Nuclear Engineering and Design 328 (2018) 321–332

Contents lists available at ScienceDirect

Nuclear Engineering and Design

journal homepage: www.elsevier.com/locate/nucengdes

A CFD analysis of the flow dynamics of a directly-operated safety relief valve T

a,⁎ b a,c a
N.L. Scuro , E. Angelo , G. Angelo , D.A. Andrade
a
Instituto de Pesquisas Energéticas e Nucleares, Avenida Lineu Prestes, 2242, Centro de Engenharia Nuclear, 05508-000 São Paulo, SP, Brazil
b
Universidade Presbiteriana Mackenzie, Rua da Consolação, 930, 01302-907 São Paulo, SP, Brazil
c
Centro Universitário FEI, Avenida Humberto de Alencar Castelo Branco, 3972-B, 09850-901 São Bernardo do Campo, SP, Brazil

A B S T R A C T

A three-dimensional numerical study on steady state was designed for a safety relief valve using several openings
and inlet pressures. The ANSYS-CFX® commercial code was used as a CFD tool to obtain several properties using
dry saturated steam revised by IAPWS-IF97. Mass flow and discharge coefficient calculated from simulations are
compared to the ASME 2011a Section 1 standard. The model presented constant behavior for opening lifts
smaller than 12 mm and is very reasonable when compared to the standard (ASME). In addition, the conven-
tional procedure to design normal disc force assumes that all the fluid mechanical energy was converted into
work; however, the CFD simulations showed that average normal disc force is about 19% lower than theoretical
ASME force, which could prevent the valve oversizing. A numerical validation was conducted for a transonic air
flow through a converging–diverging diffuser geometry to verify the solver's ability to capture the position and
intensity of a shockwave: the results showed good agreement with the benchmark experiments.

1. Introduction Vu et al. (1994) performed a three-dimensional numerical study of a

safety relief valve working with oxygen gas, aiming to better under-
PWR nuclear power plants contain a primary pressurized circuit of stand the appearance of erosion zones. This study showed that the flow
water at liquid state, which is responsible for heating a secondary steam created multiple vortices, which induced a chaotic cycle and oscillatory
generator system. Both systems must have their flow, pressure, tem- behavior. This situation caused a vibrational movement that caused
perature and other physical properties controlled. Under overheating many impacts to the nozzle, breaking and bringing the circuit to
events, the increasing pressure inside the vessel can lead to an accident danger; therefore, the valve design influenced the chaotic behavior of
if it is not controlled to be within the project limit range. Safety relief the valve. In addition, Vu et al. (1994) showed that the lack of main-
valves installed at each circuit can reduce internal pressure by dis- tenance at a safety relief valve originated little solid waste deposits in
charging water (Jan, 1980). Several factors determine the stability of the nozzle, increasing its internal friction force. The generation of heat
the safety relief valves in order to guarantee their behavior for nuclear in the throat area increased with the intensive use of the valve, resulting
or industrial use (Darby, 2013). However, six main factors are listed in little sparks. Due to the fluid used, small regions initiated the com-
below: bustion process, eroding many parts of the valve. This behavior was
posteriorly discovered as a function of disc chattering.
(i) Valve design, Kasai (1968); In order to study the influence of pressure distribution inside the
(ii) Spring pre-deformation (x), Bazsó & Hös (2013); relief valve, Francis & Betts (1998) conducted a similar study to the one
(iii) Length of the inlet pipeline (L), Thomann (1976); proposed by Vu et al. (1994). The researchers have observed the ex-
(iv) Valve opening length (lift) (δ), Song et al. (2014); istence of a critical backpressure condition of work that leads to un-
(v) Regulator rings position (x′), Green & Woods (1973); stable behavior, but due the number of instability factors, such behavior
(vi) Pressure vessel volume (ϑ ), Song et al. (2014). cannot be immediately simplified.
Among the instability factors, spring pre-deformation was noticed as
These factors have been studied, many times, analyzing the influ- one of the most delicate factors to instability (MacLeod, 1985). As ob-
ence of one factor, because, due to the large number of variables, a served in their study, the more the spring is deformed, the more diffi-
complete analysis is generally not viable. Thus, some studies are pre- cult it becomes to open it for the same designed pressure. Therefore,
sented to explain the influence of each factor on the valve stability. with the increase in spring pre-deformation, higher inlet pressure is

⁎
Corresponding author.
E-mail addresses: [email protected] (N.L. Scuro), [email protected] (E. Angelo), [email protected] (G. Angelo), [email protected] (D.A. Andrade).

https://doi.org/10.1016/j.nucengdes.2018.01.024
Received 3 May 2017; Received in revised form 8 January 2018; Accepted 9 January 2018
Available online 20 January 2018
0029-5493/ © 2018 Elsevier B.V. All rights reserved.
N.L. Scuro et al. Nuclear Engineering and Design 328 (2018) 321–332

Nomenclature ′ fluctuation component

A area (m2) Subscripts

a disc lift (m)
d diameter (m) eff effective
h specific enthalpy (J/kg) IN inlet
k turbulent kinetic energy (m2/s2) N nozzle
kd discharge coefficient [1] OUT outlet
ṁ mass flow rate (kg/s) R real
p pressure (Pa) S standards
Pk shear turbulent production (kg/m.s2) t turbulent
Prt turbulent Prandtl number [1] T throat
Ui vector of velocity (m/s) tot total
ui′ u′j Reynolds stress (m2/s2)
λ thermal conductivity (W/m.K) k−ε (constants)
ε turbulence dissipation rate (m2/s3)
μ dynamic viscosity (Pa.s) Cμ 0.09
ρ density (kg/m3) Cε1 1.44
τ shear stress (Pa) Cε2 1.92
σk 1.00
Superscripts σε 1.30

