You are on page 1of 30

Ocean Engineering 296 (2024) 117011

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

On the critical spacing between two in-line diamond cylinders at  = 100


Shravan Kumar Mishra, Subhankar Sen ∗
Department of Mechanical Engineering , Indian Institute of Technology (Indian School of Mines) Dhanbad, Dhanbad 826004, India

ARTICLE INFO ABSTRACT

Keywords: Results are presented for flow around two stationary in-line diamond cylinders at Reynolds number 100. The
Stabilized finite-element dimensionless centre-to-centre distance between cylinders is varied from 2 to 15. Discontinuities in drag plotted
Tandem arrangement against base suction are found to adequately represent the transitions. A jump discontinuity and subsequent
Diamond cylinder
trend reversal in drag indicate the first and second transitions, respectively. The first transition occurring at
Critical spacing
normalized critical spacing of 3.4 demarcates the regimes of weak vortex-shedding from the upstream cylinder
Vortex formation length
Base pressure recovery length
followed by alternate reattachment on the downstream cylinder (regime I) and co-shedding (regime II). Based
on transitions in wake mode, regime II is further sub-divided into two-layered vortex-shedding (TVS) and
secondary vortex-shedding (SVS) modes. In the reattachment regime, shear layer separating from the upstream
cylinder splits into three parts: innermost part rolls into a vortex attached to the cylinder, intermediate part
develops into an alleyway of fluids while exterior part reattaches on the forebody of downstream cylinder. The
concept of a new length scale, i.e. base pressure recovery length,  characterizing streamwise dimension of
mean wake, is introduced.  represents streamwise separation between base point and a point of recovery
of base pressure located along the wake axis.

1. Introduction the flow interference into five regimes, namely, single body flow,
alternate reattachment, quasi-steady reattachment, intermittent reat-
The flow around multiple objects is encountered in several engineer- tachment and co-shedding (simultaneous vortex-shedding from both
ing/industrial applications, such as, skyscrapers, cooling towers, trans- the cylinders) regimes.
mission lines, underwater pipelines (piggyback pipelines), etc. Around
Relative to that of an isolated cylinder, i.e.  → ∞, the study of twin
a fixed cylinder, another one can be positioned in unrestricted and 
cylinders becomes more intriguing given the associated complexities,
unlimited ways holding their axes parallel to each other (Zdravkovich,
1977). All such positional combinations can be generalized in tandem such as, diverse kinematics of flow structures, shear layer impingement
or in-line (Fig. 1), side-by-side or transverse and staggered or inclined phenomenon, formation and splitting of binary vortices, discontinuous
arrangements. In Fig. 1, diamond-section cylinders, i.e. square cylin- jump in fluid loading at the critical spacing,  | signifying transition to
 
ders at 45◦ incidence are shown that are considered in the current a different state of flow, admission of hysteresis and bistable solutions
investigation. Here,  represents the characteristic dimension of the in the neighbourhood of  | , suction pressure along whole of the rear
 
object and , the horizontal centre-to-centre spacing; the quantity cylinder for low  and consequently, its negative drag, etc. The critical

being referred to as the spacing ratio or gap ratio. The shortest 
 spacing corresponds to transition from reattachment to co-shedding or
surface-to-surface separation is denoted by  ′ . The cylinder facing the
binary vortex regime where intermittent switching between these two
upstream is designated as the upstream, leading or first cylinder and
the other one as downstream, rear or second cylinder. Leading edges types of flow occurs (Elhimer et al., 2016). In an early review work on
of the upstream and downstream cylinders are denoted by F1 and F2, flow past in-line circular cylinders, Zdravkovich (1977) emphasized on
respectively while G1 and G2 denote trailing edges/base points of the the bistable nature of flow at  | . Depending on the value of Reynolds
 
upstream and downstream cylinders. Thus, the inter-cylinder gap G1F2 number, , the gap flow for  
<  | may be steady or may exhibit
 
=  ′ . As compared to circular cylinders, the diamond cylinders find unsteady oscillations (Chatterjee and Mondal, 2012). Via experiments
much lesser applications in real environments. The cross-section of on flow past twin tandem circular cylinders in a low speed wind tunnel,
cooling towers and transmission lines is mostly circular while cylinders Okajima (1979) found that  | ceases to exist for  ≥ 3.8 × 105 . The
of diamond cross-section can be applied to the suspension rods of large-  
following sub-sections discuss the prior studies on flow around two
span bridges. For two in-line cylinders, Zdravkovich (1984) classified

∗ Corresponding author.
E-mail address: ssen@iitism.ac.in (S. Sen).

https://doi.org/10.1016/j.oceaneng.2024.117011
Received 9 October 2023; Received in revised form 11 January 2024; Accepted 1 February 2024
Available online 8 February 2024
0029-8018/© 2024 Elsevier Ltd. All rights reserved.
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

critical spacing, the shear layer reattachment could be alternating and


steady or simultaneous.
Using a streamline upwind Petrov–Galerkin formulation on hybrid
meshes, Jester and Kallinderis (2003) presented extensive results for
flow past circular cylinders in in-line, side-by-side and tandem arrange-
ments at  = 80 and 1000. By suitable choice of initial conditions,
they demonstrated hysteresis in mean drag coefficient,  as well as
bistable solutions at  = 1000. Via experiments in a closed-circuit
wind tunnel, Xu and Zhou (2004) investigated the dependence of
vortex-shedding frequency,  ( signifies the Strouhal number and
represents normalized frequency) based on streamwise velocity,  of
two in-line circular cylinders on  as well as  . Measurements were
Fig. 1. Schematic representation of flow past twin identical diamond cylinders in 
tandem arrangement with the shortest surface-to-surface separation being denoted by
conducted for  = 800−42000 and  
= 1−15. In addition to bistability
 ′ . Here,  signifies the circumferential angle measured counter-clockwise relative to involving stable reattachment and co-shedding (Igarashi, 1981) in the
the leading edge. gap, they resolved a new type of bistability aft the downstream cylinder
where co-existence of stable reattachment and roll up of shear layer
vortices was noted. For  
= 1 − 6 and  ≤ 120, Mizushima and
cylinders of circular and square shapes. With research gaps identified, Suehiro (2005) investigated the origin of discontinuities in fluid loading
research goals are set for the current study. and shedding frequency of in-line circular cylinders via linear stability
analysis and direct numerical simulations. Discontinuity in properties at
the critical gap was attributed to their solution multiplicity. Via finite-
1.1. Flow past two circular cylinders
volume computations, Sharman et al. (2005) analysed the flow past
two circular cylinders in tandem arrangement at  = 100 and a low
The early contributors to the analysis of flow around twin objects
blockage,  (ratio of  and width of domain) of 0.02. They varied
are Pannell et al. (1915) and Biermann and Herrnstein (1933). With
the normalized gap from 2 to 10 and predicted that  | lies between
focus on vortex-formation and shedding, Ishigai et al. (1972) experi-  
3.75 and 4. By performing finite-volume computations on triangular
mentally investigated the flow past two circular cylinders in tandem,
cells, Singha and Sinhamahapatra (2010) explored laminar flow past
staggered and side-by-side arrangements over  = 1500 − 15000. For
two circular cylinders in tandem arrangement at  = 40, 70, 100, 120
tandem arrangement, they determined  | = 3.8. For towed circular ′
  and 150. This study reveals that  | ≈ 3 at  = 100 and 120 and at
cylinder pair in tandem arrangement in a still-liquid tank, Tanida et al. ′
 = 150, the value of  | is slightly less than 3.
(1973) determined the values of  | at  = 80 and 3400 to be 5
  Alam and Zhou (2007) presented an empirical relationship (and its
and 3, respectively. For  = 8700 − 52000 in the sub-critical regime,
theoretical counterpart) involving vortex-shedding frequency of in-line
Igarashi (1981) experimentally investigated the characteristics of flow
circular and square cylinders, phase lag,  in vortex-shedding from
around identical in-line circular cylinders separated via  range of
 the cylinders and  . Dehkordi et al. (2011) performed finite-volume
1.03 − 5. The value of critical spacing was found to be 3.5 times the 
computations in two-dimensions to characterize the laminar (at  =
cylinder diameter. He resolved unstable flow for  = 1.18 and bistable
 100 and 200) and turbulent (at  = 22000) flows past two circular
flow for  = 3.09 and 3.53. The time interval of bistable solutions
 cylinders in tandem arrangement. For  varying from 1.2 to 10, Han
respectively corresponded to weak and strong fluctuations of surface 
et al. (2012) computed the flow past two in-line circular cylinders at
pressure,  at the leading edge of the rear cylinder. The quantity  ,
 = 200. Vu et al. (2016) numerically studied the flow past two
also known as pressure coefficient, is defined as  = 1 ∞2 where
−

2
 identical circular cylinders in tandem and side-by-side arrangements.
, ∞ ,  and  signify the static pressure, free-stream pressure, fluid For tandem cylinders separated via short distances, they noted low drag
density and free-stream speed. Zdravkovich (1987) identified the most on the downstream cylinder as an outcome of shielding effect of its
dominant interference regimes as: proximity interference, wake inter- upstream counterpart. With the objective of determining the critical
ference, proximity-wake interference and no interference. Interference spacing between two in-line circular cylinders at  ≈ 100, Yang and
flow regimes resolved for in-line circular cylinders are: single slender Stremler (2019) conducted experiments in a gravity-driven soap film
body, alternate reattachment, quasi-steady reattachment, intermittent system by steadily increasing and decreasing the inter-cylinder gap
shedding, jump discontinuity and binary vortex street. over  = 1.3 − 6.6 and holding  constant. With regard to hysteretic

Early numerical efforts to understand the flow around multiple solutions, they determined that  | = 2.9 − 4.1. Concerning flow past a
 
cylinders are due to Stansby (1981) and Chen et al. (1986). Li et al. pair of in-line circular cylinders at  = 100 and 150, Shan (2021)
(1991) used a velocity-pressure finite-element formulation with stag- found that the co-shedding regime can be further split into prime
gered arrangement of velocity and pressure and computed the flow past vortex-shedding (PVS), two-layer vortex-shedding (TVS) and secondary
an isolated and two in-line circular cylinders in two-dimensions. For vortex-shedding (SVS) sub-regimes on the basis of inter-cylinder gap.
two cylinder flow at  = 100 using  
= 2, 3, 4 and 6, vortex-shedding
from both the cylinders was noted only for  
> 3. In an early numerical 1.2. Flow past multiple square cylinders
effort, Mittal et al. (1997) employed a stabilized finite-element solver
and predicted the flow past stationary circular cylinders in tandem and Reinhold et al. (1977) measured fluid forces and moment in flow
staggered configurations at  = 100 and 1000. Meneghini et al. (2001) past two identical in-line square cylinders at various angles of incidence
employed a fractional step method on unstructured finite-element mesh including 45◦ . Lankadasu and Vengadesan (2008) conducted an early
and studied flow interference involving twin circular cylinders in tan- numerical experiment on flow past two square cylinders in tandem
dem and side-by-side configurations at  = 200. With  
varying from set-up at  = 100 to understand the effects of shear inlet as well
1.5 to 4 in tandem arrangement, they reported negative drag (thrust as inter-cylinder spacing. Yen et al. (2008) experimentally explored
or acting against the free-stream flow) on the downstream cylinder for topological aspects of flow past a pair of in-line square cylinders with


< 3. Alam et al. (2003) investigated the flow past two in-line circular angle of incidence of the rear cylinder varied from 0◦ to 45◦ . Bao et al.
cylinders in a low-speed closed circuit wind tunnel at a sole Reynolds (2012) employed a characteristic based finite-element formulation and
number of 65000 in subcritical regime. They noted that prior to the predicted the  = 100 flow past (a) an isolated square cylinder and

2
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

tandem arrays of identical square cylinders involving (b) two cylinders of a potential  | −  correspondence. The final part of the current
 
with  
= 1.25 − 11 and (c) six cylinders with  
= 1.5 − 15. For work is motivated primarily by an important observation concerning
case (b), they predicted  | = 4.5 − 4.75. For two in-line square possible dependence of the critical spacing on vortex formation length
 
cylinders, Chatterjee and Mondal (2012) predicted the critical spacing of a single cylinder by Yang and Stremler (2019) who state ‘It would
for different  and noted an inverse relationship between the critical seem reasonable for the transition from the reattachment regime to the co-
spacing and Reynolds number. shedding regime (and vice versa) to be caused by interference of the second
Zhao et al. (2016) used immersed boundary method and presented cylinder with vortex formation from the first cylinder, and thus for there to
results for flow past a pair of in-line square cylinders at  = 100 be a correlation between the critical spacing and the vortex formation length
over  = 2 − 7 so that the upstream cylinder was held fixed while of an isolated cylinder. King and Johns (1976) state that these distances
are comparable at corresponding Reynolds numbers. However, Papaioannou

angle of incidence of the downstream one was altered from 0◦ to 45◦ .
With increasing , they noted higher and lower extents of formation et al. (2006) show (see their Fig. 8) that the critical spacing is roughly
length for modes I (vortex-shedding only from the rear cylinder) and twice that of the recirculation region length (one possible definition of the
II (co-shedding), respectively. For  = 2 − 30, Kumar et al. (2019a) vortex formation length for a single cylinder). We suggest that explaining
this discrepancy is an open topic for future research.’ It may be added here

numerically explored the steady flow past two in-line square cylinders
at a sole Reynolds number of 40. Through finite-element computations, that Papaioannou et al. (2006) did not find any precise correspondence
Shui et al. (2021) investigated the far wake of a pair of identical square between vortex formation length of a circular cylinder and  | of two
 
cylinders in tandem set-up at a fixed  of 100. Depending on the in-line circular cylinders while both of these length scales were found
flow patterns resolved, six different modes of flow, i.e. single bluff to display similar type of variation with . Another objective of the
body, no vortex-shedding, shear layer reattachment, synchronization of current work is therefore to explore a possible correspondence of  |
 
vortex-shedding, two-layered vortices formation and secondary vortex with  ′ and  ′ . We also aim at understanding the implications of
formation were resolved over the considered  range of 1.5−9. Very relative magnitudes of  and  of square, circular and diamond

recently, Firdaus et al. (2023) numerically investigated the  = 3−150 cylinders. Finally, by compiling data for  | of circular and square
 
cylinders at various  from the literature,  ′ | −  and  ′ | − 

flow past a pair of in-line diamond cylinders separated via  = 0.5 − 6.  
This study analyzes the fluid loading and presents a flow regime map in relationships are investigated.

the  −  plane. Flow and heat transfer from twin square cylinders The remaining of the article is composed using the following se-
have been studied by Sohankar and Etminan (2009), Chatterjee and quence. Section 2 reviews the governing equations, depicts the problem
Mondal (2012) and Patel et al. (2018). undertaken and also includes a brief account of the finite-element
formulation used. Construction of mesh, convergence and accuracy
1.3. Research gaps and objectives of the predicted results are discussed in Section 3. Findings from
the present investigation are presented and analysed in Section 4.
Barring the recent study by Firdaus et al. (2023), the low - flow Concluding statements are recorded in Section 5.
around two stationary diamond-section cylinders in tandem arrange-
ment is mostly unexplored. In the current effort, focus is concentrated 2. Governing equations and finite-element formulation
on identification of various transitions, i.e. regimes of flow as the
spacing ratio is progressively increased from 2 to 15 holding the value 2.1. The incompressible flow equations
of  constant at 100. In absence of information on critical spacing
of two in-line diamond cylinders in the literature, our prime objec- Let us consider a spatial domain  in two-dimensional real number
tive is to determine  | via numerical flow visualization as well as space and let (0, ) signifies the time domain. The dimensional spatial
 
identification of discontinuities in characteristic flow quantities, as is coordinates are denoted by the vector  = (, ) while  denotes dimen-
conventionally done. Since the variation of  with base suction, − sional time. Let  ( , ) = (, ) and  , respectively stand for the fluid
velocity vector and stress tensor at a point. Using indicial notations
predicts flow transitions (Yadav et al., 2021), we examine if  | can
  for Cartesian tensors, strong or continuum form of the Navier–Stokes
be predicted from  -− criterion. Subsequently, we explore for a
equations of motion governing incompressible flow is expressed as:
possible relationship between  | and characteristic streamwise length
  ( )
of vortex formation of solitary circular, square and diamond cylinders; 
 
