Tunnelling and Underground Space Technology 118 (2021) 104185

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology

incorporating Trenchless Technology Research
journal homepage: www.elsevier.com/locate/tust

Computational studies on design optimization for Saccardo nozzle in

tunnel ventilation
Joo Guan Hang a, Hui An a, *, Peng Cheng Wang a, Simon Ching Man Yu b, Samuel Chan c,
Melvyn Thong c, Lean Wan Lim c, Mun Kit Cheong c, Gerald Louis c
a
Engineering Cluster, Singapore Institute of Technology, 10 Dover Drive, Singapore
b
Aeronautical and Aviation Engineering, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong
c
Land Transport Authority of Singapore, 1 Hampshire Road, Singapore

A R T I C L E I N F O A B S T R A C T

Keywords: Longitudinal ventilation system with Saccardo nozzles has been widely used to ventilate long road tunnels to
CFD maintain acceptable air quality and thermal comfort. The design of Saccardo nozzle impacts the ventilation
Saccardo nozzle power consumption, nozzle efficiency and momentum exchange in the tunnel, which inherently affects the
Tunnel ventilation
tunnel air temperature. This paper presents a detailed study of the performance of an optimised Saccardo Nozzle
Optimization study
design using Computational Fluid Dynamics (CFD) simulation. The CFD model was combined with a mono-
Nozzle efficiency
dimensional temperature prediction model to evaluate the carry-over effects in the tunnel. The impact of
various tunnel design parameters was parametrically evaluated to identify the key important geometric pa­
rameters that affect the Saccardo nozzle performance. The design of the nozzle was finally optimized under
constant power consumption conditions and compared against that in an existing baseline road tunnel. The
optimized nozzle design shows a significant reduction in the overall maximum tunnel air temperature. The
nozzle efficiency shows a 9% improvement. In addition, an alternative nozzle design was also proposed by
introducing air bleeds underneath the Saccardo nozzle. The air bleeds design could further reduce the tunnel air
temperature by 6% with a 5% improvement in the nozzle efficiency. The air bleeds design is considered to be a
feasible solution to be adopted into the existing tunnel with minimal modification work required.

Saccardo nozzle are classified as impulse systems. Jet fan works as a

1. Introduction blower where their kinetic energy is transferred to a small portion of the
tunnel air, which accelerates, and generates an induced air movement in
A tunnel ventilation system is designed to provide an acceptable the tunnel. The fresh air is induced from the entry portal, and it flows
environmental condition within the road tunnel during normal or con­ longitudinally in the direction of traffic flow and carries the heat out of
gested operation, as well as to provide a tenable environment in the the tunnel via the exit portal. For Saccardo nozzle ventilation, fresh air is
event of an emergency evacuation. The tunnels are mechanically delivered into the tunnel via the nozzle by axial fans in a ventilation
ventilated to dilute and extract both heat and pollutant emissions from building/shaft. As the ejected air moves in the designed direction, it
vehicles to maintain acceptable temperatures and pollutant concentra­ undergoes momentum exchange and entrains into the tunnel airflow. In
tion levels during normal traffic conditions. In general, tunnel ventila­ a longitudinal ventilation design, the heat emissions by vehicles are
tion systems are categorized into longitudinal, semi-transverse and full being carried by the tunnel-induced flow. An exhaust damper is placed
transverse ventilation systems. For longitudinal ventilation, the air in the downstream location to extract the hot exhaust air using axial
movement through the tunnel can be accomplished by using either jet fans. In typical long road tunnels exceeding several kilometers in length,
fans or Saccardo nozzle. Saccardo nozzles are typically employed in long there could be several pairs of Saccardo nozzles and exhaust dampers,
road tunnels due to their ability in providing fresh air intake, without where each pair is placed at an interval apart from each other in the
compromising the tunnel height, and reduce the frequency of lane tunnel. Fig. 1(a) shows the typical schematic of a tunnel with Saccardo
closure during maintenance activities as compared to jet fans (Tarada nozzles and exhaust dampers.
and Brandt, 2009; Tabarra et al., 2000; Kwa, 2004). The jet fans or Despite the aforementioned merits for Saccardo nozzle ventilation

* Corresponding author.
E-mail address: [email protected] (H. An).

https://doi.org/10.1016/j.tust.2021.104185
Received 21 December 2020; Received in revised form 26 August 2021; Accepted 6 September 2021
Available online 15 September 2021
0886-7798/© 2021 Elsevier Ltd. All rights reserved.
J.G. Hang et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 118 (2021) 104185

Nomenclature μ dynamic viscosity of air (Ns/m2)

μt turbulent dynamic viscosity
At tunnel cross-section area (m2) ρ density of air (kg/m3)
An nozzle inlet area (m2) λ thermal conductivity of air (W/m·K)
Cp specific heat capacity of air (J/kg·k) σk turbulent Prandtl number for k
C1ε, C2ε, Cμ turbulent model constants σε turbulent Prandtl number for ε
Gk production of turbulent kinetic energy ηn efficiency of nozzle (%)
k turbulent kinetic energy (m2/s2) ηe efficiency of exhaust (%)
Kmx momentum exchange coefficient of nozzle θ incline angle of nozzle (◦ )
p Pressure (Pa)
Prt turbulent Prandtl number Subscripts
Q volume air flow rate (m3/s) 1,2,3 stages in the tunnel
qheat heat emission from the tunnel floor (J) i, j indexes
t Time (s) e exhaust
T Temperature (◦ C) s static pressure
ui velocity at stage i (m/s) t total pressure
xi cartesion coordinates at i direction n nozzle
Δx mesh size (m)

Fig. 1. (a) The schematic diagram for the deployment of Saccardo nozzles and exhaust dampers in tunnel. (b) Temperature profiles along the tunnel.

