Ideas that won the community vote, handed to an AI pipeline that researches the problem, preps geometry, meshes, solves and post-processes — hands-free, on licensed Ansys tools, with an engineer reviewing before anything publishes.
130studies built & staged
1778solver runs (1158 utilized)
19Ansys tools exercised
2,800–4,070engineer-hours equivalent (est.)
Real program numbers, traced to the run ledger — sweeps counted as sweeps, dead ends included.
Of these, the 119 published studies in the record below itemize 1,088 utilized solves (1,484 counting per-study tracked dead ends); the remaining 294 solves (70 utilized) sum from retired studies — counted in the tracker, not itemized in the public record.
Loading queue…
✓ Published — every published study, newest first; filter by tool for that tool's record
Filter by industry
6solver runs (2 utilized)
4cores per solve
23k cellsmesh
We simulated a backyard grill plume in Ansys Fluent: the smoke dilutes ten-fold across the yard but is still ~200x above the nose's detection threshold at the property line -- so yes, the neighbors smell it, because noseheadline result
We built a net of 3,280 springs in Ansys Mechanical and let it sag under a weight -- its settled shape solves the same K u = F an FEA solver does, matching a direct linear solve to 0.000%: a lattice of springs and Gaussiheadline result
We dropped two office chairs in Ansys LS-DYNA at the same 65-degree lean: the narrow 320 mm base tips over while the normal 600 mm base doesn't move -- base width, not lean effort, is what keeps you uprightheadline result
We settled 526 solid-gold coins in Ansys Rocky: the packed bed hits ~11,000 kg/m3 -- about the density of solid lead and more than 11x denser than a person -- so you'd land on top like a cork (even a steel ball floats), headline result
DEM (no FE mesh): ~700 discrete particles per case; open-top cylinder jar wall imported…mesh
The lentil-vs-sphere packing question, tested in Ansys Rocky DEM: pour ~700 equal-volume candies of each shape and the flat lentils fill ~9% more of the jar than round beads (0.62 vs 0.57) -- matching the famous 2004 resheadline result
Utilized = the 5 production Monte-Carlo solves (perfect cube + shave 0.5/1/2/3 mm), 1,024 dice each (~5,120 rolls). Superseded = method-development…what ran
We rolled 1,000 dice at once in Ansys LS-DYNA (and shaved them): a perfect cube is fair (chi-square 4.8), and rigging one is harder than you'd think -- a 0.5 mm shave does nothing, and it takes ~3 mm before the two big fheadline result
Utilized = 3 production occupant-response runs (stand / brace / jump) + 1 production deformable-car run (intact frame), each to normal termination.…what ran
The 'jump right before it lands' myth, tested in Ansys LS-DYNA: jumping is the WORST option (~138 g vs ~80 g standing) because you separate from the floor and get slammed -- bracing your knees (~75 g) is best, and none oheadline result
Utilized = the one converged elevation sweep (7 sun angles + a validation trace, 200k rays each) that produced the concentration data and the plaza…what ran
The 'death-ray building' effect, recreated in Ansys: a generic concave glass facade concentrates the reflected sun ~3.3x, and an Ansys Mechanical transient thermal solve heats a stationary car with the moving hot spot --headline result
Utilized = 8 FDE gap-sweep cross-section solves (coupling length vs gap) + 1 FDTD field-hero run. Superseded = development runs: an earlier FDTD…what ran
How close can eight optical signals race on a chip before they crash? Ansys Lumerical says keep the silicon waveguides ~0.45-0.55 um apart for <-20 dB crosstalk -- squeeze to a 0.2 um gap and >80% of a beam jumps straighheadline result
Utilized = the one converged MagnetostaticZ field solve (real NdFeB PM material + far-field balloon boundary) that produced the field and the…what ran
