What’s new in Axomek 0.1.11

Results that change when a saved project is solved again:

• Linear guides: the reported pitch of an off-centre pressing load (Fz at x ≠ 0) has the opposite sign (positive now presses the +x carriages); the carriage loads of such a load are unchanged.
• ISO 286 grade-4 holes (K4, M4, N4, P4 to S4) take Δ = IT4 − IT3, also as a bearing's housingTolerance: K4 at Ø40 is +1/−6 µm instead of −2/−9 µm, which changes the operating clearance and fit pressure of such seats.
• Several supports at one node: the reaction is shared by the supports rigid in that freedom only (a spring there carries its own k·u); before, every support at the node took an equal share, springs included.
• A bearing linked to a shaft support takes the shaft's material (modulus, Poisson's ratio, expansion) and bore and the linked housing's material for its fits wherever its mounting leaves them unset (before, bearing steel and a solid shaft), which changes the fit pressure and operating clearance on non-steel shafts and housings and hollow shafts.

Also from 0.1.11:

• Planetary sets without a ringTipDiameter, refused by 0.1.11, solve again: the ring's tip is shortened just enough to clear the planets' root fillets (the standard tip where that clears them); a given value still overrides it. The starter set no longer stores Ø256 mm (which fitted only 21/33/87 × m3): its default tip is Ø255.53 mm, the ring mesh's εα 1.73 → 1.83 and the lowest root safety 4.04 → 4.02. Internal gear pairs without tip diameters likewise (the ring tip that clears the pinion instead of a refusal).
• Planet pin load: the vector sum of both meshes' tooth forces at their own working pressure angles and the centrifugal force; unchanged for z_r = z_s
  • 2·z_p without shifts, different for shifted or other tooth counts (whose radial components no longer cancel). New pinTangentialForce and pinRadialForce; a planet bearing's duty cycle scales them directly.
• A self-aligning bearing (spherical roller, self-aligning ball) on a shaft support with a tilt stiffness, refused by 0.1.11 ("loads.mz"), solves again: the support is free to tilt (its tilt stiffness ignored, with a warning), so the shaft no longer gets a support moment there, in every load case and the duty cycle.
• Vibration with coupled self-aligning bearings: their radial tangent stiffness and no tilt stiffness. With a rigid housing the secant was taken before (several times softer: critical speeds rise); with a compliant one the housing's tilt stiffness was counted as the bearing's.
• Whirl on supports that differ between the planes: both planes are solved together. The governing critical speed is the first whirl that unbalance excites, which can be an elliptic backward one well below the softer plane's round forward whirl taken before (a disc on tilt springs 1:4, Ip = Id/2: 1.47 times lower); the forward and backward lists and the Campbell diagram change accordingly. Round supports are unchanged.
• Gear load spectra: the S–N curve no longer rises between knee points (the running minimum), so stresses in such a band count damage; the spectrum safety can drop slightly. Lives the life curves cannot rate are named in the curves' notes.
• Spherical roller bearings: an element's reported load is the magnitude of its laminae's resultant (consistent with its reported angle), slightly below the laminae's sum; lives and stresses are unchanged.
• A mounted gear pair driven by its wheel: gears[].meshLoss.loss is P1·(1/η − 1), as in powerLoss (it gave P1·(1 − η)).
• aISO: eC below κ = 0.1 from κ = 0.1, as aISO (slightly higher modified lives there); a κ that is not positive (no viscosity) gives aISO 0 with a warning, not the κ = 0.1 value.
• A self-aligning ball bearing on its own with "tilt": "free" (standard rules) has its tilt released as a spherical roller bearing's: held at zero with the moments left unbalanced, instead of an all but singular free tilt.
• Ball screws: the static load with the preload lifts off at 2^1.5·Fpr exactly (2.83 before), a negligible change.
• Command line: bare numbers are read in US customary units only with an explicit --units imperial and never with --json (a machine's units setting changes the display only; before, axm --json load add --at 50 wrote 1270 mm there). axm set, sweep and optimize --vary read each field in its own unit: gear, drive, rotor, key and planetary torques in N·m (axm set p.axm.json gears.G1.torque "150 N·m" stored 150 000 before), face widths in mm, strengths in MPa, speeds in rpm. --mass reads lb and clearances mil with --units imperial; --part-length takes units.
• A part with a length reaching past the shaft's end, or a diametral inertia below its spread's m·L²/12, is reported (warnings; results unchanged).
• Robustness (from an audit of some 19 000 solves of edited templates):
  • Invalid materials are refused instead of rated: gear pairs, bevel and planetary sets, fits and bolted plates need a Young's modulus above 0, a Poisson's ratio of 0 to below 0.5, positive strength limits and densities (a wheel with E = −206 000 MPa had infinite safeties shown as passing). Shaft materials now need ν ≥ 0 (was > −1) and a density above
    1. The smallest of several safeties is NaN, and fails, when one is NaN (it was dropped). A module whose inputs fail validation shows its errors and no results or checks.
  • Loads that are not finite numbers (a belt drive with a zero pulley diameter) are refused by the shaft instead of giving NaN support loads.
  • Deep groove ball bearings with clearance under a light load (a 6207 under 0.01–15 N purely radial) did not reach equilibrium; a load continuation now solves them. Every load that solved before solves the same way (the reference cases are unchanged).
  • A bearing on a shaft support that the shaft leaves unloaded is reported as unloaded (life and static safety unlimited, no stress, the minimum-load check) instead of an error.
  • Loads can take the torque that balances their shaft ("torqueMode": "balance"); the gearbox, belt-drive and electric-motor templates use it for their input and output couplings, so small edits of the gears or drives no longer leave the shafts unbalanced. Existing files keep their explicit torques.
  • Lubricants need ν40 above ν100, both 0.5–100 000 mm²/s, and bearing temperatures −60…300 °C; a bearing whose κ cannot be computed gets no life.
  • Ball screws: a ball carrying less than 10⁻⁹ of the largest ball load counts as unloaded. The native build had counted rounding residues (some 10⁻²¹ of the largest) as loaded, so its load ratio differed from the browser's (the linear-axis template with 6.35635 mm balls: 2.052 natively, now 1.026 in both). A nut has 1–20 circuits.
  • Validation now reports what the solvers refuse: thrust-only bearings on shaft supports, single tapered roller bearings on floating supports, NU/N bearings locating a shaft that gravity pulls axially, sections of absurd size (diameters outside 0.1–10 000 mm or more than 300 times apart, sections over 100 m), a rotor stack beyond the shaft's end and a magnetic pull the shaft cannot hold (with that message instead of "not held in place radially"). The conformity of roller bearings, which is not used, is no longer checked.
  • JSON results write an unlimited value (an infinite life or safety) as the string "inf" instead of null, which now only means "not computed"; check messages say "unlimited" instead of "inf h … (0%)".