 0.8, that is, the temperature
rise rate is 0.8 ?? / min, and the slope of the third stage is 0.8667, that is, the
temperature rise rate is 0.8667 ◦ C / min. Calculate the value of each minute
through the slope, in fact, it is consistent with the value of start-up curve for
the convenience of data fitting.
Taking the data of the start-up curve as the input and the rotor surface
temperature calculated by ANSYS as the output, the system identification is
used to identify the transfer function. The input and output data are presented
in the form of pictures in fig3. Four poles and three zeros are set for the iden-
tification transfer function. Because the data difference between rotor surface
temperature and actual start-up curve is not big, four poles and three zeros
are selected for system identification, the identification accuracy is 98.2%, and
the specific transfer function is eq. 2. The identification results are shown in
Figure4.

63.17s3 + 1.433s2 + 0.02318s + 0.0005257

G1 (s) = (2)
s4 + 64.45s3 + 1.462s2 + 0.02365s + 0.0005362
6 Article Title

550
Steam temperature
500 Rotor surface temperature

450

400
temperature/¡æ

350

300

250

200

150

100
0 100 200 300 400 500 600
time/min

Fig. 3 Rotor steam temperature and surface temperature

Input and output signals

600

400
y1

200

0
0 100 200 300 400 500 600

600

400
u1

200

0
0 100 200 300 400 500 600
Time

Fig. 4 System identification temperature output

 Article Title 7

2.1 Discretization of temperature model

After the transfer function is established, the discrete data cannot be input
into the transfer function model, but the required mathematical model must
load the input data into the model for calculation, so it needs to be discretized.
The discretization first is to transform the transfer function z, using the ztrans
command in MATLAB to transform the transfer function z, and the discrete
transfer function after the above transformation is as follows eq.3:

0.9804z 3 − 1.054z 2 + 1.186z − 0.2514

G1 (z) = (3)
z 4 − 1.075z 3 + 1.21z 2 − 0.2564z + 3.243 ∗ 10−17

After transforming the transfer function z, it is necessary to change the

transformed transfer function into the form of difference equation, because
after transforming into difference equation, the model can be analyzed directly
in the form of equation[21]. The transformation rule between the transfer
function and the difference equation after z transformation is as follows:

Y (z)
G(Z) = (4)
X(z)

y(n)+a1 y(n−1)+...+an y(n−N ) = b0 x(n)+b1 x(n−1)+...+bm x(n−M ) (5)

There is formula 5 which can be obtained:

M
X N
X
y(n) = bi x(n − i) − ai y(n − i) (6)
i=1 n=1
Through the above transformation, the difference equation of the model can
be obtained as follows:

y(n) − 1.075y(n − 1) + 1.21y(n − 2) − 0.02564y(n − 3) + 3.243 ∗ 10−17 y(n − 4)

= 0.9804x(n − 1) − 1.054x(n − 2) + 1.186x(n − 3) − 0.2514x(n − 4)
(7)

2.2 Temperature model validation

Take the above transformed difference equation as the temperature model, and
input the temperature data of cold start-up curve into the model to verify the
accuracy of the model. The output results are shown in Figure5: The maximum
error is 7.2mpa, and the maximum error between the model output data and
the actual data is 1 %, which indicates that the model is accurate.
8 Article Title

550
Actual temperature value
500 model output

450

400
temperature/¡æ

350

300

250

200

150

100
0 100 200 300 400 500 600
time/min

Fig. 5 Comparison of actual temperature and model temperature output

3 Establishment of stress output model

Before establishing the stress output model of the steam turbine rotor, the
heat flux of the steam turbine rotor should be calculated first[22]. The heat
flux needs to be calculated through the temperature model. During the start-
up and shutdown of the steam turbine unit, the steam temperature and heat
release coefficient on the rotor surface will change with time. At the same time,
the values of different positions are also different, so the steam temperature can
be explained And the heat release coefficient is a function of time and space.
The calculation formula of heat release coefficient is given below, which is a
function of rotor speed and pressure. The formula is given by Westinghouse[23].
The formula is as follows:

h = (0.0633678∗N −7.71255∗10−6 ∗N 2 +0.00339265∗P ∗N )∗F +2.724 (8)

where, N is the rotation speed of the turbine rotor, P is the back pressure
value at the same spatial position, F is the adjustment factor for adjusting
the value of heat release coefficient. The values of heat release coefficient F
are different for different types of turbine rotors with different materials. The
formula is also given by Westinghouse. The calculation formula of heat release
coefficient is as follows:
 Article Title 9

