Granular flow experiment using artificial gravity generator at International Space Station

The study addresses how gravitational acceleration influences the mechanics and flow of granular regolith relevant to landers and rovers on extra-terrestrial bodies. Prior rover and lander experiences underscore the importance of accurate wheel–soil and footpad–regolith interaction models. Many models (pressure–sinkage, resistive force theory, local friction models, unified drag, modified Archimedes law) implicitly depend on gravity through hydrostatic-like forces. Discrete Element Method (DEM) simulations are widely used to model such interactions, but verification and validation require experimental data across gravity levels. Existing reduced-gravity experiments on Earth (parabolic flights, drop towers) have short durations and unstable gravity. Beverloo’s law predicts mass flow rate scales with the square root of gravity, but its validity at low gravity remains uncertain due to lack of data. This paper proposes long-duration, stable artificial gravity experiments aboard the ISS to quantify granular flow across 0.063–2.0 G and to compare against DEM, aiming to validate gravity scaling and reveal deviations, with implications for bulk density and design of space systems.

The paper surveys prior experimental and computational research on gravity-dependent granular behavior: studies show flow dynamics depend on gravity; angle of repose remains constant across gravity, but friction angle increases as gravity decreases. Wheel mobility under reduced gravity has been tested. DEM has been used extensively to predict rover mobility and granular interactions on Earth, Moon, and Mars, and in low-gravity settings. However, existing experimental platforms (parabolic flights, drop towers) provide short, limited, and unstable gravity profiles. Beverloo’s law is a foundational relation for hopper discharge, validated for wide outlets and under high gravity, but lacking validation at low gravity. The review motivates the need for long-duration, stable low-gravity data for V&V of DEM and for testing Beverloo scaling.

Experiments: Conducted aboard the Japanese Experimental Module (JEM) on the ISS using a centrifuge to generate artificial gravity (AG) levels from 0.063 to 2.0 G in steps, each held for approximately 13 minutes, over a 7-hour campaign. Eight hourglass-shaped, vacuum-sealed (<20 Pa) hoppers were built, each loaded with a different granular material. Hoppers had two cone angles enabling flow at 60° and 120° relative to the orifice. A servo motor flipped each hourglass 180° every 60 s (flip duration ~0.6 s), alternating flow direction/angle; total flips >400. Onboard camera captured flows at 25 fps (H.264 MP4, 40 Mbps). Images were undistorted and synchronized with AG profiles via calibration markers. Mass flow rate Vm was computed as Vm = m/(te − to), where m is packed mass, to is when the hourglass is perpendicular to AG during flip (flow start), and te is when motion ceases; temporal resolution 0.04 s. Pre-flight Earth (natural gravity, NG, 1.0 G) tests were also performed. Materials: Eight granular media: alumina beads (#01), silica sand No. 5 (Tohoku sand, #02), Toyoura sand (#03), and regolith simulants (FJS-1 sieved, Phobos simulant, JSC MARS-1A, silica sand No. 8, FJS-1). Representative masses packed, particle size distributions, shear strengths, and densities are provided in Supplementary Methods. Alumina beads were partly dyed black for imaging. DEM simulations: Rocky DEM software modeled particles as spheres with size distributions matching measured sieving histograms. Contact models: Hertzian spring-dashpot (normal) and Mindlin-Deresiewicz (tangential); rolling resistance model (type C) to emulate nonsphericity effects and energy dissipation. Material parameters (density, elastic, friction, restitution, rolling resistance) were tuned using minimum density tests, residual strengths from direct shear tests, static friction tests against borosilicate glass, and hourglass flow data. For NG simulations, a uniform gravity field and flipping were applied. For AG simulations, zero-gravity field with imposed revolution replicating the centrifuge created an effective centrifugal field inside the hourglass; particles passing the orifice experienced micro/zero-g while the hourglass rotated, introducing Coriolis effects. Geometry, revolution center/radius, and angular velocities matched the experiment (AG gradient effects were minor within the geometry). Particle velocities were sampled just below the orifice in a defined window for analysis. Analysis: Mass flow rates were measured across AG levels for three free-flowing media (alumina beads, silica sand No. 5, Toyoura sand). Regression to Beverloo-inspired power law Vm = α θ^(0.5+β) was performed over 0–1 G to quantify deviations from square-root scaling; α, β, and R² with p-values (F-test) were reported. Additional analyses examined intermittency due to arching, flow modes (mass/funnel), and the competition between particle weight and van der Waals adhesion as a function of particle size (thresholds ~50 μm under 1 G; >200 μm weight dominates even at 0.063 G). Particle trajectories under AG and the effect of Coriolis force were calculated using initial velocities from DEM.