The magnet-in-a-copper-pipe trick, computed in Ansys Maxwell: a moving magnet induces eddy currents that brake it (F = k*v), giving a terminal velocity of ~5.4 cm/s -- more than 40x slower than free fall -- and a colder headline result
Utilized = the 4 production solves behind the post: Mechanical free-free modal of the FORK (anti-phase tine mode tuned to 440.0 Hz), Mechanical…what ran
A shape's sound is baked into its geometry -- Ansys Mechanical reads the natural frequencies off the CAD file (fork = one pure tone tuned to A440, bronze bell = a warm ~303 Hz hum plus a chord of partials), an LS-DYNA stheadline result
Utilized = 4 production cases, all fully DEFORMABLE conforming-wedge blade: DEAD-ON (hero + survival + lead-extrusion result), OFF-CENTER 2 mm…what ran
A 9 mm bullet really does split in two on a fixed katana edge -- but the sword isn't cutting it: the soft bullet splits itself on an edge 20-80x stronger than it is, and the blade comes through it -- surviving with at moheadline result
Utilized = the two solves that produced (and cross-checked) the answer: a coarse 10-through-thickness NLGEOM run and a finer 14-through-thickness…what ran
A drying board is a dry, shrinking shell wrapped around a wet, swollen core -- the shell gets stretched to ~8x the wood's cross-grain strength and checks; dry it slower and the tension stays under the limitheadline result
Utilized = the two steady SST solves that produced the answer: tailgate UP (Cd 0.614, 311k-cell polyhedral mesh) and tailgate DOWN (Cd 0.613,…what ran
At 70 mph the drag is the same up or down (Cd 0.614 vs 0.613) -- dropping the tailgate doesn't save gas, because the raised gate traps an air bubble the wind just rides overheadline result
Utilized = the 4 conjugate Icepak field solves that back the post: desk, lap, duvet, doghair (each gives the die + copper-spreader + keyboard-deck…what ran
On a desk or a lap it's fine; bury the vents in a duvet and thermal throttling quietly takes about a third of your frame rate -- while the keyboard over the chips hits 121 Cheadline result
The best coupler is only best within +/-0.75 um of perfect aim; past that, the design that gave up half a decibel carries 2.8 dB more lightheadline result
Utilized = the five converged vent-variant solves (free-space reference 49k tets/7 passes, sealed 29k/10, 80x3 mm slot 62k/11, equal-area hole array…what ran
One cooling slot leaves 32 dB at GPS L1 and turns the box into an amplifier at cavity resonance; eight holes of equal open area buy 47 dB back for freeheadline result
Utilized = the two converged 1.8M-cell prism-mesh solves: the two-tower case (2000 iters) and the empty-domain reference that defines amplification R…what ran
Everyone expects the gap to be a wind tunnel; the solve shows it is the calmest spot, and the corners are the dangerheadline result
Utilized = the 6-snapshot burnback sweep (web 0->100%, port 15->50 mm) with the final converged recipe; every case closes mass and lands within 4-9%…what ran
Same motor, four grain shapes, four completely different thrust curvesheadline result
Utilized = 30-case laminar sweep (5 Mach x 6 nose radii) that gives the shock stand-off vs Billig, of which the 6 Mach-6 cases were re-solved with…what ran
A blunt nose survives re-entry -- the bow shock detaches and does the shieldingheadline result
Utilized (1): the final coupled solve -- a transient thermal solve (PLANE292 axisymmetric, 60 substeps over a 420 s descent) whose peak temperature…what ran
Three hand estimates of a brake disc's temperature disagree by 1000 C -- two say it meltsheadline result
Utilized = 23 board-layout solves in the 24-sample sweep (best 115 C, worst 149 C, 34 C spread) + 1 transient warm-up of the worst layout (cold start…what ran
In a fanless box, where you put the boards swings the hottest one by 34 Cheadline result
