METALLAI › Sample report

Can aluminium replace this stainless rotor?

A complete METALLAI case, anonymised and unedited in its conclusions. Four candidate alloys, fatigue at the real duty cycle, the standards checked — and the two places the platform refused to give a design answer. The obvious candidate lost.

The brief

A hollow conical component drives belts inside a continuously running production machine. It is made from precipitation-hardening stainless and works perfectly — it is simply heavy and expensive. The customer wanted something lighter, cheaper and easier to source, and arrived with 7xxx-series aluminium already in mind.

Part Hollow conical rotor, ~110 mm outside diameter, 35 mm bore

Incumbent 17-4PH, H900 condition — forged, no failures in service

Duty Belt speed 600 m/min, about 20 hours a day

Unknown at the time Belt tension, torque and the exact connection geometry

This is not a failure investigation. The part already works. It is a substitution study, which means the burden of proof sits on the replacement — and because the component turns continuously, fatigue had to be checked, not just static strength.

Step one: what load is it actually seeing?

Before any prediction runs, the duty cycle has to become a stress. This part of the report is arithmetic, and it is shown so the engineer can check it:

Rotational speed. 600 m/min ÷ (π × 0.110 m) ≈ 1,736 rpm.
Cycle count. About 2.08 million cycles a day → roughly 625 million cycles a year over a 300-day year.
Load scenarios. Belt tension was unknown, so three cases were carried through: 50 N, 150 N and 300 N tangential difference.
Worst-case alternating stress. At the thinnest section, the 300 N case gives σ_a ≈ 3.53 MPa, R = −1. Every fatigue result below uses that figure.

Stated assumption, not a hidden one. The geometry was idealised, because no detailed drawing of the connection was available. The report says so in exactly those words, and lists it again at the end under the effect of missing data. An assumption you cannot find is an assumption you cannot check.

Step two: four alloys, side by side

The incumbent plus three candidates, each run through the full chain — composition and process route in, mechanical properties, fatigue verdict, corrosion flags and standards screening out.

Alloy & condition	YS MPa	UTS MPa	HV	El %	Fatigue at σ_a = 3.53 MPa	SCC risk
17-4PH H900 stainless, forged — incumbent	1156.7	1270.2	378.4	10.8	No failure expected strong evidence, n = 126 forged rows	None
AA 7075-T6 forged, 480 °C/1 h + 120 °C/24 h	502.0	569.6	153.4	10.3	Beyond model range n = 2 on this route — extrapolation	Yes
AA 7075-T73 over-aged, 163 °C/24 h	466.2	537.5	143.9	11.6	Beyond model range n = 2 on this route — extrapolation	Reduced
AA 6082-T6 recommended forged, 530 °C/1 h + 175 °C/8 h	283.5	310.6	93.9	10.0	Beyond model range n = 2 on this route — extrapolation	None

The finding that decided it: the real load is a few MPa. All four alloys sit tens to hundreds of times below their yield strength. Strength was never the deciding criterion — which is why the strongest candidate is not the recommended one. A tool that only printed property numbers would have pointed at 7075-T6.

Step three: where the platform said no

Three of the four fatigue results came back “beyond model range” rather than as a number. That is the useful answer, and it is worth being precise about why:

The load is too far below the endurance limit. At 2.7–4.9 % of the endurance limit the Basquin term saturates at 10⁷ cycles. The engine can print a life — 7075-T6 computes to roughly 928,000 cycles, 95 % interval 467,000–1,845,000 — but it labels it as informational, not a design life, because the model is being asked a question outside where it was fitted.
The training coverage is thin. All three aluminium routes rest on two rows of forged data. The steel result rests on 126. The report says this next to every aluminium number rather than in a footnote, and recommends component testing.
The standards check is a screening level, not a certification. 6082-T6 clears EN 755-2's 260 MPa minimum — but that standard covers extrusion and rolling, and this part is forged. The report states the mismatch instead of quietly claiming a pass.

One more distinction the report insisted on. The standards table also showed FAIL rows — for 440C and X105CrMo17, which are different alloys, not conditions of 17-4PH. They do not apply to this part, and the report says so explicitly so nobody reads a red row as a problem with their material.

The recommendation

AA 6082-T6, forged, wall thickness increased to about 10 mm

Solution treat 530 °C for 1 h, age 175 °C for 8 h. Expected result: roughly 55–60 % lighter than the stainless part, lower cost, easy to source, and no stress-corrosion exposure.

Why not the alloy the customer came in with?

7075-T6 is stronger than the part needs — and strength was never the constraint.
It carries stress-corrosion risk at 5.6 % Zn in the T6 temper, in a machine that runs warm, damp and dusty. The platform raised that warning on its own.
It costs more and is harder to source. 6082 is standard structural stock.
7075-T73 stays as a fallback only if impact or wear resistance later becomes a priority.

On the wall thickness: the calculations used the existing 8.5 mm section and strength was still comfortable. But aluminium's elastic modulus is about a third of steel's, so the report recommends increasing the wall by 15–20 % to hold stiffness and keep vibration behaviour sensible. Even with that extra metal, the density ratio of 2.87 leaves the 55–60 % weight saving intact.

What was missing from the inputs — stated, not buried

Every report closes with this section. It is the part customers tell us they have never seen from a prediction tool, and it is the reason the numbers above can be trusted at the level they claim.

Unknown	Does it change the conclusion?
Rotational speed	No. It comes from physics, and belt slip would not move it materially.
Belt tension / torque	No. Three scenarios were carried through; even the worst leaves an enormous margin.
Connection geometry (keyway, interference fit, grub screw)	The critical gap. A sharp corner or keyway can raise local stress by 2–4×. The margin probably absorbs it — but probably is not a guarantee.
Aluminium fatigue coverage	All three results are extrapolation from two rows. Validate on the component. The steel result, at 126 rows, is far safer.

This is a starting point, not a final engineering decision. The report's own closing words. Before production: FEA on the real geometry for stiffness and vibration, a prototype tested in the actual machine, and a stress-concentration check once the connection detail is fixed.

Run this on your own part

Composition and process route in; properties, fatigue, weldability and standards out — each with its uncertainty and its data coverage stated. Steel and aluminium today; outside them the platform declines rather than extrapolating.

Open the app — free tier The other case: a confident 4140 run