0.03 25.817 ∗ G ∗ (1 − γ)
F = ∗[ ] (9)
φ 57735 ∗ Ddia
where, φ is the seal clearance, in inches, G is the steam flow at the high pressure
part under the maximum power working condition, Ddia is the diameter of the
steam seal, γ is the leakage factor of the blade, and the physical parameters
of the 300MW turbine rotor should be considered when selecting the value of
F[24], as shown in the following table1-table3:

Table 1 The 30Cr1Mo1V steel chemical composition 1

Element C Fe Mo Cr Si Mn P S V Ni
Wt /% 0.28 1.10 0.28 0.73 1.13 0.023 0.22 0.005 0.24 0.4

Table 2 The 30Cr1Mo1V steel chemical composition 2

Temperature Yield strengthσs Ultimate strengthσb Elongationδ Reduction of areaψ
20◦ C 629Mpa 779Mpa 20% 60%
540◦ C 465Mpa 520Mpa 29.6% 88.5%

Table 3 The physical property of 30Cr1Mo1V steel

Temperature/◦ C 20 100 200 300 400 500 600
Young modulus E(Gpa) 214 212 205 199 190 178 178
Poisson ratio µ 0.288 0.292 0.287 0.299 0.294 0.305 0.305
Thermal conductivity λs (W/m·K) 48.5 47.1 44.8 42.8 40.3 37.5 35.3
Linear expansion coefficient αl 0 11.99 12.81 13.25 13.66 13.92 14.15
Specific heat c(J/Kg) 554 574 599 624 666 720 804

After the heat release coefficient is calculated, the stress output model
can be established through the temperature model calculated in the previous
chapter. The specific steps are as follows:
(1)The heat release coefficient is calculated by the process variables during
the start-up of the steam turbine unit, including steam temperature, steam
pressure, steam flow, rotor speed, etc.

h = (0.0633678∗N −7.71255∗10−6 ∗N 2 +0.00339265∗P ∗N )∗F +2.724 (10)

(2)After Newton’s cooling law, when there is a temperature difference

between the surface and the surrounding, the heat loss per unit time from the
unit area is directly proportional to the temperature difference, and the pro-
portional coefficient is called the heat transfer coefficient, the heat flow density
q can be calculated as follows:

q = h(T − T0 ) (11)
10 Article Title

where, T is the input temperature of the rotor and T0 is the actual tem-
perature of the rotor surface. The temperature model has been found in the
previous chapter. The model is a temperature model transfer function, which
is transformed by z-trans command of MATLAB. The discrete function model
between T and T0 is:

0.9804z 3 − 1.054z 2 + 1.186z − 0.2514

G2 (z) = (12)
z4 − 1.075z 3 + 1.21z 2 − 0.2564z + 3.243 ∗ 10−17

Take G2 (z) into eq.12 and get the following formula:

0.9804z 3 − 1.054z 2 + 1.186z − 0.2514

q = h(T − T0 ) = h( − 1)T0
z 4 − 1.075z 3 + 1.21z 2 − 0.2564z + 3.243 ∗ 10−17
(13)
(3)Take the calculated q value as input and the stress value as output to
fit the transfer function of q and stress value, and finally get the stress output
model.

3.1 System Identification of Stress Model

The stress value can not be obtained by direct measurement, so the stress
value is calculated by establishing a finite element 3-D model, and the cold
start curve is given to the established 3-D model, according to the relationship
between stress and strain, geometric equation and balance equation[25], the
thermal stress distribution of rotor can be obtained.
α l E n 1 R ro 2 2 R ro o
σrr = ∗ − r2 ri trdr + r2r(r−r i
2 −r 2 ) r
trdr (14)
1−u o i i

α l E n 1 R ro 2 2 R ro o
σxx = ∗ r2 ri trdr + r2r(r2+r i
2 trdr − t (15)
1−u o −ri ) ri

αl E n 2
R ro o
σyy = ∗ r2 (ro2 −ri2 ) ri trdr − t (16)
1−u
The maximum stress curve obtained by finite element analysis software is
shown in Figure6: After calculation, the size of q is shown in Figure7: With the
above figure q value as the input and the stress value as the output, the data
information shown in Figure8 and the transfer function shown in formula 14
can be obtained through the system identification toolbox of MATLAB. After
fitting, the accuracy of the transfer function is 98.23%.

