3.1 DSMC計算による円管内流速の層流分布からの逸脱:乱流再現?(Transition from laminar to pseudo- turbulent distribution of flow velocity in a circular tube by DSMC calculation)


(The description in English is shown after the Japanese sentence.)

DSMC計算で,レイリー・テイラー不安定に似たものを外乱として円管の内壁近傍に与え,円管内流速の層流分布からの逸脱(乱流?)を算出した.この計算には,人為的な近似操作が用いられているので,実際の管内乱流にどれだけ近づいているかについては,今後,十分な検討を行う必要がある.今回は,とりあえず,この解析の導入部を記述する.

DSMC法は,希薄気体解析のために開発されたもので,多数の分子の運動を元に流れを再現するものである.したがって,低密度(温度一定とすれば低圧力)であればあるほど,あるいは,代表長さが小さければ小さいほど解析しやすくなる.また,分子の平均速度を仮に 350 m/sとしたら,それに近いか大きい流速の現象は解析しやすいが,通常の生活に現れるような 10 m/s 以下の風に起因する現象は,DSMC法では解析しにくい対象となる.そのことを踏まえ,今回,まず,Re=5000 の円管内流れをシミュレーションするにあたって,次のような条件を考えた.軸対称を仮定した流れ場は,直径d=2.5mm(これが代表長さ)で,また,管軸方向にdと同じ長さかそれ以上の長さを持つ円筒管で,管の入口と出口には周期境界条件を適用する.当然,管路壁面に摩擦が存在するので,壁面で失った運動量は,流れ場に存在する全ての分子に均等に補填して,流れ場全体の運動量すなわち平均流速は一定に保たれるようにする.失った運動量を全ての分子に均等に与えるという考え方であるが,一定流速で垂直な管の中を落下する流れは,分子に働く重力と壁面摩擦とが釣り合うことにより維持されることを考えれば妥当なものである.管路長は,長ければ長いほど興味深い結果を導けると思われるが,計算時間も長くかかることを覚悟する必要がある.圧力0.5気圧(380 mmHg),温度288 K のアルゴンAr単一気体(粘性係数 2.2e-5 Pa・s)が平均流速50 m/sで流れるとすれば,Reは約5000となる.DSMC法で用いられる壁面での境界条件は,分子が壁面で散乱反射(拡散反射)されるというものが一般的である.

初期条件として,平均流速で一様分布するとして分子速度を発生させたのち計算を続けていくと,Re=5000であっても,層流を示す放物線分布に収束する(ポアズイユ流).この結果は外乱を全く与えていないので不自然なことではない.その後,壁面で反射する分子に人為的操作(筆者は,これを表面粗さの大きなゆらぎという観点から発案したが,厳密にはそれは正確でない)を加えて分子の反射速度を決定し,それにより,結果として,管の内壁面近傍にレイリー・テイラー不安定に類似するものを発生させて外乱とすると,これが流速分布に歪みを引き起こす(DSMC用の宇佐美の疑似外乱).なお,反射分子の軸方向の運動量は互いに打ち消し合ってその総和はゼロとなるようにする.この外乱を決めるのに2つのパラメータを用いているが,その値の選択に応じて最終的な管内流速分布の形状が定まる.各図の横軸は,管中心からの距離を管直径で無次元化したもの.縦軸は,流速を平均流速で無次元化したものである.図1,2(アニメーション)は,軸方向全領域を平均して結果表示しているのに対し,図3,4(アニメーション)では,軸方向をある程度細かく分割して結果表示しているので,壁面近傍における流速変化の様子が観測できる.図5,6(アニメーション)は,両者を交互に表示していったもの.また,図7,8(アニメーション)は,壁面付近に生じる流れのうねりの様子を,軸方向流速の等値線と流線で示している.なお,流れ場全体の温度変化や密度変化は,最大で1%程度なので,結果表示は流速で行い,あえて,質量流速にはしていない.また,今回の計算にはUsys法(U_system) だけを用いており,Bird法(B_original) は使用していない.ここで採用したセル構造を用いた場合,Bird法では,層流の放物線分布ですら完全には得ることができなかった.Bird法は,Usys法より計算速度に優れるので,より細かなセル構造を採用して比較するのがフェアであるが,ここでは行わない.なお,下記結果については,断りなく転載していただいて結構ですが,出典元 ‘DSMC calculation by M.Usami (https://usamimas.net)’ を明示していただくようお願いいたします.他の章の記述についても同様です.