Utilized = the full 10 aspect-ratio x 3 frequency (1/10/50 kHz) matrix, all converged one extraction at a time; every case returns DC R, AC R (skin…what ran
A busbar's resistance can be 8.5x its DC value at switching frequencyheadline result
Utilized = 5 mount positions (top aft/mid/fwd + belly mid/fwd) each a full-wave HFSS solve; the far-field pattern + coverage extracted per position.…what ran
Same antenna, different spot: ground coverage swings from 75.0% to 92.7%headline result
Utilized (1): the final coupled solve -- a transient thermal solve (PLANE292 axisymmetric, stepped convection over a 2400 s quench) whose per-step…what ran
A quenched part squeezes its own skin -- the compression that fights fatigue, and the tensile core that is its priceheadline result
Utilized (1): the visible TiO2 atom library -- 11 pillar widths x 61 wavelengths (0.40-0.70 um) swept in Lumerical FDTD (periodic unit cell, 0.30 um…what ran
A flat metalens camera images sharply in exactly one color -- green stays crisp, red and blue smear into halosheadline result
One production solve: solve_progress.json mode 'run' ends 'flux ratio settled at -7.73% after 600 iterations... exported field_steady.csv... gate…what ran
A header does not share air equallyheadline result
Three head cases solved (0.50/1.00/1.33 design head -> h050/h100/h133 case files, each with its own export_progress + analysis + field CSV) plus one…what ran
A clean 3-configuration comparison (influent depth 0.5/4.5/8.5 m), each a steady 900-iteration precursor plus a 6362-step transient tracer march to…what ran
The ideal tank has no term for where the pipe entersheadline result
Utilized (1): the PLOW strip -- 9,261 grains of 0.7 mm sand, one 0.25 m/s pass of the 0.8 mm conical tip at 2.0 mm below the MEASURED settled…what ran
A fine pendulum point plows a V-groove that holds its shape -- swept along two detuned swing frequencies, it engraves the classic sand roseheadline result
Ansys Lumerical FDTD 2026 R1 driven headless through lumapi. 2D simulations of N-way silicon-in-oxide waveguide intersections (N = 4, 8, 12, 16) at…what ran
A four-way waveguide crossing passes 91.5% of the light straight through at 0.39 dB, and even sixteen arms meeting at one point still pass roughly 80% - beams never collide, but the merged silicon core does diffract themheadline result
Ansys LS-DYNA R16 coupled thermal-structural (*CONTROL_SOLUTION SOLN=2, implicit mechanics) on a 120 x 124 x 6 mm butt joint, 25,920 hexahedra at 1…what ran
Full-penetration fusion zone 7.11 mm wide through 6 mm plate, cooling 800-500 C in 11.5 s, leaving +386 MPa of locked-in tension at the seam against a 350 MPa yield.headline result
Ansys Mechanical APDL 2026 R1 steady-state conduction (SOLID70, enthalpy-free constant properties, k=167 W/mK aluminum) on a 60x60 mm / 6 mm base…what ran
Best buildable heat sink 59.88 °C at 16 mm fins, 1.0 mm thick, 22 of them — using 42.7 of the 45 cm³ mass budget, hard against the constraint.headline result
Utilized: a 20-width Lumerical FDTD unit-cell library (0.60-2.50 um square Si pillars, h=10 um, 201 wavelengths 8-12 um; TIME 15 ps, SHUTOFF 1e-5,…what ran
A metalens is a lens made of data -- flat, hair-thin, and sharp at exactly one color of heatheadline result
26 Ansys sessions. UTILIZED (14): five watertight mesh builds (one per configuration), five 600-iteration solves, and the two-case mesh-refinement…what ran
22 Ansys sessions. UTILIZED (12): the dimpled mesh build, the smooth mesh build, the 40-second setup probe that validated the Transition SST toggle…what ran
Finite-difference grid, mesh accuracy 4, over a single 1000 nm grating period in x and…mesh
46 solver runs across ~24 Lumerical sessions. UTILIZED (42): the FDTD validation probe at normal incidence, and the 41-angle production sweep.…what ran
No volume mesh - DEM is meshless. 40 mm rigid spheres (up to 70 simultaneous bodies in…mesh