– average value

necessary to the opening of the valve, increasing the chance of gen- the relief valve behavior. The study showed the appearance of an effect
erating a shockwave. called Hopf Bifurcation, studied by Han and Bao (2009). It rises as inlet
Another factor related to valve instability is how the inlet pressure is pressure gradually increases. If this pressure keeps increasing, it can
increasing. In other words, if the inlet pressure increases slowly with a lead to a chaotic behavior named grazing bifurcation.
constant rate, the mass spring system would have a large time interval Most of researchers showed the enormous number of variables that
to align with natural frequencies, which should be avoided. Otherwise, influences the valve stability and how complex and difficult it is to
the valve would be subjected to a longer vibrational period, that could predict the behavior of the safety relief valve and the relation between
cause severe impacts between the nozzle and the disc (Hayashi et al., each factor. But, as computer simulations have been used as a metho-
1997; Hös et al., 2014). Therefore, when the inlet pressure on safety dology to aid engineering design and performance assessment
relief valve inlet has small oscillations, it can be controlled with spring (Dominguez-Ontiveros & Hassan, 2009), the test of unviable situations
pre-deformation, thus reducing vibration on the disc (Botros et al., becomes more accessible. As an example of this, Song et al. (2014)
1997; Chabane et al., 2009; Misra et al., 2002). Nevertheless, the more performed a transient numerical simulation using CFD methodology to
the fluctuation of the developed flow leads the valve to its natural show the difference between three vessel volumes during discharge.
frequency, the greater will be the fluctuations inside the valve, (Funk, The study reported that the valve needs more time to blow down the
1964; Botros et al., 1997). same pressure difference for bigger vessels, when compared to minor
Nevertheless, valve design and components are not the only factors vessels. It also reported a relation of pressure distribution inside the
that can lead to instability. A theoretical research proposed by valve with the regulator ring position. The position of the regulator
Thomann (1976) indicates the existence of a link between the length of rings was also performed: the study showed that, as the distance be-
the inlet pipe and valve static equilibrium. According to Kasai (1968), tween rings decreases, higher pressure distribution on the disc face is
an oscillatory behavior could be aggravated or delayed by the length of observed, producing higher flow forces. Minor forces on the disc allow
the inlet pipe. This pipe length would be theoretically responsible for the valve to close at higher vessel pressures. The study conducted by
eliminating fluctuations of inlet pressure, thus eliminating abrupt Song et al. (2014) showed the possibilities to analyze many situations in
changes in the velocity field and pressure. safety relief valves as an excellent non-intrusive method.
Green and Woods (1973) conducted a study related not just to In conclusion, many factors influence a stable operation of the
geometric factor, but also to flow properties. The nonlinear behavior of safety relief valve, not just the six major factors presented here, but a
the valve could be caused by four factors: (i) laminar flow transition to combinatorial analysis of all the circuit variables (Darby, 2013), leading
turbulent during the valve process of opening and closing, (ii) balance to an extremely complex situation of predicting the main factor influ-
restoring force, (iii) hysteresis and (iv) fluctuation of the downstream encing the stability of the valve. But it is important to reinforce that the
pressure. The same researcher verified the existence of a relation be- instability is not caused just by geometric factors, but also by the var-
tween the regulator rings position and the force transmitted to the valve iations and relations that the process itself imposes on the valve, such as
disc, influencing in the stability. fluctuations in flow, waste deposits around the body, lack of main-
Another alternative to analyze the influence of many factors in tenance, low boiler efficiency and lack of operator’s capacitation.
safety relief valves is their physical/mathematical analysis, by using Other studies can be conducted by using non-intrusive methods to
specific numerical codes. Lee et al. (1982) showed the influence of a analyze flow fields. Lasers have also been used to compare numerical
safety relieve valve in a PWR primary loop by using RELAP5/MOD3 for and physical solutions for velocity field and mass flow rate according to
transient analysis. This study showed the effects of including support different turbulence models. Laser techniques as DPTV and PIV have
stiffness in piping analysis. This methodology showed a minimization of been used at Texas A&M University to benchmark velocity field for
tank nozzle loads and damping factors. PWR fuel rods, and got good quality data according to Conner et al.
A mathematical study proposed by Licsko et al. (2009) used a (2013) and Dominguez-Ontiveros & Hassan (2014). Another example is
nonlinear ordinary first-order differential equation system to analyze the particle image velocimetry (PIV) for high speed in pipe elbows,

322
N.L. Scuro et al. Nuclear Engineering and Design 328 (2018) 321–332

making possible the comparison between experimental and numerical ∂U ∂Uj ⎞ 2 ∂Uk
analysis (Ono et al., 2011). τi,j = μ ⎛⎜ i + ⎟− μδi,j
⎝ ∂x j ∂x i ⎠ 3 ∂xk (5)
The present single-phase CFD study is based on mapping the in-
tensity of outlet velocity, mass flow rate, discharge coefficient, max- μeff = μ + μt (6)
imum Mach number and disc forces using ANSYS-CFX® code. These
physical quantities are compared to standards in order to verify if they 1
k= ui′ ui′
are properly designed for industrial application. 2 (7)

