The agitator job builds an ordinary FEA case: a nozzle at the centre of a top head, the mounting flange face as the loaded end. Nothing is hidden in the scenario; it only fills the same inputs the workbench shows, and they stay editable afterwards. The loads are the vendor's, given at the mounting flange.
Which load goes in which row
- Weight, axial thrust and torque act while the machine runs. They go in the weight row (D, a sustained primary load): axial force P = thrust − weight (thrust positive away from the vessel, so the weight enters negative) and torque about the nozzle axis.
- The shaft bending moment turns with the shaft, so at a fixed point of the nozzle it reverses. It goes in the operating row as a load-controlled load (L) with reversal on, the fourth component of the six-component row. The other components are zero.
- Loads are in the attachment frame and applied at the nozzle end (the flange face). The lever of any shear is therefore not added: only what the vendor gives at the face acts, and the head sees it through the neck.
- Gussets between neck and head are recorded and hand-checked, not meshed. The flange is drawn, not meshed: it adds no stiffness.
Torque from motor power
Cycles from speed, hours and life
The code counts one bending stress cycle per revolution over the design life, and gives it to the fatigue analysis as an event that swings the operating row between −1 and +1 at operating pressure. The range is therefore twice the vendor's bending moment. Start-ups are a separate count (the design cycles, shutdown to operating).
Flange tilt and face deflection
For each load case a rigid motion u = u0 + ω × (x − c) is fitted by least squares to the displacements of the end ring (the ring is held as a plane circular section, so it does not oval). The rotation component across the nozzle axis is the face tilt. With the flange given, the face deflection is the tilt times a radius.
Default agitator of the dialog (calculated by hand from the formulas above)
| Motor 15 kW, 90 rpm, SF 1.5 | ω = 2π·90/60 = 9.425 rad/s |
| Torque | 15 / 9.425 = 1.5915 kN·m; × 1.5 = 2.387 kN·m |
| Weight 12 kN, thrust 1.5 kN | P = 1.5 − 12 = −10.5 kN (row D) |
| Bending 8 kN·m reversing | operating row ±8; event range 16 kN·m |
| Cycles, 8000 h/yr, 20 yr | 90 · 60 · 8000 · 20 = 8.64 × 10⁸ |
| Start-ups over the life | 2000 (separate event) |
| Neck 406.4 × 12.7 mm, hand statics | A = 15 708 mm², Z = 1.499 × 10⁶ mm³: bending 5.34 MPa, axial 0.67 MPa, torsion shear 0.80 MPa |
| Flange OD 597 mm, tilt limit 0.05° | 0.05° = 8.727 × 10⁻⁴ rad; δ = 8.727 × 10⁻⁴ · 298.5 = 0.260 mm |
| Face limit 0.5 mm at OD | equals a tilt of 0.096°: the 0.05° tilt limit governs |
The neck stresses are plain beam and pipe statics for a feel for size; they are not an engine output. The dev log (2026-10-10) records 0.046° tilt per case on this model, which is 0.240 mm at the OD. The limits 0.05° and 0.5 mm are typical values in the dialog, flagged to be confirmed with the vendor.
What to expect and what is not in it
The dev log (2026-10-10) records that the default agitator passes the stress checks at 34 % while the welded-joint fatigue fails, damage about 118 at 8.6 × 10⁸ cycles. That is the finding the job is built to show: a reversing bending moment at 90 rpm is a very high cycle count, and the stress ratio does not tell it.
- One cycle per revolution only. Blade-pass or other shaft harmonics, hydraulic unbalance and resonance are not computed; the vendor's bending moment must already include them.
- The loads are static and the end is a rigid plane. The vendor's gearbox and shaft stiffness are not modelled.
- The flange is not meshed; if the projection is measured to the back of the flange the neck is run to the face, which is a little soft (conservative for a tilt limit).
[UNVERIFIED] The event states list P, O and C only, so the sustained weight and torque do not appear in the two states of the bending event. Their effect on the mean-stress correction in Method C was not traced through the fatigue code.
FE results are early access and under verification. They are not a substitute for the vendor's and the engineer's review.