+   =

+  on  × (0, ), (1)
two different representations of length scale are used and a new length   
scale based on  is introduced. In the first one, the streamwise 
= 0 on  × (0, ). (2)
stretch,  of the mean recirculation regime or time-averaged wake 
bubble (see Fig. 5a) between base point of cylinder and wake stagnation In the stress-divergence form of momentum equations given by Eq. (1),
point is considered. As the second representation, following the leads  indicates the body force vector per unit volume. In this work,  is
of Bloor (1964), Bearman (1965), Griffin and Ramberg (1974) and zero-valued. For flow of a Newtonian fluid, the constitutive relation for
Williamson (1996), streamwise extent along the wake centreline,  stress at a point, as a sum of its isotropic and deviatoric components is
at which the fluctuating component of , i.e.  attains its peak and

 given by:
declines subsequently (see Figs. 2a and 5b), is used.  , similar to  , is ( )
1  
measured downstream from the base point. The wake length and vortex  = − + 2 ,  = + . (3)
formation length measured from the centre of cylinder are denoted by 2  
 ′ (indicated in Fig. 5a) and  ′ (indicated in Fig. 5b), respectively. Here  ,  and  , respectively denote the Kronecker delta, dynamic
The definition of  | is shown schematically via Fig. 2b. Fig. 2c
 
viscosity of the fluid and strain rate tensor. The conservation Eqs. (1)
reproduced from Figure 13 of Ljungkrona et al. (1991) plots the critical and (2) are solved simultaneously for unknown degrees-of-freedom 
spacing of twin circular cylinders as well as  over a wide range of and  in association with prescribed essential, natural/flux boundary
. Starting from a high value at low , the critical spacing displays a conditions and an initial condition on the velocity field. The initial
decaying wave-like variation with increasing . While the variations condition is expressed as:
of  | and  with  are quite analogous, the curves do not suggest
   ( , 0) = 0 on , (4)
for any precise relationship between  | and  . The 
 
| − relation-
 
ship rather appears to depend on discrete ranges of . Papaioannou so that 0 satisfies the zero-dilatation constraint as required by the
et al. (2006) and later, Yang and Stremler (2019) looked into this aspect incompressible continuity equation, i.e. Eq. (2).

3
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 2. Definition of (a)  and (b) 


|
 
via schematic presentation of flow past a single and a pair of in-line diamond cylinders, respectively. For unsteady flow past an isolated
circular cylinder, Figure (c) shows the variation of  | and  with  over a wide range of . This figure is a reproduction (with permission from Elsevier) of Figure 13 of
 
Ljungkrona et al. (1991) with some modifications in defining the symbols and variables.

Fig. 3. Illustration of the flow domain, dimensions and boundary conditions for analysis of laminar flow past a pair of in-line diamond cylinders.

A pair of identical diamond cylinders in tandem arrangement is formulation employed in this work, does not have an explicit equation
positioned at the mid-height of a rectangular-shaped computational for pressure and hence, does not require initial or boundary conditions
domain of length 125 and width 50 (Fig. 3). The value of width for pressure.  is varied by altering the location of the downstream
of the domain indicates that  = 0.02. The origin of the Cartesian cylinder. Irrespective of the value of , size of the computational
frame of reference is chosen so as to overlap with the centre of the domain and value of  are held fixed.
leading cylinder. Specified or essential no-slip and free-stream velocity
boundary conditions, respectively, are prescribed at the fluid-cylinder 2.2. The finite-element formulation
interfaces and inlet boundary of the domain that is located at an
upstream distance,  of 30 relative to the origin. The exit boundary The current work employs the velocity-pressure based monolithic
is assigned with traction-free natural boundary condition while free- or coupled stabilized finite-element formulation due to Tezduyar et al.
slip condition is imposed at the lateral boundaries. The monolithic (1992a,b). In this formulation, the spatial discretization is performed

4
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Table 1
Flow past a pair of diamond cylinders in tandem configuration at  = 100 and  = 0.02: summary of mesh convergence test conducted for 

= 4.
Mesh Nodes Elements Upstream cylinder Downstream cylinder
     0 −      0 −
M1 97 383 96 428 1.591 0.029 1.549 0.648 0.460 1.199 −1.008 0.454 0.063 0.370 1.310 0.939 −0.755 −0.583
M2 206 721 205 360 1.591 0.029 1.549 0.648 0.460 1.227 −1.015 0.452 0.062 0.368 1.309 0.937 −0.791 −0.559

Table 2
Flow past a pair of identical diamond cylinders in tandem set-up at  = 100 and a blockage of 0.02: illustration of effect of time step size on the results for 

= 4.
▵ ∗
Upstream cylinder Downstream cylinder
     0 −      0 −
0.01 1.590 0.029 1.549 0.648 0.461 1.193 −1.013 0.457 0.064 0.372 1.313 0.940 −0.773 −0.591
0.005 1.591 0.029 1.549 0.648 0.460 1.199 −1.008 0.454 0.063 0.370 1.310 0.939 −0.755 −0.583
0.0025 1.591 0.029 1.550 0.649 0.460 1.214 −1.011 0.453 0.062 0.370 1.310 0.938 −0.776 −0.586

via finite-element method while temporal discretization follows finite- Validation for single circular cylinder at  = 100
difference method. The time integration is performed using Crank– As part of validation, predicted  ,  ,  ,   and  of
 
 
Nicolson scheme. This formulation stabilizes the velocity and pressure an isolated circular cylinder at  = 100 and along the wake axis
fields via inclusion of streamline-upwind/Petrov–Galerkin (SUPG) and downstream from its base point are compared with those reported by
pressure-stabilizing/Petrov–Galerkin (PSPG) stabilization terms in the Qu et al. (2013). The current predictions correspond to a blockage of
weak form of momentum and continuity equations, respectively. Such 0.01 while Qu et al. (2013) used a blockage of 0.0083. Respective
inclusions preclude the non-physical oscillatory solutions of velocity comparisons are illustrated via Fig. 5a, b, c, d and e. Also compared is
and pressure, respectively. The PSPG stabilization allows implemen- the distribution of mean surface pressure (see Fig. 5f). Fig. 5 divulges
tation of identical interpolation for velocity and pressure. We opt an overall favourable comparison between the two sets of results. The
for bilinear interpolation functions for both velocity and pressure on first row of Fig. 5 indicates a satisfactory agreement of predicted 
bilinear quadrilateral elements used for domain discretization. In this and  with those obtained by Qu et al. (2013).
formulation, the variational form of Eqs. (1) and (2) is cast in a For  = 100 flow past a circular cylinder, Table 3 compares the
compact vector form that yields a single global matrix equation system fluctuating streamwise velocity and fluctuating pressure along the -
for the unknown nodal velocity and pressure. In each time step, the axis with those available in the literature. Also compared are  ,  ,
non-symmetric flow matrix is solved via non-stationary Generalized surface stagnation pressure coefficients and fluctuating lift. Overall, the
Minimal RESidual iterative solver of Saad and Schultz (1986). The comparison reveals close convergence of the predictions and results
finite-element formulation is extensively discussed in Tezduyar et al. from literature.
(1992a,b). Validation for two tandem cylinders at  = 100
Accuracy of results predicted by the current solver for flow in-
3. Domain discretization, mesh convergence and validation volving twin cylinders has already been established in Kumar et al.
(2019a). The  = 100 unsteady flow past pair of in-line circular and
The range of  
considered in this work is 2 − 15 and a total of square cylinders predicted by Kumar et al. (2019a) were compared
seven meshes, one each for  
= 2, 3, 3.1 − 3.39, 3.4, 4 − 7, 8,9 and with those documented by Sharman et al. (2005) for  = 4 and
10−15 has been considered that differ in number of nodes and elements.

Bao et al. (2012) for  = 5, respectively. Each of these comparisons
Construction of these meshes is identical; each mesh is structured and 
revealed excellent agreement for the integral parameters. Concerning
composed of ten non-uniform mesh blocks that are generated using flow around two in-line circular cylinders at  = 100, predicted mean
bilinear quadrilateral elements. For a representative  
of 2, Fig. 4a drag and vortex-shedding frequency of the cylinders for  = 0.02,  =3
illustrates the finite-element mesh that is composed of 93 173 nodes and 4 are compared with those obtained by Sharman et al. (2005) and

and 92 232 bilinear quadrilateral elements. Two internal square mesh Papaioannou et al. (2006). The comparison shown in Table 4 indicates
blocks having a constant edge length of 1.5 for each mesh, accom- a close agreement of the results.
modates the cylinders. As evident from close-up of the mesh in close As a further verification of accuracy, predicted fluid loading and
vicinity of the cylinders (see Fig. 4b), radial and circumferential grid vortex-shedding frequency for  = 100 and  = 0.02 flow past in-
lines are used to construct non-Cartesian mesh in two internal blocks. line circular cylinders separated by  = 5.5 in the co-shedding regime
In the remaining eight blocks, the mesh is Cartesian. 
are compared with those reported by Shan (2021) using a blockage of
As discussed above, meshes used for  = 4 − 6 are identical, they
 0.025. A close agreement between the two sets of results is obvious
consist of 97 383 nodes, 96 428 elements and are designated by the
from Table 5.
symbol M1. A mesh convergence test intended to explore insensitivity
of the predicted results to further mesh refinement, is conducted at
4. Results
 = 100 for  
= 4. Table 1 reports the results of mesh convergence
test and reveals that characteristic flow variables do not show any
4.1. The critical spacing
remarkable change when the flow is computed on a much refined mesh
M2 consisting of 206 721 nodes and 205 360 elements. In Table 1,
4.1.1. Prediction of  | from flow
0 indicates the mean forward stagnation pressure coefficient while  
subscripts ‘max’ and ‘rms’ indicate the maximum and root mean square The critical spacing corresponds to a transition from reattachment
values. to co-shedding regime and the transition is reflected through discon-
A further study is conducted to determine a suitable value of nor- tinuous jump/drop in characteristic flow quantities. For circular and
malized time step size, ▵ ∗ = ▵ below which solutions become square section cylinders, Table 8 indicates that  | ∈ [3, 7]; the upper
 
limit of  at being equal to 5. Thus, attention is focussed on

virtually independent of the size of time step. The flow is computed 
|   = 100
on M1 using ▵ ∗ = 0.01, 0.005 and 0.0025. Table 2 affirms that


range of 2 − 5 in determining the critical spacing of a pair of in-line
▵ ∗ = 0.005 is adequate to predict time step size independent solutions. diamond cylinders. For low  
, the gap flow is found to be unsteady

5
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 4. Laminar flow past two in-line diamond cylinders of the same size separated by 

= 2: (a) the finite-element mesh and (b) exaggerated view of mesh near the cylinders.

Table 3
Unsteady flow past an isolated circular cylinder at  = 100: comparison of predicted characteristic flow quantities with those reported in the literature. It may be recalled that
is measured from the base point while the location of maximum   is measured from the origin.


Study  along +ve  axis   along +ve  axis



 
0 −  

Maximum Maximum for maximum


 
 

Williamson and Roshko (1990) 0.70


Park et al. (1998) 0.01 1.03 0.74 0.23
Kravchenko et al. (1999) 0.0083 1.45 0.735
Sharman et al. (2005) 0.02 1.04 0.72 0.23
Posdziech and Grundmann (2007) 0.0125 0.7188 0.23
Qu et al. (2013) 0.0083 0.0913 2.4559 0.082 2.0981 1.42 0.707 0.2253
(case D9)
Chopra and Mittal (2019) 0.01 2.4820 1.4337 0.6910 0.2245
(extracted from graph)
Present 0.02 0.0931 2.4341 0.0877 2.0000 1.426 1.0966 0.6919 0.2303

with weak vortex-shedding from the leading cylinder while vortex- 



= 3.4 marks the reattachment→ co-shedding transition and hence,
shedding from the rear cylinder is strong for all  
. In particular for 
|
 
= 3.4.


= 2−3.3 or regime I (see the discussion in Section 4.1.2), flow around As stated in Section 1, several variants of reattachment regime were
the upstream cylinder is characterized by weak vortex-shedding from reported by Zdravkovich (1984). Further refinements of reattachment
it and associated alternate reattachment of shear layers separated from regime are due to Alam et al. (2003) who conducted wind tunnel
it and onto its downstream counterpart. A full-fledged vortex-shedding experiments involving in-line circular cylinders at a sub-critical  of
from the upstream cylinder, i.e. co-shedding or regime II commences 6.5×104 . They found that shear layer reattachment on the rear cylinder
when the gap ratio is increased to 3.4. The inception of co-shedding at is alternate for 

< 2 and steady or anti-symmetric for 2 ≤  
< 3; the

6
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 5. Unsteady flow past an isolated circular cylinder at  = 100: comparison of predicted distributions of (a) , (b) , (c) , (d)   and (e)  along the wake centreline
  
  
with those obtained by Qu et al. (2013) for 

= 0.5 (base point) to 8. Figure (f) compares the predicted mean surface pressure distribution ( − ) with those reported by Qu
et al. (2013). Figure (a) indicates  in −  curve while Figure (b) indicates  in curve. The results of Qu et al. (2013) for  = 0.0083 are shown by firm black
  
 
− 
curves whereas the current results for  = 0.01 are shown using red coloured circles.

value of  | being equal to 3. Regardless of steady or alternating nature


 
flow field at  
= 3.3 is analysed over a cycle of vortex-shedding.
of reattachment, the reattached shear layer was found to bifurcate into Fig. 6a illustrates time response of lift force on the downstream cylin-
two components, i.e. backward shear layer that travels downstream and der. The response is periodic with mono-frequency and the value of
forward shear layer that travels upstream (see the schematic of Fig. 7a). vortex-shedding frequency,  obtained via performing Fast Fourier
For in-line circular cylinders separated via  
= 3.7, such bifurcated Transform on this signal is 0.1162. The  is quite low relative to the
shear layers were resolved also by Elhimer et al. (2016) through standard value of 0.2 for a bluff object and hence, signifies weak vortex-
particle image velocimetry or PIV experiments at  of the order of shedding. However, the fully developed nature of  - 
profile (devoid
105 . To understand the reattachment phenomenon, the instantaneous of time intervals over which  = 0) indicates that vortex-shedding

7
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Table 4
Flow past a pair of identical circular cylinders in tandem configuration at  = 100, 

= 3 and 4: comparison of predicted mean drag and
vortex-shedding frequency with those reported by Sharman et al. (2005) and Papaioannou et al. (2006).
Study 

Upstream cylinder Downstream cylinder 

 
Sharman et al. (2005),  = 0.02 3 1.1105 −0.0238 0.1099
Papaioannou et al. (2006) 3 1.16 0.00 0.116
Present,  = 0.02 3 1.1296 −0.0140 0.1133
Sharman et al. (2005),  = 0.02 4 1.2740 0.7051 0.1483
Papaioannou et al. (2006) 4 1.31 0.75 0.152
Present,  = 0.02 4 1.2601 0.7013 0.1468

