APPENDIX H

Analytical Verification of the Solver

Equation and implementation checks for contact, rigid bodies, clumps, liquid bridges and breakage. Each result states its reference, parameters and evidence limits.

1. Purpose and methodology

Verification establishes that the selected equations are solved correctly. Validation compares predictions with independent physical data; calibration fits parameters to a set of tests. This appendix collects numerical equation checks: plots compare analytical references and solver response. Parameters and procedures accompany each case; theoretical formulation is in Chapter 3. Agreement with a law does not establish experimental material validation.

For a nonzero reference, relative error = |measured−reference|/|reference|. Tables state whether errors are fractions or percentages; probability errors use percentage points (pp). For a zero reference, use the absolute residual with its unit. Results refer to each case’s stated conditions; reproduce the same parameters and observables.

Stiffness and timestep. Control overlap and check convergence by reducing Δt. The GUI offers a suggestion based on Rayleigh time. With breakage, start from a fraction of 0.05–0.1 and account for the smallest fragment; convergence testing remains necessary. The very small timesteps of isolated numerical tests below are verification conditions rather than general performance recommendations.

2. Normal sphere–wall restitution

What this case verifies

A sphere hits the wall head on, without friction and without gravity. The restitution coefficient e = |vout/vin| measures the fraction of speed preserved. It is the most basic test and covers, at once, the normal contact force, the damping, the prescribed restitution, the time integration and the approximate conservation of energy.

Figure 2.1. The sphere arrives with vin and returns with e·vin.

Parameters to reproduce it in the GUI

Panel / fieldValue
Material · Radius0.005 m
Material · Density2500 kg/m³
Material · Young's modulus (E)1×10⁷ Pa
Material · Poisson (ν)0.30
Material · Restitution (e)0.1 to 0.9 (swept)
Material · Friction (μ)0
Contact interactionHertz–Mindlin
Setup · Gravity / Δt(0,0,0) / 1×10⁻⁶ s
Initial conditionv = (0, −1, 0) m/s above the floor

Result

Figure 2.2. Measured against prescribed restitution (dashed: ideal).
QuantityAnalyticalbrixDEMError [%]
e = 0.100.100.2526152.6
e = 0.200.200.324862.4
e = 0.300.300.397132.4
e = 0.400.400.472018.0
e = 0.500.500.550310.1
e = 0.600.600.63235.4
e = 0.700.700.71822.6
e = 0.800.800.80811.0
e = 0.900.900.90200.2
Reading the recorded values. The deviation from prescribed restitution is 10.1% at e=0.5, 5.4% at e=0.6 and ≤2.6% at e≥0.7. At low e the difference is much larger. This table documents the contact/damping behaviour under the test conditions; it does not establish accuracy over the whole range and cannot support an overall “within 5%” verdict.

3. Oblique collision with friction

What this case verifies

The sphere hits the wall with normal and tangential velocity components and no initial spin. Friction converts part of the linear momentum into rotation. The core test is the coupling between the change in tangential velocity and the change in rotation, where sign or magnitude errors of the torque would show up.

Figure 3.1. The tangential component vt generates rotation ω.

Parameters to reproduce it in the GUI

Panel / fieldValue
Material · Radius / Density0.005 m / 2500 kg/m³
Material · Young / Poisson1×10⁷ Pa / 0.30
Material · Restitution (e)0.5
Material · Friction (μ)0.10
Contact interactionHertz–Mindlin
Setup · Gravity / Δt(0,0,0) / 1×10⁻⁶ s
Initial conditionv = (1, −1, 0) m/s, ω = 0

Result

Figure 3.2. Measured/analytical ratio (green: strict test; orange: model characteristic).
QuantityAnalyticalbrixDEMError [%]
Coupling Δω/Δv_t [1/m]500.0472.45.5 (OK)
v_n,out [m/s]0.5000.55010.1 (model)
v_t,out [m/s]0.8500.8651.8 (OK)
|ω| vs μ·J_n [rad/s]75.063.715.1 (model)
Reading. The torque/translation coupling matches within 5.5 %: the sign and magnitude of the generated rotation are correct. The larger deviations come from the restitution excess (e = 0.5) and from partial tangential mobilization at the contact edges, where the normal force tends to zero. These are model characteristics, not torque errors.

