You are on page 1of 15

Manuscript details

Manuscript number IPVP_2015_71

Title On the Use of Computational Fluid


Dynamics (CFD) Simulations to Assess
the Impact of Low-Load Operation
on Heat Recovery Steam Generator
(HRSG) Tube Module Integrity

Article type Research paper

Abstract As the electricity market has evolved


with the addition of renewables
to the generation mix, Heat Recovery
Steam Generators (HRSG) that were
originally designed for base load
conditions are now frequently forced
to operate in a cycling and/or
low-load regime. This can lead
to front end tube-to-header fatigue,
creep or creep-fatigue failures,
often induced by GT flow imbalances
causing locally-elevated tube temperatures
and/or bending stresses on joints
due to large temperature differences
between tube rows. This paper focuses
on the use of Computational Fluid
Dynamics (CFD) as a tool to analyze
the risks of shifting operation
mode. Exhaust gas flow profiles
were analyzed for various low load
conditions in two power plants
with differing vertical designs.
One of the plants had already
moved into cycling mode and suffered
tube failures that were directly
related to low-load (and start-up)
exhaust flow patterns, the other
plant is projected to operate in
a frequent cycling mode in the
near future. The contribution
of CFD to identifying the conditions
that lead to failure for the first
plant is presented, along with
projections on the potential impact
of low-load operation on the second
plant design in terms of risk of
hot-end tube failures. Mechanisms
to reduce the failure risk, such
as addition of flow-conditioning
devices, are also investigated.

Corresponding Author Andreas Fabricius

Order of Authors Andreas Fabricius, Jan Rusaas,


David Moelling

Submission files included in this PDF

File Type File Name

Manuscript Tetra submission 090116.pdf

To view all the submission files, including those not included in


the PDF, click on the manuscript title on your EVISE Homepage, then
click 'Download zip file'.
On the Use of Computational Fluid Dynamics (CFD) Simulations
to Assess the Impact of Low-Load Operation on Heat Recovery
Steam Generator (HRSG) Tube Module Integrity
Andreas Fabriciusa , Jan Rusaasb , Dave Moellingc
a
Tetra Engineering Europe, Sophia-Antipolis, France
b
R&R Consult, Aalborg, Denmark
c
Tetra Engineering Group, Inc., Simsbury, USA

Abstract
As the electricity market has evolved with the addition of renewables to the generation
mix, Heat Recovery Steam Generators (HRSG) that were originally designed for base load
conditions are now frequently forced to operate in a cycling and/or low-load regime. This
can lead to front end tube-to-header fatigue, creep or creep-fatigue failures, often induced
by GT flow imbalances causing locally-elevated tube temperatures and/or bending stresses
on joints due to large temperature differences between tube rows. This paper focuses on
the use of Computational Fluid Dynamics (CFD) as a tool to analyze the risks of shifting
operation mode. Exhaust gas flow profiles were analyzed for various low load conditions in
two power plants with differing vertical designs. One of the plants had already moved into
cycling mode and suffered tube failures that were directly related to low-load (and start-up)
exhaust flow patterns, the other plant is projected to operate in a frequent cycling mode in
the near future. The contribution of CFD to identifying the conditions that lead to failure
for the first plant is presented, along with projections on the potential impact of low-load
operation on the second plant design in terms of risk of hot-end tube failures. Mechanisms
to reduce the failure risk, such as addition of flow-conditioning devices, are also investigated.
Keywords: HRSG, Computational Fluid Dynamics (CFD), Tube failures, Creep, Fatigue

1. Introduction
As the electricity market changes, the demands on boiler efficiency and availability be-
come more and more dependent on other sources of energy, e.g. renewables in the form of
wind and solar. This forces power plants originally design for base load conditions to move

Email addresses: andreas.fabricius@tetra-eng.com (Andreas Fabricius),


J.Rusaas@R-R-Consult.com (Jan Rusaas), dave.moelling@tetra-eng.com (Dave Moelling)
URL: www.tetra-eng.com (Andreas Fabricius), www.R-R-Consult.com (Jan Rusaas),
www.tetra-eng.com (Dave Moelling)

Preprint submitted to International Journal of Pressure Vessels and Piping January 9, 2016
into daily load shift, shutdowns over weekends and/or standby modes with rapid starts. This
paper investigates the use of CFD as a simulation tool when looking at the root cause for
existing HRSG tube failures at one combined cycle power plant (CCPP) and for determining
the risk associated with moving to different load patterns at another CCPP. CFD allows
detailed modeling of the GT exhaust gas flow distributions in the HRSG duct and across
tube modules comprising rows of finned heat exchanger tubes.