2. Model development 1
htot = h + Ui Ui + k
2 (8)
A schematic cross-sectional view of the relief valve contemplated in In the study of safety relief valves, the application of different tur-
this study is shown in Fig. 1 with the indication of the eight major bulence models does not provide significant variation in the fluid be-
components. havior according to studies by Dempster and Elmayyah (2013). In ad-
The mechanical operation of the safety relief valve is based on dition, three turbulence models were tested: (i) the k – ɛ model
equilibrium forces acting on the disc. In order to keep the valve closed, (Launder & Spalding, 1974), (ii) model k – ω (Wilcox, 1993), and (iii)
the fluid force must be smaller than the force originated by the internal Menter’s SST k-ω (Menter et al., 2003).
mass plus the spring force. To keep the valve open, the force imposed by Fig. 2 shows a graph for the mass flow through the valve for the
the flow must be greater than the other forces mentioned in balance. tested turbulence models. All boundary conditions (such as pressure
When the sum is different from zero, the valve is not in mechanical difference between inlet and outlet) were kept for all turbulence
equilibrium. models.
The internal pressure of the flow can be up to 10% higher than the Nevertheless, as the percentage change between the responses is
set inlet pressure to reach the maximum lift, and may reach 10% lower relatively small, about 5% for the analyzed mathematical models, the
than the set inlet pressure for the complete closure. Such non-linear k−ε turbulence model has been chosen. The k−ε model provides a
behavior can be caused by several scaling factors according to Song better numerical stability and computational time. The use of the k−ε
et al. (2014). Therefore, several simulations will be carried out at model can also be considered a conservative choice, since it had the
steady-state condition to investigate the flow dynamics for an already lowest value of the mass flow through the valve for the same pressure
open valve. So, the higher inlet pressure and backpressure effect ori- difference (Song et al., 2014).
ginated in the opening or closing process will not affect the simulated The main objective of this paper was to obtain results for compu-
geometry. tational numerical simulations using a commercially available compu-
tational tool at the disposal of engineers and designers, and to compare
3. CFD analysis of SRV the results with the ASME 2011a Section 1 standard, using the re-
cognized good practices for discretization, choice of boundary
A three-dimensional model was developed based on Fig. 1, using the
finite volume method applied to a tetrahedral and prismatic un-
structured mesh. The mathematical model considers the dry saturated
steam flow in steady state condition and time averaging in the con-
servation equations for the following hypotheses: (i) Newtonian fluid
and (ii) Buossinesq assumption for the Reynolds stresses. Thus, the
average conservation equations for the mass, the momentum and en-
ergy are given respectively by Eqs. (1), (2) and (3).
All thermodynamic properties for dry saturated steam were eval-
uated as a function of temperature and pressure. This procedure is ac-
complished by the usage of a property table based on the International
Association for the Properties of Water and Steam (IAPWS-IF97), and
the appropriate formulation for the state equations used in this work is
fully described at Wagner (1998).
∂
(ρUj ) = 0
∂x j (1)

∂ ∂p ∂
(ρUi Uj ) = − + (τi,j−ρui′ u′j )
∂x j ∂x i ∂x j (2)

∂ ∂ ⎛ ∂T μ ∂h ⎞ ∂
(ρUi htot ) = ⎜ λ + t +⎟ [Ui (τi,j−ρui′ u′j )]
∂x i ∂x i ⎝ ∂x i Prt ∂x i ⎠ ∂x j (3)

The flow in the valve has a highly compressible characteristic, thus,

∂
the work due to viscous stresses term ∂x [Ui (τi,j−ρui′ u′j )] was considered.
j
Reynolds stress tensor was calculated (−ρui′ u′j ) according to the
Boussinesq’s hypothesis Eq. (4) and the shear stress (τi,j ) for a com-
pressible gas can be expressed as indicated by Eq. (5). The effective
viscosity (μeff ) , turbulent kinetic energy (k) and the mean total enthalpy
(htot ) are respectively shown at Eqs. (6), (7) and (8).

∂U ∂Uj ⎞ 2 ⎛ ∂Uk ⎞
−ρui′ u′j = μt ⎜⎛ i + ⎟− δi,j ρk + μt
⎜ ⎟

⎝ ∂x j ∂x i ⎠ 3 ⎝ ∂xk ⎠ (4) Fig. 1. Directly-operated SRV model.

323
N.L. Scuro et al. Nuclear Engineering and Design 328 (2018) 321–332

(2001) and Wilson et al. (2001), where the error due to the dis-
cretization is minimized by a successive process solutions and refine-
ments. The numerical solution is considered independent of dis-
cretization when the results, for the same boundary conditions, have
insignificant variation with the increase of the number of elements.
Fig. 4 shows a graph for various physical quantities (mass flow,
average outlet velocity, average outlet static pressure, average outlet
density) as a function of the number of elements, for the same boundary
conditions. The dotted line in this figure indicates that, beyond this line,
the quantities remain almost constant, so that the final mesh was set
with 3.4 million elements. Fig. 5 also shows the relative mesh volume
and distribution used in k−ε turbulence model. Mesh counts with body
Fig. 2. CPU time and mass flow for different turbulence models. sizing of maximum edge length with 2 mm, and 0.8 mm for all internal
faces near the throat area.
conditions and fluid properties. In this sense, the proposed analysis does
not include an extensive investigation of the response or limitations of 3.2. Boundary conditions
the available turbulence model, which could only be done with a set of
experiments properly conducted for this purpose. The summary of the boundary conditions is listed below:
The k−ε turbulence standard model, one of the simplest and most
famous turbulence models available (Ahsan, 2014), is widely used for (i) Average outlet pressure equal to 1 bar (absolute);
the practical engineering flow problems solution Rolander et al. (2006). (ii) Inlet pressure ranging from 0.3 MPa to 1.3 MPa (0.3, 0.5, 0.7, 0.9,
Thus, the expected conclusion is to recognize how the available simu- 1.1 and 1.3 MPa);
lation design tool responds against a well-established and disclosed (iii) Inlet temperature equal to the saturation temperature for the inlet
simple algebraic mathematical mode of determination for relief valves. working pressure;
The k−ε two-equation turbulence model uses the gradient diffusion (iv) The logarithmic wall function is used (Launder & Spalding, 1974)
hypothesis to relate the Reynolds stresses to the mean velocity gradients and the profile is corrected to incorporate fluid compressibility
and the turbulent viscosity (Ansys, 2008, 2006). The turbulent viscosity effects (Huang et al., 1993), using an equivalent roughness of all
for this model is defined by Eq. (9), the calculations of turbulence ki- internal surfaces equal to 0.2 mm according to the valve design.
netic energy (k ) and turbulence dissipation rate (ε ) come, respectively, (v) Symmetry boundary condition in plane xz, Fig. 3;
from Eq. (10) and Eq. (11). (vi) Nozzle opening (a) ranging from 3 mm to 18 mm (3, 6, 9, 12, 15
k2 and 18 mm).
μt = Cμ ρ
ε (9)
The combination of the conditions ii and vi resulted in thirty-six
∂ ∂ ⎡⎛ μ ∂k ⎤ simulations. For the first boundary condition, the flow is considered
(ρUj k ) = ⎜ μ + t⎞ ⎟ + Pk−ρε
∂x j ∂x j ⎢
⎣ ⎝ σ ⎥
k ⎠ ∂x j ⎦ (10) subsonic in the outlet surface, which requires a predefined pressure

∂ ∂ ⎡⎛ μ ∂ε ⎤ ε
(ρUj ε ) = ⎜ μ + t⎞⎟ + (Cε1 Pk−Cε 2 ρε )
∂x j ∂x j ⎢
⎣ ⎝ σε ⎠ ∂x j
⎥
⎦ k (11)
The shear turbulent production (Pk ) due to viscous forces where
indicated at Eq. (12).