4. Sliding into pure rolling

What this case verifies

A sphere is launched with horizontal velocity v₀ on a floor with friction, without initial spin. Friction decelerates it and spins it up at the same time, until pure rolling (v = ω·R) sets in and the velocities become constant. It verifies Coulomb friction, torque, moment of inertia and the regime transition.

Figure 4.1. Friction lowers v and builds ω until pure rolling.

Parameters to reproduce it in the GUI

Panel / fieldValue
Material · Radius / Density0.005 m / 2500 kg/m³
Material · Young / Poisson1×10⁷ Pa / 0.30
Material · Friction (μ)0.5
Material · Rolling resistanceNone
Contact interactionHertz–Mindlin
Setup · Gravity / Δt(0, −9.81, 0) / 2×10⁻⁶ s
Initial conditionv = (1, 0, 0) m/s, ω = 0, on the floor

Result

Figure 4.2. vp and ωp·R converge to 5/7·v₀ at ts.
QuantityAnalyticalbrixDEMError [%]
v_f [m/s]0.714290.713940.05
ω_f [rad/s]142.857142.790.05
t_s [s]0.05827≈0.058< 1
Reading. Essentially exact convergence (0.05 %). One of the most complete cases in the set, exercising friction, torque and inertia at the same time.

5. Sphere on an inclined plane

What this case verifies

On a ramp the sphere has two regimes. With enough friction it rolls without sliding and the acceleration is a = (5/7)·g·sinθ. With low friction it slides and a = g(sinθ − μ·cosθ). Verifying both separately confirms that the solver tells the regimes apart and combines gravity, rotation and contact.

Figure 5.1. Acceleration along the inclined plane.

Result

Figure 5.2. Acceleration against inclination: rolling (left) and sliding (right).

Case 4, rolling without sliding (μ = 0.6):

θ [°]a analytical [m/s²]a brixDEM [m/s²]error [%]
50.61070.61040.05
101.21681.21620.05
151.81361.81270.05
202.39662.39550.05
252.96132.96000.05
303.50363.50200.04

Case 5, sliding (μ = 0.10):

θ [°]a analytical [m/s²]a brixDEM [m/s²]error [%]
202.43342.43340.00
253.25683.25680.00
304.05544.05540.00
354.82324.82320.00
405.55435.55430.00
Reading. Error below 0.1 % across the whole range, in both regimes. In the GUI, use the same material as Case 3 and create the floor as a plane with the normal rotated by θ; μ = 0.6 to roll, μ = 0.10 to slide.

6. Rolling threshold (rolling resistance)

What this case verifies

With rolling resistance active (CDT model, μr = 0.3) and high friction to prevent sliding, the sphere only starts rolling when the inclination exceeds θc = arctan(μr). Below that it stays static. It verifies the direction and the threshold of the rolling resistance torque.

Figure 6.1. The sphere rolls only when tan θ > μr.

Parameters to reproduce it in the GUI

Panel / fieldValue
Material · Friction (μ)0.60 (prevents sliding)
Material · Rolling resistanceCDT, μ_r = 0.30
Contact interactionHertz–Mindlin
GeometryPlane inclined by θ (swept 8° to 24°)
Setup · Gravity / Δt(0,−9.81,0) / 2×10⁻⁶ s
Initial conditionat rest on the ramp

Result

Figure 6.2. Acceleration against θ: zero below θc, growing above.
θ [°]Reported acceleration [m/s²]State
80.0000static
12-0.0000static
140.0000static
16-0.0000static
170.0351rolling
180.1627rolling
200.4178rolling
240.9262rolling
Reading. The observed transition falls between 16° and 17°, matching the predicted θc = 16.70°. The rolling torque acts in the correct direction.

7. Braking by rolling resistance

What this case verifies

A sphere in pure rolling on a horizontal floor, with rolling resistance (CDT, μr = 0.05), decelerates linearly until it stops. It verifies the magnitude of the deceleration imposed by the rolling torque. The correct value for this model is (5/7)·μr·g, obtained from the angular momentum balance at the contact point; the simplified form (2/5)·μr·g does not correspond to this model.

Figure 7.1. The rolling torque brakes the sphere.

Parameters to reproduce it in the GUI

Panel / fieldValue
Material · Friction (μ)0.5
Material · Rolling resistanceCDT, μ_r = 0.05
Contact interactionHertz–Mindlin
Setup · Gravity / Δt(0,−9.81,0) / 2×10⁻⁶ s
Initial conditionpure rolling: v=(1,0,0), ω_z=−v/R

Result

Figure 7.2. vp(t) decays with slope (5/7)·μr·g.
QuantityAnalyticalbrixDEMError [%]
deceleration a [m/s²]−0.3504−0.35080.1
Reading. The measured deceleration matches within 0.1 % the value (5/7)·μr·g.