2. Approach
Two different plants are analyzed. Both of these plants are vertical HRSG designs, with
GT exhaust gas flowing first horizontally and then vertically upwards across the tube banks.
One of the plants (Plant A) operates in a cyclic mode while the second plant (Plant B) is
planning to do so in the near future. Plant A already suffers from tube failures caused by
accelerated creep and fatigue that are directly related to uneven flow distribution. For Plant
A, CFD simulations are used when investigating the root cause of the tube failures, as well
as to investigate how flow corrective measures will affect the flow distribution upstream the
High-Pressure High-Temperature Superheater (HPHTSH) tube sheet. For plant B the CFD
simulation is used as part of a proactive risk assessment to determine whether uneven flow
distributions at different loads could lead to tube failures. The cause of tube failures would
be through creation of locally-elevated tube temperatures and/or large row-row temperature
differences. The commercial flow solver Fluent from ANSYS [1] is used for the simulations.
A high quality mesh with 8 million cells is used for the modeling. For improved quality
and accuracy the mesh consists mainly of structured and unstructured hexahedral elements.
A region with a dense (fine) resolution tetrahedral mesh is used in the transition section
and a denser mesh is also used around the FCD and in the HPHTSH section, in order to
capture the larger velocity gradients. The mesh is gradually made coarser downstream the
HPHTSH section, as this region does not have an upstream effect on the flow distribution.
For improved accuracy higher order discretization
schemes are used for the advection terms and for the
pressure interpolation (PRESTO) the double precision
solver is used. The realizable k-epsilon model is used
for the turbulent equations. A simplified approach is
used for the front-end tube sheet; the actual tubes are
included and a circular volume is created around every
tube that extends the finned area. A resistance is in-
cluded in the volume such that the flow straightening
effect from the fins on the flow is included.
The remaining boiler sections are included as porous
media, with flow resistances, such that the impact on
Figure 1: GT swirl at low load
the flow is included. The pressure drops over the tube
sheets are calculated using the boiler thermal perfor-

2
mance simulation software PPS&D1 .
The low load conditions results in more swirl from the GT than base load conditions,
leading to a more uneven flow distribution, Figure 1. This load case is analyzed in detail
for both of the plants. Results from base load or cycling modes are not presented in this
paper. It should be noted that side to side flow uniformity was observed when base load flow
conditions were analyzed for both plants, indicating that the FCD designs were sufficient
for this load case.
The flow is characterized by a Vulcan shape, typical for axial GT machines, with a lower
axial velocity in the center and increasing velocity outwards until a maximum, and then
decreasing towards the outer wall. The calculations are carried out with a GT mass flow
of 358.6 kg/s at a temperature of 628C for plant A and with a GT mass flow of 411.6 kg/s
at a temperature of 628C for Plant B. The swirl angle of the axial GT varies with the load
and from GT data a swirl angle of 30 degree is found for part load operation. The axial
flow component is scaled to the reduced mass flow and a tangential velocity is added that is
30 degrees counter-clockwise from the axial velocity. This GT configuration is used on both
plants. Both designs had flow correction devices (FCD) made of horizontal tubes installed.
The geometry of plant A can be seen in Figure 2 and the geometry of plant B can be seen
in Figure 3.

Figure 2: Left: Plant A 3D model original design. Right: Plant A flow improvements, Lattice (green) and
Perforated plates (red).

