Documente Academic
Documente Profesional
Documente Cultură
Zhang
Post Doctoral Fellow.
Numerical Simulation of Different
A. C. M. Sousa
Types of Steam Surface
Professor,
Mem. ASME Condensers
A numerical procedure is developed to simulate the fluid flow and heat transfer
J. E. S. Venart processes in the shell-side of steam surface condensers. The governing equations are
Professor. solved in primitive variable form using a semi-implicit consistent control-volume
formulation in which a segregated pressure correction linked algorithm is employed.
Fire Science Center, The procedure is applied to three different types of surface condenser. The numerical
and Department of Mechanical Engineering, predictions are critically assessed by comparison to available experimental data for
University of New Brunswick, condensers, and in general, the solutions are in good agreement with the experimental
Fredericton, New Brunswick, Canada data.
I Introduction
Failure and/or off-design performance of power plant steam CALICO developed at Electricite de France, which uses a two-
surface condenser, as pointed out by Diaz-Tous [6], have re- dimensional, pseudo-transient numerical algorithm. Its pre-
sulted in substantial losses due to plant down-time and/or low dictions may prove useful in optimizing geometric parameters
efficiency. Values of 3.8 and 1.5 to 2 percent are reported for with respect to flow-induced vibration. Al-Sanea et al. [1]
loss of unit availability and performance loss, respectively, of employ a commercial code to solve the partial-differential
installed power in the US utility industry. The costs of these equations governing the two-dimensional flow field and steam
losses during the period 1974-1984 was estimated to be more concentration. The densities for the steam and air, possibly
than US$ 18 billion (1974 US$). These losses indicate that there for computational expediency, were assumed to be constant.
is a clear need of design methods, which can predict the shell- Shida et al. [12] and Takahashi et al. [13] analyze several
side flow distribution and heat transfer. This information is condenser configurations by using the numerical technique first
of critical importance in performing analyses related to flow- reported by Shida et al. [11], In this technique, the mesh is
induced vibration, heat transfer rates and pressure losses, based on a triangular pattern as in the finite element method,
which, eventually, will lead to enhanced unit performance and and the calculations are performed using upwind two-step La-
reliability. For instance, an average 1-in. Hg increase of back grangian and Eulerian time-marching techniques.
pressure costs US$ 1 million and US$ 2 million annually for The procedure developed employs coupled heat transfer and
typical 600 MWe oil-fired and nuclear units, respectively. fluid flow calculation techniques combined with a continuum
To attain the goal of developing new methods, a viable approach, where the steam/air mixture is assumed to be flow-
approach is through advanced numerical models, which are ing through a porous medium. Densities for steam and air vary
based on governing differential equations and appropriate con- according to the local conditions, and in addition the procedure
stitutive relations. In this way the designer can investigate the incorporates a model, which can accommodate multipass con-
effects of different shapes of tube nest, placement of baffles, densers with the assumption that at the end of each pass, the
steam lanes and venting points. outlet cooling water temperature is well mixed and equal to
The open nonproprietary literature dealing with numerical that into the next pass. The present study is a further extension
modeling of condensers is not extensive, and, in many cases, of recent work by Zhang and Sousa [16,17], in which single-
critical details of the models and/or relevant results are omit- phase flow was modeled for a shell-and-tube heat exchanger
ted. In the present work a numerical procedure is proposed, [16], and preliminary computations were carried out for a
which combines some of the features suggested by Davidson condenser with a simple geometric configuration [17].
and Rowe [5], Davidson [4], Caremoli [3], Al-Sanea et al. [1], Three different steam surface condenser configurations as
Shida et al. [11,12], and Takahashi et al. [13]. Davidson and described by Fuji et al. [8], Naviglio et al. [9], and Al-Sanea
Rowe [5] and Davidson [4] outline the basis of the continuum et al. [1] are utilized in the evaluation of the code. The results
modeling approach for the thermal design of condensers, and are compared with available experimental data so to bench-
review some related work. Caremoli [3] describes the program mark the procedure and on a preliminary basis to validate the
program.