0.06245s4 + 0.006419s3 + 0.0002455s2 + 0.00002362s − 0.00000006247

G2 (s) =
s5 + 0.01564s4 + 0.003776s3 + 0.00005561s2 + 0.000001003s + 0.00000003642
(17)
The above transfer function is discretized to verify the accuracy of
the transfer function, use ztrans command in MATLAB to make eq.17
 Article Title 11

450

400

350

300
stress(Mpa)

250

200

150

100

50

0
0 100 200 300 400 500 600
time(min)

Fig. 6 Maximum stress of rotor

200

150

100
q/W

50

−50

−100
0 100 200 300 400 500 600
time/min

Fig. 7 q value of heat flux

12 Article Title

Input and output signals

600

400
y1

200

0
0 100 200 300 400 500 600

150

100
u1

50

0
0 100 200 300 400 500 600
Time

Fig. 8 Input and output of stress model system identification

Z-transform:

1.2835z 4 − 0.7825z 3 + 1.0224z 2 + 2.21z − 0.002325

G2 (z) =
z 5 + 0.45217z 4 − 0.08254z 3 + 0.23345z 2 − 1.2117z − 0.00004214
(18)
When the calculated q value is input into the discrete model as the input
data, a group of stress output data can be obtained, and the data is compared
with the stress value under the actual working condition, as shown in Figure9:
These two sets of data are continuous, it is impossible to directly take
points to calculate the error, so the determination coefficient method is used
to evaluate the accuracy of the model. When the determination coefficient
method is used, the closer the value is to 1, the better the proof result is. The
formula of the determination coefficient is as follows:
Pn ′ ′ 2
2 (si − s̄i )
R = Pi=1
n 2 (19)
i=1 (si − s̄i )
′
where, n is the number of model output data, si is the actual data, si is the
model output data, s̄i is the mean value of the data. After calculation, the
coefficient of determination R2 = 0.9972, it shows that the model has very high
availability.
 Article Title 13

450
model output
400 Actual stress value

350

300
stress/Mpa

250

200

150

100

50

0
0 100 200 300 400 500 600
time/min

Fig. 9 Comparison of actual stress value and model output

3.2 State space equation of stress output system

In this paper, the control method is model predictive control, model predic-
tive control calls the mathematical model in the form of state space equation
is more convenient, so in the next use of model predictive control, the stress
output model needs to be transformed into the state space format, the trans-
formation of the two uses the SS function of Matlab. The mathematical model
form of the specific state space equation is eq.20-eq.24, it is transformed from
eq.17 through SS function of MATLAB:
 dx
dt = Ax + Bu (20)
y = Cx + Du
 
−0.1085 −0.0033 −0.0002872 −0.00000242 −0.00000007159 −0.000000002774

 1 0 0 0 0 0 

 0 1 0 0 0 0 
A= 

 0 0 1 0 0 0 

 0 0 0 1 0 0 
0 0 0 0 1 0
(21)
14 Article Title

 
1
0
 
0
B=
0
 (22)
 
0
0

C = [−0.0707; −0.0033; 0.0006738; 0.0000189; 0.000002293; 0.000000007125]

(23)

D=0 (24)
The above state space model is verified by using the lsim function of
M AT LAB, and q is input into the model as an input variable to obtain a set
of stress curves. The specific stress curves are shown in Figure10:

450

400

350

300
stress/Mpa

250

200

150

100

50

0
2 4 6 8 10 12 14 16
Sampling point 9
x 10

Fig. 10 Stress output of state space model

It can be seen from the figure that the trend of the model in the form of
state space equation is consistent with that of the discrete transfer function
model, which proves that the mathematical model in the form of state space
can be used.
 Article Title 15

4 Stress-speed feedback model predictive

control
4.1 Calculation conditions of effective stress
The stress of steam turbine rotor mainly includes two aspects, one is the
thermal stress caused by the drastic change of material temperature of steam
turbine rotor due to the application of temperature load during start-up, and
the other is the centrifugal stress generated with the increase of speed during
start-up. The thermal stress and centrifugal stress can be calculated by the
fourth strength criterion formula[25]. The fourth strength criterion formula of
effective stress is shown in eq.25.
p
σeq = σth 2 + σt 2 + σth σt (25)

where, σeq is the effective stress, σth is the thermal stress and σt is the centrifu-
gal stress of the rotor. The calculation formula of rotor speed and centrifugal
stress[25] is shown in eq.26:

N N
σt = σe = σe (26)
Ne 3000

where, σe is the value of the rated stress, N is the speed in the formula, Ne is
the rated speed.