Fig. 1 Velocity distribution obtained by averaging the entire area (A).
Fig. 2 Velocity distribution obtained by averaging the entire area (B).
Fig. 3 Velocity distribution obtained at several sections (A).
Fig. 4 Velocity distribution obtained at several sections (B).
Fig. 5 Velocity distributions obtained by two different averagings (A).
Fig. 6 Velocity distributions obtained by two different averagings (B).
Fig. 7 Contour lines of the axial flow velocity.
Fig. 8 Undulations of flow near the wall shown with streamlines and contour lines (Usami’s pseudo disturbance for DSMC).

 

In the present DSMC analysis, something similar to the Rayleigh-Taylor instability has been applied as a disturbance near the inner wall of the circular tube, and the transition of the flow velocity in the tube from a laminar distribution to a turbulent distribution has been calculated. Since this calculation uses artificial approximation, sufficient consideration will be required in the future to determine how close it can be to the actual turbulence in the tube.

This time, I will first explain the introductory part of this analysis. The DSMC method has been developed for rarefied gas dynamics and it reproduces the flow based on the motion of a large number of molecules. Therefore, the lower the density (lower the pressure if the temperature is constant) and/or the smaller the characteristic length, the easier the analysis is. Furthermore, if the average speed of molecules is assumed to be 350 m/s, it is easy for the DSMC method to analyze phenomena with flow velocities close to or larger than that, but difficult to solve phenomena caused by winds of 10 m/s or less in daily life. Based on this, we first considered the following conditions when simulating the flow in a circular tube with Re=5000. The axisymmetric flow field is assumed, which consists of a cylindrical tube with a diameter d = 2.5 mm (this is the characteristic length) and a length equal to or longer than d in the tube axis direction. Periodic boundary condition is also assumed at the inlet and outlet of the tube. Naturally, since there is friction on the tube wall, the momentum lost on the wall is compensated equally for all molecules present in the flow field, so that the momentum of the entire flow field, that is, the average flow velocity, is kept constant. The idea that the lost momentum is equally imparted to all molecules is reasonable if you consider that a flow falling at a constant flow rate in a vertical tube is maintained by balancing the gravity acting on the molecules and the wall friction. It seems that the longer the tube length, the more interesting the results will be, but you need to be prepared that the calculation time will also be longer. If a simple argon gas (viscosity coefficient 2.2e-5 Pa_s) with a pressure of 0.5 atm (380 mmHg) and a temperature of 288 K flows at an average velocity of 50 m/s, Re will be approximately 5000. The boundary condition at the wall surface used in the DSMC method is generally such that molecules are diffusely reflected by the wall surface (diffuse reflection).

If the molecular velocity is generated from a uniform distribution at the average flow velocity as an initial condition and the calculations are continued, even if Re=5000, it will converge to a parabolic distribution indicating laminar flow (Poiseuille flow). This result is not unnatural since no disturbance was applied. From then on, the reflection velocities are determined by artificially manipulating them of the molecules from the wall (the author proposed this from the perspective of undulating at long positional intervals in surface roughness, but strictly speaking, this is not accurate). As a result, if a disturbance similar to Rayleigh-Taylor instability is generated near the inner wall of the tube, this causes distortion in the flow velocity distribution (Usami’s pseudo disturbance for DSMC). Note that the axial momenta of the reflected molecules cancel each other so that their total sum becomes zero. Two parameters are used to determine this disturbance, and the final shape of the flow velocity distribution in the circular tube is determined depending on the selection of their values. The horizontal axis of each figure is the distance from the center of the tube, which is normalized by the tube diameter. The vertical axis is the flow velocity normalized by the average flow velocity. Figures 1 and 2 (animation) indicate the results by averaging the entire area in the axial direction, on the other hand, figures 3 and 4 (animation) show the results by dividing the axial direction into several sections so that it is easy to investigate how the flow velocity changes near the wall surface. Figures 5 and 6 (animation) are those in which both are displayed alternately. In addition, figures 7 and 8 (animation) show the undulations of the flow that occur near the wall using streamlines and contour lines of the axial flow velocity. Note that the temperature and density changes in the entire flow field are approximately 1% at most, so the results are shown in terms of flow velocity and intentionally not in mass flux (rate of mass flow). In addition, only the Usys method (U_system) has been used in this calculation, and the Bird method (B_original) is not used. Using the cell structure adopted here, the Bird method could not completely obtain even the parabolic distribution of laminar flow. Since the Bird method has excellent calculation speed than the Usys method, it would be fair to use a more fine cell structure for comparison, but this is not done here. You may reproduce these results without permission, but please cite the source, ‘DSMC calculation by M.Usami (https://usamimas.net)’. The same applies to the descriptions in other chapters.