8. JKR adhesion between two spheres

What this case verifies

Two cohesive spheres (JKR model) are slowly pulled apart; the largest tensile force before rupture (snap-off) is measured. It verifies the sign and magnitude of the adhesion and the separation criterion. JKR cohesion keeps the bonded state from first touch and sustains tension across an already open gap until it breaks at pull-off — behavior that holds for the particle-to-particle pair and for wall contact alike (Case 8b).

Figure 8.1. Cohesive force between the spheres; pull-off peak before snap-off.

Parameters to reproduce it in the GUI

Panel / fieldValue
Material · Radius / Density0.005 m / 2500 kg/m³
Material · Young1×10⁷ Pa
Material · CohesionJKR
Material · Surface energy (γ)0.10 J/m²
Contact interactionHertz–Mindlin + JKR
Setup · Gravity / Δt(0,0,0) / 5×10⁻⁷ s
Initial conditiontwo spheres in contact, separating slowly

Result

Figure 8.2. Attractive force against distance: peak at pull-off and snap-off.
QuantityAnalyticalbrixDEMError [%]
F_pull-off [mN]2.35622.35360.11
Reading. The tensile peak matches within 0.11 % the value 3π·γ·Reff.

Extension: JKR adhesion at a wall (Case 8b)

The same pair from Case 8 was repeated replacing one sphere with a flat wall carrying JKR cohesion and the same γ. The wall model uses the full JKR normal force: it keeps the bonded state during separation and breaks at the analytical snap-off, exactly like the particle-to-particle pair. The only difference is the effective radius — R_eff = R/2 between two equal spheres, R_eff = R against the flat wall — which doubles the pull-off force for the same γ.

Figure 8.3. Analytical JKR curves (stable branch) with the Case 8 parameters (R = 5 mm, E = 1×10⁷ Pa, ν = 0.30, γ = 0.10 J/m²). Both contacts sustain tension across an open gap and break on reaching pull-off; the markers land exactly on the values measured by the suite.

Result

QuantityAnalyticalbrixDEMError
Particle–wall F_pull4.7124 mN4.7124 mN< 0.01 %
Particle–wall rupture gap1.4526 µm1.4513 µm0.09 %
Particle–wall F(δ) curveanalytical solutionmax relative error 4.19×10⁻⁶—
Particle–particle F_pull (reference)2.3562 mN2.3562 mN< 0.01 %
Particle–particle rupture gap1.1529 µm1.1521 µm0.07 %

9. Free-body 6-DOF: torque-free rotation and DEM coupling

What this case verifies

The Free-body 6-DOF mode (Chapter 7.11) has two independent halves, and this appendix covers both. The first is the integrator: torque-free rotation of a body with three distinct principal moments, which only comes out right if the gyroscopic term ω × (I ω) exists and the quaternion is normalized. The second is the coupling: transmission of force and torque from the contact point to the center of mass, and the corresponding reaction on the particle.

Part A — torque-free rotation (Dzhanibekov effect)

Part A status: reproduction reference. The RK4 curve and invariants below are independent references. This section does not contain a measured solver series establishing that this torque-free rotation was executed and passed. The measured Part B results are separate coupling evidence.

A body spinning about the intermediate axis sits in unstable equilibrium and must flip periodically, with a closed-form period — the tennis racket theorem. Since the case uses no particles, a failure points straight at the integrator, never at the contact model.