4.2 Centrifugal stress control

The centrifugal stress control of the steam turbine rotor is actually the speed
control of the steam turbine rotor. By adjusting the speed, the centrifugal
stress is within the required reasonable range. The speed control of steam
turbine can directly affect the value of centrifugal stress caused by the change
of speed. In the previous researchers, the speed model of steam turbine given is
the model[26] shown in eq.27 and eq.28, which is composed of hydraulic motor
and steam volume. W (s) is the transfer function of hydraulic motor and Q(s)
is the transfer function of steam volume.
1
W (s) = (27)
0.08s

1
W (s) = (28)
6.2s + 1
The speed model is simulated. Under the condition of step response as
input, the speed output response of the speed model is shown in Figure11.
It can be seen from the figure that the overshoot of the speed model is
large and the stable time is long. On this basis, a PID controller is connected
in series to the speed model, and the parameters of the controller are set as
16 Article Title

1.8

1.6

1.4

1.2
response

0.8

0.6

0.4

0.2

0
0 200 400 600 800 1000 1200
time

Fig. 11 Step response of original speed model

follows: P = 0.02967, I = 7.73867, D = 0.11769. The unit step response of

PID controller is shown in Figure12.
It can be clearly seen from the figure that the overshoot of the model
is reduced and the stabilization time is shorter. Subsequent control will be
studied on the basis of this model. In order to more visually see the control
effect, the model output after adding the speed controller and the original
speed control model are compared, as shown in Figure13:

4.3 Load monitoring of steam turbine connected to grid

The system model used in the process of steam turbine grid connection is
composed of hydraulic motor module and steam volume module as shown in
eq.27 and eq.28, and reheater module and steam turbine rotor module are also
used. The model of reheater module D(s) and steam turbine rotor module[26]
is shown in eq.29 and eq.30.
1 1
D(s) = ∗ (29)
0.3s + 1 0.08s
1 8/3s + 1 1
G3 (s) = ∗ ∗ (30)
0.35s + 1 8s + 1 0.2s + 1
The turbine can not be directly loaded after grid connection, and it needs to
be increased step by step. This paper analyzes the corresponding speed output
under the unit step input condition when the load is mounted at 30% load
and 60% load. When the load is 30%, the output response of turbine speed is
shown in Figure14.
 Article Title 17

1.4

1.2

0.8
response

0.6

0.4

0.2

0
0 200 400 600 800 1000 1200
time

Fig. 12 Speed response after adding PID controller

2
Add PID controller
1.8 Original model speed response

1.6

1.4

1.2
response

0.8

0.6

0.4

0.2

0
0 200 400 600 800 1000 1200
time

Fig. 13 Speed model output comparison before and after adding controller
18 Article Title

1.4

1.2

0.8
response

0.6

0.4

0.2

−0.2
0 200 400 600 800 1000 1200
time

Fig. 14 Speed response at 30% load

It can be seen from the above figure that when the turbine is loaded with
30% load, the overshoot increases and the stable time increases. When the
turbine is loaded with 60% load, the speed output response of the turbine is
shown in Figure15.
When the load is 60%, the turbine speed output overshoot increases and
the stable time is long, which is convenient for intuitive analysis. The output
response of 30% and 60% load is given, as shown in Figure16.
The stress speed model predictive control in this paper is mainly aimed at
the speed control of the steam turbine unit system. Through the speed control,
the centrifugal stress of the steam turbine rotor is within a reasonable range.
The effective stress online detection model of the steam turbine rotor has been
given by previous researchers[25]. In this paper, the existing stress monitoring
model is stripped out by deducing the fourth strength criterion formula In the
centrifugal stress part of the effective force, the centrifugal stress is fed back to
the rotor speed model through the feedback of rotor speed data. At the same
time, PID controller is added to the speed model for comparison to realize the
predictive control of centrifugal stress. The improved system block diagram is
shown in Figure17.
The stress speed controller is mainly deduced by the fourth strength cri-
terion formula shown in eq.25, which makes the centrifugal stress as shown in
eq.31. √
−σth + −3σth 2 − σeq 2
σt = (31)
2
 Article Title 19