Table 5
Flow past a pair of identical circular cylinders in tandem configuration at  = 100 and 

= 5.5: comparison of predicted fluid loading and vortex-shedding frequency with those
reported by Shan (2021).
Study  Upstream cylinder Downstream cylinder 
         
Shan (2021) 0.025 1.2718 0.2602 0.3681 0.5676 0.9617 1.3389 0.1567
Present 0.02 1.2513 0.2542 0.3598 0.5539 0.9430 1.3105 0.1508

from the upstream cylinder is continuous and not intermittent. The shedding of an opposite signed vortex from the downstream cylinder
instantaneous flow is studied at seven instants of time of a shedding is also obvious from Fig. 6g. Finally, instantaneous surface vorticity
cycle of period  that are indicated in Fig. 6b. The reattachment is distribution (see Fig. 6h) clearly establishes the switching of reattach-
discussed with the aid of two representative cases corresponding to ment region from near the lower shoulder at ∗ = 3 6
to close to the
∗ = 36
and  . Fig. 6c, d and e, respectively illustrate the instantaneous upper shoulder at ∗ =  . The location of zero-vorticity points on the
streamlines, vorticity field of the cylinders and instantaneous surface surface of the downstream cylinder are: 1 = 77.31◦ , 2 = 271.59◦ and
vorticity of the downstream cylinder at ∗ = 3 6
. The streamlines 2 = 263.66◦ .
indicate the presence of twin vortices in the inter-cylinder gap. These The shear layer reattachment phenomenon discussed above for  =
vortices are separated by fluid alleyway that develops from part of the 100 flow differs from those of Alam et al. (2003) and Elhimer et al.
positive shear layer at the lower surface of the upstream cylinder. The (2016) noted at sub-critical Reynolds numbers. Unlike Alam et al.
counter-clockwise rotating vortex attached to the lower afterbody of (2003) (their Figs. 4 and 7; schematic in current Fig. 7a) and Elhimer
the upstream cylinder is an outcome of rolling of the same shear layer. et al. (2016) (their Fig. 16), the current study does not find a forward
Part of the positive shear layer reattaches on the lower forebody surface shear layer and growth of a gap vortex ahead of the downstream
of the downstream cylinder. A clockwise vortex shed previously from cylinder post reattachment. Besides, formation of an alleyway in be-
the upstream cylinder impinges on upper forebody of the downstream tween the attached and shed vortices of the upstream cylinder appears
cylinder and is mostly enveloped by the fluid alleyway. The shear layer to be a highly distinguishing feature. Such alleyways are visible in
forming at upper surface of the upstream cylinder is prevented by the Figs. 13 through 17 of Slaouti and Stansby (1992) who investigated the
fluid alleyway from reattaching on its downstream counterpart. Thus, flow past two in-line circular cylinders at  = 200 (Also see Fig. 5b of
shear layer reattachment on lower forebody of the rear cylinder occurs Deng et al. (2006) for  = 220 flow past twin in-line circular cylinders
at an instant when the counter-clockwise vortex attached on the lower and Figure 5b of Choi et al. (2012) for  = 170 flow past two in-
afterbody of the upstream cylinder continues to grow. The reattached line square cylinders). The topological departures in the reattachment
shear layer subsequently bifurcates into a backward shear layer as phenomenon are therefore attributed to Reynolds number effects rather
reported previously by Alam et al. (2003). Shedding of a clockwise than object cross-sections. Development of the shear layer separated
vortex from the downstream cylinder (see Fig. 6d) is also coincident from top surface of the upstream cylinder is depicted schematically in
with reattachment of the positive (or lower) shear layer. Concerning Fig. 7b for a representative time instant of ∗ =  .
the rear cylinder, synchronization of reattachment and vortex-shedding The vorticity field and streamlines for 
= 3.3 are shown in Fig. 8a
was earlier noted by Igarashi (1981) and Elhimer et al. (2016). In and c, respectively using a longer window of domain. The Karman
the instantaneous surface vorticity distribution on the rear cylinder vortex street aft the rear cylinder is obvious from Fig. 8a. Following
shown in Fig. 6e, circumferential angles of the zero-vorticity points of the terminology pioneered by Williamson and Roshko (1988) in context
separation and reattachment located close to the lower shoulder are of moving objects, the wake mode of the rear cylinder is 2S where S
indicated via symbols 1 and 1 , respectively. The backward shear signifies a single vortex. A minute increment of the spacing ratio sees
layer originates at 1 . From this figure, 1 = 88.3◦ and 1 = 95.18◦ , a drastic alteration in the flow; full-fledged Karman vortex-shedding
i.e. reattachment and separation from the lower end of the downstream from the upstream cylinder (Fig. 8c and d) ensues for the first time
cylinder occur from ahead of and behind the lower shoulder. Similarly, at a normalized spacing of 3.4. This is the  | of diamond cylinders
 
the attachment angle 2 corresponding to the attachment point located at  = 100 and signifies the onset of co-shedding regime. While the
close to the upper shoulder (or  = 270◦ ) is associated with the value short wake of the upstream cylinder is 2S, its downstream cylinder
of 282.78◦ . counterpart departs appreciably from 2S. Little downstream of near
A switch in location of reattachment occurs at the end of shedding wake of the rear cylinder, the shed vortices organize themselves in
cycle as evident from Fig. 6f. In this case, upper shear layer of the two distinct rows composed of like-sign vortices. Such wakes developed
upstream cylinder splits into three parts − the part immediately next via amalgamation were referred to as C(2S) wakes by Williamson and
to the cylinder shoulder rolls into a clockwise vortex attached to the Roshko (1988). The C(2S) mode generally develops in the wake of an
cylinder, an intermediate part (for which streamfunction,  ≠ 0) that excited cylinder (moving) during resonance. The streamlines of the rear
develops into an alleyway of fluid and the remote part reattaching cylinder (Fig. 8d) exhibit stronger waviness than those at  
= 3.3. This
at upper forebody of the rear cylinder. Thus, Fig. 6c and f ascertain feature is indicative of stronger vortex-shedding from the downstream
alternating nature of shear layer reattachment on the downstream cylinder and associated high lift force (see Table 6 for   ). Further
cylinder. Synchronization of reattachment of negative shear layer with insightful features of flow in the immediate neighbourhood of  |
 

8
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 6. Two-dimensional laminar flow past a pair of identical diamond cylinders in tandem arrangement at  = 100 and 

= 3.3: (a) time series of  of downstream cylinder,
(b) a vortex-shedding cycle for analysing instantaneous flow, instantaneous (c) streamlines, (d) vorticity and (e) rear cylinder’s surface vorticity at ∗ = 3
6
. Figures (f), (g) and (h)
show instantaneous streamlines, vorticity field and rear cylinder’s surface vorticity at ∗ =  .

are unveiled by the time-averaged streamlines at  


= 3.3 (Fig. 8e) for 

= 3.4 reveals drastic topological departures relative to those at
and 3.4 (Fig. 8f), respectively. The mean streamlines at  
= 3.3 

= 3.3. The upstream cylinder at  
= 3.4 possesses its own mean
are reminiscent of single body flow regime or extended body regime recirculation regime and a sub-wake or smaller wake within the main
resolved at low gap ratios in which both the cylinders are connected wake (Yadav et al., 2021). The mean flow of the rear cylinder involves
via zero streamlines bounding the gap flow and overall, act as an an anti-wake (Mishra et al., 2022) forming at the leading edge and a
elongated streamlined object exposed to free-stream. The mean flow pair of distinct separation bubbles on its afterbody. In context of steady

9
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 7. Flow past two in-line cylinders of the same size: streamline patterns proposed by (a) Alam et al. (2003) for closely spaced cylinders of circular cross-section at sub-critical
 and (b) the current study for diamond shaped cylinders at  = 100 and  
= 3.3. The schematic corresponds to ∗ =  . In Figures (a) and (b), r and s signify reattachment and
separation points, respectively.

Table 6
Flow past two in-line diamond cylinders at  = 100 and  = 0.02: summary of characteristic flow parameters for 

= 3.3, 3.4, 9 and 10.
Corresponding results for an isolated diamond cylinder are also included.
Cylinder 

     0 −
Upstream 3.3 0.1162 1.2895 0.0495 0.0004 1.1828 0.5790
3.4 0.1516 1.5577 0.4521 0.0335 1.2050 1.0081
Downstream 3.3 0.1162 −0.0256 0.3548 0.0256 −0.6993 0.3917
3.4 0.1516 0.5590 0.9962 0.1274 −1.0764 0.8887
Single →∞ 0.1773 1.7317 0.4720 0.0281 1.2602 1.2366
Upstream 9 0.1700 1.6539 0.4535 0.0272 1.2084 1.1135
10 0.1715 1.6839 0.4605 0.0276 1.2091 1.1537
Downstream 9 0.1700 0.4244 0.4173 0.0435 0.0479 0.3217
10 0.1715 0.5020 0.4338 0.0441 0.0640 0.4068

separated flow around a stand-alone diamond cylinder, Kumar et al. least vortex-shedding frequency at  
= 3.3 and a jump in shedding
(2018b) and Kumar et al. (2019b) resolved such distinct separation frequency at  = 3.4. Papaioannou et al. (2006) also reported the
bubbles along afterbody edges. Time responses of  of the cylinders

least value of  at the upper extremity of reattachment regime and
are shown in Fig. 8g for  
= 3.3 and Fig. 8h for 

= 3.4. These figures attributed it to poor generation of surface vorticity. The transition of 2S
indicate that vortex-shedding (or lift) from the cylinders is anti-phase wake to two-layered C(2S) wake, i.e. TVS mode at  = 3.4 is reflected
synchronized for  = 3.3 and in phase synchronized for 

 
= 3.4. by highly dense vorticity contours generated via amalgamation of like-
In phase synchronization of vortex-shedding at the critical gap is an sign vortices downstream of the rear cylinder. In the reattachment
essential trait of flow past two in-line objects (Alam and Zhou, 2007). regime, the mean streamwise velocity in the gap is negative (Fig. 9c)
The transition from reattachment to co-shedding state is evident consistent with the presence of gap recirculation. The onset of vortex-
also from the contours of instantaneous as well as mean flow field shedding at  = 3.4 corresponds to a shorter recirculation in the gap
variables. For  = 3.3 and 3.4 respectively, the first and second

 (Fig. 9h; also see Fig. 14i). Absence of mean recirculation aft the rear
columns of Fig. 9 illustrate the contours of mean vorticity, mean ,  cylinder for  = 3.3 and 3.4 both, is obvious from the contours of
and . In each case, symmetry is preserved about the -axis. For com-

. For  > 10 in co-shedding regime,  along and around the wake
pleteness of flow visualization, corresponding instantaneous vorticity centreline is much higher for 2S wake than TVS wake, implying that the
fields are shown again in Fig. 9a and f, respectively. A dramatic increase Karman vortices travel faster than the chain of amalgamated vortices.
in concentration of mean vorticity around the cylinders at  
= 3.4 The contours of mean transverse velocity reveal anti-symmetry about
(Fig. 9g) relative to those at  
= 3.3 (Fig. 9b) is consistent with the the wake centreline. For  
= 3.4, the contours of negative pressure

10
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 8. Two-dimensional laminar flow past a pair of identical diamond cylinders in tandem arrangement at  = 100: the first column shows (a) instantaneous vorticity, (c)
instantaneous streamlines, (e) mean streamlines and (g)  - 

variation of the cylinders for 

= 3.3. The second column shows the respective plots in Figures (b), (d), (f) and (h)
for 

= 3.4.

(Fig. 9j) are densely concentrated around the cylinders than those for slope of  in  -− plane is 42.09◦ . An approximate linear  -−


= 3.3 (Fig. 9e). At the onset of co-shedding regime, a strong suction relationship is apparent also from the unsteady part of Figure 14b of
therefore prevails in the gap and also around the cylinders, leading Yadav et al. (2021) in context of flow around a diamond cylinder. In
to higher drag on the cylinders allied to an abrupt jump in  (see Figure 14b of Yadav et al. (2021), increase in the value of  is reflected
Table 6). by simultaneous increase in  as well as − . The  -− data of
Yadav et al. (2021) for  = 50 − 150 are used to obtain a line of best
4.1.2. Prediction of transitions and flow regimes from  -− relationship fit given by
Base suction coefficient is a highly sensitive flow parameter the
variation of which reflects the transitions in a flow (Roshko, 1993).  = 1.4325 + 0.2878(− ). (6)
Concerning flow around a solitary diamond cylinder for  ≤ 150,
Yadav et al. (2021) found that characteristic flow quantities, such as  The slope of this line is 16.06◦ . For twin cylinders, the sensitivity
and its components,  , etc when plotted against − , reveal an ob- of − is found to reflect discontinuous changes in characteristic
vious trend change at the onset of vortex-shedding. Papaioannou et al. quantities (for instance, drag coefficient) at the critical spacing. For
(2006) for mean flow around an isolated circular cylinder (Equation


= 2 − 15, Fig. 10a and b, respectively depict the  -− relationship
3.1 of their paper) and Roshko (1993) for a vertical flat plate with an of the upstream and downstream cylinders.
attached splitter plate, presented linear empirical relationships linking The  -− relationship of the cylinders differs drastically, both
 and − . Eq. (1) of Roshko (1993) is expressed as in terms of range of variables and nature of curves. As the spacing
ratio is increased progressively from 2, jump discontinuity (indicated
 = 0.88(1 + 0.11 ) −  = 0.88 + 0.9032(− ). (5)
via dotted lines) in  as well as − of the cylinders occurs at
In consistency with Eq. (5), Figure 4 of Roshko (1993) indicates that the normalized critical spacing of 3.4 that corresponds to the first
 increases linearly with increasing − and vice versa. Besides, the transition. Thus, the flow around a pair of in-line diamond cylinders

11
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 9. Two-dimensional laminar flow past a pair of identical diamond cylinders in tandem arrangement at  = 100: contours of (a) instantaneous vorticity, (b) mean vorticity,
(c) mean streamwise velocity, (d) mean transverse velocity and (e) mean pressure in the reattachment regime corresponding to  
= 3.3. The second column shows respective
contours in Figures (f), (g), (h), (i) and (j) for 

= 3.4 corresponding to the co-shedding regime.

Fig. 10. Two-dimensional laminar flow past two in-line identical diamond cylinders at  = 100: relationship between  and − of the (a) upstream and (b) downstream
cylinders.

at  = 100 can be classified into two broad regimes, i.e. regime I (for leads to an increase in − as well as  of the upstream cylinder (sim-


= 2 − 3.3) characterized by weak continuous vortex-shedding and ilar to an isolated cylinder). In contrary, in the reattachment regime,
alternate reattachment and regime II (for 

= 3.4 − 15) associated with both  and − decrease as the spacing is increased. However, in both
co-shedding phenomenon. In the co-shedding regime, an increase in  
cases, slope of the  -− curve is 41.02◦ ; quite different from the

12
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 11. Two-dimensional laminar flow past a pair of identical diamond cylinders in tandem arrangement at  = 100: contours of (a) instantaneous vorticity, (b) mean vorticity,
(c) mean streamwise velocity, (d) mean transverse velocity and (e) mean pressure at  
= 9 corresponding to the TVS mode. The second column shows respective contours in
Figures (f), (g), (h), (i) and (j) for 

= 10 corresponding to the SVS mode.

one of an isolated diamond cylinder but close to the one predicted by 2021). Interestingly, the PVS mode detected by Shan (2021) for in-line
Roshko (1993) for a vertical flat plate. The linear  -− relationship circular cylinder pair appears to be absent for the diamond counterpart.
with increasing  noted by Roshko (1993) and Yadav et al. (2021) The non-linear  -− profile of the rear cylinder in regime I and TVS
and currently observed linear  -− relationship with increasing  segment of regime II indicate that the rear cylinder is under strong
≤ 9. The spacings over which 

indicate that the role played by  for an isolated cylinder is equivalent influence of the leading cylinder for 

to that of 
for tandem cylinders at fixed . The fact that hysteresis varies linearly with − , represent the regions along which solutions
can be resolved by altering either  or  
(Zdravkovich, 1987) also are closer to those of an isolated cylinder than the ones associated
indirectly supports the equivalence of  (at fixed  
) and 
(at fixed with non-linear  -− relationship. For the rear cylinder, spacings
). Concerning a pair of square cylinders in tandem set-up, hysteresis with SVS mode provide such solutions. The second transition is hardly
was investigated by holding  constant and varying  
(Liu and Chen, recognizable from the  -− curve of the upstream cylinder. This is
2002) and alternately, fixing  and varying  (Choi et al., 2012). a natural outcome of the absence of a far wake of this cylinder.