Figure 9.1. The T-handle spins about the intermediate axis X. The principal moments satisfy IY < IX < IZ, and it is that ordering that makes the spin about X unstable.
Panel / fieldValue
Body · geometry / densityT-handle: bar 120 × 20 mm, stem 24 × 80 mm, thickness 20 mm, recentred on the center of mass / 2650 kg/m³
Body · Mass and inertia0.228960 kg; Ixx = 2.074773×10⁻⁴; Iyy = 1.651565×10⁻⁴; Izz = 3.573698×10⁻⁴ kg·m²
Body · Motion typeFree-body 6-DOF, preset Free — 6 degrees of freedom; quaternion 1,0,0,0; no lock, spring or limit
Body · Initial angular velocityω = (20.00; 0.10; 0.10) rad/s — without the 0.10 perturbation on the other two axes the body stays in unstable equilibrium and never flips
Setup · Gravity / Δt / duration(0,0,0) / ≤ 1×10⁻⁵ s / 6 s, no particles and no contact
Figure 9.2. Reference series of ω in the body frame (independent integration of the Euler equations, RK4, Δt = 2×10⁻⁵ s). ω_X flips between +20 and −20 rad/s every 3.4930 s. This is the curve the solver output must overlay.
QuantityReference valueTolerance
\|L\| in world frame4.149733418×10⁻³ kg·m²/srelative drift < 1×10⁻⁴ over 6 s
Rotational energy4.149807930×10⁻² Jrelative drift < 1×10⁻⁴ over 6 s
Flip period3.4930 s (zeros at 0.8156 / 2.5621 / 4.3086 s)2 %
Check the inertia tensor before running. Use the Compute mass, center and inertia from the STL button (Chapter 7.11) on a block of known dimensions and compare against the closed form I = m/12·(a² + b²). For a 60 × 40 × 20 mm block at 2650 kg/m³: mass 0.127200 kg, Ixx 2.120000×10⁻⁵, Iyy 4.240000×10⁻⁵, Izz 5.512000×10⁻⁵ kg·m² — acceptable error below 0.1 %. With that step correct, the numbers in the case below are trustworthy.

Part B — DEM coupling (measured)

The coupling is covered by 25 free-body checks — gravity and quaternion normalization, full inertia tensor, locks and limits, simultaneous STL contacts (n = 2, 3 and 8), persistent contact history, a sphere in a granular bed, dense particle fields with three seeds, springs and dampers, an external hinge, and regression of the legacy motions. All passed. Two quantitative results summarize the set:

QuantityAnalyticalbrixDEMError
Hinged gate · impulse J0.764575 N·s0.765049 N·s0.18 % (relative error of the set)
Hinged gate · particle v−0.52915 m/s−0.530099 m/s
Hinged gate · gate ω−0.344059 rad/s−0.34427 rad/s
Hinged gate · L residual05.33×10⁻¹⁴—
Hinge pendulum · period1.63795 s1.6382 s0.015 %
Hinge pendulum · energy drift04.60×10⁻¹¹—
Why this case is decisive. The eccentric impact is the test that separates a correct coupling from an apparently correct one: if only the force reached the body, without the lever arm, the gate would translate without rotating and ω would come out zero. The angular-momentum residual on the order of 10⁻¹⁴ confirms the lever arm is measured from the center of mass, not the mesh's geometric center.

10. Clumps: rigid-body dynamics and mass properties

What this case verifies

A clump is a rigid body made of overlapping spheres (Chapter 3.14). The verification covers the whole chain: the mass properties obtained from the geometric union of the spheres, the load distribution among the pebbles in contact, the response to constant force and torque, the invariants of central and eccentric impact, JKR cohesion on the clump, and the convergence order of the orientation integration.

Figure 10.1. Left, the dimer: the intersection lens belongs to both spheres and can only be counted once in the volume. Right, the consequence — the naive sum of the two spheres only coincides with the exact union when they just touch (d/r = 2); at typical overlap the difference is several percent, and that is what the mass-scaling check rules out.

Result

0 failures across 17 checks.

QuantityExpectedbrixDEM
Mass scaling M ~ ρ·s³00
Inertia scaling I ~ ρ·s⁵00
Hertz equilibrium: overlap2.699789×10⁻⁵2.699789×10⁻⁵
Hertz equilibrium: total load0.2055380.205538
Load distribution among pebbles00
Constant force: Δv = F·t/M0.54998630.5499862
Constant torque: ΔL = τ·t5.382208×10⁻⁶5.386167×10⁻⁶
Central impact: restitution0.70.7203921
Central impact: P / L conservation0 / 00 / 1.34×10⁻¹⁹
Eccentric impact: restitution0.70.7202585
Eccentric impact: P / L conservation0 / 00 / 2.05×10⁻¹⁴
JKR on clump: pull-off force5.654867×10⁻³ N5.654954×10⁻³ N
JKR on clump: rupture distance1.640308×10⁻⁶ m1.640452×10⁻⁶ m
Rigidity: distance between pebbles01.39×10⁻¹⁷
Orientation: \|q\| − 100
Temporal order of orientation integration22.001639
What the numbers show. The mass and inertia scalings match exactly (error 0), confirming that the volume integration subtracts the overlap and that inertia scales with ρ·s⁵. The distance between pebbles constant at 10⁻¹⁷ shows the clump moves as a rigid body, with no internal drift. The measured temporal order of 2.0016 matches the expected order 2 of the orientation integrator. The restitution deviations (0.72 against 0.70) are the same Tsuji damping overshoot already documented in Case 1, not a clump effect.