∂U ∂Uj ⎞ ∂Ui 2 ∂Uk ⎛ ∂U

Pk = μt ⎜⎛ i + ⎟ − 3μt k + ρk ⎞
⎜ ⎟

⎝ ∂x j ∂x i ⎠ ∂x j 3 ∂x k ⎝ ∂xk ⎠ (12)

3.1. Computational domain and spatial discretization

The computational domain is considered symmetrical, so as to re-

duce the required number of elements to be used without impairing the
accuracy of the mathematical model. The existence of a complex sec-
ondary flow in outlet region is known, as rotational flow and cross flow
through mid-plane. This simplification fails to capture those behaviors;
however, the symmetrical consideration did not significantly influence
the determination of mass flow through the valve or even the force
exerted by the fluid on the disc.
Fig. 3 shows schematically an isometric view of the computational
domain and indicates major regions of the device, which are: (i) inlet,
(ii) nozzle, (iii) throat, (iv) disc, (v) outlet, (vi) increased domain. The
increase in the downstream computational domain is an artificial
strategy to ensure that all components of the velocity field, near to the
outlet region, are pointed out of the computational domain, so as to
avoid a numeric instability due to recirculation and heterogeneity in
velocity field at this region.
The methodology used in this study aimed to verify the appropriate
Fig. 3. Isometric view of the domain and nomenclature for principal regions.
numerical mesh discretization, based on that proposed by Stern et al.

324
N.L. Scuro et al. Nuclear Engineering and Design 328 (2018) 321–332

Fig. 4. Number of elements as a function of mass flow, average outlet velocity, average outlet density and average outlet pressure.

length of the outlet pipe has been increased to guarantee subsonic

outlet.

3.3. Solver options and convergence criteria

The high-resolution implicit total variation diminishing (TVD)

model showed an improvement in shock-capturing schemes, either for
steady or unsteady calculations (Yee et al., 1990). These schemes
showed to be an efficient discretization method and sufficiently accu-
rate for very complex hypersonic inviscid and viscous shock interac-
tions. So, the same type of high-resolution scheme was used in simu-
lations as advection and turbulence scheme discretization.
All steady simulations were performed with spatial convergence
errors lower than 10−5 for root mean square (RMS) for the mass, mo-
mentum, energy and turbulence.

Fig. 5. Relative mesh volume for 3.4 million of elements designed for k-epsilon turbu-
3.4. Numerical validation
lence model.

In the computational domain, under supersonic conditions, it is

value. However, when the flow reaches sonic or supersonic velocity possible to observe values of Mach number greater than one (M > 1). In
near the throat area, the outlet condition needs to be changed to su- such conditions, an intense gradient in physical properties should be
personic outlet and the outlet pressure value is no longer necessary. The observed, which would indicate the existence of a shockwave.
detailed description of this mathematical behavior can be obtained for A validation test was conducted to verify the ability of the solver to
nozzle studies by Anderson (2002). Mathematically, these two flow adequately predict shockwave formation, position and intensity of the
possibilities, subsonic and supersonic, impact on the mathematical be- physical phenomenon. The test was performed for a transonic air flow
havior of the conservation equations, changing the solution from el- through a converging-diverging diffuser geometry due to an extensive
liptic to hyperbolic. In order to avoid mathematical inconsistencies, the experimental data provided and a wide range of the flow conditions

325
N.L. Scuro et al. Nuclear Engineering and Design 328 (2018) 321–332

(Chen et al., 1979; Sajben et al., 1981; Bogar 1986; Bogar et al., 1983; where, dT is the inlet throat diameter and a is the disc lift.
Salmon et al., 1983). The geometry of the convergent-divergent was Fig. 10 shows the behavior of the throat area towards inlet and
generated according to the equation presented by Bogar et al. (1983). outlet areas. It was found that, with the increase of the disc lift (a ), the
A three-dimensional model for the transonic diffuser was developed throat area tends to be equal to the inlet area. This occurs for disc lift
based on the finite volume method applied to an unstructured mesh in a equal to 14.25 mm.
steady state condition. The mass, the momentum and energy con- It is observed that, for openings smaller than 14.25 mm
servation equations are considered. The turbulence models used are the (a < 14.25 mm), the mass flow through the valve is limited by the
same ones presented in the valve case. The mesh was verified in- throat area. If the throat area is greater than the inlet area, the mass
dividually for each turbulence model according to Stern et al. (2001) flow through the valve is limited by the valve inlet region. This char-
and Wilson et al. (2001) criteria. acteristic justifies the behavior of the constant average outlet velocity
The air was considered as calorically-perfect gas and Sutherland’s for a disc lift higher than 12 mm.
law (Sutherland, 1893) was used for the viscosity correction with
temperature. The boundary conditions applied to the validation test are 4.2. Mass flow
schematically indicated at Fig. 6 and listed below:
ASME 2011a Section 1 provides standards for relief and pressure
(i) subsonic inlet with total absolute pressure equal to 135 kPa and valves for the calculation of the theoretical flow for both smaller
temperature equal to 300 K; openings ( AIN > AN ) and bigger openings ( AN > AIN ). In ASME 2011a
(ii) outlet static pressure equal to 101.8 kPaabs; standards, the mass flow calculation (theoretical condition) for smaller
(iii) no-slip wall condition for upper and lower surfaces; and openings is classified as “Flat Seat”, and the one for bigger openings is
(iv) two symmetry conditions (planes α and β ). classified as “Nozzle”.
The mass flow calculations for the Flat Seat and Nozzle conditions
The use of symmetry on both side surfaces is required since ANSYS- are respectively indicated in Eqs. (14) and (15).
CFX® is not able to solve purely two-dimensional problems.
ṁ S = 5,25·AT pIN (14)
Fig. 7 presents the results obtained by simulations compared to the
experiments (Sajben et al., 1981). The comparison is performed for a ṁ S = 5,25·AIN pIN (15)
horizontal line located on the lower surface of the computational do-
main, with y = 0 in the Cartesian system positioned according to Fig. 6. where, pIN is the inlet pressure (SI).
In Fig. 7, the length (x) and absolute pressure (p) was normalized, Fig. 11 shows a comparison between the theoretical mass flow rate
where hT (40 mm) is the throat height and p0 an absolute reference calculated by the standard and mass flow rate obtained by the simu-
pressure (135 kPa). lation. It is important to note that the same behavior observed in
Analyzing Fig. 7, it was possible to verify that the solver can capture average outlet velocity is verified for mass flow. It is also possible to
the existence of the shockwave independently of the turbulence model observe that in all simulated cases, the mass flow for the same inlet
used. It was noticed that among the turbulence models, the Shear Stress pressure is lower than that calculated by the standards. This behavior is
Transport model predicted more adequately the phenomenon, pre- also expected since mechanical energy losses are not considered in Eqs.
senting an absolute error of 7% for the shockwave position (14) and (15).
( x / hT = 2.413) while k−ε (10.5%, x / hT = 2.492 ) and k−ω (15.7%,
x / hT = 2.609 ) delay the formation of the shockwave. 4.3. Discharge coefficient
Fig. 8 presents a contour map for each tested turbulence model,
indicating the normalized absolute static pressure. In this figure, it is Calculations of the valve discharge coefficient were based on the
possible to notice an abrupt pressure variation in the shockwave posi- ASME 2011a formulation, Eq. (16). The real mass flow rate (ṁ R ) was
tion. obtained by the simulations and standards mass flow (ṁ S ) according to
Eqs. (14) and (15).
4. Results kd = ṁ R / ṁ S (16)