system, there still exist technical challenges that need to be addressed Saccardo nozzle were measured with various nozzle designs. The air
for a more viable and people-center tunnel transportation system. The blowing from the Saccardo nozzle caused pressure rises in the tunnel,
findings from Kwa (2004) showed that the temperature and pollutant which in turn affects the efficiency of the ventilation system. There are
levels increase in the direction of traffic along the tunnel and then drop two momentum exchanges in the tunnel, which happen at the Saccardo
upon passing each ventilation building/shaft. The temperature and nozzle and the exhaust locations, hence resulting in a 2-stage pressure
pollutant level rise in each subsequent tunnel section forming an rise. Long the tunnel section, a steady pressure drop trend was observed
ascending saw-tooth profile as shown in Fig. 1b, due to the carry-over due to the flow resistance of the tunnel. Their experiment result
effect in the tunnel where the heat and pollutants are not effectively (Yoshizawa et al., 1991) shows that the ventilation efficiency was highly
exhausted by the fans in ventilation buildings /shafts. The carry-over influenced by the nozzle aspect ratio at the high flow ratio region.
effect poses a challenge to tunnel ventilation system designers as the Furthermore, the momentum exchange coefficients were not signifi­
tunnel air temperature could become higher than their desired value. cantly affected by the flow ratio. Costeris and Sweetland (1994) showed
Therefore, in this present study, it is imperative to investigate the that the shape of the Saccardo nozzle has a large influence on the airflow
effectiveness of Saccardo nozzle design, in terms of its tunnel-induced pattern in the tunnel. A bad design in nozzle shape would lead to an
airflow, power consumption, nozzle efficiency, and the maximum tem­ unstable and uneven flow in the nozzle opening. The turbulent flow was
perature resulted from the carry-over effect. believed to have led to a high loss in momentum energy and thus lower
Yoshizawa et al. (1991) experimentally tested a longitudinal venti­ nozzle efficiency.
lation system that has a Saccardo nozzle and exhaust in a model tunnel. Computational Fluid Dynamics (CFD) simulation has been used to
The ventilation efficiency and air momentum exchange coefficient of the analyse tunnel ventilation air flow under various objectives. Wang et al.

2
J.G. Hang et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 118 (2021) 104185

Table 1 Saccardo nozzle design in a straight tunnel section. Under the assump­
Spatial discretization method used in simulation. tion of a single fan operating condition, a parametric study was per­
Governing Equations Discretization method formed to optimize the design of the Saccardo nozzle. The objective of
this study was to optimize the design of Saccardo nozzle for higher
Gradient Least Square Cell Based
Pressure Second Order ventilation effectiveness, reducing the tunnel temperature while main­
Velocity SIMPLE taining the total power consumption of the nozzle. The optimized and
Turbulent kinetic energy Second Order Upwind proposed nozzle designs were assessed by the mono-dimensional model
Turbulent Dissipation Rate Second Order Upwind to predict the carry-over effect and the maximum temperature in the
Energy Second Order Upwind
tunnel.

2. Simulation methodologies

2.1. Numerical model

The commercial simulation software, ANSYS Fluent (ANSYS, 2017a,

2017b) was employed in the present study. The Reynolds-averaged
Navier-Stokes (RANS) equations were solved using the finite volume
method. The RANS equations were derived to describe the mean air flow
quantities and turbulent Reynolds stress. The two-equation approach
was adopted in this model for the closure of Reynolds stresses terms. The
standard k-epsilon turbulent model (Launder and Spalding, 1972) was
chosen to resolve the turbulent kinetic energy and turbulent dissipation
rate. In addition, the energy conservation equation was enabled to
resolve the energy transfer in the tunnel. The governing equations were
being discretized and they were solved using the pressure-based segre­
Fig. 2. The geometry of Saccardo nozzle and exhaust damper. gated algorithm on a structured hexahedral grid. The discretization
methods for each of the governing equations are shown in Table 1. The
SIMPLE scheme (Patankar, 1980) was chosen for pressure–velocity
(2010) applied Reynolds-Averaged Navier-Stokes (RANS) models with
coupling calculation. The air flow motion in the tunnel was assumed to
the standard k-epsilon turbulence model to examine its turbulent char­
be in steady-state condition and the fluid was defined as incompressible
acteristics. They investigated the interaction of jet fans and the wall
flow. The governing equations solved in Fluent for present simulation
surfaces in a curved tunnel, which caused the asymmetrical jet flow
are written as below,
downstream. The static pressure increased on the convex wall while the
Continuity equation:
pressure reduced on the concave wall. Similarly, Betta et al. (2009)
employed the k-epsilon turbulence model to optimize the pitch angle of ∂ui
=0 (1)
jet fans in a longitudinal tunnel. Betta et al. (2010) further extended the ∂x
simulation model to evaluate the heat and air flow patterns when heavy
Momentum Equation:
good vehicles burning in the tunnel. Vega et al. (2008) used CFD com­
( ) [ ( )]
mercial software, ANSYS FLUENT (ANSYS, 2017a, 2017b), to reproduce ∂ui ∂ui uj ∂p ∂ ∂ui ∂uj 2
ρ + = + (μ + μt ) + − ρk (2)
the Memorial Tunnel fire test, and showed that the numerically pre­ ∂t ∂xj ∂xi ∂xj ∂xj ∂xi 3
dicted temperature contours were in good agreement with the experi­
Turbulent kinetic Energy Equation:
mental results. Costantino et al. (2014) considered the effect of surface
( ) [( ) ]
roughness on tunnel ceiling and floor in the simulation model. The au­ ∂k ∂kui ∂ μ ∂k
ρ + = μ+ t + Gk − ρε (3)
thors showed that the wall surface roughness did affect the tunnel ve­ ∂t ∂xi ∂xj σk ∂xj
locity development and distribution. Furthermore, Król et al. (2017)
examined the critical air velocity in the tunnel using the anemometer. Turbulent Dissipation Rate Equation:
The on-site measurement method showed the feasibility of identifying (
∂ε ∂εui
)
∂
[( ) ]
μ ∂ε ε ε2
the air flow distribution in the tunnel. The experimental results were ρ + = μ+ t + C1ε Gk − C2ε ρ (4)
∂t ∂xi ∂xj σε ∂xj k k
then used to validate their simulation results. Colella et al. (2009)
developed a two-scale modelling approach for tunnel simulation where Energy Equation:
the solution of the simulation was coupled with the mono-dimensional (
∂T ∂Tuj
)
∂
[(
CP μt
)( ) ]
∂T
model to predict the overall tunnel ventilation. The mono-dimensional ρC P + = λ+ (5)
∂t ∂xj ∂xj Prt ∂xj
model simplifies the 3D simulation model. The air flow in these sec­
tions was assumed to behave as a fully developed flow; the air flow k2
properties can be calculated based on the conservation of mass and μt = ρ C μ (6)
ε
momentum. This two-scale modelling approach could reduce compu­
tational time, while preserving an acceptable accuracy for the tunnel Where
ventilation performance. C1ε = 1.44, C2ε = 1.92, Cμ = 0.09, σk = 1.0, σε = 1.3, Prt = 0.85
The paper in this work modelled the Saccardo nozzle ventilation
scheme in a Singapore underground road tunnel which consists of 5 In the tunnel ventilation model studied, there are 5 ventilation
ventilation buildings. The overall tunnel length is about 7 km and the buildings spaced equally along the tunnel. Each of the ventilation
distance between each ventilation building is about 1.4 km. The design buildings consists of three Saccardo nozzles and exhaust dampers. In the
of Saccardo nozzle is employed in the present study as the baseline simulation model, these ventilation buildings are set up as repeated
design. Measurement of velocity and temperature in the tunnel has also designs so that the mono-dimensional model can be used. A ventilation
been conducted in the actual tunnel for validation purposes. The present building was modelled to be in the middle of the tunnel. The tunnel has a
paper used a CFD simulation model to evaluate the effectiveness of length of 1 km, a width of 14.2 m and a height of 6 m, with an equivalent