11. Liquid bridge: separation, rupture and viscosity

What this case verifies

Two spheres form a bridge on touching and then separate. Force at each separation is compared with the Mikami correlation, together with the distance at which the bridge ceases to act. A second test adds normal and tangential motion to check Nase’s viscous resistance. Equations and conventions are in 3.15.

Figure 11.1. Two spheres joined by a pendular bridge of volume V and contact angle θ are pulled apart at speed v; the capillary force F(S) is recorded against the separation S until rupture.
Figure 11.1. Two spheres joined by a pendular bridge of volume V and contact angle θ are pulled apart at speed v; the capillary force F(S) is recorded against the separation S until rupture.

Parameters and procedure

Panel / conditionValue
SpheresR = 1 mm; ρ = 1×10⁶ kg/m³; E = 1×10⁷ Pa; ν = 0.30; e = 1; μ = 0
Prescribed bridgeγ = 0.072 N/m; θ = 0°; V̂ = 0.002; V = 2×10⁻¹² m³; CGF = 1
Initial conditionOverlap 2×10⁻⁷ m; axial velocities −0.01 and +0.01 m/s; zero gravity.
TimestepΔt = 1×10⁻⁷ s
Viscous testη = 1000 mPa·s; S₀ = 1×10⁻⁶ m; v₂,t = 0.01 m/s
Sphere–plane variantR = 1 mm; θ = 20°; V̂ = 0.002

In the GUI, use two spheres, zero gravity and prescribed bridge volume with zero liquid contents (10.9); record force and surface separation until rupture. In the second test, enable viscosity and tangential velocity. The high density is artificial: it limits acceleration while reading the curve and is not an ore property. Pointwise comparison requires controlled position and force recording; the quick plotter does not directly provide a surface-separation axis.

Result

Figure 11.2. Capillary force against separation. Dashed: reference; symbols: brixDEM. Equal spheres: θ = 0°; sphere–plane: θ = 20°.
Figure 11.2. Capillary force against separation. Dashed: reference; symbols: brixDEM. Equal spheres: θ = 0°; sphere–plane: θ = 20°.
Figure 11.3. Total normal and tangential force magnitudes in the viscous test. Dashed: reference; symbols: brixDEM.
Figure 11.3. Total normal and tangential force magnitudes in the viscous test. Dashed: reference; symbols: brixDEM.
QuantityReferencebrixDEM / errorCriterion
F(h)Mikami, Eqs. (3.56)–(3.58)4.42×10⁻¹⁶ maximum relative error≤ 1×10⁻⁶
Rupture: equal spheres119.670 µm119.669 µm; 0.00084%≤ 0.01%
Normal viscous forceNase, Eq. (3.60)5.61×10⁻⁵ maximum relative error≤ 1×10⁻⁴
Tangential viscous forceNase, Eq. (3.60)6.65×10⁻⁶ maximum relative error≤ 1×10⁻⁴
Interpretation. Force decreases with separation and vanishes at rupture; viscous resistance opposes motion. These curves check law evaluation, not fitting to an experimental material. Errors are computed before rounding; the force check excludes the tail below 2% of the maximum and the final rupture neighbourhood.

Additional check: liquid conservation

In migration mode, repeat bridge formation and rupture with dry-basis contents of 5% and 1%, liquid density 1000 kg/m³ and relative bridge-volume limit 0.05. Liquid should move from film to bridge and return without changing the inventory of a closed system.

QuantityReference / result
Initial films2.094395×10⁻⁷ m³; 4.188790×10⁻⁸ m³
Formed bridge volume5×10⁻¹¹ m³
Maximum relative inventory drift2.11×10⁻¹⁶; criterion ≤ 1×10⁻¹²

For compatibility with clumps and breakage, see 28.7.