4.1. Average outlet velocity

The first behavior that indicates consistency in the results is ob-

served in Fig. 9. In this figure, the average outlet velocity as a function
of the throat opening for different inlet pressures is shown. The increase
in the inlet pressure results in an increase in the average outlet velocity.
However, for the same inlet pressure, the increase of the throat opening
results in a velocity increase for opening values of less than 12 mm and
becoming almost constant from beyond this point.
The behavior shown in Fig. 10 can be justified by observing the
three main regions of the flow passage:

– Domain inlet area ( AIN ), coincident with the valve inlet area;
– Outlet region area ( AOUT ), previously shown in the Fig. 3 and equal
to the outlet area of the valve;
– Throat area (AT ) , flow area formed between nozzle and disc.

Eq. (13) indicates the approximate throat area. This equation ne-
glects the effects of vena contracta formed in the region of the throat.
The flow area in this region is assumed as the lateral area of a cylinder.
Fig. 6. Indication of the principal dimensions in millimeters of the convergent-divergent
AT = π·dT ·a (13) geometry and boundary conditions.

326
N.L. Scuro et al. Nuclear Engineering and Design 328 (2018) 321–332

Fig. 7. Pressure variation as a function of length on the lower surface of the computa-
Fig. 10. Comparison between inlet area, outlet area and throat area as a function of the
tional domain.
disc lift.

Fig. 8. Contour map indicating absolute static pressure ( p/ p0 ) for: (a) Shear Stress
Transport; (b) k−ε and (c) k−ω .

Fig. 11. Mass flow rate as a function of disc lift.

Fig. 9. Behavior of the average outlet velocity in function of the disc lift (a) for each
simulation.

Values for the discharge coefficient as a function of the disc lift are
shown in Fig. 12. The bars indicate the maximum and minimum dis-
persion in the discharge coefficient for the same disc lift. The average
values of the discharge coefficient (kd ) are indicated in Table 1. This
table also shows the standard deviation (σ ) and the maximum deviation
defined by Eq. (17). Fig. 12. Discharge coefficient as a function of the disc lift.

327
N.L. Scuro et al. Nuclear Engineering and Design 328 (2018) 321–332

Table 1 such as velocity, density, pressure. However, their maximum and

Average discharge coefficient, standard deviation and maximum deviation as a function minimum values were not computed, since it is out of the scope of this
of the disc lift.
study.
a [mm] kd σ δMAX [%] In this case, obtaining the maximum values of the Mach number is
of great interest, since this value exhibits the valve behavior. Mach
3 0.975 0.00609 1.64 average values for the outlet region in the valve can hide a possible
6 0.957 0.01799 5.07 supersonic behavior.
9 0.951 0.01234 3.14
To verify the existence of shockwaves, Fig. 13 indicates the max-
12 0.939 0.01719 4.79
15 0.803 0.01081 3.52 imum Mach number in the complete computational domain according
18 0.773 0.01092 3.62 to the inlet Reynolds number.
Fig. 14 shows streamlines behavior nearby the throat area. The
existence of higher velocity profiles, coupled with generation of large
turbulent frequencies, prevails in the downstream to the throat area.
This behavior is known and widely reported by Kasai (1968). This re-
circulating structure promotes instability behavior on the discs ex-
perimentally demonstrated by Hős et al. (2014). Depending on the
magnitude due to collisions between disc and valve seat, according to
Vu et al. (1994), the instability can cause premature wear.