3
J.G. Hang et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 118 (2021) 104185

Fig. 3. The side (a) and top (b) views of Saccardo nozzle design geometry.

uniform wall roughness thickness of 0.005 m. For thermal boundary

Table 2
conditions, the nozzle and tunnel inlet have an identical ambient air
Boundary conditions of CFD model.
temperature of 29 ◦ C for parametric investigation. In addition, the heat
Boundary Surfaces Boundary Types Units Value flux value of 241 W/m2 released by the tunnel floor represented the
Nozzle Inlet Mass Flow Inlet kg/s 202 vehicle heat emission during congested traffic. For validation purposes,
Exhaust Outlet Mass Flow Outlet kg/s 202 only Nozzle 3 was turned on. The tunnel inlet temperature was set as
Tunnel Inlet Pressure Inlet Pa 0 32.7 ◦ C with a heat flux of 19.35 W/m2 emitting from the tunnel floor.
Tunnel Outlet Pressure Outlet Pa 0
Tunnel Surfaces Wall Heat Flux W/m2 19.35
Tunnel Floor Wall Heat Flux W/m2 19.35/ 241.5* 2.2. Grid independence test
Tunnel Surfaces Wall Roughness m 0.005

*It represents heat generation by congested traffic which has been used for Hexahedral mesh was created with five inflation layers on the wall
parametric study. boundaries to capture the near-wall region accurately. In the simulation
domain, the tunnel has a length of 1 km. The tunnel section itself has
hydraulic diameter of 8.44 m. The geometries of Saccardo nozzles and occupied a significantly larger amount of meshes. Mesh independence
exhaust dampers are shown in Figs. 2 and 3. The three separated nozzles study was conducted to design an adequate mesh resolution to accu­
were placed side by side, identified as Nozzle 1, Nozzle 2 and Nozzle 3 rately capture the flow phenomenon in the tunnel. The mesh indepen­
respectively. Each of these nozzles has an individual supply fan, which dence study was performed with mesh sizes of 0.4, 0.3, 0.2 and 0.15 m in
can provide 202 kg/s of fresh air into the tunnel. The exhaust dampers, the tunnel section. Fig. 4 shows the different mesh resolutions at the
highlighted as red surfaces, are connected to exhaust fans that extract tunnel cross-section surface 50 m downstream of the Saccardo nozzle,
the same amount of air flow as the supply fans. During tunnel operation, capturing the velocity magnitude. The result shows that sufficiently fine
these three nozzles operate individually following a pre-defined duty mesh resolution is necessary to resolve the air momentum transfer from
cycle. In some situations, e.g. high traffic on an extremely hot day, the nozzle. Bigger mesh size tends to overpredict the bulk flow velocity
congested traffic, or fire emergencies, two or three nozzles will be turned magnitude, such as the mesh size of 0.4 m where a velocity magnitude of
on concurrently to enhance the ventilation rate. The fans will be acti­ 18 m/s was predicted. By reducing the mesh size, the velocity magnitude
vated when the tunnel temperature exceeds the threshold value during of bulk flow was reduced to 16.9, 16.55 and 16.5 m/s with mesh sizes of
normal operation. For the mesh independence test and validation sec­ 0.3, 0.2 and 0.15 m correspondingly. The velocity magnitude with
tion, Nozzle 3 was operating alone which align with the working con­ respect to the total amount of mesh elements is shown in Fig. 5. The
ditions during the on-site measurement. In the parametric investigation, reduction of mesh size resulted in a drastic growth in mesh count, and
Nozzle 2 was assumed to be operating only. It would simplify the the velocity magnitude was also observed to gradually reduced to a
evaluation of the Saccardo nozzle performance by minimizing the effect stable value. In summary, the mesh independence study suggested that
of the sidewall. the mesh size of 0.2 m was adequate to resolve the air flow in the tunnel
The boundary conditions for the CFD model were defined as in with the changes in velocity magnitude lower than 0.3%.
Table 2 and the inlets of Nozzle 2 or 3 were defined as mass flow inlet
conditions. The corresponding mass flow outlet with the same amount of 2.3. Validation of CFD mode
air flow was set at the damper outlet. The pressure inlet and outlet have
been chosen for the tunnel inlet and outlet boundary conditions to An onsite measurement was conducted in the actual tunnel (the
simulate the carry-over effect along the tunnel. The air flow rate at the tunnel was closed with no traffic during the time of measurement). Both
tunnel outlet was monitored throughout the simulation to verify that the air velocity magnitude and temperature profiles were measured
there was no reverse flow. The tunnel walls and ground surface were set using Testo 440 hot wire probe. The measurement system has an accu­
as walls with no-slip boundary conditions and are assumed to have a racy of ± 0.1 m/s for velocity magnitude and ± 0.3 ◦ C for temperature
reading. The location of measurement is shown in Fig. 6. A total of 17

4
J.G. Hang et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 118 (2021) 104185

Fig. 4. Front view of tunnel with cross-section velocity contour and meshes. (a) Mesh size = 0.4 m, (b) Mesh size = 0.3 m, (c) Mesh size = 0.2 m, and (d) Mesh size =
0.15 m.

