International Communications in Heat and Mass Transfer 136 (2022) 106185

Contents lists available at ScienceDirect

International Communications in Heat and Mass Transfer

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

Experimental and numerical approach for characterization and

performance evaluation of cryogenic turboexpander under
rotating condition
Manoj Kumar a, *, Rasmikanti Biswal b, Suraj Kumar Behera b, Ranjit Kumar Sahoo b, d,
Sandip Pal a, c
a
Cryogenic Plant and Instrumentation Section, Variable Energy Cyclotron Centre, Department of Atomic Energy, Kolkata 700064, India
b
Department of Mechanical Engineering, National Institute of Technology Rourkela, 769008, India
c
Homi Bhabha National Institute, Kolkata 700064, India
d
National Institute of Technology Meghalaya, 793003, India

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

Keywords: A small-scale radial expansion turbine is distinguished by its ease of production, higher efficiency, and reliability.
Nitrogen turboexpander Such turbines have been successfully used in cryogenic turboexpander systems for refrigeration and liquefaction
Sobol sensitivity of the process gas. This paper presents a methodology for design optimization, numerical, and experimental
Experimental and numerical analysis
investigation of a radial turbine and nozzle (turboexpander), the most necessary and extortionate component of a
Claude cycle
nitrogen liquefaction system based on Claude cycle. The initial investigation starts with the preliminary design of
a turboexpander using an in-house developed Matlab® code. Then, a Sobol method-based optimization approach
has been proposed to determine the normalized sensitivity index and optimal range of major non-dimensional
and geometrical variables for the better off-design performance of the turboexpander. After that, different los­
ses of the turbine have been determined using an optimum set of loss correlations which is incorporated into the
preliminary design process. This approach can overcome the optimization issues caused by the high sensitive
design parameters, which may not be addressed through conventional methods, and ameliorates the off-design
performance of the turboexpander (improves power output, total-to-static efficiency, and diminishing the turbine
losses by 14.28%, 3.89%, and 9.61% respectively) as compared to the initial design. Based on this, three tur­
boexpander systems are designed and a comparative numerical study has been conducted to study the flow field
phenomenon and their thermodynamic performance at three operating pressure and cryogenic temperature (16
bar & 150 K, 8 bar & 120 K, and 4.5 bar & 95 K) using ANSYS CFX®. Finally, the Claude cycle-based experi­
mental facility has been established to determine the thermal performance of the turboexpander at various
operating pressure (16 and 8 bar), rotational speeds (120,612, 102,419, and 80,914 rpm), inlet temperatures
(150 and 120 K), and mass flow rates (0.01–0.10 kg/s). The results illustrate that the predicted performance from
the numerical simulation shows good agreement with the experimental results. Additionally, error analysis of
experimental parameters has also been discussed.

1. Introduction nozzle (turboexpander) are the key components for such types of
refrigeration and liquefaction cycles which can be used as an expansion
To accomplish the emerging demand for cryogenic fluids, it is widely device in many applications, such as gas turbines, organic Rankine cy­
accepted that the cryogenic turboexpander-based gas liquefaction cycles cles, air separation, and liquefaction systems [1,2,41]. Generally, these
would be the mainstream for various industrial and research appliances. turboexpanders have been designed to operate in a gaseous state for
It has been a promising alternative to traditional cryogenic systems to durable and safe operation. The performance of the gas liquefaction
generate ultra-low temperature and gas liquefaction. Radial turbine and system depends on the efficient design of such turboexpander.

* Corresponding author.
E-mail addresses: [email protected], [email protected] (M. Kumar), [email protected] (S.K. Behera), [email protected] (R.K. Sahoo), sandip@
vecc.gov.in (S. Pal).

https://doi.org/10.1016/j.icheatmasstransfer.2022.106185

Available online 24 June 2022

0735-1933/© 2022 Elsevier Ltd. All rights reserved.
M. Kumar et al. International Communications in Heat and Mass Transfer 136 (2022) 106185

Consequently, improving the efficiency and performance of such tur­ Table 1

boexpander still has a prime target in the research and development and Details of turboexpander used in cryogenic applications.
is a prerequisite for the existing state-of-the-art. Authors Field Pressure Fluid
Ghosh et al. [3] suggested that the performance of small and micro- ratio
scale turboexpanders would be increased by considering loss correlation Kumar et al. [9] Cryogenics (Experimental) 2.5–4.0 Air
in the preliminary design process. The procedure provided the opti­ Ghosh et al. [25] Cryogenics (Experimental) 1.8–6.0 Air
mized turboexpander profile having better off-design aerodynamic and Balaji [26] Cryogenics (Experimental) 2.0–5.8 Air
thermophysical performance. Aungier [4] demonstrated the preliminary Kumar et al. [8] Cryogenics (Experimental and 4.0 Nitrogen
Numerical)
design procedure to increase the efficiency of a radial turbine. Sauret Ino et al. [27] Cryogenics (Experimental) 7.0–12.0 Helium
et al. [5,6] suggested that the reduction in efficiency of radial turbo­ Kumar et al. Cryogenics (Numerical) 1.8–3.6 Helium
machines at off-design conditions may occur due to aerodynamic pro­ [2,10,11]
file, operating conditions, and other relevant losses. Therefore, it has Li et al. [28] Cryogenics (Numerical) 5.53 Helium
been necessary to design a turbine by considering its operating cycles to
incorporate the reasons (losses) for decrement in efficiency. Addition­
loss correlations in the design equations is a challenging task. In this
ally, the physical characteristics such as Coriolis force and secondary
regard, Ssebabi et al. [16] proposed a design methodology (using air as a
flows have been very complex to incorporate in the preliminary design
working fluid) considering various loss correlations. The study was
equations which may determine the reliability of the model. Hence,
extended to perform an experimental investigation to analyze the
their effect must be introduced in the design procedure in the form of
thermal performance of an organic Rankine cycle (ORC) using R-123 as
constants. Baines [7] summarized these losses for different operating
a process fluid. Jensen and Quale [17] studied the thermal performance
conditions using meanline loss modeling approach to obtain an opti­
and fluid flow analysis of a high-pressure ratio turboexpander
mized turboexpander geometry. Kumar et al. [8,42–44] followed the
(approximately 4) in the subsonic flow scheme. The analysis shows
aforementioned approaches and proposed a design methodology by
eminent consistency with the experiments. Dadone and Pandolfi [18]
considering well-established loss correlations for nitrogen turboex­
carried out the off-design performance of large and medium-scale tur­
pander in the design procedure and optimizing the design variables
boexpanders. The computational outcomes were validated with low-
using artificial intelligence techniques. The design methodology in­
pressure ratio and low rotational speed turboexpander using ideal gas
creases the efficiency (up to 4%) and refrigerating capacity (up to 18%)
approach. But the thermophysical properties of the fluids at cryogenic
of the turboexpander. The study further introduces the experimental and
temperatures behave differently. Therefore, the real gas approach has
numerical approach to investigate the flow characteristics of a tur­
been required for accurate numerical results.
boexpander. However, the operating pressure can be increased to
It is widely acknowledged, the three-dimensional unsteady state
improve the performance and yield of the cycle.
numerical simulation has been quite important to characterize the most
The turboexpander works at transient state conditions, several flow
reliable thermodynamic characteristics, flow physics, and other perfor­
physics appears from the wave dynamics. The 0-D models cannot predict
mance parameters along the streamwise and spanwise direction of tur­
these flow physics as well as off-design performance explicitly because
boexpander [19,20]. In this regard, Niu et al. [21] studied the inside
of the inadequacy of the spatial parameters. This problem can be over­
fluid flow characteristics of a wet type cryogenic turboexpander using
come using appropriate one-dimensional design procedures and opti­
CFX®. The investigation was further extended to characterize the
mization of non-dimensional and geometrical design parameters.
condensation flow (non-equilibrium) of an air turboexpander using
Several studies have been reported about the 0-D, 1-D, and 3-D perfor­
mean streamline off-design approach [22]. The study shows that the
mance of turboexpanders. For instance, Kumar et al. [9] predicted the
numerical analysis has been quite beneficial because of the lower cost,
optimal range of several design variables of nitrogen turboexpander
high repeatability, and understanding of the internal fluid flow physics
(operating pressure is 8 bar) using artificial intelligence techniques. The
inside the turboexpander which are very difficult to analyze from the
work was further extended to 1-D design and 3-D numerical investiga­
experimental investigation.
tion to analyze the flow physics and thermodynamic performance of a
Shao et al. [23] experimentally examined the thermodynamic per­
modified Collins cycle-based helium turboexpander operating at three
formance of a turboexpander (radial type) using R-123 fluid. Recently,
different pressure and cryogenic temperature [10,11]. Mounier et al.
Alshammari et al. [24] conducted an experimental analysis to study the
[12] suggested that the best design procedure of a turboexpander con­
repercussions of rotational speed on the efficiency of a radial turboex­
sists of four stages starting from 0-D (provide baseline design) to 1-D
pander. Several experimental and numerical studies were conducted for
(mean line or preliminary design) analysis (provide geometrical de­
thermodynamic performance evaluation and flow field behavior of
tails). The third stage stands for flow calculation through which blade
cryogenic turboexpanders which are presented in Table 1. However,
profile has been optimized and finally three-dimensional numerical
these studies do not present a detailed experimental and numerical study
analysis which is the most accurate and computationally expensive
about the thermodynamic and fluid flow characteristics of a cryogenic
procedure to attain the optimized turboexpander geometry. The study
turboexpander at various operating pressure, mass flow rate, and rota­
also proposed a new performance map of the small-scale turboexpander.
tional speeds. Hence, there is still a necessity for extensive numerical
The one-dimensional (1-D) analysis was mostly used to obtain the inlet
and experimental study for a detailed internal fluid flow physics and
and outlet velocity triangle and efficiency had been calculated from
thermodynamic characteristics of a cryogenic turboexpander.
enthalpy loss correlations. The process consumed less time and was
To the author’s best knowledge, some literatures are available which
advantageous to determine the off-design performance of the turboex­
consider the loss correlations in the one-dimensional design procedure
pander. Several works were reported to design a radial type turboex­
for Brayton and organic Rankine cycles. They also consider real gas
pander based on this approach [13–15]. However, the performance of
properties of the working fluid [40]. However, such techniques are very
turboexpander could be increased by decreasing the various losses such
few at cryogenic turboexpanders. Also, sensitivity analysis of important
as tip clearance, incidence, passage, blade loading, vortex, secondary,
design parameters, a comparative numerical analysis to investigate the
exit, and other losses. Generally, these losses had been considered for
insights of thermal and flow characteristics of the turboexpander at
large-scale turboexpanders. However, it may also be considered for
different operating conditions and experimental analyses to measure
small-scale cryogenic turboexpanders for enhancing their thermody­
such performance parameters at different rotational speed, mass flow
namic and aerodynamic performances.
rate, and operating pressure need to be investigated. Present work is an
For accurate design, the 1-D design of turboexpander operating at
attempt to fulfill the existing state-of-the-art. The originality of paper
cryogenic temperature needs real gas properties and the integration of