• Stable AG experiments (0.063–2.0 G) over ~7 hours with >400 flips produced high-quality granular flow data; three media (alumina beads, silica sand No. 5, Toyoura sand) exhibited clear mass flow. Materials with many fine particles experienced adhesion/aggregation and complex flow modes.
• Mass flow rate increased with gravity following a power law consistent with Beverloo’s law (∝ √g), with measurable deviations at low gravity captured by exponent correction β. Hopper angle 60° yielded faster discharge than 120°.
• Quantitative regression (0–1 G) for Vm = α θ^(0.5+β): • 60°: Alumina beads α=15.41 g/s, β=0.0836 (p=1.63×10⁻⁴), R²=0.972; Silica sand No. 5 α=8.73 g/s, β=−0.0178 (p=0.222), R²=0.977; Toyoura sand α=7.67 g/s, β=0.0289 (p=0.014), R²=0.989. • 120°: Alumina beads α=8.32 g/s, β=0.0909 (p=1.16×10⁻⁴), R²=0.969; Silica sand No. 5 α=3.50 g/s, β=0.0144 (p=0.387), R²=0.975; Toyoura sand α=3.69 g/s, β=0.0770 (p=3.87×10⁻⁶), R²=0.986.
• Flow durations ranged from 0.40 s (alumina beads, 1.0 G) to 6.56 s (silica sand No. 5, 0.063 G).
• DEM reproduced experimental trends and relative magnitudes of mass flow rates across gravity and hopper angles; NG (uniform g) and AG (centrifuge) produced similar velocity fields near the orifice in mass flow regime, supporting equivalence for flow-rate determinants.
• Particle velocities just below the orifice scaled ∝ √g under both NG and AG; differences between NG and AG were minor at low g and/or 120° hopper, with some divergence at 60° and higher g due to post-orifice dynamics and Coriolis effects.
• Intermittent flow under low gravity was linked to transient arching near the orifice; no permanent clogs occurred in the three free-flowing media.
• Adhesive forces dominate for particles <~50 μm at 1 G; for particles >~200 μm, gravitational weight dominates even at 0.063 G. Media containing many fines (#04–#08) showed adhesion, deposition, aggregation, bridging, and ratholing.
• Deviations (positive β) imply effective bulk density decreases with decreasing gravity (increased porosity during fall), suggesting bulk density should be tuned downward in low-g simulations for accurate predictions.

The experiments validate that granular discharge rates scale with gravity consistent with Beverloo’s law while quantifying deviations under low gravity. By operating in a long-duration, stable AG environment, the study overcomes limitations of short, unstable reduced-gravity platforms, providing robust datasets for V&V of DEM models. DEM analyses show that, within the hopper’s upper region where flow rate is set, AG and NG produce similar velocity fields, supporting the use of AG data to infer NG behavior and vice versa for mass flow phenomena. The measured positive β for several media indicates that low gravity increases porosity and reduces effective bulk density in the flowing region, leading to slightly enhanced gravity dependence relative to √g. For silica sand No. 5, larger particle size reduces adhesive effects, yielding β not significantly different from zero and closer adherence to classic Beverloo scaling. These findings directly inform design and analysis of landers and rovers: models should incorporate gravity-dependent bulk density to avoid overestimating traction or underestimating sinkage in low-g environments. The dataset provides a benchmark for calibrating DEM (including parameters like rolling resistance and friction) and for validating mechanical interaction models under low gravity.

The study introduces and demonstrates an ISS-based artificial gravity experimental approach enabling long-duration, stable investigation of granular flow from 0.063 to 2.0 G. Measured discharge rates follow Beverloo-type scaling with quantifiable deviations at low gravity, attributable to reduced bulk density (increased porosity) of flowing media. DEM simulations corroborate experimental trends and show that near-orifice flow fields are similar under AG and NG, validating the AG approach for studying gravity-dependent granular flows. The results provide actionable guidance for space robotics: tune bulk density downward in low-g simulations to improve predictions of rover traction and lander sinkage. Publicly available datasets support V&V of DEM and other models. Future work should incorporate van der Waals and electrostatic forces explicitly in DEM, investigate non-spherical particle shapes and dilatancy effects, expand material sets and geometries, and refine measurements across broader gravity ranges to further constrain scaling laws.

• DEM models did not explicitly include van der Waals or electrostatic forces; these effects likely influenced fine-particle media (#04–#08) and adhesion to glass walls under vacuum.
• Regression was focused on 0–1 G due to dataset sufficiency; behavior above 1 G less constrained by regression.
• Differences between AG and NG exist in post-orifice dynamics (Coriolis effects, gravity field gradients), and NG vs AG discharge at 120° showed discrepancies; however, near-orifice fields were similar for mass flow.
• Hourglass geometry, orifice size relative to particle size, and limited camera frame rate (25 fps; 0.04 s temporal resolution) constrain detection of rapid transients.
• Only three media exhibited clear mass flow across all conditions; materials with wide size distributions and substantial fines exhibited adhesion, bridging, and ratholing, limiting generalizability to cohesive or polydisperse regolith.
• Spherical particle assumption in DEM approximates non-spherical grains via rolling resistance; detailed shape effects and dilatancy were not fully captured.