Fig. 5. The velocity magnitude of bulk flow with respect to the mesh.

points were measured along the longitudinal direction at a height of 1.6 turbulence of incoming air flow. The induced air travels along the tunnel
m above ground level. All the measurement points were taken at the and is later exhausted at the exhaust damper, thus the velocity magni­
centre of tunnel width and the average values of air velocity were tude dropped to around 2 m/s before it merged with the fresh air supply
recorded. The fluctuation of measurement is defined by the variation from the Saccardo nozzle. The velocity magnitude increased signifi­
between the maximum and minimum readings. cantly at the region where the nozzle airflow transferred the momentum
The velocity profiles of the onsite measurements and CFD simulation energy to the tunnel air flow. The velocity magnitude thus varies be­
results are shown in Fig. 7(a). The mean velocity at the tunnel inlet was tween 2 m/s and 8 m/s, with a distinguishable peak velocity captured.
about 5 m/s with 10% of fluctuation, possibly attributed to the As the flow moved towards the downstream of the nozzle, the air flow

5
J.G. Hang et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 118 (2021) 104185

Fig. 6. The locations of measurement along the tunnel from side view.

Fig. 7. The velocity and temperature profiles of onsite measurement and simulation results.

indicating that the simulation model was capable of capturing the

overall velocity profiles as observed in the experiment.
The temperature profiles of the onsite measurements and CFD
simulation are shown in Fig. 7(b). The inclined temperature gradient in

Table 3
The dimensions of Saccardo nozzle.
Parameters Units Baseline Range of study
(Validation)
(Lower (Upper
limit) limit)

P1 - Incline.Angle degree 10 8 15
P2 - Nozzleout. m 27.5 13.7 41.2
Length
P3 - Nozzlein. m 6 3 9
Length
P4 - Nozzle.Width m 2.8 2 3.6
P8 - Exhaust.Width m 3 2 4
P9 - Exhaust.Length m 3 2 4
Fig. 8. The static pressure (Ps,i) profile along the tunnel with baseline nozzle in P14 - Exhaust. m 16 10 20
Distance
the simulation.
P15 - Nozzle.Height m 2 1 3
P16 - Nozzlein. m 3 2 4
further developed to a much more stable air flow and its fluctuation Height
dropped to 0.5 m/s. The velocity profiles show that the simulation result P17 - Nozzle.Bend m 0.1 8 15
P23 - Nozzle.Flow kg/s 202 101 303
has a comparable flow behaviour with the measured readings,

Fig 9. Schematic diagram of heat carry-over effect in the tunnel section.

6
J.G. Hang et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 118 (2021) 104185

Table 4 temperature which was lower during the period of measurement. The
The range of dimension for parametric study. simulation model was validated by comparing the simulation result
Parameters Units Lower Limit Upper Limit against the measured velocity and temperature. The air velocity at 500
m downstream of the nozzle was compared with the experiment
Slot Angle degree 15 75
Slot Width m 0.2 0.4 measured value. The relative error is defined by Eq. (7) below,
⃒ ⃒
⃒Simulation(u) − Experiment(u)⃒
Error(%) = ⃒⃒ ⃒
⃒ (7)
the tunnel implies that there is a heat exchange rate in the tunnel cor­ Experiment(u)
responding to a heat flux of 19.35 W/m2 on the tunnel surfaces. The 4.7% and 1.7% relative errors were observed in the velocity and
simulated temperature downstream of the Saccardo nozzle was higher temperature at the tunnel outlet, respectively. The errors could be
than the experiment result, and the maximum observed deviation was caused by the experiment uncertainty or actual tunnel circumstances
0.7 ◦ C. The temperature deviation could be due to the fresh air

Fig. 10. The temperature profiles for 1, 2 and 3 nozzles operating conditions.

Fig. 11. The sensitivity analysis of each output parameters with respect to the changes on input parameters.

7
J.G. Hang et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 118 (2021) 104185

Fig. 12. The contours of (a) nozzle power, (b) tunnel induced flow and (c) nozzle efficiency with respect to Nozzle Height.

that were not considered in the simulation, such as the curvature of the air movement. Using the Eq. (8) below, the power is equal to the total
tunnel and obstacles in the tunnel. pressure multiplied by the air volume flow rate at each of the tunnels.
The total pressure at each stage is the summation of the static and dy­
3. Results and discussion namic pressures. The air volume flow rate at each stage can also be
extracted from the simulation result. To evaluate the power consump­
3.1. Characteristics of Saccardo nozzle in a longitudinal ventilation tion for the Saccardo nozzle system, the power at nozzle inlet and
system exhaust outlet is calculated using Eq. (8).
( )
1
Yoshizawa et al. (1991) suggested that the pressure increase in the Poweri = Pt,i × Qi = Ps,i + ρu2i × Qi (8)
2
tunnel is an important consideration in the design of a longitudinal
ventilation system. It is practically useful in determining the power Yoshizawa et al. (1991) proposed the ventilation efficiencies of
consumption of the Saccardo nozzles, and it is easily measurable nozzle and exhaust defined as Eqs. (9) and (10) below,
through an experimental approach. Similarly, the static pressure dis­
Pt,3 × Q3 − Pt,2 × Q2
tribution is attainable in the present simulation result. The static pres­ ηn = (9)
Pt,n × Qn
sure profile for the baseline design of Saccardo nozzle is shown in Fig. 8.
The important pressure values are defined as Ps,1, Ps,2 and Ps,3, with the
Pt,1 × Q1 − Pt,2 × Q2
corresponding locations shown in the tunnel schematic diagram in ηe = (10)
Pt,e × Qe
Fig. 6. Ps,1 is the minimum static pressure in the tunnel, and it represents
the pressure before the air flow is sucked out from the exhaust. Ps,2 is the By applying momentum conservation between the intersection area
static pressure between the nozzle and exhaust point. Ps,3, which is the of nozzle and tunnel, the momentum exchange coefficient can be
maximum pressure in the tunnel, denotes the maximum momentum at derived using Eq. (11),
the tunnel downstream received from the Saccardo nozzle. The pressure ( ) ( )
Ps,3 − Ps,2 At + ρ u23 − u22 At
profile can be divided into 3 sections, (1) the pressure increases between Kmx = (11)
Ps,2 and Ps,1 that is caused by the exhaust effect; (2) the pressure in­ ρun An cosθ
2