Contributed by the Petroleum Division for publication in the JOURNAL OF II Numerical Method
ENERGY RESOURCES TECHNOLOGY. Manuscript received by the Petroleum Di-
vision, July 20, 1990; revised manuscript received April 11, 1991. Two different approaches may be used for condenser mod-
Total volume
- 3Puv) + - ( M = - ^ - j +- ^ -
This method has been shown by Rhodes and Carlucci [10] and -!3^-l3mv + f3Fv (4)
Theodossiou et al. [14] to describe adequately the flow in tube dy
banks. Conservation of Air Mass Fraction
The shell-side and water-side flows are treated as steady
state. On the steam side the model considers the fluid to be a
mixture of noncondensable gases (air) and steam, behaving as l ^U) + Ty ^V) = YX ( * fx) + Ty (* fy) (5)
a Newtonian fluid with constant viscosity. The mixture of
steam and noncondensable gases, for simplicity, is assumed to 2.2 Auxiliary Relationships
be a perfect gas, although other equations of state can be
considered, and its density varies locally according to the per- (/) Momentum Source Term. The local hydraulic flow
fect gas state equation. The steam is assumed at its saturation resistances, Fu and F, in the momentum equations caused by
temperature throughout the condenser. This assumption is the tube bundle and/or baffles, are related to the pressure loss
based on the observation reported by Bush et al. [2] that su- coefficients, and , by
perheating has little effect upon heat transfer. Fu = kuuUp (6)
The governing equations are the equations of conservation
Fv=^vUp (7)
of mass, momentum and air mass fraction, and they are solved
by a modified control-volume formulation based on the al- Since there is no general expression for the pressure loss in
gorithm suggested by Van Doormaal and Raithby [15], and to condensing tube banks, as a first approximation, the expres-
be described in Section 2.4. sions proposed by Rhodes and Carlucci [10] for the loss coef-
ficients and are used; namely,
2
2.1 Conservation Equations. The two-dimensional fu\ P(3 1-/
steady-state porous medium volume-averaged conservation iu=2 (8)
P-D,
equations of mass, momentum and air mass fraction, with
flow, heat and mass transfer resistances are written in the P(3 1-/3
Cartesian coordinate system, as follows: f. = 2ft P-Dr 1-A.
(9)
Nomenclature
*--J@
(ii) Mass Source Term. The mass condensation rate per
unit volume, m, can obtained by a simple energy balance equat-
Fig. 1 Grid arrangement
ing the phase change enthalpy with the heat transfer rate
* S -* W
mLV= (10) p = pCT=PaCaT+PsCsT (16)
R
where T is the saturation temperature determined by partial
where the cooling water temperature Tw is obtained from the
steam pressure, and C, Ca and Cs are the gas constants for the
heat balance between the steam and cooling water for each
mixture, air and steam, respectively.
control volume. The overall thermal resistance for each control
volume, R, is calculated from various semi-empirical heat (iv) Effective Viscosity. The concept of an effective vis-
transfer correlations, as described in the forthcoming. cosity is used, which is defined as the sum of the laminar and
For the water-side thermal resistance, the relationship sug- turbulent viscosity, namely,
gested by Dittus and Boelter [7] is employed; namely,
He = H + H, (17)
-^- = 0.023 R e / 8
Pr 04 For all simulations the turbulent viscosity, /*,, mainly to save
computer time in this exploratory stage of the study, is assumed
The wall resistance for each tube is obtained with the as- to be constant. Its value, based on previous work by Zhang
sumption of one-dimensional, steady-state conduction and Sousa [17,18], is kept at a range of 10 to 100 times the
dynamic viscosity, /x.
/},= [AJog(ZVA)]/(2*,) (12)
In this paper, since its purpose is to demonstrate the nu- 2.3 Boundary Conditions. The boundary conditions for
merical calculation method rather than to conduct a detailed the inlet, outlet and solid walls are:
analysis of the process, filmwise condensation is assumed. The
resistance due to condensation is calculated based on the work Inlet: The velocity and air mass fraction are specified at the
of Fujii et al. [8] inlet boundary.
Outlet: The pressure is fixed at the outlet and a mass imbalance
J_ correction scheme is used to update the normal outflow ve-
KX[ 1 +0.276/(x 4 Fr H)}lM Re m 1/2 k/D (13)
R, locity.
Walls: The shell walls of the condenser are assumed to be non-
where
K=0.% for in-line arrangement slip, impervious to flow, and adiabatic.