1.8

1.6

1.4

1.2

1
response

0.8

0.6

0.4

0.2

−0.2
0 200 400 600 800 1000 1200
time

Fig. 15 Speed response at 60% load

1.8
Output response of 30% load mounted speed
1.6 Output response of 60% load mounted speed

1.4

1.2

1
ÏìÓ¦

0.8

0.6

0.4

0.2

−0.2
0 200 400 600 800 1000 1200
time

Fig. 16 Speed response of 30% and 60% load respectively

20 Article Title

Fig. 17 Structure diagram of stress-speed feedback control model

When there is no stress speed feedback controller, the stress output response
of steam turbine rotor is shown in Figure18.
In order to facilitate comparison, the output response without feedback
controller is compared with the stress output with feedback controller, as
shown in Figure19. As shown in Figure 19, after adding the feedback controller,
the overshoot is reduced by nearly 23%, and the start-up time is shortened by
nearly 7%, which indicates that the response speed and stability of the system
are better than the original system after adding the feedback controller.

5 Conclusion
In this paper, stress control problem of steam turbine rotor is solved by the
model identification and predictive control. Firstly, a three-dimensional model
of steam turbine rotor is established using finite element technology. Com-
bined with the working condition, then the simulation data is obtained and
identified to get the transfer function. Parameters of this transfer function are
determined according to the working condition and simulation results. After
discretization the transfer function can be converted to dynamic matrix control
model, and then the stress-feedback controller is designed. Structure diagram
of stress-speed feedback control model is drawn to realize the stress of rotor
model predictive control (MPC), at the same time, it is compared with orig-
inal controller system. Comparison of actual stress value and model output
are conducted. Speed response by adding PID controller improves the results
greatly. On basis of the speed model, the speed response of 30% and 60% are
given respectively, and the stress output response without feed back controller
 Article Title 21

1.8

1.6

1.4

1.2
response

0.8

0.6

0.4

0.2

0
0 500 1000 1500 2000 2500 3000
time

Fig. 18 Stress output response without feedback controller

1.8
Output response after adding feedback controller
1.6 Output response of original control model

1.4

1.2
response

0.8

0.6

0.4

0.2

0
0 500 1000 1500 2000 2500 3000
time

Fig. 19 Stress output response with feedback controller

22 Article Title

and the stress output response with feedback controller are obtained and com-
pared. Finally, the feasibility of the this control scheme is verified. This work
supplies a new view to solve the stress control problem in thermal engineering,
and the gap between the theory and engineering may be overcome.

6 Author Declarations
Ethical Approval: not applicable. This article does not involve human subjects
or animals.
Competing interests: the author has no relevant financial or non-financial
interests that need to be disclosed, and no competitive interests related to this
article.
Authors’ contributions: Wang Wei is responsible for building the framework
and writing of the article. Li Shaohui is responsible for collecting and sorting
documents. Sun Yongjian is responsible for providing ideas and formulating
research directions.
Funding: this work is supported by National Natural Science Foundation
(NNSF) of China under Grant 61603149, by Shandong Key Research and
Invention Program under Grant 2019GSF111046, by the Science Foundation
for Post Doctorate Research from the University of Jinan.
Availability of data and materials: data sharing is not applicable to this
article as no new data were created or analyzed in this study.

References
[1] Ahmed A., Robert E., Daniel L., et al. Low cycle fatigue life modelling
using finite element strain range partitioning for a steam turbine rotor
steel, Theoretical and Applied Fracture Mechanics, 2020, 113: 311-323.

[2] Jakub, P., Dominik, G., Janisaw, Z., et al. Elastic limit load resource
when reaming holes in turbine rotor discs, Engineering Failure Analysis,
2020, 113: 104555.

[3] Zhewen, C., Yanjuan, W., Xiaosong, Z., et al. Energy and exergy analyses
of S?CCO2 coal-fired power plant with reheating processes, Energy, 2020,
211: 118651.

[4] G.L. Hou, L.J. Gong, C.Z. Huang, J.H. Zhang. Novel fuzzy modeling and
energy-saving predictive control of coordinated control system in 1000
MW ultra-supercritical unit, ISA Trans, 2019, 86: 48-61.