creases between Ps,3 and Ps,2 that is due to the nozzle effect; and (3) the
pressure decreases from the tunnel inlet (at 500 m) to Ps,1, and from Ps,3
3.2. The Mono-Dimensional temperature prediction model
to the tunnel outlet (at − 500 m) that is due to friction loss in the tunnel.
The static pressure at each stage can be translated into the power of
In order to assess the carry-over effect along the tunnel, the simu

8
J.G. Hang et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 118 (2021) 104185

Fig. 13. The contours of (a) nozzle power, (b) tunnel induced flow and (c) nozzle efficiency with respect to Nozzle Width.

lated result was used to evaluate the five repeated tunnel sections. The exchange with the tunnel surface (qheat ) and the tunnel outlet tempera­
schematic of the heat carry-over effect is shown in Fig. 9 below. Firstly, ture is Tout. The energy conservation can be written as Eq. (14) below,
the Saccardo nozzle momentum exchange was assumed to be constant at qheat
each repeated section and the tunnel-induced air flow was calculated = ρCP Q3 (Tout − T3 ) (14)
2
from the simulation result. Secondly, the heat was transferred from the
present tunnel section outlet to the next tunnel section inlet. The where the total energy equals to the heat flux in Table 3 multiply by the
temperate affected by the heat carry-over effect at each section was surface area of the tunnel.
calculated as follows. The tunnel-induced flow (Q1) with temperature Tin The outlet temperature at the current tunnel section was calculated
moves into the tunnel, while the heat was discharged from tunnel sur­ and applied to the inlet temperature for the next tunnel section. This
faces. The temperature of the tunnel induced air increases until it mono-dimensional model was used to compute for all repeated tunnel
reached the exhaust damper, where the temperature at stage 1 was sections. Subsequently, the average temperature in the tunnel was pre­
calculated by the energy conservation for the first half of the tunnel, dicted and the carry-over effect showed the oscillatory increment
written as Eq. (12) below, pattern. Fig. 10 shows the temperature profile during congested traffic
qheat conditions. During one nozzle operation, the maximum temperature
= ρCP Q1 (T1 − Tin ) (12) observed in the tunnel was 53 ◦ C, which has exceeded the temperature
2
threshold limit. The second nozzle would need to be operated to enhance
*Divided by 2 as half of the heat in the tunnel is considered in each the ventilation rate. With two nozzles operating, the mono-dimensional
section for Eqs. (12) and (14). model estimated that the maximum temperature would drop 11 ◦ C from
Then the heat and mass exchange happened at the exhaust damper 53 ◦ C to 42 ◦ C. Extending this further, with three nozzles operating
and Saccardo nozzle, where some of the tunnel induced flow (Qe) would concurrently, the maximum temperature predicted reduced to 38 ◦ C.
be discharged from the exhaust damper while the remaining air flow
(Q2) travelled towards the Saccardo nozzle. After that, the remaining air
flow (Q2) mixed with the fresh air from the Saccardo nozzle (Qn). The 3.3. Parametric study on the Saccardo nozzle design
temperature of the tunnel air after mixing (T3) was calculated by Eq.
(13) below, To characterize the key important parameters of the Saccardo nozzle
design, a parametric study has been performed, with the assumption
0 = ρCP Qn (T3 − Tn ) + ρCP Q2 (T3 − Te ) (13) that only Nozzle 2 is operating. The objective of the parametric study
After the mixing at stage 3, the air flow (Q3) undergoes heat was to reduce the average tunnel temperature while maintaining the
constant power consumption of the nozzle. Ten (10) geometrical

9
J.G. Hang et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 118 (2021) 104185

Fig. 14. The contours of (a) nozzle power, (b) tunnel induced flow and (c) nozzle efficiency with respect to Nozzle Bend Radius.

consumption was associated with the mass flow rate and total pressure
Table 5 of the nozzle. The parametric study was conducted using ANSYS soft­
The comparison of nozzles performance. ware with 264 design points simulated and generated by the Latin hy­
Nozzle Normalized Normalized Nozzle Nozzle percube sampling method. The output parameters are defined as the
Design Nozzle Flow Tunnel Induced Efficiency Kmx tunnel-induced flow, tunnel temperature, the power consumption of
Flow (%) the nozzle, the efficiency of the nozzle and momentum coefficient (Kmx)
Baseline 1.0 1.00 16.5 0.976 of the nozzle.
Nozzle The sensitivity analysis is shown in Fig. 11 and the magnitude of
Optimized 1.5 1.21 26 0.927
sensitivity ranged from 0 to ±100%. It represented the influence of each
Nozzle
Air Bleeds 1.2 1.09 22 0.86 input parameter on the respective output parameters. The first factor
Design occurred at the sensitivity limit (+100% or − 100%) and it represented
the most important factor among others. The second and third factors
were based on their magnitude. The sensitivity value closer to zero
dimensions representing the structure design of the Saccardo nozzles meant that the input parameter has an insignificant effect on the output
and exhaust dampers were defined as shown in Fig. 3. The dimensions of parameters. The sensitivity analysis assisted in identifying the important
the baseline design and the ranges of the investigations are summarized parameters of the nozzle. The result shows that the tunnel-induced flow
in Table 3, with the upper and lower ranges defined in consultation with was highly affected by these three most important parameters: nozzle
the tunnel ventilation design specialists of the KPE road tunnel. These width, nozzle height and nozzle flow. The positive sensitivity value
parameters selected were also shown to directly affect the tunnel represented a positive correlation between the changes of input and
ventilation performance as shown in the sensitivity analysis that will be output parameters, for example, a higher nozzle flow would lead to
presented later. The geometrical dimensions from P1 to P17 were set to larger tunnel-induced flow. Conversely, the negative sensitivity value
evaluate their effect on the ventilation performance. The mass flow rate indicated the negative correlation between the input and output pa­
of the nozzle (P23) was considered as an independent variable that is rameters, for example, a larger nozzle width and height would result in
related to the nozzle ventilation performance under the constraint of lower tunnel-induced flow. Subsequently, the sensitivity of tunnel
constant power consumption. With the changes in nozzle dimensions, temperature shows an inverse correlation to the tunnel induced flow, i.
the power consumption of the nozzle would vary significantly. The in­ e., the increment in tunnel-induced flow leads to lower tunnel temper­
crease or reduction in nozzle mass flow rate was required to compensate ature. The analysis shows that there are two direct ways to increase the
for the power consumption loss or rise because the nozzle power