Based on this equivalence, we infer that linearity of  -− in the For TVS (at  
= 9) and SVS (at  
= 10) modes, respectively
co-shedding regime is a testimony of the upstream cylinder behaving the first and second columns of Fig. 11 show the contours of field
in a manner similar to that of an isolated cylinder for  
≥ 3.4. A variables, such as, instantaneous and mean vorticity, ,  and . For
kink in  -− curve of the downstream cylinder (and not of the the TVS mode, existence of a pair of distinct shear layers aft the rear
upstream one) at  = 9 indicates a second transition in the flow cylinder is obvious from the contours of instantaneous vorticity shown
in Fig. 11a. Wake transition associated with the SVS mode is obvious

in which base suction and mean drag both increase with increasing
spacing. This transition is found to reflect alterations in wake modes from Fig. 11b. For both modes, the regions of twin layers of vorticity are
of the rear cylinder. Therefore, the co-shedding regime or regime II identified with close packing of mean vorticity contours (see Fig. 11b
of the rear cylinder can be further split into a couple of sub-regimes for TVS and 11g for SVS modes); in case of SVS mode, the contours
spanning over 3.4 ≤  
≤ 9 and 9 <  
≤ 15, respectively. In the are less dense for single row of vortices. In addition,  attains small
first sub-regime, two-layered arrangement of shed vortices aft the rear positive values over  
associated with twin layers of shed vortices
cylinder corresponds to the TVS mode discussed by Shan (2021). In this (Fig. 11c for TVS and 11h for SVS modes). The contours of  (Fig. 11d
regime,  exhibits least sensitivity to changes in − . The second for TVS and 11i for SVS modes) are analogous around the leading
sub-regime is characterized by a brief regime of two-layered vortices cylinder while they reveal marked differences downstream of the rear
in the near wake followed by a single row of large-sized vortices of one. Red coloured pressure contours near leading edge of both the
alternate sign. This regime is associated with the SVS mode (Shan, cylinders (Fig. 11e and j) are indicative of positive pressure ( < 0)

13
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 12. Two-dimensional laminar flow past a pair of identical diamond cylinders in tandem arrangement at  = 100: (a)   − 

relationship of the upstream and downstream
cylinders and (b) ratio of the maximum lift of the downstream cylinder to that of the upstream cylinder.

Fig. 13. Two-dimensional laminar flow past a pair of identical diamond cylinders in tandem arrangement at  = 100: distribution of mean surface pressure of the upstream and
downstream cylinders for 

= (a) 3.3 and (b) 3.4. For comparison,  −  distribution of an isolated diamond cylinder is also included.

directed towards the cylinder surface. The highly concentrated suction pressure at cylinder shoulders. According to Arie et al. (1983), weak
pressure contours aft the upstream cylinder for  
= 9 and 10 both, correlation between surface pressure at the shoulders for  < | lead
suggest sharp gradient of pressure in the wake and consequent high to low   . Over 
= 2 − 8,   or alternately, amplitude of lift of
drag on the cylinder. In contrary, suction pressure contours in the wake the downstream cylinder is higher than those of the upstream cylinder
of downstream cylinder are distant/less dense and hence, suggestive and a trend reversal is noted thereafter. Thus, for  
= 2 − 8, vortex-
of weak pressure gradient. Under this circumstance, the drag is low. shedding from the downstream cylinder is stronger than its upstream
The value of pressure aft the upstream cylinder for both the spacing counterpart. Li et al. (1991) suggest that higher asymmetry of flow
ratios is found to be much smaller (more negative) than those aft (about  axis) ahead the downstream cylinder than that of the upstream
the downstream one. This corroborates with the noteworthy difference cylinder is reflected in stronger lift force of the former than the latter.
in drag force acting on the upstream and downstream cylinders (see   of the downstream cylinder increases with  for  <  | , attains
 
Table 6). its peak value of 0.9962 at  =  | corresponding to an abrupt jump,
 
subsequently decays linearly up to  
| = 9 and remains fairly constant
4.1.3. Variation of fluctuating lift with spacing thereafter. A very close peak   value of 0.9855 (extracted from
At a fixed  , the twin cylinder system possesses a single vortex- Figure 14d of Sharman et al. (2005)) at  | = 4,  = 100 is recorded
  
shedding frequency. However, owing to the interference effect, strength by the downstream cylinder of a twin tandem circular cylinder system.
of shedding from the cylinders differs and accordingly, r.m.s. lift of the The relationship between the amplitude of lift and inter-cylinder gap is
cylinders do not overlap at all (see Fig. 12a). Below and above the illustrated via Fig. 12b. Up to  = 9, the profiles of   curves are

critical spacing,   of the upstream cylinder exhibits insignificant similar to respective   curves and afterwards, a departure in trend
sensitivity to ; the   beyond  | being close in magnitude to that
 
is observed where   of the rear cylinder continues to marginally
(= 0.4717 for  = 0.02) of a single diamond cylinder. The insignificant surpass those of the leading cylinder. As observed by Alam et al. (2003)
  of the upstream cylinder below critical spacing is an outcome earlier, in-phase vortex-shedding from the cylinders or in-phase of the
of lack of organized vortex-shedding from it (Arie et al., 1983). In lift coefficients leads to the maximum value of the fluctuating lift of the
context of vortex-induced vibrations of an elliptic cylinder, Kumar et al. upstream cylinder. The minimum lift on the other hand, is associated
(2018a) illustrated the correspondence between r.m.s. lift and surface with anti-phase vortex-shedding or lift force. In the reattachment zone,

14
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

anti-phase vortex-shedding from the cylinders or anti-phase lift of the (Fig. 13b),  − profile of the upstream cylinder undergoes noteworthy
cylinders relate to insignificant   or   of the leading cylinder. modifications around the base region and also along the forebody.
This however, does not stand valid in whole of the co-shedding regime; Convex downward nature of mean pressure at the base is consistent
the magnitude of   or   at higher  appears to be insensitive to with the appearance of a sub-wake aft the cylinder. Besides, location

the phase of flow around the cylinders. For instance, the valley in   of the maximum pressure of the downstream cylinder shifts upstream
or   of the leading cylinder around  = 6 stems from anti-phase of its shoulder and hence surface pressure in the forebody region

vortex-shedding from the cylinders. However, no further drop is noticed undergoes an initial rise (associated with adverse pressure gradient)
in lift despite anti-phase shedding at normalized gaps of 9 and beyond. followed by a fall (associated with favourable pressure gradient) till the
The amplification in the lift of the rear cylinder due to shear layer shoulder is reached. The transition of pressure gradient from adverse to
impingement from the leading cylinder is represented via the quantity favourable at the location of maximum pressure indicates the presence
  , i.e. ratio of the amplitude of lift of the rear to the leading cylinder. of a pair of attachment points on the forebody of the downstream
The vortex impingement phenomenon induces an extra unsteady force cylinder. These attachment points are associated with the anti-wake.
on the rear cylinder that pulsates with the frequency of vortex-shedding Apart from these alterations, location of minimum pressure shifts from
of the downstream cylinder (Shan, 2021). For  < 9, the impingement the leading edge to the shoulder.
For  = 3.4, Fig. 14 illustrates time variation of surface pressure

effect is dominant and hence,   > 1. Prior to the critical spacing, 
constancy of   for the upstream cylinder and linear increase for of the cylinders at stagnation points (Fig. 14a and e for  = 0◦ ,
the downstream cylinder renders a rapid increase in   . At  | ,   Fig. 14d and h for  = 180◦ ), lower shoulder (Fig. 14b and f) and at an
 
drops sharply followed by an approximate linear decay up to  | = 8. angular position of 174.86◦ (Fig. 14c and g) close to the rear stagnation

The decay of   to unity is indicative of the diminishing effect of shear point. Each of the time traces reveals periodicity; time period being
layer impingement as the critical spacing is exceeded. For  ≥ 9, the smaller at the stagnation points than at the rest of locations. For
 the upstream cylinder, amplitude of  is maximum at the shoulder
closeness of   to unity implies that vortex-shedding from the leading
(Fig. 14b) whereas for the downstream one, maximum amplitude is
cylinder does no longer influence the vortex-shedding from the rear
noted at the forward stagnation point (Fig. 14e). Formation of an anti-
cylinder.
wake at leading edge F2 of the rear cylinder (Fig. 8f) is a testimony of
the existence of a highly decelerated flow in that region. Compared to
4.1.4. Mean and fluctuating surface pressure
the reattachment regime at  = 3.3, an extremely low velocity close
Vortex-shedding alters the instantaneous surface pressure and hence 
to the leading edge F2 in the co-shedding regime is apparent also from
highly dissimilar flow patterns at  = 3.3 and 3.4 are expected to ren-
 the variation of  along the gap G1F2 (Fig. 14i). In the co-shedding
der contrasting distribution of surface pressure at these spacing ratios.
regime, advent of a tiny sub-wake aft the upstream cylinder is indicated
Fig. 13a and b, respectively illustrate the mean surface pressure along
by positive  just downstream of the base point (shown in close-up).
lower half surface of the cylinders at  = 3.3 and 3.4, respectively.
 In line with the principle of Bernoulli, the magnitude of  is high at
For comparison,  of an isolated diamond cylinder is also shown. The and around the forward stagnation point of the rear cylinder. This is
range of  of the isolated cylinder is wider than those of each of obvious from  −  variation along the inter-cylinder gap (Fig. 14j).
the in-line cylinders. This is indicative of higher pressure drag of an The variations of mean streamwise velocity and mean pressure
isolated cylinder than those of the upstream and downstream cylinders along the gap G1F2 provide subtle information on the number of
at 
= 3.3 and 3.4. recirculation regimes accommodated within G1F2. The number of 
 −  profile of the upstream cylinder in general resembles the one crossings, i.e. attainment of the same value of  in  -  variation
of the isolated cylinder while a significant divergence is visible for the or alternately, zero-crossings of  in -  variation matches the number
downstream cylinder. For  
= 3.3 and 3.4 both, mean pressure of the of stagnation points existing in G1F2 excluding the terminal stagnation
rear cylinder is negative or suction-type. In context of flow around an points G1 and F2. Absence of a  crossing for  = 3.3 (Fig. 14j) is

isolated diamond cylinder, Yadav et al. (2021) showed that declining consistent with the absence of zero-crossing of  and hence, absence
surface pressure close to the base involves a favourable pressure gra- of a stagnation point (or a mean wake) in G1F2. For  
= 3.4 in
dient and presence of a recirculation. They reported the presence of a contrary, three zero-crossings of  and an equal number of  crossings
sub-wake in association with convex downwards  −  variation close affirm the presence of three stagnation points within the gap, i.e. points
to the base region. Convex downwards pressure at the base indicates M (for sub-wake; see the first inset of Fig. 14i), M (for main or
the presence of reattachment points either due to a sub-wake or a primary wake) and M (for anti-wake); each stagnation point being
separation bubble. At  
= 3.3 (Fig. 13a),  −  curve of the upstream equivalent to wake stagnation point of respective wake (see Fig. 8f).
cylinder is concave upwards at the base. This implies continuation of The corresponding  crossing points are identified as R , R and
adverse pressure gradient from near the lower shoulder ( = 90◦ ) of R , respectively (Fig. 14j). We identify the distance between base
the cylinder and absence of a sub-wake or reattachment (see Fig. 8e). point G1 and base pressure recovery point R as a characteristic
In contrary, for the downstream cylinder, slightly convex downward  streamwise length scale of the mean wake. This length scale, introduced
at the base hints towards the existence of a pair of adjacent attachment as the ‘base pressure recovery length’ and denoted by  , exhibits a
points. These attachment points are associated with separation bubbles strong influence on the location of wake stagnation point M of the
forming along the afterbody edges of this cylinder. The maximum and mean wake (see Section 4.1.5 for details). It is found that the wake
minimum pressures of the rear cylinder are obtained near the shoulder stagnation point appears downstream of the recovery point, yielding
and at the leading edge, respectively. The abrupt drop in  near the  <  . Thus,  serves as the least estimate of  . For  =

lower shoulder of the rear cylinder is indicative of the presence of a 3.4, Fig. 14k and l, respectively compare 0 of an isolated diamond
brief regime of favourable pressure gradient close to the shoulder. For cylinder with those of the upstream and downstream ones. 0 of the
low gaps, occurrence of maximum surface pressure on the downstream downstream cylinder departs from that of an isolated cylinder due
cylinder at the point of reattachment of shear layers separated from the to interference whereas 0 of the upstream cylinder is modified via
upstream cylinder was earlier reported by Ishigai et al. (1972), Igarashi feedback effect of its downstream counterpart.
(1981), Li et al. (1991), Alam et al. (2003) and Shan (2021), etc. For For  = 3.3, fluctuating or r.m.s. pressure along the lower half sur-



= 3.3, presence of a local pressure peak at the lower shoulder of face of the upstream and downstream cylinders are plotted in Fig. 15a
the rear cylinder indicates reattachment of shear layers separated from and b, respectively. Fig. 15c and d illustrate the respective quantities at
the upstream cylinder on the downstream cylinder (also see Igarashi the critical spacing of 3.4. In consistency with the findings of Igarashi
(1981) close to  = 90◦ . At the normalized critical spacing of 3.4 (1981) at sub-critical Reynolds numbers,   of the upstream cylinder

15
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 14. Two-dimensional laminar flow past a pair of identical diamond cylinders in tandem arrangement at  = 100: time traces of  for 

= 3.4 at  = (a) 0◦ , (b) 90◦ , (c)
174.86◦ and (d) 180◦ of the upstream cylinder. Figures (e), (f), (g) and (h) show respective time traces for the downstream cylinder. For 

= 3.3 and 3.4, Figures (i) and (j) plot
 and  , respectively along the gap G1F2. The deviation of 0 of the upstream and downstream cylinders from that of an isolated diamond cylinder is illustrated via Figures (k)
and (l), respectively.