Code accuracy × model accuracy

The errors in the table above compare the solver with the same Mikami equation: they prove the law is implemented correctly, not that it reproduces a real bridge. The uncertainty of the model itself is much larger and can be estimated by comparing Mikami with other published solutions for the same case (equal spheres, R = 1 mm, V̂ = 0.002, θ = 0°):

QuantityMikami (used in BrixDEM)Independent referenceDifference
Force at contact, F/(2πRγ)0.935Derjaguin limit: 1 (V → 0)−6.5%
Rupture distance119.67 µmLian et al. (1993): (1+θ/2)·V1/3 = 125.99 µm−5.0%
Rupture distance119.67 µmWillett et al. (2000): (1+θ/2)(V̂1/3+V̂2/3/10)·R = 127.58 µm−6.2%
Rupture at V̂ = 0.001 / 0.01 / 0.05—Willett et al. (2000)−6.4% / −6.0% / −6.4%
How to read this. Differences of 5–7% between published correlations are the typical uncertainty of the pendular bridge model; the implementation error (≤ 10⁻⁶) is negligible next to them. Neither case 11 nor 12 is an experimental validation. To validate a wet material, compare with measured force × separation curves (for example, Willett et al., 2000), with the viscous resistance measured between moving spheres (Pitois et al., 2000) and with a macroscopic test (heap, drum or shear) as a function of liquid content, as in 18.6.
Artificial values in this test. ρ = 1×10⁶ kg/m³ (case 11) and Δt = 1×10⁻¹² s (case 12) exist only to isolate the force law. In real simulations, use the material density and a Δt from the Rayleigh suggestion with a convergence test.

12. Liquid bridge with coarse-graining

What this case verifies

The same local interaction is compared at CGF 1, 2 and 4. The reference is the scaling of Eq. (3.64): forces four and sixteen times larger, respectively, at equal relative surface velocities. This supplements case 11; it does not compare equipment throughput or mixing.

Parameters and procedure

FieldPhysical value at CGF = 1
ParticlesR₁ = R₂ = 1 mm; ρ = 2500 kg/m³; E = 1×10⁵ Pa; e = 1; μ = 0
Liquidγ = 0.072 N/m; η = 1 mPa·s; θ = 17.1887°; V̂ = 0.002; S₀ = 1 µm
Measurement stateh/rₚ = 0.02; v = 0.02 m/s; ω = 0; Δt = 1×10⁻¹² s

Form a bridge, set a controlled separation and measure the capillary and each viscous contribution separately. Repeat with enlarged geometry at CGF 2 and 4 while retaining physical input properties: conversion is applied once (10.7). The numerical test repositions the bridge after formation to isolate the law; it is not a free trajectory reproduced through animation alone.

Result

Figure 12.1. Capillary, normal viscous and tangential viscous forces. Dashed: reference; symbols: brixDEM.
Figure 12.1. Capillary, normal viscous and tangential viscous forces. Dashed: reference; symbols: brixDEM.
CGFCapillary [mN]Normal viscous [mN]Tangential viscous [mN]
10.2640050.004712390.000504326
21.056020.01884960.00201730
44.224080.07539820.00806922
Interpretation. All three contributions follow quadratic scaling. Maximum relative errors for the illustrated family are 2.05×10⁻¹⁶, 2.88×10⁻¹⁵ and 5.18×10⁻⁹, respectively. The supplementary matrix of 2,805 checks includes unequal sizes, walls and migration; its largest error is 6.98×10⁻⁹, below the force criterion of 2×10⁻⁷. This checks local scaling and does not establish macroscopic process equivalence.
Scope. This case confirms that each local force scales as CGF² with preserved velocities. It does not show that flow rate, mixing or angle of repose of a wet material are the same at CGF 1, 2 and 4; that equivalence must be checked on the equipment or in a reference test.

13. Ab-T10/PRM breakage: DWT impact and progeny

What this case verifies

A breakable sphere rests on an anvil and is struck by a steel sphere. Work assigned to the event determines t10 and the fragment distribution; the product masses must sum to the parent mass. This virtual DWT connects impact mechanics to PRM replacement, while supplementary checks isolate selection, size and damage. Formulation is in 3.16.

Figure 13.1. Virtual DWT: the impactor of diameter di hits, at speed v, the breakable body of diameter d resting on the anvil; the absorbed work sets damage, fracture and the t10 of the fragments.
Figure 13.1. Virtual DWT: the impactor of diameter di hits, at speed v, the breakable body of diameter d resting on the anvil; the absorbed work sets damage, fracture and the t10 of the fragments.