10
J.G. Hang et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 118 (2021) 104185

Fig. 15. The geometry of Saccardo Nozzle with air bleeds design.

Fig. 16. The temperature contours on symmetry plane of the tunnel with various slot designs.

tunnel-induced flow, which are (1) by increasing the nozzle flow; or (2) nozzle flow, height, width and bend appeared as the top four important
by reducing the nozzle width or height. By comparing these two ways, factors that would impact the nozzle efficiency. All of them have a
the tunnel temperature has the strongest influence on the nozzle flow, as positive correlation in improving the nozzle efficiency.
it blows additional fresh air to the tunnel. On another hand, the From the sensitivity study, the nozzle height, width and nozzle bend
reduction of nozzle width or height induces more air flow from the radius were identified as the key parameters that would affect the nozzle
tunnel entrance. Simultaneously, the nozzle power consumption in­ power, tunnel-induced flow and nozzle efficiency. Three of these pa­
creases with the suggested changes above. From the sensitivity analysis, rameters have been selected for design optimization at a fixed power
an interesting finding suggested that the nozzle flow rate was the first consumption. Fig. 12(a) shows the contours of normalized nozzle power
factor for tunnel temperature and the second factor for the nozzle power. under various nozzle flow rate on Y-axis and nozzle heights on X-axis.
Conversely, the nozzle height was the first factor for nozzle power and The nozzle power was normalized by the power of the baseline nozzle
the second factor for tunnel temperature. In order to achieve a reduction and was limited to one. As a result, the constant power curve can be
in tunnel temperature while maintaining the power consumption, an extracted and is shown in Fig. 12(a). Under the constant power condi­
increase in nozzle flow and nozzle height are recommended. Besides, the tion, the nozzle flow rate can be determined from the nozzle height.
nozzle bend was identified as the fourth factor that would significantly With this relationship, the constant power curve is superimposed onto
reduce the nozzle power by increasing its bend radius. The nozzle bend Fig. 12(b) to evaluate its impact on the tunnel-induced flow. The
has an insignificant impact on the tunnel temperature. Concurrently, the intersection of the constant power curve and the contours of tunnel-

11
J.G. Hang et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 118 (2021) 104185

Fig. 17. The contours of (a) nozzle power, (b) tunnel induced flow and (c) nozzle efficiency with respect to slot angle.

induced flow shows the increase of tunnel-induced flow by increasing efficiency would improve with a bigger bend radius. With the constraint
the nozzle height. Fig. 12(c) shows the contours of nozzle efficiency with in nozzle power consumption, the optimization study recommends
respect to nozzle height and nozzle flow. The constant power curve revising the nozzle height and width to larger sizes while fixing the
demonstrates the improvement in nozzle efficiency by increasing in nozzle bend radius at 0.735 m in order to enhance the tunnel-induced
nozzle height. flow and tunnel temperature. The optimized nozzle design is summa­
The contours with respect to the nozzle width are shown in Fig. 13. rized in Table 5 and it has a nozzle width of 3.2 m and nozzle height of
The constant power curve behaves more linearly with respect to the 2.75 m. With the same amount of nozzle power consumption, the nozzle
changes in nozzle width under various nozzle flow conditions, as shown flow operates at 1.5 times the baseline nozzle flow rate. The tunnel-
in Fig. 13(a). At constant nozzle power, Fig. 13(b) shows that the con­ induced flow increases by 21% from the baseline design, and the
tours of tunnel-induced flow intersecting with the constant power curve. nozzle efficiency improves from 17% to 26%. At the same time, the
The tunnel-induced flow would increase by increasing the nozzle width. nozzle momentum exchange coefficient is observed to drop from 0.976
A similar observation on the nozzle efficiency is shown in Fig. 13(c), to 0.927.
both the width and height of the nozzle react similarly to the nozzle An alternative nozzle design was proposed based on the presented
efficiency as the dimension of height and width are a function of the results. The air bleeds design consisted of three slot openings underneath
nozzle cross-section area. Changes in either nozzle width or height the baseline nozzle design. The design geometry is shown in Fig. 15 with
would increase the cross-section area of the nozzle. The nozzle height each slot opening having a fixed length of 2 m and symmetrically
plays a more important influencing factor as compared to the nozzle positioned with the nozzle. The dimensions of angle and width of the
width as the nozzle flow inlet direction is perpendicular to the nozzle slots were determined by using another parametric study, and the range
height. The nozzle flow would have to undergo a bend in order to change of dimensions for the optimization study is shown in Table 4. The air
the flow direction. bleeds were positioned underneath the nozzle inlet, as the fresh air from
The bend radius has a significant impact on the nozzle power. The the nozzle inlet moves in the vertical direction and undergoes a 90-de­
contours of nozzle power with respect to nozzle bend radius are shown gree bend. The air bleeds were meant for reducing the nozzle power
in Fig. 14(a), with the constant power curve behaving parabolically and consumption before the bend by redirecting a small amount of air
it has an optimum value at the radius of 0.735 m. The intersection of through the slot openings. With an angled slot design, the momentum
constant power curve and tunnel induced flow in Fig. 14(b) shows that from air bleeds would improve the tunnel-induced flow. The air bleeds
the highest tunnel induced flow could be obtained with the optimum size would reduce the power consumption of the nozzle. The tempera­
bend radius. Nozzle bend radius exceeding 0.735 m would have insig­ ture contours of the tunnel symmetry plane for different slot configu­
nificant influences on tunnel-induced flow. Similarly, the nozzle rations are shown in Fig. 16. First, the implementation of air bleeds

12
J.G. Hang et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 118 (2021) 104185

Fig. 18. The contours of (a) nozzle power, (b) tunnel induced flow and (c) nozzle efficiency with respect to slot width.