in the reattachment regime (Fig. 15a) is extremely low in magnitude. of separation. The fluctuating pressure of the rear cylinder (Fig. 15b)
From the weak fluctuations of  , one can predict very low magnitudes is an order of magnitude higher than those of the leading cylinder.
of   and   of the upstream cylinder in the reattachment regime.   −  profile of the downstream cylinder differs from that of its
Besides, a sharp peak or horn appears in   close to the lower upstream counterpart in that the former is associated with a pair of
shoulder or lower separation point (also see Alam et al. (2003)), consis- peaks; a relatively flat peak in the forebody region ( < 90◦ ) and a
tent with high amplitude oscillations in  (Fig. 15b). As evident from sharp peak just next to the shoulder and along the afterbody surface.
Fig. 8e (for  = 3.3) and f (for  = 3.4), the mean separation angle These peaks respectively correspond to shear layer reattachment on
 
of the upstream cylinder is about 90◦ . The occurrence of maximum the rear cylinder and separation of backward shear layer from it,
fluctuating pressure at the location of separation points of a circular respectively. In co-shedding regime (Fig. 15d), in absence of shear
cylinder was earlier established by Batham (1973) through experiments layer reattachment on the rear cylinder, the first peak is no longer
at  ≈ 1.1 × 105 . For flow around a fixed circular cylinder at  = realizable while the second one continues to represent the separation
4.1 × 104 , Norberg (1986) found that out of two maxima in   −  of backward shear layer. The r.m.s. pressure of the upstream cylinder
distribution, one appears approximately at  = 77◦ around the location for 
= 3.4 (Fig. 15c) is one order of magnitude higher than its

16
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 15. Two-dimensional laminar flow past a pair of identical diamond cylinders in tandem arrangement at  = 100: the distribution of fluctuating surface pressure of the
upstream cylinder for 

= (a) 3.3, (c) 3.4 and of the downstream cylinder for (b) 

= 3.3, (d) 3.4.

counterpart at  
= 3.3. Irrespective of spacing, extremely small   the leading edge of the rear cylinder (Fig. 16c showing the associated
at the leading edge ( = 0◦ ) of the upstream cylinder stems from streamline plot) leaves a profound impact on the magnitude of  
the approaching undisturbed free-stream flow (Qu et al., 2013). The at  = 0◦ . The higher velocity of flow around this region (compared
dramatic rise in   of both the cylinders at  
= 3.4 is attributed to to  = 3.4) is responsible for relatively weaker fluctuations in  and

the transition phenomenon. Vortices shed from the upstream cylinder consequent low magnitude   at  = 9 and 10. Near the leading
enhance the fluctuations in  along the forebody of the downstream

edge (  = 8.5) of the rear cylinder, the mean streamwise velocity is of
cylinder leading to large   as compared to the afterbody surface. 
high magnitude followed by a steep fall while the mean pressure also
The more number of   peaks resolved by Igarashi (1981) and
exhibits a declining trend (Fig. 16d). A pair of zero-crossings of  at
Alam et al. (2003) at subcritical  are attributed to Reynolds number
M (shown in inset) and M coupled with a pair of  crossings
effects where a backward shear layer develops on the rear cylinder. The
at R and R is evident from this figure. The first zero-crossing of
generation of alleyway at  = 100 prevents the formation of backward
 at M corresponds to a sub-wake behind the upstream cylinder (see
shear layer and hence   of the upstream cylinder shows only one
the inset) while the second zero-crossing at M is associated with the
prominent peak while the downstream one shows two (in regime I)
main wake. In consistency with the case of  = 3.4, G1M >G1R ,
or one peak (in regime II). Fig. 15 indicates that regardless of flow 
regime,   |=0◦ <   |=180◦ for the upstream cylinder whereas for i.e.  >  . Beyond a cylinder width downstream of the leading
the downstream cylinder, an opposite trend is noted. cylinder,  continues to diverge from  and hence, a third 
Fig. 16a compares   of the upstream cylinder for  = 9 and 10. crossing does not substantiate. This affirms the absence of anti-wake

The r.m.s. surface pressures virtually overlap indicating very similar ahead of the rear cylinder. For  
= 9, Fig. 16e plots  as well as 
values of fluctuating fluid forces on the cylinder at these gap ratios. aft the rear cylinder along the -axis. A single zero-crossing of  is
In a similar vein,   of the downstream cylinder reveals a close consistent with the appearance of a mean wake and absence of its sub-
comparison for  = 9 and 10 (16b). Form here, one can predict very wake. The mean pressure curve intersects the -axis twice in extremely

similar values of   (as well as   ) of the rear cylinder at these close vicinity of base point of the rear cylinder and does not aid in the
spacings. Comparison with   −  distribution of in-line circular prediction of relevant  . Similar observation is made at the highest
cylinders reported by Sharman et al. (2005) for  = 100 and  = 10 spacing ratio of 16 as well (not shown). Thus, in context of flow around

indicate that fluctuating pressure in a diamond cylinder pair is much twin in-line diamond cylinders, we could define  solely for the
stronger than those of a circular cylinder pair. Absence of anti-wake at upstream cylinder.

17
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 16. Two-dimensional laminar flow past a pair of identical diamond cylinders in tandem set-up at  = 100: distribution of   of the (a) upstream and (b) downstream
cylinders for 

= 9 and 10. For comparison, respective   of two in-line circular cylinders reported by Sharman et al. (2005) for 

= 10 are also included. Figure (c) confirms
absence of anti-wake of the rear cylinder at 

= 9. The  − 

and  − 

variations (d) in the gap G1F2 and (e) downstream of the rear cylinder correspond to 

= 9.

For 
= 3.3 and 3.4 both, Fast Fourier Transform of  is carried out upstream cylinder (Fig. 17a) results from modified  due to feedback
at certain locations along lower surface of the cylinders. Corresponding effect of the rear cylinder. A weak second harmonic is also seen for
frequency distributions are shown in Fig. 17a (  
= 3.3, upstream the downstream cylinder. At  = 3.4, the general trend of frequency

cylinder), b (  = 3.3, downstream cylinder), c ( 
= 3.4, upstream distribution of the cylinders does not undergo any significant alteration.
 
cylinder) and d (  = 3.4, downstream cylinder). When the upstream The magnitudes of  and  both increase to 0.1509 and 0.3041,
respectively with  = 2 relationship being preserved. For the

and downstream cylinders are separated via  = 3.3, Fast Fourier

Transform of  at several points excluding the stagnation points, downstream cylinder, the second harmonic becomes more prominent
reveals the existence of a pair of peaks; a dominant peak corresponding than those at 
= 3.3.
to  =  = 0.1160 and its harmonic with  = 0.2319 = 2 . The diminishing feedback effect on the upstream cylinder (and
It is found that the harmonic frequency matches the frequency of  . hence, behaviour more like an isolated cylinder) becomes apparent
In contrast,  at the stagnation points is associated with a single from the relatively clean spectral content at  
= 9 (Fig. 18a) and 10
dominant frequency (shown in red), the value of which matches the (Fig. 18c). Each spectrum is identified mostly with a single dominant
frequency of  (also see Yadav et al. (2021) for a single diamond cylin- peak. For these cases, the wake mode within the gap is 2S. At  
= 9,
der). The harmonic frequency is indicative of alternate reattachment i.e. at the upper extremity of TVS mode, a weak harmonic of vortex-
phenomenon (Alam et al., 2003). For flow past an isolated diamond shedding frequency appears that matches the frequency of  of the
cylinder at  = 100, Figure 10d of Yadav et al. (2021) shows the rear cylinder (Fig. 18c). At  
= 10 corresponding to the SVS mode,
existence of a single dominant peak in Fast Fourier Transform of  the power spectra turn to be the most complex (Fig. 18d) with an
at the shoulders. The dominant peak in  of the downstream cylinder additional sub-harmonic frequency appearing along the afterbody half
at its shoulder therefore corresponds to the vortex-shedding frequency of the rear cylinder. A very interesting observation from the power
whereas the minor harmonic peak (Fig. 17b) is an outcome of alter- spectra (Figs. 17 and 18) at spacing ratios bracketing the transitions
nate reattachment phenomenon. The corresponding harmonic in the suggest that the frequency of  at the stagnation points represent the

18
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 17. Two-dimensional laminar flow past a pair of identical diamond cylinders in tandem arrangement at  = 100: power spectra of  at different angular locations along
the lower surface of the upstream cylinder for 

= (a) 3.3, (c) 3.4 and along the lower surface of the downstream cylinder for 

= (b) 3.3, (d) 3.4.

frequency of drag. On the other hand, barring this set of diametrically transition.  being a primary unknown quantity, serves better than
opposite points,  at all other surface points yield the vortex-shedding  (drag is a quantity derived by integrating local  ) in locating the
frequency as the fundamental frequency. transitions.
The values of certain characteristic flow quantities are listed in
Table 6 for spacing ratios bracketing the transitions. For reference, data 4.1.5. Base pressure recovery length
for an isolated diamond cylinder are also included. The first transition For Stokes flow between a pair of counter-rotating cylinders posi-
occurring at  = 3.4 is associated with abrupt jump in magnitude tioned in-line, Gaskell et al. (1998) analytically predicted flow struc-
tures and saddle-stagnation points in the gap region. The results were

of characteristic quantities of both the cylinders. In consistency with
the trend reported in the literature, drag of the rear cylinder in reat- further supported with experimental and finite-element simulated re-
tachment zone, i.e. at  = 3.3 is negative or upstream directed. The sults. Here, we focus on four-way saddle-stagnation points in the gap

mean drag of a cylinder is proportional to the difference of line integral at  = 100. To understand the role of  crossings in determining
of mean surface pressure between its forebody and afterbody. The locations of four-way stagnation points,  along the lower surface
pressure distribution shown in Fig. 13a indicates that the average of of the leading cylinder and gap G1F2 is combined at the junction,
mean pressure along the lower forebody of the downstream cylinder i.e. base point G1 and shown in Fig. 19a for a representative  
of
is little higher (less negative) than that along its lower afterbody. 3.4 belonging to the co-shedding regime. Coupled with geometrically
This elucidates the low magnitude thrust force on the rear cylinder symmetric location of base point, dominance of base pressure (or base
in the reattachment regime. Owing to shielding effect of the upstream suction) along the rear half of the cylinder and early part of its wake
cylinder and consequent weak generation of vorticity at the cylinder renders it an ideal pressure to analyse the downstream flow along the
walls, vortex-shedding frequency of the cylinders attains its least value wake axis. As one travels downstream from point G1, four  crossings
at the upper extremity of reattachment regime (Papaioannou et al., denoted by points R, R , R and R are identified. Also identified
2006). This is reflected by the low magnitude of  at  = 3.3 are five alternate crests (and peaks) of local minimum (and maximum)

of  and denoted by points A1, A2, A3, A4 and A5, respectively. Points
as compared to that at  = 3.4. In the first transition, anti-phase
 R and A1 are indicated in the inset of Fig. 19a. The regions G1A1,
vortex-shedding ( = 105.91◦ ) corresponding to  = 3.3 turns in-phase
 A2A3 and A4A5 along which  decays, correspond to favourable
( = 5.74◦ ) at 
= 3.4. However, in the second transition occurring pressure gradient while regions A1A2 and A3A4 of adverse pressure
between  
= 9 and 10, the shedding is anti-phase, i.e.  = 112.65 ◦ at
gradient are linked to enhancing  . It may be noted that regions


= 9 and  = 188.08 at  = 10. The value of  at a given gap ratio
◦ 
A2A3, A3A4 and A4A5 accommodate  crossing points R , R
has been obtained from  signals of the cylinders. Unlike in the first and R , respectively. The relative locations of stagnation points and
transition, flow parameters for the upstream cylinder do not exhibit  crossings in the gap as a function of governing pressure gradient is
any noteworthy jump during the second transition. However, for the illuminated schematically by Fig. 19b. The mean streamwise velocity
downstream one, − and consequently,  reveal sensitivity to the (in blue colour) and recirculation zones are also shown to facilitate the

19
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 18. Two-dimensional laminar flow past a pair of identical diamond cylinders in tandem arrangement at  = 100: power spectra of  at different angular locations along
the lower surface of the upstream cylinder for 

= (a) 9, (c) 10 and along the lower surface of the downstream cylinder for 

= (b) 9, (d) 10.

identification of stagnation points. For favourable pressure gradient, it intervals of time, regime I dominates while over some other intervals,
is found that separatrices emanating from near the shoulders converge regime II acts as the primary solution. Thus, a bistable solution admits
at the four-way saddle point M located along the  axis next to two stable solutions at a constant spacing ratio, see for example,
point R . For adverse pressure gradient in contrary, two stream- Figure 2 of Elhimer et al. (2016). Figure 7e of Zhao et al. (2013)
lines emanate from four-way saddle points M and M that are concerning free vibrations of a square cylinder at 45◦ incidence also
located upstream of points R and R , respectively. The favourable reveals bistability in  . To resolve bistability, the flow is computed at
pressure gradients along G1A1 and A4A5 might contribute to the certain  , such as 3.32, 3.33, 3.34 and 3.39 between 3.3 and  | = 3.4.
  
reattachment of sub-wake and anti-wake streamlines on the leading and Fig. 20a, b, c and d depict the time response of lift of the upstream
rear cylinders, respectively. The (four-way saddle point,  crossing) cylinder for  = 3.32, 3.33, 3.34 and 3.39, respectively. Corresponding

pairs appear alternately as (M , R ), (R , M ) and (M , time traces of  of the downstream cylinder are shown in Fig. 20e,
R ) along the wake centreline. For three such pairs, the  crossing f, g and h, respectively. For each case, time trace reveals periodicity.
of the second pair (R , M ) provides the base pressure recovery For  = 3.32, starting from zero value, the lift attains its saturation
length, i.e. G1R =  . The directions of streamlines constructing

amplitude and preserves it as time progresses. The constancy of  
the four-way saddle points M , M and M within the fluid confirms that the flow corresponds to a single stable state belonging to
medium are apparent from Fig. 19c, d and e, respectively; an alternate regime I (the associated streamlines and vorticity contours are shown).
sequence of directions is noted. At a slightly higher  of 3.33, the time response reveals a stark
 ′
For  = 3.4 − 15, Table 7 lists the values of  and base pressure

 departure; the response of fluid forces displays dual nature of flow. At
early time, the flow corresponds to regime I and as time progresses, a
 ′
recovery length,  measured from the centre of the cylinder. For
each spacing ratio, it is found that  ′ <  ′ , i.e. the location of wake transition from regime I to regime II takes place. While both the states
stagnation point of mean wake of the upstream cylinder is preceded by are stable, regime II persists for sufficiently large time and a reverse
location of recovery of base pressure along the wake axis. transition to regime I is not observed. Absence of continuous and mu-
tual intermittent switching between the reattachment and co-shedding
4.2. The transitional regimes regimes at  
= 3.33 does not render the associated flow truly bistable.
The unsaturated nature of the amplitude of  in reattachment regime
In context of flow around twin in-line cylinders, bistability in the is apparent from inset (i) for upstream and inset (ii) for downstream
neighbourhood of  | is associated with sustained competition be- cylinder. At  of 3.34, regimes I and II co-exist with shortened stretch
  
tween two modes, i.e. reattachment and co-shedding. Over certain of the reattachment zone. At  
= 3.39 immediately below   
| , the

20
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 19. Two-dimensional laminar flow past a pair of identical diamond cylinders in tandem arrangement at  = 100 and  
= 3.4: (a) distribution of  along lower half of the
upstream cylinder and gap G1F2, (b) schematic presentation of zero streamlines and critical points of the upstream cylinder and gap flow, illustration of streamline directions for
four-way saddle points (c) M , (d) M and (e) M .

Fig. 20. Two-dimensional laminar flow past a pair of identical diamond cylinders in tandem arrangement at  = 100: time series of  of the upstream cylinder for 

= (a)
3.32, (b) 3.33, (c) 3.34 and (d) 3.39. Corresponding time traces for the downstream cylinder are shown in Figures (e), (f), (g) and (h), respectively. For 
= 3.33, close-up of
time traces in reattachment regime of the upstream and downstream cylinders are shown in Sub-figures (i) and (ii), respectively. The instantaneous streamlines and vorticity field
shown reveal reattachment nature of flow for  
= 3.32.