[5] L. Kong, J.Q. Yuan. Disturbance-observer-based fuzzy model predictive

control for nonlinear processes with disturbances and input constraints,
ISA Trans, 2019, 90: 74-88.
 Article Title 23

[6] H. Gouta, S.H.J. Said, A. Turki, F. M??Sahli. Experimental sensorless

control for a coupled two-tank system using high gain adaptive observer
and nonlinear generalized predictive strategy, ISA Trans, 2019, 87: 187-
199.

[7] L. Ju, Q. Tan, R. Zhao, S. Gu, J. Yang, W. Wang. Multi-objective electro-

thermal coupling scheduling model for a hybrid energy system comprising
wind power plant, conventional gas turbine, and regenerative electric
boiler, considering uncertainty and demand response, J Clean Prod, 2019,
237: 1-16.

[8] David C., Mazen A., Guillaume S.. Data-driven fatigue-oriented MPC
applied to wind turbines Individual Pitch Control, Renewable Energy,
2021, 170: 1008-1019.

[9] Juan W., Canbing L., Sheng H., et al. MPC-based DC-link voltage control
for enhanced high-voltage ride-through of offshore DFIG wind turbine,
International Journal of Electrical Power Energy Systems, 2021, 126:
106591.

[10] Dettori, S., Maddaloni, A., Colla, V., et al. Nonlinear Model Predictive
Control strategy for steam turbine rotor stress, Energy, 2019, 158: 5653-
5658.

[11] Jianxi Yu, Pei Liu, Zheng Li. Hybrid modelling and digital twin develop-
ment of a steam turbine control stage for online performance monitoring,
Renewable and Sustainable Energy Reviews, 2020, 133: 110077.

[12] S.Dettori, V.Iannino, V.Colla, A.Signorini. An adaptive Fuzzy logic-based

approach to PID control of steam turbines in solar applications, Applied
Energy, 2018, 227: 655-664.

[13] Janusz B., Mateusz B.. Accelerated start-up of the steam turbine by
means of controlled cooling steam injection. Energy, 2019, 173: 1242-1255.

[14] Hugo A., Pipino., Marcelo M., et al. Nonlinear temperature regulation
of solar collectors with a fast adaptive polytopic LPV MPC formulation.
Solar Energy, 2020, 209: 214-225.

[15] Yan Z., Fanlin M., Jianmai S.. Uncertainty-resistant stochastic MPC
approach for optimal operation of CHP microgrid. Energy, 2019, 179:
1265-1278.

[16] Shifan W., Wenjun X., Jianglong Q.. MPC-based frequency control strat-
egy with a dynamic energy interaction scheme for the grid-connected
microgrid system. Journal of the Franklin Institute, 2020, 357: 2736-2751.
24 Article Title

[17] Soroush R., Sebastien G., Luca F.. MPC approaches for modulating
air-to-water heat pumps in radiant-floor buildings. Control Engineering
Practice, 2020, 95: 104209.

[18] Luwei Y., Tao L., David J.. Distributed MPC-based frequency control for
multi-area power systems with energy storage. Electric Power Systems
Research, 2019, 173: 1242-1255.

[19] Morteza M., Ali R., Milad E.. Design and implementation of MPC for
turbofan engine control system. Aerospace Science and Technology, 2019,
173: 1242-1255.

[20] Martin K., Michal K.. Control of a boiler-turbine unit using MPC-based
reference governors, Applied Thermal Engineering, 2017, 110: 1437-1447.

[21] Salem B., Roshan S., Nasibeh Z.. An MPC-based power management of
standalone DC microgrid with energy storage, International Journal of
Electrical Power Energy Systems, 2020, 120: 105949.

[22] Guolian H., Linjuan G., Jianhua Z.. Fuzzy modeling and fast model
predictive control of gas turbine system, Energy, 2020, 200: 117465.

[23] Westinghouse Electric Corporation Process Control Division. Standard

control algorithm user manual, 2000??(7).

[24] Yongjian S., Xiaoqiang L., Xueyan T.. Online life estimation for steam
turbine rotor, Journal of Loss Prevention in the Process Industries, 2013,
26: 272-279.

[25] Ji, D., Sun, J., Sun, Q., et al. Optimization of start-up scheduling and life
assessment for a steam turbine, Energy, 2018, 160: 19-32.

[26] Lu. S., Hogg. B. W. Predictive Co-orienated Control for Power-Plant

Steam Pressure and Power Output. Control Eng.Practice,1997, 5(1):
79-84