2
M. Kumar et al. International Communications in Heat and Mass Transfer 136 (2022) 106185

techniques, and numerical methodology. Then the results of the inves­

tigation are introduced and explained to provide an insight into the
performance of cryogenic turboexpander using an experimental-
numerical approach. It is worthwhile to mention that the present
investigation accentuates the design and sensitivity analysis of radial
turboexpander, development of a Claude cycle-based experimental test-
rig, the numerical investigation, and a methodology to produce the
liquid nitrogen at laboratory scale.
The paper is organized as: the design methodology, loss analysis, and
Sobol method are presented in section 2. Next, the experimental tech­
niques and error analysis are described in section 3. The numerical
methodology, boundary conditions, and grid independence test are
presented in section 4. The experimental and numerical results are
presented in section 5 and concluding remarks are mentioned in section
6.

2. Preliminary design procedure

The turboexpander is the most important component of the gas

liquefaction cycle. The overall performance of the liquefaction cycle, in
general, and off-design performance of the turboexpander in particular,
depend on the optimum configuration of the turboexpander blade. In
addition, the inlet and outlet boundary conditions are the deciding
factors for preliminary design procedure. The design process has been
disseminated in Matlab® domain. The energy equation solver (EES)
solver is integrated with the design code to obtain the real fluid prop­
erties of nitrogen at different pressure and ultra-low temperature.
The tangential velocity component (Cθ3) at the leading edge of the
turbine is computed as Eq. (1).
Cθ3 r3 = Cθ4 r4 (1)
After that, the meridional velocity component is computed with the
Fig. 1. (a) Schematic of turboexpander geometry (b) Velocity triangles. initial guess (Cm3 = Cm4). The remaining velocity components are
computed based on the velocity triangle (Fig. 1). The mass flow rate is
sprawls in the optimal turboexpander configuration, loss quantification, calculated using continuity equation as: (Eq. (2)).
and a comparable numerical and experimental investigation that lucidly
ṁ = 2πr4 ρ4 b4 Cm4 (2)
contemplates: proposing a novel preliminary design procedure inte­
grated with Sobol method and real gas approach; loss quantification of Initially, the exit velocity of the turbine has been assumed to be zero.
high, medium, and low-pressure turboexpander; sensitivity study of However, at the off-design conditions, the Euler equation (Eq. (3)) has
eight design and non-dimensional parameters; and numerical and been used to compute the exit velocity:
experimental techniques for thermodynamic and fluid flow performance
Δhact = U3 Cθ3 − U4 Cθ4 (3)
measurement at different cryogenic conditions.
In this paper, the authors investigated the thermodynamic perfor­ The isentropic and actual enthalpy, and power output are computed
mance and fluid flow physics of a Claude cycle-based nitrogen tur­ as Eqs. (4), (5), and (6).
boexpander operating at three different operating conditions. Initially,
Δhis = h03 − h4s (4)
the preliminary design procedure incorporated with different loss cor­
relations has been carried out using nitrogen (real gas) as a working Δhact = ηt− s × Δhis = h03 − h04 (5)
fluid. After that, the Sobol method has been used to determine the
optimal range and normalized sensitivity indices of non-dimensional P = m × Δhact (6)
and geometrical design parameters for higher efficiency and refriger­
ating capacity, and lower loss. The perspective ameliorates power The expansion of the fluid through the turboexpander has been
output, total-to-static efficiency, and diminishing the losses by 14.28%, assumed to be isentropic. Based on which Δhis and Δhact are computed.
3.89%, and 9.61% respectively. Based on the aforementioned methods, After that, the loss correlations have been incorporated to compute the
three turboexpander model have been developed. Furthermore, the various losses of the turbine.
three-dimensional numerical simulations have been conducted to
envisage the flow field and thermodynamic aspects of the designed 2.1. Loss models
turboexpander at three different pressure, temperature, and rotational
speed. Finally, the Claude cycle-based experimental facility has been The turbine internal losses such as incidence loss (Δhincidence, Whit­
augmented to distinguish the thermal performance of turboexpander at field [29]), passage and tip clearance loss (Δhpassage and Δhtip, Kumar
various operating pressure (16 and 8 bar), inlet temperature (150 and et al. [30]), exit energy loss (Δhexit, Suhrmann et al. [31]), trailing edge
120 K), and rotational speed (120,612, 102,419, and 80,914 rpm). The loss (ΔhTEL, Ghosh et al. [3]), and other losses must be considered.
validation of the numerical analysis with the experimental results has The incidence loss obtains because of the blade loading phenomenon
also been introduced. A detailed description of design methodology is which instigates the fluid flow inside the turbine passage at an angle
explained first, which includes turboexpander’s preliminary design different from the optimal angle. Such loss is computed as (Eq. (8)):
procedure, Sobol sensitivity analysis, loss analysis, experimental

3
M. Kumar et al. International Communications in Heat and Mass Transfer 136 (2022) 106185

1[ ( ) ]2
Δhincidence = W4 sin β4 − β4,opt (8)
2
The incidence loss is minimized by decreasing the relative flow angle
(β4) between incoming flow and turbine inlet angle whereas the turbine
exit losses can be controlled by increasing the exit flow area.
The passage loss is obtained due to a combination of viscous and
secondary flow loss within the blade passage. Generally, this type of loss
has been strongly influenced by Coriolis and centripetal forces. The
shape of flow passage and fluid density at design conditions (tempera­
ture and pressure) are also responsible for such losses [32]. Wasswe­
bauer and Glassman [33] proposed that this loss would be computed by
averaging the inlet and outlet relative velocity. The passage loss is
calculated as: (Eq. (9)).
( )
Δhpassage = 0.5K W32 cos2 i + W42 (9)

where K is a calibration factor.

