2.1. Methodology to determine stable postures

In addressing the problem, we developed a methodology that is based on the equilibrium relationship for a two-dimensional object. This methodology operates under the assumption that at low Reynolds numbers, the stable falling of an object is effectively identical to a scenario where uniform flow passes around a stationary object. As illustrated in figure 1, we consider an object of arbitrary shape, with the origin of the coordinate system ($O$–$xy$) positioned at its geometric centre. Notably, while the centre of buoyancy (CoB) of the object is invariably located at its geometric centre, the centre of gravity (CoG) is contingent on the mass distribution of the object and may not coincide with the geometric centre. Here, we define the Reynolds number as $Re= U_rd/\nu$, where $U_r$ represents the relative velocity of the incoming flow, $d$ is the characteristic length of the object and $\nu$ is the kinematic viscosity of the fluid. Here $d$ is calculated as $\sqrt {S}$, with $S$ being the area of the object. The $d$ used in the validation process is consistent with other articles and will be described separately. The angle of attack $\alpha$ is defined as the direction of incoming flow with respect to the $x$-axis.

Figure 1.

Schematic of the force ($B$, $G$ and $F$) and moment ($M$) acting on a falling object.

For a steady falling object in a fluid, the force equilibrium equations can be expressed as $F=G-B$ and $M=GL$, where $F$ and $M$ are the total fluid force and moment acting on the object, respectively. Here $B$ and $G$ denote the buoyancy and gravity forces, respectively, and $L$ represents the arm of the gravity force. Here $L=m^*C_M/C_F$ can be derived, where $C_M$ and $C_F$ are moment and total force coefficients, respectively. The equations for $C_M$ and $C_F$ can be expressed as

(2.1)\begin{gather} {C_M}= \frac{2M}{\rho_fU_r^{2}d^2}, \end{gather}

(2.2)\begin{gather}{C_F}= \frac{2F}{\rho_fU_r^{2}d}, \end{gather}

where $\rho _f$ is the density of the fluid. And one of the influence parameters is $m^*=1-\rho _f/\rho _s$, where $\rho _s$ is the density of the object. According to the force balance equation $F = G - B$ and (2.2), another influence parameter Archimedean number $Ar$ can be derived:

(2.3)\begin{equation} {Ar}= \frac{Fd}{\rho_f\nu^{2}}= \frac{gd^3}{\nu^{2}}\left(\frac{\rho_s}{\rho_f}-1\right).\end{equation}

Thus, for any specific combination of $m^*$, $Ar$ and $\alpha$, we can consistently derive a line of solutions for the CoG that adheres to the force equilibrium relationship. It is noted that only the lower half of this CoG solution line represents static stable conditions for a falling object. Consequently, the upper half, which signifies unstable conditions, is not depicted in figure 1. The critical point between the upper and lower half-lines refers to the CoG location where the restoring moment equals zero. The direction of gravity is indicated by an arrow in the diagram. By systematically adjusting the angle of attack through a full $360^{\circ }$ rotation, we can map out all possible stable solution lines for the CoG. This approach enables us to thoroughly understand and predict the range of stable falling postures for an object under different conditions of fluid dynamics.

3.1. Experimental set-up

Free-falling experiments are conducted in a static fluid tank for objects of different shapes, namely, elliptical, triangular and L-shaped objects. The dimensions of the experimental tank are 300 mm long, 250 mm wide and 1000 mm tall. Images of the object models and the cross-section are shown in figure 2. The dimensions of the cross-section of the objects are $a_1 = 7$ mm, $a_2 = 10$ mm and $a_3 = 1.25$ mm. The height of the tank is 100 times $a_2$ to ensure that the object has enough space to achieve stability. The spanwise length of the model is 150 mm, indicating a sufficiently long object to minimize the end effect. The object models are made of copper with a uniform mass distribution ($\rho _s = 8400.34\ {\rm kg}\ {\rm m}^{-3}$); thus, the CoG of the object is located at its geometric centre. The rectangular glass tank is filled with a mixture of glycerin and water. The mixture is stirred thoroughly to ensure that the liquid is sufficiently uniform. The appropriate $Ar$ and $m^*$ can be achieved by adjusting the viscosity and density of the fluid with different ratios of water and glycerin. Each object model is released at different initial angles from $0^\circ$ to $360^\circ$ with an interval of approximately $30^\circ$. The object model is released at rest from the centre of the tank and at a depth of approximately 50 mm below the fluid surface. The distance between the object and the walls of the tank is maintained at no less than 50 mm to minimize the effect of the wall on the falling mode. The temperature of the room is approximately 21–25 $^\circ$C during the experiments, and the variation in temperature in the experiment for one model is maintained within 0.6 $^\circ$C to ensure that the variation in the fluid viscosity is no more than 22.5 cP (centipoise, $1~{\rm cP} = 10^{-3}$ Pa s). The fluid viscosity was measured using a Brookfield DVS digital display viscometer. A camera-based motion capture system is developed to measure the falling postures of the objects with high accuracy. The NikonD850 was used to record the video of the falling motion. The shooting resolution is $3840 \times 2160$, and the shooting frame rate is 30 f.p.s. (frames per second). The camera is calibrated before measurement, and the distortion parameter obtained by the calibration is used to correct the video results. The Hough transform is used to identify the outlines of the models. Then, the position, orientation and speed of the objects are calculated.