4.5. Shockwave

Even if the flow does not behave as a purely one-dimensional flow, a

simple analysis of the average general behavior of the flowing fluid
reveals a peculiar pattern. Fig. 15 indicates a graph constructed for a
vertical line from the valve inlet to the center of the disc. It is not the
Fig. 13. Maximum Mach number in the computational domain as a function of the inlet analysis of a streamline, but a generic analysis of some flow properties
Reynolds number. of interest in that region. In Fig. 15 it is possible to observe, in the flow
direction, the initial Mach number in the region where there is the first
decrease of the diameter (alpha region indicated in Fig. 17 and Fig. 18).
Being a subsonic flow, the area reduction really should lead to an ac-
celeration of the fluid, as observed. Subsequently, a second area re-
duction (beta region indicated in Fig. 17 and Fig. 18) again increases
fluid velocity. The complex feature of the three-dimensional flow begins
to be noted just upstream of the section “x-x” to the valve disc region.
The central flow in the latter region results in a Mach number decrease
(because of the valve disc), which can configure the presence of a
shockwave. The results of the mathematical model indicate a dispersion
of this shockwave that looks like a bowl shape, located in a detached
position and in front of the valve disc. The more peripheral flow in the
region of section “x-x” remains sonic, showing an increase in the area
caused by the geometry of the valve in this region, as a function of
chamfer (with the possible presence of Prandtl Meyer expansion
waves), which allows supersonic fluid velocity. In sequence, radial flow
(and increasing areas) further accelerates the supersonic vapor.
Fig. 14. Streamlines at full lift and pressure. A horizontal line, also constructed for flow analysis, is shown with
quantities of interest indicated in Fig. 16. The radial acceleration of the
fluid is locally influenced by the geometric details on the valve disc
ṁ R (max )−ṁ R (min)
δMAX = 100 (grooves), though, after passage of the fluid by the region with the
ṁ R (max ) (17)
grooves, there is a continuous increase in the Mach number to the re-
Considering all the uncertainties in this study, the values showed gion after the disc where the valve diameter is constant, which tends to
similarities with studies carried out by Kim et al. (2006) and Schmidt equalize the flow quantities.
et al. (2009) for the discharge coefficients. However, some particula- The Mach equal to one should occur in the section of the throat of
rities should be briefly discussed, such as the maximum deviation be- the valve (lower area between the valve seat and the disc), admittedly
havior observed for each valve opening. Between the total opening the section with smaller area in the flow. However, the presence of the
(18 mm) up to 9 mm, there were no large variations of the maximum disc and the possible shockwave, imposed a restriction, forced a three-
deviation, however, for 6 mm, it reached 5.07%, followed by the dimensional flow pattern, causing an unexpected and atypical central
minimum statistical value (1.64%). Among the possibilities, it is be- hydraulic block, advancing the smallest flow area (real geometric
lieved that the position of the regulation rings has a delicate influence throat area) to a virtual throat area located upstream of the actual
on the smaller openings, because, as discussed in item 4.6, the 6 mm geometric throat area. A phenomenon comparable to a kind of vena
aperture obtained the largest portion of force transmitted to the disc. contracta. Thus, the pressure distribution on the disc front is lower than
the one obtained if the flow is without the shockwave detached pre-
sence, and the anticipation of the smaller area of free flow for a section
4.4. Mach number downstream valve, justifying the lowest force disc results found in the
simulations (results compared to the empirical model).
Average values were obtained for the calculation of some quantities

328
N.L. Scuro et al. Nuclear Engineering and Design 328 (2018) 321–332

Fig. 15. Physical properties in vertical line (zeta

line).

Fig. 16. Physical properties in horizontal line (eta

line).

4.6. Disc force neglects any other three-dimensional behavior of the flow, regardless of
the flow conditions. The force obtained by this procedure is used to
The conventional manner to determine the force applied on the disc calibrate the opening/closing condition, which is used in the spring
is to convert all the work done by the pressure at the inlet into a normal design. However, this procedure implies that the force acting on the
force applied on the disc. It does not consider vortex formations and disc for the same inlet pressure remains constant.

329
N.L. Scuro et al. Nuclear Engineering and Design 328 (2018) 321–332

Fig. 17. Gray map for absolute pressure in symmetrical plane.

Fig. 18. Gray map for Mach number in symmetrical plane.

330
N.L. Scuro et al. Nuclear Engineering and Design 328 (2018) 321–332

disc. The CFD simulations showed that average normal disc force is
about 19% lower than theoretical ASME force, which could prevent the
valve oversizing.

5. Conclusions

A three-dimensional numerical study is presented for a security

relief valve using the ANSYS-CFX® commercial code. The study mapped
various physical situations – particularly, the transonic and supersonic
conditions.
The mathematical model considered steady state flow and included
dry saturated steam at extremely compressible condition. All thermo-
dynamic properties were corrected International Association for the
Properties of Water and Steam. The discretization of the computational
domain was appropriately analyzed using recommended techniques in
scientific and technical literature.
This study, conducted in accordance with several opening positions
and inlet pressure conditions, resulted in 36 steady state simulations. To
obtain the transonic and/or supersonic behavior, the increment of the
outlet domain allows the use of a subsonic boundary condition for the
outlet. It allows the model to capture all the phenomena inside the
valve using the same outlet boundary conditions, including a possible
shockwave condition.
One of the properties responsible for predicting relief valve behavior
efficiency is the discharge coefficient. As shown in item 4.2, the valve
presented coefficients and mass flow according to the ASME 2011a
Section 1 safety standard.
Due to the number of Mach reaching high values in certain simu-
lations (M > 0.7), the turbulence originated by transonic and super-
sonic situations could induce vibrations in the disc, which could affect
stability of the valve, as shown by Bardina et al. (1997).
The minimum flow area, according to the results of the simulations,
was anticipated, not corresponding to the smaller flow area between
the disc and the valve seat, an effect possibly caused by the presence of
the disc (central flow restriction causing a possible bowl-shaped
shockwave) of the valve and a small divergent anterior to the seat re-
gion, forming a specimen of the vena contracta.
The relief valve operated by spring requires several adjustments at
the time of installation, which must be in accordance with the re-
spective standard. Nuclear and industrial hydraulic circuits operate
with different patterns and variations. Working with steam, PWR and
Fig. 19. Normal force at the disc sealing. BWR nuclear reactors have their physical quantities constantly mon-
itored, and their variation must be predicted to guarantee safety op-
Fig. 19 presents a comparison between the normal force on the disc eration.
sealing by the theoretical condition, which considers that all static inlet The factors that may alter significantly the operation of the valve
pressure is converted into mechanical work in the disc area, and the are presented in item 1, where: Valve design (Kasai, 1968); Spring pre-
force obtained by the simulation as a function of the disc lift. Fig. 19 deformation (x) (Bazsó & Hös, 2013); Length of the inlet pipe line (L),
was separated into groups, to facilitates the visualization according to (Thomann, 1976); Valve opening length (δ) (Song et al., 2014) and
the inlet pressure. This figure also revealed that the force acting on the (Singh, 1982); Regulator rings position (x’) (Green & Woods (1973);
disc presents a non-linear behavior and cannot be considered constant. and Pressure vessel volume (ϑ ) (Song et al., 2014). These six variables
It is possible to observe that the magnitude of the inlet pressure does are defined as the most critical in its operation.
not matter. For a given disc lift, the force transmitted to the sealing disc As studied by Bazsó and Hös (2013), it is impossible to get a “perfect
is always lower than the theoretical value. valve adjustment” where the valve will not result in vibrations and
This study can support the spring design analysis in pressure relief instabilities by such many variables. It is important to mention that
valves in order to avoid oversizing. It is known that an incorrect spring valve maintenance and testing practices influence directly the valve
selection can cause intense vibrational behavior or the valve remains behavior, McElhaney (2000).
closed at the design condition (Bazsó and Hös, 2013). As the most influential variable, the course of the disc was the major
Fig. 19 shows an interesting behavior for each opening. Starting object of this study; all the other variables were kept constant.
from a 3 mm opening, the transmitted mechanical force portion grows Theoretical analysis showed a critical behavior in the 14.25 mm
to a peak at 6 mm, decreasing linearly up to a 18 mm opening. This opening, which would present an equality between the inlet area and
behavior may be tied to the position of the adjusting rings, where, for throat area, thus providing a constant behavior after this opening.
minimum openings, both rings remain near to each other, increasing Simulations showed this behavior for lift smaller than 12 mm. It could
the area with which the fluid can convert its kinetic energy to me- help in the discharge coefficient design and also avoid errors during
chanical. However, for larger openings, the superior adjusting ring hydraulic sizing.
become far away from nozzle, reducing the force transmitted to the The vibrational elements were compared to Francis & Betts (1998)
and Green & Woods (1973) papers to determine similarities with the