The passage loss has been reduced by decreasing the L/D ratio and
increasing the specific speed (ns).
The tip leakage loss arises because of the pressure gradient between
the two surfaces (pressure and suction) and the allowance between the
turboexpander and outer casing. The allowance confides in the
manufacturing tolerances and other aerodynamic limits. The tip clear­
ance loss is calculated as (Eq. (10)):

U43 Zr ( √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ )
Δhtip = Kx εx Cx + Kr εr Cr + Kx,r εx εr Cx Cr (10)
8π
where Cx (axial) and Cr (radial) are the coefficients which are
calculated from Eqs. (11) and (12) respectively.
r4,sh
1−
(11)
r3
Cx =
Cm3 b3
(r ) Z − b
(12)
4,sh r 3
Cr =
4 Cm4 r4 b4
The tip leakage loss is related to the tip gaps from which mass flow Fig. 2. Flowchart of the one-dimensional design methodology.
rate is reduced in between the blade passages. It is also associated with
turboexpander geometry. This loss has been curtailed by minimizing the
radial clearance at the turboexpander outlet.
The exit energy loss appears because of energy conversion at the
turboexpander exit. Such losses generally introduce swirls. Since that
kinetic energy is unworkable, it is considered as a loss. The exit energy
loss is calculated as (Eq. (13)):
1
Δhexit = C42 (13)
2
It is considered that the fluid density does not affect this loss. How­
ever, it has been minimized by increasing the cross-sectional area at the
turboexpander exit.
These losses have been integrated with the preliminary design code
to update the initial prediction of total-to-static efficiency (ηts) (Eq. (14))
[34].
ΔhTotal,loss = Δhincidence + Δhpassage + Δhtip + Δhexit + ΔhTEL + ΔhO (14)

After incorporating the aforementioned losses, the attained effi­

ciency from Eq. (15) is used to update the initial guess.
Δhact
ηt− s = (15)
Δhact + Δhloss,turbine

Lastly, the mass flow rate at the turboexpander outlet is calculated Fig. 3. Various losses of the turbine.
using Eq. (16) and compared with the assumed value at the nozzle inlet.
The new mass flow rate is updated and the procedure is continued till velocity diagram. Fig. 2 represents the flowchart of preliminary design
convergence. optimization procedure used in the present study. The additional details
of the design procedure for other components are described in our
m4 = ρ4 Cm4 A4 (16)
previous study [8].
Fig. 1 represents the schematic diagram of turboexpander and its

4
M. Kumar et al. International Communications in Heat and Mass Transfer 136 (2022) 106185

Table 2 Table 3
Normalized sensitivity index of important design parameters and their optimum Input parameters of a radial expansion turbine.
range. Design variables Unit Range
Non-dimensional Normalized Normalized Optimum
Inlet total pressure bar 4–17
design variables sensitivity index sensitivity index range
Outlet static pressure bar 1.2–8.2
(ηts) (Total loss)
Inlet total temperature K 90–180
Pressure ratio (rp) 0.72 0.69 2.68–3.96 Mass flow rate kg/s 0.01–0.10
Rh/R3 0.62 0.74 0.25–0.28 Specific speed (ns) – 0.48–0.60
Rs/R3 0.44 0.51 0.86–0.92 Specific diameter (ds) – 2.81–3.38
Static head 0.17 0.14 0.79–0.88 Rotational speed rpm 0.6–1.8 (×105)
coefficient Ψ) Tip clearance mm 0.1–0.5
Blade speed ratio 0.87 0.43 0.65–0.69 Flow coefficient (ϕ) – 0.82–0.96
(vs) Stage loading (ψ) – 0.72–0.88
Flow coefficient (ϕ) 0.21 0.13 0.23–0.34 Number of blades (Turbine) – 7–17
Number of blades 0.58 0.64 10–13 Number of blades (Nozzle) – 15–31
(Z) Turbine inlet flow angle degree 70–83
Specific speed (ns) 0.29 0.22 0.48–0.55 Turbine outlet flow angle degree − 2.8 to − 3.8
Hub radius ratio – 0.20–0.32
Shroud radius ratio – 0.74–0.96
2.2. Loss analysis

Fig. 3 represents the details of different losses obtained for high- Table 4
pressure turboexpander (HPT), medium pressure turboexpander Specifications of designed turbine.
(MPT), and low-pressure turboexpander (LPT). It is noticed that passage Parameters High- Medium Low-
loss (approximately 40%), tip clearance loss (approximately 23%), and pressure pressure pressure
incidence loss (approximately 7%) contribution to HPT is higher.
Flow coefficient (ϕ) 0.32 0.28 0.26
However, exit energy loss (approximately 28%) and trailing edge loss Specific speed (ns) 0.55 0.53 0.52
(approximately 15%) are maximum for LPT. This may happen because Rotational speed (ω) (krpm) 180 160 130
of higher kinetic energy loss at the exit of LPT as compared to MPT and Turbine inlet diameter (d3) (mm) 24.76 26.89 28.74
HPT. It is also observed that among all the losses, passage loss contrib­ Turbine outlet diameter (d4) 15.42 16.20 17.96
(mm)
utes the maximum for all the turboexpanders. This may happen due to Blade height (turbine inlet) (b3) 1.68 2.41 2.63
higher secondary loss contribution inside the blade passages. However, (mm)
the tip clearance, exit energy, trailing edge, and incidence losses Blade height (turbine outlet) (b4) 0.84 1.26 1.38
contribute in the decreasing order for HPT, MPT, and LPT respectively. (mm)
Mach number at turbine inlet 0.89 0.85 0.80
The other losses for HPT, MPT, and LPT are approximately 4, 3, and 5%
(M3)
respectively. Blade speed (U3) (m/s) 273.85 250.22 224.80
Absolute velocity (C3) (m/s) 194.49 181.12 175.50
Axial velocity (W3) (m/s) 86.62 82.64 76.09
2.3. Sensitivity analysis Absolute velocity angle (α3) 80 80 78
(degree)
Sobol method is a model-independent sensitivity analysis technique Head coefficient (ψ) 0.85 0.82 0.80
to measure the total output variances of input parameters. It can solve Absolute flow angle (β3) (degree) − 55.89 − 61.13 − 61.01
Number of blades (Z) 13 13 11
the non-linear and non-monotonic functions efficiently. The relation is Mass flow rate (m.) (kg/s) 0.08 0.06 0.05
expressed in functional form as Eq. (17): Total losses (Lt) 0.23 0.20 0.19
( ) Total-to-static efficiency (ηts) 81.21 79.42 77.31
Y = f (X) = f X1 , …..,Xp (17) Power (kW) 5.41 4.51 3.81

where Y is the objective function (model output) and X = (X1, …. ., Xp) is

the parameter. The function f is decomposed as follows (Eq. (18)): Vij
Sij = (21)
( ) p
∑ p
∑ V
f X1 , …..,Xp = f0 + fi (Xi ) +
i=1 i=1 Total sensitivity index (Eq. (22)):
p
∑ ( ) ( ) ∑
× fij Xi Xj + ….. + f1,…..p X1 , …..,Xp (18) STi = Si + Sij + … (22)
j=1+1 j∕
=i

where f0 is a constant which is approximately equal to the expected Table 2 illustrates the normalized sensitivity indices and optimal
value (output). ranges of important non-dimensional and geometrical design parame­
The variance decomposition has been expressed as (Eq. (19)): ters of the turboexpander. It is observed that the blade speed ratio (vs),
pressure ratio (rp), number of blades (Z), Rh/R3, and Rs/R3 have a
maximum sensitivity index to total-to-static efficiency (ηts) and total
p
∑ p− 1 ∑
∑ p
V(Y) = Vi + Vij + ….. + V1,…..p (19)
i=1 i=1 j=1+1 loss. Therefore, the value of these parameters has been updated in the
preliminary design process. Based on these ranges, three turboexpanders
Following the aforementioned procedure, the total output variance (High, medium, and low-pressure) are developed. Major preliminary
of each parameter has been determined in terms of sensitivity indices design parameters and obtained results of the designed turboexpanders
(ratio of partial to total variance). are mentioned in Tables 3 and 4.
First-order sensitivity index (Eq. (20)):
Vi 3. Experimental techniques
Si = (20)
V
The experimental test rig has been developed using different indoor
Second-order sensitivity index (Eq. (21)):