Figure 2.

Physical image (a) and the cross-sectional dimensions (b–d) of the object models.

3.2. Free-fall tests

The properties of liquids are measured in experiments to obtain the viscosity and density of the liquid. The elliptical and L-shaped objects are released in a liquid with a viscosity of 386.25 cP and a density of 1332 kg m$^{-3}$, while the triangular object is released in a liquid with a viscosity of 757.5 cP and a density of 1349.8 kg m$^{-3}$. The corresponding $Ar$ values are 34.79 (ellipse), 27.22 (triangle) and 49.44 (L-shape). In addition, $m^*$ is 0.841 (ellipse), 0.839 (triangle) and 0.841 (L-shape). For a steadily falling object, there are three parameters that can completely express its state, including $\theta _g$, $Re$ and $\theta _m$. The direction $\theta _g$ is defined as the relative direction of gravity with respect to the $x$-axis. Here $\theta _m$ is the direction of motion of the object, which corresponds to the direction of incoming flow from the stationary object and is defined as the direction relative to the x-axis. Figure 3(a–c) shows the variations in the gravity direction, $Re$ and motion direction over time. Two stable solutions are achieved for all three objects. The object tends to stabilize at approximately 1.5 s. All the results are in perfect agreement with the interpolation results based on the proposed approach (described later in § 4 in detail). The results obtained from the experiment were superimposed. The typical fall postures and trajectories are shown in figure 3(d). The snapshot intervals are 5, 3 and 4 frames for the elliptical, triangular and L-shaped objects, respectively. Therefore, the existence of multiple stable falling postures is demonstrated. The details are shown in Supplementary movie 1 is available at https://doi.org/10.1017/jfm.2024.557.

Figure 3.

Experimental verification results of the multiple falling postures of the objects. (a–c) The variations in $\theta _g$, $Re$ and $\theta _m$ over time for the elliptical, triangular and L-shaped objects released at different initial angles. (d) Results showing the fall postures and trajectories of the elliptical (left-hand), triangular (middle) and L-shaped (right-hand) objects.

4.1. Numerical methods

The two-dimensional Navier–Stokes equations governing the incompressible fluid flow are

(4.1)\begin{gather} \frac{\partial{u_i}}{\partial{x_i}}= 0, \end{gather}

(4.2)\begin{gather}\frac{\partial{u_i}}{\partial{t}}+u_j\frac{\partial{u_i}}{\partial{x_j}}={-}\frac{1}{\rho_f}\frac{\partial{p}}{\partial{x_i}}+ \nu\frac{\partial^2{u_i}}{\partial{x_j}\partial{x_j}}, \end{gather}

where $i,j = 1, 2, p$ is the pressure. For clarity, the velocity components $u_1$ and $u_2$ are also denoted by $u_x$ and $u_y$, respectively. We use STAR-CCM+ software to solve these equations using the finite volume method. Here, a second-order upwind scheme is used to discretize the convective term, and steady and implicit unsteady solvers are used to perform transient calculations for stationary and falling objects, respectively. A six-degree-of-freedom motion solver is applied to simulate the migration of an object under flow-induced forces as well as gravity.