Fig. 19. The comparison of nozzle efficiency between original nozzle, optimized nozzle and air bleeds design.

(Fig. 16(a) and (b)) shows a cold air jet stream releasing from the bleeds nozzle, thus producing higher nozzle flow under the constant power
with the bleeds dividing air flow into two mainstreams, which would conditions. The distribution of air flow rates passing through the bleed
cause the momentum loss on the Saccardo nozzle mainstream air flow. slot openings and the Saccardo nozzle was determined by the slot
At the same time, it compensated for the power consumption of the opening width (Fig. 16(b) and (d)). Excessive slot openings would

13
J.G. Hang et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 118 (2021) 104185

Fig. 20. The predicted temperature profile along the tunnel with different nozzle design and air bleeds.

reduce the effectiveness of the Saccardo nozzle. Secondly, the slot angle baseline KPE Saccardo nozzle design shows comparable nozzle effi­
will affect the airflow direction exiting the air bleeds immediately ciency with the reference nozzle design from Yoshizawa et al. (1991).
(Fig. 16(b) and (c)). A higher angle design tends to blow air toward the However, the optimized nozzle design shows an improvement in nozzle
road which will result in higher resistance to the tunnel flow. Further­ efficiency, i.e. 26% as compared to the 17% efficiency of the baseline
more, the air bleeds design shows the feasibility of reducing the tem­ nozzle. The air bleeds nozzle design has a 5% improvement in nozzle
perature underneath the nozzle area. efficiency of 22%. In addition, the optimized nozzle is able to provide
The results of the parametric study were analysed under the same 50% more nozzle flow rate; hence, the nozzle enhances the tunnel-
constant power consumption of the nozzle. Fig. 17(a) shows the con­ induced air flow by 21% more than the baseline nozzle. The air bleeds
tours of nozzle power against the slot angle and nozzle flow. The slot design could supply about 20% more nozzle flow rate which results in a
angle has the least influence on the power consumption as the constant 9% increase in tunnel flow.
power curve shows a nearly horizontal line along with the changes of Apart from the above, the momentum exchange coefficient (Kmx) has
slot angle. However, the slot angle plays an important part in inducing decreased for both the optimized nozzle and air bleeds design from
air flow movement in the tunnel. The constant power curve superposing 0.976 to 0.927 and 0.86 respectively. Despite several tunnel designs are
with the contour of tunnel-induced flow is shown in Fig. 17(b). From the targeting on Kmx coefficient as the design specification while neglecting
figure, it can be seen that the tunnel-induced flow can be improved by the nozzle efficiency, there is a strong correlation between the defined
19% by varying the slot angle from 75 to 15◦ . The lower slot angle in­ Kmx coefficient and nozzle efficiency. Both Eqs. (9) and (11) have an
creases the transfer of momentum to the air flow underneath the nozzle, identical definition on the numerator, which is the total energy transfer
pre-accelerates the tunnel airflow before it enters the Saccardo nozzle in the tunnel. The denominator in the Kmx definition is the dynamic
section. A slot angle of 15 degrees can improve the nozzle efficiency to pressure of the nozzle. The quantity of dynamic pressure can be achieved
23%. The contour of nozzle efficiency in Fig. 17(c) shows that the higher by varying the total nozzle flow rate under various sizes of the nozzle,
efficiency area at low angle and low nozzle flow regions. Conversely, the but the static or total pressure would differ significantly according to the
slot width has a significant effect on the nozzle power consumption. The different sizes of the nozzle. The total pressure directly influences the
constant power curve in Fig. 18(a) shows that the power gradually de­ power consumption by the nozzle. In the present result, the slight
creases by reducing the nozzle width. The intersection of the constant reduction in the Kmx coefficient is caused by the lower nozzle velocity,
power curve and tunnel induced flow in Fig. 18(b) shows that the bigger due to the change in nozzle size and implementation of air bleeds. The
slot width would result in a larger portion of air escaping from the supplement of nozzle flow rates could not reproduce the same air ve­
bleeds, therefore reducing the tunnel induced flow and lowering nozzle locity at the nozzle. Moreover, the lower the air velocity magnitude from
efficiency. The contours of nozzle efficiency shown in Fig. 18(c) indicate the Saccardo nozzle, the safer it would be for the vehicles in the tunnel.
that the highest nozzle efficiency that occurs on the constant power More importantly, the tunnel temperature reduces while maintaining
curve is 23% with at a 0.2 m slot width dimension. In summary, the air the same level of power consumption by the nozzle. The air bleeds
bleeds design with a 15◦ slot angle and a 0.2 m slot width is adopted to design did not have a significant improvement as compared to the
the baseline Saccardo nozzle. optimized nozzle; however, it has the advantage of easy implementa­
The performances of the baseline nozzle, optimized nozzle and air tion. It does not require major modification to the existing design.
bleeds designs are summarized in Table 5. Both the recommended With the proposed nozzle designs presented above, the tunnel tem­
optimized nozzle and air bleeds designs consume the same amount of perature profiles were evaluated using the mono-dimensional temper­
nozzle power as compared to the baseline nozzle design. The efficiencies ature prediction model under the traffic congestion condition. The
of the original nozzle design, optimized nozzle design and air bleed predicted carry-over effect in Fig. 20 shows that the highest temperature
design were further plotted and compared with the reference nozzle of 53 ◦ C occurs for the baseline nozzle design at around 6.3 km from the
design from Yoshizawa et al. (1991) as shown in Fig. 19 below. The tunnel entrance. Meanwhile, the optimized nozzle shows that the

14
J.G. Hang et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 118 (2021) 104185

maximum temperature in the tunnel could reduce by up to 6.6 ◦ C. On the interests or personal relationships that could have appeared to influence
other hand, the air bleeds design shows a promising result by reducing the work reported in this paper.
the highest temperature and overall tunnel temperature by 3.2 ◦ C and
1.8 ◦ C respectively. For the normal operating conditions, the maximum Acknowledgement
temperature would trigger second nozzles to be operated under these
congested traffic conditions. The predicted temperature in Fig. 19 shows The authors would like to thank the Land Transport Authority (LTA)
that the optimized nozzle and air bleeds design could reduce the of Singapore for its financial support for this investigation under the
maximum temperature to below 40 ◦ C. Land Transport Innovation Fund (LTIF).