5
M. Kumar et al. International Communications in Heat and Mass Transfer 136 (2022) 106185

Fig. 4. Schematic of the experimental test rig.

particles, dust particles, and other atmospheric impurities which may be

the reason for diminishing the performance of the turboexpander. Then,
a freon precooler has been situated for dehumidification and reducing
the temperature of compressed fluid. After that, the precooled and
dehumidified fluid passes through a cryogenic heat exchanger (plate-fin
type) for further cooling in the presence of liquid nitrogen (coolant). All
the supply lines (heat exchanger and turboexpander) have been con­
nected with flexible and adaptable stainless-steel pipes (Grade SS304).
These supply lines have been insulated with a 2 cm thick nitrile rubber
pipe and perlite powder to attain the adiabatic condition. Before entry
into the turboexpander unit, pressure gauge, rotameter, and ball valves
have been situated to control and measure the mass flow rate and
operating pressure.
The height of the experimental turbine blade is 22.04 mm, with a hub
and shroud diameter of 22.14 mm and 30.16 mm respectively. There are
thirteen turbine blades (aluminum alloy Al 6082) and 21 nozzles. The
turbine blade tip clearance and shaft diameter are 1.0 mm and 16.08
mm respectively. The turboexpander assembly (includes bearing) has
been situated in the bearing housing (160.50 mm diameter and 157.50
Fig. 5. Experimental set-up. mm height) which is fabricated from austenitic stainless steel (Grade-
316). The other stationary components (nozzle, bearings, shaft, and
(plate-fin heat exchanger, turboexpander set-up, reservoir, measuring brake compressor) have been fabricated from chromium‑vanadium steel
instruments) and outdoor (Screw compressor, Freon precooler, purifi­ (AISI-6118H). The assembly is made leak-proof by using O-ring
cation, and filtration unit) equipment, as shown in Figs. 4 and 5. The (neoprene). Fig. 6 represents the various components of a turboexpander
description of all the equipment is explained herewith: unit. Table 5 represents the additional details of fabricated turboex­
The high-pressure fluid has been supplied from the screw compressor pander (HPT) for experimental analysis. The dimensions of the fabri­
(KAESER SIGMA BSD 72). The compressed fluid passes through the cated turboexpander have been based on the obtained results from high-
filtration and purification unit to remove carbon particles, moisture, oil pressure turboexpander. The developed experimental set-up is a Claude
cycle-based turboexpander system used for refrigeration and

6
M. Kumar et al. International Communications in Heat and Mass Transfer 136 (2022) 106185

Fig. 6. Turboexpander components and bearing housing.

Table 5 Table 8
Specifications of experimental turboexpander. Node allocations in different sections (HPT / MPT / LPT).
Turbine inlet diameter (mm) 24.90 Inlet block Passage block Outlet block
Turbine outlet diameter (mm) 15.60
Streamwise Blade-to- Streamwise Blade-to- Streamwise Blade-to-
Number of nozzles 21
blade blade blade
Nozzle pitch circle diameter (mm) 40.17
Nozzle outlet diameter (mm) 25.54 Turbine 30 / 30 / 30 64 /64 / 54 / 52 / 52 82 / 82 / 30 / 30 / 28 60 / 60 /
Shaft diameter (mm) 16.08 60 80 60
Number of turbine blades (Z) 13 Nozzle 25 / 25 / 25 60 / 60 / 64 / 64 / 60 52 / 52 / 25 / 25 /25 44 / 42 /
60 50 44

Table 6 ADAM®. The mass flow rate is measured using rotameter. The specifi­
Specifications of measuring instruments. cations of the measuring devices are mentioned in Table 6.
Instrument Company Range Accuracy Additionally, the four important dimensionless parameters are
Temperature sensor ADAM 70–800 K – calculated from experimental results. The error analysis has been con­
Flow meter Alflow 1–30 m3/h ± 2.5% ducted to analyze the effect of measurement errors on performance
Speed sensor Emerson 0.07–2500 Hz ± 0.005% (dimensionless) parameters. Table 7 represents the uncertainty analysis
Oscilloscope Tektronix 20–100 MHz ± 2% which clearly shows that the isentropic efficiency, specific speed, pres­
Digitel caliper Mitutoyo 150 mm 0.013 mm
Pressure gauge Swagelok 0–25 bar ± 0.2%
sure ratio, and blade speed ratio have maximum uncertainty of 5.41%,
Pressure transducer Endevco 0–35 bar ± 0.05% 5.17%, 3.22%, and 8.22% respectively.

4. Numerical procedure
Table 7
Error propagation analysis (experimental). 4.1. Grid generation

Parameters Value Uncertainty Referred value (%)

The coordinates obtained from the preliminary design process have
2
Isentropic efficiency 0.79 4.62 × 10− 5.41 been used to generate the three-dimensional model of the turboex­
2
Specific speed 0.54 3.96 × 10− 5.17
2 panders using BladeGen®. After that, Turbo-Grid® is utilized to create
Pressure ratio 2.16 2.17 × 10− 3.22
Blade speed ratio 0.72 5.96 × 10− 2
8.22 the computational mesh. The O/H type topology has been opted to
create the mesh in the inlet, passage, and outlet blocks using ATM to­
pology method to avoid the occurrence of negative volume [35]. The
liquefaction of process fluid. boundary layer refinement has been controlled using first element
method. The Reynolds number (to achieve the near-wall treatment) for
3.1. Measuring instruments and error analysis

To minimize the measurement uncertainties, three experimental test Table 9

y+ values at different locations (HPT / MPT / LPT).
data have been taken in succession from all the test runs. The temper­
ature and pressure sensors have been calibrated to reduce the bias errors Turbine y+ Nozzle y+
during the measurement. The inlet and outlet pressure of the turboex­ Inlet 0.82 / 0.81 / 0.84 Inlet 0.94 / 0.97 / 0.96
pander has been measured using a pressure gauge. A pressure transducer Adjacent to blade 0.76 / 0.76 / 0.82 Adjacent to blade 0.88 / 0.91 / 0.96
has been used to measure the tip pressure. The inlet and outlet tem­ Passage block 1.41 / 1.49 / 1.62 Passage block 1.84 / 1.92 / 1.96
Outlet 0.92 / 0.96 / 0.95 Outlet 0.83 / 0.86 / 0.89
perature has been measured using a resistance thermometer and

7
M. Kumar et al. International Communications in Heat and Mass Transfer 136 (2022) 106185

Table 10
Boundary conditions (HPT / MPT / LPT).
Total Total Rotational Flow
pressure temperature speed (rpm) regime
(bar) (K)

Turboexpander 16 / 8 / 4.5 150 / 120 / 95 1.21 / 1.02 / Subsonic

inlet 0.81 (×105)
Turboexpander 8 / 4.5 / 1.8 – – Subsonic
outlet

Table 11
Grid independence analysis (HPT / MPT / LPT).
Nozzle (in Turbine (in ηis Power output CPU time
Fig. 7. Computational domain in CFX® pre. million) million) (kW) (hrs)

0.28/0.26/ 0.57/0.54/ 0.682/0.734/ 3.091/2.152/ 92/90/91

satisfying the y+ criteria of the turbulence model has been set to 5 × 106. 0.29 0.55 0.729 1.864
The y+ value of first grid away from the leading and trailing edge is 0.59/0.57/ 0.84/0.82/ 0.725/0.769/ 3.794/2.771/ 126/129/
0.60 0.84 0.757 2.142 128
varying between 0.70 and 0.94, with a mean value of approximately 1.0
0.88/0.89/ 1.35/1.34/ 0.751/0.794/ 4.211/3.369/ 151/153/
whilst the remainder of the blade has an average value of 1.28. It is 0.85 1.36 0.775 2.796 153
considered to be sufficient to solve the viscous sub-layer and aero­ 1.29/1.31/ 1.98/1.99/ 0.768/0.801/ 4.821/3.742/ 176/179/
dynamic results are agreed well with experimental results. The addi­ 1.31 1.97 0.783 2.987 181
tional details of node distribution and corresponding y+ values are 1.46/1.48/ 2.24/2.25/ 0.771/0.805/ 4.833/3.761/ 231/234/
1.48 2.25 0.788 2.994 235
mentioned in Tables 8 and 9 respectively.
The computational mesh in the turbine and nozzle are created
separately. Therefore, boundary nodes may not be aligned with each
other at the interface of both components. To avoid this, a general grid
interface (GGI) has been applied for better connection and permits fluid
flow through the interface nodes. Thereafter, the nozzle outlet and
turbine inlet interfaces have been created (using general grid interface
(GGI)) in CFX®. The computational domain consists of one stationary
component (nozzle) and one rotating component (turbine). The rotation
of turbine in steady-state conditions has been considered using stage
interface (mixing frame model) in which only reference frame changes
but the relative position of the components is fixed at the interface. In
this condition, there is no need for grid movement. However, transient
blade row simulations are carried out with an automatic pitch change
option to characterize the fluid flow along the streamwise location of the
turboexpander passages in the rotating coordinate system. This model is
realistic for solving the turbomachinery problems under rotating con­
ditions. The solution obtained from the steady-state condition has been
used to initialize the transient blade row simulation. Fig. 7 represents
the details of the computational domain in CFX® pre.

