1 Sequence of bifurcations of natural convection of air in a laterally heated cube with perfectly insulated horizontal and spanwise boundaries

2025-04-30 0 0 5.12MB 52 页 10玖币
侵权投诉
1
Sequence of bifurcations of natural convection of air in a laterally heated cube
with perfectly insulated horizontal and spanwise boundaries
Alexander Gelfgat
School of Mechanical Engineering, Faculty of Engineering, Tel-Aviv University, Ramat
Aviv, Tel-Aviv, Israel, 6997801, gelfgat@tau.ac.il
Abstract
A sequence of three steady oscillatory transitions of buoyancy convection of air in a laterally
heated cube with perfectly thermally insulated horizontal and spanwise boundaries is studied. The
problem is treated by Newton and Arnoldi methods based on Krylov subspace iteration. The finite
volume grid is gradually refined from 1003 to 2563 finite volumes. It is shown that the primary
instability is characterized by two competing eigenmodes, whose temporal development results in
two different oscillatory states that differ by their symmetries. Bifurcations due to both modes are
subcritical. These modes develop into different oscillatory and then stochastic flow states, which,
at larger Grashof number, stabilize and arrive to single stable steady flow. With further increase
of the Grashof number this flow loses it stability again. It is argued that in all the three transitions,
the instabilities onsets, as well as reinstatement of stability, take place owing to an interaction
between a destabilizing centrifugal mechanism and stabilizing effect of thermal stratification.
Key words: natural convection, instability, Krylov methods, SIMPLE iteration
2
1. Introduction
A primary goal of this study is to examine a chain of three steady oscillatory bifurcations
that take place in a three-dimensional benchmark configuration of natural convection of air in a
laterally heated cube with perfectly insulated horizontal and spanwise boundaries. The two-
dimensional analog of this problem, i.e., convection in a square laterally heated cavity with
perfectly thermally insulated (adiabatic) horizontal boundaries, is one of the earliest CFD
benchmarks [1] proposed for validation of steady flow calculations. It was then extended for
comparison of calculated critical parameters of the primary bifurcation corresponding to steady to
oscillatory transition. The details and additional references can be found in [2]. Later, the two-
dimensional problem was extended to a three-dimensional one in a cube, assuming the spanwise
boundaries to be also perfectly thermally insulated. The results for 3D steady flows have been
published and cross-verified [3-8], so that the 3D steady flows benchmark is also well established.
However, numerical investigation of the instabilities of 3D steady flows is significantly more
challenging and had been addressed only by straight-forward time integration of the 3D time-
dependent governing equations [9-14].
Recently, a comprehensive linear stability analysis was applied to the primary steady -
oscillatory transition of air convection in a laterally heated cube in two simpler configurations, for
a cube with perfectly thermally conducting horizontal and spanwise boundaries [15], and with
perfectly insulated horizontal and perfectly conducting spanwise boundaries [16]. The second case
appeared to be noticeably more complicated because the critical Grashof number becomes almost
two orders of magnitude larger, and the steady oscillatory transition is preceded by a symmetry
breaking steady bifurcation. A replacement of the perfectly conducting spanwise boundaries by
the perfectly insulated ones leads to a qualitatively different transition, which exhibits a sequence
of bifurcations described below. This finding is in line with recent computational and experimental
study [17], where the authors concluded that the flow can be stabilized or destabilized by variation
of the heat transfer conditions at the horizontal boundaries.
The time-dependent calculations of [14] showed that with the increase of the Grashof number
, the steady flow becomes oscillatory unstable at  > 4 10, then the stability reinstates at
 > 7 10, and the resulting steady flow becomes oscillatory unstable at  > 2 10.
Following a series of the cited above previous studies, we consider the fixed value of the Prandtl
number,  = 0.71, characteristic for air. The flows with very large [18] and very small [19,20]
3
Prandtl numbers were also considered for the current configuration, however dependence of the
transition on the Prandtl number is out of scope of the present study.
In this study, we examine the instabilities by applying linear stability analysis, using the
numerical approach and the visualization technique of [15,16]. For calculation of steady flows we
apply the Newton method, whose corrections are calculated by the biconjugate gradient stabilized
(BiCGstab(2)) method combined with generalized minimal residual (GMRES) method. Leading
eigenvalues are computed by the Chebyshev preconditioned Arnold iteration [21]. The
corresponding Krylov vectors, which are divergence free and satisfy all the boundary conditions
are computed by the SIMPLE-like technique proposed in [15,22].
Applying the linear stability analysis to the calculated 3D steady flows, we confirm three
transitions predicted by the time-dependent computations [14]. We discuss which physical
mechanisms destabilize the flow, then stabilize it, and then destabilize it again. Along with that,
we confirm existence of two different most unstable modes of the primary instability, as was
predicted by [11]. We show that these modes differ by broken flow symmetries and become
unstable at close Grashof numbers. Additional time-dependent calculations revealed that beyond
the stability limits, there exist two different oscillatory flow states, which also differ by their
symmetries. It was shown also that the steady oscillatory transitions caused by any of the two
modes are subcritical. With further increase of the Grashof number, these modes turn into
stochastic regimes with different phase space attractors. Surprisingly, when the Grashof number is
increased to approximately 7 × 10, the two attractor collapse into the same stable focus, thus
producing a single stable steady flow state. This flow state remains stable up to the Grashof number
2.9 × 10, after which the flow becomes oscillatory and then turbulent.
Analyzing the most unstable disturbances and steady flow patterns near the critical points, we
argue that in all three cases the destabilization and stabilization result from an interaction of two
main factors: destabilizing centrifugal instability mechanism and stabilizing thermal stratification.
The centrifugal instability sets in at the cube corners where the flow direction turns from horizontal
to vertical, and vice versa, similarly to what is observed in the 3D lid driven cavity flow (see [22]
and references therein). This mechanism can be enhanced by reverse circulations that appear near
the top and bottom borders, and increase the curvature of streamlines of the main convective
circulation. On the other hand, a strong convective mixing tends to make the isotherms far from
the vertical isothermal boundaries almost horizontal, so that colder and heavier fluid is located
4
below the warmer and lighter one. This results in a stable stratification, which tends to suppress
all possible instabilities.
2. Formulation of the problem
As in our previous studies [14-16], we consider the natural convection in an incompressible
fluid in a cubic cavity, whose opposite sidewalls are maintained at constant but different
temperatures,  and . The horizontal and spanwise boundaries are perfectly thermally
insulated, as is defined for the AA AA case in [14]. The Boussinesq approximation is applied.
To render equations dimensionless we choose
,  
,
as scales of the length, time
, the velocity =(,,) and the pressure , respectively, where is the fluid kinematic
viscosity and is the density. The temperature is rescaled to a dimensionless function by  
(  ) (  ). Additionally, the dimensionless time, velocity and pressure are
scaled, respectively, by  
,  
, and , where  =(  )
is the
Grashof number, is the acceleration due to gravity, and is the thermal expansion coefficient.
The resulting system of energy, momentum, and continuity equations is defined in a cube 0
,,  1 as:

 +(  )=
 
 (1)

 +(  )= +
 
 + (2)
   = 0 . (3)
where  = 
is the Prandtl number, and is the thermal diffusivity. All the boundaries are
assumed to be no-slip. Two vertical boundaries at = 0,1 are kept isothermal, so that
(= 0, ,)= 1, (= 1, ,)= 0. (4)
The absence of the heat flux at the horizontal and spanwise boundaries yields:

 =
 = 0 (5)

 =
 = 0 (6)
For discussion purposes, the areas adjacent to the cube edges (0,y,0), (0,y,1), (1,y,0) and (1,y,1) are
called below the lower left, upper left, lower right, and upper right corners, respectively.
5
The problem is additionally characterized by three symmetries [11]: (i) reflection symmetry
with respect to the midplane = 0.5, {,,,}(,,)={,,,}(, 1  ,), (ii) 2D
rotational symmetry with respect to rotation through 180° around the line x = z = 0.5,
{,,,}(,,)={,,,}(1 ,, 1  ), and (iii) 3D centro-symmetry
{,,,}(,,)={,,,}(1 , 1  , 1  ), where, =  (1 ). Obviously,
these symmetries are characteristic of steady state flows at relatively small subcritical Grashof
numbers. They can be broken by instability, so that supercritical oscillatory flows can maintain
only one type of symmetry or be fully non-symmetric. Following [16], to examine whether the
steady state solution is symmetric, we use the temperature field and define
 =|(,,) (, 1  ,)| (7)
 =|(,,)+(1 ,, 1  )| (8)
 =|(,,)+(1 , 1  , 1  )| (9)
as measures of the reflection, 2D rotational and 3D centro-symmetries, respectively. For the
symmetric solutions, these values are of the order of 10. Deviation of any of these expressions
from numerical zero indicates symmetry breaking.
3. Numerical and visualization techniques
The numerical approach is the same as in [14-16], where the reader is referred for the details.
Here we only mention that this approach allows for calculation of divergence free Krylov vectors
satisfying all the boundary conditions, which, in its turn, allows for a direct calculation of the
leading eigenvalue [21], without applying the Arnoldi method in shift-and-inverse mode like in
[23].
As mentioned in [16], convergence of the critical parameters for a similar 2D case of
convection in a laterally heated square cavity with insulated horizontal boundaries was established
in [23], the convergence and comparison with independent results for 3D steady states were
reported in [8], and the convergence of slightly supercritical oscillatory states was studied in [14].
Three-dimensional velocity fields are visualized by the method proposed in [24,25] and
applied to similar problems in [15,16], where all the definitions and computational details are
discussed. Here, to clarify the plots below, we only mention that 3D flow is visualized by
摘要:

1SequenceofbifurcationsofnaturalconvectionofairinalaterallyheatedcubewithperfectlyinsulatedhorizontalandspanwiseboundariesAlexanderGelfgatSchoolofMechanicalEngineering,FacultyofEngineering,Tel-AvivUniversity,RamatAviv,Tel-Aviv,Israel,6997801,gelfgat@tau.ac.ilAbstractAsequenceofthreesteady–oscillator...

展开>> 收起<<
1 Sequence of bifurcations of natural convection of air in a laterally heated cube with perfectly insulated horizontal and spanwise boundaries.pdf

共52页,预览5页

还剩页未读, 继续阅读

声明:本站为文档C2C交易模式,即用户上传的文档直接被用户下载,本站只是中间服务平台,本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私,请立即通知玖贝云文库,我们立即给予删除!
分类:图书资源 价格:10玖币 属性:52 页 大小:5.12MB 格式:PDF 时间:2025-04-30

开通VIP享超值会员特权

  • 多端同步记录
  • 高速下载文档
  • 免费文档工具
  • 分享文档赚钱
  • 每日登录抽奖
  • 优质衍生服务
/ 52
客服
关注