4. Conclusions References

In the present study, a CFD simulation model combined with a mono- ANSYS Fluent User’s Guide, Release 18.2, ANSYS Inc, August 2017.
dimensional temperature prediction model was developed to mimic the ANSYS Fluent Theory’s Guide, Release 18.2, ANSYS Inc, August 2017.
Betta, V., Cascetta, F., Musto, M., Rotondo, G., 2009. Numerical study of the optimization
carry-over effects in a road tunnel equipped with the Saccardo nozzle of the pitch angle of an alternative jet fan in a longitudinal tunnel ventilation system.
longitudinal ventilation system. The impacts of ten selected Saccardo Tunnel. Undergr. Space Technol. 24 (2), 164–172.
nozzle design parameters were evaluated in terms of the tunnel venti­ Betta, V., Cascetta, F., Musto, M., Rotondo, G., 2010. Fluid dynamics performances of
traditional and alternative jet fans in tunnel longitudinal ventilation systems.
lation power consumption, nozzle efficiency, and the momentum ex­ Tunnel. Undergr. Space Technol. 25, 415–422.
change coefficient. The developed CFD model was validated against Colella, F., Rein, G., Borchiellini, R., Carvel, R., Torero, J.L., Verda, V., 2009. Calculation
onsite experimental results in terms of the tunnel velocity and temper­ and design of tunnel ventilation systems using a two-scale modelling approach.
Build. Environ. 44, 2357–2367.
ature profiles. Subsequently, two alternative Saccardo nozzle designs Costantino, A., Musto, M., Rotondo, G., Zullo, A., 2014. Numerical Analysis for reduced-
namely: (1) optimized Saccardo nozzle design with a larger nozzle scale road tunnel model equipped with axial jet fan ventilation system. In: 68th
height, nozzle width and a nozzle bend radius of 0.735 m, and (2) air Conference of the Italian Thermal Machines Engineering Association, ATI2013,
Energy Procedia, vol. 45, pp. 1146–1154.
bleeds design by introducing air bleeds underneath the Saccardo nozzle
Costeris, N.P., Sweetland, I.A., 1994. Longitudinal Ventilation by Injection. In: 8th
were proposed and studied. International Symp. on Aerodynamics and Ventilation of Vehicle Tunnels, BHRG,
The CFD validation study shows an estimated average discrepancy of Liverpool, UK, pp. 696–715.
10% against the onsite experimental study. From the CFD sensitivity Król, M., Król, A., Koper, P., Wrona, P., 2017. Full scale measurements of the operation of
fire ventilation in a road tunnel. Tunnel. Undergr. Space Technol. 70, 204–213.
analysis, it was observed that the nozzle height, nozzle width, and nozzle Kwa, G.S., 2004. The Design of Tunnel Ventilation System for a Long Vehicular Tunnel.
bend radius were the key design parameters that would significantly In: Tunnelling and Underground Space Technology, Underground Space for
affect the nozzle power, tunnel-induced flow, and nozzle efficiency. Sustainable Urban Development, Proceedings of the 30th Ita-Aites World Tunnel
Congress Singapore, vol. 19, pp. 312–319.
Subsequently, these three parameters were selected for the design Launder, B.E., Spalding, D.B., 1972. Lectures in Mathematical Models of Turbulence.
optimization study under constant power consumption conditions. The Academic Press, London, England.
CFD simulation results show the optimized nozzle design needs to Patankar, S.V., 1980. Numerical Heat Transfer and Fluid Flow. Hemisphere, Washington,
DC.
operate at a higher nozzle flow rate than the baseline condition to Tabarra, M., Matthews, R.D., Kenrick, B.J., 2000. The Revival of Saccardo Ejectors –
achieve the constant nozzle power consumption condition, consequently History, Fundamentals, and Applications. In: 10th International Symposium of
showing that the tunnel-induced flow increases by 21% and the nozzle Aerodynamics and Ventilation of Tunnels, BHRG, Boston, USA, 1-3 November 2000.
Tarada, F., Brandt, R., 2009. Impulse ventilation for tunnels – a state of the art review. In:
efficiency improves from 17% to 26% from the baseline design. In 13th International Symposium on Aerodynamics and Ventilation of Vehicle Tunnels,
addition, the nozzle momentum exchange coefficient is observed to drop New Brunswick, New Jersey, USA.
from 0.976 to 0.927. Finally, the proposed air bleeds design could Vega, M.G., Argüelles Díaz, K.M., Fernández Oro, J.M., Tajadura, R.B., Santolaria
Morros, C., 2008. Numerical 3D simulation of a longitudinal ventilation system:
further reduce the tunnel air temperature by 6% with a 5% improvement
memorial tunnel case. Tunnel. Undergr. Space Technol. 23 (5), 539–551.
in the nozzle efficiency. Wang, F., Wang, M., He, S., Zhang, J., Deng, Y., 2010. Computational study of effects of
jet fans on the ventilation of a highway curved tunnel. Tunnel. Undergr. Space
Declaration of Competing Interest Technol. 25 (4), 382–390.
Yoshizawa, I., Komatsu, K., Ohta, Y., Hagiwara, K., 1991. Characteristics of blowing and
exhaust nozzles used in longitudinal ventilation system. Aerodynamics and
The authors declare that they have no known competing financial Ventilation of Vehicle Tunnels, pp. 689–705.

15