21
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Table 7
Two-dimensional laminar flow past a pair of identical diamond cylinders in tandem configuration at  = 100 in the
 ′  ′
co-shedding regime: summary of  , 
and 
.


Upstream cylinder Downstream cylinder
 ′  ′ 
  
 

3.4 −1.0081 1.6025 1.4498 −0.8887 –


4 −1.0083 1.4364 1.3234 −0.5833 –
5 −1.0142 1.4136 1.3843 −0.4610 0.5395
6 −1.0310 1.4403 1.4221 −0.4108 1.1006
7 −1.0638 1.4393 1.4076 −0.3720 1.4335
8 −1.0982 1.4279 1.3885 −0.3432 1.9596
9 −1.1135 1.4242 1.3916 −0.3217 2.5260
10 −1.1537 1.4278 1.3805 −0.4068 1.7523
11 −1.1592 1.4200 1.4085 −0.4293 1.5345
12 −1.1743 1.4157 1.3973 −0.4451 1.4689
13 −1.1829 1.4175 1.3818 −0.4574 1.4476
14 −1.1913 1.4105 1.3846 −0.4718 1.4592
15 −1.1908 1.4153 1.3912 −0.4827 1.4609

reattachment regime further narrows down significantly before the flow cylinder pairs are compiled from the literature and listed in Table 8. For
evolving fully to the co-shedding regime at  = 3.4. The co-existence comparison, the currently predicted  | data for a diamond cylinder
 

of regimes I and II over  = 3.32 − 3.39 indicates that flow bistability pair is also included. While the range of Reynolds number extends
from 50 to 1000, the normalized critical spacing lies between 3 and

exists at certain spacing ratios close to 3.4. Partial co-existence of two
different flow states are apparent also from Figure 8 of Meneghini et al. 7. As obvious from Table 8, for fixed  of 100 and fixed blockage of
(2001) where they showed time variation of fluid forces on two in-line 0.02, the least critical spacing is associated with the diamond-cylinder
circular cylinders at  = 200. Co-existence of stable reattachment as system, followed by circular and square cylinder pairs, respectively.
well as co-shedding regimes are obvious also from Figure 2 of Borazjani The variable location of separation points of a circular cylinder as
and Sotiropoulos (2009) where they presented time trace of drag of the opposed to fixed separation points of a square cylinder renders the
downstream cylinder of a twin fixed in-line circular cylinder system at flow past a circular cylinder array more prone to turn unstable than
 = 200 for  = 4. For ∗ up to ≈430, drag force on the rear cylinder its square counterpart (Bao et al., 2012). Thus, when  is held fixed,

is negative and gap flow quasi-steady as revealed by vorticity field. At co-shedding phenomenon in tandem circular cylinder pair initiates at
∗ > 430, a sharp rise is observed in  in association with vortex- a lower gap ratio than a square cylinder pair.
The placement of downstream cylinder inhibits the development of
shedding from both of the cylinders. In addition, flow around the rear
vortices, i.e. the vortex formation process from the upstream cylinder.
cylinder is not time-periodic.
The characteristic length of wake along the flow direction, i.e. vortex
The attainment of fully developed co-shedding regime via transition
formation length of a single cylinder therefore turns out to be a key
of an initial regime of reattachment is explored for  = 3.33 (Fig. 21).
 parameter governing the extent of critical spacing (Yang and Stremler,
Fig. 21a for the upstream cylinder and Fig. 21b for the downstream
2019). Roughly, the mean recirculation length of an object may also
cylinder, depict the time traces of  . The initial reattachment regime
represent the formation length. As highlighted in Yang and Stremler
and subsequent co-shedding regime are indicated in the figures. Over-
(2019) in context of flow past twin in-line circular cylinders, King
all, time traces of drag of the cylinders appear quite similar to each
and Johns (1976) reported similar magnitudes of  | and  whereas

other barring the higher fluctuations in the reattachment zone of the  
Figure 8 of Papaioannou et al. (2006) indicates that  | ≈ 2  over

rear cylinder than the leading one. Besides,  of the rear cylinder is  
 = 100 − 350.
negative in the reattachment regime similar to that found by Borazjani
The variation of  along the wake axis and downstream of the
and Sotiropoulos (2009) for circular cylinders in tandem arrangement.
base point shown in Fig. 22a for circular, square and diamond cylin-
The reattachment and co-shedding natures of flow at early and late
ders provides a quantitative comparison of associated  values at
phases of time are also revealed by the  −  curves (Fig. 21c and d,
 = 100 and  = 0.02. The onset of steady laminar boundary layer
respectively) as well as instantaneous vorticity fields (Fig. 21e and f, re-
separation occurs at a critical Reynolds number, commonly referred to
spectively). The power spectra of  corresponding to the reattachment
as the separation Reynolds number. For square, circular and diamond
(Fig. 21g for upstream cylinder and Fig. 21h for downstream cylinder)
cylinders, the values of separation Reynolds number are 1.15 (Sen
and co-shedding regimes (Fig. 21i for upstream cylinder and Fig. 21j
et al., 2011), 6.19 (Sen et al., 2009) and 7.3 (Kumar et al., 2018b),
for downstream cylinder) reveal single dominant frequency of vortex- respectively. Thus, at an  above the separation Reynolds number,
shedding in each case. The Fast Fourier Transform of  stretching expected sequence on the magnitude of  is  | >  | >
over the reattachment and co-shedding regimes reveals the co-existence  | . As per the expectation, the largest and smallest values of 
of the low and high frequency peaks for the upstream (Fig. 21k) and at  = 100 are found to be associated with the square and diamond
downstream (Fig. 21l) cylinders. cylinders, respectively;  of the circular cylinder residing in between
(also see Table 4 of Sourav et al. (2020)). Owing to the extra cross-
4.3. Comparison of critical spacing of circular, square and diamond cylin- sectional area of a square cylinder than a circular cylinder of the same
ders characteristic dimension, the solid mass containing fixed separation
points of a square cylinder is protruded further in the fluid medium.
The locations of separation of shear layer from the upstream cylin- Consequently, a fluid particle at the point of separation of a square
der and reattachment on its downstream counterpart display a strong cylinder travels with a wider angle of trajectory than with a circular
bearing on cylinder cross-section and contour. Accordingly, the value of cylinder and results in a bigger recirculation zone (Shui et al., 2021).
critical spacing for tandem cylinders of common geometric shapes, such Unlike a square cylinder, the corners of a diamond cylinder are located
as, circular, square and diamond, is expected to differ under identical along the  and  axes. Besides, its area of cross-section is smaller
flow conditions. For a wide range of Reynolds number, i.e.  = 50 − than those of circular and square ones. These factors contribute to less
2.5×105 , Figure 3 of Yang and Stremler (2019) plots the critical spacing deflection of separation streamlines of a diamond cylinder and lead to
of a circular cylinder. The  | data for in-line circular and square
 
the formation of a small recirculation zone.

22
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 21. Two-dimensional laminar flow past a pair of identical diamond cylinders in tandem arrangement at  = 100 and 

= 3.33: time series of  of the (a) upstream and (b)
downstream cylinders, (c) and (d)  and (e) and (f) instantaneous vorticity in reattachment and co-shedding regimes, respectively, power spectra of  of the upstream cylinder
in (g) reattachment, (i) co-shedding, (k) combined regimes and of the downstream cylinder in (h) reattachment, (j) co-shedding, (l) combined regimes.

Table 8
Summary of the values of critical gap ratio for pairs of in-line circular, square and diamond-section cylinders reported in the literature.
Study Geometry   
|
 

Huhe-Aode et al. (1985) Circular – 100 4.5−5


300 3.5−4
1000 3−3.5
Li et al. (1991) Circular – 100 3−4
Slaouti and Stansby (1992) Circular – 200 3.5−4
Sharman et al. (2005) Circular 0.02 100 3.75−4
Deng et al. (2006) Circular 0.0833 220 3.5−4
Ayyappan and Vengadesan (2008) Circular 0.0625 100 4.25−4.75
Han et al. (2012) Circular 0.025 200 ≈3.6
Vu et al. (2016) Circular 0.05 60 5.5−6
100 4−4.2
200 3.9−4
1000 3−3.2
Bao et al. (2012) Square 0.02 100 4.5−4.75
Chatterjee and Mondal (2012) Square 0.05 50 5−7
75 5−7
100 4−5
125 3−4
150 3−4
Shui et al. (2021) Square 0.0556 100 4.4−4.5
Present Diamond 0.02 100 3.4

23
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 22. Two-dimensional laminar flow past isolated cylinders of square, circular and diamond cross-sections at  = 100 using  = 0.02: variation of (a) and (b) along the
 
 
wake centreline downstream from the base point.

Table 9
Flow past isolated square, circular and diamond-section cylinders at  = 100 using  = 0.02: summary of mean wake length, vortex
formation length and ratio of vortex formation to mean wake length. The last row of this table lists the  | for each cross-section.
 

Square Circular Diamond


1.942 1.426 0.923


3.224 2.434 1.234




1.6601 1.7069 1.3369





|
 
4.5−4.75 (Bao et al., 2012) 3.75−4 (Sharman et al., 2005) 3.4

Fig. 22b showing the distribution of  along wake axis, divulges cylinder is placed in-line, it interacts with flow in the wake of the up-
that the sequence of magnitude of  of the cylinders is identical to stream cylinder and imparts a stabilizing effect (Zhang and Su, 2021),
those of associated  , i.e.  | >  | >  | . Besides, i.e. obstructs the formation of a fully developed wake. In a hypothetical
the value of  of a given cross-section at  = 100 surpasses  . The situation, where the mean gap recirculation reaches the downstream
proportionality of  | with  reported by King and Johns (1976) and cylinder, the cylinders and gap flow act as a single extended solid
 
Papaioannou et al. (2006) ascertains that (  | ) > ( | ) > object (indicated via blue colour) defined by the surface streamline
     
(  | ) , very much in consistency with the trend indicated by
  = 0. The rear cylinder acts as a splitter plate (Roshko, 1955) and
Table 8. By compiling data from the literature, Yadav et al. (2021) obstructs/delays the onset of vortex-shedding from the leading cylin-
found that  | >  | >  | where  signifies the der. Owing to the largest extent of wake length, square cylinder wake
is the most stabilized due to interference from a downstream cylinder
critical Reynolds number marking the onset of primary wake instability
while the diamond cylinder wake is the least affected. An uninhibited
or laminar vortex-shedding from an isolated cylinder. The sequence of
growth of the diamond cylinder wake is therefore conducive for an
magnitude changes in  ,  and  of square, circular and diamond
early transition of flow state as compared to the circular and square
cylinders is alike while the sequence reverses for separation Reynolds
ones. The hypothetical single extended body illustrated in Fig. 23a
number. The normalized values of  ,  and   ratio of the cylinders

 ascertains that for a given object, the critical spacing must exceed the
corresponding to  = 100 are listed in Table 9. The value of

ratio of length of mean recirculation at the same , i.e.  | >  .

  
each cylinder surpasses unity so that

| >

| >

| ; For a representative gap ratio of 3 pertaining to regime I, Fig. 23b
     
compares the extent of recirculation within the gap at  = 100. The
this trend reveals an obvious departure from those followed by  and cylinder contours as well as bounding streamlines correspond to  = 0.
 . The last row of Table 9 affirms that irrespective of cross-section, Since the recirculation zone is unable to elongate along the streamwise
>  ,  . direction, it is expected that stronger the recirculation, higher is the
  
|
 
The vortex-shedding frequency and width of near wake share an extent,  of recirculation across the streamwise direction. The trajec-
inverse relationship (Zdravkovich, 1987). As noted by Roshko (1954), tory of separation streamlines is governed by the location of separation
higher the bluffness, lower the vortex-shedding frequency and short- points on the cylinders. The widest wake is associated with diamond
ening of the formation region boosts the vortex-shedding frequency. cylinder, followed by square and circular ones. The order of  of the
Accordingly, at a given , the magnitude of vortex-shedding frequency cylinders does not precisely match with the (  | )
  
> ( | )
  
>
of an isolated diamond cylinder is expected to be the highest, fol- (  | ) sequence or its reverse. Considering the upstream cylinder,

lowed by those of circular and square cylinders, respectively. Close  | = 225.81◦ <  | = 244.81◦ <  | = 268.72◦ and
proximity of shear layers and consequently, their enhanced interactions this is precisely in reverse of the (  | ) > ( | ) > ( | )
        
shorten the period or enhance the frequency of vortex-shedding (Ger- sequence. Thus, closeness of separation points of the upstream cylinder
rard, 1966). Thus, under identical conditions, the diamond cylinder is (or an isolated cylinder) to its shoulders, i.e.  = 90◦ or 270◦ appears to
more prone to vigorous vortex-shedding than its circular and square be more stronger a criterion than a larger  to yield a low  | . From
 
counterparts. A representation of the mean wakes is shown schemat- here we infer that  (or equivalently,  ) and  strongly govern the
ically in Fig. 23a for square (Sub-figure i), circular (Sub-figure ii) extent of critical spacing; the critical spacing is low if  (or  ) is low
and diamond (Sub-figure iii) cross-sections. Thus, when a downstream and  high.

24
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 23. Flow past a pair of identical in-line cylinders of square, circular and diamond shapes: (a) (i) schematic presentation of a hypothetical situation when the upstream and
downstream square cylinders along with the gap flow form an extended body, corresponding zero streamlines for the mean flow of (ii) circular and (iii) diamond cylinder pairs.
Figure (b) illustrates (by zero streamlines) the relative positions of mean separation-attachment points of the cylinder pairs separated via 

= 3 for  = 100 flow.

Fig. 24. Two-dimensional laminar flow past a pair of identical square cylinders in tandem arrangement at  = 100 and 

= 3: instantaneous (a) streamlines and (b) vorticity.