4.2. Turbulence model

The Reynolds-Averaged Navier-Stokes (RANS) equations are

resolved which is based on finite volume approach. The upwind
advection scheme (second-order) has been selected to discretize the
RANS equations. The shear stress turbulence (SST) model has been
recommended to solve the turbomachinery-related numerical issues
because it possesses a combined feature of k-ω and k-ε model [6,10,38].
This model accomplishes k-ω model to show the turbulent boundary Fig. 8. Computational mesh of HPT, MPT, and LPT.
layer flows near the wall surfaces whereas the k-ε model solves the free
shear flow in the mainstream. A blending function is specified for the model for better accuracy of the numerical results. The convergence
smooth transition between the k-ω and k-ε models, which varies from criteria are set when the constants of root-mean-square of mass, mo­
one to zero in the logarithmic boundary layer region and free shear layer mentum, and energy equations are <10− 6. All numerical simulations are
regions respectively. SST model combines the assimilating features of performed on a Dell workstation with 64 GB RAM (3.00 GHz).
these two turbulence models. Additionally, the eddy viscosity has been
modified to consider the issues related to turbulence shear stress
transportation which is responsible to predict the adverse pressure 4.3. Boundary conditions
gradient flows. The scalable wall function has been selected for SST
turbulence model to resolve the velocity profile adjacent to the solid Total pressure and total temperature have opted at the turboex­
boundary which is generally used in turbomachinery-related problems pander inlet. Static pressure is imposed at the outlet of the turboex­
and recommended by [36,37]. The thermodynamic properties of nitro­ pander. The fluid flow direction is normal to the nozzle inlet and inlet
gen have been obtained from the Peng-Robinson (equation of state) turbulence intensity is fixed to be 5%. The solid boundaries are supposed

8
M. Kumar et al. International Communications in Heat and Mass Transfer 136 (2022) 106185

Fig. 9. Variation of isentropic efficiency and mass flow rate with pressure ratio
(a) Isentropic efficiency with pressure ratio.
(b) Mass flow rate with pressure ratio.

to be no-slip, adiabatic, and hydraulically smooth. Since the simulations 5. Results and discussions
are carried out using a single turboexpander domain, the cyclic peri­
odicity has been imposed along the circumferential direction to decrease 5.1. Experimental results
the computational cost. The additional details of boundary conditions
have been mentioned in Table 10. In this section, the experimental and numerical results are presented
and discussed with proper validation. The important basic parameters
4.4. Grid sensitivity analysis used to describe the performance of a turboexpander system includes
rotational speed, pressure ratio, mass flow rate, power output, and
The isentropic efficiency and power output of the turboexpander isentropic efficiency. The pressure ratio (rp), specific speed (ns), and
have been compared for five different grid resolutions to check the blade speed ratio (vs) have been used as independent variable while
reliability of the numerical simulations for all turboexpander systems as isentropic efficiency (ηi), mass flow rate, and power output are depen­
revealed in Table 11 and Fig. S1. It shows that the fourth and fifth rows dent variables.
have a very small variation of output power and isentropic efficiency Fig. 9 represents a comparison between isentropic efficiency and
(maximum variation <1%) however, the computational time is rela­ mass flow rate with respect to pressure ratio at three different rotational
tively higher. Therefore, the grid resolution of the fourth row of Table 11 speeds (80,914 rpm, 102,419 rpm, and 120,612 rpm). It is noticed that
provides satisfactory results and has been considered for all the nu­ the isentropic efficiency increases with an increase in pressure ratio
merical simulations as shown in Fig. 8. reach its maximum value and then decreases. The highest isentropic
efficiency has been achieved for a pressure ratio of 1.84, 2.70, and 3.75

Fig. 10. Isentropic efficiencies as a function of (a) specific speed (ns) (b) Blade speed ratio (vs).

9
M. Kumar et al. International Communications in Heat and Mass Transfer 136 (2022) 106185

Fig. 11. Variations of (a) rotational speed (b) temperature drop with power output.

Fig. 12. (a) Pressure (b) Mach number (c) Temperature (d) Static entropy contours at 50% blade height.
(a) Pressure contours.
(b) Mach number contours.
(c) Temperature contours.
(d) Static entropy contours.

10
M. Kumar et al. International Communications in Heat and Mass Transfer 136 (2022) 106185

Fig. 12. (continued).

at 80,914 rpm, 102,419 rpm, and 120,612 rpm respectively. At low isentropic efficiency as a function of blade speed ratio and specific
rotational speed, the isentropic efficiency is highest at a pressure ratio of speed. It shows that the highest efficiency has been obtained for specific
1.8–2.0 whereas it is highest for medium and high rotational speed for a speeds in the range of 0.52–0.58 (approximately 0.54). Fig. 10 (b) rep­
pressure ratio of 2.4–2.8 and 3.5–4.0 respectively. It is also noticed that resents that the maximum efficiency has been obtained for LPT, MPT,
the maximum efficiency is obtained at a lower rotational speed. It may and HPT when the blade speed ratio is 0.58, 0.64, and 0.67 respectively.
happen because minimum losses are obtained at this speed. However, at Fig. 11 represents the variation of rotational speed and temperature
high-pressure ratios where isentropic efficiency decreases, the losses are drop with respect to power output. It is prominent that the power output
higher which diminishes the efficiency of the turboexpander. increases with increase in rotational speed of the turboexpander. It
Fig. 9 (b) represents the variation of pressure ratio with mass flow happens because of work done by the fluid on the turboexpander blade.
rate at different rotational speeds. It is obvious that the pressure ratio In this process, the kinetic energy of the fluid has been transformed into
increases with mass flow rate as it is directly related to square of the the rotational energy of the turboexpander blade, consequently, tem­
mass velocity under turbulent flow conditions. It is noticed that the perature decreases. It is distinguished that the highest power output has
pressure ratio increases with increase in mass flow rate, reaches its been obtained at a rotational speed of 1.21 × 105 and 1.02 × 105 rpm for
maximum value at a specific mass flow rate. Further, an increase in mass which approximately 25 and 20 K temperature drop occurs for high
flow rate decreases the pressure ratio. It can be seen that the specific (HPT) and medium (MPT) pressure turboexpander respectively.
mass flow rate and pressure ratio for LPT, MPT, and HPT (from nu­ The results conclude that the HPT has better performance (ηis, P, and
merical analysis) are approximately 0.09 kg/s and 3.65, 4.00, and 4.30 temperature drop) as compared to MPT and LPT. However, the LPT has
respectively. However, the experimental results show the specific mass higher efficiency at low-pressure ratio ranges (approximately
flow rate and corresponding pressure ratio for LPT, MPT, and HPT are 1.80–2.10).
0.85 kg/s for rp = 2.45, 0.86 kg/s for rp = 3.25, and 0.82 kg/s for rp =
3.60 respectively. The difference in numerical and experimental results 5.2. Numerical results
appears due to limitation of the experimental facility.
The performance of turboexpander can also be represented in terms Fig. 12 represents the area-averaged contours of thermophysical
of specific speed and blade speed ratio. Fig. 10 illustrates the variation of characteristics such as pressure, Mach number, temperature, and static