331
N.L. Scuro et al. Nuclear Engineering and Design 328 (2018) 321–332

relief valve of this study. The mathematical model used by Darby 258–271.
(2013) assisted in the prediction of studied factors such as the vibra- Green, W.L., Woods, G.D., 1973. Some causes of chatter in direct acting spring loaded
poppet valve. In: The 3rd International Fliud Power Symposium. Turin.
tional behavior of the valve. So, the presence of high rotational eddy at Han, W., Bao, Z., 2009. Hopf bifurcation analysis of a reaction–diffusion Sel'kov system. J.
Fig. 14 could indicate instabilities (Galbally et al., 2015). Math. Anal. Appl. 356 (2), 633–641.
Although this CFD study has not been validated experimentally, a Hayashi, S., Hayase, T., Kurahashi, T., 1997. Chaos in a hydraulic control valve. J. Fluids
Struct. 11 (6), 693–716.
numerical validation has been compared to extensive experimental data Hős, C.J., Champneys, A.R., Paul, K., McNeely, M., 2014. Dynamic behavior of direct
provided and a wide range of the flow conditions (Chen et al., 1979; spring loaded pressure relief valves in gas service: model development, measure-
Sajben et al., 1981; Bogar, 1986; Bogar et al., 1983; Salmon et al., ments and instability mechanisms. J. Loss Prev. Process Ind. 31, 70–81.
Huang, P., Bradshaw, P., Coakley, T., 1993. Skin friction and velocity profile family for
1983). The numerical code proves to be capable of predicting shock- compressible turbulent boundary layers. AIAA J. 31 (9).
wave phenomena as illustrated in Fig. 17 and Fig. 18 for safety relief Jan, C., 1980. Dynamic effects on structures and equipment due to safety relief valve
valve and Fig. 8 for the converging-diverging diffuser geometry. Other discharge loads. Nucl. Eng. Des. 59, 171–183.
Kasai, K., 1968. On the stability of a poppet valve with an elastic support: 1st report,
studies also achieved similar results for wave formations for relief valve
considering the effect of the inlet piping system. Bull. JSME 11 (48), 1068–1083.
using CFD methodology (Bassi et al., 2011, 2014). An important study Kim, H.-D., et al., 2006. A study of the gas flow through a LNG safety valve. J. Therm. Sci.
also revealed the shockwave formations after safety/relief valve dis- 15 (4), 355–360.
charge (Moody, 1982). Launder, B.E., Spalding, D.B., 1974. The numerical computation of turbulent flows.
Comput. Methods Appl. Mech. Eng. 3 (2), 269–289.
Lee, M., Chou, L., Yang, S., 1982. PWR safety/relief valve blowdown analysis experience.
Acknowledgments Nucl. Eng. Des. 72 (3), 421–427.
Licsko, G., Champneys, A., Hös, C., 2009. Nonlinear analysis of a single stage pressure
relief valve. Int. J. Appl. Math 39 (4), 12–26.
The authors would like to acknowledge the Nuclear and Energy MacLeod, G., 1985. Safety valve dynamic instability: an analysis of chatter. J. Press.
Research Institute, IPEN-CNEN/SP, for allowing the use of its infra- Vessel Technol. 107 (2), 172–177.
structure, particularly the computer laboratory. McElhaney, K., 2000. An analysis of check valve performance characteristics based on
valve design. Nucl. Eng. Des. 169–182.
Menter, F.R., Kuntz, M., Langtry, R., 2003. Ten years of industrial experience with the SST
References turbulence model. Turbulence Heat Mass Transfer 4 (1).
Misra, A., Behdinam, K., Cleghorn, W.L., 2002. Self-excited vibration of a controlvalve
due to fluid-structure interaction. J. Fluids Struct. 16 (5), 649–665.
Ahsan, M., 2014. Numerical analysis of friction factor for a fully developed turbulent flow
Moody, F., 1982. Unsteady piping forces caused by hot water discharge from suddenly
using k-epsilon turbulence model with enhanced wall treatment. Beni-Suef Univ. J.
opened safety/relief valves. Nucl. Eng. Des. 72 (2), 213–224.
Basic Appl. Sci. 3 (4), 269–277.
Ono, A., Kimura, N., Kamide, H., Tobita, A., 2011. Influence of elbow curvature on flow
Anderson, J., 2002. Modern Compressible Flow: With Historical Perspective, third ed.
structure at elbow outlet under high Reynolds number condition. Nucl. Eng. Des. 241
McGraw-Hill Education, s.l.
(11), 4409–4419.
Ansys, 2008. ANSYS CFX-Pre User's Guide, 15.0 ed. Ansys, Canonsburg.
Rolander, N., et al., 2006. An approach to robust design of turbulent convective systems.
Ansys, C., 2006. Solver Theory Guide. Ansys CFX Release, Issue 2006.
Am. Soc. Mech. Eng. 128, 844–855.
Bardina, J., et al., 1997. Turbulence Modeling Validation. National Aeronautics and Space
Sajben, M., Bogar, T., Kroutil, J., 1981. Forced oscillation experiments in supercritical
Administration, pp. 1–100.
diffuser flows with application to ramjet instabilities. AIAA Paper 1, 487.