2 Rocky DEM solves utilized - the single-trap launch KERNEL calibration and the final 36-trap HERO cascade (70 balls, clean headless extract). 4…what ran
Frame: 12 perimeter columns 0.5 m square x 96 m + ring beams 0.4 m square every 16 m…mesh
4 utilized: shunt-at-24m, shunt-at-48m, post-quake damage, PEC control - the final set, each adapted at 5 MHz + 121 exact sweep points (~16-24 min…what ran
84 m x 3 m tied-arch bowstring: CHS 273x8 arches, CHS 244.5x8 ties, CHS 168.3x6 floor…mesh
5 utilized: final modal (84 m design, f1 2.48 Hz), static standing crowd, self-weight-only static (DAF baseline), and the two 40 s crowd transients…what ran
Segmented minifigure-style figure: 7 hex-meshed ABS parts (head, torso, hips, two legs,…mesh
8 utilized: the impact matrix (4 landing attitudes x concrete/grass) on the reworked deformable-concrete model. 9 superseded: a first impact matrix…what ran
No mesh (ray tracing). Non-sequential scene: collimated Source Rectangle (CosineExponent…mesh
8 utilized: the density sweep (tau550 0.2-10), each an OpticStudio non-sequential ray trace of 250k rays x 18 wavelengths that produced the final…what ran
960 FLUID220 20-node hexahedral acoustic elements (240 axial x mapped), 7221 nodes; ~10…mesh
2 utilized: (1) modal acoustic, 26 modes to 1.7 kHz; (2) full harmonic sweep, 400 frequency steps 4-1600 Hz. 5 superseded during feasibility/method:…what ran
Utilized: 3 CFX steady SST + Thermal Energy solves of the same 168,000-cell mixing tee, identical mesh, differing only in branch inlet speed. (A)…what ran
Utilized: 6 SIwave AC (SYZ full-wave -> Touchstone Z11) solves of the PDN self-impedance — bare plane; 1, 2, 4, 8 ceramics; and a bulk+ceramic mix —…what ran
Utilized: 1 Q3D adaptive extraction of the two-wire 20 AWG copper power pair (60 mm section, 27,106 tets) that solved all three matrices in one pass…what ran
Utilized: 3 CFX rotating-frame SST solves of the same 277k-node passage at 14,324 rpm, one per back-pressure rung, forming the speed-line and the…what ran
Utilized: 4 Icepak steady conjugate natural-convection + Discrete-Ordinates radiation solves of the same phone, identical mesh (89,015 cells),…what ran
Utilized: 5 full build-and-cutoff print simulations of the Ti-6Al-4V bridge, each 20 super-layers grown by SOLID185 element birth/death with a 0.5%…what ran
Utilized: 2 Maxwell 2D AC-magnetic (eddy-current) solves of the coil+pan, identical except pan material. (1) Steel base -> 541 W absorbed, 0.113 mm…what ran
Utilized: 1 HFSS solve of the inset-fed patch (16-pass adaptive + interpolating 2.0-2.8 GHz sweep + infinite-sphere far field) producing the S11…what ran
Utilized: 2 solves. m6 - quasi-static equatorial belt-pressure ramp on the layered melon; first rind tear at ~0.8 MPa belt pressure at the belt edge…what ran
Utilized: 3 converged 2D-axisymmetric VOF transients (v = 4.1 / 8.1 / 12.2 m/s), each 3000 steps to 3.0 s flow time with 60 autosaved snapshots, plus…what ran
Round 1 utilized: 3 elastic head-press solves (nc=8/12/16) for the Hertz-overlay + mesh-convergence control, and 3 drop solves (T1 tail-force ramp,…what ran
Utilized: 12 elastic sweep solves (Models A & B, V=2-12 m/s) for the true peak-stress-vs-speed chart + 1 erosion run (temperhard) for the shatter…what ran
Mechanical corrected frictionless solve + LS-DYNA explicit run used; first-pass bonded Mechanical solve + several LS-DYNA decks with IC/connectivity…what ran
UV-sphere membrane, R 60 mm, 56 x 76 lat/long -> 4,256 quad membrane shells (ELFORM5) + a…mesh
7 explicit LS-DYNA solves during the rework, 1 utilized (the final seam-guided pop). The 6 superseded runs were the debug/tuning ladder: (1) a…what ran
Fluent coarse 2D triangle mesh of the exact 20x20 maze (mm scale: 160x160 mm block, 6 mm…mesh
The reaching-the-outlet VOF flood + the 3 Laplace solves (flow/thermal/electric) are the 4 utilized results. The 5 superseded VOF runs were the…what ran