Table 10
Two-dimensional laminar flow past a pair of identical cylinders in tandem set-up at  = 100 and a blockage of 0.02: summary of characteristic flow quantities for square, circular
and diamond shaped cylinders for 
= 3.
Geometry Upstream cylinder Downstream cylinder
     0 −       0 − 
Square 0.0000 1.2739 0.0000 1.0571 0.4303 0.0000 0.0000 −0.0965 0.0000 −0.4000 0.2294 0.0000
Circle 0.00014 1.1296 0.0123 1.0841 0.4597 0.1118 0.00179 −0.0140 0.0985 −0.4195 0.2611 0.1118
Diamond 0.00126 1.3058 0.0465 1.1820 0.5991 0.1181 0.0220 −0.0868 0.3149 −0.7074 0.3774 0.1181

For 
= 3, Table 10 lists the values of characteristic flow quantities  = 60 − 1000, over most of which the flow turns three-dimensional.
of in-line cylinder pairs at  = 100 and a  = 0.02. For square- Thus, depending on the value of  and associated two- or three-
section cylinders, the flow is steady in nature and accordingly,   , dimensional flow, corresponding values of  | and  available in
 
  and  are zero-valued. This feature is apparent also from Fig- the literature are used. Figure 2 of Roshko (1993) reproduced here as
ures 3b, 3c and 4 of Bao et al. (2012). Fig. 24a and b depict the Fig. 25a illustrates the relationship between − and  of a circular
associated instantaneous streamlines and vorticity, respectively. The cylinder for  ≤ 1000. We have inserted some texts in the original
closed and anti-symmetric streamlines in the gap region as well as figure of Roshko (1993). This figure aids in locating the onset of vortex-
aft the downstream cylinder further confirm that the flow is steady. shedding as well as transition of flow from two- to three-dimensional.
Besides, shear layers separated from the cylinders (Fig. 24b) appear The onset of vortex-shedding at  corresponds to the first occurrence
straight and devoid of waviness as in a steady flow. The streamlines of divergence of − obtained from steady and unsteady solutions.
and vorticity field are very similar to those presented by Kumar et al. Beyond  , the mean base suction increases with  till the upper
(2019a) in context of steady flow past a pair of in-line square cylinders limit of two-dimensional flow. At the transition from two- to three-
at  = 40 and 2 ≤  
< 5 (see their Figure 4a). For the circular cylinder dimensional flow at  = 180 (Williamson, 1988), − exhibits a
pair in contrast, the flow is unsteady and fluctuating fluid forces are prominent drop of low magnitude, rises thereafter until a peak is
small. Fluctuating fluid forces of diamond cylinders are one order of reached at  ≈ 260 (Williamson, 1996) and decays slowly thereafter
magnitude higher compared to those of respective circular cylinders. A till  ≈ 1000. These events are obvious also from Fig. 25a. In the
highly noteworthy observation from Table 10 for tandem cylinders is unsteady vortex-shedding regime, the length of mean recirculation
that for each characteristic quantity but for the mean drag, the smallest exhibits an inverse relationship with base suction (Figure 4 of Roshko
and largest magnitudes relate to the square (largest  | ) and diamond
 
(1993); Figure 14a of Yadav et al. (2021)). This is evident from Fig. 25b
(smallest  | ) cylinders, respectively, in reverse sequence for  ,  that is inspired by Figure 6 of Jiang and Cheng (2017) for flow around
 
and  of isolated cylinders. a circular cylinder. The three-dimensional direct numerical simulation
results of Jiang and Cheng (2017) resolves a sharp rise in  ′ at the
4.4. Relationship between critical spacing and vortex formation length  corresponding to transition to three-dimensionality. For evaluating
 ′
the relationship between  | and  , i.e. the  ′ | ratio, we use two-
  
Finally, we investigate for a possible relationship between the criti- dimensional  ′ data of Jiang and Cheng (2017) for  < 200 and their
cal spacing and characteristic length of vortex-formation ( and  ) three-dimensional data for  > 200.
of circular, square and diamond-section cylinders. Concerning a pair For a circular cylinder, results available in the literature for normal-
of circular cylinders in tandem, the data listed in Table 8 span over ized critical spacing at  = 60, 100, 160, 250, 300, 500 and 1000 are

25
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 25. Flow past a circular cylinder for  ≤ 1000: the variation of (a) − (reproduced from Roshko (1993) with permission from Elsevier) and (b)  ′ with . The schematic
shown in Figure (b) is inspired by Figure 6 of Jiang and Cheng (2017).

considered. As indicated in Fig. 26, the  | at  = 60, 100 and 300 are
 
 ′ data are taken from Bai and Alam (2018). Similar to the circular
compiled from Vu et al. (2016), Sharman et al. (2005) and Huhe-Aode cylinder, the  ′ | ratio in general, increases with . For a diamond
et al. (1985), respectively whereas the data at  = 160, 250, 500 and

cylinder, the presently computed data at  = 100 yield 
| =
1000 are taken from Papaioannou et al. (2006). It may be noted that  ′ 
3.4
= 2.39 and 
| = 3.4
= 1.96. Figs. 26 and 27 establish that a
each of these studies provide a range of  | rather than a single 
 
| .
 
1.423  ′  1.734
The subscripts l, u and m appearing in the figure respectively signify the definitive relationship between the characteristic streamwise dimension
lower, upper and average values of the ratio. As apparent from Fig. 26, of vortex-formation of a single cylinder and critical spacing of two in-

| is not a simple multiplier of  ′ . The  ′ | ratio in contrary
  line cylinders does not exist. This is also evident from Figure 13 of
displays a strong dependence on ; the value of the ratio essentially

Ljungkrona et al. (1991) that is reproduced in this work as Fig. 2c. The
exceeding unity. For the upper as well as lower limits of  | , the loci
 

| or  ′ | ratio in general depends on . For a circular cylinder,
 ′ 
of  ′ | ratios form a band (also see Papaioannou et al. (2006)). The
 


monotonic decay of  | with  and that of  ′ for  above ≈ 300
 
mean 
|
 ′ 
line passing through midway of the band does not indicate is consistent with a decaying trend of the  ′ | with  with  ′ |
surpassing unity. Based on Fig. 2c concerning circular cross-section, we
 
any specific 
|
 ′ 
−  relationship. The mean 
|
 ′ 
is rather found
conjecture that regardless of cross-section,  ′ | must exceed unity till

to be a piecewise continuous function of  whose slope continues to 

alter over different ranges of . In particular,  ′ | ratio increases  ≈ 2 × 105 , i.e. the shear layer transition regime (Williamson, 1996).

non-monotonically till  = 300 (where  ′ is minimum for the 
5. Conclusions
considered, see Fig. 25b) and subsequently, continues to decay. The
apparent  | ≈ 2 relationship reported in Yang and Stremler (2019)
  With primary focus on identification of transitions, the present work
in relation to Figure 8 of Papaioannou et al. (2006) for  = 100 − 350
numerically explores the unsteady flow past two identical diamond
does not deem valid. The observation by Yang and Stremler (2019) is
cylinders in tandem arrangement at a fixed Reynolds number of 100
based on two-dimensional data of Papaioannou et al. (2006) whereas
by varying the normalized centre-to-centre separation from 2 to 15.
Fig. 25b clearly indicates the onset of three-dimensionality at  < 200
With the aid of field variables as well as variation of mean drag with
and departure of  ′ − profiles of two- and three-dimensional results
mean base suction, a pair of transitions are resolved over the parameter
at  < 300. The  | data of Huhe-Aode et al. (1985) and Vu et al.
  space. The first transition indicated by a jump discontinuity in  -− ,
(2016) listed in Table 8 indicate that the critical spacing decreases with
occurs at the dimensionless critical spacing of 3.4 demarcating the
increasing  for  at least up to 1000 (also see Fig. 2c of this paper). reattachment and co-shedding regimes while the second one indicated
However, as suggested by Fig. 25a and obvious from Fig. 25b, the mean by the appearance of a kink and not a jump in  -− , occurs at
wake length exhibits repeated non-monotonic variations over this range a certain  between 9 and 10 within the co-shedding regime. The
of Reynolds number. Such a variation of  ′ renders a peak in  ′ | to 
 reattachment regime (  
= 2−3.3) is characterized by weak, continuous
appear at  = 300, followed by decline in the ratio. The inset of Fig. 26 vortex-shedding from the leading cylinder, alternate reattachment of its
plots the  ′ | ratio for  = 60, 100 and 160. The  ′ data have been separated shear layers on the rear cylinder and formation of 2S mode in
the rear cylinder wake. The associated distribution of separated shear

taken from Chopra and Mittal (2019). The 
| divulges slow increase
 ′  layers in three parts (rolling into a vortex, alleyway formation and
with . The most interesting feature of the inset is the oneness of  |
  reattachment) differs from those depicted by Alam et al. (2003) for
and  ′ at  = 60. This can be explained by the fact that  ′ <  ′ flow around a pair of in-line circular cylinders at sub-critical Reynolds
at low  close to the onset of vortex-shedding (see Figure 9 of Chopra numbers. At the critical spacing, vortex-shedding from the cylinders
and Mittal (2019)). is in-phase synchronized (as opposed to anti-phase synchronization at
The relationship of critical spacing with characteristic length of 
= 3.3), the rear cylinder wake transits to two-layered C(2S) mode
vortex-formation is explored next for a square cylinder at  = 50,

or TVS mode and the mean wake accommodates a sub-wake (for the
75, 100, 125 and 150 within the regime of two-dimensional flow. The leading cylinder) and an anti-wake (for the rear cylinder). Between
mean wake length of an isolated square cylinder measured from the 
= 3.3 and 3.4, the flow unveils transitional behaviour − a stable

centre of the cylinder is extracted from Chatterjee and Mondal (2012) reattachment at small time finally settles to stable co-shedding. For all
for  = 50, 100, 150 and from Bai and Alam (2018) for  = 75 and gap ratios, vortex-shedding frequency of the rear cylinder locks onto
125. Similar to that of the circular cylinder, the value of mean  ′ | those of the leading cylinder.
ratio exceeds unity and reveals strong sensitivity with . The ratio of This study for the first time demonstrates the importance of the  -

critical spacing and  is plotted with  in the inset of Fig. 27. The
′ − relationship in identifying the transitions occurring in flow around

26
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Fig. 26. Flow past a pair of identical circular cylinders in tandem arrangement for  = 60 − 1000: variation of 
| with . The data for 
 ′ 
| and  ′ are
 
compiled from the
literature. The subscripts l, u and m signify the lower, upper and average values of the ratio. The inset shows the  ′ | −  relationship for  = 60 − 160.

Fig. 27. Two-dimensional laminar flow past a pair of identical square cylinders in tandem arrangement for  = 50 − 150: variation of 
|
 ′ 
with . The data for 
|
 
and  ′
are compiled from the literature. The inset shows the 
|
 ′ 
−  relationship.

27
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

a pair of in-line objects. This curve is also capable of predicting the Nomenclature
range of spacing ratios over which the upstream cylinder acts as an Symbols
isolated object and over which solutions for the downstream cylinders  Blockage
are closer to those of an isolated cylinder. Discontinuities in the  -  Instantaneous drag force coefficient
− curve are found to represent transitions from one flow regime to  Instantaneous lift force coefficient
another. The sensitivity of − to changes in flow conditions (here,   Ratio of maximum lift of downstream and upstream
spacing ratio) adequately predicts the absence and presence of TVS and cylinders
SVS modes for the upstream and downstream cylinders, respectively.  Instantaneous surface pressure coefficient
Besides, a linear variation of  with − is a testimony of single 0 Instantaneous forward stagnation pressure coefficient
cylinder-like behaviour whereas a departure from linearity signals the  Instantaneous base pressure coefficient
influence of a cylinder on the flow around the other.  Characteristic dimension of an object
Several features of mean and fluctuating pressure reported by earlier  Base pressure recovery length measured from the
studies for in-line circular cylinders at sub-critical  are resolved for base point
tandem diamond cylinders at  = 100. The   −  distribution of ′ Base pressure recovery length measured from the
the upstream cylinder for  
= 3.3 is associated with the presence of centre of the cylinder
a single sharp peak close to the shoulder that also happens to be the  Vortex-formation length measured from the base
location of separation. For the downstream cylinder in contrary, a pair point
of peaks appears in the fluctuating pressure, one ahead and one behind  ′ Vortex-formation length measured from the centre of
the shoulder. The peak upstream of the shoulder corresponds to the the cylinder
reattachment of shear layer separated from the leading cylinder (the  Upstream extent of the computational domain
mean pressure also attains its peak at the location of reattachment)  Wake length measured from the base point
while the location of downstream peak matches the location of sep-  ′ Wake length measured from the centre of the
aration of backward shear layer. In the co-shedding regime, i.e.  = cylinder

3.4, absence of reattachment is reflected by the disappearance of the  Fluid static pressure at a point
upstream peak.   curves of the upstream cylinder are virtually  Reynolds number
identical for 
= 9 and 10, yielding very similar values of   (as well  Critical Reynolds number marking the onset of
as   ). Same holds true for the rear cylinder as well. A new length vortex-shedding
scale in the wake, i.e. base pressure recovery length, denoted by the  Strouhal number
symbol  , is introduced.  signifies the distance between the base  Dimensional time
point and a point of recovery of base pressure along the wake axis. This ∗ Non-dimensional time
length scale is however, found to exist for the upstream cylinder alone.  Dimensional time period of a vortex-shedding cycle
Interestingly, the wake stagnation point of the upstream cylinder in the  Fluid velocity vector at a point
co-shedding regime invariably forms downstream of the base pressure  Streamwise velocity component at a point
recovery point indicating that  >  . Thus,  can serve as a  Free-stream speed
characteristic length scale of the wake and yields the lowest estimate  Cross-stream velocity component at a point
of  . The number of  crossings within the inter-cylinder gap  ′  Centre-to-centre spacing between cylinders
or G1F2 matches the number of saddle-stagnation points located in the ′ Surface-to-surface spacing between cylinders
gap. In the reattachment regime, absence of a  crossing ascertains 
Centre-to-centre spacing ratio or gap ratio

non-existence of a saddle-stagnation point within the gap. 
|
 
Centre-to-centre critical spacing ratio
The extent of critical spacing of tandem cylinders is largely gov-
Greek symbols
erned by characteristic dimensions of mean recirculation, i.e.  (or
equivalently,  ) and  of an isolated cylinder. The magnitude of  Phase shift between vortex-shedding from the
critical spacing is low if  (or equivalently,  ) is low and  high. In upstream and downstream cylinders
light of a hypothetical scenario, it is argued that  | must sufficiently  Streamfunction
 
exceed  of an isolated cylinder, else the flow regime will fall under  Density of the fluid
the category of extended single body regime where  | does not  Circumferential angle measured clockwise from the
 
exist. For circular cross-section, Figure 13 of Ljungkrona et al. (1991) forward stagnation point
suggests that  ′ | > 1 up to the shear layer transition regime. The  |  Reattachment angle measured clockwise from the
 
 forward stagnation point
of square, circular and diamond cylinders is compared. For  = 0.02
 Separation angle measured clockwise from the
flow at  = 100, it is found that (  | ) > ( | ) > ( | ) ,
         forward stagnation point
in line with  | >  | >  | ,  | >  | >
 | and  | >  | >  | relationships. By Subscripts
extracting data from the literature for  | ,  ′ and  ′ of circular and
 
rms Root mean square
square cylinders, existence of a possible relationship between critical max Maximum
spacing and characteristic streamwise dimension of recirculation zone Abbreviations
has been explored. The data for circular cross-section over  = 60 −
SUPG Streamline-upwind/Petrov-Galerkin
1000 reveal that each of  | ,  ′ and  ′ depends on . While
  PSPG Pressure-stabilizing/Petrov-Galerkin
the  | −  relationship is monotonic,  ′ and  ′ reveal non-

PVS Prime vortex-shedding
monotonicity with . Accordingly,  | does not appear to be a simple
  TVS Two-layered vortex-shedding
multiplier of  ′ or  ′ . Similar conclusion is reached in conjunction
SVS Secondary vortex-shedding
with square cross-section as well.