1
Power Plant Simulator & Designer, KED GmbH, Rodenbach, Germany

3
Figure 3: Left: Plant B 3D original design. Right: Plant B Lattice (green) and FCD (yellow)

3. Results
3.1. Plant A - Original Design
Figure 4 shows the velocity magnitude on the horizontal and vertical mid-plane at the
inlet section and through the FCD. The swirling motion of the GT flow forces the flow
outwards to the wall and a low velocity, reversed flow zone is established in the center of
the duct.

Figure 4: Velocity magnitude on vertical (left) & horizontal (right) mid plane at the inlet

Figure 5 shows the normal velocity distribution at the entrance and at the exit of the
HPHTSH section, respectively. The distribution is clearly not symmetric around the mid
plane, which is due to the swirling motion of the GT flow. A higher velocity (mass flow) is
seen at the lower side of the duct in the figure (right hand side of duct, as viewed from the
GT), in the main part of the area towards the wall, the velocity exceeds 50% above mean.

4
Figure 5: Normal velocity distribution at the entrance (top) and exit (bottom) of the HPHTSH section
(Left: Range 0 to max. Right: Range 50% from mean)

The normalized mass flow distribution is shown in Table 1. The variation in the sections
varies from 0.66 to 1.58, corresponding to a mass flow of 34% below mean to 58% above
mean. The row average is reasonable evenly distributed, however a large variation can be
seen in the column average, column D sees an average mass flow of 42% above average.
Hence, tubes in the right hand side in the flow direction will be exposed to a considerably
greater heat flux than tubes in column areas A-C.

Table 1: Normalized mass flow distribution, row/column explanation in Figure 6


A B C D Row.avg
7 1.20 0.78 0.86 1.41 1.06
6 1.11 0.86 0.94 1.27 1.04
5 0.84 0.77 0.97 1.44 1.00
4 0.74 0.68 0.93 1.58 0.98
3 0.73 0.66 0.94 1.56 0.97
2 0.81 0.68 0.96 1.41 0.96
1 0.96 0.83 0.82 1.28 0.97
Col.Avg 0.91 0.75 0.92 1.42

5
Figure 6: Row and Column distribution

3.2. Plant A - Including Lattice and Perforated Plates


In order to reduce the swirling motion of the flow a lattice is introduced just upstream
the originally installed FCD. The lattice consists of nine vertical and horizontal plates,
equally distributed over the width and height of the duct. The outline of the lattice is
shown in Figure 2. In addition, perforated plates with an opening fraction of 60% is added
downstream the lattice to further stabilize the flow. It can be seen from Figure 7 how the
effects of the swirl are reduced immediately upstream the perforated plates (compare to the
original design, Figure 4). The plates pushes the flow from the boiler sides towards the
centre, creating a more evenly distributed flow profile upstream the HPSH section.

Figure 7: Velocity magnitude on vertical (left) & horizontal (right) mid plane at the inlet

The normal velocity distribution at the entrance to the HPSH section showed an im-
proved distribution, compared to the original design case, Figure 5. The region with higher
velocity towards the right hand side wall is reduced and the low velocity region in the center
is improved, Figure 8. A similar flow pattern can be seen at the exit of the HPSH section.

6
Figure 8: Normal velocity distribution at the entrance (top) and exit (bottom) of the HPSH section (Left:
Range 0 to max. Right: Range 50% from mean)

The normalized mass flow distribution is shown in Table 2 . The column average at the
right hand side, column D, is reduced from 42% to 12% above average. The column average
for B is also significantly improved, now at 1% below average, whereas it was 25% below
average for the original design (Table 1). The maximum value is found in D7 (close to the
outlet header) where it is 24% above average. This is still an improvement from the previous
58% above average.