K= 1.0 for staggered arrangement 2.4 Solution Procedure. The discretization of the differ-
ential equations, equations (2)-(5), is carried out by integrating
X = 0.9 [l + \/(rH)]xn
liner piecewise profiles over the control volumes defined on a
Fr = M2/psg D0; H=cpc(Ts- Tw)/PrcL staggered grid as shown in Fig. 1. In this grid configuration,
the velocity components are stored on the center of the control
Re, = pcM D0/psjic; r= (pciic/ps/j,s)wl volume surfaces, while all other variables, such as pressure,
The fouling resistance (R/) is taken as 3.5*10~5 m 2 K/W as temperature, and air mass fraction are located on the center
suggested by Naviglio et al. [9] and the resistance due to the of the control volume. The control volumes for each velocity
presence of an air film is evaluated via a mass transfer coef- component and for the scalar quantities do not coincide; this
ficient as reported by Caremoli [3] arrangement has the convenient feature of storing the velocity
components at points where they are required for the calcu-
1/3 lation of their convective contribution. Also, the important
(14)
advantage of this type of grid is that the pressure difference
where between two adjacent grid points relates the driving force for
the velocity component located between these grid points.The
= 0.52 and 6 = 0.7 for Re,< 350 resulting discretized equations are solved in primitive variables
a = 0.82 and 6 = 0.6 forRe s >350 using the algorithm proposed by Van Doormaal and Raithby
[15]. Since these equations are coupled together and are highly
The overall resistance to heat transfer for each control vol- nonlinear, an iterative approach is used for their solution. A
ume, R, is the sum of the individual resistances just outlined; cyclic outer iteration is employed comprising the following
thus, R, when related to the outer surface of the tube, can be sequence of operations:
written as
(/') The momentum equations, equations (3) and (4) after
R = Rw-^ + R, + Rc + Rf+Rs (15) discretization, are solved based on a pressure field taken from
the previous iteration.
(Hi) Equation of State. The air and steam mixture is (ii) A Poisson equation for the pressure correction, derived
assumed to behave as an ideal gas; namely, from the continuity equation, equation (2), is solved, and at
Cooling water
outlet
Table 1 Experimental and computational conditions for condenser con- (vi) The overall energy balance for cooling water, steam,
figuration no.1
and condensate is checked.
Arrangement of Tube In-line
2.5 Convergence Criteria. Based on numerical tests, the
Test No. Run No.3 Run No.6
convergence criteria for the overall computational procedure
Test
Conditions
Exp. Cal. Exp. Cal. have been specified as:
Inlet Pressure Pa 5476 b
5424 a 6481 b 6508 a j
a
(/) the Euclidean error norm for the momentum equations
Inlet Temperature C 34.48 34.27 37.55 37.59a|
is less than 10~5;
Inlet Velocity m/s 16.74 16.74 28.21 28.21 J (//) the total continuity error is less than 0.01 percent of the
1 Steam Outlet Pressure Pa 5365 b 5365 b 6345 b 6345 b l mass inflow.
a
Outlet Temperature C 34.11 34.07 37.16 37.123] III Experimental Condenser: Application Example
M'*10+3 kg/s 7.12 6.82 a 14.05 14.09 a l No.
Inlet Temperature C 7.75 7.75 10.72 10.72 I 3.1 Configuration of Condenser No. 1. The experimental
Water Inlet Velocity m/s 0.505 0.505 1.350 1.350 data of fluid flow and heat transfer in the steam surface con-
denser of Fujii et al. [8] is used as one test of the proposed
Outlet Temperature C 14.16 13.86 a 15.44 15.43 a
numerical method. This small steam condenser with a five-
pass layout was designed to study the heat transfer and pressure
8 Arrangement of Tube Staggered B loss in the steam side for the two tube banks of in-line and
staggered arrangements. Figure 2 shows the configurations of
I Test No. Run No.12 Run No.14 I
two tube banks. Each bank was composed of brass tubes of
I Test
Conditions
Exp. Cal. Exp. Cal. 1 14.0 mm o.d., and 10.4 mm i.d., and 100 mm heat transfer
Inlet Pressure Pa 3413 b 3414 a 4739 b 4694 a I length, the tubes were arranged in 15 rows normal to the
a
direction of the steam flow and with 22 mm normal spacing.