A rectangular computational domain is adopted, and the size of the domain is large enough to eliminate the boundary effect. Since the dimensions and shapes of the objects are different, the computational domain sizes used here are different for different shapes. The distances from the upstream, downstream and lateral boundaries to the centre of the object are no less than $49d$, $149d$ and $62d$, respectively. For the numerical simulation of the falling object, a moving computing domain is adopted based on the overset grid method. The computational domain moves at the same speed as the object. This approach ensures a sufficiently long development process for simulating the falling object. The overset mesh scheme is employed to solve the problem of rotational and large amplitude horizontal migration of an object during its fall.

In order to verify the accuracy of the calculations, the numerical simulation results for the fixed ellipse were verified. Table 1 is a comparison of $C_D$, $C_L$ and $C_M$ for fixed ellipses between the present numerical simulations and the literature. Drag coefficient $C_D$ and lift coefficient $C_L$ are

(4.3)\begin{gather} {C_D}= \frac{2F_D}{\rho_fU_r^{2}d}, \end{gather}

(4.4)\begin{gather}{C_L}= \frac{2F_L}{\rho_fU_r^{2}d}, \end{gather}

where $F_L$ and $F_D$ are lift and drag. The results at different Reynolds numbers are in good agreement with the literature. In numerical simulations, the computational domain is partitioned into a set of unstructured trimmed cells. The mesh is initially generated at a basic size, it is refined in the region containing the object. The effects of the grid resolution on the calculated results are verified for the elliptical object using four meshes with different numbers of elements from 51 670 to 105 211. As shown in table 2, good agreement is obtained for the cases with different meshes, indicating good grid convergence. In the simulation of this work, we all use the M2 which is shown in figure 4. Here M2 is used for all the following simulations in this study.

Table 1.

Comparison of results for $C_D$, $C_L$ and $C_M$ for $Re=15$ and 30 (the characteristic length is the length of the major axis of the ellipse), $\alpha =45$.

Table 2.

Effect of the mesh resolution of the ellipse on the drag coefficient $C_D$ when $Re = 25$ (the characteristic length is the length of the major axis of the ellipse), and $\alpha =90$.

Figure 4.

Grid distribution of ellipse in a falling view. (a) The mesh in the whole computational domain. (b) Close-up view of the mesh around the object.

4.2. Interpolated results

As in the experiments, we consider elliptical, triangular and L-shaped objects as shown in figure 2(b–d). Based on the extensive numerical simulations for flow past the fixed objects, the fluid force and moment for the objects at different $Re$ and $\alpha$ are obtained. The $Re$ ranges from 1 to 15 at an interval of 0.5 or 1, while $\alpha$ ranges from $0^\circ$ to $360^\circ$ at an interval of $10^\circ$. Figure 5 shows the distribution of the CoG lines for elliptical, triangular and L-shaped objects, respectively, at $Ar = 150$ and $m^* = 0.5$. For each shape, the stable solutions of the CoG line are depicted for the incidence angle $\alpha$, ranging from $0^\circ$ to $360^\circ$ at an interval of $10^\circ$. The solution map for each shape can be classified into several base families according to the distribution. Figure 5(a) illustrates two base families of CoG lines for an elliptical object, denoted in yellow (family 1) and blue (family 2), respectively. Each base family of solution is of a trumpet shape with its mouth facing outward and the tip facing inward. Therefore, in regions where different base families overlap, the CoG lines intersect, indicating the existence of multiple stable falling solutions. The overlapped multiple-solution region is represented in figure 5(b) using a mixed colour (green) of the base family colours (blue and yellow). A similar procedure is employed for the more complex cases with additional base families. As shown in figure 5(c–e), a pink colour is introduced to represent a third base family. The dark brown area represents the convergence of three base families, indicating that three different falling postures can occur there. These regions are named as SS fall posture regions, BS fall posture regions and TS fall posture regions, respectively. As shown in these figures, the ellipse has no more than two stable falling postures. However, up to three stable solutions can be obtained for the triangular and L-shaped objects. It is noted that the single-solution area still dominates the majority of the map.

Figure 5.

Distribution of CoG lines for elliptical, triangular and L-shaped objects at $Ar = 150$ and $m^* = 0.5$. (a) Two base families of CoG lines of elliptical objects. (b) The single stable (SS) fall posture region and bistable (BS) fall postures region of elliptical. (c) Three base families of CoG lines of triangle objects. (d,e) The SS, BS and tristable (TS) fall postures regions of (d) triangular and (e) L-shaped objects.