Parameters and procedure

Panel / conditionValue
Breakable bodyd = 5.5 mm; ρ = 2930 kg/m³; E = 52 GPa; ν = 0.25
Steel impactor and anvildᵢ = 12 mm; ρᵢ = 7800 kg/m³; Eᵢ = Ewall = 182 GPa; νᵢ = νwall = 0.30
Ab-T10/PRMS = 0.002 kg/J; dref = 5 mm; Emin,ref = 100 J/kg; A = 67.7%; seed = 0
Contact / integrationHertz–Mindlin; e = 1; μ = 0; g = 0; Δt = 5×10⁻⁸ s
Energy sweep200 independent pairs; initial impactor kinetic energy of 100–2000 J/kg of body mass; 4000 steps.

Set the breakable material to Damage + replacement (10.10), keep the impactor unbreakable and set its velocity for the target energy. Comparison uses work actually assigned to the body, not nominal impactor energy. The campaign contains 139 fractured parents. Statistical reproduction across energies requires independent impacts and event recording; a single animation does not reproduce the full population.

Result: t10 and mass

Figure 13.2. t10 against specific event work and mass-balance residual per parent. Dashed: Ab-T10 law; symbols: brixDEM.
Figure 13.2. t10 against specific event work and mass-balance residual per parent. Dashed: Ab-T10 law; symbols: brixDEM.
QuantityReference / criterionResult
t10Eq. (3.69), using event workMaximum difference ≤ 1.5×10⁻¹⁴ pp
Child and fines massesRecover parent mass; relative error ≤ 1×10⁻⁹2.76×10⁻¹⁵ maximum error

Fragment-distribution check

To separate PRM sampling from impact mechanics, fix t10 = 20% and generate 10,000 families for each progeny law. Compare mean mass passing with the prescribed distribution, rather than sphere count.

Figure 13.3. Mean progeny mass passing for Beta and Gaudin–Schuhmann; dashed: prescribed distribution.
Figure 13.3. Mean progeny mass passing for Beta and Gaudin–Schuhmann; dashed: prescribed distribution.

The maximum absolute difference in fractional passing is 0.0046 for Beta and 0.0047 for Gaudin–Schuhmann: approximately 0.46 and 0.47 percentage points. Fragment discretisation and sampling explain the difference between a continuous curve and a finite population.

Supplementary checks: selection, size and damage

For selection, use 1,000 fresh bodies per energy with the tabulated properties and replacement disabled. For size effects, repeat at diameters of 2.5, 5.5 and 11 mm. For repeated impacts, retain damage across five impacts with mean specific work of 194.96 J/kg.

Figure 13.4. First-fracture frequency against specific work. Dashed: reference probability; symbols: brixDEM.
Figure 13.4. First-fracture frequency against specific work. Dashed: reference probability; symbols: brixDEM.
Figure 13.5. Size effect: results normalised by energy exposure from Eq. (3.67).
Figure 13.5. Size effect: results normalised by energy exposure from Eq. (3.67).
Figure 13.6. Cumulative fracture under repeated impacts. Dashed: reference; symbols: brixDEM.
Figure 13.6. Cumulative fracture under repeated impacts. Dashed: reference; symbols: brixDEM.
Event E [J/kg]Reference P [%]Fractured [%]Difference [pp]
140.6310.3610.20−0.16
256.5930.5530.60+0.05
411.9250.6551.20+0.55
647.9670.6470.90+0.26
1156.0590.4090.90+0.50
Interpretation. First-fracture frequencies lie within three binomial standard deviations of the reference; no individual decision differed from its sampled threshold. Under repeated impacts, retained damage increases the fractured fraction. These results check equations and fragment replacement. Ore validation requires comparing probability and PSD with independent experiments after calibration under the same energy protocol (18.4 and 18.6).