Inlet Temperature c 26.22 26.18 31.90 31.69 a
Steam temperature and bulk mean temperature of cooling water
Inlet Velocity m/s 16.85 16.85 27.22 27.22 entering and leaving the tubes were measured by copper-con-
Steam Outlet Pressure Pa 3371 b 3371 b 4560 b 4560b stantan sheath thermocouples of 16 mm diameter. The flow
rate of cooling water through each row was measured by 15
Outlet Temperature C 26.01 25.97 a 31.22 31.18 a
sets of an inverse U-tube water manometers and an orifice,
,
M *10 +:J
kg/s 4.59 4.67 a
10.10 9.18 a and was distributed uniformly within an accuracy of 2 percent.
Inlet Temperature C 7.70 7.70 8.01 8.01 The total condensation rate of the tube bank was evaluated
through the accumulation rate of the condensation in a meas-
Water Inlet Velocity m/s 0.634 0.634 1.186 1.186
uring glass tube attached to a drain separator.
Outlet Temperature C 11.36 10.91 a 11.91 11.52a
3.2 Results and Discussion. The experimental conditions
a - Predicted results; Saturation pressure. of four filmwise condensation test cases (run no.3, run no.6,
run no. 12 and run no. 14 (Fujii et al. [8])), as shown in Table
the end of each outer iteration loop, pressures and velocities 1 were chosen to evaluate the numerical model. In this case,
are corrected. the steam inlet velocity and outlet pressure are given, and the
steam inlet pressures are calculated from the outlet pressure
(Hi) The air mass fraction <j> is obtained from the discretized and pressure drop.
form of its transport equation, equation (5).
(iv) The temperatures of mixture and cooling water, den- (/') Grid-Independence Tests. Calculations were carried
sity, mass source term, and momentum source term are then out in grids of 40x20 in the main and cross-flow directions,
updated. respectively. To test the grid-independence of the solutions,
(v) A new cycle is started unless the prescribed accuracy three different grids, 21 x 10, 40x20 and 59x30, were used.
has been reached. Convergence trends indicated an asymptotic behavior. Com-
between the heat obtained by cooling water and the heat re-
leased from condensed steam, which was always kept at a value
less then 1 percent.
Fig. 6 Grid used for the simulation of condenser configuration no.2 Fig. 10 Shell-side heat transfer coefficient distribution in condenser
configuration no.2
Tube bank j
Vent 1
Steam
inlet
(a) Geometry no.1
0.1m
3m+ 0.7m-
Steam
inlet
this lane, and turns towards the tube bundle due to combined VI Closing Remarks
effects of the rear wall and the "suction" of the tube bundle.
The lowest velocities on the centerline of condenser occur at The numerical procedure proposed in the present study has
x = 0.8 and x=0.1 m for geometry nos. 1 and 2, respectively. shown its capability of predicting the performance of con-
Figure 15 shows the pressure (p) distribution for geometry densers of different geometries. The predictions have produced
no.1, while Figs. 16 and 17 present the distributions of shell- physically meaningful results, but full benchmarking of the
side heat transfer coefficient (a) and heat flux (<?). The lowest procedure, however, requires further testing against experi-
pressure, heat transfer coefficient and heat flux values occur mental data for condensers with different configurations and
in the venting region. It is noted from the contour maps that operating conditions. This information will guide future de-
both the pressure and heat transfer coefficient distributions velopments of the model, particularly in what concerns the
are similar. In both distributions the gradients of pressure and constitutive relations. This particular aspect is of primary im-
shell-side heat transfer coefficient are larger at the tube bundle portance, since vapor superheating, condensate subcooling,
boundaries. Figure 17 shows that the highest heat flux occurs flooding and condensation conditions will dictate the relations
on the first row of tubes as may be expected since the steam for the heat transfer and pressure loss parameters to be em-
velocity is highest there. The heat flux is largest at the tube ployed.
bundle periphery and falls rapidly to lower values in the venting
region. In Fig. 16, it can be seen that the shell-side heat transfer
coefficient decreases sharply from the tube bundle boundaries Acknowledgment
to the inside of bundle. Financial support for this work was received from the Uni-
(a) Geometry no.1 Fig. 16 Shell-side heat transfer coefficient distribution for condenser
configuration no.3 (geometry no.1)
q(KW/i