Table 2: Normalized mass flow distribution


A B C D Row.avg
7 0.95 1.02 0.93 1.24 1.03
6 1.01 1.14 0.95 1.13 1.06
5 1.04 1.06 0.93 1.15 1.05
4 0.89 0.94 0.99 1.18 1.00
3 0.82 0.91 1.00 1.04 0.94
2 0.85 0.90 0.87 1.01 0.91
1 1.03 0.97 0.97 1.08 1.01
Col.Avg 0.94 0.99 0.95 1.12

The combination of a lattice and perforated plates improves the flow distribution con-
siderably compared to the original design. The column average variation ranges from -6%
to 12% from mean, whereas the original design ranges from -25% to 42% from mean. It is
estimated from the CFD calculation that the lattice and the perforated plate will increase
7
the pressure drop with some 75 Pa. The effects of the pressure drop on the overall thermal
performance can be analyzed by the use of thermal simulation tools.

3.3. Plant B - Original Design


The low load calculation is carried out with a GT mass flow of 411.6 kg/s and a temper-
ature of 628.3 C. The swirl of the GT flow is set to 30 degrees. The axial flow component
is scaled to the reduced mass flow and a tangential velocity is added of 30 degrees counter-
clockwise from the axial velocity. Figure 9 shows the velocity magnitude on the horizontal
and vertical mid-plane at the inlet section and through the FCD. The swirling motion of
the GT flow forces the flow outwards to the wall, as was observed for Plant A, and a low
velocity flow zone is created in the center of the duct.

Figure 9: Velocity magnitude on vertical (left) & horizontal (right) mid plane at the inlet

Figure 10 shows the normal velocity distribution at the exit and the entrance of the first
boiler section, respectively. It can be seen that the distribution is highly uneven, with zones
of both high mass flow and reversed flow. Before the first boiler section several zones can
be seen to exceed 50% above mean. Furthermore, the distribution is non-symmetrical along
both axes.
The uneven flow distribution is also clearly seen in Table 3. The section flow variation
is from 0.71 to 1.88, corresponding to 29% below average and 88% above average. Similar
variations can also be seen in both the column- and row averages, where an average deviation
of 35% and 24% is predicted, respectively. The results show that the tubes at the right hand
side, column D, will be exposed to a considerably higher heat flux, than the remaining tubes.

8
Figure 10: Normal velocity distribution at the entrance (top) and exit (bottom) of the HPSH section (Left:
Range 0 to max. Right: Range 50% from mean)

Table 3: Normalized mass flow distribution, row/column explanation in Figure 11


A B C D Row.avg
8 0.86 0.97 1.08 1.16 1.02
7 0.80 0.86 1.03 1.28 0.99
6 0.80 0.80 0.94 1.27 0.95
5 0.80 0.81 0.92 1.18 0.93
4 0.75 0.80 0.94 1.20 0.92
3 0.71 0.76 0.95 1.28 0.92
2 0.75 0.74 1.06 1.54 1.02
1 1.12 0.72 1.23 1.88 1.24
Col.Avg 0.82 0.81 1.02 1.35

9
Figure 11: Row and Column distribution

3.4. Plant B - Including Lattice


The previous low load case showed an uneven flow distribution upstream the boiler
section, especially the side-side-uniformity where the variation in mass flow for the column
average was in the range of -19% to +35% from mean, resulting in a large variation of heat
flux affecting the upstream tubes. In an attempt to create a more uniform distribution, a
lattice was included upstream of the existing FCD. Figure 12 shows the velocity magnitude
on a horizontal and vertical mid plane at the inlet section and through the FCD. The location
of the grid and the FCD is shown in black. The swirling motion of the GT flow forces the
flow outwards to the wall and a low velocity flow zone is created in the center of the duct.

Figure 12: Velocity magnitude on vertical (left) & horizontal (right) mid plane at the inlet

Figure 13 shows the normal velocity distribution 100 mm downstream and 100 mm
upstream the first boiler section, respectively. A zone of reversed flow can be seen closest
to the GT, which is due to the 90 degree turn of the duct. Further away from the wall, the
flow appears more uniform than in the load case without a lattice presence.