Bassi, F., et al., 2011. High-order discontinuous Galerkin computation of axisymmetric
Salmon, J., Bogar, T., Sajben, M., 1983. Laser Doppler velocimeter measurements in
transonic flows in safety relief valves. Comput. Fluids 49 (1), 203–213.
unsteady, separated, transonic diffuser flows. AIAA J. 21 (12), 1690–1697.
Bassi, F., et al., 2014. Investigation of flow phenomena in air–water safety relief valves by
Schmidt, J., Peschel, W., Beune, A., 2009. Experimental and theoretical studies on high
means of a discontinuous Galerkin solver. Comput. Fluids 90, 57–64.
pressure safety valves: sizing and design supported by numerical calculations (CFD).
Bazsó, C., Hös, C.J., 2013. An experimental study on the stability of a direct spring loaded
Chem. Eng. Technol. 32 (2), 252–262.
poppet relief valve. J. Fluids Struct. 42, 456–465.
Singh, A., 1982. An analytical study of the dynamics and stability of a spring loaded safety
Bogar, T., 1986. Structure of self-excited oscillations in transonic diffuser flows. AIAA J.
valve. Nucl. Eng. Des. 72 (2), 197–204.
24 (1), 54–61.
Song, X., et al., 2014. A CFD analysis of the dynamics of a direct-operated safety relief
Bogar, T., Sajben, M., Kroutil, J., 1983. Characteristic frequencies of transonic diffuser
valve mounted on a pressure vessel. Energy Convers. Manage. 81, 407–419.
flow oscillations. AIAA J. 21 (9), 1232–1240.
Stern, F., Wilson, R., Coleman, H.W., Paterson, E.G., 2001. Comprehensive approach to
Botros, K.K., Dunn, G.H., Hrycyk, J., 1997. Riser-relief valve dynamic interactions. J.
verification and validation of CFD simulations—part 1: methodology and procedures.
Fluids Eng. 119 (3), 671–679.
J. Fluids Eng. 123 (4), 793–802.
Chabane, S., et al., 2009. Vibration and chattering of conventional safety relief valve
Sutherland, W., 1893. LII. The viscosity of gases and molecular force. London Edinburgh
under built up back pressure. In: 3rd IAHR International Meeting of the WorkGroup
Dublin Philos. Mag. J. Sci. 36 (223), 507–531.
on Cavitation and Dynamic Problems in Hydraulic Machinery and Systems, pp.
The American Society of Mechanical Engineers, 2011. ASME BPVC Section I - Rules for
281–294.
Construction of Power Boilers - Certification of Capacity of Pressure Relief Valves -
Chen, C., Sajben, M., Kroutil, J.C., 1979. Shock-wave oscillations in a transonic diffuser
PG-69.2.3. ASME, New York.
flow. AIAA J. 17, 1076–1083.
Thomann, H., 1976. Oscillations of a simple valve connected to a pipe. Z. Angew. Math.
Conner, M.E., Hassan, Y.A., Dominguez-Ontiveros, E.E., 2013. Hydraulic benchmark data
Phys. 27 (1), 23–40.
for PWR mixing vane grid. Nucl. Eng. Des. 264, 97–102.
Vu, B., Wang, T.-S., Shih, M.-H., Soni, B., 1994. Navier-Stokes flow field analysis of
Darby, R., 2013. The dynamic response of pressure relief valves in vapor or gas service,
compressible flow in a high pressure safety relief valve. Appl. Math. Comput. 65 (1),
part I: mathematical model. J. Loss Prev. Process Ind. 26 (6), 1262–1268.
345–353.
Dempster, W., Elmayyah, W., 2013. Two phase discharge flow prediction in safety valves.
Wagner, W., 1998. Properties of Water and Steam: The Industrial Standard IAPWS-IF97
Int. J. Press. Vessels Pip. 110, 61–65.
for the Thermodynamic Properties and Supplementary Equations for Other
Dominguez-Ontiveros, E.E., Hassan, Y.A., 2009. Non-intrusive experimental investigation
Properties: Tables Based on These Equations. Springer Verlag, s.l.
of flow behavior inside a 5 × 5 rod bundle with spacer grids using PIV and MIR. Nucl.
Wilcox, D.C., 1993. Turbulence Modeling for CFD.–Glendale: DCW Industries, first ed.
Eng. Des. 239 (5), 888–898.
DCW Industries, Inc., California.
Dominguez-Ontiveros, E., Hassan, Y.A., 2014. Experimental study of a simplified 3 × 3
Yee, H.C., Klopfer, G.H., Montagne, J.-L., 1990. High-resolution shock-capturing schemes
rod bundle using DPTV. Nucl. Eng. Des. 279, 50–59.
for inviscid and viscous hypersonic flows. J. Comput. Phys. 88 (1), 31–61.
Francis, J., Betts, P.L., 1998. Backpressure in a high-lift compensated pressure relief valve
Wilson, R., Stern, F., Coleman, H., Paterson, E., 2001. Comprehensive approach to ver-
subject to single phase compressible flow. J. Loss Prev. Process Ind. 11 (1), 55–66.
ification and validation of CFD simulations. 2. Application for RANS simulation of a
Funk, J.E., 1964. Poppet valve stability. J. Basic Eng. 86 (2), 207–212.
cargo/container ship. J. Fluids Eng. 123 (4), 803–810.
Galbally, D., et al., 2015. Analysis of pressure oscillations and safety relief valve vibra-
tions in the main steam system of a Boiling Water Reactor. Nucl. Eng. Des. 293,

332