11
M. Kumar et al. International Communications in Heat and Mass Transfer 136 (2022) 106185

Fig. 13. (a) Density variation (area-averaged) along the streamwise location (b, c) Variation of different properties inside the turboexpander.

entropy for all the turboexpander at mid-span (50% blade height). The higher at the pressure surface (up to 1.30 streamwise distance from the
expansion of the process fluid throughout the flow passage results in leading edge) in contrast to that of the suction surface. It may happen by
pressure drop and corresponding increase in kinetic energy of the fluid. striking of the high kinetic energy expanded fluid imminent to the
In this process, the rotational speed of the turbine has been increased leading edge. After 1.30 streamwise location, the kinetic energy di­
which is the prime reason for increase in Mach number and corre­ minishes and Mach number decreases and finally stabilized. However,
sponding enthalpy and temperature drop. In all the cases, the flow is there is some instantaneous increase in Mach number near the trailing
smoothly accelerated on the first half of the pressure side because of an edge. This happens due to sudden contraction in the flow passage area
increase in Mach number (near the turbine leading edge). However, the and static pressure drop in these regions.
suction side experiences a local increase in Mach number at the trailing Fig. 12 (c) represents the temperature contours at 50% blade height.
edge. It may happen because of instantaneous pressure drop in these The temperature drop for HPT is more coherent than other cases. It is
regions or rotation of the turbine blade. It is noticed that the pressure noticed that the temperature decreases along the flow passages and
drop in the turbine blade passages are higher up to 1.40 streamwise becomes minimum at the turbine outlet. It happens because of contin­
location thereafter the variation is smooth which shows the designed uous pressure drop along the flow passage. It is also observed that the
models (HPT, MPT, and LPT) are adequate. instantaneous temperature drop occurs near the trailing edge of the
Fig. 12 (b) represents the Mach number contours at 50% blade height turboexpander because of the static pressure drop in these regions. After
which shows that the fluid flow throughout the turboexpander is in the that, the kinetic energy of the fluid decreases, as a result, the pressure
subsonic regime. The maximum Mach number of HPT, MPT, and LPT are energy increases, and finally a small increase in temperature (approxi­
0.86, 0.80, and 0.77 respectively, which occur near the leading edge of mately 2 K) takes place. It is observed that the maximum temperature
the turbine blade. It is noticed that the intensity of Mach number is drop (24 K) takes place for high-pressure turboexpander. In medium and

12
M. Kumar et al. International Communications in Heat and Mass Transfer 136 (2022) 106185

Fig. 14. Velocity vector at different blade heights (spans) for high-pressure turboexpander.

low-pressure turboexpander, 20 and 16 K temperature drop takes place. Fig. 13 (c) represents the temperature Vs static entropy variation to
This happens because of higher enthalpy drop in HPT (Fig. 13 (b)). In understand the thermodynamic conditions (expansion process) of the
MPT and LPT, the decrease in temperature drop may happen due to inlet fluid inside the turboexpander. For an ideal process, the line should be
fluid temperature advancing towards its boiling point. Under these vertical (isentropic process), but for a real process, it deviates. The in­
conditions, the thermodynamic properties of fluid-like density, molec­ crease in inclination shows that the entropy generation during the
ular viscosity, etc. have severe changes that may affect the corre­ expansion process inside the turboexpander. It is observed that the
sponding enthalpy drop inside the turboexpander. maximum entropy generation takes place inside the turbine where
Fig. 12 (d) characterizes the static entropy contours at 50% blade higher temperature gradients have been obtained.
height. It shows that the increase in entropy inside the nozzle is mini­
mum whereas it increases at the leading edge of the turbine due to
passage vortices and boundary layer separation in these regions 5.3. Internal flow field analysis
(Figs. 14-16). Additionally, the increase in shear strain of the fluid
because of the blending of flow stream from pressure and suction side The streamline anatomy of the turbine blade profile and boundary
may also be responsible for increase in entropy in these regions. It is layer formation near the leading and trailing edge, and hub and shroud
visualized that; the static entropy is highest at the outlet of the turbine. layers are responsible for secondary flow (especially passage and
This may happen due to viscous dissipation, flow reversal, and other scraping flow vortices). Also, the meridional curvature has been
non-ideal conditions in these regions. responsible for those flow patterns (vortices) at pressure and suction
Fig. 13 (a) demonstrates the variation of density (area-averaged) surfaces, which globally form the boundary layers in those regions.
inside the turboexpander along the streamwise location. It depicts the These types of flow behaviors have prevailed inside the turbine flow
density variation is strongly influenced by the pressure drop of the fluid. passage. Therefore, the fluid flow structures of turbines of HPT, MPT,
Fig. 13 (b) depicts the pressure variation with static enthalpy (p-h dia­ and LPT have been represented and explained in detail.
gram) throughout the turboexpander. The diagram indicates the static Figs. 14-16 represents the velocity vectors at 10, 50, and 90% blade
enthalpy drop for corresponding change in pressure inside the tur­ height. It is interesting to observe that the secondary flow regime has
boexpander, which gives the refrigerating capacity of the system. It has been obtained at the pressure surface of the turbine blade passage
been distinguished that the HPT has maximum enthalpy drop, as a because of the pressure gradient induced by the rotation of the turbine.
result, a higher temperature drop occurs (Fig. 12 (c)). For all the designed turboexpanders, it is noted that the fluid flows
steadily and there are few regions where flow separation exists. The

13
M. Kumar et al. International Communications in Heat and Mass Transfer 136 (2022) 106185

Fig. 15. Velocity vector at different blade heights (spans) for medium pressure turboexpander.

secondary flow appears near the blade wall because of Coriolis accel­ which further transforms into the passage vortices. These vortices have
eration and dominance of the radial pressure gradient of the fluid. An been generated mostly at the pressure side (PS) (50% blade height) at
analogous phenomenon has been observed and explained in our previ­ the leading edge which disappears after mixing with the mainstream
ous study [39]. However, the appearance of such flow characteristics flow and rotation of the turbine blade. However, at the pressure side
near the pressure side is because of pressure non-uniformities in these (90% blade height), scraping flow exists on the leading edge which
regions. Moreover, at higher blade heights, a relatively weak secondary disappears at higher streamwise locations (between 1.50 and 1.60) and
flow exists at the suction surface which mixes with the mainstream flow may be the reason for flow separation in these regions. Similarly, the tip
at higher span and then disappeared. These flow characteristics increase leakage flow exists adjacent to the suction side (SS) adjacent to the mid-
the entropy of the fluid which diminishes the performance of the tur­ section of the turbine blade passage. The passage vortices appear at 50%
boexpander (Fig. 12 (d)). It is also noticed that the velocity vectors are blade height along the leading edge on the pressure side of the flow
directed towards the suction side for all the cases. This happens because passage. Along the further downstream direction, the passage vortices
of difference in magnitude of velocity vectors is greater along the disappeared by mixing with the mainstream (fully developed) flow.
pressure side. Also, a relatively higher pressure has been obtained near In Fig. 15, the passage vortices have been obtained at 10 and 50%
the tip regions induces tip leakage flow mostly near the suction side of blade height near the leading edge which further changes into the
all the turbines. scraping flow at a higher span. The intensity of passage vortices and
In Fig. 14, the flow separation occurs adjacent to the leading edge at scraping flow have been decreased at higher streamwise locations (50
10% blade height. The induced pressure gradient in the flow passage is and 90% blade height). However, at 10% blade height, the scraping flow
the main reason for flow separation along the downstream direction appears at a higher streamwise location (at PS between 1.35 and 1.75).

14
M. Kumar et al. International Communications in Heat and Mass Transfer 136 (2022) 106185

Fig. 16. Velocity vector at different blade heights (spans) for low-pressure turboexpander.