10
Figure 13: Normal velocity distribution at the entrance (top) and exit (bottom) of the HPSH section (Left:
Range 0 to max. Right: Range 50% from mean)

The results of the simulations with added lattice can be seen in Table 4. The variation in
the sections is between 0.71 and 1.21, corresponding to 29% below average and 21% above
average. This is comparatively better than the previous simulation (no lattice), where the
variation was 29% below and 88% above. Variations in the row averages are between 15%
below and 14% above, compared to 19% below to 35% above, previously. In the column
averages the variation is between 8% below and 5% above, compared to 8% below to 24%
above previously. The results are generally improved, indicating that the guide plates has a
pronounced positive effect on the distribution at the lead HRSG tube module.

11
Table 4: Normalized mass flow distribution, row/column explanation in Figure 11
A B C D Row.avg
8 1.09 0.21 1.02 0.86 1.04
7 1.05 1.19 1.21 1.11 1.14
6 0.75 0.96 1.17 1.21 1.02
5 0.71 0.85 1.10 1.09 0.94
4 0.96 0.96 1.04 0.96 0.98
3 0.86 0.80 0.79 0.95 0.85
2 0.88 0.93 0.98 1.05 0.96
1 1.08 1.02 1.10 1.09 1.07
Col.Avg 0.92 0.99 1.05 1.04

4. Discussion
4.1. Plant A
The swirling motion of the GT flow results in a flow pattern that varies considerable in
the span-wise direction, where the tubes on the right hand side in the flow direction are
exposed to a mass flow that is 42% above average (see Table 1). As a consequence the tubes
see a higher heat flux from the GT exhaust gas, leading to higher local metal temperature
and/or large temperature differences between front and back-end tubes. The uneven flow
distribution is thought to be a contributing factor to the fatigue (and creep) failures seen in
the HRSG over the last couple of years. Other load cases, like purging and cold/warm/hot
starts will also affect the tube integrity. These have not been simulated, however the re-
sults from the low load simulation gives an indication of the gas flow distribution for this
combination of turbine and HRSG duct design.
The combination of introducing a lattice and perforated plates (at 60% opening fraction)
upstream the existing FCD showed a considerable improvement to the overall flow distribu-
tion. The column average variation was reduced from -25% to +40% from mean at current
design, Table 1, to from -6% to +12% from mean, Table 2.

4.2. Plant B
The flow distribution at low load has been shown and high variations in flow distribution
were observed. The tubes in column D where exposed to an average mass flow of 35%
above mean. The highest variation was found in row 1, where the mass flow is as much
as 88% above mean. Introducing a lattice upstream the existing FCD stabilized the flow
significantly, reducing the section mass flow from 88% to 21% above mean, Table 4.

12
5. Conclusions
The following conclusions can me drawn from the study:

• Computational Fluid Dynamics simulations can efficiently be used to determine the


risks of changing operational modes, with respect to thermo-mechanically induced tube
failures in the hot end of the HRSG

• Neither the current Plant A Flow Correction Device (FCD) nor the Plant B FCD
design are appropriate for low load (and/or cyclic) operations given the installed GT
exhaust profile. The use of a lattice in combination with perforated plates significantly
improved the flow distribution

• The GT flow distribution will depend on both the type of turbine used and the HRSG
duct design. Hence, the effects on front-end HPSH and/or RHT tubes is highly plant
specific.

• Plant B showed very similar flow characteristics through the boiler flow path, as was
shown in Plant A. Given the fact that plant A continuously suffers from tube failures
(attributed to both creep and fatigue), it can be concluded that Plant B has an elevated
risk of similar types of failures unless the flow distribution is significantly improved,
by introducing a lattice and/or perforated plates to reduce the effects of the GT swirl.

• Flow distributions upstream the Duct Burners are normally required to be kept at
+-15% from mean, however acceptable flow distributions above/below mean upstream
the hottest tube sheet are normally not addressed in the design stage. This could be
established by combining simulations using CFD , HRSG thermohydraulic models and
thermo-mechanical Finite Element Analysis .

6. Recommendations
• Use CFD as a first screening tool for a plant planning to move into cycling mode.

• Use CFD in the design phase for new plants, analyzing all different load cases.

References
[1] ANSYS Fluent Theory Guide. ANSYS Inc, version 14.5, November 2011.

13

You might also like