28
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

CRediT authorship contribution statement Han, Z., Zou, D., Gui, X., 2012. Flow past two tandem circular cylinders using Spectral
element method. In: International Colloquium on Bluff Body Aerodynamics and
Applications. Shanghai, China, 2–6.
Shravan Kumar Mishra: Writing – original draft, Validation, For-
Huhe-Aode, Tatsuno, M., Taneda, S., 1985. Visual studies of wake structure behind two
mal analysis, Data curation, Conceptualization. Subhankar Sen: Writ- cylinders in tandem arrangement. Rep. Res. Inst. Appl. Mech. (Kyushu Univ. Jpn.)
ing – review & editing, Supervision, Software, Formal analysis. 32 (99), 1–20.
Igarashi, T., 1981. Characteristics of the flow around two circular cylinders arranged
Declaration of competing interest in tandem: 1st report. Bull. Japan Soc. Mech. Eng. 24 (188), 323–331.
Ishigai, S., Nishikava, E., Nishimura, K., Cho, K., 1972. Experimental study on structure
of gas flow in tube banks with tube axis normal to the flow: Part 1, Karman vortex
None flow from two tubes at various spacing. Bull. JSME 15 (86), 949–956.
Jester, W., Kallinderis, Y., 2003. Numerical study of incompressible flow about fixed
Data availability cylinder pairs. J. Fluids Struct. 17, 561–577.
Jiang, H., Cheng, L., 2017. Strouhal–Reynolds number relationship for flow past a
circular cylinder. J. Fluid Mech. 832, 170–188.
Data will be made available on request.
King, R., Johns, D.J., 1976. Wake interaction experiments with two flexible circular
cylinders in flowing water. J. Sound Vib. 45 (2), 259–283.
Acknowledgement Kravchenko, A.G., Moin, P., Shariff, K., 1999. B-Spline method and zonal grids for
simulations of complex turbulent flows. J. Comput. Phys. 151, 757–789.
The Authors are grateful to senior research scholar of the group, Mr. Kumar, D., Mittal, M., Sen, S., 2018a. Modification of response and suppression of
vortex-shedding in vortex-induced vibrations of an elliptic cylinder. Int. J. Heat
Pavan Kumar Yadav, Department of Mechanical Engineering, IIT(ISM)
Fluid Flow 71, 406–419.
Dhanbad for useful discussions in explaining some of the results and Kumar, D., Sourav, K., Sen, S., 2019a. Steady separated flow around a pair of identical
also for providing the  = 0.01 results for a circular cylinder used in square cylinders in tandem array at low Reynolds numbers. Comput. & Fluids 191,
Fig. 5. 104244.
Kumar, D., Sourav, K., Sen, S., Yadav, P.K., 2018b. Steady separation of flow from
an inclined square cylinder with sharp and rounded base. Comput. & Fluids 171,
References
29–40.
Kumar, D., Sourav, K., Yadav, P.K., Sen, S., 2019b. Understanding the secondary
Alam, M.M., Moriya, M., Takai, K., Sakamoto, H., 2003. Fluctuating fluid forces acting separation from an inclined square cylinder with sharp and rounded trailing edges.
on two circular cylinders in tandem arrangement at a subcritical Reynolds number. Phys. Fluids 31 (7), 073607.
J. Wind Eng. Ind. Aerodyn. 91 (1–2), 139–154. Lankadasu, A., Vengadesan, S., 2008. Interference effect of two equal-sized square
Alam, M.M., Zhou, Y., 2007. Phase lag between vortex shedding from two tandem bluff cylinders in tandem arrangement: With planar shear flow. Internat. J. Numer.
bodies. J. Fluids Struct. 23 (2), 339–347. Method Fluids 57 (8), 1005–1021.
Arie, M., Kiya, M., Moriya, M., Mori, H., 1983. Pressure fluctuations on the surface of Li, J., Chambarel, A., Donneaud, M., Martin, R., 1991. Numerical study of laminar flow
two circular cylinders in tandem arrangement. ASME J. Fluids Eng. 105, 161–167. past one and two circular cylinders. Comput. & Fluids 19 (2), 155–170.
Ayyappan, T., Vengadesan, S., 2008. Influence of staggering angle of a rotating rod on
Liu, C.H., Chen, J.M., 2002. Observations of hysteresis in flow around two square
flow past a circular cylinder. J. Fluids Eng. 130 (3), 031103.
cylinders in a tandem arrangement. J. Wind Eng. Ind. Aerodyn. 90 (9), 1019–1050.
Bai, H., Alam, M.M., 2018. Dependence of square cylinder wake on Reynolds number.
Ljungkrona, L., Norberg, C., Sundén, B., 1991. Free-stream turbulence and tube spacing
Phys. Fluids 30 (1), 015102.
effects on surface pressure fluctuations for two tubes in an in-line arrangement. J.
Bao, Y., Wu, Q., Zhou, D., 2012. Numerical investigation of flow around an inline
Fluids Struct. 5 (6), 701–727.
square cylinder array with different spacing ratios. Comput. & Fluids 55, 118–131.
Meneghini, J.R., Saltara, F., Siqueira, C.L.R., Ferrari, Jr., J.A., 2001. Numerical
Batham, J.P., 1973. Pressure distributions on circular cylinders at critical Reynolds
simulation of flow interference between two circular cylinders in tandem and
numbers. J. Fluid Mech. 57 (2), 209–228.
side-by-side arrangements. J. Fluids Struct. 15 (2), 327–350.
Bearman, P.W., 1965. Investigation of the flow behind a two dimensional model with
Mishra, S.K., Yadav, P.K., Sarkar, H., Sen, S., 2022. Correspondence between the
blunt trailing edge and fitted with splitter plates. J. Fluid Mech. 21 (2), 241–255.
number of no-slip critical points and nature of rear stagnation point of a symmetric
Biermann, D., Herrnstein, Jr., W.H., 1933. The Interference Between Struts in Various
object. Phys. Fluids 34 (11), 111702.
Combinations. NACA Technical Report 468.
Mittal, S., Kumar, V., Raghuvanshi, A., 1997. Unsteady incompressible flows past two
Bloor, M.S., 1964. The transition to turbulence in the wake of a circular cylinder. J.
cylinders in tandem and staggered arrangements. Internat. J. Numer. Methods
Fluid Mech. 19 (2), 290–304.
Fluids 25 (11), 1215–1344.
Borazjani, I., Sotiropoulos, F., 2009. Vortex-induced vibrations of two cylinders in
Mizushima, J., Suehiro, N., 2005. Instability and transition of flow past two tandem
tandem arrangement in the proximity–wake interference region. J. Fluid Mech.
circular cylinders. Phys. Fluids 17 (10), 104107.
621, 321–364.
Norberg, C., 1986. Interaction between freestream turbulence and vortex shedding for
Chatterjee, D., Mondal, B., 2012. Forced convection heat transfer from tandem square
a single tube in cross-flow. J. Wind Eng. Ind. Aerodyn. 23, 501–514.
cylinder for various spacing ratios. Numer. Heat Transfer A 61 (5), 381–400.
Chen, C.-K., Wong, K.-L., Cleaver, J.W., 1986. Finite element solutions of laminar flow Okajima, A., 1979. Flow around two tandem circular cylinders at very high Reynolds
and heat transfer of air in a staggered and an in-line tube bank. Int. J. Heat Fluid numbers. Bull. JSME B22, 504–511.
Flow 7 (4), 291–300. Pannell, J.R., Griffiths, E.A., Coales, J.D., 1915. Experiments on the Interference
Choi, C.-B., Jang, Y.-J., Yang, K.-S., 2012. Secondary instability in the near-wake past Between Pairs of Aeroplane Wires of Circular and Lenticular Cross Section. Reports
two tandem square cylinders. Phys. Fluids 24, 024102. and Memoranda No. 208. Annual Reports 7, British Advisory Committee for
Chopra, G., Mittal, S., 2019. Drag coefficient and formation length at the onset of Aeronautics, pp. 219–221.
vortex shedding. Phys. Fluids 31, 013601. Papaioannou, G.V., Yue, D.K., Triantafyllou, M.S., Karniadakis, G.E., 2006. Three-
Dehkordi, B.G., Moghaddam, H.S., Jafari, H.H., 2011. Numerical simulation of flow dimensionality effects in flow around two tandem cylinders. J. Fluid Mech. 558,
over two circular cylinder in tandem arrangement. J. Hydrodyn. 23 (1), 114–126. 387–413.
Deng, J., Ren, A.L., Zou, J.F., Shao, X.M., 2006. Three-dimensional flow around two Park, J., Kwon, K., Choi, H., 1998. Numerical solutions of flow past a circular cylinder
circular cylinders in tandem arrangement. Fluid Dyn. Res. 38 (6), 386–404. at Reynolds numbers up to 160. Korean Soc. Mech. Eng. Int. J. 12, 1200–1205.
Elhimer, M., Harran, G., Hoarau, Y., Cazin, S., Marchal, M., Braza, M., 2016. Coherent Patel, C.G., Sarkar, S., Saha, S.K., 2018. Mixed convective vertically upward flow past
and turbulent processes in the bistable regime around a tandem of cylinders side-by-side square cylinders at incidence. Int. J. Heat Mass Transfer 127, 927–947.
including reattached flow dynamics by means of high-speed PIV. J. Fluids Struct. Posdziech, O., Grundmann, R., 2007. A systematic approach to the numerical calcula-
60, 62–79. tion of fundamental quantities of the two-dimensional flow over a circular cylinder.
Firdaus, A.F., Nguyen, V.L., Zuhal, L.R., 2023. Investigation of the flow around two J. Fluids Struct. 23 (3), 479–499.
tandem rotated square cylinders using the least square moving particle semi-implicit Qu, L., Norberg, C., Davidson, L., Peng, S.-H., Wang, F., 2013. Quantitative numerical
based on the vortex particle method. Phys. Fluids 35 (2), 027117. analysis of flow past a circular cylinder at Reynolds number between 50 and 200.
Gaskell, P.H., Savage, M.D., Thompson, H.M., 1998. Stagnation−saddle points and flow J. Fluids Struct. 39, 347–370.
patterns in Stokes flow between contra-rotating cylinders. J. Fluid Mech. 370, Reinhold, T.A., Tieleman, H.W., Maher, F.J., 1977. Interaction of square prisms in two
221–247. flow fields. J. Wind Eng. Ind. Aerodyn. 2 (3), 223–241.
Gerrard, J.H., 1966. The mechanics of the formation region of vortices behind bluff Roshko, A., 1954. On the Drag and Shedding Frequency of Two-Dimensional Bluff
bodies. J. Fluid Mech. 25 (2), 401–413. Bodies. NACA TN No. 3169.
Griffin, O.M., Ramberg, S.E., 1974. The vortex-street wakes of vibrating cylinders. J. Roshko, A., 1955. On the wake and drag of bluff bodies. J. Aeronaut. Sci. 22 (2),
Fluid Mech. 66 (3), 553–576. 124–132.

29
S.K. Mishra and S. Sen Ocean Engineering 296 (2024) 117011

Roshko, A., 1993. Perspectives on bluff body aerodynamics. J. Wind Eng. Ind. Aerodyn. Vu, H.C., Ahn, J., Hwang, J.H., 2016. Numerical simulation of flow past two circular
49 (1–3), 79–100. cylinders in tandem and side-by-side arrangement at low Reynolds numbers. KSCE
Saad, Y., Schultz, M.H., 1986. GMRES: A generalized minimal residual algorithm for J. Civ. Eng. 20, 1594–1604.
solving nonsymmetric linear systems. SIAM J. Sci. Stat. Comput. 7 (3), 856–869. Williamson, C.H.K., 1988. The existence of two stages in the transition to three
Sen, S., Mittal, S., Biswas, G., 2009. Steady separated flow past a circular cylinder at dimensionality of a cylinder wake. Phys. Fluids 31, 3165–3168.
low Reynolds numbers. J. Fluid Mech. 620, 89–119. Williamson, C.H.K., 1996. Vortex dynamics in the cylinder wake. Annu. Rev. Fluid
Sen, S., Mittal, S., Biswas, G., 2011. Flow past a square cylinder at low Reynolds Mech. 28, 477–539.
numbers. Internat. J. Numer. Methods Fluids 67 (9), 1160–1174. Williamson, C.H.K., Roshko, A., 1988. Vortex formation in the wake of an oscillating
Shan, X., 2021. Effect of an upstream cylinder on the wake dynamics of two tandem cylinder. J. Fluids Struct. 2 (4), 355–381.
cylinders with different diameters at low Reynolds numbers. Phys. Fluids 33 (8), Williamson, C.H.K., Roshko, A., 1990. Measurements of base pressure in the wake of
083605. a cylinder at low Reynolds numbers. Z. Flugwiss. Weltraumforsch. 14 (1), 38–46.
Sharman, B., Lien, F.S., Davidson, L., Norberg, C., 2005. Numerical predictions of low Xu, G., Zhou, Y., 2004. Strouhal numbers in the wake of two inline cylinders. Exp.
Reynolds number flows over two tandem circular cylinders. Internat. J. Numer. Fluids 37, 248–256.
Methods Fluids 47, 423–447. Yadav, P.K., Sourav, K., Kumar, D., Sen, S., 2021. Flow around a diamond-section
Shui, Q., Duan, C., Wang, D., Gu, Z., 2021. New insights into numerical simulations cylinder at low Reynolds numbers. Phys. Fluids 33 (5), 053611.
of flow around two tandem square cylinders. AIP Adv. 11, 045315. Yang, W., Stremler, M.A., 2019. Critical spacing of stationary tandem circular cylinders
Singha, S., Sinhamahapatra, K.P., 2010. High resolution numerical simulation of low at  ≈ 100. J. Fluids Struct. 89, 49–60.
Reynolds number incompressible flow about two cylinders tandem. ASME J. Fluids Yen, S.C., San, K.C., Chuang, T.H., 2008. Interactions of tandem square cylinders at
Eng. 132 (1), 011101. low Reynolds numbers. Exp. Therm Fluid Sci. 32, 927–938.
Slaouti, A., Stansby, P.K., 1992. Flow around two circular cylinders by the Zdravkovich, M.M., 1977. Review of the interference between two circular cylinders in
random-vortex method. J. Fluids Struct. 6 (6), 641–670. various arrangements. J. Fluids Eng. 99 (4), 618–633.
Sohankar, A., Etminan, A., 2009. Forced−convection heat transfer from tandem square Zdravkovich, M.M., 1984. Classification of flow-induced oscillations of two parallel
cylinders in cross flow at low Reynolds numbers. Internat. J. Numer. Methods circular cylinders in various arrangements. In: The American Society of Mechanical
Fluids 60 (7), 733–751. Engineers Proceeding, Flow Induced Vibration Symposium. Vol. 2, pp. 1–18.
Sourav, K., Kumar, D., Sen, S., 2020. Undamped transverse-only VIV of a diamond Zdravkovich, M.M., 1987. The effect of interference between circular cylinders in cross
cylinder at low Reynolds numbers. Ocean Eng. 197, 106867. flow. J. Fluids Struct. 1 (2), 239–261.
Stansby, P.K., 1981. A numerical study of vortex shedding from one and two circular Zhang, W., Su, X., 2021. Effect of surface curvature on destabilization and unsteadiness
cylinders. Aeronaut. Q. 32 (1), 48–71. of low- flow across two tandem elliptic cylinders. Proc. Inst. Mech. Eng. C 235
Tanida, Y., Okajima, A., Watanabe, Y., 1973. Stability of a circular cylinder oscillating (22), 6080–6098.
in uniform flow or in a wake. J. Fluid Mech. 61 (4), 769–784. Zhao, X., Cheng, D., Zhang, D., Hu, Z., 2016. Numerical study of low-Reynolds-
Tezduyar, T.E., Behr, M., Liou, J., 1992a. A new strategy for finite element number flow past two tandem square cylinders with varying incident angles of
computations involving moving boundaries and interfaces—The deforming-spatial- the downstream one using a CIP-based model. Ocean Eng. 121, 414–421.
domain/space–time procedure: I. The concept and the preliminary numerical tests. Zhao, M., Cheng, L., Zhou, T., 2013. Numerical simulation of vortex-induced vi-
Comput. Methods Appl. Mech. Engrg. 94 (3), 339–351. bration of a square cylinder at a low Reynolds number. Phys. Fluids 25,
Tezduyar, T.E., Behr, M., Mittal, S., Liou, J., 1992b. A new strategy for finite element 023603-1–023603–25.
computations involving moving boundaries and interfaces—the deforming-spatial-
domain/space–time procedure: II. Computation of free-surface flows, two-liquid
flows, and flows with drifting cylinders. Comput. Methods Appl. Mech. Engrg. 94
(3), 353–371.

30

You might also like