Progeny law × published family

The Beta law (default) uses the incomplete beta function with the parameters of Tavares & das Chagas (2021, Table 3), fitted to 40 materials (rocks, ores and clinker, excluding iron ore): tn = 100·I(t10/100; αn, βn). The Gaudin–Schuhmann option uses a single exponent obtained from t10, tn = (1/n)α with α = ln(t10)/ln(0.1). Percent passing at 1/2, 1/4, 1/25 and 1/75 of the parent size:

t10Beta (published family) — t2 / t4 / t25 / t75Gaudin–Schuhmann — t2 / t4 / t25 / t75
10 %47.5 / 20.3 / 5.7 / 3.250.0 / 25.0 / 4.0 / 1.3
20 %73.8 / 39.9 / 11.5 / 6.461.6 / 38.0 / 10.5 / 4.9
30 %88.0 / 56.8 / 17.6 / 9.869.6 / 48.4 / 18.6 / 10.5

For t10 ≥ 20%, Gaudin–Schuhmann underestimates the coarse part by up to 12–19 percentage points (t2). Prefer Beta unless material data justify the other law. Iron ores fall outside the published family (lower t2 and higher t75 for the same t10); for them, the default parameters are not an adequate estimate. Fines below the smallest fragment are represented by dummy particles; the power-law extension suggested in the same reference is not applied.

Energy assigned to the body and stiffness

Ab-T10 uses the work absorbed by the body, split between both sides of the contact by η = kpartner/(kbody+kpartner), not the nominal impactor energy. The parameters are therefore valid only at the stiffness at which they were calibrated. Example with fixed S and Emin, sphere against a 200 GPa wall, 72 J/kg per impact and 10 impacts:

Body Young's modulusη (body share)Fraction fractured
5.2 GPa0.97576.2%
52 GPa0.79866.0%
208 GPa0.49844.5%
Practical consequence. Reducing Young's modulus of the particle only (or of the impactor/wall only) changes η and alters breakage by tens of percentage points. Calibrate S and Emin with the same stiffness ratio that will be used in the simulation.

Uniform reduction of Young's modulus. When Young's modulus of all materials — particles, impactor and walls — is multiplied by the same factor, η does not change and the response is preserved. Same test (10 impacts of 72 J/kg, 128 bodies per shape, calibration not refitted; particle/wall 52/200 GPa as reference):

ShapeFactor 1 (52/200 GPa)Factor 0.1Factor 0.01
Sphere60.16%60.16% (+0.00 p.p.)60.16% (+0.00 p.p.)
Pair clump62.50%62.50% (+0.00 p.p.)63.28% (+0.78 p.p.)
Tetra clump59.38%60.16% (+0.78 p.p.)61.72% (+2.34 p.p.)

η = 0.798 for all three factors. For spheres, the fraction with a first fracture is identical up to a 100× reduction; for clumps, the difference stays within 1–3 bodies out of 128. Uniform reduction is therefore the consistent way to gain timestep without recalibrating Ab-T10 for this type of impact. The test covers repeated isolated impacts; it does not cover confined beds or the fragment distribution.

14. Evidence summary and limits

#CaseVerified quantityResult
1Normal restitutione = |v_out/v_in|10.1% at e=.5; ≤2.6% at e≥.7; large deviations at low e
2Oblique collisioncoupling Δω/Δv_tOK 5.5 % (torque)
3Sliding into rollingv_f, ω_f, t_sOK 0.05 %
4Ramp rollinga = 5/7·g·sinθOK < 0.1 %
5Ramp slidinga = g(sinθ−μcosθ)OK < 0.01 %
6Rolling thresholdθc = arctan(μ_r)OK (16 to 17°)
7Braking by rollinga = 5/7·μ_r·gOK 0.1 %
8JKR adhesion (particle)F = 3π·γ·R_eff and rupture gapOK < 0.1 %
8bJKR adhesion (wall)pull-off, rupture gap and F(δ) curveOK < 0.1 %
9Free-body 6-DOFtorque-free rotation; eccentric impact and reactionsMeasured coupling: 25/25; Part A free rotation: reference without solver result
10Clumpsmass/inertia, invariants, JKR and orientationOK; 0 failures across 17 checks
11Liquid bridgeForce, rupture, viscosity and inventoryRupture: 0.00084%; viscosity: relative error ≤ 5.61×10⁻⁵ (code × equation). Model × published correlations: 5–7%; no experimental validation
12Bridge + coarse-grainingLocal force scalingMaximum relative error 6.98×10⁻⁹; no macroscopic validation
13Ab-T10/PRMDWT, progeny, selection, size and damageMass: relative error 2.76×10⁻¹⁵; progeny ≤ 0.5 p.p. from the prescribed curve; uniform Young reduction preserves selection. PSD of a specific material not validated

These cases check laws and balances under the stated conditions. When reproducing a test, compare the same quantity, population and measurement interval. Agreement with an equation does not replace experimental validation of a material or equipment; calibration and physical-comparison procedures are in Chapter 18.