Table 12
Comparison of 1-D, numerical simulation and experimental results.
(1-D / numerical / HPT MPT LPT
experiment)

Isentropic efficiency (%) 81.21/80.32/ 79.42/78.82/ 77.31/

79.07 77.26 76.46/−
Mass flow rate (kg/s) 0.08 0.06 0.05
Power output (kW) 5.41/5.28/5.01 4.51/4.42/4.34 3.81/3.69/−

Flow separation appears near the pressure side (10% blade height)
because of the adverse pressure gradient which mixes with the main­
stream flow and transformed into scraping flow on higher streamwise
location. Additionally, tip leakage flow exists at pressure surface of 50%
blade height. It occurs because of pressure differences along the down­
stream direction.
In Fig. 16, the flow separation occurs adjacent to the leading edge of
the turbine blade for 10% blade heights and at the leading and trailing
edge for 50% blade height. At 90% blade heights, scraping flow exists
adjacent to the leading edge of the turbine blade.
Fig. 17 represents the area-averaged turbulence kinetic energy (TKE)
variation inside the turboexpander. It is noticed that the TKE increases
after 0.80 streamwise location inside the nozzle. After that, the intensity
Fig. 17. Area-averaged TKE distribution along the streamwise location. of TKE increases reaches its maximum value at 1.30, 1.20, and 1.10
streamwise location for HPT, MPT, and LPT respectively. The main
reason for the sudden increase of TKE is highly turbulent flow inside the
turbine governed by the blade curvature.
The performance parameters of all the turboexpanders from

15
M. Kumar et al. International Communications in Heat and Mass Transfer 136 (2022) 106185

preliminary (1-D), numerical, and experimental results are shown in interest.

Table 12.
Appendix A. Supplementary data
6. Conclusions
Supplementary data to this article can be found online at https://doi.
The flow field and thermodynamic analysis of cryogenic turboex­ org/10.1016/j.icheatmasstransfer.2022.106185.
pander are numerically investigated and experimentally verified. The k-
ω SST model is used to investigate the thermophysical behavior of References
process fluid nitrogen at different operating pressure, temperature, and
rotational speed. The important findings of the present study are: [1] Wan Sun, Yu Shuangtao Chen, Shanshan Bu Hou, Zaiyong Ma, Luteng Zhang,
Liangming Pan, Numerical studies on two-phase flow in cryogenic radial-inflow
turbo-expander using varying condensation models, Appl. Therm. Eng. 156 (2019)
1. The preliminary design procedure of a radial expansion turbine 168–177.
incorporated with different loss correlations is explained. A Sobol [2] Manoj Kumar, A comparative investigation of thermodynamic performance and
flow field structure of a modified Collins cycle-based helium turboexpander
sensitivity analysis has been conducted to determine the normalized system, Int. Comm. Heat Mass Transf. 127 (2021), 105554.
sensitivity index of different design variables. The optimum range of [3] Subrata K. Ghosh, R.K. Sahoo, Sunil K. Sarangi, Mathematical analysis for off-
these variables is also determined for higher performance of the design performance of cryogenic turboexpander, J. Fluids Eng. 133 (3) (2011),
031001.
turbine. [4] Ronald H. Aungier, Preliminary aerodynamic design of axial-flow turbine stages,
2. The performance of the turboexpander is predicted by developing an in: Turbine Aerodynamics: Axial-Flow and Radial-Flow Turbine Design and
experimental test facility with extended measuring points. The Analysis, ASME press, 2006.
[5] Emilie Sauret, Andrew S. Rowlands, Candidate radial-inflow turbines and high-
variation of mass flow rate, blade speed ratio, pressure ratio, and
density working fluids for geothermal power systems, Energy 36 (7) (2011)
rotational speed on isentropic efficiency, power output, and tem­ 4460–4467.
perature drop are discussed. Additionally, the effect of specific speed [6] Emilie Sauret, Gu Yuantong, Three-dimensional off-design numerical analysis of an
and blade speed ratio variations on the isentropic efficiency is also organic Rankine cycle radial-inflow turbine, Appl. Energy 135 (2014) 202–211.
[7] N.C. Baines, A meanline prediction method for radial turbine efficiency, in:
explained. The error analysis has been conducted to measure the IMECHE Conference Transactions vol. 11, 1998, pp. 45–56.
effect on performance parameters which shows that the blade speed [8] Manoj Kumar, Suraj K. Behera, Amitesh Kumar, Ranjit K. Sahoo, Numerical and
ratio and isentropic efficiency have maximum uncertainty (8.22% experimental investigation to visualize the fluid flow and thermal characteristics of
a cryogenic turboexpander, Energy 189 (2019), 116267, https://doi.org/10.1016/
and 5.41%) during the experiments. j.energy.2019.116267.
3. Finally, a comparative numerical study has been carried out to [9] Manoj Kumar, Debashis Panda, Suraj K. Behera, Ranjit K. Sahoo, Experimental
characterize the fluid flow field and thermodynamic behavior of investigation and performance prediction of a cryogenic turboexpander using
artificial intelligence techniques, Appl. Therm. Eng. 162 (2019), 114273, https://
nitrogen fluid inside the turboexpander at three operating condi­ doi.org/10.1016/j.applthermaleng.2019.114273.
tions. A comparison of performance parameters between 1-D, nu­ [10] Manoj Kumar, Debashis Panda, Amitesh Kumar, Ranjit K. Sahoo, Suraj K. Behera,
merical, and experimental results are also summarized. The highest A methodology for the performance prediction: flow field and thermal analysis of a
helium turboexpander, J. Braz. Soc. Mech. Sci. Eng. 41 (11) (2019) 484.
temperature drop and isentropic efficiency are 24 K and 80.32% for [11] Manoj Kumar, Suraj K. Behera, Amitesh Kumar, Ranjit K. Sahoo, Rotational effect
high-pressure turboexpander. on flow field and thermal characteristics of a Turboexpander for helium
liquefaction system: a numerical perspective, J. Heat Transf. 142 (2020) 11.
[12] Violette Mounier, Luis Eric Olmedo, Jürg Schiffmann, Small scale radial inflow
Credit author statement
turbine performance and pre-design maps for organic Rankine cycles, Energy 143
(2018) 1072–1084.
Dr. Manoj Kumar [13] Kiyarash Rahbar, Saad Mahmoud, Raya K. Al-Dadah, Nima Moazami, Parametric
analysis and optimization of a small-scale radial turbine for organic Rankine cycle,
Postdoctoral Research Associate, Cryogenic Plant and Instru­
Energy 83 (2015) 696–711.
mentation Section, Variable Energy Cyclotron Centre, Department of [14] Nick C. Baines, A. Whitfield, Design of Radial Turbomachines, Longman Scientific
Atomic Energy, Kolkata 700064, India and Technical, Essex, UK, 1990.
Credit statement: Conceptualization, Data curation, Investigation, [15] A.W. Costall, A. Gonzalez Hernandez, P.J. Newton, R.F. Martinez-Botas, Design
methodology for radial turbo expanders in mobile organic Rankine cycle
Validation, Visualization, Writing - original draft, Writing - review. applications, Appl. Energy 157 (2015) 729–743.
Rasmikanti Biswal [16] B. Ssebabi, R.T. Dobson, A.B. Sebitosi, Characterising a turbine for application in
External research student, Department of Mechanical Engineering, an organic Rankine cycle, Energy 93 (2015) 1617–1632.
[17] W. Jansen, E.B. Qvale, A rapid method for predicting the off-design performance of
National Institute of Technology Rourkela, Odisha, India, 769008 radial-inflow turbines, in: ASME 1967 Winter Annual Meeting: GT Papers, 1967 pp.
Credit statement: Formal Analysis, Software. V001T01A002.
Prof. Suraj K. Behera [18] A. Dadone, M. Pandolfi, A method for evaluating the off-design performance of a
radial inflow turbine and comparison with experiments, Int. J. Mech. Sci. 11 (3)
Assistant Professor, Department of Mechanical Engineering, Na­ (1969) 241–252.
tional Institute of Technology Rourkela, Odisha, India, 769008 [19] A.K. Gupta, M. Kumar, R.K. Sahoo, S.K. Sarangi, Analytical and experimental
Credit statement: Data curation, Funding acquisition, Project investigation of a plate fin heat exchanger at cryogenics temperature, Int. J. Heat
and Technol. 39 (4) (2021) 1225–1235, https://doi.org/10.18280/ijht.390420.
administration, Resources, Writing - review.
[20] Peter L. Meitner, Arthur J. Glassman, Off-Design Performance Loss Model for
Prof. Ranjit K. Sahoo Radial Turbines with Pivoting, Variable-Area Stators, National Aeronautics and
Emeritus Professor, Department of Mechanical Engineering, Na­ Space Administration Cleveland Oh Lewis Research Center, 1980.
[21] Lu Niu, Yang Meng, Shuangtao Chen, Xiaojiang Li, Qiang Zeng, Fang Lou, Yu Hou,
tional Institute of Technology Rourkela, Odisha, India, 769008, Adjunct
Off-design performance analysis of cryogenic turbo-expander based on mathematic
Professor, National Institute of Technology Meghalaya, India, 793003. prediction and experiment research, Appl. Therm. Eng. 138 (2018) 873–887.
Credit statement: Supervision, Conceptualization. [22] Lu Niu, Xingya Chen, Wan Sun, Shuangtao Chen, Yu Hou, Non-equilibrium
Dr. Sandip Pal spontaneous condensation flow in cryogenic turboexpander based on mean
streamline off-design method, Cryogenics 98 (2019) 18–28.
Scientist G, Cryogenic Plant and Instrumentation Section, Variable [23] Long Shao, Jie Zhu, Xiangrui Meng, Xinli Wei, Xinling Ma, Experimental study of
Energy Cyclotron Centre (VECC), Department of Atomic Energy, Kol­ an organic Rankine cycle system with radial inflow turbine and R123, Appl.
kata, India, 700064 Therm. Eng. 124 (2017) 940–947.
[24] Fuhaid Alshammari, Apostolos Pesyridis, Apostolos Karvountzis-Kontakiotis,
Credit statement: Writing - review. Ben Franchetti, Yagos Pesmazoglou, Experimental study of a small scale organic
Rankine cycle waste heat recovery system for a heavy duty diesel engine with focus
Declaration of Competing Interest …, Appl. Energy 215 (2018) 543–555.