4.3. Parameter influence

In the methodology for determining stable postures, it is found that the direction of the CoG line is only determined by $Ar$. This is because the direction of the CoG line is dependent on the direction of the force on the object, and the direction of the force on a fixed object is determined by the angle of attack and the $Re$ together. The $Re$ is embodied in $Ar$, so the direction of the CoG line at a certain angle of attack is determined by $Ar$. The magnitude of $L$ is determined by $m^*$, $C_F$ and $C_M$, where $C_F$ and $C_M$ are also determined by $Ar$ in the same way as the direction of the CoG line. Therefore, the distance $L$ of the CoG line from the origin is determined by $m^*$ and $Ar$ together.

Figure 6(a) shows the distribution of the boundary of the different regions under different $m^*$ when $Ar=100$. The force and the moment coefficient on a stationary object remain constant when $Ar$ is constant, thus, the relationship between $L$ and $m^*$ is linear. The entire boundary is enlarged by a corresponding factor centred on the coordinate origin as $m^*$ increases. Therefore, the area of the BS and TS region is linear with ${m^*}^2$. Figure 6(b) shows the boundary distribution between different regions under different $Ar$ when $m^*=0.5$. When $m^*$ is constant, the area of the BS and TS region increases with the increase of $Ar$. To the extent that the force on the fixed objects is stabilized, $Ar$ has less effect on the distribution in the SS, BS and TS region which is due to the small variation of forces and moments on fixed objects over a small range of $Re$.

Figure 6.

The influence of $Ar$ and $m^*$ on the boundary of SS, BS and TS region. (a) The boundary of SS, BS and TS region changes with $m^*$ when $Ar = 100$. (b) The boundary of the SS, BS and TS region changes with $Ar$ when $m^* = 0.5$.

4.4. Free-fall tests

The experiment faced limitations due to material selection and manufacturing accuracy. To ensure that the geometric centre is positioned within the multisolution region, copper was selected as the material for the object, resulting in $m^*$ being greater than 0.8. In contrast, the convenience of numerical computation allows for the selection of appropriate parameters to discover three stable falling postures. Therefore, the free-falling simulation is conducted for the cases at $Ar = 150$ and $m^* = 0.5$. As shown in figure 7(a), three shapes same to the experiments, i.e. the ellipse, triangle and L-shape, are considered, and the coordinates of their CoGs (CoG1, CoG2 and CoG3) that are listed in table 3 are selected in SS, BS and TS regions. For the initial state of a falling object, parameters such as initial velocity, angular velocity and initial release angle are considered. In this study, the object is released from a state of rest, so the initial state is characterized solely by the initial release angle. Therefore, 12 independent free-falls are simulated at different initial angles with an interval of 30$^\circ$ for each object. Figures 7(b) and 8 show the variations in the gravity direction, $Re$ and motion direction over time. All the terminal stable falling postures perfectly coincide with interpolation results based on the proposed method, confirming the existence of multiple stable falling postures.

Table 3.

The dimensionless coordinates of CoG for the objects selected in numerical simulations.

Figure 7.

Numerical verification of the multiple falling postures of the objects. (a) The selected CoGs in different regions and the $\theta _g$ obtained by the proposed method of these CoGs. (b) The change in gravity direction with time for selected CoGs released at various initial angles and the potential energy of the chosen CoGs at different $\theta _g$ of (i) elliptical, (ii) triangular and (iii) L-shaped objects when $Ar = 150$ and $m^* = 0.5$.

Figure 8.

The change in $Re$ (a) and $\theta _m$ (b) with time for selected CoGs released at various initial angles of elliptical (i), triangular (ii) and L-shaped (iii) objects when $Ar = 150$ and $m^* = 0.5$.

In terms of the potential energy, a multiple stable system has multiple local minimums corresponding to the static stable equilibrium solutions and local maximums corresponding to the unstable equilibrium solutions. Here, we define the potential energy as

(4.5)\begin{equation} {E}={-}\int_{0}^{\theta_g}N\, {\rm d} \theta, \end{equation}

where $N$ represents the restoring moment caused by gravity at $\theta _g$. Specifically, $N=Gd'$, where $d'$ is the distance from the CoG of the object to the CoG line at $\theta _g$. As shown in figure 7(b), the number and locations of potential wells agree well with those of the interpolation results based on the proposed method. The deeper the potential well corresponds the more easily attainable falling posture. It also corresponds to a wider range of initial release angles. Therefore, it is concluded that the proposed method based on the static flow simulation can interpolate to obtain the stable posture solution of the free-falling body in fluid at low $Re$.