The corresponding author declares that there is no conflict of

16
M. Kumar et al. International Communications in Heat and Mass Transfer 136 (2022) 106185

[25] Subrata K. Ghosh, Ranjit K. Sahoo, Sunil K. Sarangi, Experimental performance Nomenclature
study of cryogenic turboexpander by using aerodynamic thrust bearing, Appl.
Therm. Eng. 30 (11–12) (2010) 1304–1311.
[26] Balaji Kumar Choudhury, Design and Construction of Turboexpander Based
Nitrogen Liquefier, Ph.D. Dissertation, NIT Rourkela, India, 2013.
[27] N. Ino, A. Machida, K. Ttsugawa, Y. Arai, M. Matsuki, H. Hashimoto, A. Yasuda, Variables
Development of high expansion ratio helium turbo expander, in: Advances in A: Area (m2)
Cryogenic Engineering, Springer, 1992, pp. 835–844. b: Blade height (m)
[28] Xiaoming Li, Cui Lv, Shaoqi Yang, Jian Li, Bicai Deng, Qing Li, Preliminary design C: Absolute velocity (m/s)
and performance analysis of a radial inflow turbine for a large-scale helium Cx: Axial coefficient
cryogenic system, Energy 167 (2019) 106–116. Cr: Radial coefficient
[29] A. Whitfield, F.J. Wallace, Study of incidence loss models in radial and mixed-flow D: Turbine diameter (m)
turbomachinery, in: Conference on Heat and Fluid Flow in Steam and Gas Turbine d: Diameter (m)
Plant, Coventry, England, 1973, pp. 122–128. ds: Specific diameter
[30] M. Kumar, R. Biswal, A.K. Gupta, S.K. Behera, R.K. Sahoo, Effect of wall heat flux h: Enthalpy (kJ/kg)
on fluid flow and thermal characteristics of a turbulent dual jet, Mathematical K: Calibration factor
Modelling of Engineering Problems 7 (1) (2020) 127–134, https://doi.org/ L: Axial length (m)
10.18280/mmep.070116. Lt: Total loss
[31] Jan F. Suhrmann, Dieter Peitsch, Marc Gugau, Tom Heuer, Uwe Tomm, Validation M: Mach number
and development of loss models for small size radial turbines, in: Turbo Expo: S: sensitivity index
Power for Land, Sea, and Air vol. 44021, 2010, pp. 1937–1949. U: Blade speed (m/s)
[32] Samuel M. Futral, Charles A. Wasserbauer, Off-Design Performance Prediction with vs: Blade speed ratio
Experimental Verification for a Radial-Inflow Turbine: Samuel M. Futral Jr. and W: Axial velocity (m/s)
Charles …, National Aeronautics and Space Administration, 1965.
[33] Charles A. Wasserbauer, Arthur J. Glassman, FORTRAN Program for Predicting off-
Design Performance of Radial-Inflow Turbines, 1975. Acronyms
[34] M. Jankowski, A new indicator for minimizing size of an ORC power plant based on CFD: Computational fluid dynamics
heat exchanger and turbine design parameters, Appl. Thermal Eng. 201 (2020), ATM: Automatic topology meshing
117750, https://doi.org/10.1016/j.applthermaleng.2021.117750. EoS: Equation of state
[35] Ya Zheng, Hu Dongshuai, Yue Cao, Yiping Dai, Preliminary design and off-design GGI: General grid interface
performance analysis of an organic Rankine cycle radial-inflow turbine based on HPT: High-pressure turbine
mathematic …, Appl. Therm. Eng. 112 (2017) 25–37. MPT: Medium pressure turbine
[36] P.A. Costa Rocha, H.H. Barbosa Rocha, F.O. Moura Carneiro, M.E. Vieira da Silva, LPT: Low-pressure turbine
A. Valente Bueno, k–w SST (shear stress transport) turbulence model calibration: a ORC: Organic Rankine cycle
case study on a small scale horizontal axis wind turbine, Energy 65 (2014) PS: Pressure surface
412–418. SST: Shear stress turbulence
[37] Manoj Kumar, R.K. Sahoo, S.K. Behera, Design and numerical investigation to RANS: Reynolds-Averaged Navier-Stokes
visualize the fluid flow and thermal characteristics of non-axisymmetric SS: Suction surface
convergent nozzle, Eng. Sci. Technol. Int. J. 22 (1) (2019) 294–312.
[38] Ayad M. Al, Raya K. Jubori, Saad Mahmoud Al-Dadah, Ahmed Daabo, Modelling
and parametric analysis of small-scale axial and radial-outflow turbines for organic Greek symbols
Rankine cycle applications, Appl. Energy 190 (2017) 981–996. ψ : Stage head coefficient
[39] Manoj Kumar, Debashis Panda, Ranjit K. Sahoo, Suraj K. Behera, Performance ρ: Fluid density (kg/m3)
prediction, numerical and experimental investigation to characterize the flow field ηts: Total-to-static efficiency
and thermal behavior of a cryogenic turboexpander, Heat Mass Transf. 56 (3) β: Absolute flow angle (degree)
(2020) 1015–1036. α: Absolute velocity angle (degree)
[40] Sangkyoung Lee, Hal Gurgenci, A comparison of three methodological approaches ηi: Isentropic efficiency
for meanline design of supercritical CO2 radial inflow turbines, Energy Convers. ω: Rotational speed (rpm)
Manag. 206 (2020), 112500. ϕ: Flow coefficient (− )
[41] Manoj Kumar, Debashis Panda, Ranjit K. Sahoo, Suraj K. Behera, Preliminary
design, flow field, and thermal performance analysis of a helium turboexpander: a
numerical approach, SN Appl. Sci. 1 (11) (2019) 1–30. Subscripts
[42] Manoj Kumar, Rasmikanti Biswal, Suraj K. Behera, Amitesh Kumar, Ranjit h: Hub
K. Sahoo, Experimental and numerical analysis to predict the performance of a sh: Shroud
turboexpander at cryogenic temperature, Eng. Rep. (2022) e12346. s: Isentropic
[43] Manoj Kumar, Ranjit K. Sahoo, Debashis Panda, Suraj Kr Behera, CFD analysis to Ti: Total
envisage the fluid flow inside a Turboexpander operating at cryogenic act: Actual
temperature, in: Advances in Air Conditioning and Refrigeration, Springer, i: First-order
Singapore, 2021, pp. 79–87. ij: Second-order
[44] Manoj Kumar, Pankaj Kumar, Ranjit Kumar Sahoo, Design and numerical analysis is: Isentropic
to visualize the fluid flow pattern inside cryogenic radial turbine, in: Advances in m: Meridional
Fluid and Thermal Engineering, Springer, Singapore, 2019, pp. 179–187. opt: Optimum
TEL: Trailing edge loss
x: Axial
r: Radial

17